Ascertaining causes of neonatal deaths using verbal autopsy: current methods and challenges

STATE-OF-THE-ART

Ascertaining causes of neonatal deaths using verbal autopsy:current methods and challengesN Thatte, HD Kalter, AH Baqui, EM Williams and GL Darmstadt

Department of International Health, International Center for Advancing Neonatal Health, Johns Hopkins Bloomberg School of PublicHealth, Baltimore, MD, USA

Objective: ‘Verbal autopsy’ (VA) is used to ascertain cause of death in

countries where vital registration systems are lacking. Current VA methods

for neonatal deaths vary widely and suffer from several limitations. We

aimed to: (1) review current neonatal VA methods, (2) identify gaps and

limitations, (3) illustrate some limitations using VA data and (4) identify

new approaches in methodology and analysis.

Study Design: Rolling techniques and database search terms were used

to identify articles that described neonatal VA administration, validation

and cause of death assignment.

Result: Current VA interviews include open and close-ended modules and

are administered by trained interviewers. Causes of death are determined

using physician review and/or computer algorithms for various neonatal

causes of death. Challenges include lack of a standardized VA instrument

and administration of methods, difficulty in identifying gold standards for

validation studies, lack of validated algorithms for causes of death, poor

existing algorithms, lack of standardized death classification terminology

and the use of hierarchy to assign causes of death. Newer probabilistic

methods of analysis such as Bayes Theorem or the Symptom Pattern

method may improve accuracy for cause of death estimation and alleviate

some of the challenges with traditional physician and algorithmic

approaches, although additional research is needed.

Conclusion: Given the continued reliance on VA to determine cause of

death in settings with inadequate registration systems, it is important to

understand the gaps in current VA methods and explore how methods can

be improved to accurately reflect neonatal disease burden in the global

community.

Journal of Perinatology (2009) 29, 187–194; doi:10.1038/jp.2008.138;

published online 25 December 2008

Keywords: verbal autopsy; cause of death; neonatal mortality;algorithms; hierarchy; challenges

Background

Data on causes of death are important for health sector planning,including assessing programmatic needs, monitoring progress ofinterventions and reassessing health priorities. However, little isknown about causes of death in many developing countriesbecause vital registration systems are lacking. Many deaths occur athome, outside the formal health sector, and few are attended byqualified medical professionals. Verbal autopsy (VA) has been usedto assign cause of death in such settings.

VA is a post-mortem in-depth interview with the primarycaregiver of the deceased. In the case of child deaths, this is usuallythe mother. Underlying assumptions are that (1) each cause ofdeath investigated has a set of observable symptoms that can berecognized and recalled by the primary caregiver, and (2) thecharacteristics of one cause of death can be distinguished fromthose of all others.

The majority of the estimated 4 million annual global neonataldeaths occur in developing countries outside of the formal healthcare system.1 Previous research on cause of death for children <5often excluded neonatal cases due to poor reporting and smallnumbers of neonatal deaths. Neonatal deaths were grouped with‘other’ childhood or ‘perinatal’ causes, thus limiting opportunitiesto validate neonatal VA methods;2–5 or failed to identify specificcauses of neonatal deaths.6 Neonatal causes of death are alsoparticularly difficult to classify due to nonspecific signs andsymptoms in sick newborns.2,7–9 Recently, several studies havefocused exclusively on the use of VA in newborns, making this anopportune time to assess the current state of knowledge and providedirection for future research efforts.

We reviewed current literature on neonatal VA that (1) identifiedspecific causes of death during the neonatal period, (2) describedVA methodology used and/or (3) reported on validation of VA forneonatal deaths. Using this information, we aimed to: (1) identifycurrent neonatal VA methods, including administration andcause-of-death assignment, (2) identify limitations of currentmethods, (3) illustrate limitations using examples from currentVA data and (4) identify new methods to measure causes ofneonatal death.

Received 15 February 2008; revised 13 July 2008; accepted 19 July 2008; published online

25 December 2008

Correspondence: Dr GL Darmstadt, Department of International Health, International Center

for Advancing Neonatal Health, E-8153, Johns Hopkins Bloomberg School of Public Health,

615 North Wolfe Street, Baltimore, MD 21205, USA.

E-mail: [email protected]

Journal of Perinatology (2009) 29, 187–194

r 2009 Nature Publishing Group All rights reserved. 0743-8346/09 $32

www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.138

mailto:[email protected]

http://www.nature.com/jp

Current methodsVA instrument design and administrationNeonatal VA instruments consisted of open-history narratives andclosed-ended questions about signs, symptoms and events leadingto death. Social autopsy modules to assess care-seeking or othercultural influences were uncommon. Some studies reported care-seeking information from open histories10 and one recent VAinstrument included a care-seeking module,11 however, its use hasnot been validated. VA interviewer training ranged from severaldays to 1–3 weeks.12–15 Local women with secondary levelschooling12–19 were primary interviewers, however, nurses,physicians20–24 and men14,23,25–27 were sometimes used.Interviews were conducted from several days24 or weeks after thedeath10,12,13,15,19,20,23,25,26,28 –30 to 5 years postmortem.31 Themother was generally the main respondent, however, interviewswith the grandmother or other close relatives were also reported incases where both the mother and baby had died.

VA cause of death assignmentPhysician review with 2–3 physicians was the most common methodused to assign cause of death from the VA.10,13,14,18,20–30,32–34 In theevent of conflicting diagnoses among physician reviewers, discussionsto reach consensus7,16,26,32 or a third party ‘majority rules’ wereapplied.22,25 In cases where all three disagreed, the cause was oftenlabeled as ‘undetermined’.18,21

Predefined diagnostic categories and clinical guidelines,7,21,24

prescribed lists of cause of death criteria,26 and InternationalClassification of Disease, 10th Edition (ICD-10) codes,23,29,34 orreference articles,13 helped guide cause of death assignment;however, in some studies, criteria to reach the diagnosis was left tothe physician’s discretion.28

Computer-programmed combinations of signs and symptoms(computer algorithms) of the fatal illness were also used to assignneonatal causes of death.12,15,26 Cases that met the algorithm’sdefinitions for each cause were identified.

Discrepancies between physician review and computer algorithmswere common. Campbell et al.32 found that computer algorithmsresulted in 48% of cases having an ‘unidentified’ cause of deathcompared with only 13% from the physician review. Physician reviewattributed more deaths to sepsis (20%) compared with computeralgorithms (7%).32 This result was different from that reported byFreeman et al. who found that physicians assigned only 1.2% ofdeaths to sepsis, while computer algorithms attributed 52.2% of deathsto sepsis.16 Lee et al. also reported disagreement in birth asphyxiadiagnosis between physician review and computer algorithms(kappa¼ 0.15); 85% of birth asphyxia deaths identified by computeralgorithm were unidentified by physician review.35 For other causes ofdeath such as prematurity and acute lower respiratory infection,Freeman et al. found that the percentage of deaths assigned wassimilar using both physician review and computer algorithms(26.4 and 20.4% versus 23.5 and 18.4%, respectively).16

Open histories were also used to assign causes of death.7,26

Freeman et al. reported that 12% of physician diagnoses reliedexclusively on the open-history narrative; and for conditions such assudden death and malnutrition, where computer algorithms wereunavailable, the open history provided critical information forphysician diagnosis.26 The ability of physicians to incorporate open-history information into their review was indicated as a potentialreason for the discrepancy between physician review and computeralgorithms for certain conditions like sepsis.26,32 However, Freemanet al. developed computer algorithms that included coded symptomsfrom the open history and compared these diagnoses with those madeby physician review. They found no increase in agreement betweenthe computer algorithms and physician review,26 raising questionsabout the usefulness of the open history. However, for conditionswhere computer-based algorithms were unavailable, such as suddendeath, open histories provided important information. They alsoprovided useful information about the social context of the fatalillness such as care-seeking and local beliefs, their conduct providedan opportunity to build rapport with the respondent, and may haveeven provided a cathartic effect for the mother after her loss of anewborn infant.

In addition to using algorithms and open histories, physicianreviews often followed a hierarchy to assign a single cause ofdeath.12,16,17,32,33 A hierarchy arranges causes of death in aparticular order based primarily on the specificity of the causaldefinition or perceived certainty of each diagnosis, as well as thesensitivity of the causal definition and the presumed physiologicalprecedence of each cause’s contribution to a death. The ChildHealth Epidemiology Research Group (CHERG) has identified acommonly used hierarchy (congenital abnormality-neonataltetanus-preterm-birth asphyxia-sepsis/pneumonia-diarrhea-other) to estimate the cause distribution of neonataldeaths.36 Conditions like congenital malformation and neonataltetanus with highly specific symptoms (for example malformation,for the former and spasms, stiff neck and jaw for the latter) can bediagnosed by VA with reasonable certainty16 and are therefore oftenplaced high in the hierarchy. As each cause of death is identified,cases assigned to the cause are sequentially removed from thesample and cannot be assigned another cause of death. Identifyingsingle causes of death allows distribution of causes to be presentedas a pie chart, common when depicting disease burden.

Despite progress to date, there are several limitations to currentVA methods such as the administration, validation and methods toassess cause of death including terminology and hierarchy ofcauses.

ChallengesLack of standardized VA instrument and administrationThere is currently no commonly used VA instrument for neonataldeaths. The WHO (World Health Organization) has incorporated a

Current methods and challenges of neonatal VAN Thatte et al

188

Journal of Perinatology

neonatal section into their VA for childhood deaths; however, itsuse has not been standardized. Administration of the VA instrumentposes additional challenges. First, identifying deaths in thecommunity where a VA is required is difficult. Though somecommunities without a functional vital registration system mayoperate informal surveillance through community leaders andhealth workers, obtaining accurate estimates about who has diedand who requires a VA is a continuing challenge. VAadministration requires substantial interviewer training andretraining in both clinical and bereavement issues as well asregular follow-up to obtain the VA from family members after adeath. Reporting bias may occur due to cultural factors and stigmaassociated with death, whereby families may not wish to reportdeaths of infants <1 month old.10 This may be further biased withmale interviewers. In addition, interviewers with some clinicalbackground may unknowingly bias results to reflect the diseaseburden in their particular setting.37 VA relies on the ability ofcaregivers to recall signs and symptoms exhibited by the newbornbefore death. Recall periods between 1 and 12 months have beenaccepted in adult studies37 and the validity of mother’s responsesfor a period up to 20 months after a child’s death has remainedhigh, possibly due to enhanced memory for details surrounding thetragedy of losing a child.6 Still, the possibility of recall bias mayoccur, especially in cases where a maternal death has alsooccurred. In these cases, validity of the caregiver response may belessened due to the inability to recall specific events from twodeaths (mother and newborn) or the inability to recall signs/symptoms in the same detail as the mother may have been able to.

Assigning cause of death using neonatal VA poses even morechallenges. Identification of gold standard criteria for validationstudies, lack of validated algorithms, overlapping signs andsymptoms for different causes of death, use of a hierarchy toassign single cause of death, assigning single versus multiplecauses of death, and varied cause of death terminology allcontribute to the many challenges to assigning a cause of deathusing this method.

Identification of gold standards and validationVA relies on the use of reference or ‘gold’ standard definitions todevelop accurate diagnoses. Given that VA is based on subjectiveaccounts from caregivers, validation studies are critical todetermine how close to the ‘truth’ VA diagnoses are.

Validation studies compare the diagnosis made by VA with areference standard diagnosis reached at the time of death. Medicalautopsy is rarely available so reference standard diagnoses aremade under conditions (i.e. in a hospital) in which the cause ofdeath can be determined with a degree of certainty based onphysician examination and laboratory testing. The validity of eachalgorithm for assigning cause of death is measured usingsensitivity and specificity estimates and the reference standarddiagnoses as gold standard.

Few validation studies for neonatal conditions have beenconducted, and those that have included neonatal causes have notused comparable methods. Reported sensitivity and specificity forcommon neonatal conditions differ considerably across studies(Table 1). This may be due, in part, to varying definitions for thereference standards. For example, Kalter et al.15 developed referencestandard criteria based on physicians’ diagnoses combined withobjective illness signs and laboratory and radiologic findings,whereas Setal et al.29 provided physicians with ICD-10 codes toidentify reference standard cases. Table 2 illustrates the variationsin the reference standard diagnoses used for the three primarycauses of neonatal deathFsepsis, prematurity, and birth asphyxia.

Validation studies have also used different methods to assign theVA cause of death. Kalter et al.15 applied computer algorithmsbased on the closed-ended questions with the highest specificityand sensitivity profiles for each cause of death. In Pakistan, Marshet al. had physicians review the open history alone, the openhistory plus closed-ended modules, and then the closed-endedmodules alone. The open history was reviewed using a ‘sign andsymptom duration matrix’ to guide the analysis, and diagnosesmade from the closed-ended modules were based on predeterminedfield case definitions or algorithms.7 Setal et al.29 employed

Table 1 Sensitivity and specificity of validated algorithms for stillbirths andneonatal causes of death

Sensitivity (%) Specificity (%) Cases (n) Reference country

Stillbirths

80 76 135 Iriya (2002)14 Tanzania

61 84 243 Setal (2006)29 Tanzania

Neonatal tetanus

84 99 19 Marsh et al.7 Pakistan

90 79 30 Snow (1992)30 Kenya

83 89 20 Kalter (1999)15 Bangladesh

Prematurity



48 95 41 Setal (2006)29 Tanzania

Sepsis



61 81 16 Snow (1992)30 Kenya


Birth asphyxia





189


physician review of the closed-ended modules and assignment ofdiagnoses based on current ICD-10 codes.

Findings from validation studies demonstrate high sensitivityand specificity for neonatal tetanus. However, algorithms forprematurity, neonatal sepsis and birth asphyxia, the top threecauses of neonatal deaths, have variable validity. In Bangladesh,the VA algorithm for preterm, ‘pregnancy ended early or babysmall or smaller than usual at birth,’ was validated against thereference standard with high sensitivity (97%), but lower specificity(65%).15 In part, this may be because the algorithm leaves roomfor subjectivity. A woman who is primigravida or living in an areawith a high proportion of low birth weight (LBW) is unlikely toknow if her baby was ‘smaller than usual at birth’.

All validation studies use hospital diagnoses to validatealgorithms used in the community. Cases presenting in a hospital,however, are likely to have different socioeconomic status, health-seeking behaviors, disease symptomatology and possibly cause-of-death distribution than those dying without effective treatment incommunity. This raises further questions about the generalizabilityof findings from validation studies.38

Lack of validated algorithmsStillbirths account for an additional, estimated 3.2 million deathsglobally each year,39 however, there have been no well-validatedalgorithms to identify stillbirths from a VA interview, let alonespecific causes of stillbirths. Setal et al.29 included stillbirths in

Table 2 Comparison of reference standard diagnoses for neonatal sepsis, prematurity/LBW, birth asphyxia from validation studies

Kalter et al. (1999)15 Marsh et al.7 Setal et al. (2006)29

Neonatal sepsis Rectal temperature >38 or <36 1C, plus:

Positive blood culture, plus:

Nonconsolable irritability, abnormally

sleepy or difficult to wake, mottled and

cool extremities, or pale and shocky on

examination

minus

reference standard diagnosis of

pneumonia, bacterial meningitis, acute or

persistent diarrhea and local bacterial

infection

Death after 1 day, plus:

Any two symptoms of jaundice,

fever or hypothermia, convulsions

or vomiting

ICD-10 Code P36a

Bacterial sepsis of newborn (congenital septicemia)

Sepsis of newborn due to streptococcus, group B

Sepsis of newborn due to other and unspecified streptococci

Sepsis of newborn due to Staphylococcus aureus

Sepsis of newborn due to other and unspecified staphylococci

Sepsis of newborn due to Escherichia coli

Sepsis of newborn due to anaerobes

Other bacterial sepsis of newborn

Bacterial sepsis of newborn, unspecified

Prematurity/low

birth weight

Medically documented preterm birth

(gestational age <37 weeks)

Weight for age Z-score p3 on admission

to hospital

Pregnancy <37 weeks gestation

according to Ballard criteria

Weight <2500 g at birth

ICD-10 codes P05:P07a

Slow fetal growth and fetal malnutrition

Light for gestational age (usually referred to as weight below but

length above 10th centile for gestational age. Light-for-dates

Disorders related to short gestation and LBW, not elsewhere

classified

Extremely LBW (birth weight p999 g)

Other LBW (birth weight 1000–2499 g)

Extreme immaturity [<28 completed weeks (<196 completed

days) of gestation]

Other preterm infants (28 weeks or more but <37 completed

weeks (196 completed days but <259 completed days) of

gestation)

Prematurity NOS

Birth asphyxia Medical history of birth asphyxia,

evidenced by documented failure to

breathe spontaneously at birth or 20-min

Apgar score <4, plus:

either lethargy, coma, hypotonia or

seizures on examination, plus:

38 1C<rectal temperature never <36 1C

Persistent central nervous system

depression, with or without a

history of difficult labor or

delivery, in the absence of

hypoglycemia or demonstrated

infection

ICD-10 codes P20:P24a

Intrauterine hypoxia (includes: abnormal fetal heart rate,

fetal or intrauterine: acidosis, anoxia, asphyxia, distress,

hypoxia, meconium in liquor passage of meconium)

Birth asphyxia

Respiratory distress of newborn

Congenital pneumonia due to viral agent

Neonatal aspiration syndromes (includes: neonatal

pneumonia resulting from aspiration)

aWorld Health Organization (1993) International Statistical Classification of Diseases and Related Health Problems, 10th Revision, vol. 2. World Health Organization: Geneva,Switzerland.


190


their validation study; however, VA underestimated the stillbirthrate in this setting with relatively low sensitivity but good specificity(61 and 84% respectively). Iriya et al. also included stillbirths intheir validation study and reported 80% sensitivity and 76%specificity,14 illustrating the need for additional research onstillbirth modules for use in VA. In addition, subcategorization ofstillbirths according to their cause, such as infection orintrapartum hypoxia, is needed to help understand stillbirths as acause of death. Stillbirths pose an additional challenge asmisclassification can occur between stillbirths and early neonataldeaths, and between miscarriage and stillbirth.1 In some settings,the underlying stigma of stillbirths may lead to underreporting.40

In other settings, birth attendants may overreport stillbirths toprotect their reputation as good delivery attendants in thecommunity41,42 or to avoid investigation procedures required forneonatal deaths.29

In addition to the lack of a validated algorithm for stillbirths,deaths due to some causes are often included as a subset of anothercondition due to a lack of cases. For example, birth injury is oftengrouped with birth asphyxia,12,18,21,32 Thus, the data for a singlecategory of death may reflect more than one distinct condition withdifferent risk factors. Similarly, deaths attributed to prematuritymay involve a variety of disease processes including respiratorydistress syndrome (RDS), intracranial hemorrhage, necrotizingenterocolitis, hypothermia and hypoglycemia, for which differentinterventions are needed. For many subcategories of prematurity,clear case definitions for use in VA have not been identified. Amongthese, RDS may be the most distinctive and thus amenable toidentification with VA.

Poor existing algorithmsExisting algorithms remain poor, in part, due to overlapping signsand symptoms for many causes of neonatal deaths. Overlap of signsbetween pneumonia, sepsis and meningitis is common innewborns43 and Kalter et al.15 reported that ‘convulsions,’ a sign ofneonatal tetanus in VA, was also associated with other newborncauses of death such as birth asphyxia and neonatal sepsis. Othersigns like ‘fever’ and ‘difficulty breathing’, associated withpneumonia and sepsis, may also be associated with other causes liketetanus. In a WHO memorandum, experts deduced that based on thefew available validation studies, ‘it would be almost impossible todistinguish between sepsis and pneumonia in the newborn based onverbal autopsy’.44 Overlapping signs and symptoms relate to otherchallenges in neonatal VA such as misclassification of causes,assigning multiple causes and differentiating between direct,underlying and contributory causes of death.

Lack of standardized death classification terminologyTerminology used to classify cause of death has not been clearlystandardized and makes cause of death identification confusing.Most current VA studies identify the direct cause of death, however,

some include antecedent or underlying causes. Trying todifferentiate between direct, antecedent, and contributory causesremains a challenge, given the lack of standardized vocabulary andreporting in the literature. Currently, there are six recognized directcauses of neonatal deaths identifiable by VA: (1) serious infection(including sepsis, pneumonia), (2) birth asphyxia, (3)prematurity, (4) tetanus, (5) congenital malformation and (6)diarrhea.36,45 The ICD-10 specifies the listing of the followinginformation on causes of death on the standard International Formof Medical Certification of Death:46 (1) disease or condition directlyleading to death, generally known as the direct cause of death, (2)antecedent causes (morbid conditions, if any, giving rise to thedirect cause), otherwise known as the underlying cause(s), and (3)other significant conditions contributing to the death, but notrelated to the disease or conditions causing it, also known as thecontributory cause(s).47 These definitions can be difficult tointerpret when we consider causes of neonatal death. For example,in the case of a premature, LBW infant dying of RDS, the deathmight be classified as follows: direct cause of death-RDS;antecedent (or underlying) cause of death-prematurity;contributory cause of death-LBW. However, given that RDS is notone of the six direct causes of neonatal deaths recognized byCHERG (that is prematurity not RDS), another classification mightidentify prematurity as the direct cause.

In Jordan, ICD-10 codes were assigned to 128 infant andneonatal deaths. Contributory causes were identified in 50% ofdeaths, and prematurity was identified as the contributory cause in38% of deaths.48 Conditions often associated with prematurity suchas RDS, necrotizing enterocolitis and others were grouped into acategory, ‘conditions originating in the perinatal period,’ whichwas the most common cause of neonatal death in this sample.48 Inthis example, the authors considered prematurity a contributorycause; however, according to the ICD-10 definitions, contributorycauses are ‘not related to the disease causing it (death)’. Thisexample highlights the confusion in terminology regarding theclassification of neonatal deaths, particularly prematurity. Inaddition, although the International Form of Medical Certificationof Death is most commonly used for all death certificates, there isanother death certificate dedicated for use in cases of stillbirths anddeaths occurring within 168 h (1 week) from birth, called theCertificate of Cause of Perinatal Death.47 This form differs slightlyfrom the standard death certificate in that it considers maternalfactors in addition to the ‘main’ and ‘other’ diseases or conditionsof the infant. This additional form uses slightly differentterminology and its use is not standardized, adding to theconfusion on how early neonatal deaths should be classified andthe terminology used. In addition to the confusion withterminology, many physicians may not be properly trained oncompleting death certificates. This lack of standardized terminologyand the challenges of death certification, including physiciantraining, reflect limitations even among vital registration systems.


191


Limitations of hierarchy to assign causes of deathUse of a hierarchy has limitations, as the cause of deathdistribution can vary widely depending on how the hierarchy isselected. Using the CHERG hierarchy which was used to estimate 4million causes of neonatal deaths36 (congenital abnormality-neonatal tetanus-preterm-birth asphyxia-sepsis/pneumonia-diarrhea-other), Baqui et al.12 found a similarcause of death distribution in India. However, using the same data,and reordering the hierarchy by placing prematurity last, we foundthat the percentage of deaths due to preterm birth decreased from27 to 9%. When sepsis/pneumonia was placed first in the hierarchy,the percentage of deaths due to sepsis increased from 24 to 42%(all distributions; Figure 1). This highlights the variation inreporting that can occur depending on where each cause of deathis placed in the hierarchy.

Other studies have applied different hierarchies, resulting indifferent cause-of-death distributions. For example, the recentlypublished Nepal 2006 Demographic Health Survey used a hierarchy(neonatal tetanus, congenital abnormality, birth asphyxia, birthinjury, diarrhea, acute respiratory infection, possible seriousinfection, preterm/LBW, and so on) which resulted in significantlylower estimates for prematurity (6% compared with 27% as reportedby Lawn et al.1).

Implications of current VA methods and limitationsGiven the use of VA data for policy and program development, it isimportant explore to new methods to assign causes of death thatovercome limitations in conventional VA methods. The influence ofthe hierarchy to assign cause of death, for example, has importantimplications. Those with specific disease interests may apply ahierarchy that results in a cause of death distribution with a largernumber of deaths attributed to their disease of interest. Forconditions such as birth asphyxia and serious infections, for whicheffective interventions have been proven to reduce mortality, theuse of a hierarchy that illustrates their importance in neonatalcause of death distributions can help promote action.49

Assigning multiple causes of death may alleviate this issue ofassigning precedence to particular causes of death. In fact,researchers have recommended identifying multiple causes of deathas many childhood deaths result from more than one cause andmay be prevented if one of the causes is treated.38 However,assigning multiple causes of death also presents challenges. Marshet al. reported some differences in cause-of-death structure whenmultiple causes were assigned. For example, LBW increased from26% of deaths when a single cause was assigned to 39% whenmultiple causes were allowed, but birth asphyxia showed littlechange, from 14 to 15%.24 Physician assignment of multiple causesfrom VA may lack sensitivity due to the need for consensus amongseveral physicians. Snow et al.30 found that 27% of neonates werediagnosed with two causes of death based on hospital records,compared with only 7% assigned two causes through physicianreview of VA data. Assigning multiple causes can also result inoverlapping causes of death. In Nepal, Lee et al.35 reported that29% of birth asphyxia deaths were classified as being prematureand 42% also met criteria for serious infections. Baqui et al.12

reported similar findings in India, where 23% of deaths attributedto sepsis or pneumonia also met criteria for preterm birth.

New methods to measure causes of neonatal deathNew probabilistic models with the potential to overcome many ofthe biases of hospital-based validation studies and that allow formultiple causes of death have been used to analyze VA data.50–52

Fantahun et al. compared physician review of VA using ICD-10codes to identify main and underlying causes of death with theInterVA probabilistic model. In this method, signs and symptoms,including patient history, were extracted from the VA closed-endedand open-history modules. Indicators were loaded into a computerdatabase and the model was run to allow up to three causes ofdeath, each with a probability estimate. Though they did notexamine neonatal deaths exclusively, for deaths in infants under 1year of age, physician review and the InterVA method resulted insimilar cause-of-death estimates with the exception of ‘perinatalproblems,’ which physicians reported more often than the InterVAmodel.51 King and Ying52 have also proposed a probabilisticmodel using symptom profiles to determine the mortality fractionsfor all causes of death in the community at once. In this model,multiple causes for an individual are handled by joining two ormore causes together into a single category. Byass et al. reportedanother model based on Bayes theorem that identified variousdisease indicators and defined the probability of a particularcause based on the presence of specific indicators. This studyreported consensus for 75% of cases between the model-assignedand physician review-assigned causes of death.50 Finally,combining the methods of King and Byass, Murray et al.53

validated the ‘Symptom Pattern’ (SP) method with the standardphysician coded VA method and showed that SP correctly estimatedcause specific mortality fractions with less error than physician

0%

10%

20%

30%

40%

50%

Conge

nital

Anom

aly NNT

Prete

rm BA

Sepsis

Diarrh

ea

Uniden

tified

Lawn (2005)

Baqui (2006)

Baqui (2006) PretermLastBaqui (2006) SepsisFirst

Figure 1 Changes in cause of neonatal death distributions based on modifiedhierarchies.


192


coded VA at both the population and individual level. Thesemethods have advantages in that they do not rely on algorithms,require less time to analyze and do not require the time, effort andcost of physician reviewers. It is still unclear how these methodswill vary across cultural and language barriers, however, validationstudies are currently being conducted in multisite global fieldsettings (www.gcgh.org).

Conclusion

International standards for VA methodology have beenrecommended.54 These include improving and standardizing theVA questionnaire, applying methods for cause of death certificationand coding of VA according to current ICD-10 classifications, anddeveloping a cause of death list for VA according to the ICD-10.

With respect to neonatal deaths, additional methods to improveVA should be explored. The use of a social autopsy to address issuesof care seeking should be incorporated. Research about the stigmabehind stillbirths is needed to determine culturally sensitive ways toascertain the burden of stillbirths and develop specific VAalgorithms.

A standardized method of assigning causes of death should beadopted. Physician review can leave room for subjectivity,depending on who is making the diagnosis; and computeralgorithms used vary considerably.

The use of a hierarchy to assign a single cause of death imposesfurther limitations, as a standard hierarchy has not been agreedupon, resulting in varied cause of death distributions depending onwhich hierarchy is used. Further research and clear guidelines areneeded to define a standard approach to the use of a hierarchy toassign single cause-of-death estimates. The ability to assignmultiple causes is also important. Many newborns die frommultiple causes, and efforts to distinguish between direct,antecedent or underlying, and contributory causes should also beexplored.

Larger validation studies of stillbirths and neonatal deaths inareas with diverse disease mixes should be conducted usingstandardized VA instruments and methods to help develop newalgorithms and improve and validate existing algorithms. Thiscould lead to improvements in assigning cause of death usingcomputer algorithms. In addition, attempts should be made toconduct validation studies that do not rely solely on hospital goldstandard diagnoses, which may reflect a biased cause of deathprofile. New probabilistic models for assigning cause of death fromVA data overcome many of the biases of hospital-based validationstudies, and also may improve the ascertainment of multiple andunderlying causes of death. One possible method to address thisissue is to create cause of death categories with multiple causes, forexample pretermþ birth asphyxia.

Given the implications for future programs and policies, it isimportant to further explore how VA methods can be improved

and/or explore other means of obtaining accurate cause of deathestimates for newborn infants in settings with inadequateregistration of vital events.

References

1 Lawn JE, Cousens S, Zupan J. 4 million neonatal deaths: when? Where? Why? Lancet

2005; 365(9462): 891–900.

2 Awasthi S, Pande VK, Glick H. Under fives mortality in the urban slums of Lucknow.

Indian J Pediatr 1996; 63(3): 363–368.

3 Morris SS, Black RE, Tomaskovic L. Predicting the distribution of under-five deaths by

cause in countries without adequate vital registration systems. Int J Epidemiol 2003;

32(6): 1041–1051.

4 Nykanen M, Tamaona W, Cullinan T, Van Oosterzee V, Ashorn P. Verbal autopsy as a

technique to establish causes of infant and child mortality. East Afr Med J 1995;

72(11): 731–734.

5 Yassin KM. Indices and sociodemographic determinants of childhood mortality in rural

Upper Egypt. Soc Sci Med 2000; 51(2): 185–197.

6 Mirza NM, Macharia WM, Wafula EM, Agwanda RO, Onyango FE. Verbal autopsy: a

tool for determining cause of death in a community. East Afr Med J 1990; 67(10):

693–698.

7 Marsh DR, Sadruddin S, Fikree FF, Krishnan C, Darmstadt GL. Validation of verbal

autopsy to determine the cause of 137 neonatal deaths in Karachi, Pakistan. Paediatr

Perinat Epidemiol 2003; 17(2): 132–142.

8 Spika JS, Munshi MH, Wojtyniak B, Sack DA, Hossain A, Rahman M et al. Acute lower

respiratory infections: a major cause of death in children in Bangladesh. Ann Trop

Paediatr 1989; 9(1): 33–39.

9 Terra de Souza AC, Peterson KE, Andrade FM, Gardner J, Ascherio A. Circumstances of

post-neonatal deaths in Ceara, Northeast Brazil: mothers’ health care-seeking

behaviors during their infants’ fatal illness. Soc Sci Med 2000; 51(11): 1675–1693.

10 Hammer GP, Some F, Muller O, Kynast-Wolf G, Kouyate B, Becher H. Pattern of cause-

specific childhood mortality in a malaria endemic area of Burkina Faso. Malar J 2006;

5: 47.

11 Bakshi R. UNICEF unveils new tool to combat maternal mortality in India. In New

Delhi: UNICEF 2006.

12 Baqui AH, Darmstadt GL, Williams EK, Kumar V, Kiran TU, Panwar D et al. Rates,

timing and causes of neonatal deaths in rural India: implications for neonatal health

programmes. Bull World Health Organ 2006; 84(9): 706–713.

13 Chowdhury ME, Akhter HH, Chongsuvivatwong V, Geater AF. Neonatal mortality in

rural Bangladesh: an exploratory study. J Health Popul Nutr 2005; 23(1): 16–24.

14 Iriya N, Manji KP, Mbise RL. Verbal autopsy in establishing cause of perinatal death.

East Afr Med J 2002; 79(2): 82–84.

15 Kalter HD, Hossain M, Burnham G, Khan NZ, Saha SK, Ali MA et al. Validation of

caregiver interviews to diagnose common causes of severe neonatal illness. Paediatr

Perinat Epidemiol 1999; 13(1): 99–113.

16 Baqui AH, Black RE, Arifeen SE, Hill K, Mitra SN, al Sabir A. Causes of childhood

deaths in Bangladesh: results of a nationwide verbal autopsy study. Bull World Health

Organ 1998; 76(2): 161–171.

17 Baqui AH, Sabir AA, Begum N, Arifeen SE, Mitra SN, Black RE. Causes of childhood

deaths in Bangladesh: an update. Acta Paediatr 2001; 90(6): 682–690.

18 Fikree FF, Azam SI, Berendes HW. Time to focus child survival programmes on the

newborn: assessment of levels and causes of infant mortality in rural Pakistan. Bull

World Health Organ 2002; 80(4): 271–276.

19 Shrivastava SP, Kumar A, Kumar Ojha A. Verbal autopsy determined causes of neonatal

deaths. Indian Pediatr 2001; 38(9): 1022–1025.

20 Bhandari N, Bahl R, Taneja S, Martines J, Bhan MK. Pathways to infant mortality in

urban slums of Delhi, India: implications for improving the quality of community- and

hospital-based programmes. J Health Popul Nutr 2002; 20(2): 148–155.


193


www.gcgh.org

21 Garrib A, Jaffar S, Knight S, Bradshaw D, Bennish ML. Rates and causes of child

mortality in an area of high HIV prevalence in rural South Africa. Trop Med Int Health

2006; 11(12): 1841–1848.

22 Hinderaker SG, Olsen BE, Bergsjo PB, Gasheka P, Lie RT, Havnen J et al.

Avoidable stillbirths and neonatal deaths in rural Tanzania. BJOG 2003; 110(6):

616–623.

23 Khoury SA, Massad D, Fardous T. Mortality and causes of death in Jordan 1995–96:

assessment by verbal autopsy. Bull World Health Organ 1999; 77(8): 641–650.

24 Marsh DR, Husein K, Lobo M, Shah MSL. Verbal autopsy in Karachi slums:

comparing single and multiple causes of child deaths. Health Policy Plan 1995;

10(4): 395–403.

25 Adazu K, Lindblade KA, Rosen DH, Odhiambo F, Ofware P, Kwach J et al. Health and

demographic surveillance in rural western Kenya: a platform for evaluating

interventions to reduce morbidity and mortality from infectious diseases. Am J Trop

Med Hyg 2005; 73(6): 1151–1158.

26 Freeman JV, Christian P, Khatry SK, Adhikari RK, LeClerq SC, Katz J et al. Evaluation of

neonatal verbal autopsy using physician review versus algorithm-based cause-of-death

assignment in rural Nepal. Paediatr Perinat Epidemiol 2005; 19(4): 323–331.

27 Nongkynrih B, Anand K, Kapoor SK. Use of verbal autopsy by health workers in under-

five children. Indian Pediatr 2003; 40(8): 766–771.

28 Etard JF, Le Hesran JY, Diallo A, Diallo JP, Ndiaye JL, Delaunay V. Childhood mortality

and probable causes of death using verbal autopsy in Niakhar, Senegal, 1989–2000.

Int J Epidemiol 2004; 33(6): 1286–1292.

29 Setal PW, Whiting DR, Hemed Y, Chandramohan D, Wolfson LJ, Alberti KG et al.

Validity of verbal autopsy procedures for determining cause of death in Tanzania. Trop

Med Int Health 2006; 11(5): 681–696.

30 Snow RW, Armstrong JR, Forster D, Winstanley MT, Marsh VM, Newton CR et al.

Childhood deaths in Africa: uses and limitations of verbal autopsies. Lancet 1992;

340(8815): 351–355.

31 Ministry of Health and Population (MOHP) N, New ERA and Macro International Inc.

Nepal Demographic and Health Survey 2006. Kathmandu, Nepal; 2007 2007.

32 Campbell O, Gipson R, el-Mohandes A, Issa AH, Matta N, Mansour E et al. The Egypt

National Perinatal/Neonatal Mortality Study 2000. J Perinatol 2004; 24(5): 284–289.

33 Perry HB, Ross AG, Fernand F. Assessing the causes of under-five mortality in the Albert

Schweitzer Hospital service area of rural Haiti. Rev Panam Salud Publica 2005;

18(3): 178–186.

34 Whiting DR, Setal PW, Chandramohan D, Wolfson LJ, Hemed Y, Lopez AD. Estimating

cause-specific mortality from community- and facility-based data sources in the United

Republic of Tanzania: options and implications for mortality burden estimates. Bull


35 Lee AC, Mullany LC, Tielsch JM, Katz J, Khatry SK, LeClerq SC et al. Verbal autopsy

methods to ascertain birth asphyxia deaths in a community-based setting in Southern

Nepal. Pediatrics 2008 (under review).

36 Lawn JE, Wilczynska-Ketende K, Cousens SN. Estimating the causes of 4 million

neonatal deaths in the year 2000. Int J Epidemiol 2006; 35(3): 706–718.

37 Soleman N, Chandramohan D, Shibuya K. Verbal autopsy: current practices and

challenges. Bull World Health Organ 2006; 84(3): 239–245.

38 Anker M, Black RE, Coldham C, Kalter HD, Quigley M, Ross D. A Standard Verbal

Autopsy Method for Investigating Causes of Death in Infants and Children. World

Health Organization: Geneva, 1999.

39 Stanton C, Lawn JE, Rahman H, Wilczynska-Ketende K, Hill K. Stillbirth rates:

delivering estimates in 190 countries. Lancet 2006; 367(9521): 1487–1494.

40 Hsu MT, Tseng YF, Banks JM, Kuo LL. Interpretations of stillbirth. J Adv Nurs 2004;

47(4): 408–416.

41 Benara SK, Singh P. Validity of causes of infant death by verbal autopsy. Indian J

Pediatr 1999; 66(5): 647–650.

42 Datta N, Mand M, Kumar V. Validation of causes of infant death in the community by

verbal autopsy. Indian J Pediatr 1988; 55(4): 599–604.

43 Darmstadt GL, Black RE, Santosham M. Research priorities and postpartum care

strategies for the prevention and optimal management of neonatal infections in less

developed countries. Pediatr Infect Dis J 2000; 19(8): 739–750.

44 WHO/UNICEF Informal Consultation. Measurement of overall and cause-specific

mortality in infants and children: memorandum from a WHO/UNICEF meeting. Bull


45 Jamison DT, Shahid-Salles SA, Jamison J, Lawn JE, Zupan J. Incorporating deaths near

the time of birth into estimates of the global burden of disease. In: Lopez AD, Mathers

CD, Ezzati M, Jamison DT, Murray CJL (eds). Global Burden of Disease and Risk

Factors, ed.. Oxford University Press: New York, 2006, pp 427–463.

46 WHO. International Statistical Classification of Diseases and Related Health

Problems, 10th revision, vol. 2. World Health Organization: Geneva, Switzerland.

47 WHO. ICD-10: International Statistical Classification of Diseases and Related Health

Problems, 10th revision, 2nd edn. 2004.

48 Khoury SA, Mas’ad DF. Causes of infant mortality in Jordan. Saudi Med J 2002; 23(4):

432–435.

49 Darmstadt GL, Bhutta ZA, Cousens S, Adam T, Walker N, de Bernis L. Evidence-based,

cost-effective interventions: how many newborn babies can we save? Lancet 2005;

365(9463): 977–988.

50 Byass P, Fottrell E, Dao LH, Berhane Y, Corrah T, Kahn K et al. Refining a probabilistic

model for interpreting verbal autopsy data. Scand J Public Health 2006; 34(1):

26–31.

51 Fantahun M, Fottrell E, Berhane Y, Wall S, Hogberg U, Byass P. Assessing a new

approach to verbal autopsy interpretation in a rural Ethiopian community: the InterVA

model. Bull World Health Organ 2006; 84(3): 204–210.

52 King G, Ying L. Verbal autopsy mehtods with mulitple causes of death. Stat Sci 2008;

23(1): 78–91.

53 Murray CJL, Lopez AD, Feehan D, Peter ST, Yang G. Validation of the symptom

pattern method for analyzing verbal autopsy data. PLOS Med 2007; 4(11):

1739–1753.

54 Baiden F, Bawah A, Biai S, Binka F, Boerma T, Byass P et al. Setting international

standards for verbal autopsy. Bull World Health Organ 2007; 85(8): 570–571.


194


ORIGINAL ARTICLE

Digibind attenuates cytokine TNFa-induced endothelialinflammatory response: potential benefit role of Digibind inpreeclampsiaY Wang1, DF Lewis1, CD Adair2, Y Gu1, L Mason2 and JH Kipikasa2

1Department of Obstetrics and Gynecology, Louisiana State University Health Sciences Center, Shreveport, LA, USA and 2Section of MaternalFetal Medicine, Department of Obstetrics and Gynecology, University of Tennessee College of Medicine, Chattanooga, TN, USA

Objective: Exaggerated inflammatory response occurs in preeclampsia.

Preeclampsia is also associated with elevated endogenous digoxin-like

factors (EDLFs). Clinical data suggest that Digibind (a polyclonal sheep

digoxin binding Fab fragment) binds to EDLF and may have the potential

to attenuate vasoconstriction and other clinical symptoms of preeclampsia.

This study was undertaken to determine if Digibind could attenuate

increased endothelial inflammatory response induced by tumor necrosis

factor-a (TNFa).

Study Design: Confluent endothelial cells were treated with TNFa at

different concentrations with or without Digibind in culture. Endothelial

adhesion molecule ICAM, VCAM and E-selectin expressions were

determined by an immunoassay directly detected on the endothelial

surface. Effects of Digibind on TNFa-induced extracellular signal-

regulated kinase and Naþ /Kþ -ATPase expressions were also examined.

Result: (1) TNFa induced dose-dependent increases in ICAM, VCAM and

E-selectin expressions in endothelial cells; (2) Digibind could attenuate and

reduce TNFa-induced upregulation of endothelial E-selectin, ICAM and

VCAM expressions. The blocking effect was in a concentration dependent

manner; (3) Digibind had no effects on TNFa-induced upregulation of

extracellular signal-regulated kinase phosphorylation, but could block

TNFa-induced downregulation of Naþ /Kþ -ATPase b1 expression.

Conclusion: Digibind may exert beneficial effects by preserving cell

membrane Naþ /Kþ -ATPase function and consequently to offset

increased inflammatory response in endothelial cells.


published online 15 January 2009

Keywords: Digibind; endothelial cells; Naþ /Kþ -ATPase; preeclampsia;inflammatory response

Introduction

It was reported that circulating levels of endogenous digoxin-likefactor(s) such as ouabain, bufadienolide, marinobufagenin, andcardenolide were elevated in women with preeclampsia,1–3 ahypertensive and multiple system disorder unique to humanpregnancy. Studies have shown that digoxin-like factors extractedfrom plasma from women with preeclampsia could inhibiterythrocyte Naþ /Kþ -ATPase activity3 and Naþ /Kþ -ATPasepurified from human mesenteric arteries.2 Inhibition of the sodiumpump could result in an increase in intracellular calcium levelsand lead to vasoconstriction in the systemic vasculature. Therefore,the vasoconstrictive property of digoxin-like factors is believed tocontribute to maternal hypertension in preeclampsia.

Endothelial activation/dysfunction is a centralpathophysiological feature in the maternal vascular system inpreeclampsia.4 Pregnancy is also an inflammatory state andpreeclampsia is considered to be an exaggerated inflammatoryresponse during pregnancy.5 It is believed that altered endothelialfunction constitutes the exaggerated inflammatory response inthis pregnancy disorder,5 which includes activation of leukocytesand platelets, increased circulating cytokine levels of tumornecrosis factor-a (TNFa), interferon-g and interleukin-66,7 andincreased endothelial adhesion molecule levels such asintercellular adhesion molecule (ICAM) and vascular cell adhesionmolecule (VCAM).8,9

Digibind is a polyclonal-fragmented Digoxin-immune Fabantibody raised in sheep. Previous published work has shown thatin vitro treatment of erythrocytes from preeclamptic patients withDigibind could restore the cell Naþ /Kþ -ATPase activity.3

Furthermore, administration of Digibind to both antepartum andpostpartum women with preeclampsia could improve maternalsymptoms and increase fetoplacental perfusion10,11 (Dr Adair’sunpublished data), which suggest that Digibind could be apotential therapy for preeclampsia. To study if Digibind exertsbeneficial effects on endothelial cells, we examined the role ofDigibind in TNFa-induced inflammatory response in endothelialcells. Endothelial surface adhesion molecule expressions ICAM,

Received 17 April 2008; revised 15 November 2008; accepted 21 November 2008; published

online 15 January 2009

Correspondence: Dr Y Wang, Louisiana State University Health Sciences Center, Department

of Obstetrics and Gynecology, PO Box 33932, Shreveport, LA 71130, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.222



VCAM and E-selectin were used as the endpoint readout. Effects ofDigibind on endothelial extracellular signal-regulated kinases(ERKs) and Naþ /Kþ -ATPase expressions affected by cytokineTNFa were also examined.

MethodsEndothelial isolation and cultureHuman umbilical vein endothelial cells were isolated bycollagenase digestion as previously described.12 Umbilical cordswere collected from normal pregnant women after delivery atLouisiana State University Health Sciences Center in Shreveporthospital. Normal pregnancy was defined as a pregnancy in whichthe mother had normal blood pressure (p140/90 mm Hg),absence of medical and obstetrical complications. This study wasapproved by the Institutional Review Board for Human Research atLSUHSC-Sh, LA.

Isolated cells were incubated with endothelial cell growthmedium (BioWhittaker Inc., Walkersville, MD, USA). Only the first-passage (P1) endothelial cells were used in this study. Cells usedfor adhesion molecule expression experiments were grown in 48wells per plate and cells used for protein extraction were grown in25 cm2 culture flasks. Confluent endothelial cells were treated withTNFa (Sigma, St Louis, MO, USA) or combined with Digibind(GlaxoSmithKline, Research Triangle Park, NC, USA).

Endothelial surface molecule expression assayCellular surface molecule expressions for ICAM, VCAM andE-selectin were determined as we previously described.12 Briefly,after endothelial cells were treated with TNFa or combined withDigibind in culture, cells were fixed with 1% paraformaldehyde andthen incubated with a primary antibody (mouse anti-human) toICAM-1 (CD54), VCAM-1 (CD106) or E-selectin (CD62E),respectively. Horseradish peroxidase-goat anti-mouseimmunoglobulin G (Sigma) was used as the secondary antibody.Hydrogen peroxide (0.003%) and 3,30,5,50-tetramethybenzidine(TMB) (0.1 mg ml�1) were used as substrate and color generation.The reaction was terminated by 8 N H2SO4. Cells that reacted withsecondary antibody only were used as background. After reaction,plates were read at 450 nm by an autoplate reader (MolecularDevices, Sunnyvale, CA, USA). All samples were tested in triplicate.

Western blot analysisAt the end of each experiment, total cellular protein was extractedwith an ice-cold lysis buffer that contained 50 mmol l�1

Tris-HCl (pH7.6), 1% Triton X-100, 0.5% NP-40, 1 mmol l�1

phenylmethylsufonyl fluoride and 0.5%mmol l�1 dithiotheritol.The lysate was centrifuged at 14 000 r.p.m. at 4 1C for 15 min toremove insoluble materials. All samples were stored at �70 1C. Thetotal endothelial cell protein extract (10mg per sample) wassubjected to electrophoresis on 12% polyacrylamide gels by using

the Mini-protein 3 gel running system (Bio-Rad, Hercules, CA,USA) and then transferred to nitrocellulose membrane. Themembranes were probed with a primary monoclonal antibodyagainst ERK (Santa Cruz, San Diego, CA, USA), pERK (SantaCruz), Naþ /Kþ -ATPase b1 (Santa Cruz) or b-actin (Sigma).The secondary antibody was horseradish-linked anti-mouseantibody. The bound antibodies were visualized with an enhancedchemiluminescent deletion Kit (Amersham Corp., ArlingtonHeights, IL, USA). Nitrocellulose membranes were stripped andblocked before they were probed again with different primaryantibodies.

Statistical analysisData are presented as mean±s.e. Statistical analysis wasperformed with analysis of variance by a computer softwareprogram StatView (SAS Institute Inc., Cary, NC, USA). A probabilitylevel less than 0.05 was considered statistically significant.

ResultsDigibind attenuates TNFa-induced endothelial surface adhesionmolecule expressionsEndothelial inflammatory response was induced by cytokine TNFa.Confluent endothelial cells were treated with TNFa atconcentrations of 1, 10 and 100 pg ml�1 for 2 h, then endothelialadhesion molecule ICAM, VCAM and E-selectin expressions weredetermined. TNFa at a concentration of 100 pg ml�1 was relativecompatible with the TNFa levels in the maternal plasma in womenwith preeclampsia.13 Figure 1 shows dose-dependent increase inendothelial ICAM, VCAM and E-selectin expressions induced byTNFa.

TNFa at lower dose (1 pg ml�1) had no effects on endothelialICAM, VCAM and E-selectin expressions compared to the controls,ICAM: 0.645±0.028 vs 0.564±0.036; VCAM: 0.101±0.005 vs0.090±0.002; and E-selectin: 0.073±0.011 vs 0.070±0.014,respectively. Endothelial ICAM, VCAM and E-selectin expressionswere significantly increased when TNFa concentrations were usedat 10 pg ml�1 (P<0.05) and 100 pg ml�1 (P<0.01)FICAM:1.261±0.067 and 1.789±0.143; VCAM: 0.236±0.032 and0.663±0.072; and E-selectin: 0.128±0.020 and 0.345±0.007,respectively.

To determine if Digibind could attenuate TNFa-induced ICAM,VCAM and E-selectin expressions in endothelial cells, cells werepretreated with Digibind for 1 h and then TNFa at a concentrationof 100 pg ml�1 for 2 h. Two concentrations of Digibind (50 and100mg ml�1) were used. Figure 2 shows that Digibind could dose-dependently attenuate TNFa-induced upregulation of endothelialICAM, VCAM and E-selectin expressions, Digibind at50 pg ml�1 þ TNFa and at 100 pg ml�1 þ TNFa vs TNFaaloneFICAM: 1.214±0.103 and 1.074±0.086 vs 1.704±0.147,P<0.01; VCAM: 0.271±0.080 and 0.226±0.069 vs 0.531±0.147,

Digibind in preeclampsiaY Wang et al

196


P<0.05; E-selectin: 0.220±0.034, P<0.05 and 0.150±0.028,P<0.01 vs 0.342±0.034, respectively.

Digibind reduces TNFa-induced endothelial surface adhesionmolecule expressionsTo further determine if Digibind could reduce cytokine TNFainduced endothelial activation, ECs were first treated with TNFa for1 h and then Digibind was added to the cell culture. The cells werecontinuously cultured for 1 h and ICAM and VCAM and E-selectinexpressions were determined. In this experiment, twoconcentrations of TNFa (10 and 100 pg ml�1) were used.Interestingly, endothelial ICAM, VCAM and E-selectin expressionswere downregulated in cells even when Digibind was added to thecell culture after TNFa treatment (Figure 3), addition of Digibind

100 pg ml�1 vs TNFa at 100 pg ml�1 aloneFICAM: 0.896±0.12vs 1.435±0.192, P<0.05; VCAM: 0.413±0.062 vs 0.584±0.071,P<0.05; and E-selectin: 0.256±0.034 vs 0.498±0.086, P<0.01,respectively. These data suggest that Digibind exerts protectiveeffects on endothelial cells against cytokine TNFa inducedendothelial activation.

Digibind attenuates TNFa-induced downregulation ofNaþ /Kþ -ATPase b1 expression, but has no effect on TNFa-induced ERK phosphorylation in endothelial cellsIt is known that cytokine induced inflammatory response isinvolved in extracellular signal-regulated kinases (ERKs)activation. Using ERK expression as a comparison, we examined ifincreased endothelial adhesion molecule expression induced byTNFa is associated with alteration of Naþ /Kþ -ATPase expressionin endothelial cells. Endothelial cells were treated with TNFa atdifferent concentrations and then ERK, pERK and Naþ /Kþ -ATPase expression were examined by western blot analysis.

**

## ##

ICA

M E

xpre

ssio

n(O

D 4

50n

m)

VC

AM

Exp

ress

ion

(OD

450

nm

) **

#

**

#

##E

-sel

ecti

n E

xpre

ssio

n(O

D 4

50n

m)

0.5

0.4

0.3

0.2

0.1

0(-) (-) 50pg ml-1 100pg ml-1

TNFαα 100pg/ml(-) (+) (+) (+)

1.0

0.8

0.5

0.2

0

2.5

2.0

1.5

1.0

0.5

0

Digibind

Figure 2 Digibind-attenuated tumor necrosis factor-a (TNFa)-induced ICAM,VCAM and E-selectin expressions in endothelial cells (ECs). ECs were pretreatedwith Digibind (50 and 100 mg ml�1) for 1 h and then TNFa at a concentration of100 pg ml�1 for 2 h. Digibind could dose-dependently attenuate TNFa-inducedupregulation of ICAM, VCAM and E-selectin expression. Data are means from sixindependent experiments, each in triplicate. **P<0.01: TNFa-treated only vscontrol; #P<0.05 and ##P<0.01: cells treated with Digibindþ TNFa vs TNFa-treated only, respectively.

*

**

VC

AM

Exp

ress

ion

(OD

450

nm

)E

-sel

ecti

n E

xpre

ssio

n(O

D45

0nm

)

0.5

0.4

0.3

0.2

0.1

0

*

**

1.0

0.8

0.5

0.2

0

ICA

M E

xpre

ssio

n(O

D45

0nm

)2.5

2.0

1.5

1.0

0.5

0

*

**

TNFαα Concentration

Control 1pg ml-1 10pg ml-1 100pg ml-1

Figure 1 Tumor necrosis factor-a (TNFa) dose-dependently increased inendothelial ICAM, VCAM and E-selectin expressions. Data are means from sixindependent experiments, each in triplicate. *P<0.05, **P<0.01, respectively.


197


As shown in Figure 4, TNFa induced upregulation of ERKphosphorylation (pERK) was in a dose-dependent manner.In contrast, Naþ /Kþ -ATPase b1 expression was downregulatedin endothelial cells treated with TNFa. The TNFa inducedinhibitory effect on Naþ /Kþ -ATPase b1 expression was also in a

dose-dependent manner. We further determined if Digibind-exertedprotective effects against TNFa-induced downregulation ofNaþ /Kþ -ATPase b1 expression in endothelial cells. As shown inFigure 5a, pretreatment of Digibind did not affect pERK expressionin endothelial cells stimulated with TNFa. In contrast, pre-treatment of Digibind could attenuate TNFa-induced down-regulation of Naþ /Kþ -ATPase b1 expression (Figure 5b).

Discussion

In this study, using TNFa as an inflammatory stimulator weexamined potential beneficial effects of Digibind on vascularendothelial cells. We found that TNFa could not only induce anendothelial inflammatory response as evidenced by upregulation ofendothelial surface adhesion molecule expression and induction ofERK phosphorylation, but also downregulate Naþ /Kþ -ATPaseexpression in endothelial cells. Naþ /Kþ -ATPase is important cellmembrane ionic machinery, which keeps membrane potential in aproper order and maintains intracellular and extracellular ionicbalance. TNFa induced upregulation of endothelial adhesionmolecule ICAM, VCAM and E-selectin expressions, accompanied bydownregulation of Naþ /Kþ -ATPase expression, suggest thatenhanced endothelial inflammatory response may associate withaltered Naþ /Kþ -ATPase activity or function in endothelial cells.

One of the most important findings of our study is that Digibindmay have protective effects to offset endothelial inflammatoryresponse as demonstrated not only by attenuating but also byreducing TNFa-induced increased endothelial inflammatoryresponses determined by E-selectin, ICAM and VCAM expressions onthe endothelial surface. Both in vivo and in vitro studies haveshown that enhanced adhesion molecule expression in endothelialcells is directly related to activation of leukocytes, increasedleukocyte–endothelial adhesion and leukocyte extravasation.14,15

In preeclampsia, maternal TNFa levels, as well as soluble levels ofE-selectin, ICAM and VCAM are elevated compared to normalpregnant controls.9,16 As maternal circulating levels of endogenous

*

**

#

ICA

M E

xpre

ssio

n(O

D 4

50n

m)

VC

AM

Exp

ress

ion

(OD

450

nm

)E

-sel

ecti

n E

xpre

ssio

n(O

D 4

50n

m)

2.0

1.5

1.0

0.5

0

*

**

#

*

**

##

(-) 10pg ml-1 100pg ml-1 TNFαα(-) (+) (-) (+) Digibind 100pg ml-1(-)

0.8

0.6

0.4

0.2

0

0.8

0.6

0.4

0.2

0

Figure 3 Digibind reduced tumor necrosis factor-a (TNFa)-induced ICAM,VCAM and E-selectin expressions in ECs. In this experiment, Digibind was addedto the cell culture 1 h after TNFa treatment. Data are means from six independentexperiments, each in triplicate. *P<0.05 and **P<0.01: TNFa-treated vs controlcells; #P<0.05 and ##P<0.01: cells treated with TNFaþ Digibind vs TNFa-treatedonly, respectively.

ββ-actin

Na+/K+ATPase β1

0 0.1 0.5 1 10 ng ml-1 TNFα

pERK

ERK

Figure 4 Effects of tumor necrosis factor-a (TNFa) on extracellular signal-regulated kinase (ERK), pERK and Naþ /Kþ -ATPase b1 expressions in ECs.b-Actin expression was used as control. TNFa dose-dependently induced pERKupregulation and Naþ /Kþ -ATPase b1 downregulation in ECs. The blots arerepresentative from three independent experiments.

- - + + Digibind- + - + TNF

pERK

- - + Digibind- + + TNF

Na+/K+ATPase 1

Figure 5 Effects of Digibind on pERK and Naþ /Kþ -ATPase b1 expression inECs. Digibind had no effect on tumor necrosis factor-a (TNFa)-inducedupregulation of pERK (a), but could block TNFa-induced downregulation ofNaþ /Kþ -ATPase b1 (b) in ECs. The blots are representative from threeindependent experiments.


198


digoxin-like factor(s) are also increased in women withpreeclampsia,1,2 Digibind may not only be able to neutralizeendogenous digoxin-like factor(s) in the circulation but also exertprotective effects on vasculature by reducing endothelialinflammatory response in preeclampsia.

In this study, we also found that TNFa down-regulatedendothelial Naþ /Kþ -ATPase b1 expression, suggesting thatincreased inflammatory response is accompanied by altered sodiumpump activity or function on the cell membrane. Regarding thefact of reduced Naþ /Kþ -ATPase activity in erythrocytes fromwomen with preeclampsia,1,2 our finding of downregulation ofNaþ /Kþ -ATPase b1 expression induced by TNFa indicated thatincreased endothelial inflammatory response may directly orindirectly influence endothelial Naþ /Kþ -ATPase activity orfunction. This concept is supported by our data that TNFa-induceddownregulation of endothelial Naþ /Kþ -ATPase b1 expressioncould be blocked by pretreatment of endothelial cells with Digibindin culture.

At the present time, we do not know the direct relationship ofupregulation of endothelial adhesion molecule expression todownregulation of Naþ /Kþ -ATPase b1 expression stimulated byTNFa in endothelial cells. Our data showed that upregulation ofE-selectin, ICAM and VCAM expression was related to upregulationof phosphorylated EKR expression in endothelial cells after TNFastimulation, which indicates that TNFa-induced endothelialinflammatory response is an transcriptional factor regulated event,at least in part, through the ERK pathway regulation. However,Digibind could attenuate or reduce TNFa-induced endothelialadhesion molecule expression, but had no effects on TNFa-inducedpERK upregulation. These observations suggest that attenuation ofTNFa-induced endothelial adhesion molecule expression byDigibind is not mediated through the ERK pathway regulation.

This study further supports the potential function of Digibind fora possible clinical application for preeclampsia patients. Therecently reported Digibind Efficacy Evaluation in Preeclampsia(DEEP) study, Gov trial no. NCT00158743, showed improved renalhemodynamic effects of Digibind on preterm severe preeclampsia,which correlated with improved erythrocyte sodium potassiumATPase pump function.17,18 As pump inhibition maintains severalkey cellular membrane functions, Digibind potentially may exert itsobserved clinical benefit by reversal of sodium pump inhibitionfollowed by improving cell membrane function.18 These observedeffects would not have limitation to preeclampsia only but couldhypothetically be extrapolated to other diseases related to theincreased inflammatory response and would imply a potentialplatform technology for Digibind.

Naþ /Kþ -ATPase is a highly conserved ubiquitous membraneprotein, which is composed of three subunits: a, b and g. Wefound that TNFa had no effect on Naþ /Kþ -ATPase a-subunitexpression (data not shown), but downregulation of b-subunit wasobserved. Interestingly, Digibind could block the downregulation of

b-subunit expression induced by TNFa in endothelial cells.Although the exact mechanism of Digibind preservation of theb-subunit is not clear, studies have shown that b-subunit may bemore intimately involved in the mechanism of active transportfunction of Naþ /Kþ -ATPase,19 as the cation affinity of theNaþ /Kþ -ATPase can be affected by changes in theb-subunit,19,20 which indirectly supports our data that Digibindcould preserve the b-subunit function in vascular endothelium.

In summary, in this study we found that Digibind couldattenuate cytokine TNFa-induced increased endothelial surfaceadhesion molecule expression and decreased Na/K-ATPase b1expression in cultured endothelial cells. As it is impossible to obtainmaternal systemic vessels during pregnancy, it limits us to directlystudy the endothelial response to Digibind in an in vivo situationsuch as in preeclampsia. However, the DEEP study result thatDigibind could reverse erythrocyte Na/K-ATPase pump function18

supports the idea that Digibind may exert protective effects onvascular endothelial function by restoring Na/K-ATPase pumpfunction and increase the pump activity. Although the Digibindaction on endothelial function is largely unknown, our data dosuggest that Digibind may exert antiinflammatory effects onvascular endothelial cells, the mechanism of which warrantsfurther investigation.

References

1 Graves SW, Williams GH. An endogenous ouabain-like factor associated with

hypertensive pregnant women. J Clin Endocrinol Metabol 1984; 59: 1070–1074.

2 Lopatin DA, Ailamazian EK, Dmitrieva RI, Shpen VM, Fedorova OV, Doris PA et al.

Circulating bufodienolide and cardenolide sodium pump inhibitors in preeclampsia.

J Hypertens 1999; 17: 1179–1187.

3 Averina IV, Tapilskaya NI, Reznik VA, Frolova EV, Fedorova OV, Lakatta EG et al.

Endogenous Na/K-ATPase inhibitors in patients with preeclampsia. Cell Mol Biol

(Noisy-le-grand) 2006; 52: 19–23.

4 Roberts JM, Taylor RN, Musci TJ, Rodgers GM, Hubel CA, McLaughlin MK.

Preeclampsia: an endothelial cell disorder. Am J Obstet Gynecol 1989; 161:

1200–1204.

5 Redman CWG, Sacks GP, Sargent IL. Preeclampsia: an excessive maternal

inflammatory response to pregnancy. Am J Obstet Gynecol 1999; 180: 499–506.

6 Kupferminc MJ, Peaceman AM, Wigton TR, Rehnberg KA, Socol ML. Tumor necrosis

factor-a is elevated in plasma and amniotic fluid of patients with severe preeclampsia.

Am J Obstet Gynecol 1994; 170: 1752–1759.

7 Vince GS, Starkey PM, Austgulen R, Kwiatkowski D, Redman CWG. Interleukin-6,

tumour necrosis factor and soluble tumour necrosis factor receptors in women with

pre-eclampsia. Bri J Obstet Gynecol 1995; 102: 20–25.

8 Lyall F, Greer IA, Boswell F, Macara LM, Walker JJ, Kingdom JCP. The cell adhesion

molecule, VCAM-1, is selectively elevated in serum in pre-eclampsia: does this indicate

the mechanism of leucocyte activation? Br J Obstet Gynaecol 1994; 101: 485–487.

9 Chaiworapongsa T, Romero R, Yoshimatsu J, Espinoza J, Kim YM, Park K et al.

Soluble adhesion molecule profile in normal pregnancy and pre-eclampsia. J Matern

Fetal Neonatal Med 2002; 12: 19–27.

10 Adair CD, Buckalew V, Taylor K, Ernest JM, Frye AH, Evans C et al. Elevated endoxin-

like factor complicating a multifetal second trimester pregnancy: treatment with

digoxin-binding immunoglobulin. Am J Nephrol 1996; 16: 529–531.


199


11 Adair CD, Hinshaw H, Russell G, Rose J, Veille J, Buckalew V. Effects of digoxin specific-

antibody on mean arterial pressure in severe preeclampsia. Presented at the 12th

Scientific Meeting of the American Society of Hypertension, San Francisco, CA, May 29,

1997.

12 Wang Y, Adair CD, Coe L, Weeks JW, Lewis DF, Alexander JS. Activation of endothelial

cells in preeclampsia: Increased neutrophil-endothelial adhesion correlates with up-

regulation of adhesion molecule P-selectin in human umbilical vein endothelial cells

isolated from preeclampsia. J Soc Gynecol Investig 1998; 5: 237–243.

13 Canzoneri BJ, Lewis DF, Zhang Y, Gu Y, Philibert L, Groome LJ et al. Maternal

circulating TNFa levels are highly correlated with IL-10 levels, but not IL-6 and IL-8

levels, in women with preeclampsia. Am J Obstet Gynecol 2006; 193(6).

14 Ohno N, Ichikawa H, Coe L, Kvietys PR, Granger DN, Alexander JS. Soluble selectins

and ICAM-1 modulate neutrophil-endothelial adhesion and diapedesis in vitro.

Inflammation 1997; 21: 313–324.

15 Silber A, Newman W, Reimann KA, Hendricks E, Walsh D, Ringler DJ. Kinetic

expression of endothelial adhesion molecules and relationship to leukocyte recruitment

in two cutaneous models of inflammation. Lab Invest 1994; 70: 163–175.

16 Kupferminc MJ, Peaceman AM, Aderka D, Wallach D, Socol ML. Soluble tumor necrosis

factor receptors and interleukin-6 levels in patients with severe preeclampsia. Obstet

Gynecol 1996; 88: 420–427.

17 Lam G, Johnson D, Robinson C, Saade G, Lewis DF, Porter K et al. Antepartum

administration of a Digoxin immune fab (Digibind) improves renal function in

patients with severe preeclampsia. Presented at the XVI World Congress of International

Society for the Study of Hypertension in Pregnancy, 2008; Washington, DC, USA, Sept.

20–24, 2008, pp. 1–2.

18 Hopate M, Graves S, Adair CD, Lam G, Johnson D, Saade G et al. In-vivo reversal of

functional sodium pump inhibition with Digibind Treatment. Presented at the XVI

World Congress of International Society for the Study of Hypertension in Pregnancy,

2008; Washington, DC, USA, Sept. 20–24, 2008.

19 Kaplan JH. Biochemistry of Na,K-ATPase. Annu Rev Biochem 2002; 71:

511–535.

20 Eakle KA, Kabalin MA, Wang SG, Farley RA. The influence of beta subunit structure on

the stability of Na+/K(+)-ATPase complexes and interaction with K+. J Biol Chem

1994; 269: 6550–6557.


200


ORIGINAL ARTICLE

Fetal macrocrania: diagnosis, delivery and outcomesMR Laye1, BC Moore2, MA Kosek3, LK Bufkin4, JC Morrison4 and JA Bofill4

1Regional Maternal-Fetal Medicine, Spartanburg Regional Medical Center, Spartanburg, SC, USA; 2Department of Obstetrics andGynecology, The University of Tennessee College of Medicine Chattanooga, Chattanooga, TN, USA; 3Division of Neonatology,Department of Pediatrics, University of Mississippi Medical Center, Jackson, MS, USA and 4Division of Maternal-Fetal Medicine,Department of Obstetrics and Gynecology, University of Mississippi Medical Center, Jackson, MS, USA

Objective: To describe fetal macrocrania including prenatal diagnosis,

delivery considerations and clinical outcomes.

Study Design: A retrospective case series was developed by reviewing

26 885 ultrasounds performed between 1 March 2003 and 30 June 2007

for the prenatal diagnosis of macrocrania. Medical records of each

mother/infant pair were reviewed for demographic information,

ultrasound findings, obstetric management and outcomes.

Result: Twenty-three fetuses were diagnosed with macrocrania. Median

gestational age at diagnosis was 31.1 weeks (range 18.3–38.1) and at

delivery was 36.9 weeks (range 30.7–39.9). Fifteen patients (65%)

underwent amniocentesis for karyotype; none were aneuploid but one had

a duplication on chromosome 7. All the 23 infants were liveborn. Twenty-

one deliveries were by Cesarean (91%), with thirteen of these by classical

incision (62%). Of the infants, 5 (22%) died shortly after birth, 16 (70%)

were stabilized in the neonatal intensive care unit and were discharged

alive and 2 (8%) were transferred to another center and subsequently

died. Eighteen babies required ventriculoperitoneal shunting (78%).

Conclusion: Macrocrania is a diagnosis usually made in children but

can also be made prenatally. Fetal macrocrania is usually a result of

ventriculomegaly due to an obstructive process to cerebrospinal fluid flow.

Abdominal delivery is usually required, often necessitating a classical

uterine incision. Targeted ultrasonography, extensive counseling of

parents and delivery at a tertiary care center with availability of

neurosurgery is recommended.



Keywords: prenatal diagnosis; ultrasound; anomaly; ventriculomegaly

Introduction

Terms used to describe a large head are often used interchangeably.However, this is incorrect because the different terms represent

different entities. Making the distinction is important because use ofthe correct term may yield information about the underlying processcausing the clinical finding, and may confer information aboutprognosis and/or management. Macrocrania is defined as anabnormal increase in the size of the skull, with the facial area beingdisproportionately small in comparison. In contrast, macrocephalyis excessive size of the whole head, whereas megalencephalyrepresents overgrowth of the brain.1

Macrocrania is a term rarely encountered in obstetrics orprenatal diagnosis but is a very common diagnosis in pediatricsand in radiology, affecting up to 5% of pediatric patients.2

Unfortunately, diagnostic criteria are inconsistent. A review of theliterature on the topic shows criteria that are both objective andsubjective. Objective measures are inconsistent from study to studyand include children with a head circumference >95th–98thpercentile for age, disproportionate head size compared with bodylength and weight, and rapidly enlarging head circumference onserial measurements.3,4 The descriptors ‘disproportionate’ and‘rapidly enlarging’ are not well defined. A large-appearing head byobservation is an example of a subjective measure.4

In the pediatric population, macrocrania is most commonly dueto abnormal fluid collections in the head.3 These most commonlyinclude abnormalities in the ventricular system of the centralnervous system (ventriculomegaly) or external hydrocephalus(abnormal fluid collection in the subarachnoid space with normalcerebral ventricles).2 When ventriculomegaly is noted, this is mostoften due to obstruction in the flow of cerebrospinal fluid, such asis seen with stenosis or atresia of the aqueduct of Sylvius.4 Otherdifferential diagnostic considerations when macrocrania isencountered include space occupying lesions such as tumors,hemorrhages, large arachnoid cysts and myelomeningoceles.2,5

It has long been advocated that children considered to havemacrocrania be evaluated with imaging studies to determine theneed for surgical versus nonsurgical management.2,6 Whichmodality to use is the subject of some debate. The classic way toevaluate children with macrocrania was to use ventriculographyand cerebral angiography. Due to the desire to avoid these invasivetests in children, these were limited to only the most severe cases.

Received 1 June 2008; revised 21 October 2008; accepted 24 October 2008; published online

4 December 2008

Correspondence: Dr MR Laye, Regional Maternal-Fetal Medicine, 853 North Church Street,

Suite 610, Spartanburg, SC 29303, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.196



The introduction of computed tomography (CT) permittednoninvasive evaluation of these cases, allowing those with lesserdegrees of head enlargement to be included.3 Cranial ultrasoundhas since been shown to correlate well with CT, with no significantabnormality missed using this modality. This, coupled with thedesire to avoid ionizing radiation in children, led to cranialultrasonography being advocated as a first line method of imagingthe head in pediatrics.4 CT or magnetic resonance imaging is nowused in cases that are unclear, finer discrimination is needed, orfor imaging the external subarachnoid spaces.6

Given the frequency of ultrasonography during pregnancy andthat evaluation of the head and intracranial contents is a part of eachtargeted sonogram, many of the same diagnoses made in childrenwith macrocrania should be made in the fetus. The paucity ofinformation in the obstetric literature regarding the prenatal diagnosisof macrocrania, the associated obstetric management and outcomes isstriking. Accordingly, we created this retrospective case series at ourtertiary referral center so that we could better counsel patients andreferring physicians when a fetus with an abnormally large head isencountered on ultrasound. Our main interests includeddemographics of patients whose fetuses were diagnosed withmacrocrania, obstetric management and delivery considerations inthese pregnancies and outcomes including duration of stay in theneonatal intensive care unit (NICU), need for surgery shortly afterdelivery and ultimate survival.

Methods

The Institutional Review Board of the University of MississippiMedical Center (UMC), the primary perinatal referral center for thestate of Mississippi, approved this retrospective review of deidentifieddata (IRB File no. 2007–0016). Sonographers, maternal–fetalmedicine fellows and attending physicians performed allultrasounds in the Antenatal Diagnostic Unit at UMC. Ultrasoundstudies performed during a 52-month period from March 2003 toJune 2007 were reviewed and those cases in which the prenataldiagnosis of macrocrania was suspected were identified. This wasfollowed by a review of both maternal and neonatal medicalrecords. Demographic variables, fetal information includingdiagnoses, obstetric management including amniocentesis andresults, timing and mode of delivery and neonatal data includingsurvival were recorded for each mother/infant pair.

Diagnostic criteria were based on those used for children afterbirth. However, choosing a particular percentile above which wecould make the diagnosis would have been arbitrary given thediscrepancies by different authors in the pediatric literature asdescribed above. Therefore, we considered a fetus to havemacrocrania if either or both of the following were met:

(1) A rapidly enlarging head as evidenced by the headmeasurements (head circumference and biparietal diameter)

outpacing the other biometric parameters by greater than orequal to 4 weeks size.

(2) Fetal head size greater than would likely be successfullydelivered vaginally as evidenced by a biparietal diameter ofgreater than or equal to 11 cm (given that a completely dilatedcervix is usually 10 cm).

Results

A total of 26 885 ultrasound studies were reviewed and 23 fetusesdiagnosed with macrocrania were identified. None were lost tofollow-up and all deliveries were at our institution. Thedemographics of this cohort are summarized in Table 1. Fifteenpatients (65%) underwent genetic amniocentesis; none wereaneuploid but one had a duplication on chromosome 7. All the 23fetuses were liveborn. Continuous variables regarding infant birthdata are summarized in Table 2. Fifteen infants (65%) were male.

Table 1 Demographic variables of patients diagnosed with macrocrania

Demographic variable Median result (range)

Maternal age 23 years (15–36)

Gravidity 2 (1–7)

Parity 1 (0–4)

Gestational age at diagnosis 31.1 weeks (18.3–38.1)

Gestational age at delivery 36.9 weeks (30.7–39.9)

Table 2 Infant birth data of patients diagnosed with macrocrania

Variable Median result (range)

Birth weight 3308 g (2230–4305)

APGAR at 5 min 9 (2–9)

Head circumference 42.5 cm (35.5–59)

Table 3 Underlying causes of macrocrania by diagnosis

Diagnosis Number of infants

(% of all infants)

Stenosis/atresia of the aqueduct of Sylvius 10 (43.5)

Meningomyelocele 3 (13.0)

VATER anomalad 3 (13.0)

Communicating hydrocephalus 1 (4.3)

Hydranencephaly 1 (4.3)

Dandy–Walker defect 1 (4.3)

Goldenhar syndrome 1 (4.3)

Pseudoporencephaly 1 (4.3)

Intracranial hemorrhage 1 (4.3)

Agenesis of corpus callosum and interhemispheric cyst 1 (4.3)

Total 23 (B100)

Fetal macrocraniaMR Laye et al

202


Twenty-one deliveries were by Cesarean (91%), with thirteen ofthese by classical uterine incision (62%). Of the infants, 5 (22%)died shortly after birth, 16 (70%) were stabilized in the NICU withsubsequent discharge and 2 (8%) were transferred to anothercenter and subsequently died. Median length of stay in the NICUwas 21 days (range 3–101). Eighteen babies (78%) requiredventriculoperitoneal shunting; the median day of life for thisprocedure was day 2 (range 1–20).

The underlying causes of the macrocrania encountered onprenatal sonography are listed in Table 3. All diagnoses were

confirmed after delivery by appropriate imaging, consultation andautopsy where appropriate. A patient-by-patient summary ofultrasound findings, diagnoses (before and after delivery),outcomes and length of NICU stay can be seen in Table 4.

Discussion

This case series documents our experience with fetalmacrocrania, and to our knowledge is the first to specificallydescribe the diagnosis prenatally. Therefore, our data may be

Table 4 Summary of ultrasound findings, diagnoses, outcomes, and NICU length of stay (LOS) for each patient. The diagnoses in parentheses were made after deliveryand were not documented antepartum

Pt no. Ultrasound findings Diagnoses Outcome NICU LOS

(days)

1 Midline falx, no neural tissue above brainstem Hydranencephaly Died 4

2 Dilation of lateral and third ventricles Aqueductal stenosis Survived 21


4 Dilation of lateral and third ventricles Aqueductal stenosis (VATER anomalad:

imperforate anus,

tracheal stenosis, single kidney, hemivertebrae)

Transferred, Died 15

5 Dilation of lateral and third ventricles, posterior fossa cyst,

single umbilical artery

Dandy–Walker malformation, (ear tags,

perimembranous VSD)

Survived 50

6 Dilated lateral and third ventricles, polyhydramnios Aqueductal stenosis, (periauricular skin tags,

truncus arteriosus)

Survived 20

7 Dilated lateral and third ventricles Aqueductal stenosis (Goldenhar syndrome:

left anotia, right microtia, hypertelorism,

cleft palate)

Survived 45

8 Absence of CSP, intrahemishperic cyst, hypertelorism, absent clavicles Agenesis of the corpus callosum,

intrahemispheric cyst, (pachygyria)

Survived 4

9 Unilateral absence of brain parenchyma extending to skull table Pseudoporencephaly Survived 22


11 Dilated lateral and third ventricles, cleft lip and palate, single umbilical artery,

hemivertebrae, clenched right hand

VATER anomalad (imperforate anus,

ambiguous genitalia)

Died 27

12 Dilated lateral and third ventricles, thoracolumbar ONTD Meningomyelocele Survived 34

13 Dilated lateral and third ventricles Aqueductal stenosis Survived 44

14 Disorganized intracranial tissue, visible clot and fibrin stranding in skull Intracranial hemorrhage Died 4

15 Dilated lateral and third ventricles Aquedutal stenosis Survived 13

16 Dilated lateral and third ventricles, dilated loops of small bowel, ventricular

septal defect, polyhydramios

Aqueductal stenosis, jejunal atresia,

(interrupted aortic arch)

Died 23

17 Dilated lateral and third ventricles, lumbosacral ONTD, talipes

equinovarus deformity

Meningomyelocele Survived 31

18 Dilation of lateral and third ventricles, polyhydramnios Aqueductal stenosis Survived 50

19 Dilation of lateral and third ventricles, hypoplastic cerebellum Aqueductal stenosis Survived 21

20 Dilation of lateral and third ventricles, lumbosacral ONTD Meningomyelocele Survived 19

21 Dilated lateral and third ventricles, esophageal atresia, hemivertebrae,

bifid great toe, pulmonary hypoplasia, preaxial polydactyly,

malformed tibia and fibula on right, ventriculoseptal defect, short ribs

VATER anomalad,

(tracheoesophageal fisulta)

Died 101

22 Dilated lateral and third ventricles Aqueductal stenosis Survived 22

23 Tetraventricular dilation, polyhydramnios, micrognathia, atriventricular

septal defect, hypoplastic left ventricle

Communicating hydrocephalus,

AV canal defect,

(coarctation of aorta, interrupted IVC)

Transferred, died 3

Abbreviations: AV, atrioventricular; IVC, inferior vena cava; NICU, neonatal intensive care unit;ONTD, open neural tube defect; VSD, ventricular septal defect.


203


useful in counseling patients and detailing physicians on what toexpect once the diagnosis is made. However, our data must beviewed in light of what has been reported in infants after birth.

Our median gestational age at diagnosis (31.1 weeks) issomewhat later than expected. However, in our study the mostfrequent cause of macrocrania was stenosis/atresia of the aqueductof Sylvius. This leads to progressive ventriculomegaly and althoughthe diagnosis of ventriculomegaly can be made earlier in gestation,the diagnosis of macrocrania is only possible after weeks or monthswhen the ventriculomegaly has become severe enough to causemarked expansion of the fetal head.

The median gestational age at delivery (36.9 weeks) is more areflection of the management of these patients than the naturalhistory of infants with macrocrania. Management of prenatalpatients with suspected fetal macrocrania at the University ofMississippi includes observation with monthly scans to follow thesize of the head and amniocentesis at 37 weeks followed by deliveryif lung maturity is documented. This strategy allows for patientsfrom our fairly rural state to deliver at a tertiary institution. Indeed,70% of infants (16/23) in this series underwent amniocentesis forfetal lung maturity, whereas the other 30% had preterm labor or anindicated preterm delivery before undergoing the amniocentesis asplanned.

The underlying diagnoses leading to fetal macrocrania in thisseries were similar to that recorded in pediatric populations. Donatreported in 1981 that hydrocephalus from ventriculomegaly is theleading cause of macrocrania in children,3 and the current caseseries echoes this finding. Babcock et al. reported in 1988 that ininfants with macrocrania, 5 of 11 (45.5%) infants withsignificantly abnormal findings on cranial sonography hadaqueductal stenosis.4 This is very similar to the 43.5% incidence ofaqueductal stenosis reported in this series.

We noted a majority of male fetuses diagnosed withmacrocrania. Our 65% male predominance is very close thatpreviously reported in children. Medina reported 59 of 88 (67%)children diagnosed with macrocrania were male,2 whereas Donatreported 53 of 72 (73.6%) children were male.3 This might beexplained by the high number of diagnoses of aqueductal stenosis,which is known to have an X-linked form.

Ventriculoperitoneal shunting was required in this series in 78%of infants. However, this is much higher than expected, based onpediatric literature. Medina et al. reported in 2001 that only 18% ofchildren with macrocrania required surgical treatment.2 Thismarked difference is likely due to worse ventriculomegaly in thoseinfants diagnosed prenatally. Further, some of the infants shuntedin this series did not survive and therefore may not have beenincluded in pediatric investigations.

There were no intrauterine fetal demises. However, our overallsurvival rate with prenatally diagnosed macrocrania was only 70%.This reflects the diagnoses made underlying the macrocrania andwill vary from case to case. For example, all three of the infants

with the VATER anomalad and the infant with hydranencephalydied shortly after birth but all cases with isolated aqueductalstenosis survived. Long-term outcomes for survivors will alsovary based upon the underlying diagnoses and are a potentialfurther course of investigation. The pediatric literature reflects thatinfants do well if the macrocrania is due to external hydrocephalusor mild ventriculomegaly.2,4,6,7 However, when moderate tosevere ventriculomegaly is noted, especially when associated withcerebral atrophy, outcomes are poor. Bosnjak et al. reportedonly 2 of 20 patients with cerebral atrophy were neurologicallynormal at presentation and in most cases the initial abnormalfinding persisted at follow-up.6 Babcock et al. reported that19% of term infants and 25% of preterm infants withmacrocrania were neurologically or developmentally abnormalon follow-up that ranged from 1 month to 5 years.4 The diagnosisof macrocrania has also been linked to epilepsy and autisticdisorders.8

In summary, the diagnosis of macrocrania is possibleprenatally. Fetal macrocrania is usually a result ofventriculomegaly due to an obstructive process to cerebrospinalfluid flow. Suspicion for macrocrania necessitates a targetedsonographic evaluation by personnel experienced in prenataldiagnosis to evaluate for the underlying pathology leading to themacrocrania. This allows extensive counseling of parents,arrangements for delivery at a tertiary care center with availabilityof neurosurgery, and evaluation for mode of delivery. Abdominaldelivery is usually required, often necessitating a classical uterineincision.

Acknowledgments

No financial support was obtained for this study.

References

1 Dorland, WA, Newman. Dorland’s Illustrated Medical Dictionary, 29th edn, WB

Saunders Company: Philadelphia, PA, 2000, pp 1043–1044, 1073.

2 Medina LS, Frawley K, Zurakowski D, Buttros D, DeGrauw AJ, Crone KR. Children with

macrocrania: clinical and imaging predictors of disorders requiring surgery. AJNR Am J

Neuroradiol 2001; 22(3): 564–570.

3 Donat JF. Evaluation of macrocrania using computed tomography. Am J Dis Child 1981;

135(12): 1118–1121.

4 Babcock DS, Han BK, Dine MS. Sonographic findings in infants with macrocrania. AJR

Am J Roentgenol 1988; 150(6): 1359–1365.

5 Park SW, Cho KH, Shin YS, Kim SH, Ahn YH, Cho KG et al. Helmetlike skull deformity

with a large arachnoid cyst. Surg Neurol 2006; 65(1): 95–98.

6 Bosnjak V, Besenski N, Marusic-Della Marina B, Kogler A. Cranial ultrasonography

in the evaluation of macrocrania in infancy. Dev Med Child Neurol 1989; 31(1):

66–75.

7 Parmeggiani A, Posar A, Giovanardi-Rossi P, Andermann F, Zifkin B. Autism,

macrocrania and epilepsy: how are they linked? Brain Dev 2002; 24(5): 296–299.

8 Hamza M, Bodensteiner JB, Noorani PA, Barnes PD. Benign extracerebral fluid

collections: a cause of macrocrania in infancy. Pediatr Neurol 1987; 3(4): 218–221.


204


ORIGINAL ARTICLE

Acute hemodynamic effects of caffeine administrationin premature infantsV Soloveychik1, A Bin-Nun2, A Ionchev2, S Sriram2 and W Meadow2

1Neonatology Section, Department of Pediatrics, Comer’s Children Hospital, The University of Chicago, Chicago, IL, USA and2Neonatology Section, Department of Pediatrics, South Bend Memorial Hospital, South Bend, IN, USA

Objective: Administration of caffeine citrate (caffeine) has been a central

component of the treatment of apnea of prematurity. However, given its

multiple pharmacologic effects, caffeine might be expected to produce

hemodynamic changes in heart rate, stroke volume, cardiac output and

vascular resistance.

Study Design: In this prospective observational study, we report the

hemodynamic effects of intravenous caffeine administration in a

population of premature infants who received caffeine to correct or

prevent apnea of prematurity.

Methods: Hemodynamic effects of caffeine were determined in 31

infants. Stroke volume was measured via echocardiogram, using velocity

time integral at the aortic root diameter. Statistically univariate analyses

were performed parametrically using paired t-test and nonparametrically

(sign test). Multivariate linear regression models were used to identify

subgroup covariate effects.

Results: After intravenous caffeine, cardiac index increased in 31 of 31

trials, by an average of 14.6±16.3% (s.d.); stroke volume increased in 24

of 31 trials, by 7.8±12.2%; heart rate increased in 28 of 31 trials by

7.7±7.2 beats per min; and blood pressure increased in 25 of 31 trials, by

4.1±5.8 mm Hg (all P<0.001). Multivariate linear regression revealed

no significant effect of dose, birth weight, gestational age or postnatal age.

Conclusions: Intravenous caffeine consistently increases cardiac output

and blood pressure in relatively stable premature infants, when given to

treat or prevent apnea of prematurity. We speculate that there may be

a role for caffeine in the hemodynamic treatment of hypotensive/

hypoperfused infants.



Keywords: cardiac output; caffeine; newborn; prematurity

Introduction

Caffeine citrate (caffeine), a trimethylxanthine, is widely used asfirst-line therapy in neonates to treat or prevent apnea ofprematurity. The beneficial effects of caffeine in these contextsare widely documented in both randomized and observationaltrials.1–3 However, caffeine has other physiologic effects, includingstimulating the central nervous and cardiovascular system,4

enhancing catecholamine secretion,4 and increasing basalmetabolic rate.4 At the pharmacologic level, caffeine acts byblocking adenosine receptors A1 and A2a, increasing cyclic 3, 5adenosine monophosphate by inhibiting phosphodiesterase, andtranslocating intracellular calcium.5 At least plausibly, some ofthese pharmacologic actions might interact with regulationof systemic hemodynamics. Small studies with variablemethodology have reported conflicting effects of caffeineon systemic blood pressure (BP), cardiac output (CO) andvascular resistance. However, no large study has lookedsystematically at the acute hemodynamic effects of caffeine inpremature infants.

We performed a prospective observational study of the acutehemodynamic effects of administration of caffeine on strokevolume (SV), heart rate (HR), CO and BP.

MethodsStudy designWe identified 31 infants in our neonatal intensive care unit (NICU)for whom caffeine had been ordered by the attending neonatologist(not one of the experimenters) as therapy or prophylaxis for apneaof prematurity. As our study was observational in nature, noconstraints were put on the dose of caffeine citrate (5, 10,20 mg kg�1), gestational age, birth weight or postnatal age ofpatient. All caffeine doses were administered intravenously over15–20 min. For each infant, HR and arterial oxygen saturationwere monitored continuously, and BP was determined eithercontinuously by indwelling transducer or intermittently by cuff. COwas determined echocardiographically by integration of bloodflow velocity at the aortic root (see below).

Received 5 March 2008; revised 17 August 2008; accepted 29 September 2008; published online

4 December 2008

Correspondence: Dr V Soloveychik, South Bend Memorial Hospital, 615 N. Michigan Street,

Suite NICU Office, South Bend, IN 46601, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.193



Measurement of hemodynamic parametersWe first assessed the presence of a patent ductus arteriosus (PDA)in each of our study patients using color Doppler. Any patient witha PDA was excluded from further study, as left ventricular outputwould not be an accurate measure of systemic blood flow (SBF) inthe presence of either a right-to-left or left-to-right aortic-pulmonary shunt.

Cardiac output was calculated using the standardechocardiographic method. A Hewlett-Packard 2500 ultrasoundscanner with 7.5 MHz probe was used to determine aortic rootdiameter (AOD), measured in the parasternal long axis view bytwo-dimensional echocardiography, frame by frame between thebases of cusps at the end of the diastole. Aortic cross-sectionalarea (ACSA) was calculated using the formula 3:14�ðAODÞ2=4.Because a small change in the AOD will change the COexponentially and AOD is not likely to change between pre- andpost-caffeine administration, we used the same AOD for both pre-and post-caffeine CO calculations. Aortic velocity time integral(VTI) was recorded in the apical view by continuous Doppler wave,where the cursor was placed across the aortic annulus close to thevalve leaflets. Care was taken to align the Doppler beam with thelong axis of the aorta, so that the maximal velocities are recordedand no correction for intercept angles was needed. Dopplerwaveformswere analyzed by the software of the ultrasound system (HP 2500)for peak velocity, mean velocity and average VTI. For each of theinfants in the study, the average value of VTI five cardiac cycleswas determined in the pre- and post-caffeine condition, and theseaverage values were used in subsequent calculations ofhemodynamic function. SV was calculated as the product of ACSAand VTI divided by body weight. CO was calculated as the productof HR and SV. Systemic vascular resistance (SVR) was calculatedas the ratio of mean BP/CO.

The collection and analysis of hemodynamic data for thepurposes of publication has been approved by the Internal ReviewBoard of the Division of Biological Sciences at the University ofChicago. Parental consent was waived, in accordance with our IRBdirectives, as bedside echocardiography is standard medical care inour NICU.

Sample size estimationThe number of patients needed to confirm a 20% differencebetween pre- and post-caffeine CO using an estimation for theb (power) of 0.90 and P-value (a) of 0.05 was 25 patients(pre- and post-caffeine).

Statistical analysisOne caffeine administration per patient was analyzed.Hemodynamic values pre- and post-caffeine were comparedstatistically using both univariate analysis by Student’s t-testfor paired samples and nonparametric analysis (sign test).

Multivariate linear regression analysis was performed for the effectsof covariates: gestational age, birth weight, postnatal age and doseof caffeine on the difference between pre- and post-caffeine valuesfor each hemodynamic variable of interest. Because this was anobservational study, no constraints were put on the dose of caffeine(5 mg kg�1, n¼ 18; 10 mg kg�1, n¼ 12; 20 mg kg�1, n¼ 13),gestational age, birth weight or postnatal age of patient (ranges1–38 days). Consequently, multivariate linear regression analyseswere possible for each of these potential variables. Statisticalsignificance for these analyses was accepted at an a-value of 0.05.

All data are presented as mean±standard deviation unlessotherwise specified.

ResultsDemographicsGestational age of the 31 infants in our study population was28.4±2.8 (s.d.) weeks; birth weight was 1.19±0.4 kg; day of lifeof caffeine administration was 9.2±10.6 days; dose was10.2±6.6 mg kg�1. Post-caffeine measurements were obtained at47±20 min after caffeine administration.

Hemodynamic effects of caffeineThe hemodynamic effects of 31 doses of caffeine are summarizedin Table 1 and Figure 1. Cardiac index increased in 31 of 31 trials,by 14.6±16.3% (s.d.); SV increased in 24 of 31 trials, by7.8±12.2%; HR increased in 28 of 31 trials, by 7.7±7.2 beats per

Table 1 Demographics

n¼ 43 GA (wks) BW (kg) Wt (kg) DOL (days) Dose (mg kg�1)

Average 27.62 1.08 1.23 16.43 10.85

s.d. 2.94 0.42 0.41 17.82 6.40

Median 27.00 1.00 1.18 7.00 10.00

Abbreviations: BW, body weight; DOL, days of life; GA, gestational age; Wt, weight;wks, weeks.

0

100

200

300

400

Pre

ml k

g-1

per

min

Post

Figure 1 Effects of caffeine administration on cardiac output in 31 prematureinfants. Pre- and post-caffeine values are indicated for each infant individually.Caffeine increased cardiac output in 31 of 31 trials, by an average of 15%(P<0.001).

Administration of caffeine in premature infantsV Soloveychik et al

206


min and BP increased in 25 of 31 trials, by 4.1±5.8 mm Hg (allP<0.001). Caffeine had no significant effect on systemic vascularresistance. Multivariate linear regression revealed no significanteffect of dose, birth weight, gestational age or postnatal age.Specifically, there was no tendency for either more mature, or moreimmature infants to respond differentially to caffeine. In total, 20infants were <27 weeks, change in CO 209±264, P<0.001. Intotal, 23 infants were >27 weeks, change in CO 173±197,P<0.001 (Table 2).

Discussion

Caffeine is widely used in NICU care for prevention, or prophylaxisapnea of prematurity. It has been extensively studied in this role,and appears both effective and safe. In a large placebo-controlledstudy, caffeine toxicity (measured as tachycardia, tachypnea,jitteriness, tremors and unexplained seizures) was noticed in only1.8% of preterm infants, with no statistical significance between thestudy and control groups, and no increase in mortality.1 However,other aspects of caffeine’s effects, in particular its effects onhemodynamic performance have not been well studied. We tookadvantage of the routine use of bedside echocardiographicassessments of CO in our NICU to evaluate in a prospective,systematic fashion the acute hemodynamic impact of caffeineadministration.

We report here the largest study of which we are awaredescribing the acute hemodynamic effects of intravenous caffeinein premature infants. We note that caffeine increased leftventricular CO in 31 of 31 administrations, by an average ofapproximately 15%. Both SV and HR increased equally, byapproximately 7%. BP rose significantly, by 4 mm Hg. Caffeine hadno significant effect on systemic vascular resistance. Multivariatelinear regression revealed no significant effects of birth weight,gestational age, chronologic age and dose. Though the statisticalsignificance of the hemodynamic impact of caffeine is clear, we

recognize that the clinical significance of these effects remainsuncertain.

Previous studies, with smaller numbers and less rigorousmeasurement protocols, have reported conflicting cardiovasculareffects of methylxanthines (either caffeine or aminophylline) inpremature infants. Hoecker et al.4,6 examined the effect of caffeineloading in two regimens in two groups of 16 infants and showedno change in left ventricular CO (LVCO), HR or BP when measured1 and 2 h after in one regimen and 1, 2 and 20 h later in thesecond regimen. Walther et al.5 showed an increase in LVCO andHR in 11 infants treated with aminophylline, however post-aminophylline determinations were made 4 h after maintenancedose. SV was increased by 15% only in the first 3 days of treatment,but returned to base line by the seventh day of treatment. Waltheret al.7 also showed an increase in LVCO (B27%), SV (B35%)and BP in 10 treated infants, compared to controls, measured 1 hbefore the maintenance dose. An increase in LVCO (B47%), SV(B44%) and HR was also shown in a study of theophylline effecton 15 infants by Fesslova et al.8 performed 30 min afteradministration.

The effect of caffeine on LVCO and SV noted in our patients wasmore significant, and more consistent, than any previouslydescribed pharmacologic agent in premature infants. Previousstudies describing the pharmacological effects of dopamine,dobutamine, milrinone, epinephrine or hydrocortisone have notshown consistent changes in the LVCO and SV.9–16

Hydrocortisone,9 commonly used to increase BP in infantsunresponsive to dopamine, failed to increase LVCO significantly ina study of 15 hypotensive infants. The hemodynamic effect ofdopamine has been examined in several studies: A trend ofincreasing RVCO, did not reach statistical significance, withsignificant increase of the shortening fraction in 8 hypotensivepremature infants;10 in 50 hypotensive patients treated withdopamine, LVCO dropped in 40, compared to 10 others in whomLVCO increased11: BP did increase in the low LVCO group, withconcomitant increase in the SVR. When comparing dopamine anddobutamine in the treatment of low SBF in 42 patients, 40% ofpatients did not respond at all, while dobutamine achieved betterSBF increase and dopamine achieved BP increase in the others.12

Dopamine was more effective than dobutamine in raising BP in 20hypotensive infants however dopamine did not increase LVCO.13

Both dopamine and dobutamine failed to increase myocardialcontractility in 106 infants with low SBF in the first day of life, withdopamine increasing LV stress.14 Milrinone failed to increase rightventricular CO (RVCO) in a prospective trial.16

In sum, intravenous caffeine administration appears to increaseCO in normotensive premature infants by approximately 15%. Thiseffect appears to be reliable, and appears to be insensitive tovariations in birth weight, gestational age, chronologic age anddose. The increase in CO is mediated equally by an increase in SVand HR. These observations provoke an intriguing hypothesisFif

Table 2 Hemodynamic effects of caffeine administration in 31 premature infants

Pre Post P

Mean BP 37.2 41.3 <0.001

(s.d.) 6 6.5

Mean SV 1.32 1.68 <0.001

(s.d.) 0.44 0.77

Mean HR 148 155 <0.001

(s.d.) 16.1 14.3

Mean CO 197 224 <0.001

(s.d.) 79.3 88.5

Mean SVR 0.21 0.21 F

(s.d.) 0.08 0.08

Abbreviations: BP, blood pressure (mm Hg); CO, cardiac output (ml min�1 per kg); HR,heart rate (beats per min); SV, stroke volume (ml per beat); SVR, systemic vascularresistance (mm Hg/ml min�1 per kg).


207


caffeine improves CO and BP in relatively stable infants, mightthere be a role for caffeine in improving hemodynamics in sickneonates?

References

1 Schmidt B, Roberts RS, Davis P, Doyle LW, Barrington KJ, Ohlsson A, et al. Caffeine for

apnea of Prematurity Trial Group. Caffeine therapy for apnea of prematurity. N Engl J

Med 2006;354(20):2112–2121.

2 Aranda JV, Gorman W, Bergsteinsson H, Gunn T. Efficacy of caffeine in treatment of

apnea in the low-birth-weight infant. J Pediatr 1977;90(3):467–472.

3 Murat I, Moriette G, Blin MC, Couchard M, Flouvat B, De Gamarra E, et al. The

efficacy of caffeine in the treatment of recurrent idiopathic apnea in premature infants.

J Pediatr 1981;99(6):984–989.

4 Hoecker C, Nelle M, Beedgen B, Rengelshausen J, Linderkamp O. Effects of a divided

high loading dose of caffeine on circulatory variables in preterm infants. Arch Dis

Child Fetal Neonatal Ed 2006;91(1):F61–F64.

5 Walther FJ, Sims ME, Siassi B, Wu PY. Cardiac output changes secondary to

theophylline therapy in preterm infants. J Pediatr 1986;109(5):874–876.

6 Hoecker C, Nelle M, Poeschl J, Beedgen B, Linderkamp O. Caffeine impairs cerebral and

intestinal blood flow velocity in preterm infants. Pediatrics 2002;109(5):784–787.

7 Walther FJ, Erickson R, Sims ME. Cardiovascular effects of caffeine therapy in preterm

infants. Am J Dis Child 1990;144(10):1164–1166.

8 Fesslova V, Caccamo ML, Salice P, Marimi A. Assessment of cardiovascular effects to

theophylline in premature newborns by means of serial echocardiography. Acta

Paediatr Scand 1984;73(3):404–405.

9 Noori S, Friedlich P, Wong P, Ebrahimi M, Siassi B, Seri I. Hemodynamic changes

after low-dosage hydrocortisone administration in vasopressor-treated preterm and

term neonates. Pediatrics 2006;118(4):1456–1466.

10 Clark SJ, Yoxall CW, Subhedar NV. Right ventricular performance in hypotensive

preterm neonates treated with dopamine. Pediatr Cardiol 2002;23(2):167–170.

11 Zhang J, Penny DJ, Kim NS, Yu VY, Smolich JJ. Mechanisms of blood pressure increase

induced by dopamine in hypotensive preterm neonates. Arch Dis Child Fetal Neonatal

Ed 1999;81(2):F99–F104.

12 Osborn D, Evans N, Kluckow M. Randomized trial of dobutamine versus dopamine in

preterm infants with low systemic blood flow. J Pediatr 2002;140(2):183–191.

13 Roze JC, Tohier C, Maingueneau C, Lefevre M, Mouzard A. Response to dobutamine

and dopamine in the hypotensive very preterm infant. Arch Dis Child 1993;

69(1 Spec No):59–63.

14 Osborn DA, Evans N, Kluckow M. Left ventricular contractility in extremely

premature infants in the first day and response to inotropes. Pediatr Res

2007;61(3):335–340.

15 Lundstrom K, Pryds O, Greisen G. The haemodynamic effects of dopamine and volume

expansion in sick preterm infants. Early Hum Dev 2000;57(2):157–163.

16 Paradisis M, Evans N, Kluckow M, Osborn D, McLachlan AJ. Pilot study of milrinone

for low systemic blood flow in very preterm infants. J Pediatr 2006;148(3):

306–313.


208


ORIGINAL ARTICLE

Apnea, bradycardia and desaturation in preterm infants beforeand after feedingC Slocum1, M Arko2, J Di Fiore2, RJ Martin2,3 and AM Hibbs2,3

1Carillion Roanoke Community Hospital, Roanoke, VA, USA; 2Rainbow Babies and Children’s Hospital, University Hospitals,Cleveland, OH, USA and 3Department of Pediatrics, Case Western Reserve University, Cleveland, OH, USA

Objective: A common clinical impression is that both gastroesophageal

reflux (GER) and cardiorespiratory events increase after feeding in preterm

infants. We aimed to measure objectively the effects of feeding on GER,

apnea, bradycardia and desaturations.

Study Design: We conducted a retrospective review of premature infants

with a gestational age of 23 to 37 weeks at birth and a post-conceptional

age of 34 to 48 weeks, who were referred for multichannel intraluminal

impedance (MII), pH probe and 12-h apnea evaluation. Cardiorespiratory

and GER event rates during pre- and post-feeding intervals were

compared.

Result: Thirty-six infants met the inclusion criteria. More GER events

occurred after a feed than before (P¼ 0.012). After feeds, reflux was less

acidic and higher in the esophagus (P<0.05). In contrast, the rates of

apnea, bradycardia and desaturations were not altered by infant feeding.

Apnea of >5 s occurred at a median frequency of 0 (range 0 to 3) events

per hour before a feed and 0 (0 to2) events per hour after a feed

(P¼ 0.61).

Conclusion: The frequency, height and pH of GER are significantly

altered by feedings in preterm infants. However, the common clinical

impression that apnea, bradycardia and desaturations are more prevalent

after feeding is not supported.


published online 15 January 2009

Keywords: gastroesophageal reflux; apnea; esophageal pH monitoring;multichannel intraluminal esophageal impedance

Introduction

A causal relationship between gastroesophageal reflux (GER) andapnea at prematurity continues to be a topic of significant debateand investigation. Although early studies suggest that such an

association might exist, subsequent investigations have yieldedconflicting results.1–8 Nevertheless, premature infants frequentlyundergo diagnostic evaluation for GER, and are widely prescribedanti-reflux medications.9,10

A common clinical impression is that GER may be more severeand apnea may be more significant in some infants after feeding.In this scenario, a recently fed infant who may or may not haveclinical evidence of reflux, experiences an apnea, bradycardia ordesaturation. A causal relationship between GER and apnea isinferred, and the practitioner is called upon to initiate anti-refluxtherapy. Although periods of prolonged airway closure have beenshown during active bottle feeding, and measurable changes inpulmonary mechanics may occur during or after feeds, it remainsunclear whether these physiological changes translate intoincreased cardiorespiratory instability after feeding.11–13

We aimed to empirically assess the putative increase incardiorespiratory events in preterm infants after feeds by comparingGER, apnea, bradycardia and desaturation frequencies before andafter feeding. We hypothesized that feeding will alter GER patternsowing to acid buffering and gastric distention as follows: incidenceof GER will increase after feeds, acidic reflux will predominatebefore feeds, non-acidic reflux will predominate after feeds, and thebolus height of refluxate will increase after feeds. Similarly, wehypothesized that the incidence of apneas, bradycardias anddesaturations would increase after feeds.

Methods

This study is a retrospective review of 36 inpatient preterm infants,who were referred, owing to clinical indications, for a 12-h bedsideapnea monitoring study with multichannel intraluminal impedance(MII) and pH monitoring at the Rainbow Babies and Children’sHospital from January to November 2006. Inclusion criteria were agestational age of 23 to 37 weeks and a post-conceptional age of 34to 48 weeks. Infants with major congenital anomalies were excluded.The sample size was determined by the number of clinical studiesperformed at our institution before the review.

Cardiorespiratory monitoring was carried out by inductanceplethysmography (Respitrace, Viasys Respiratory Care, Yorba Linda,

Received 27 May 2008; revised 15 October 2008; accepted 2 December 2008; published online

15 January 2009

Correspondence: Dr AM Hibbs, Division of Neonatology, Department of Pediatrics, Rainbow

Babies & Children’s Hospital, Case Western Reserve University, 11100 Euclid Ave, Suite 3100,

Cleveland, OH 44106, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.226



CA, USA), in conjunction with electrocardiogram and oxygensaturation (Ohmeda, Boulder, CO or Masimo, Irvine, CA, USA)recording. GER monitoring was conducted using a combinedMII and pH system (Sleuth Monitoring System, SandhillScientific, Highlands Ranch, CO, USA). MII utilizes esophagealmeasurements of electrical impedance between multiple closelyspaced electrodes to identify fluid in the esophagus, and enablesone to detect reflux regardless of acidity.14 The impedance/pHcatheter was initially placed by measuring the nasal–ear–xiphoiddistance. Placement was verified by chest radiography, withadjustment as needed to position the catheter tip approximately1 cm above the gastroesophageal junction. Apnea monitoring, MIIand pH studies were reviewed and manually scored by a singletrained reviewer, with the interpreter blinded to infant feedingtimes. A chart review was conducted to obtain demographicinformation.

Common definitions for GER and respiratory events wereused. A pH probe event was defined as pH <4 for >5 s. An MIIevent was defined as a 50% reduction in baseline impedancedetected by at least two sequential channels, and was classifiedas acidic (pH<4) or non-acidic (pH>4). To obtain the totalnumber of acidic events, those detected by impedance and acidevents detected only by pH were totaled. The height of refluxatetoward the proximal esophagus was approximated from theimpedance/pH catheter tip through imbedded electrodes spaced at1.5-cm intervals. The height of the most proximal electrode toidentify a reflux bolus was recorded. An apnea of >15 s wasscored and classified as central, obstructive or mixed, depending onthe absence or presence of obstructed respiratory efforts.Bradycardia at <85 beats min–1 and desaturations of <85% wererecorded. These values were selected as they may indicate apotentially clinically significant cardiorespiratory event, and arecommonly reported in neonatal literature. Shorter respiratorypauses of 10 and 5 s in length were also evaluated as alternativemeasures.

One-hour windows before and after feeding were identified foreach infant and used for the analysis. The period of time duringwhich the infant was actually feeding was excluded from thisanalysis, as the goal of the study was to compare cardiorespiratorystability and GER with the stomach in its most full state post-feedversus in its most empty state post-feed, but not to compare thefeeding versus non-feeding states, which have been describedearlier.15,16 All infants were on full enteral bolus feeds at the timeof study. Nasogastric or orogastric tubes, if utilized for feeding,remained in place for the duration of the study. The Wilcoxonsigned rank test for paired non-parametric data was used tocompare average pre- and post-feeding cardiorespiratory and GERevent rates for each infant.

This study was approved by the University Hospitals institutionalreview board and a waiver of consent was granted for chart reviewwithout patient contact.

Results

Thirty-six infants met the inclusion criteria and none of them wereexcluded. Apnea monitoring data were unavailable for three infantswho were included in the GER analysis. One infant was evaluatedtwice, and only the first evaluation is included in the analyses. Theaverage gestational age was 30.7±4.3 weeks, with a birth weightof 1617±834 g (Table 1). The average post-conceptional age atthe time of study was 38.6±3.0 weeks, with a weight of2766±720 g. The primary reasons for referral for evaluation wereapnea (n¼ 4), bradycardia (n¼ 10), desaturations or cyanosis(n¼ 12), acute life-threatening event (n¼ 6), suspected reflux(n¼ 3) and others (n¼ 1). Twenty-two infants were evaluatedduring their intitial hospitalization after birth, and 14 were studiedduring a subsequent admission. Only one infant was being treatedwith anti-reflux medications at the time of study. That infant wason both ranitidine and metoclopramide, but still had a pH <4 forover 30% of the tracing. No infants were on proton-pump inhibitorsor other histamine blockers or prokinetics. One infant wasreceiving caffeine therapy. The average duration of monitoring was14.0±1.8 h, with 3.2±0.9 clearly defined feedings per infant.

The character of GER was markedly altered by infant feeding(Table 2). Significantly more GER events, as detected by either MIIor pH probe, occurred after a feed than before. Acidic GER eventswere more frequent before a feed, whereas non-acidic GER eventswere more frequent after a feed. The median percentage of refluxevents that were acidic fell from 100 (0 to 100%) during the pre-feeding period to 4 (0 to 100%) during the post-feeding period(P<0.001). The average height of refluxate toward the proximalesophagus, as measured from the impedance/pH catheter tip, washigher after feeding than before.

In contrast, the rates of apnea, bradycardia and desaturationswere not altered by infant feeding (Table 3). An apnea of >15 s

Table 1 Demographic statistics for the study population (n¼ 36)

Gestational age (weeks) 30.7±4.3

Post-conceptional age (weeks) 38.6±3.0

Birth weight (g) 1617±834

Study weight (g) 2766±720

Race

White 22 (61%)

Black 14 (39%)

Male sex 18 (50%)

Supplemental oxygen at study 5 (14%)

Breast milk at study 8 (22%)

NG or OG feeds during study 8 (22%)

Caffeine at study 1 (3%)

IVH X3 (n¼ 29) 2 (7%)

PVL (n¼ 29) 3 (10%)

History of necrotizing enterocolitis or perforation 6 (19%)

Abbreviations: IVH, intraventricular hemorrhage; NG, nasogastric; OG, orogastric; PVL,periventricular leukomalacia.

Feeds, GER and cardiorespiratory eventsC Slocum et al

210


occurred at a frequency of 0 (0 to 3) events per hour before a feedand 0 (0 to 2) events per hour after a feed (P¼ 0.61). As nodifference in apnea of 15 s duration was detected, shorterrespiratory pauses were also analyzed. Apnea of >10 s and >5 salso did not occur at statistically different rates. Similarly, the ratesof bradycardia at <85 beats min–1 and desaturations of <85%were not affected by feeding.

Discussion

We hypothesized an increase in the rates of GER, apnea,bradycardia and desaturations after feeding. As predicted, feedingsignificantly increased the frequency, pH and bolus height of GERin preterm infants. Nevertheless, concomitant increases in apnea,bradycardia or desaturation were not found.

The increase in the median bolus height of refluxate afterfeeding is a new finding. As the sensors on the MII probe arespaced 1.5 cm apart, a median height increase from 6 (0 to 9) cmbefore feeding to 6.75 (3.75 to 9) cm after feeding simplycorresponds to an increase in the proportion of reflux boluseshitting higher sensors. In smaller infants, this may mean that therefluxate more frequently reached the level of the larynx; however,we were unable to pinpoint the anatomical location of the reflux inindividual infants in this retrospective study. The anatomicalheight of the refluxate may be relevant in triggering certainmorbidities. For example, perhaps only a refluxate reaching thelaryngeal area triggers apnea via protective reflex pathways. Thiscould theoretically account for the increased symptoms noted in asubset of infants after feeding. Nevertheless, in the group of infantsstudied, even though the higher post-feed reflux height after feedslikely meant that more reflux reached the upper esophagus afterfeeds, we did not find evidence of increased apnea, bradycardia ordesaturations.

Although not statistically significant, we found a higher rate ofapnea before feeding. In some studies, the acidity and osmolality offluid instilled into the laryngeal area have been found to be animportant factor in triggering apnea.17,18 Furthermore, a recentinvestigation of the temporal relationship between reflux andapnea of prematurity noted a significant relationship with weaklyacidic reflux events.19 If acidic reflux were a more potent triggerthan non-acidic events, this could explain the non-significantlyhigher rate of apneic events before feeds when acidic reflux is mostprevalent.

We found a marked change in the acidity of reflux events frompre- to post-feeding periods. A high percentage of reflux events,particularly after feeding, were non-acidic, and may have beenmissed by traditional pH monitoring. This finding agrees withpreviously published studies in infants showing that 53 to73% of reflux events are non-acidic and that gastric pH maybe >4 for 69 to 75% of the total monitoring time.20–22 Feedingappears to be a principal factor for the high rates of non-acidicreflux given the marked change in acidity from pre- to post-feedingintervals.

The population for this study consisted of inpatient preterminfants who were primarily referred for cardiorespiratory events andclinical suspicion of GER. If a selection bias were present in thisretrospective cohort, one would expect an overestimation of theassociation between feeding and cardiorespiratory events. However,no increases in apnea, bradycardia or desaturations after feeds wereseen. There was some heterogeneity in the population with regardto such characteristics as the presence of a nasogastric tube,diagnoses and severity of early neonatal illness. However, theanalytic approach was to compare each infant’s post-feed valueswith his/her own pre-feed values. Each infant served as his/herown comparator, thereby limiting the potential bias andconfounding from other patient characteristics. This does not rule

Table 2 Rates of reflux events before and after feeding, median (range)

Total GER

(events per hour)

Acidic GER

(events per hour)

Non-acidic GER

(events per hour)

Median height of

GER (cm)

Before feed 2 (0–11) 2 (0–11) 0 (0–4) 6 (3–9)

After feed 4 (0–16) 0 (0–10) 4 (0–11) 6.75 (3.75–9)

P-value 0.012 <0.001 <0.001 0.025

Table 3 Rates of cardiorespiratory events during pre- and post-feeding intervals, median (range)

Apnea X15 s

(events per hour)

Apnea >10 s

(events per hour)

Apnea >5 s

(events per hour)

Bradycardia

<85 beats min–1

Desaturations

<85% (events per hour)

Pre-feed 0 (0–3) 0 (0–23) 7 (0–86) 0 (0–4) 0 (0–25)

Post-feed 0 (0–2) 0 (0–6) 6 (0–60) 0 (0–4) 0 (0–40)

P-value 0.61 0.22 0.30 0.42 0.23


211


out the possibility that a particular subset of patients arepredisposed to increased cardiorespiratory events after feeds.

Furthermore, because the aim of this study was to comparecardiorespiratory stability and reflux in the pre-feed state, when thestomach is most empty, with those in the post-feed state, when thestomach is most full, we excluded the active feeding period itselffrom the analysis. Other studies have shown increaseddesaturations during feeding, particularly in infants withbronchopulmonary dysplasia.15,16 As our study was not designed toaddress the physiology of feeding and specifically excluded thistime period, our finding that cardiorespiratory stability was notincreased after feedings should not be interpreted as a claim thatthe infants did not have events during the active feeding perioditself. Other factors, such as pulmonary reserve with exertion andpossible aspiration, are likely to impact the active feeding periodmore significantly than the resting pre- and post-feed periodsaddressed in this study.

Another limitation of this study may be the power. In a post-hocpower calculation, assuming alpha 0.05, power 0.8 and the standarddeviation of the difference in pre- and post-feeding rates rangingfrom 0.050 to 0.50 events per hour, the detectable difference rangedfrom 0.024 to 0.24 events per hour. However, we found a non-significantly higher rate of apnea before feeding, suggesting that wewould not have found a significant post-feed increase in apnea witha larger sample size.

These findings argue against the common clinical impressionthat cardiorespiratory events are more frequent in preterm infantsafter feeding. Although an increase in the rate and height of GERwas observed after feeds, a corresponding increase was not found inapnea, bradycardia or desaturations. The discrepancy between thesefindings and a clinical impression of increased cardiorespiratoryevents after feeding may stem from clinicians interpretingpost-feeding events differently from those occurring at other times.Anecdotal accounts of increased cardiorespiratory events followingfeeds are commonly treated by anti-reflux therapies. As there is nodifference in the incidence of cardiorespiratory events before and afterfeeding, these findings suggest that routine treatment of post-feedingcardiorespiratory events with anti-reflux therapies is not warranted.

Acknowledgments

This research was approved by the Institutional Review Board of University

Hospitals.

Disclosure

The authors have no financial disclosures or conflict of interest to report.

References

1 Herbst JJ, Minton SD, Book LS. Gastroesophageal reflux causing respiratory distress and

apnea in newborn infants. J Pediatr 1979; 95: 763–768.

2 Spitzer AR, Boyle JT, Tuchman DN, Fox WW. Awake apnea associated with

gastroesophageal reflux: a specific clinical syndrome. J Pediatr 1984; 104: 200–205.

3 Menon AP, Schefft GL, Thach BT. Apnea associated with regurgitation in infants.

J Pediatr 1985; 106: 625–629.

4 Harris P, Munoz C, Mobarec S, Brockmann P, Mesa T, Sanchez I. Relevance of

the pH probe in sleep study analysis in infants. Child Care Health Dev 2004; 30(4):

337–344.

5 Peter CS, Sprodowski N, Bohnhorst B, Silny J, Poets CF. Gastroesophageal reflux and

apnea of prematurity: no temporal relationship. Pediatrics 2002; 109(1): 8–11.

6 Mousa H, Woodley FW, Metheney M, Hayes J. Testing the association between

gastroesophageal reflux and apnea in infants. J Pediatr Gast Nutr 2005; 41: 169–177.

7 Wenzl TG, Schenke S, Peschgens T, Silny J, Heimann G, Skopnik H. Association of

apnea and nonacid gastroesophageal reflux in infants: investigations with the

intraluminal impedance technique. Pediatr Pulmonol 2001; 31(2): 144–149.

8 Di Fiore JM, Arko M, Whitehouse M, Kimball A, Martin RJ. Apnea is not

prolonged by acid gastroesophageal reflux in preterm infants. Pediatrics 2005; 116:

1059–1063.

9 Clark RH, Bloom BT, Spitzer AR, Gerstmann DR. Reported medication use in the

neonatal intensive care unit: data from a large national data set. Pediatrics 2006;

117: 1979–1987.

10 Malcolm WF, Gantz M, Martin RJ, Goldstein RF, Goldberg RN, Cotten CM. Use of

medications for gastroesophageal reflux at discharge among extremely low birth weight

infants. Pediatrics 2008; 121(1): 22–27.

11 Koenig JS, Davies AM, Thach BT. Coordination of breathing, sucking, and swallowing

during bottle feedings in human infants. J Appl Physiol 1990; 69: 1623–1629.

12 Yu VY, Rolfe P. Effect of feeding on ventilation and respiratory mechanics in newborn

infants. Arch Dis Child 1976; 51: 310–313.

13 Shivpuri CR, Martin RJ, Carlo WA, Fanaroff AA. Decreased ventilation in preterm

infants during oral feeding. J Pediatr 1983; 103: 285–289.

14 Vandenplas Y, Salvatore S, Vieira MC, Hauser B. Will esophageal impedance replace pH

monitoring? Pediatrics 2007; 119(1): 118–122.

15 Poets CF, Langer MU, Bohnhorst B. Effects of bottle feeding and two different methods

of gavage feeding on oxygenation and breathing patterns in preterm infants. Acta

Paediatr 1997; 86: 419–423.

16 Singer L, Martin RJ, Hawkins SW, Benson-Szekely LJ, Yamashita TS, Carlo WA. Oxygen

desaturation complicates feeding in infants with bronchopulmonary dysplasia after

discharge. Pediatrics 1992; 90: 380–384.

17 St-Hilaire M, Nsegbe E, Gagnon-Gervais K, Samson N, Moreau-Bussiere F, Fortier PH

et al. Laryngeal chemoreflexes induced by acid, water, and saline in nonsedated

newborn lambs during quiet sleep. J Appl Physiol 2005; 98(6): 2197–2203.

18 Downing SE, Lee JC. Laryngeal chemosensitivity: a possible mechanism for sudden

infant death. Pediatrics 1975; 55: 640–649.

19 Magista AM, Indrio F, Baldassarre M, Bucci N, Menolascina A, Mautone A et al.

Multichannel intraluminal impedance to detect relationship between gastroesophageal

reflux and apnoea of prematurity. Dig Liver Dis 2007; 39(3): 216–221.

20 Lopez-Alonso M, Moya MJ, Cabo JA, Ribas J, del Carmen Macias M, Silny J et al.

Twenty-four-hour esophageal impedance-pH monitoring in healthy preterm neonates:

rate and characteristics of acid, weakly acidic, and weakly alkaline gastroesophageal

reflux. Pediatrics 2006; 118(2): e299–e308.

21 Condino AA, Sondheimer J, Pan Z, Gralla J, Perry D, O’Connor JA. Evaluation of

infantile acid and nonacid gastroesophageal reflux using combined pH monitoring and

impedance measurement. J Pediatr Gastroenterol Nutr 2006; 42(1): 16–21.

22 Mitchell DJ, McClure BG, Tubman TR. Simultaneous monitoring of gastric and

oesophageal pH reveals limitations of conventional oesophageal pH monitoring in milk

fed infants. Arch Dis Child 2001; 84: 273–276.


212


ORIGINAL ARTICLE

Splanchnic tissue oxygenation, but not brain tissue oxygenation,increases after feeds in stable preterm neonates tolerating fullbolus orogastric feedingV Dave, LP Brion1, DE Campbell, M Scheiner, C Raab and SM Nafday

Division of Neonatology, Department of Pediatrics, Children’s Hospital at Montefiore, Bronx, NY, USA

Objective: The objective of this prospective, observational study was to test

the hypothesis that tissue oxygenation in the splanchnic bed compared with

tissue oxygenation in the cerebral circulation changes after feeding in

preterm neonates who are tolerating full bolus orogastric feeds.

Study Design: Clinically stable premature neonates with postmenstrual

age between 32 and 356/7 weeks who were tolerating full bolus feedings

were studied before feeding and 1 h after feeding using near-infrared

spectroscopy. The ratio of oxygenated to reduced hemoglobin (tissue

oxygenation index, TOI) in the splanchnic circulation bed was divided by

the TOI in the cerebral circulation, thereby yielding the cerebro-

splanchnic oxygenation ratio (CSOR). We compared TOI and CSOR

before and after feeding. As the changes in TOI and CSOR had non-

Gaussian distribution, nonparametric statistics were used.

Result: Among 32 infants, CSOR increased significantly after feeding

(median difference 0.08; range �0.48, þ 0.58; P¼ 0.011), whereas

pulse oximetry did not change significantly (P¼ 0.600). The change in

CSOR with feeding was associated with a significant increase in

splanchnic TOI (preprandial median 43.8, range 25.2–68.4 vs

postprandial 47.5, range 25.8–70.8; P¼ 0.013), without any significant

change in brain TOI (preprandial median 64.9, range 44.5–75.4 vs

postprandial 58.9, range 42.2–72.3; P¼ 0.153).

Conclusion: This study indicates that CSOR and splanchnic TOI, but

not brain TOI, increase significantly after feeding in stable preterm

infants who are tolerating full orogastric feeds.


published online 20 November 2008

Keywords: preterm neonates; feeding; near-infrared spectroscopy;splanchnic circulation; cerebral circulation

Introduction

Doppler ultrasonography is the method currently used for theclinical assessment of velocity of superior mesenteric artery bloodflow.1–5 Several reports have shown that enteral feeding induces asignificant and progressive increase in blood flow velocity inthe superior mesenteric artery that peaks 45–60 min after themeal.6–12 The effect of feeding increases with the volume of milkand the interval between feedings. The increase in velocity peaksearlier after breast milk than after formula feeding.13

The possible association between the increase in blood flowvelocity and change in tissue oxygenation is expected. Greaterunderstanding of the rate of oxygen delivery and uptake in sickpreterm and term infants undergoing intensive care is animportant aim of neonatal medicine. Near-infrared spectroscopy(NIRS) is a continuous, noninvasive and portable technique,which can be used to measure oxygenation in living tissue. Thisnoninvasive method for assessing tissue oxygenation has hadlimited use in infants. NIRS has been used to monitor oxygenationof the brain in neonates by measuring the ratio of oxygenated todeoxygenated hemoglobin (termed ‘tissue oxygenation index’,TOI).14

Regional tissue oxygenation of other vascular beds is also beingstudied in tissues such as the splanchnic area,15,16 renal tissues17

and as a screening tool for the diagnosis of patent ductus arteriosusin extremely low birth weight neonates.18

NIRS has been reported to be useful in detecting changes insplanchnic oxygen delivery during apneic episodes15 and inpredicting splanchnic ischemia in neonates by measuring a ratio ofsplanchnic to cerebral TOI, cerebro-splanchnic oxygenation ratio(CSOR).16 Splanchnic oxygenation was compared with brainoxygenation as a reference, because under most physiologicalconditions cerebral blood flow autoregulation minimizes changesin brain oxygenation during events affecting splanchniccirculation.19 NIRS has also been utilized to measure knownphysiologic changes in tissue oxygenation of the liver in newborninfants during and after feeding via a nasogastric tube.20

As reliable data are not available on splanchnic oxygenationchanges after feeding, this study was designed to test the hypothesis

Received 22 June 2008; revised 17 September 2008; accepted 29 September 2008; published

online 20 November 2008

Correspondence: Dr SM Nafday, Division of Neonatology, Children’s Hospital at Montefiore,

Weiler Division, Albert Einstein College of Medicine, 1825 Eastchester Road, Suite #725,

Bronx, NY 10461, USA.

E-mail: [email protected] address: Division of Neonatal-Perinatal Medicine, University of Texas Southwestern

Medical Center at Dallas, Dallas, TX, USA



www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.189



that splanchnic oxygenation increases relatively to brainoxygenation after bolus feeds in growing preterm infants. Weexpected to see a postprandial increase in splanchnic oxygenationand no change in brain oxygenation.

MethodsStudy subjectsPremature neonates with postmenstrual age between 32 and 356/7

weeks were eligible for the study if they were clinically stable andtolerating full bolus feeds with a total daily amount between 120and 150 cm3 kg�1 body weight, without intravenous support.Eligibility for the study also include tolerance of orogastric tubefeeding by gravity over a period of 5–10 min given every 3 h, onthe day of the study. These criteria were used to control for feedingvolume and the amount of time over which the feeding isdelivered.

Exclusions from the study were: requirement for any form ofcardiovascular support, signs of abdominal distension, infection ornecrotizing enterocolitis developing within 1 week before or afterenrollment, gastroschisis or congenital diaphragmatic hernia.These exclusion criteria were selected because of concerns over theintegrity of splanchnic perfusion following pathologicalgastrointestinal disorders21–22 and because catecholamineadministration may reduce perfusion by pathologicalvasoconstriction.23

Routine care of the infants was not interrupted.All patients born between March 2006 and May 2007 were

enrolled in the study if they met all inclusion criteria and had noexclusion criteria, and if informed consent was obtained from theparent(s) or legal guardian.

Study protocolAll tracings were recorded by a single investigator (VD). We usedthe NIRO 300 (Hamamatsu Photonics, London, UK) niroscope fordata recording. Using this device, near-infrared light is transmittedthrough the region of interest. The emitter and receiver probes wereplaced flush against the skin. Two sets of optodes were placed intheir carriers on the forehead and abdomen, just below theumbilicus as demonstrated by Fortune et al.16 Distance between theemitter and the receiver probes was 4 cm. Changes in the signalsreceived from simultaneously recorded four infrared frequencieswere used to measure and calculate the proportion of thehemoglobin in oxygenated and reduced states. Thus, the TOI, thatis, oxygenated hemoglobin/total hemoglobin, was obtainedsimultaneously at two locations. For each occasion, the TOImeasurements were only recorded if a steady state (defined as acoefficient of variation less than 50%) was reached. The TOImeasurement recordings were obtained continuously for 5 minfrom both the head and abdomen. Preprandial measurements weremade at least 3 h after the last feeding. Another recording was

obtained for 5 min, 1 h (58–62 min) after the end of feedingbecause pilot data (not shown) suggested that peak CSOR wasobserved around 60 min after feeding. Data were transferred byRS232 cable to a personal computer on a comma separated valuefile, which was converted to a spreadsheet file for data analysis. Theaverage values of TOI for each 5-min tracing (500 measurements)were used for statistical analysis. Those measurements were thencombined as a ratio of abdominal TOI over brain TOI to produce aCSOR (TOIabdomen/TOIbrain).

Infants were fed either fortified own mother’s breast milk or 24-calories per ounce formula, either Special Care Formula (AbbottNutritions, Columbus, OH, USA) or Enfamil premature Lipil 24(Mead-Johnson & Co., Evansville, IN, USA).

Arterial hemoglobin oxygen saturation, measured by pulseoximetry, and vital signs were recorded during the NIRS tracing.NIRS tracings were used only in the absence of bradycardia, apneaor desaturation (<88% saturation).

Sample size and statistical analysisWe used SPSS version 13 (SPSS Inc., Chicago, IL, USA) with an aerror of 0.05 and two-tailed tests. Reliability was assessed by thecoefficient of variation and by the intraclass correlation coefficient.Intraassay reliability was assessed by the coefficient of variationwithin 5-min tracings and, for those patients with tracings longerthan 15 min, by the intraclass correlation coefficient between threeindividual segments of 5-min measures of CSOR. Interassayreliability was assessed by the intraclass correlation coefficientbetween pre- and postprandial measurement of brain TOI, whichwe expected would not be significantly different from each other.

Continuous variables with Gaussian distribution were analyzedusing correlation, and multiple linear regression, and presented asmean±s.d. Variables without Gaussian distribution were analyzedusing Wilcoxon signed rank test (paired values, exact test) ornonparametric correlation (Spearman’s r) and presented asmedian and range.

We used sample power version 2 (SPSS Inc.) to calculate thesample size. On the basis of preliminary data, we determined thatwe would need 31 cases to have a 90% power to detect a 20%change in CSOR from baseline to postprandial measurement(mean difference 0.14, s.d. of the difference 0.23), with an a errorof 0.05 and a two-tailed paired Student’s t-test.

Protection of human subjectsThe institutional review board of Montefiore Medical Centerapproved this study. We obtained written, informed consent of theinfant’s parent(s) or legal guardian before enrolling any patient.

Results

During the duration of the study, 40 patients were eligible, andparental consent was obtained in all patients. Seven patients’

CSOR and splanchnic TOI increase in preterm infant after feedsV Dave et al

214


recordings were excluded because steady state was not reached. Oneinfant who developed neonatal necrotizing enterocolitis the dayafter recording was excluded. None of the babies included in theanalysis had necrotizing enterocolitis before recruitment, during orafter the study. Thus, 32 infants’ data were included in theanalysis.

Baseline patient characteristics are presented in Table 1. None ofthe patients had a patent ductus arteriosus at the time of therecording, and no patients received treatment with indomethacinor dopamine at the time of the study. However, among four babieswith a history of indomethacin administration for patent ductusarteriosus, three had required surgical ligation after failure ofindomethacin therapy. The median time to full feeds was 7 days(range 1–60 days, interquartile range 5–17.5 days). Seven babieshad feeding intolerance leading to delayed establishment of fullfeeds beyond 30 days. Gastroesophageal reflux disease was noted in10 (approximately 31%) patients based on clinical symptomatologyand treated with metoclopramide. No radiological/pH probe studieswere conducted to establish this diagnosis. One baby had recoveredfrom Staphylococcus epidermidis bacteremia earlier during thehospital stay.

In the 5-min tracings used for the main analysis, the averageintraassay coefficient of variation of baseline 5-min tracings ofCSOR was 17% (range 8–32%, n¼ 32) and the coefficient ofvariation of postprandial CSOR was 16% (range 6–31%). Amongthe seven recordings with tracings greater than 15 min, theintraassay intraclass correlation coefficient for three individualsegments of 5-min measures of CSOR was 0.896 (95% confidenceinterval 0.673–0.979). Corresponding values for splanchnic TOIand brain TOI were, respectively, 0.838 (95% confidence interval0.533–0.967) and 0.986 (95% confidence interval 0.948–0.997).The interassay intraclass correlation coefficient between pre- andpostprandial measurements of brain TOI was 0.817 (95%confidence interval 0.625–0.911, n¼ 32).

Both baseline (0.71±0.20 (mean±s.d.)) and postprandial(58–62 min postfeeding) CSOR (0.80±0.16) had a normaldistribution. Because neither changes in CSOR nor measurementsof TOI were normally distributed, we used nonparametric statisticsto assess the changes in CSOR and TOI after feeding. BaselineCSOR increased significantly after feeding (Table 2), whereasoxygen saturation did not change significantly (P¼ 0.600). Thechange in CSOR with feeding was associated with a significantincrease in splanchnic TOI, without any significant change inbrain TOI (Table 2).

Post hoc analysis showed a significant correlation betweenpreprandial CSOR and body weight at the time of the study(r¼ 0.427, P¼ 0.015); this relationship was not significantlyaffected by intrauterine growth restriction. In contrast, there was nosignificant correlation between preprandial CSOR and eithergestational age (r¼�0.061, P¼ 0.738), postnatal age(r¼ 0.179, P¼ 0.328) or postmenstrual age (r¼ 0.240,

Table 1 Baseline patient characteristics

Characteristics (N¼ 32)

Birth weight, mean±s.d. (g) 1476±397

Gestation, mean±s.d. (week) 30.8±2.5

Female gender, n (%) 18 (56.3)

Mode of delivery, n (%)

Cesarean section 19 (59.4)

Vaginal delivery 13 (40.6)

Race, n (%)

White 7 (21.9)

Hispanic 14 (43.8)

Black 9 (28.1)

Asian 2 (6.3)

Type of gestation, n (%)

Single 24 (75)

Twin 8 (25)

Prenatal steroids, n (%) 23 (71.9)

Size for age, n (%)

Appropriate 29 (90.6)

Small 3 (9.4)

Feeding issues, n (%) 7 (21.9)

Gastroesophageal reflux disease, n (%) 10 (31.3)

Metoclopramide, n (%) 10 (31.3)

Cholestasis, n (%) 2 (6.3)

Bronchopulmonary dysplasia, n (%) 4 (12.5)

At the time of the measurement

Postnatal age, mean±s.d. (days) 34.1±1.1

Postmenstrual age, mean±s.d. (weeks) 25.4±18.0

Weight, mean±s.d. (g) 1838±359

Volume per feeding (ml kg�1 body weight) 19.6±2.0

Feeding, n (%)

Formula feeding 24 (75)

Breast milk feeding 8 (25)

Ventilatory support, n (%)

None (room air) 28 (87.5)

Nasal canula 2 (6.3)

Nasal continuous positive airway pressure 2 (6.3)

Oxygen saturation (preprandial), mean±s.d. (%) 98.7±1.6

Oxygen saturation (postprandial), mean±s.d. (%) 98.5±1.8

Caffeine, n (%) 2 (6.3)

Hematocrit, mean±s.d. (%) 37.2±7.0)

Temperature, mean±s.d. (1C) 36.9±0.1

Systolic blood pressure, mean±s.d. (mm Hg) 68.1±6.3

Diastolic blood pressure, mean±s.d. (mm Hg) 37.8±5.6


215


P¼ 0.187). There was no significant correlation between postnatalage and either postprandial change in splanchnic TOI (Spearman’sr¼�0.169, P¼ 0.36, postprandial change in brain TOI(Spearman’s r¼ 0.166, P¼ 0.36) or postprandial change inCSOR (Spearman’s r¼ 0.182, P¼ 0.32). Similarly, there was nosignificant correlation between postmenstrual age and eitherpostprandial change in TOI (splanchnic r¼ 0.068, P¼ 0.712 andbrain r¼ 0.047, P¼ 0.797, respectively) or change in CSOR(r¼�0.013, P¼ 0.946). Postprandial change in CSOR wasinversely related to baseline value (Spearman’s r¼�0.632,P<0.001) and to body weight (r¼�0.449, P¼ 0.01), thusCSOR increased most in those infants with a low baseline value.Postprandial change in CSOR was not related to the type of milkfeeding: formula, 0.09 (range �0.48, þ 0.58, n¼ 24) vs breastmilk, 0.05 (range �0.02, þ 0.41, n¼ 8, P¼ 0.685) or to thevolume of feeding (P¼ 0.191). A sensitivity analysis excludingpatients on nasal continuous positive pressure (NCPAP) (Table 2)had only minimal effect on changes in TOI and CSOR.

Discussion

Our study is the first, to the best of our knowledge, to systematicallyassess the change in CSOR after bolus feeding in preterm infantstolerating full orogastric feeding. Our data indicate that 5-minNIRS recordings can be used to measure CSOR at the bedsidewithout interrupting the routine care of the patient. Preprandialvalues of CSOR were significantly correlated with body weight atthe time of the measurement, but not with gestational age,postnatal age or postmenstrual age. The data in this study showthat CSOR increases 1 h after orogastric feeding in stable preterminfants with a postmenstrual age of 32–36 weeks, who toleratebolus orogastric feedings at the amount of 20 ml kg�1 body weightevery 3 h. This change in CSOR results from an increase insplanchnic oxygenation in the absence of any significant changein brain oxygenation. In our study, infants with lower baseline gutoxygenation, which is associated with lower body weight at thetime of the study, had the greatest degree of redistribution of tissueoxygenation after feeding. The significance of the changes in CSORafter feeding in preterm infants will need to be determined byadditional studies.

The data in our study are in agreement with previouspublications. Petros et al.,15 reported one infant with postprandialincrease in gastrointestinal blood volume and oxygen delivery tothe gut. The changes in splanchnic oxygenation after feeding inour study parallel those previously observed for the liver.20 Theobserved postprandial increase in splanchnic tissue oxygenation isexpected given the postprandial increase in superior mesentericartery blood flow velocity observed using Doppler studies.6–12

Nevertheless, as we did not obtain simultaneous Doppler data, wecould not determine the relationship between blood flow velocityand tissue oxygenation.

There have been theoretical concerns regarding the use of NIRSover the abdomen, primarily related to the path length of the bowelstudied, insufficient infrared light traversing to the intestines, andgut movement within the abdomen and peristalsis leading toscattering of infrared light. With advances in signal manipulation,expression of a ratio of oxyhemoglobin to reduced hemoglobin inthe tissues is possible, thus a need to know the path length and theproblems of movement artifacts have largely been eliminated.16,24

Nevertheless, variability of splanchnic TOI was greater than brainTOI in our study as shown by a lower intraassay intraclasscorrelation.

Median preprandial CSOR obtained on infants in our study waslower than that in control infants reported by Fortune et al.16 (0.70vs 0.96). As we observed in our study, a significant positivecorrelation between preprandial CSOR and body weight at the timeof the measurement, the most likely explanation is that the lowervalue of CSOR in our study may be related to the lower body weightin our patients, 1.8 kg (1.4–3.0) vs 2.4 kg (1.4–3.4) in controls inFortune’s study. The average gestational age in our study was31±3 weeks (range 26–34) vs 38 weeks (range 29–40) incontrols in Fortune’s study.

Although interindividual variability in CSOR in our study waslarge, with a range of 0.36–1.15, the interquartile range 0.59–0.83 was similar to that in Fortune’s study, 0.88–1.03. Our datashow that large Interindividual variability was in part related topatient heterogeneity in weight at the time of the study. Althoughwe were unable to show any relationship between CSOR andintrauterine growth restriction or postnatal age, our study was notpowered for these secondary analyses. A previous study has reported

Table 2 Niroscopic data

Preprandial Postprandial (58–62 min postfeeding) Difference: median (range) P-valuea

CSOR, mean±s.d. 0.71±0.20 0.80±0.16 0.08 (�0.48, +0.58) 0.011*

Splanchnic TOI, median (range) 43.8 (25.2–68.4) 47.5 (25.8–70.8) 2.2 (�26.9, +40.5) 0.013**

Brain TOI, median (range) 64.9 (44.5–75.4) 58.9 (42.2–72.3) 0.0 (�14.2, +10.8) 0.153***

Abbreviations: CSOR, cerebro-splanchnic oxygenation ratio; TOI, tissue oxygenation index.aWilcoxon signed ranks test, exact value.Sensitivity analysis conducted after excluding the two patients on NCPAP at the time of the study led to similar results: preprandial CSOR 0.71±0.20 vs postprandial CSOR 0.80±0.16,P¼ 0.025*; Splanchnic TOI 43.2 (25.2–68.4) vs 47.5 (25.8–70.8), **P¼ 0.032 and brain TOI 62.4 (44.5–75.4) vs 60.3 (42.2–72.3), ***P¼ 0.347.


216


changes in mean cerebral blood flow velocity after NCPAP25 therebypotentially altering the measurements of CSOR; nevertheless,excluding patients on NCPAP in sensitivity analysis did not affectresults of TOI or CSOR. Other studies have reported changes incerebral blood flow velocity after initiation of phototherapy;26,27 noinfant in our study was undergoing phototherapy at the time of themeasurement. Our study is limited by lack of serial measurementsand lack of power for secondary analyses. As Doppler studies haveshown cerebral and mesenteric region-specific changes during thefirst week of life in very low birth weight infants,28 we would haveexpected to find a relationship between preprandial CSOR andpostnatal age. As Doppler data13 have shown that the time to peakvelocity after feeding was shorter after breast milk than afterformula feeding, it is possible that CSOR values for breast milkmay have peaked earlier and that we may have missed a truedifference in CSOR values between formula and breast milk. Thus,serial data are needed to assess the relationship between CSOR,gestational age, size for age at birth, weight and postnatal age; andthe exact timing of postprandial increase in CSOR in preterminfants.

In conclusion, whereas Doppler testing in splanchnic bedprovides a useful tool to evaluate blood flow velocity to the gut,NIRS is suitable for monitoring changes in the tissue oxygenationin the splanchnic region in preterm infants and its changes relativeto cerebral circulation after feeding.

As the study was completed, we have obtained pilot data in otherinfants. An ex-28 week preterm neonate who had an NIRSperformed on day of life 35 (postmenstrual age 33 weeks) had aprefeeding CSOR of 1.05 and dropped his postfeeding CSOR to 0.57.The next day the baby developed modified Bell stage IIB NEC, andthe CSOR value was 0.3. Four patients (average gestational age,27.8±2.9 weeks; postmenstrual age, 32.5±2.1 weeks; birthweight, 1052±429 g and weight at the time of the measurement,1363±470 g) with hemodynamically significant patent ductusarteriosus had baseline (unfed) CSOR values of 0.39±0.06. Thislow CSOR compared with that in our study (P<0.005, Student’st-test) might be related in part with their low average weight atthe time of the measurement (P<0.05).

Splanchnic TOI (or oxygen saturation) is positively correlatedwith systemic venous saturation and gastric tonometry andnegatively correlated with serum lactate in infants followingsurgery for congenital heart disease.29 Splanchnic oxygenation, lowin patients with hemodynamically significant patent ductusarteriosus,19 increased after ductus ligation in one patient describedby Meier et al.30 It would be of interest to prospectively and seriallystudy neonates with risk factors for abnormal splanchnic bloodflow such as prematurity, hemodynamically significant patentductus arteriosus, intrauterine growth retardation31 and to monitortheir feeding protocols using NIRS. Furthermore, studies in babieswith feeding intolerance are required to determine whether CSORcan be used to differentiate disorders such as NEC, ‘CPAP belly’

(abdominal distention caused by CPAP overflow into thegastrointestinal tract) or sepsis. In a previously reported study onthe use of NIRS in patients with a surgical abdomen, the techniquewas shown to be highly sensitive and specific in predicting intra-abdominal pathology.16 However, in that study, infants with NEChad a median gestational age that was 10 weeks younger thanpatients with other disorders and their controls. Feeding decisions,that is, enteral feeds vs bowel rest in patients with abdominaldistension and the optimal timing for the initiation or reinitiationof enteral feeding in patients with NEC are very difficult. Studiesutilizing niroscopic evaluation of the splanchnic bed maypotentially offer important guidance in an area of clinical carewhere there are unclear recommendations.

Acknowledgments

Support for the study was provided from the Division of Neonatology, Children’s

Hospital at Montefiore, Bronx, NY, USA. Preliminary results were presented at the

2008 Eastern Society for Pediatric Research Annual Meeting, Philadelphia, PA,

USA, 28–30 March 2008 and at the Annual Meeting of the PAS-SPR, Honolulu,

Hawaii, 5 May 2008.

References

1 Akinbi H, Abbasi S, Hilpert PL, Bhutani VK. Gastrointestinal and renal blood flow

velocity profile in neonates with birth-asphyxia. J Pediatr 1994; 125: 625–627.

2 Quamar MI, Read AE, Skidmore R, Evans JH, Wells PNT. Transcutaneous Doppler

ultrasound measurement of superior mesenteric artery blood flow in man. Gut 1986;

27: 100–104.

3 Weir FJ, Fong K, Ryan ML, Myhr T, Ohlsson A. Superior mesenteric artery and renal

artery blood flow velocity measurements in neonates: technique and interobserver

reliability. Pediatr Radiol 1995; 25: 145–148.

4 Martinussen M, Brubakk AM, Linker DT, Vik T, Yao AA. Mesenteric blood flow velocity

and its relation to circulatory adaptation during the first week of life in healthy term

infants. Pediatr Res 1994; 36: 334–339.

5 Leidig E. Doppler analysis of superior mesenteric artery blood flow in preterm infants.

Arch Dis Child 1989; 64: 476–480.

6 Kempley ST, Gamsu HR. Superior mesenteric artery blood flow velocity in necrotising

enterocolitis. Arch Dis Child 1992; 67: 793–796.

7 Kempley ST, Gamsu HR, Vyas S, Nicolaides K. Effects of intrauterine growth retardation

on postnatal visceral and cerebral blood flow velocity. Arch Dis Child 1991; 66:

1115–1118.

8 Leidig E. Pulsed Doppler ultrasound blood flow measurements in the superior

mesenteric artery of the newborn. Pediatr Radiol 1989; 19: 169–173.

9 Ozkan H, Oren H, Erdag N, Cevik N. Breast milk versus infant formulas: effects on

intestinal blood flow in neonates. Indian J Pediatr 1994; 61: 703–709.

10 Hsu CH, Lee HC, Huang FY. Duplex ultrasonographic assessment of gut blood flow

velocity: effect of meal composition in normal full-term newborns after first feed.

J Ultrasound Med 1994; 13: 15–18.

11 Coombs RC, Morgan ME, Durbin GM, Booth IW, McNeish AS. Doppler assessment of

human neonatal gut blood flow velocities: postnatal adaptation and response to feeds.

J Pediatr Gastroenterol Nutr 1992; 15: 6–12.

12 Fang S, Kempley ST, Gamsu HR. Prediction of early tolerance to enteral feeding in

preterm infants by measurement of superior mesenteric artery blood flow velocity. Arch

Dis Child 2001; 85: F42–F45.


217


13 Lane AJP, Coombs RC, Evans DH, Levin RJ. Effect of feed interval and feed type on

splanchnic hemodynamics. Arch Dis Child Fetal Neonatal Ed 1998; 79(1): F49–F53.

14 Suzuki S, Takasaki S, Ozaki T, Kobayashi Y. A tissue oxygenation monitor using NIR

spatially resolved spectroscopy. In: Chance B, Alfano RR, Tromberg BJ (eds). SPIE Proc

(Optical Tomography and Spectroscopy of Tissue III, SPIE Digital Library,

Bellingham, WA, USA) 1999; 3597: 582–592.

15 Petros AJ, Heys R, Tasker RC, Fortune PM, Roberts I, Kiely E. Near infrared spectroscopy

can detect changes in splanchnic oxygen delivery in neonates during apnoeic episodes.

Eur J Pediatr 1999; 158: 173–174.

16 Fortune PM, Wagstaff M, Petros AJ. Cerebro-splanchnic oxygenation ratio (CSOR) using

near infrared spectroscopy may be able to predict splanchnic ischemia in neonates.

Intensive Care Med 2001; 27: 1401–1407.

17 Petrova A, Mehta R. Near-infrared spectroscopy in the detection of regional tissue

oxygenation during hypoxic events in preterm infants undergoing critical care. Pediatr

Crit Care Med 2006; 7(5): 449–454.

18 Underwood MA, Milstein JM, Sherman MP. Near-infrared spectroscopy as a screening

tool for patent ductus arteriosus in extremely low birth weight infants. Neonatology

2007; 91(2): 134–139.

19 Martinussen M, Brubakk A-M, Vik T, Yao AC. Mesenteric blood flow velocity and its

relation to transitional circulatory adaptation in appropriate for gestational age preterm

infants. Pediatr Res 1996; 39(2): 275–280.

20 Teller J, Schwendener K, Wolf M, Keel M, Bucher HU, Fanconi S et al. Continuous

monitoring of liver oxygenation with near infrared spectroscopy during naso-gastric

tube feeding in neonates. Schweiz Med Wochenschr 2000; 130(18): 652–656.

21 Coombs RC, Morgan ME, Durbin GM, Booth IW, McNeish AS. Abnormal gut blood flow

velocities in neonates at risk of necrotizing enterocolitis. J Pediatr Gastr Nutr 1992;

15: 13–19.

22 Kempley ST, Gamsu HR. Superior mesenteric artery blood flow velocity in necrotizing

enterocolitis. Arch Dis Child 1992; 67: 93–796.

23 Schwartzopff B, Hennersdorf M. Influence of cardiac circulation and medication on the

perfusion of the intestine Zentralbl Chir. Zentralbl Chir 2005; 130(3): 218–222.

24 Matcher SJ, Kirkpatrick P, Nahid K, Cope M, Delpy DT. Absolute quantification methods

in tissue near infrared spectroscopy. Optical tomography, photon migration, and

spectroscopy of tissue and model media: theory, human studies and instrumentation.

Proc SPIE 1995; 2389: 486–495.

25 Havranek T, Madramootoo C, Carver J. Nasal continuous positive airway pressure

affects pre- and postprandial intestinal blood flow in preterm infants. J Perinat 2007;

27(11): 704–708.

26 Benders MJ, van Bel F, van de Bor M. The effect of phototherapy on cerebral blood flow

velocity in preterm infants. Acta Paediatr 1998; 87(7): 786–791.

27 Dani C, Bertini G, Martelli E, Pezzati M, Filippi L, Prussi C et al. Effects of

phototherapy on cerebral haemodynamics in preterm infants: is fibre-optic different

from conventional phototherapy? Dev Med Child Neurol 2004; 46(2): 114–118.

28 Yanowitz TD, Yao AC, Pettigrew KD, Werner JC, Oh W, Stonestreet BS. Postnatal

hemodynamic changes in very-low-birth weight infants. J Appl Physiol 1999; 87(1):

370–380.

29 Kaufman J, Almodovar MC, Zuk J, Friesen RH. Correlation of abdominal site near

infrared spectroscopy with gastric tonometry in infants following surgery for congenital

heart disease. Pediatr Crit Care Med 2008; 9(1): 62–68.

30 Meier SD, Eble BK, Stapleton GE, Morales DL, Chang AC, Andropoulos DB. Mesenteric

oxyhemoglobin desaturation improves with patent ductus arteriosus ligation.

J Perinatol 2006; 26(9): 562–564.

31 Dorling J, Kempley S, Leaf A. Feeding growth restricted preterm infants with abnormal

antenatal Doppler results. Arch Dis Child Fetal Neonatal Ed 2005; 90(5): F359–F363.


218


ORIGINAL ARTICLE

Early neutropenia is not associated with an increased rate ofnosocomial infection in very low-birth-weight infantsR-J Teng, T-J Wu, RD Garrison, R Sharma and ML Hudak

Division of Neonatology, Department of Pediatrics, University of Florida Health Science Center at Jacksonville, Jacksonville, FL, USA

Background: Evidence is contradictory whether very low-birth-weight

(VLBW, birth weight <1500 g) infants with early neutropenia (NP),

especially those born to mothers with preeclampsia experience a greater

incidence of nosocomial infection (NI).

Objective: To investigate whether NP within the first 7 days of life is a

risk factor for NI in VLBW infants.

Methods: Over a 42-month period, we identified all VLBW infants born

at p34 weeks gestation who survived for more than 72 h. Infants who

had no evidence of early infection, who had at least one complete blood

count performed in the first week of life, and who were not given

prophylactic recombinant human granulocyte colony-stimulating factor

(rhG-CSF) were included in this retrospective study. Early NP was defined

as an absolute neutrophil count less than 1500 per ml at any time during

the first week of life. NI was defined as the culture of a bacterial or fungal

pathogen from a sterile body fluid that was obtained after 72 h of life in

an infant with one or more clinical signs of infection.

Results: A total of 338 VLBW infants were reviewed. Of those, 51 infants

were excluded because of death or onset of an infection before 72 h of age,

lack of a complete blood count in the first week of life or treatment

with rhG-CSF. Of the remaining 287 infants, NI occurred in 11 of 77

(14.3%) infants with early NP compared to 42 of 210 (20.0%) infants

without early NP (P¼ 0.31). Infants who developed NI were smaller and

less mature, had lower Apgar scores, were more frequently instrumented

with central lines and required a longer duration of parenteral nutrition

compared to infants without NI. Infants with NI also had a higher

mortality and a greater incidence of necrotizing enterocolitis, severe

intraventricular hemorrhage and threshold retinopathy of prematurity.

However, using stepwise multivariate logistic regression analysis, only the

duration of parenteral nutrition and gestational age were significant risk

factors for NI.

Conclusion: Our data do not support the hypothesis that early NP

increases the risk for NI in VLBW infants.



Keywords: neutropenia; nosocomial infection; very low birth weightinfant

Introduction

Nosocomial infection (NI) is a common morbidity in very low-birth-weight (VLBW) infants. About 20% of VLBW (birth weight<1500 g) infants develop at least one episode of NI during theirhospitalization.1–3 It has been demonstrated that NI is associatedwith prolonged hospital stay and increased medical costs,4

decreased survival5,6 and increased likelihood of thresholdretinopathy of prematurity.7 For these reasons, prevention of NIin VLBW infants has been an important objective in the clinicalpractice of neonatology for the past two decades.8,9

Very low-birth-weight infants are thought to be at increased riskfor NI for multiple factors that include invasive line(s), prolongedparenteral nutrition,10 prolonged endotracheal intubation,2

treatment with postnatal steroids for the prevention chronic lungdisease11 and neutropenia (NP).12,13 In the hope of reducing theincidence of NI among VLBW infants several strategies have beenstudied including prophylactic treatment with intravenousimmunoglobulin,14 antimicrobial agents15 and treatment of NPwith recombinant human granulocyte colony-stimulating factor(rhG-CSF)16,17 but without any proven success.

It has been suggested that prophylactic treatment with rhG-CSFor other hematopoietic growth factors might reduce the likelihoodof NI in VLBW infants, especially those with NP.17 However, manyinfants exhibit NP in the first week of life18 when VLBW infants arelikely being treated with broad-spectrum antibiotics empirically.These early episodes of NP usually resolve within 24–48 h and theefficacy of rhG-CSF remains unknown. Although the associationbetween NP and NI has been reported in infants born tohypertensive mothers by 4- to 10-fold19–21, the incidence of NP inVLBW infants born to normotensive mothers has not beenquantified. The association between maternal hypertension andrisk for NI was also challenged by Mouzinho et al.22 in their study.Our intention was to investigate whether early NP in VLBW infants,

Received 5 July 2008; revised 21 September 2008; accepted 29 September 2008; published online

11 December 2008

Correspondence: Dr R-J Teng, Division of Neonatology, Department of Pediatrics, Medical

College of Wisconsin, Suite 410, Children Corporate Center, 999 North 92nd Street,

Milwaukee, WI 53226, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.202



born to mothers with or without hypertension and not treated withrhG-CSF, is a significant risk factor for NI.

Methods

Computerized medical records of the University of Florida HealthScience Center at Jacksonville Neonatal Intensive Care Unit werereviewed retrospectively. All premature infants with birth weightsbetween 350 and 1500 g with gestational age p34 weeks, whosurvived for more than 72 h without a culture-proven bacterial orfungal infection with onset within that time, and who were nottreated with rhG-CSF were included (Figure 1). Gestational age wasdetermined by prenatal ultrasound before 20 weeks gestation or onthe basis of the last menstrual period. All laboratory data andmedication records were computerized and could be obtained uponrequest to link with the computerized medical records for crossreference.

In our unit a complete blood count and blood culture wasroutinely performed upon admission before initiation of treatmentwith ampicillin and gentamicin. Subsequently, a repeat completeblood count was ordered at the discretion of the care teams. Theabsolute neutrophil count (ANC) was obtained by multiplying thecombined percentage of mature and immature neutrophil forms bythe corrected total white count. NP was defined as a corrected ANC<1500 per ml.23–25 Other definitions of NP, including Mouzinhoet al.26 revised reference ranges, were also used in our analyses.

Antibiotics were discontinued if the blood culture was negativeby 48–72 h. Treatment of NP with rhG-CSF in the first week wasprovided at the discretion of the attending neonatologists. Thehighest ANC in rhG-CSF-treated infants was 1420 per ml. Umbilicallines were placed by either a resident physician or a physicianextender while percutaneous central venous lines were placed by ateam that included five neonatologists, one nurse practitioner andfour registered nurses. The indwelling catheters were removedwithin 24 h after discontinuation of parenteral nutrition.

Infants with one or more signs of sepsis after 72 h of life wereevaluated by culturing peripheral blood (two sets), cerebrospinal

fluid obtained by lumbar puncture and urine obtained bysuprapubic aspiration or sterile catheterization. If any one of thesecultures was positive for a bacterial or fungal pathogen, a diagnosisof NI was made with the exception that a diagnosis of coagulasenegative staphylococcal sepsis required two positive cultures. Amodified Bell’s classification was used to stage necrotizingenterocolitis (NEC)27 and Papile’s classification was used to gradethe severity of intraventricular hemorrhage (IVH).28 IVH grades IIIand IV were considered as severe IVH.

Sample size calculationWe assumed an increase in the incidence of NI from 20% in VLBWinfants without NP to 40% (2� increase) with NP, and 27% ofVLBW developed early NP,1,2 with a two-sided a-error of 0.05 and80% power (1-b), we needed 223 (163 non-NP and 60 NP) VLBWinfants for the retrospective study.

Statistic analysisContinuous variables were expressed as mean±standard deviation.The Student’s t-test or Mann–Whitney U-test was used forcomparison of continuous variables as appropriate. Categoricalvariables were compared using Fisher’s exact test. A univariateanalysis was used to determine the association between variousclinical characteristics and invasive procedures and the diagnosis ofNI. A multivariate stepwise logistic regression was performed, usingNI as the dependent variable, with factors significantly associatedwith NI under univariate analyses used as the independentvariables. Multiple definitions of NP were tested to identify whetherany definition of NP was independently associated with NI. Areceiver operating characteristic (ROC) curve was constructedusing the lowest ANC in the first 7 days of life to predict thepossibility of NI with sensitivity served as the x axis and (100specificity) as the y axis. The area under curve (AUC) of ROC curveabove 0.725 is usually considered as significant. A P-value <0.05was considered statistically significant. Odds ratios (ORs) wereexpressed with 95% confidence interval (CI). AUC of the ROC curveis expressed as mean±standard error (95% CI). MedCalcs

version 9.2.0.1 was used for statistic analyses.

210 non-NP 77 NP

287 no early infection 3 early infection 2 early infection

332 with CBC

338 VLBW

8 non-NP 3 NP

11 died in 3 days 2 survived > 3 days

6 without CBC

4 died in 3 days

1 with rhG-CSF290 rhG-CSF (-) 31 rhG-CSF (+)

321 survived > 3 days

Figure 1 The flowchart of patient exclusion. NP, neutropenia.

NI in very low-birth-weight infantsR-J Teng et al

220


The study received approval from the Institutional Review Boardof the Shands Hospital at Jacksonville. Informed consent waswaived because of the retrospective nature of this study and its lackof intervention on the patient population.

Results

During the study period, from January 2002 through June 2005,338 VLBW infants were admitted to our neonatal intensive care unit<34 weeks gestation. Of which, 51 (15.1%) infants were excludedbecause of death or an infection with onset before 72 h, the lack ofcomplete blood count or treatment with rhG-CSF (Figure 1). The51 excluded infants were significantly less mature (25.9±2.9versus 27.9±2.7 weeks, P<0.01) and had lower birth weights(720±223 versus 1061±280 g, P<0.01) compared to the 287infants who comprised the study population. Of the 332, 113(34.0%) VLBW infants with a complete blood count available in thefirst week of life experienced early NP with an ANC less than 1500per ml.

Of the 287 infants included in this study, 77 (26.8%) developedearly NP. The NP infants had a lower mean birth weight and lowermedian 1-min Apgar score than their non-NP counterparts(Table 1). All 52 (67.5%) NP infants were male compared to 95 ofthe 210 (42.5%) non-NP infants (P<0.01). NI was absolutely butnot significantly less frequent in NP compared to non-NP infants(14.3 versus 20.0%, P¼ 0.31; Table 2). The median age at the firstNI among NP infants (11 days, ranges 6–81 days) did not differfrom the non-NP group (17 days, ranges 6–67 days). Of interestwas that early NP is significantly more common in male infantswith all different definitions of NP except the most restrictivedefinition (ANC less than 500 per ml).

The incidence of NP was not different among infants born tomothers with hypertension (preeclampsia, eclampsia, chronichypertension, pregnancy-induced hypertension and HELLPsyndrome) compared to infants of normotensive mothers (24 of 74,or 32.4% versus 53 of 213 or 24.9%, P¼ 0.23). Among infants bornto mothers with hypertension, only 2 of the 24 (8.3%) infants withNP developed NI compared to 6 of the 50 (12%) infants without NP(P¼ 1.00). We also did not find any significant difference in theincidence of NP using either the definition of ANC less than 500 perml or that of ANC less than 1100 per ml. Using Mouzinho’s revisedreference ranges of ANC, there were 9 of 59 infants (15.3%) born tohypertensive mothers who fulfilled the diagnosis of NP compared to23 of 228 infants (10.1%) born to normotensive mothers(P¼ 0.25).

Of the 287 infants, 53 (18.5%) developed at least one episode ofNI and they were significantly (P<0.001) less mature (25.9±2.0versus 28.4±2.6 weeks) and had lower birth weights (877±270versus 1103±266 g) than their noninfected counterparts. Themedian 5-min Apgar score was also significantly lower in the NIgroup. All 53 (100%) infected infants had indwelling central

venous lines and 14 (26.4%) had severe IVH, both of which rateswere significantly higher than in the noninfected infants. Infantswith NI were on parenteral nutrition significantly longer thaninfants without NI (29.5±25.6 versus 10.6±7.5 days, P<0.01).Multivariate logistic regression showed that a lower gestational agein weeks (OR 1.32, CI: 1.11–15.6) and days of parenteral nutrition(OR 1.10, CI: 1.06–1.14) were the only two factors associated withNI (r2 ¼ 0.40). NP, regardless of the definition, was not associatedwith the development of NI. Threshold retinopathy of prematuritywas significantly more common in infants with NI (11 of 44, or25%, versus 7 of 214, or 3.3%, P<0.01). About half (40 of 81) ofthe identified pathogens were Gram-negative bacilli and one-eighth(10 of 81) were fungal species (Table 3).

We also analyzed the data by using different criteria of NP butagain failed to find an increased risk for NI. Eight infants had ANCless than 500 per ml between 2 h and 6 days of age; none developedNI and all survived to discharge (Table 4). Of 48, 8 infants withANC less than 1100 per ml died in hospital but only 4 (8%)developed NI; 2 developed NI within 7 days of death. The ROCcurve that used the lowest ANC within 7 days to predict thedevelopment of NI revealed the AUC was only 0.57±0.04

Table 1 Demographic data of neutropenic (ANC <1500 per ml) and non-neutropenic (ANC X1500 per ml) very low-birth-weight infants

Non-neutropenic

(N¼ 210)

Neutropenic

(N¼ 77)

P-value

Birth weight (gm) 1083.1±281.0 1001.9±271.2 0.028

Gestational age (wk) 28.0±2.7 27.6±2.6 0.309

Small for date 28/210 (13.3%) 11/77 (14.3%) 0.847

Male 95/210 (45.2%) 52/77 (67.5%) 0.001

TPN (days) 13.0±12.6 16.9±19.5 0.105

10-Apgar score 1–9 (median¼ 5) 0–9 (median¼ 4) 0.006

50-Apgar score 1–9 (median¼ 7) 1–10 (median¼ 7) 0.252

Central line 193/210 (91.9%) 75/77 (97.4%) 0.114

Symptomatic PDA 29/207 (14.0%) 12/76 (15.8%) 0.706

Severe IVH 28/207 (13.5%) 17/76 (22.4%) 0.097

NEC 8/208 (3.8%) 6/74 (8.1%) 0.208

Mortality 26/210 (12.4%) 10/77 (13.0%) 0.844

Maternal hypertension 50/210 (23.8%) 24/77 (31.2%) 0.225

Abbreviations: IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; PDA,patent ductus arteriosus; TPN, total parenteral nutrition; wk, week.

Table 2 The number of episode(s) of nosocomial infection among very low-birth-weight infants with and without early neutropenia

Episode(s) Non-neutropenic (N¼ 210) Neutropenic (N¼ 77)

0 168 (80.0%) 66 (85.7%)

1 28 (13.3%) 9 (11.7%)

2 9 (4.3%) 1 (1.3%)

3 4 (1.9%) 0 (0%)

4 1 (0.5%) 1 (1.3%)


221


(0.51–0.63) and not significantly different (P¼ 0.10) from theline of identity (Figure 2).

All 20 infants born to preeclamptic mothers developed early NPbut only 1 had NI as compared to 38 out of 170 infants born tonormotensive mother without early NP developed NI (P¼ 0.08).There was no difference between two groups regarding gestationalage, birth weight, Apgar scores and total days on PN, central lineusage, IVH, NEC, BPD, CLD and gender (data not shown). The onlydifference between the two groups were less premature rupture ofmembrane in infants born to preeclamptic mothers (2 of 20 versus72 of 170, P<0.01).

Discussion

Neonatal mortality has gradually fallen in the past 20 years and NIremains an issue. Unfortunately, the NI rate in these vulnerablepremature infants has not changed commensurately. About 20% ofVLBW infants still experience at least one episode of bacterial orfungal infection during hospitalization.1,2 One report found athreefold higher mortality rate among infants who experienced atleast one episode of NI compared to infants who remained free of

infection.3 NI has also been reported to increase the length of stayand hospital charges,29–33 and has been associated with greaterdegrees of various morbidities. Hence, prevention of NIs in VLBWinfants might reasonably be expected to improve the outcome ofthese vulnerable infants.

Several strategies, including the prophylactic intravenousadministration of immune globulin,14 treatment of NP with rhG-CSF16,17 and empirical antibiotic treatment,16 have been studied toreduce NI rates. To our knowledge, no controlled intervention withthe exception of implementation of a strict infection controlprotocol has been shown to be effective in reducing the rate of NI.

Neutropenia, defined as an ANC <1500 per ml, affects 6–58%of premature infants in the first week of life.22,34–36 It was reportedthat about half of the premature infants born to mother withhypertension develop NP20,37 and at that point are more likely todevelop either early-onset sepsis37 or NI.20 However, reports byMouzinho et al.26 and Paul et al.38 failed to confirm theassociation between NP and NI.

Of the 332, 113 (34.0%) in our series developed early NP,consistent with the rates of NP reported by others. We previouslyhave demonstrated that the G-CSF level in the cord blood tends tobe lower in premature infants with early NP especially when themother is preeclamptic.39 In a preliminary report, Kocherlakotaand La Gamma found that rhG-CSF might be able to prevent NI inneutropenic infants born to preeclamptic mother within their first4 weeks of hospitalization by increasing the ANC.16 They suggestedthat prophylactic treatment with rhG-CSF might be effective inreducing the incidence of NI. To date, no controlled trial has been

Table 4 Sensitivity, specificity, positive predictive value and negative predictivevalue for predicting nosocomial infection according to different criteria ofneutropenia

ANC<500 ANC<1100 ANC<1500 Mouzinho

Sensitivity 0/53 (0%) 8/53 (13.1%) 11/53 (20.8%) 5/53 (9.4%)

Specificity 226/234 (96.6%) 194/234 (82.9%) 168/234 (71.8%) 207/234 (88.5%)

PPV 0/8 (0%) 8/48 (16.7%) 11/77 (14.3%) 5/32 (15.6%)

NPV 226/279 (81.0%) 194/239 (81.2%) 168/210 (80.0%) 207/255 (81.2%)

Abbreviations: ANC, absolute neutrophil count; NPV, negative predictive value; PPV,positive predictive value.

Lowest ANC

100

80

60

40

20

0

0 20 40 60 80 100100-Specificity

Sen

sitiv

ity

Figure 2 The receiver operating characteristic (ROC) curve of using the lowestabsolute neutrophil count (ANC) to predict the development of nosocomialinfection(s). Solid line: ROC curve; broken lines: 2 s.d. of the ROC curve.

Table 3 Microbiologic identification of nosocomial infection in very low-birth-weight infants

Gram positive Number Gram negative Number

CONS 13 E. coli 8

MRSA 6 Klebsiella 11

Staph. aureus 3 Enterobacter 8

Enterococcus 6 Pseudomonas 7

GBS 1 Serratia 3

Pneumoccus 1 Proteus 2

Bacillus spp. 1

Fungus Aeromonas 1

Candida 9

Yeast 1

Abbreviations: CONS, coagulase negative staphylococcus; E. coli, Escherichia coli; GBS,group B streptococcus; MRSA, methicillin resistant staphylococcus aureus.


222


published to examine this hypothesis. There is also no basis to beconfident that rhG-CSF treatment would be effective in thetreatment of early NP in VLBW infants born to normotensivemothers.

In out unit, most (74%) of the VLBW infants with early NP wereborn to normotensive mothers, and NP was more commonlyobserved in male infants. Contrary to other reports,20,39 our data donot show a higher incidence of NP in infants born to hypertensivemothers. We also could not demonstrate in our study populationthat early NP was a risk factor for NI. Only 20 infants born topreeclamptic mothers developed early NP in our series but theirrisk of NI was not higher than the 170 infants that born tonormotensive mothers without early NP. This further supports ourbelief that early NP does not increase the risk of NI but the limitedcase number does not provide enough power for the conclusion.

Conclusion

Our study showed that early NP that developed in VLBW infantswithout congenital infection during the first week of life did notresult in a higher likelihood of NI among infants born to eitherpreeclamptic or normotensive mothers. This observation does notsupport the routine use of rhG-CSF for early onset NP to prevent NIin VLBW infants.

Acknowledgments

Part of the results were presented as a poster in 2005 Society of Pediatric Research

meeting in Washington, DC.

References

1 Fanaroff AA, Korones SB, Wright LL, Wright EC, Poland RL, Bauer CB et al.

A controlled trial of intravenous immune globulin to reduce nosocomial infections

in very-low-birth-weight infants. N Engl J Med 1994; 330: 1107–1113.

2 Stoll BJ, Hansen N, Fanaroff AA, Wright LL, Carlo WA, Ehrenkranz RA et al. Late-onset

sepsis in very low birth weight neonates: the experience of the NICHD Neonatal

Research Network. Pediatrics 2002; 110: 285–291.

3 Kaufman D, Fairchild KD. Clinical microbiology of bacterial and fungal sepsis in

very-low-birth-weight infants. Clin Microbiol Rev 2004; 17: 638–680.

4 Stoll BJ, Gordon T, Korones SB, Shankaran S, Tyson JE, Bauer CR et al. Late-onset

sepsis in very low birth weight neonates: a report from the National Institute of Child

Health and Human Development Neonatal Research Network. J Pediatr 1996; 129:

63–71.

5 Makhoul IR, Sujov P, Smolkin T, Lusky A, Reichman B. Epidemiological, clinical, and

microbiological characteristics of late-onset sepsis among very low birth weight infants

in Israel: a national survey. Pediatrics 2002; 109: 34–39.

6 Auriti C, Maccallini A, Di Liso G, Di Ciommo V, Ronchetti MP, Orzalesi M. Risk factors

for nosocomila infections in a neonatal intensive-care unit. J Hosp Infect 2003; 53:

25–30.

7 Manzoni P, Maestri A, Leonessa M, Mostert M, Farina D, Gomirato G. Fungal and

bacterial sepsis and threshold ROP in preterm very low birth weight neonates.

J Perinatol 2006; 26: 23–30.

8 Schelonka RL, Infante AJ. Neonatal immunology. Semin Perinatol 1998; 22: 2–14.

9 Yang KD, Chen MZ, Teng RJ, Yang MY, Liu HC, Chen RF et al. A model to study

antioxidant regulation of endotoxemia-modulated neonatal granulopoiesis and

granulocyte apoptosis. Pediatr Res 2000; 48: 829–834.

10 Zafar N, Wallace CM, Kieffer P, Schroeder P, Schootman M, Hamvas A. Improving

survival of vulnerable infants increases neonatal intensive care unit nosocomial

infection rate. Arch Pediatr Adolesc Med 2001; 155: 1098–1104.

11 Kumar P. Effect of decreased use of postnatal corticosteroids on morbidity in extremely

low birthweight infants. Am J Perinatol 2005; 22: 77–81.

12 Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood count in health

and disease. I. Reference values for neutrophilic cells. J Pediatr 1979; 95: 89–98.

13 Davies NP, Buggins AGS, Snijders RJM, Jenkins E, Layton DM, Nicolaides KH. Blood

leukocyte count in the human fetus. Arch Dis Child 1992; 67: 399.

14 Ohlsson A, Lacy JB. Intravenous immunoglobulin for preventing infection in preterm

and/or low-birth-weight infants. Cochrane Database Syst Rev 2004: CD000361.

15 Craft AP, Finer NN, Barrington KJ. Vancomycin for prophylaxis against sepsis in

preterm neonates. Cochrane Database Syst Rev 2000: CD001971.

16 Kocherlakota P, La Gamma EF. Preliminary report: rhG-CSF may reduce the incidence

of neonatal sepsis in prolonged preeclampsia-associated neutropenia. Pediatrics 1998;

102: 1107–1111.

17 Funke A, Berner R, Traichel B, Schmeisser D, Leititis JU, Niemeyer CM. Frequency,

natural course, and outcome of neonatal neutropenia. Pediatrics 2000; 106: 45–51.

18 Juul SE, Hayes JW, McPherson RJ. Evaluation of neutropenia and neutrophilia in

hospitalized preterm infants. J Perinatol 2004; 24: 150–157.

19 Cadnapaphornchai M, Faix RG. Increased nosocomial infection in neutropenic low

birth weight (2000 g or less) infants of hypertensive mothers. J Pediatr 1992; 121:

956–961.

20 Koenig JM, Christensen RD. Incidence, neutrophil kinetics, and natural history of

neonatal neutropenia associated with maternal hypertension. N Engl J Med 1989; 321:

557–562.

21 Gray PH, Rodwell RL. Neonatal neutropenia associated with maternal hypertension

poses a risk for nosocomial infection. Eur J Pediatr 1999; 158: 71–73.

22 Mouzinho A, Rosenfeld CR, Sanchez PJ, Risser R. Revised reference ranges for

circulating neutrophils in very-low-birth-weight neonates. Pediatrics 1994; 94:

76–82.

23 Baley JE, Stork EK, Warkentin PI, Shurin SB. Neonatal neutropenia: clinical

manifestations, cause and outcome. Am J Dis Child 1988; 142: 1161–1165.

24 Engle WD, Rosenfeld CR. Neutropenia in high-risk neonates. J Pediatr 1984; 105:

982–986.

25 Gessler P, Luders R, Konig S, Haas N, Lasch P, Kachel W. Neonatal neutropenia in low

birth weight premature infants. Am J Perinatol 1995; 12: 34–38.

26 Mouzinho A, Rosenfeld CR, Sanchez PJ, Risser R. Effect of maternal hypertension on

neonatal neutropenia and risk of nosocomial infection. Pediatrics 1992; 90: 430–435.

27 Walsh MC, Kliegman RM. Necrotizing enterocolitis: treatment based on staging criteria.

Pediatr Clin North Am 1986; 33: 179–201.

28 Papile LA, Burnstein J, Burnstein R, Koffler H. Incidence and evolution of

subependymal and intraventricular hemorrhage: a study of infants with birthweight

less than 1500 gm. J Pediatr 1978; 92: 529–534.

29 Girard R, Fabry J, Meynet R, Lambert DC, Sepetjan M. Costs of nosocomial infection in

a neonatal unit. J Hosp Infect 1983; 4: 361–366.

30 Freeman J, Epstein MF, Smith NE, Platt R, Sidebottom DG, Goldmann DA. Extra

hospital stay and antibiotic usage with nosocomial coagulase-negative staphylococcal

bacteremia in two neonatal intensive care unit populations. Am J Dis Child 1990; 144:

324–329.

31 Baker CJ, Melish ME, Hall RT, Casto DT, Vasan U, Givner LB. Intravenous immune

globulin for the prevention of nosocomial infection in low-birth-weight neonates. The

multicenter Group for the Study of Immune Globulin in Neonates. N Engl J Med 1992;

327: 213–219.

32 Gray JE, Richardson DK, McCormick MC, Goldmann D. Coagulase-negative

staphylococcal bacteremia among very low birth weight infants: relation to admission

illness severity. Pediatrics 1995; 95: 225–230.


223


33 Sohn AH, Sinkowitz-Cochran RL, Stover BH, Siegel JD, Jarvis WR. Prevalence of

nosocomial infections in neonatal intensive care units: results from the first national

point-prevalence survey. J Pediatr 2001; 139: 821–827.

34 Engle WD, Rosenfeld CR. Neutropenia in high-risk neonates. J Pediatr 1984; 105:

982–986.

35 Baley JE, Stork EK, Warkentin PI, Shurin SB. Neonatal neutropenia: clinical

manifestation, cause and outcome. Am J Dis Child 1988; 142: 1161–1165.

36 Gessler P, Luders R, Konig S, Haas N, Lasch P, Kachel W. Neonatal neutropenia in low

birth weight premature infants. Am J Perinatol 1995; 12: 34–38.

37 Doron MW, Makhlouf RA, Katz VL, Lawson EE, Stiles AD. Increased incidence of sepsis

at birth in neutropenic infants of mothers with preeclampsia. J Pediatr 1994; 125:

452–458.

38 Paul DA, Leef KH, Sciscione A, Tuttle DJ, Stefano JL. Preeclampsia does not increase the

risk for culture proven sepsis in very low birth weight infants. Am J Perinatol 1999; 16:

365–372.

39 Tsao PN, Teng RJ, Tang JR, Yau KI. Granulocyte colony-stimulating factor in the cord

blood of premature neonates born to mothers with pregnancy-induced hypertension.

J Pediatr 1999; 135: 56–59.


224


ORIGINAL ARTICLE

Comparison of serum amyloid A concentrations with those ofC-reactive protein and procalcitonin in diagnosis and follow-upof neonatal sepsis in premature infantsM Cetinkaya1, H Ozkan1, N Koksal1, S Celebi2 and M Hacımustafaoglu2

1Division of Neonatology, Department of Pediatrics, Uludag University, Bursa, Turkey and 2Division of Pediatric Infectious Disease,Department of Pediatrics, Uludag University, Bursa, Turkey

Objective: The purpose of this study was to determine the role of serum

amyloid A (SAA) in diagnosis of neonatal sepsis and evaluation of clinical

response to antibiotic therapy. We also aimed to compare the efficiency of

SAA with that of C-reactive protein (CRP) and procalcitonin (PCT) in

diagnosis and follow-up of neonatal sepsis in preterm infants.

Study Design: A total of 163 infants were enrolled in this prospective

study. The infants were classified into four groups: group 1 (high

probable sepsis), group 2 (probable sepsis), group 3 (possible sepsis) and

group 4 (no sepsis, control group). Blood samples for whole blood count,

CRP, PCT, SAA and culture were obtained before initiating antibiotic

treatment. This procedure was repeated three times at 48 h, 7 and 10 days.

Result: Initial CRP, PCT and SAA levels were found to be positive in 73.2,

75.6 and 77.2% of all infants, respectively. Sensitivities of CRP, PCT and

SAA at 0 h were 72.3, 74.8 and 76.4%, respectively. Although it was not

statistically significant, SAA was found to be more sensitive than CRP and

PCT in diagnosis of neonatal sepsis. The area under the curve (AUC) for

CRP, PCT and SAA at 0 h were 0.870, 0.870 and 0.875, respectively.

Although the AUC for SAA at 0 h was higher than PCT and CRP, the

difference was not statistically significant.

Conclusion: SAA is an accurate and reliable marker for diagnosis and

follow-up of neonatal sepsis. It is especially useful at the onset of

inflammation for rapid diagnosis of neonatal sepsis and can be safely and

accurately used in combination with other sepsis markers such as CRP

and PCT in diagnosis and follow-up of neonatal sepsis in preterm infants.



Keywords: serum amyloid A; C-reactive protein; procalcitonin; neonatalsepsis; newborn

Introduction

Neonatal sepsis remains as an important cause of neonatalmorbidity and mortality, despite the major advances in themanagement of newborn infants.1 Early warning signs andsymptoms of neonatal sepsis are often nonspecific and subtleespecially at the onset of infection, and can easily be confused withother common noninfectious causes.2 The incidence of sepsisranges from 1 to 10 cases in every 1000 live births, with highmortality rates despite antibiotic treatment.3 One of the mostimportant aspects is the ability to obtain an early diagnosis andthus early therapy.3 Therefore, antibiotics are started immediatelyin newborn infants who have nonspecific findings of infection andare continued until the final result of the blood culture isobtained.4,5 Blood culture is the most valuable and gold-standarddiagnostic method, but it may yield false-positive results because ofcontamination. Also, blood culture can remain negative despitegeneralized bacterial infection. Body fluid cultures, determinationof bacterial antigens, white blood cell count, acute-phase proteins(C-reactive protein (CRP), haptoglobin, fibrinogen, a1-antitrypsin), interleukin (IL) and inflammatory cytokines andprocalcitonin (PCT) are other laboratory tests, which are used tosupport the neonatal sepsis diagnosis.6–8

Serum amyloid A (SAA) term groups a family of 12 to 14 kDapolymorphic apolipoproteins that are mainly produced by the liver.They are regarded as acute-phase proteins because they increaseconsiderably during infection. They can show as much as a 1000-fold increase in 8 to 24 h after the onset of sepsis. SAA has beenshown to be useful in various acute diseases (bacterial, viral,traumatic, rheumatic and ischemic heart disease) and also in thediagnosis of sepsis in neonates.3,9–11 Although the data about itsuse in diagnosis of neonatal sepsis is limited, recently it has beensuggested as a superior marker compared with CRP.12

CRP is a protein produced by the liver in response toinflammatory and/or infectious stimuli, and for this reason it isregarded as an acute-phase protein. It may increase 12 to 24 h afterthe exposure to endotoxins. CRP may also increase in a number ofprenatal conditions such as fetal distress, stressful delivery and

Received 6 August 2008; revised 27 October 2008; accepted 6 November 2008; published online

11 December 2008

Correspondence: Dr M Cetinkaya, Uludag Universitesi Tıp Fakultesi, Cocuk Saglıgı ve

Hastalıkları ABD, Gorukle, Bursa 16059, Turkey.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.207



maternal fever in the absence of systemic infection. Therefore, itsspecificity is accepted low and it is preferably used in combinationwith another serum marker.12

PCT is the precursor protein of calcitonin and is produced bymonocytes and hepatocytes. It begins to rise 2 to 4 h after exposureto bacterial endotoxins, increases rapidly peaking in 6 to 8 h, andreaches a plateau and then decreases to normal levels after 24 h.Serum PCT levels appear to correlate with the severity of microbialinvasion and they decrease rapidly after appropriate antibiotictherapy. Normal serum and plasma levels of PCT are less than0.5 ng ml�1. Levels above this value have been accepted to bepathological.13–16 It has also been suggested that PCT levels maybe increased in neonatal sepsis and its diagnostic utility iscomparable that of with CRP.12

As stated previously, clinical signs of neonatal sepsis arenonspecific and there is no sensitive and specific laboratory test fordetecting an early-stage infection. The purpose of this study was todetermine the role of SAA in diagnosis of neonatal sepsis and inevaluation of clinical response to antibiotic therapy. In addition, wealso aimed to compare the efficiency of SAA with that of CRP andPCT in diagnosis and follow-up of neonatal sepsis in preterminfants.

Methods

In this prospective study, 185 premature infants born betweenJanuary 2006 and January 2008 in the neonatal intensive care unit(NICU) of the Pediatric Department of Uludag University, Facultyof Medicine were initially planned to be enrolled. However, 22infants were excluded from the study population because parents of14 infants (4 in study group and 10 in control group) rejectedinclusion of their children and 8 infants (all from the study group)were already receiving antibiotics on admission. Therefore, a totalof 163 premature infants (123 infants with neonatal sepsis and 40controls) were included in this study. This study was originallyplanned to investigate 120 infants in each group. However, thenumber of patients in control group did not reach the initiallyplanned level because most infants hospitalized during the studyperiod had major respiratory, cardiac and neurological problemspartly because of the fact that our NICU is a tertiary care neonatalreferral unit in South Marmara region of Turkey. The infants wereclassified into four groups according to the criteria defined by Gittoet al.17 group 1 (high probable sepsis), group 2 (probable sepsis),group 3 (possible sepsis) and group 4 (no sepsis, control group).Infants with neonatal convulsion, neonatal hypoglycemia,neonatal hyperbilirubinemia who had no signs of clinicaland laboratory infection were referred to as the control group.Table 1 lists the criteria for classifying study groups.

The study protocol was approved by the Ethics Committee ofUludag University, Faculty of Medicine. Informed parental consentwas obtained for all infants. Exclusion criteria included

administration of antibiotic therapy at admission and refusal ofparental consent. Gestational age, birth weight, gender, mode ofdelivery, Apgar score at 1 and 5 min, prenatal demographics,premature rupture of membranes (PROM) and history ofchorioamnionitis were all recorded. Temperature instability, apnea,need for supplemented oxygen, need for ventilation, tachycardia/bradycardia, hypotension, feeding intolerance, abdominaldistension, necrotizing enterocolitis were considered among clinicalsigns of sepsis. The changes in the hematologic parameters wereprocessed according to the Manroe and Rodwell scoringsystems.18,19 Leukopenia was defined as leukocyte count<5000 mm�3; leukocytosis was defined as leukocyte count>25 000 mm�3 at birth, >30 000 mm�3 at 12 to 24 h and>21 000 mm�3 after the second day. Thrombocytopenia wasdefined as platelet count <150 000 mm�3. Normal absoluteneutrophil count was accepted as 7800 to 14 500 mm�3 in the first60 h and 1750 to 5400 mm�3 after 60 h. Before initiating theantimicrobial therapy, blood samples for whole blood count, CRP,PCT, SAA and culture were obtained both from neonates with sepsisand from control patients.

This procedure was repeated three times at 48 h, 7 and 10 days.Cerebrospinal fluid (CSF), urine, and tracheal and gastricmaterials were also sent out for culture, if obtained. Blood smearsof all infants were also evaluated for the findings of sepsis when

Table 1 Criteria employed for defining the sepsis score

Groups Criteria

Group 1 At least 3 sepsis-related clinical signsa

High probable sepsis CRP >1 mg per 100 ml

At least 2 other altered serum parameters in

addition to CRPb

Blood culture; positive or negative

Group 2 Less than 3 sepsis-related clinical signsa

Probable sepsis CRP > 1 mg per 100 ml

At least 2 other altered serum parameters in

addition to CRP

Blood culture; negative

Group 3 Less than 3 sepsis-related clinical signsa

Possible sepsis CRP <1 mg per 100 ml

Less than 2 other altered serum parameters


Group 4 No sepsis-related clinical signsa

No sepsis CRP <1 mg per 100 ml

No altered serum parameters


Abbreviation: CRP, C-reactive protein.aSepsis-related clinical signs: temperature instability, apnea, need for supplementedoxygen, need for ventilation, tachycardia/bradycardia, hypotension, feeding intolerance,abdominal distension, necrotizing enterocolitis.bSerum parameters other than CRP: white blood cell count, absolute neutrophil count,platelet count.

SAA, CRP and PCTM Cetinkaya et al

226


blood samples obtained. Meningitis was diagnosed according tothe cell count, glucose and protein levels of CSF along withCSF culture.

Whole blood count, PCT, CRP, SAA levels and cultures werestudied immediately. Whole blood count was performed using anautomated counter, Cell Dyn 3700 (Abbott Diagnostics Division,Santa Clara, CA, USA). CRP and SAA were determined by animmunonephelometric method using BN II device (Dade BehringMarburg GMBH, Marburg, Germany). Detection limits were 0.5 and6.8 mg per 100 ml for CRP and SAA, respectively. PCT wasmeasured by monoclonal immunoluminometric assay (LumitestPCT; Brahm Diagnostica GMBH, Berlin, Germany), which isspecific for PCT molecule. In this assay, two different antibodies,one directed to calcitonin and other directed to katacalcin wereused. Levels greater than 0.5 ng ml�1 were accepted aspathological. Levels lower than 0.5 ng ml�1 for PCT and CRP and6.8 mg per 100 ml for SAA were accepted as 0 for statistical analysis.Blood and CSF cultures were analyzed using fully automatedBACTEC method by BACTEC 9240 device (Becton Dickinson,Heidelberg, Germany).

Infants were treated with appropriate antibiotic therapies.Neonates who had positive cultures were treated with antibioticsaccording to the culture antibiogram. The antimicrobial therapywas stopped after clinical and laboratory improvement. All of thegroups were compared according to demographic features, clinicaland laboratory findings.

SPSS 12.0/Windows program was used for data analyses.Descriptive statistics were given as mean, median, standarddeviation, minimum, maximum and percentage. The significancebetween groups were evaluated with w2-test and McNemar’s test for

qualitative data and with Friedman and Wilcoxon tests forquantitative data. Correlations between quantitative data wereanalyzed by Spearman’s correlation test. Values of P<0.05 wereconsidered to be significant. In subgroup comparisons, Bonferronicorrection was used for P significance levels.

Results

A total of 163 preterm infants were enrolled in this study. Thedemographic and clinical characteristics of the study populationare shown in Table 2. There were 108, 5, 10 and 40 neonates ingroups 1, 2, 3 and 4, respectively. But, as the number of infants ingroups 2 and 3 were too small for statistical analysis, they weregrouped together as group 2þ 3. There were no differencesbetween the groups with respect to gestational age, birth weight,gender, Apgar scores at 1 and 5 min, the mode of delivery andpresence of PROM. Hypoxia, apnea, jaundice, feeding intoleranceand chorioamnionitis were also found to be significantly higher ingroup 2þ 3 than those in group 1. However, all these differenceswere associated with the small number of infants in group 2þ 3compared with those in group 1. There were no differences betweeninfants with sepsis with respect to other characteristics.

In group 1, 10 infants (9.3%) had early-onset sepsis (EOS) and98 infants (90.7%) had late-onset sepsis (LOS). There were eightinfants (53.3%) with EOS and seven infants (47.7%) with LOS ingroup 2þ 3. Pneumonia was diagnosed in 52.8 and 33% of theinfants in group 1 and group 2þ 3, respectively. Neonatalmeningitis was determined in 23.1 and 38.5% of infants in group 1and group 2þ 3, respectively. There were no statisticallysignificant differences between two groups with respect to

Table 2 Birth and clinical characteristics of study group

Group 1 (n¼ 108) Group 2+3 (n¼ 15) Group 4 (n¼ 40)

Gestational age (week)a 31.2±3.19 (25–37) 31.6±2.9 (26–36) 32.7±1.56 (32–35)

Birth weight (g)b 1745 (690–2700) 1685 (750–2550) 1725 (1200–2250)

Males/females 58/50 10/5 22/18

Apgar min 1b 5.7 (1–9) 5.6 (2–9) 7.4 (4–9)

Apgar min 5b 6.7 (5–9) 7.0 (5–9) 7.5 (6–10)

Cesarean deliveryc 72 (66.7) 11 (73.3) 20 (50)

PROMc 12 (11.1) 2 (13.3) 0 (0)*

Choriamnionitisc 4 (3.7) 1 (6.7) 0 (0)*

Hypoxiac 6 (5.6) 6 (40) 0 (0)*

Apneac 6 (5.6) 4 (26.7) 0 (0)*

Feeding intolerancec 36 (33.3) 4 (26.3) 0 (0)*

Jaundicec 22 (20.4) 9 (60) 1 (2.5)*

Mortalityc 25 (23.1) 1 (6.7) 0 (0)*

Abbreviation: PROM, premature rupture of membrane.*P<0.05 between sepsis and nonsepsis group.aMedian±standard deviation (range).bMedian (range).cn (%).


227


meningitis and pneumonia. Total 72 infants had positive bloodculture for gram-positive sepsis (66 Staphylococcus epidermidis, 4group B Streptococcus, 1 Enterococcus faecalis, 1Corynebacterium matruchotii), 13 for gram-negative sepsis(6 Escherichia coli, 4 Pseudomonas aeruginosa, 3 Klebsiellapneumonia) and 14 for fungal sepsis (10 Candida parapsilosis,4 Candida albicans). Out of 123, 26 (21.1%) infants in sepsisgroup died during the study period. Nine of them died because ofnoninfectious causes. Twenty-five of them were in group 1, onlyone was in group 2þ 3. The difference between mortality rates wasfound to be statistically significant.

Mean initial leukocyte counts were determined as14 377±11 065, 14 913±7788 and 12 575±2763 mm�3 ingroups 1, 2þ 3 and 4, respectively. No significant differences werefound between sepsis groups in terms of mean leukocyte counts.Although the mean leukocyte counts in sepsis group was higherthan those in control group, the difference was not statisticallysignificant. Mean leukocyte counts in sepsis group significantlydecreased at 48 h, 7 and 10 days. There were no differences betweensepsis groups with respect to platelet counts and thrombocytopenia.There were also no differences between infants with positive ornegative blood cultures with respect to CRP, PCT and SAA levels(P>0.05).

Initial CRP levels were positive in 73.2% of infants in sepsisgroup. CRP levels reached to peak levels at 48 h and then decreasedsignificantly. PCT levels were found to be positive in 75.6% ofinfants in sepsis group and these levels declined throughout thestudy period. SAA levels were positive in 77.2% of infants with sepsisand it also returned to normal levels after 48 h. Figure 1 shows thebaseline levels of CRP, PCT and SAA with their trends throughoutthe study period.

SAA values of 92.2% of patients with CRP values beyond upperlimits were higher than normal limits, however, the results werenot statistically significant (P¼ 0.359). SAA values of 7 (7.8%)patients with CRP values beyond upper limits were within normallimits, and CRP values of 12 (12.6%) patients with SAA valuesbeyond upper limits were within normal limits. CRP results showed87.8% correlation with PCT being beyond upper limits(P¼ 0.690). SAA levels of 93.5% of patients with PCT valuesbeyond upper limits also were higher than normal, and there wasno statistical difference between groups (P¼ 0.791). Statisticalanalyses demonstrated that the rate (73.2%) of sepsis diagnosiswith only CRP positivity would increase to 82.9 and 84.6%following additional analysis of SAA and PCT, respectively.Therefore, it is estimated that the combination of SAA, CRP andPCT for the diagnosis of neonatal sepsis would increase thediagnosis rate from 72 to 76% to 82 to 85%.

There were positive correlations between laboratory values ofCRP, PCT and SAA (that is, r: 0.627 for CRP and SAA, P<0.001;r: 0.602 for PCT and SAA, P<0.001 and r: 0.563 for PCT and CRP,P<0.001) at the onset of the study. Mean percent increases in CRP,PCT and SAA compared with their normal laboratory values were71.4, 75.8 and 70.8%, respectively. There were no statisticaldifferences in terms of mean percent increases between CRP andSAA, or CRP and PCT, whereas mean percent increase in PCT wasstatistically higher (P¼ 0.004) than that in SAA according to thecoefficient of variations (Figure 2). When CRP, PCT, SAA levels andthe percentage variations at baseline, 48 h, 7 and 10 days wereevaluated in group 1 and group 2þ 3 separately, the results inboth groups were similar to the total patient group and there wereno significant differences between groups.

The sensitivity of CRP, PCT and SAA at 0 h were 72.3, 74.8 and76.4%, respectively. Although it was not statistically significant, SAAwas found to be more sensitive than CRP and PCT in the diagnosisof neonatal sepsis (Table 3). Initial CRP, PCT and SAA levels were

CRP PCT SAA

0.00

10.00

20.00

30.00

40.00

50.00

60.00

95% CI

0 hours48 hoursDay 7 Day 10

Figure 1 Figure showing C-reactive protein (CRP), procalcitonin (PCT) andserum amyloid A (SAA) levels measured baseline, and trends of these levels during48 h, day 7 and day 10.

2.6

2.1

1.81.61.3

1.91.7

1.4

3.73.1

2.12.5

0

1

2

3

4

Baseline 48 hours day 7 day 10

Coefficient of variation of SAA Coefficient of variation of CRP Coefficient of variation of PCT

Figure 2 Figure showing the coefficient of variations of C-reactive protein(CRP), procalcitonin (PCT) and serum amyloid A (SAA).


228


significantly higher in sepsis group than those in control group(Table 4). Significant differences were observed between sepsis(groups 1þ 2þ 3) and control (group 4) groups with respect toinitial CRP, PCT and SAA levels, as expected (Table 5).

Although the area under the curve (AUC) for SAA at 0 h washigher than that in PCT and CRP, the difference was not

statistically significant. There were no statistical differences between

CRP, SAA and PCT according to PPV, NPV, AUC at 0, 48 h and day

7. Table 3 shows the sensitivity, specificity, PPV, NPV and AUC

values for CRP, PCT and SAA at 0, 48 h and day 7. Figure 3 shows

the comparison of AUC for CRP, PCT and SAA at 0, 48 h and day 7.

Discussion

Systemic infection is a devastating and important cause ofmorbidity and mortality in both term and preterm infants.Although the initial clinical signs and symptoms are subtle, theclinical course can rapidly progress and worsen, and may lead todisseminated intravascular coagulation and death within hours.2,20

Also, noninfected infants such as those with transient tachypnea ofthe newborn, meconium aspiration syndrome, respiratory distresssyndrome, apnea of prematurity and acute exacerbation of chroniclung disease are often clinically indistinguishable from infants whoare really in the initial stages of bacterial infection.21 If the absence

Table 3 Comparison of CRP, PCT and SAA according to sensitivity, specificity, PPV, NPV and AUC at 0, 48 h and 7 days

Marker Time Sensitivity (%; 95% Cl) Specificity (%; 95% Cl) PPV (%) NPV (%) AUC P-value

CRP, >0.5 mg per 100 ml 0 h 72.3 (63.6–80) 100 (91.1–100) 100 54 0.870

48 h 71.5 (62.4–79.5) 100 (91.1–100) 100 54.8 0.861

7 days 50 (40.5–59.5) 100 (91.1–100) 100 41.2 0.758

PCT, >0.5 mg per 100 ml 0 h 74.8 (66.2–82.2) 100 (91.1–100) 100 56.3 0.870 >0.05

48 h 57.7 (48.2–66.9) 100 (91.1–100) 100 44.9 0.807

7 days 28.4 (20.5–37.6) 100 (91.1–100) 100 32.5 0.657

SAA, >6.8 mg per 100 ml 0 h 76.4 (67.9–83.6) 100 (91.1–100) 100 58 0.875

48 h 66.3 (57–74.9) 100 (91.1–100) 100 50.6 0.837

7 days 44.7 (35.4–54.3) 100 (91.1–100) 100 38.8 0.712

Abbreviations: AUC, area under the curve; CI, confidence interval; CRP, C-reactive protein; NPV, negative predictive value; PCT, procalcitonin; PPV, positive predictive value; SAA, serumamyloid A.

Table 4 Table showing the mean CRP, PCT and SAA levels in three groups at 0, 48 h, 7 and 10 days

Time after sepsis onset High probably sepsis (n¼ 108) Probable sepsis+possible sepsis (n¼ 15) No sepsis (n¼ 40)

CRP mg per 100 ml

0 h 3.9±5.1* 1.7±2.9* 0.3±0.5

48 h 4.8±8.1 1.9±2.9 N/A

7 days 2.3±4.2 0.9±1.0 N/A

10 days 1.1±2.3 0.3±0.6 N/A

PCT mg per 100 ml

0 h 6.2±15.7* 5.1±7.5* 0.1±0.4

48 h 4.2±9.1 4.0±6.8 N/A

7 days 1.1±3.6 1.1±1.9 N/A

10 days 0.3±1.1 0.1±0.2 N/A

SAA mg per 100 ml

0 h 44.4±57.3* 37.9±41.7* 3.2±3.4

48 h 39.5±67.1 40.8±34.5 N/A

7 days 15.7±28.7 22.0±41.7 N/A

10 days 5.8±15.0 2.9±5.9 N/A

Abbreviations: CRP, C-reactive protein; N/A, not available; PCT, procalcitonin; SAA, serum amyloid A.*P<0.05.


229


of systemic infection can be detected early, then the number ofinfants started on antibiotics could be reduced, the length ofhospitalization could be shortened, and the potential for emergenceof resistant organisms could be lessened. However, to date, althoughinfection markers might help to diagnosis, no single laboratory testhas provided rapid and reliable identification of early infectedneonates.22 Hence, it is important to diagnose neonatal sepsis in arapid and accurate way especially in preterm infants. Therefore, weperformed this prospective study to determine possible diagnosticvalue of SAA in neonatal sepsis and also compared its efficiency withmore widely used markers CRP and PCT.

CRP is the most commonly used acute-phase reactant inneonates.23 As CRP increases 12 to 24 h after the onset of infection,

it is usually used in combination with other markers. PCT isanother marker, which has been used recently in combination withCRP in the diagnosis of neonatal sepsis. High PCT levels werereported in neonates with early- or late-onset neonatal sepsis.24

There are also studies comparing CRP and PCT in diagnosisand follow-up of neonatal sepsis. There are conflicting resultsabout their superiority to each other in diagnosis of neonatal sepsisin different studies. In our previous study, serum PCT levels seemedto be superior to serum CRP levels in terms of early diagnosis ofneonatal sepsis, in detecting the severity of the illness, and inevaluation of the response to antibiotic treatment.25 This wassimilar to findings of some,26,27 but not all28,29 studies. Therefore,in some studies multiple comparisons were made between CRP,PCT and other cytokines such as IL-6, IL-8 and tumor necrosisfactor-a. As a result of these conflicting results, we aimed tocompare SAA with PCT and CRP in the diagnosis and follow-up ofneonatal sepsis.

SAA has inhibitory effects on inflammation by reducing theproduction of prostaglandin E2 and oxidative respiration ofneutrophils, counteracting the pyrogenic effect of a number ofcytokines, inhibiting platelet activation, negatively controlling theproduction of antibodies and inducing the secretion of collagenaseby fibroblasts.3 Therefore, it has been started to use in diagnosis ofneonatal sepsis with other laboratory tests.

Arnon et al.30 reported that SAA could be used as a reliablemarker for early detection of LOS in preterm infants. These authorsalso stated that SAA levels had prognostic value in the first 24 hafter the onset of neonatal sepsis.31 They established that SAA hadhigher levels, and rose earlier and sharper than CRP.10 In ourstudy, we found that SAA had higher levels during the initial and48-h evaluation of infants with sepsis. We have also shown that itdeclined faster than CRP and PCT during the follow-up of sepsis.These findings are all concordant with literature and we suggestthat it can be used safely in diagnosis of neonatal sepsis.

Shortland et al.32 reported that CRP remained normal in 54% ofvery low-weight premature infants during culture positive neonatalsepsis. No differences between septic infants with positive or

Table 5 Table showing the comparison of mean CRP, PCT and SAA levels insepsis (groups 1+2+3) and control (group 4) groups at 0, 48 h, 7 and 10 days

Time after

sepsis onset

Sepsis group

(groups 1+2+3, n¼ 123)

Control group

(group 4, n¼ 40)

CRP mg per 100 ml

0 h 3.6±4.9* 0.3±0.5

48 h 4.4±7.6 N/A

7 days 2.1±3.9 N/A

10 days 1.0±2.1 N/A

PCT mg per 100 ml

0 h 6.0±14.9* 0.1±0.4

48 h 4.1±8.8 N/A

7 days 1.1±3.4 N/A

10 days 0.2±1.0 N/A

SAA mg per 100 ml

0 h 43.6±55.5* 3.2±3.4

48 h 39.6±64.2 N/A

7 days 16.5±30.5 N/A

10 days 5.3±14.1 N/A

Abbreviations: CRP, C-reactive protein; N/A, not available; PCT, procalcitonin; SAA, serumamyloid A.

*P<0 05

CRP0PCT0

SAA0

0 20 40 60 80 100

100

80

60

40

20

0

100-Specificity

Sensitivity CRP48

PCT48SAA48

0 20 40 60 80 100

100

80

60

40

20

0

100-Specificity

Sensitivity CRP7

PCT7SAA7

0 20 40 60 80 100

100

80

60

40

20

0

100-Specificity

Sensitivity

Figure 3 Graphics showing the comparison of the area under curve (AUC) for C-reactive protein (CRP), procalcitonin (PCT) and serum amyloid A (SAA) at onset, 48 hand day 7.


230


negative cultures with respect to CRP, PCT and SAA levels werefound in our study. These findings may suggest that culturepositivity does not increase CRP, PCT or SAA levels in infants withneonatal sepsis.

Arnon et al.11 found that SAA had significantly largest AUCcompared with CRP at 0 h of sepsis evaluation. Enguix et al.22

reported that SAA, CRP and PCT had the same diagnostic efficiencyand the same area under the ROC curve in neonates with bacterialsepsis. In our study, we also found no significant differencesbetween CRP, PCT and SAA in terms of AUC, PPV, NPV andspecificity.

In our hospital, the cost of measuring each CRP or SAA ($4) isone-fourth that of PCT. Therefore, although combining SAA withother two tests for the diagnosis of sepsis may not be consideredcost-effective, our finding that combination of the three methodscould increase the rate of sepsis diagnosis by about 10% suggeststhat measuring all three parameters is reasonable to consider.

The limitation of our study is the small number of patients inthe control group. We believe that the power of the study would beincreased considerably under such circumstance that we were ableto recruit adequate number of infants in the control group.

As a result, SAA is an accurate and reliable marker for diagnosisand follow-up of neonatal sepsis. It is especially useful at the onsetof inflammation with rapid diagnosis of neonatal sepsis. It can alsohelp clinicians to follow the clinical course of neonatal sepsis up.The rapid diagnosis of neonatal sepsis will reduce the morbidityand mortality of sepsis by starting the antibiotic therapy as soon aspossible. Also, it can be used to establish the duration of treatmentand response to treatment. Therefore, SAA can be safely andaccurately used either alone or in combination with other sepsismarkers such as CRP and PCT in diagnosis and follow-up ofneonatal sepsis in preterm infants. Future studies with largernumber of patients are warranted for determining the best markerfor the diagnosis of neonatal sepsis.

References

1 Ng PC, Lam HS. Diagnostic markers for neonatal sepsis. Curr Opin Pediatr 2006; 18:

125–131.

2 Ng PC. Clinical trials for evaluating diagnostic markers of infection in neonates. Biol

Neonate 2005; 87: 111–112.

3 Pızzını C, Mussap M, Plebanı M, Fanos V. C-reactive protein and serum amyloid a

protein in neonatal infections. Scand J Infect Dis 2000; 32: 229–235.

4 Remington JS, Klein JO. Infectious Diseases of the Fetus and Newborn Infant. WB

Saunders Company: Philadelphia, 2001, pp 943–985.

5 Chiesa C, Panero A, Rossi N, Stegagno M. Reliability of procalcitonin concentrations

for the diagnosis of sepsis in critically ill neonates. Clin Infect Dis 1998; 26: 664–672.

6 Folli HL, Poole RL, Benitz WE, Russo JC. Medication error prevention by clinical

pharmacist in two pediatric hospitals. Pediatrics 1987; 79: 718–722.

7 Philip AG, Hewitt JR. Early diagnosis of neonatal sepsis. Pediatrics 1980; 65: 1036–1041.

8 Rodwell RL, Leslie AL, Tudehope DI. Early diagnosis of neonatal sepsis using a

hematological scoring system. J Pediatr 1988; 112: 761–767.

9 Laurenti F, Fassi F, Campi E, Ceri E, Ligi L, Donato AI et al. Origine e significato delle

proteine ‘maggiori’ della fase acuta. Aggiornamenti in Neonatologia 1996; 4: 173–193.

10 Arnon S, Litmanovitz I, Regev RH, Bauer S, Shainkin-Kestenbaum R, Dolfin T. Serum

amyloid A: an early and accurate marker of neonatal early-onset sepsis. J Perinatol

2007; 5: 297–302.

11 Arnon S, Litmanovitz I, Regev R, Bauer S, Lis M, Sha R. Serum amyloid a protein

is a useful inflammatory marker during late-onset sepsis. Biol Neonate 2005; 87:

105–110.

12 Arnon S, Litmanovitz I. Diagnostic tests in neonatal sepsis. Curr Opin Infect Dis 2008;

21: 223–227.

13 Whicher J, Bienvenu J, Monneret G. Procalcitonin as an acute phase marker. Ann Clin

Biochem 2001; 38: 483–493.

14 Dandona P, Nix D, Wilson MF, Aljada A, Love J, Assicot M et al. Procalcitonin increase

after endotoxin injection in normal subjects. J Clin Endocrinol Metab 1994; 79:

1605–1608.

15 Assicot M, Gendrel D, Carsin H, Raymond J, Guilbaud J, Bohuon C. High serum

procalcitonin concentrations in patients with sepsis and infection. Lancet 1993; 341:

515–518.

16 Ghillani PP, Motte P, Troalen F, Jullienne A, Gardet P, Le Chevalier T et al.

Identification and measurement of calcitonin precursors in serum of patients with

malignant diseases. Cancer Res 1989; 49: 6845–6851.

17 Gitto E, Karbownik M, Reiter RJ, Tan DX, Cuzzocrea S, Chiurazzi P et al. Effects of

melatonin treatment in septic newborns. Pediatr Res 2001; 50: 756–760.

18 Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood count in health

and disease. I. Reference values for neutrophilic cells. J Pediatr 1979; 95: 89–98.

19 Rodwell RL, Leslie AL, Tudehope DI. Early diagnosis of neonatal sepsis using a

hematological scoring system. J Pediatr 1988; 112: 761–767.

20 Ng PC, Cheng SH, Chui KM, Fok TF, Wong MY, Wong W et al. Diagnosis of late onset

neonatal sepsis with cytokines, adhesion molecule, and C-reactive protein in preterm

very low birthweight infants. Arch Dis Child 1997; 77: 221–227.

21 Ng PC. Diagnostic markers of infection in neonates. Arch Dis Child Fetal Neonatal Ed

2004; 89: 229–235.

22 Enguix A, Rey C, Concha A, Medina A, Coto D, Dieguez MA. Comparison of procalcitonin

with C-reactive protein and serum amyloid for the early diagnosis of bacterial sepsis in

critically ill neonates and children. Intensive Care Med 2001; 27: 211–215.

23 Couto RC, Barbosa JA, Pedrosa TM, Biscione FM. C-reactive protein-guided approach

may shorten length of antimicrobial treatment of culture-proven late-onset sepsis: an

intervention study. Braz J Infect Dis 2007; 11: 240–245.

24 Chiesa C, Pacifico L, Rossi N, Panero A, Matrunola M, Mancuso G. Procalcitonin as a

marker of nosocomial infections in the neonatal intensive care unit. Intensive Care

Med 2000; 26: 175–177.

25 Koksal N, Harmanci R, Cetinkaya M, Hacimustafaoglu M. Role of procalcitonin and

CRP in diagnosis and follow-up of neonatal sepsis. Turk J Pediatr 2007; 49: 21–29.

26 Hatherill M, Tibby SM, Sykes K, Turner C, Murdoch IA. Diagnostic marker of infection:

comparison of procalcitonin with C-reactive protein and leucocyte count. Arch Dis

Child 1999; 81: 417–421.

27 Prat C, Domınguez J, Rodrigo C, Gimenez M, Azuara M, Blanco S et al. Use of

quantitative and semiquantitative procalcitonin measurement to identify children with

sepsis and meningitis. Eur J Clin Microbiol Infect Dis 2004; 23: 136–138.

28 Lapillonne A, Basson E, Monneret G, Bienvenu J, Salle BL. Lack of specificity of

procalcitonin for sepsis diagnosis in premature infants. Lancet 1998; 351: 1211–1212.

29 Janota J, Stranak Z, Belohlavkova S, Mudra K, Simak J. Postnatal increase of

procalcitonin in premature newborns is enhanced by chorioamnionitis and neonatal

sepsis. Eur J Clin Invest 2001; 31: 978–983.

30 Arnon S, Litmanovitz I, Regev R, Lis M, Shainkin-Kestenbaum R, Dolfin T. Serum

amyloid A protein in the early detection of late-onset bacterial sepsis in preterm infants.

J Perinat Med 2002; 30: 329–332.

31 Arnon S, Litmanovitz I, Regev R, Lis M, Shainkin-Kestenbaum R, Dolfin T. The

prognostic virtue of inflammatory markers during late-onset sepsis in preterm infants.

J Perinat Med 2004; 32: 176–180.

32 Shortland DB, MacFayden U, Elston A, Harrison G. Evaluation of C-reactive protein

values in neonatal sepsis. J Perinat Med 1990; 18: 157–163.


231


ORIGINAL ARTICLE

Factors affecting postnatal changes in serum creatinine inpreterm infants with gestational age <32 weeksS Iacobelli1, F Bonsante1, C Ferdinus2, M Labenne1 and J-B Gouyon1,2,3,4

1Neonatal Intensive Care Unit, Department of Paediatrics, University of Dijon, Dijon, France; 2Department of Statistics, Centred’Epidemiologie des Populations EA 4184, University of Dijon, Dijon, France; 3Department of Statistics, Inserm, CIE1, Dijon, France and4Department of Statistics, Centre d’Investigation CliniqueFEpidemiologie Clinique/Essais Cliniques, University of Dijon, Dijon, France

Objective: The Aim of this study was to investigate maternal and neonatal

factors associated with serum creatinine (SeCr) changes in a representative

cohort of preterm newborns during their first week of life.

Study Design: Retrospective study. All the infants born less than 32

weeks of gestational age (GA) and cared for in our neonatal intensive care

unit between January 2001 and December 2005 were eligible for the

analysis. Epidemiological data of all mother–infant pairs and neonatal

SeCr values were recorded.

Result: A total of 652 infants were studied. Multivariate regression

analysis showed that the main independent factors associated with high

SeCr at day 1 were hypertensive disease of pregnancy (P<0.0001) and

advancing hour of life (P<0.0001), with minimal contribution of

placental abruption (P<0.05) and higher GA (P<0.05). Lower GA

(P<0.0001) and ibuprofen-treated patent ductus arteriosus (PDA;

P<0.0001) were the main analyzed factors independently associated with

higher SeCr peak (defined as the highest SeCr during the week), with less

contribution of respiratory distress syndrome (P<0.01) and early onset

infection (P<0.05). In infants with hemodynamically significant PDA

(hsPDA) SeCr before ibuprofen administration was higher when

compared to GA-matched controls without hsPDA (P< 0.0001).

Conclusion: SeCr peak was inversely correlated to GA in preterm infants born

less than 32 weeks of GA. Neonatal rather than maternal morbidity affected

SeCr peak. In hsPDA, SeCr increase preceded ibuprofen administration.



Keywords: preterm infant; renal function; serum creatinine; patentductus arteriosus

Introduction

Serum creatinine (SeCr) concentration is usually the sole availablemarker of glomerular filtration rate (GFR) in clinical practice,especially in neonatal intensive care units (NICUs).

At birth, the newborn SeCr reflects maternal concentrations andthis is because maternal creatinine equilibrates with fetalconcentrations across the placenta.1–3 However, preterm birth isoften associated with gestational diseases affecting placentalfunction and/or maternal renal function. Furthermore, somegestational diseases may also affect fetal glomerular developmentby inducing intrauterine growth restriction.4 To our knowledge, noprevious study has specifically explored the changes in SeCr atbirth according to gestational diseases. In preterm infants born lessthan 32 weeks of gestational age (GA), previous reports showed thatSeCr concentration had a postnatal increase with a peak betweenthe second and the fourth day of life that was followed by a SeCrdecrease.5 Both the increased peak and the slow decrease of SeCr insmaller premies reflect decreased renal function associated with lowGA.6 Postnatal SeCr changes in preterm infants may result fromtubular creatinine reabsorption in the immature kidney7 and froma physiologically low GFR during the first days of life.8 Moreover,in sick preterm infants, many clinical conditions inducinghypotension, hypovolemia and hypoxemia and the use ofnephrotoxic drugs may lead to further reduction of GFR.9 Inparticular, nonsteroidal anti-inflammatory drugs (NSAID) that aregiven for treatment of patent ductus arteriosus (PDA) have beenreported to impair renal perfusion10 and the immature kidney isregarded as especially sensitive to NSAID renal effects. However,hemodynamically significant PDA (hsPDA) itself can be associatedwith poor renal perfusion and we could speculate that the NSAIDrenal effects may be exacerbated by a low GFR preceding NSAIDadministration.

We used a regional perinatal database, to obtain additionalinformation about maternal and neonatal factors influencing SeCrat birth and postnatal SeCr increase in a large cohort or preterminfants born less than 32 weeks of GA.

Received 13 July 2008; revised 30 September 2008; accepted 7 October 2008; published online

11 December 2008

Correspondence: Dr S Iacobelli, Neonatal Intensive Care Unit, Department of Paediatrics,

University of Dijon, 10, Bd Mal de Lattre de Tassigny, BP 77908, Dijon Cedex, Bourgogne

21079, France.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.203



MethodsStudy populationAll the infants born less than 32 weeks of GA from January 2001 toDecember 2005 and hospitalized in our NICU were eligible for thisstudy.

Data collectionData on the population were obtained from our regional perinataldatabase11,12 with prospective recording of mother–infant clinicalinformation, which was set up with the approval of the NationalCommittee of Informatics and Liberty. A total of 661 medical records ofpreterm infants born less than 32 weeks of GA were identified.Information was collected about antenatal history: maternal age,hypertensive disease of pregnancy (HDP), placental abruption, placentaprevia, diabetes, preterm prelabor rupture of membranes (PPROM),preterm labor, antenatal steroids administration, singleton or multiplepregnancy, abnormal fetal heart rate patterns and postnatal life: modeof delivery, 1 min Apgar score, GA, birth weight (BW), gender,appropriateness of BW for GA, early onset sepsis (EOS), need formechanical ventilation, respiratory distress syndrome (RDS) requiringsurfactant administration, hsPDA requiring medical and/or surgicaltreatment. The GA in completed weeks was assessed on the basis of themother’s last menstrual period and confirmed or modified whennecessary by routine early antenatal ultrasound examination. Theoccurrence of hsPDA was defined according to echocardiographiccriteria (left atrial to aortic root ratio, pulsed Doppler signal within theduct and ductal diameter).13,14 Owing to the NICU common standards,all the infants in the study had an echocardiography for PDA screeningbetween days 2 and 4 of life. In case of clinically symptomatic PDA,echocardiography was performed before day 2 and similarly repeatedafter day 4 in case of clinical symptoms.

According to previous recommendations for water andelectrolytes prescription in very low birth weight (VLBW) infants,15

babies born less than 32 weeks of GA had daily measurements ofserum electrolytes concentrations in our NICU, over the first days oflife. SeCr was concurrently measured to allow a prompt diagnosisof clinically significant reduction in GFR and also to anticipateconsequences of renal failure, as previously described.16

Creatinine values available between days 1 and 7 were collectedretrospectively for each patient from the hospital laboratoryelectronic database. Day 1 was considered to be the first morningafter birth. Baseline SeCr was defined as SeCr at day 1 and SeCrpeak as the highest SeCr value during the first week. Creatininewas measured by multi-enzymatic assay(amidohydrolaseþ oxidase and peroxidase) with Ortho ClinicalDiagnostics analyzer (Rochester, NY, US).

Statistical analysisThe normality of the distributions of SeCr was assessed withKolmogorov–Smirnoff and Shapiro–Wilks tests. Univariateanalyses were performed to correlate SeCr values and perinatal

variables using one-factor analysis of variance. Variables significantat a P-level <0.20 at the univariate analysis were entered into abackward selection analysis of variance. The r2 and theincremental r2 were calculated. Statistical analyses were performedusing SAS 8.2 (SAS Institute Inc.). All hypotheses were tested at thetwo-tailed 0.05 significance level. To further evaluate theindependent effect of PDA on SeCr before ibuprofen treatment, ourstudy population was divided into two groups: newborns whoreceived ibuprofen treatment (exposed group, cases) and newbornswho did not receive ibuprofen (non-exposed group, controls).These two groups were matched for GA and sex.

ResultsStudy populationOf the 661 infants eligible, 5 were excluded because they had majorcongenital malformations and 4 because of their mother’s terminal orsevere acute renal failure. So, the study population included 652 infants.

Clinical dataTable 1 shows antenatal and postnatal characteristics of the studypopulation.

Creatinine values and trendsSerum creatinine values were available for 89, 87, 80, 71 and 66%of infants progressing from days 1 to 5 and 53% for days 6 to7 of life (results in Table 2). Figure 1 shows creatinine profileduring the first week according to GA at birth. For infantsp28 weeks GA, SeCr was found to decrease after reaching apeak at day 5. This peak was at day 2 for infants born between29 and 31 weeks of GA.

Correlation with perinatal variablesFactors associated with high baseline SeCr at the univariate analysiswere: cesarean section, placental abruption, abnormal fetal heart ratepatterns, HDP, antenatal steroids, higher GA, SGA and advancinghour of life. Preterm labor and PPROM were associated with lowbaseline SeCr (data not shown). Independent variables associatedwith baseline SeCr at the multivariate analysis were: HDP, advancinghour of life, placental abruption and GA (Table 3). All these variableswere positively related to baseline SeCr.

At univariate analysis, there was no statistically significant effect ofthe explored antenatal factors on SeCr peak (data not shown); GA atbirth, ibuprofen-treated PDA, RDS requiring surfactant administrationand EOS were postnatal factors associated with SeCr peak. They wereall identified as independent significant variables at the multivariateanalysis (Tables 2 and 4). GA showed a negative association, whereasPDA, RDS and EOS were positively related to SeCr peak.

Within the whole population, children with hsPDA requiringtreatment showed a different SeCr profile during the first week, asthey presented a SeCr peak that was higher and occurred laterwhen compared to children without hsPDA. However, GA was

Creatinine changes in preterm infantsS Iacobelli et al

233


significantly lower in the hsPDA group (27.9±1.9 weeks) whencompared to infants without hsPDA (29.2±1.7 weeks; P<0.001).So, a case–control analysis was performed to eliminate the GAconfounding effect.

Case–control analysisOf 162 infants with hsPDA, 155 had been treated by ibuprofen, 3 byindomethacin and 4 by surgical ligature (because ofcontraindication of medical treatment). In the subgroup analysisof 155 ibuprofen-exposed, 11 infants treated at day 1 and 21infants without a GA-matched control were excluded. So, theremaining 123 ibuprofen-exposed infants were compared to 246 GAand sex-matched controls: SeCr value the day before ibuprofenadministration was significantly higher in cases when compared tocontrols of similar postnatal age (P¼ <0.0001; Table 5); casesand controls had the same characteristics (described in Table 1)

except for incidence of RDS requiring surfactant that was higher ininfants with hsPDA: 83.7 versus 65% (P<0.005, data not shown).

Discussion

Several studies have explored creatinine changes in preterm infantsafter birth.5,17–19 These studies were limited to population sampleseither small or not having frequent creatinine levels estimatedduring the first week. Among them, only one specificallyinvestigated the role of perinatal factors in influencing SeCrvalues.17

It has long been known that newborn SeCr reflects maternalSeCr at birth.20 So, maternal creatinine concentration is expectedto contribute to a great portion of the first day neonatal SeCr. Afterbirth, SeCr increases in the first 48 h in preterm infants.7,8

Consistent with these data, our results showed that the main factorsanalyzed found to be related to baseline SeCr were HDP, acondition that is often associated with high maternal levels of SeCr,and advancing hour of life. Placental abruption and GA explainedonly a very small proportion of the variability of the first dayneonatal SeCr in our study. Moreover, a positive correlationbetween SeCr at admission in NICU and GA was not observed inprevious investigations.5,6,8,17,21 Therefore, the finding of a positivecorrelation between GA and baseline SeCr, though statisticallysignificant, should be interpreted with caution.

The SeCr course was strongly influenced by GA in our cohortduring the first day of life. SeCr increased after birth of the preterminfant, reaching a peak that was at the second day for infants bornat 29 to 31 weeks of GA and at the fifth day for those born at p28weeks of GA. SeCr peak was found to be higher for lower GA. Ourresults are consistent with others: Bueva and Guignard18 showed anegative correlation between plasma creatinine and GA in onestudy comparing preterm infants versus term infants between thefirst and the second day of life; Gallini et al.5 found that during thefirst week of life, the maximum value of SeCr was significantlyhigher for GA lower than 27 weeks than for 27 to 28, 29 to 30 and31 to 32 weeks GA; recently Auron and Mhanna6 showed that inVLBW infants younger than 29 weeks or smaller than 1000 g therewas a delay in the decrease of their SeCr that extended beyond thefirst day of life. So, gestation and age-based reference charts havebeen proposed for the interpretation of creatinine values inextremely premature babies <28 weeks GA.21

Our study also disclosed EOS, RDS and ibuprofen-treated hsPDAas other independent factors associated with high postnatal SeCrincrease. One recent study from Cuzzolin et al.17 in newbornsranging from 22 to 36 weeks GA underlined that in neonates withimpaired renal function, a significant increase of SeCr occurredfrom the third day of life, this being related to many risk factors,RDS and ibuprofen treatment being among them. These findingsare consistent with the current knowledge that various clinicalconditions occurring in NICU may impair neonatal renal function:

Table 1 Characteristics of 652 infants <32 weeks of GA admitted in a regionaltertiary care center

Characteristics

Characteristics at birth

Male gender (%) 55.1

BW (g) mean±s.d. 1211±335

GA (weeks) mean±s.d. 28.9±1.8

SGA (%) 16.6

Apgar score <3 at 1 min of life (%) 22.2

Prenatal characteristics

Singletons births (%) 68.4

Maternal diabetes (%) 6.7

Preterm labor (%) 42.1

HDP (%) 23.1

PPROM (%) 38.1

Placental abruption (%) 12.2

Placenta previa (%) 6.2

Abnormal fetal heart rate patterns (%) 20.7

Antenatal steroids (%) 78.8

Cesarean section (%) 65.1

Postnatal diseases

RDS requiring surfactant (%) 61.9

Early onset sepsis (%) 2.9

PDA requiring treatment (%) 24.8

Oxygen dependency beyond 28 days (%) 31.1

Necrotizing enterocolitis (%) 8.1

Severe abnormal cerebral ultrasound (%)a 9.5

Death (%) 9.6

Abbreviations: BW, birth weight; GA, gestational age; HDP, hypertensive disease ofpregnancy; PDA, patent ductus arteriosus; PPROM, preterm prelabor rupture ofmembranes; RDS, respiratory distress syndrome; SGA, small for gestational age.aIntraventricular hemorrhage grade 3 or 4 and/or periventricular leukomalacia.


234


RDS, mechanical ventilation, systemic hemodynamic compromise,acute anemia, dehydration, hypoxemia, acidosis and nephrotoxicdrugs including ibuprofen.22 We have also found that in infantswith hsPDA, SeCr increase preceded ibuprofen administration, thissuggesting that a renal impairment because of PDA and/orassociated conditions exists before starting NSAID therapy.Concerning this point, many studies demonstrated thatprophylactic or therapeutic ibuprofen, as with other cyclooxygenaseinhibitors, may not be exempt from causing renal adverseeffects,23–25 but few investigations have explored the effect on renal

function of PDA before and independently from ibuprofentreatment. Vanpee et al.26 showed that sick VLBW infants having aPDA and requiring mechanical ventilation had lower creatinineclearances and significantly higher fractional sodium excretionthan controls; however, their study also included few infants whohad received NSAID for ductus closure before evaluating renalfunction. Shimada et al.27,28 studied the cardiocirculatory effects ofhsPDA in VLBW and extremely low BW preterm with RDS and theyfound a significant decrease in renal blood flow in infants withhsPDA compared to controls before pharmacological closure of

Table 2 SeCr values and peaks (mmol l�1) in 652 infants <32 weeks of GA during the first week of life

<27 weeks (n¼ 86) 27–28 weeks (n¼ 168) 29–30 weeks (n¼ 227) 31 weeks (n¼ 171) P-value

Day 1 (SeCr baseline) 73.0±17.0 (71; 31) 75.3±15.4 (73; 25) 76.9±15.4 (75; 20) 81.6±16.5 (80; 22) 0.0006a

Day 2 96.9±16.2 (94; 22) 96.4±18.1 (95; 20) 89.7±15.9 (88; 19) 91.7±17.4 (89; 21) 0.0006b

Day 3 97.3±20.4 (94; 26) 94.9±20.6 (92; 26) 86.0±19.2 (84; 20) 85.1±20.0 (82; 24) <0.0001b

Day 4 100.2±19.7 (98; 28) 98.3±23.8 (95; 27.5) 85.1±21.1 (83; 25) 82.4±21.2 (77; 25) <0.0001b

Day 5 105.0±21.9 (106; 33) 98.0±25.9 (93; 33) 83.4±20.9 (81; 26) 79.8±24.7 (74; 23) <0.0001b

Day 6 102.0±19.5 (98; 28) 96.0±27.3 (94; 37) 80.2±21.2 (79; 25) 76.9±24.1 (72; 26) <0.0001b

Day 7 99.6±22.0 (97; 30) 93.0±26.8 (87; 32) 76.0±21.3 (73; 21) 72.4±24.5 (68; 17) <0.0001b

SeCr peak 109.0±20.8 (107; 32) 109.2±23.7 (105; 31) 95.7±18.4 (94; 32) 96.3±21.6 (92; 31) <0.0001a

Abbreviation: SeCr, serum creatinine.Data are expressed as mean ±s.d. (median; quartile range).aMultifactor ANOVA.bOne-way ANOVA.

< 27 weeks 27-28 weeks 29-30 weeks 31 weeks

110

100

90

80

70

600 24 48 72 96 120 144 168

Hours of life

Cre

atin

ine

mic

rom

ol/l

Figure 1 Creatinine profile during the first week of life according to gestationalage (GA) in 652 infants <32 weeks of GA.

Table 3 Factors associated with baseline SeCr (mmol l�1) at multivariateanalysis in 652 infants <32 weeks of GA

P-value Incremental r2

HDP <0.0001 0.14

Hour of life <0.0001 0.11

Placental abruption 0.01 0.009

GA 0.01 0.006

Abbreviations: GA, gestational age; HDP, hypertensive disease of pregnancy.General r2 ¼ 0.26.

Table 4 Factors associated with SeCr Peak (mmol l�1) at multivariate analysisin 652 infants <32 weeks of GA

P-value Incremental r2

GA <0.0001 0.06

PDA <0.0001 0.06

RDS 0.006 0.01

EOS 0.01 0.008

Abbreviations: EOS, early onset sepsis; GA, gestational age; PDA patent ductus arteriosus;RDS, respiratory distress syndrome.General r2 ¼ 0.17.

Table 5 Case–control analysis

Cases Controls P-value

Day 2 96.5±16.5 (n¼ 44)a 93.5±16.5 (n¼ 88) 0.34

Day 3 99.2±22.3 (n¼ 39)b 86.5±18.3 (n¼ 78) 0.003

Day 4 101.8±22.9 (n¼ 40)c 87.3±23.2 (n¼ 80) 0.003

99.2±20.6 (n¼ 123) 89.5±19.5 (n¼ 246) <0.0001

Infants are divided in groups according to the day of treatment. SeCr values (mmol l�1)preceding ibuprofen administration are compared to SeCr values at the same day incontrols. Data expressed as mean±s.d.aTreated on day 2.bTreated on day 3.cTreated on day 4 or >4.


235


PDA. Abnormal renal blood flows reverted after treatment andseemed less severe when early pharmacological closure of PDA wascarried out. This is consistent with our finding that SeCr valuebefore ibuprofen administration is significantly higher in infantswith hsPDA compared to controls, this difference being strongerwhen treatment of PDA occurred later. This point needs to befurther investigated in future prospective studies.

There are many limitations in our study. First of all, maternalSeCr values were not available and this limits the interpretation ofthe multiple regression analysis. Moreover, the lack of thisinformation limits the interpretation of the detected correlationbetween baseline SeCr and HDP, a condition in which maternalSeCr may be elevated. Second, precise information about maternalexposure to drugs potentially affecting infant’s renal function wasnot accessible for review. Third, the collection of newbornscreatinine data was retrospective and we could not measurecreatinine clearance or other biological parameters of renalfunction to correlate them with GA and with all the perinatalfactors explored in this cohort. The lack of data about fluid intake,water balance and hypernatremic dehydration during the first weekof life also limits the interpretation of SeCr values. Finally, someclinical data such as neonatal hypoxia, hypotension, acute anemiaat birth and infants exposure to nephrotoxic drugs such asgentamicin were not assessed and they could represent factorsinfluencing changes in SeCr in our population.

There are limitations to physiopathological interpretation of ourdata: the higher rate of RDS in preterm infants with hsPDA is apossible, confounding factor in the analysis of the results of thepostnatal SeCr increase, as RDS represents a risk factor for elevatedSeCr itself. However, the relationship between postnatal SeCrincrease and hsPDA before any NSAID treatment remains valid, ashsPDA, together with GA, was the main factor analyzed in thisstudy that was found to be related to SeCr peak, with a lesscontribution of RDS and EOS.

Of course, future prospective studies are necessary to explore thedevelopment of renal function in very preterm infants and also todetermine whether multifactors events acting early in postnatal lifecould have long-term consequences on renal outcome in later life.

References

1 Gordjani N, Burghard R, Leititis JU, Brandis M. Serum creatinine and creatinine

clearance in healthy neonates and prematures during the first days of life. Eur J

Pediatr 1998; 148: 143–145.

2 Wilkins BH. Renal function in sick very low birth weight infants: 1. Glomerular

filtration rate. Arch Dis Child 1992; 67: 1140–1145.

3 Lao TT, Loong EP, Chin RK, Lam YM. Renal function in the newborn. Newborn

creatinine related to birth weight, maturity and maternal creatinine. Gynecol Obstet

Invest 1989; 28(2): 70–72.

4 Manalich R, Reyes L, Herrera M, Melendi C, Fundora I. Relationship between weight at

birth and the number and size of renal glomeruli in humans: a histomorphometric

study. Kidney Int 2000; 58(2): 770–773.

5 Gallini F, Maggio L, Romagnoli C, Marrocco G, Tortorolo G. Progression of renal

function in preterm neonates with gestational age p32 weeks. Pediatr Nephrol 2000;

15: 119–124.

6 Auron A, Mhanna MJ. Serum creatinine in very low birth weight infants during their

first days of life. J Perinatol 2006; 26: 755–760.

7 Guignard JP, Drukker A. Why do newborn infants have a high plasma creatinine?

Pediatrics 1999; 103(4): e49.

8 Miall LS, Henderson MJ, Turner AJ, Brownlee KG, Brocklebanck JT, Newell SJ et al.

Plasma creatinine rises dramatically in the first 48 h of life in preterm infants.

Pediatrics 1999; 104(6): 1–4.

9 Toth-Hayn P, Drukker A, Guignard JP. The stressed neonatal kidney: from

pathophysiology to clinical management of neonatal vasomotor nephropathy. Pediatr

Nephrol 2000; 14(3): 227–239.

10 Ohlsson A, Walia R, Shah S. Ibuprofen for the treatment of patent ductus arteriosus in preterm

and/or low birth weight infants. Cochrane Database Syst Rev 2008; 23(1): CD003481.

11 Cornet B, Gouyon JB, Binquet C, Sagot P, Ferdynus C, Metral P et al. Using discharge

abstracts as a tool to assess a regional perinatal network. Rev Epidemiol Sante

Publique 2001; 49(6): 583–593.

12 Quantin C, Allaert FA, Gouyon B, Cohen O. Proposal for the creation of a European

healthcare identifier. Stud Health Technol Inform 2005; 116: 949–954.

13 Iyer P, Evans N. Re-evaluation of the left atrial to aortic root ratio as a marker of

patent ductus arteriosus. Arch Dis Child 1994; 70: 112–117.

14 Kluckow M, Evan N. Early echocardiographic prediction of symptomatic patent ductus

arteriosus in preterm infants undergoing mechanical ventilation. J Pediatr 1995; 127:

774–779.

15 Lorenz JM. Assessing fluid and electrolyte status in the newborn. National Academy of

Clinical Biochemistry. Clin Chem 1997; 43(1): 205–210.

16 Choker G, Gouyon JB. Diagnosis of acute renal failure in very preterm infants. Biol

Neonate 2004; 86(3): 212–216.

17 Cuzzolin L, Fanos V, Pinna B, di Marzio M, Perin M, Tramontozzi P et al. Postnatal

renal function in preterm newborns: a role of diseases, drugs and therapeutic

interventions. Pediatr Nephrol 2006; 21: 931–938.

18 Bueva A, Guignard JP. Renal function in preterm neonates. Pediatr Res 1994; 36: 572–577.

19 Sonntag J, Prankel B, Waltz S. Serum creatinine concentration, urinary creatinine

excretion and creatinine clearance during the first 9 weeks in preterm infants with a

birth weight below 1500 g. Eur J Pediatr 1996; 155(9): 815–819.

20 Manzke H, Spreter von Kreudenstein P, Dorner K, Kruse K. Quantitative measurements of the

urinary excretion of creatinine, uric acid, hypoxanthine and xanthine, uracil, cyclic AMP, and

cyclic GMP in healthy newborn infants. Eur J Pediatr 1980; 133: 157–161.

21 Thayyil S, Sheik S, Kempley ST, Sinha A. A gestation- and postnatal age-based

reference chart for assessing renal function in extremely premature infants. J Perinatol

2008; 28: 226–229.

22 Gouyon JB, Guignard JP. Management of acute renal failure in newborns. Pediatr

Nephrol 2000; 14: 1037–1044.

23 Giniger RP, Buffat C, Millet V, Simeoni U. Renal effects of ibuprofen for the treatment

of patent ductus arteriosus in premature infants. J Matern Fetal Neonatal Med 2007;

20(4): 275–283.

24 Allegaert K, Vanhole C, de Hoon J, Guignard JP, Tibboel D, Devlieger H et al.

Nonselective cyclo-oxygenase inhibitors and glomerular filtration rate in preterm

neonates. Pediatr Nephrol 2005; 20(11): 1557–1561.

25 Allegaert K, Rayyan M, Anderson BJ. Impact of ibuprofen administration on renal drug

clearance in the first weeks of life. Methods Find Exp Clin Pharmacol 2006; 28(8): 519–522.

26 Vanpee M, Ergander U, Herin P, Aperia A. Renal function in sick, very low birth weight

infants. Acta Paediatr 1993; 82(10): 714–718.

27 Shimada S, Kasai T, Konishi M, Fujiwara T. Effects of patent ductus arteriosus on left

ventricular output and organ blood flows in preterm infants with respiratory distress

syndrome treated with surfactant. J Pediatr 1994; 125(2): 270–277.

28 Shimada S, Kasai T, Hoshi A, Murata A, Chida S. Cardiocirculatory effects of patent

ductus arteriosus in extremely low-birth-weight infants with respiratory distress

syndrome. Pediatr Int 2003; 45(3): 255–262.


236


ORIGINAL ARTICLE

Limitations of single-channel EEG on the forehead for neonatalseizure detectionCJ Wusthoff1, RA Shellhaas2 and RR Clancy1,3

1Division of Neurology, Children’s Hospital of Philadelphia, Philadelphia, PA, USA; 2Division of Pediatric Neurology, Department ofPediatrics and Communicable Diseases, Mott Children’s Hospital, University of Michigan Medical School, Ann Arbor, MI, USA and3Departments of Pediatrics and Neurology, University of Pennsylvania School of Medicine, Philadelphia, PA, USA

Objectives: In amplitude-integrated EEG, lead placement across the

forehead is convenient, but this location has unknown effects on neonatal

seizure (NS) detection sensitivity. This study describes the limits of NS

detection by a single forehead EEG channel.

Study Design: Records were taken from a digital library of conventional

EEGs (CEEGs) with NS, previously characterized at a bicentral channel,

C3-C4. We analyzed electrographic characteristics in a single forehead

channel, Fp3-Fp4.

Result: A total of 330 seizures from 125 CEEGs were included. With

Fp3-Fp4, at least one NS was detected in 66% of records vs 90% using

C3-C4 (P<0.0001). Of 330 seizures, 46% appeared in Fp3-Fp4 vs 73%

in C3-C4 (P<0.0001). Seizures appeared briefer in Fp3-Fp4 than

C3-C4 (P<0.006) and CEEG (P<0.0001).

Conclusion: NSs are significantly more difficult to detect with a single

forehead channel than bicentrally or on CEEG. In Fp3-Fp4, a third of

records with seizures were missed and over half of seizures were undetected.



Keywords: amplitude-integrated EEG; cerebral function monitor;neonate; seizure

Introduction

Despite advances in obstetrics and neonatology, neonatal seizures(NSs) remain a significant problem in the newborn period. Theincidence of clinically diagnosed NSs ranges from 1.5 to 3.5 per1000 live term born babies.1–3 Because most electrographicseizures in the neonate are subclinical, these figures mayunderestimate their true incidence.4–7 Clinical NSs have beencorrelated with significant neurological morbidity and mortality.7,8

Because of the concern for missing difficult to recognize or frankly

subclinical electrographic NSs, diagnostic monitoring tools areincreasingly used in intensive care units for high-risk neonates.

The gold standard for NS diagnosis and quantification isconventional electroencephalogram (CEEG), which uses a fullarray of electrodes placed according to the International 10–20system modified for neonates (Figure 1). These examinations areperformed by trained technologists and interpreted byelectrophysiologists expert in neonatal EEG. The logistic limitationsof CEEG often include their brief duration (usually under 60 min)and limited access to appropriately trained electrophysiologists.They are not routinely available around the clock at mosthospitals. The need for immediately available, continuousmonitoring that is interpretable in real time at the bedside hasinspired the development of amplitude-integrated EEG (aEEG).

aEEG uses one or two highly processed and time-compressedEEG channels displayed at the bedside for real-time interpretationby care providers to assess long-term trends in the EEG backgroundand to detect some electrographic seizures. Although this methodhas the advantage of bedside utility, there are known limits to itssensitivity in seizure detection.9–11

For single-channel aEEG, most advocate lead placement at thebiparietal areas (P3-P4 channel), which overlies a vascularwatershed.12 The closest corresponding electrode pair in neonatalCEEG using the 10–20 system for neonates is the adjacent bicentralarea, represented by the electrode pair C3-C4 (Figure 1). However, inpractice, some clinicians prefer lead placements on the foreheadbecause of the convenience of avoiding scalp hair and compatibilitywith selective head-cooling devices. The closest correspondence toforehead electrodes in CEEG is represented by the frontopolar electrodepair, Fp3-Fp4 (Figure 1). It is unknown whether this alternative leadplacement affects the sensitivity of NS detection. This study sought tocompare the sensitivity for NS detection of a single EEG channel atFp3-Fp4 with a single channel at C3-C4 and with CEEG.

Methods

The Children’s Hospital of Philadelphia’s Institutional ReviewBoard approved this study. We reviewed a sample of NSs from a

Received 16 April 2008; revised 23 June 2008; accepted 14 July 2008; published online

4 December 2008

Correspondence: Dr CJ Wusthoff, Division of Neurology, Children’s Hospital of Philadelphia,

6th Floor Wood Building, 34th Street and Civic Center Boulevard, Philadelphia, PA 19104,

USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.195



digital library of neonatal CEEGs with electrographic seizures thatwere recorded as part of prior research protocols or for clinicalpurposes. These records were previously stripped of patientidentifiers and converted to Persyst format (Persyst Corp.,Rochester, MN, USA). Each CEEG had been analyzed by twoelectroencephalographers (RS and RC) for backgroundcharacteristics, the duration of seizures and site of seizure onset, asdescribed in a previous study.10

An electrographic seizure was defined as a specific eventrecognized by a sudden, repetitive, evolving and stereotyped ictalpattern with a clear beginning, middle and ending; a minimumduration of 10 s; and amplitude of at least 2mV. To be counted asdistinct events, individual seizures had to be separated by at least 10 s.

In addition to the CEEG, two additional single channels of EEGwere created and simultaneously displayed. The channel C3-C4

was chosen to closely approximate the P3-P4 location fromwhich aEEG is usually generated (Figure 1). The Fp3-Fp4

channel was created to simulate the forehead location preferred bysome neonatologists. For each CEEG, up to the first three seizureswere detected by visual analysis, their durations measured and siteof origin determined. The maximal peak-to-peak amplitude ofeach seizure was measured electronically and the mean of theseamplitudes was calculated using Matlab (Mathworks Inc., Natwick,MA, USA). Each CEEG was then reviewed again to determinewhether those seizures also appeared in the single C3-C4 orFp3-Fp4 channels. The durations and maximal peak-to-peakamplitudes of seizures detected by the single channels weresimilarly measured and recorded. Finally, we compared thetechnical quality of single-channel EEGs recorded at C3-C4

and Fp3-Fp4.A power analysis yielded a sample size estimate of 125 CEEGs,

assumed to have three seizures per study. This was calculated toobtain a 95% confidence interval of ±5% around the truepercentage of seizures detectable with a single channel atFp3-Fp4. Continuous variables were compared using two-tailedstudent’s t-test and ANOVA statistics. Categorical variables werecompared using a Pearson’s w2-test. P-values of less than 0.05 wereconsidered statistically significant.

Results

A total of 125 CEEGs were recorded from 121 neonates whoseconceptional ages (legal age plus estimated gestational age) rangedfrom 34 to 50 weeks. Altogether there were 330 seizures detected.Electrographic characteristics of the seizures are described inTables 1 and 2. The mean length of the CEEG tracings analyzedwas 26±20.6 min (mean±s.d.). The duration of CEEG seizureswas 147±254 s. The mean of their maximal peak-to-peakamplitudes was 81.6±74.2 mV. The most common site of originwas central (C3 or C4), occurring in 38.2%. Seizures were leastlikely to begin in the frontopolar areas (Fp3 or Fp4), and accountedfor only 7.3% of the total.

A total of 73% (242/330) of the CEEG seizures were also detectedin the single C3-C4 channel. Mean seizure duration at C3-C4

was 111±241 s, not significantly different from seizure durationin CEEG (P¼ 0.061). The maximal ictal peak-to-peak amplitude

Table 1 Characteristics of EEG seizures recorded in CEEG compared to the single channels C3-C4 and Fp3-Fp4

CEEG C3-C4 Fp3-Fp4 P-values

CEEG vs C3-C4 CEEG vs Fp3-Fp4 C3-C4 vs Fp3-Fp4

Number (and %) of EEG records with at least one

visible seizure

125 (100%) 113 (90.4%) 82 (65.6%) P¼ 0.0003 P<0.0001 P<0.0001

Number of seizures detected 330 (100%) 242 (73%) 153 (46.4%) P<0.0001 P<0.0001 P<0.0001

Mean seizure duration (±s.d.) (s) 147±254 111±241 64±192 P¼ 0.061 P<0.0001 P<0.006

Mean maximal p–p ictal amplitude (±s.d.) (mV) 81.6±74.2 90.0±78.4 54.0±44.6 P¼ 0.14 P<0.0001 P<0.0001

Abbreviations: CEEG, conventional electroencephalogram; EEG, electroencephalogram.

Fp1 Fp2

F7

F3 Fz F4

F8

T3 C3 Cz C4 T4

T5

P3

Pz

P4

T6

O1 O2

A1 A2

Fp3 Fp4

Figure 1 The International 10–20 system of electroencephalogram (EEG) leadplacement. In neonates, only the shaded electrodes are placed on the smallerneonatal scalp for conventional EEG. Fp3-Fp4 was used to simulate a singleforehead channel as suggested by some for amplitude-integrated EEG (aEEG).C3-C4 was used to simulate the recommended biparietal lead placement. In atypical neonate with a head circumference of 36 cm, the distance between Fp3 andFp4 (anterior dotted line) is 7.2 cm, the distance between C3 and C4 (central dottedline) is 10.4 cm.

Single-channel forehead EEGCJ Wusthoff et al

238


of seizures detected at C3-C4 was 90±78.4 mV, not significantlydifferent from CEEG (P¼ 0.14).

Only 46% (153/330) of the CEEG seizures were detected with thesingle Fp3-Fp4 channel. This was significantly less sensitive thanseizure detection at C3-C4 (P<0.0001). The mean duration ofseizures recorded at Fp3- Fp4 was 64±192 s, significantly brieferthan seizures detected at C3-C4 (P<0.006) or CEEG (P<0.0001).The mean maximal ictal p–p amplitude of seizures appearing atFp3-Fp4 was 54±44.6mV, also significantly lower than C3-C4

(P<0.0001) or CEEG (P<0.0001) (Table 1). When reading theFp3-Fp4 channel alone, at least one seizure was detected in 66%(82/125) of the records (Table 1, Figure 2). This was significantlyless sensitive than the single C3-C4 channel, which detected atleast one seizure in 90% (113/125) of the CEEG records(P<0.0001).

The single channels also differed in their ability to detectseizures originating from different locations (Table 2). In CEEG,seizures arose most commonly centrally (38.2%), followed bytemporal (25.5%), occipital (17.3%), midline (11.8%) andfrontopolar (7.3%). The single Fp3-Fp4 channel detectedsignificantly fewer seizures than C3-C4 in those originating inthe central (C3 or C4) or temporal (T3 or T4) regions.

An unexpectedly high number of records contained artifact inthe single Fp3-Fp4 channel, which obscured seizure detection(Figure 3). Fourteen of the 125 studies (11%) were at least partiallyobscured by recording artifact at Fp3, Fp4 or both. In each case, theartifact rendered the record uninterpretable at Fp3-Fp4 alone forat least one seizure. These records commonly had no prominentartifacts at other leads.

Discussion

The single electrode pair P3-P4 is considered the ideal theoreticalsite for aEEG lead placement. However, many practitioners are

tempted to use Fp3-Fp4 due to its practical advantages. Leads atthe frontal locations avoid neonates’ hair and avert the need forscalp shaving or colloid adhesives. Likewise, the forehead locationis highly visible to bedside personnel, who can easily visuallyconfirm lead security without manipulating the patient. Finally,the forehead location for aEEG lead placement is compatible withselective head-cooling devices, which are increasingly common inthe neonatal intensive care unit. However, recommendations forforehead lead placement might adversely affect the sensitivity of NSdetection.

This is the first study to quantitatively compare the effects oflead placement on the sensitivity of single-channel EEG for NSdetection. Previously, we demonstrated that the ‘raw’ EEG recordedfrom a single pair of central electrodes (C3-C4) has relativelygood sensitivity for NS detection; however, sensitivity decreasedsignificantly when the raw EEG was processed to create aEEGtracings.9,10 The current study shows that many NSs are missedwhen a single forehead channel of raw EEG is recorded from a pairof frontopolar electrodes. These results have important implicationfor electrode placement during the use of aEEG in seizure detectionfor high-risk neonates.

Despite the practical advantages of forehead electrodes, our datasuggest that the Fp3-Fp4 EEG channel has significantlimitations. The sensitivity of detecting individual seizures droppedfrom the gold standard CEEG value of 100 to 73% at C3-C4 and46% at Fp3-Fp4. Similarly, 33% of the records had no seizuresidentified when using only the forehead channel, compared withonly 10% when using C3-C4 alone. This probably represents theminimum value for NS detection sensitivity, because it is possiblethat longer recording times, as occurs in clinical practice, couldresult in the detection of at least one seizure. Furthermore, theseizures in the Fp3-Fp4 channel had significantly loweramplitudes and briefer durations than C3-C4. It is reasonablyanticipated that further signal processing of raw EEG from theFp3-Fp4 channel to create aEEG tracings would similarly furtherreduce NS detection sensitivity.

There are several possible explanations for these differences.Cerebral electrical activity is generally poorly developed in the farfrontal regions of the neonatal brain. Indeed, to facilitate EEGrecordings, the International 10–20 system designates thatneonatal frontopolar electrodes should be placed midway betweenthe conventional Fp1/Fp2 and F3/F4 locations. These unique hybridlocations are designated ‘Fp3’ and ‘Fp4’. In CEEGs, NSs are leastlikely to originate in Fp3-Fp4. Nevertheless, they may first arise inother regions and then migrate there.13,14 Because most NSs do notsecondarily generalize, a single EEG channel has only a limitedchance to detect them. Specifically, Fp3-Fp4 was less likely todetect CEEG seizures arising in the central (C3 or C4) or temporal(T3 or T4) regions than the single C3-C4 channel. This maypartially explain the reduced sensitivity of a single channel of EEGin this and other studies.11

Table 2 Seizure detection sensitivity using single EEG channels alone, bylocation of seizure onset in CEEG

Seizure origin in CEEG CEEG C3-C4 Fp3-Fp4 P-valuesa

N¼ 330 seizures N (%) N (%) N (%)

Frontopolar (Fp3 or Fp4) 24 (7.3) 13 (54.2) 15 (62.5)b 0.558

Central (C3 or C4) 126 (38.2) 115 (91.3)b 53 (42.1) <0.0001

Midline (Cz) 39 (11.8) 28 (71.8) 21 (53.8) 0.101

Temporal (T3 or T4) 84 (25.5) 57 (67.9) 43 (51.2) 0.0278

Occipital (O1 or O2) 57 (17.3) 29 (50.9) 21 (36.8) 0.131

Total 330 (100) 242 (73.3) 153 (46.4) <0.0001

Abbreviations: CEEG, conventional electroencephalogram; EEG, electroencephalogram.aCompares percent detection between C3-C4 and Fp3-Fp4.bSignal cancellation or technical artifact affected the detection of some electrographicseizures using a single channel, even when the channel was the site of seizure origin.


239


Another contributing factor was the unexpectedly higherpercentage of artifact in the frontal leads, which occurred in11% of the studies. Artifact in frontal electrodes mayarise from electromyographic activity of the frontalis muscles,

drying of electrode gel by radiant warmers, or physicaldisruption during patient care. The presence of obscuringartifact did contribute to the difficulty in visualizing some seizuresat Fp3–Fp4.

T3-O1

Fp3-C3

C3-O1

Fp4-C4

C4-O2

T3-C3

C3-CZ

CZ-C4

C4-T4

Fp3-Fp4

C3-C4

Fp3-T3

Fp4-T4

T4-O2

Fp3-T3

T3-O1

Fp4-T4

T4-O2

Fp3-C3

C3-O1

Fp4-C4

C4-O2

T3-C3

C3-CZ

CZ-C4

C4-T4

Fp3-Fp4

C3-C4

Figure 2 (a) Example of a conventional electroencephalogram (CEEG) seizure also seen in both single channels (C3-C4 and Fp3-Fp4). (b) Example of a CEEGseizure visible in the C3-C4 channel, but not in the single Fp3-Fp4 channel.


240


Finally, it is possible that the sensitivity of the Fp3-Fp4

channel was reduced by the shorter interlead distance at thislocation. EEG displays the difference in voltage between twoelectrode locations. If the electrodes are extremely close together,the measured voltage difference will be small as both locationsrecord very similar electrical signals. In contrast, voltage differenceswill be greater when the recording electrodes are more widelyseparated. For example, if electrodes are properly placed accordingto the International 10–20 system in a typical neonate with a headcircumference of 36 cm, C3 and C4 would be separated by adistance of 10.4 cm, whereas Fp3 and Fp4 would be only7.2 cm apart (Figure 1). This possibility is further supported by ourfinding that seizures detected at Fp3-Fp4 alone were alsosignificantly lower in amplitude than those detected by CEEG orat C3-C4.

Similarly, the durations of NSs detected at the forehead weresignificantly briefer than those detected by CEEG or at C3-C4.From a clinical or research perspective, these findings indicate thatthe use of a single forehead EEG channel would result in missingmany electrographic seizures, or underestimating their trueduration if they were detected. Further processing of the raw EEGsignal from Fp3-Fp4 to create an aEEG tracing would be expectedto even further reduce the sensitivity of seizure detection. This isparticularly important if aEEG is used for NS detection andmeasuring the efficacy of antiepileptic drugs. If this single foreheadchannel alone were used to generate aEEGs to judge response to

treatment, the practitioner might be misled to perceive a lowerseizure burden than actually exists.

There were strengths and limitations to this study. Among thestrengths were the use of full CEEG as the basis of the study,providing a ‘gold standard’ for seizure detection and spatiallocalization. CEEGs were interpreted by experienced neonatalencephalographers. These records were obtained in the course ofclinical recordings of high-risk neonates, and thus would beexpected to faithfully reflect those obtained in similar clinicalsettings. Finally, although 26% of the CEEGs had fewer than thethree seizures each anticipated in our sample size calculation, wehad a sufficiently large sample to provide adequate power toobserve significant differences between groups. Likewise, this studyincluded relatively brief CEEG tracings, which might have lowersensitivity compared with monitoring of longer duration indetermining if a given patient has seizures. Nonetheless, theconclusion that lead placement affects sensitivity would remaintrue in both brief EEG recordings as well as in prolongedmonitoring.

One limitation of this study is that the single channel atFp3-Fp4 used here was placed by trained technologists whostrictly adhere to the International 10–20 system. It is possible thatforehead electrodes placed at the bedside in clinical practice varyfrom these locations. Another limitation was that our study onlyconsidered sensitivity of Fp3-Fp4 in seizure detection, notspecificity. Also, the records reviewed were selected for known

Fp3-T

3

T3-O

1

Fp4-T

4

T4-O

2

Fp3-C

3

C3-O1

Fp4-C4

C4-O2

T3-C3

C3-CZ

CZ-C4

C4-T4

Fp3-Fp4

C3-C4

Figure 3 Example of a record obscured by artifact in the Fp3 lead and consequently the single forehead channel, Fp3-Fp4; the conventional electroencephalogram(CEEG) seizure is easily visible in the other channels.


241


presence of seizures. Furthermore, the single-channel EEGs wereanalyzed simultaneously with the CEEG. This may bias readers tomore easily detect seizures in the Fp3-Fp4 or C3-C4 channelsthan if the single-channel tracings were read entirely in isolationof CEEG. Finally, the EEGs in this study did not include tracingsfrom neonates younger than 34 weeks gestational age, which maylimit the generalizability of our results to the preterm infant.

Our study sought only to examine the effect of single-channellead location on NS detection. We did not examine the effects oflead placement on other features of EEG, such as backgroundcharacteristics. Future work is needed to clarify if lead placementaffects sensitivity or specificity of EEG as used for purposes otherthan NS detection.

NSs detected by a single EEG channel at the forehead appearwith significantly lower frequency, duration and amplitude thanthose simultaneously recorded by a single C3-C4 channel orCEEG. Lead placement significantly affects the sensitivity of seizuredetection in single-channel raw EEG. The forehead location mightbe adequate for routine monitoring of low risk neonates but is notlikely be suitable for most high-risk neonates or those withconfirmed seizures due to its poor sensitivity for seizure detection.We urge neonatologists who use aEEG to avoid forehead leadplacement for NS detection.

Acknowledgments

We acknowledge Damon Lees (Moberg Medical, Ambler, PA) for technical support

and the generous assistance of the EEG technologists at the Children’s Hospital of

Philadelphia. The authors have no conflict of interest to disclose.

References

1 Eriksson M, Zetterstrom R. Neonatal convulsions. Incidence and causes in the

Stockholm area. Acta Paediatr Scand 1979; 68(6): 807–811.

2 Lanska MJ, Lanska DJ, Baumann RJ, Kryscio RJ. A population-based study of neonatal

seizures in Fayette County, Kentucky. Neurology 1995; 45(4): 724–732.

3 Ronen GM, Penney S, Andrews W. The epidemiology of clinical neonatal seizures in

Newfoundland: a population-based study. J Pediatr 1999; 134(1): 71–75.

4 Weiner SP, Painter MJ, Geva D, Guthrie RD, Scher MS. Neonatal seizures:

electroclinical dissociation. Pediatr Neurol 1991; 7(5): 363–368.

5 Clancy RR, Sharif U, Ichord R, Spray TL, Nicolson S, Tabbutt S et al. Electrographic

neonatal seizures after infant heart surgery. Epilepsia 2005; 46(1): 84–90.

6 Clancy RR, Legido A, Lewis D. Occult neonatal seizures. Epilepsia 1988; 29(3):

256–261.

7 Scher MS, Painter MJ, Bergman I, Barmada MA, Brunberg J. EEG diagnoses of

neonatal seizures: clinical correlations and outcome. Pediatr Neurol 1989; 5(1):

17–24.

8 McBride MC, Laroia N, Guillet R. Electrographic seizures in neonates correlate with

poor neurodevelopmental outcome. Neurology 2000; 55(4): 506–513.

9 Shellhaas RA, Soaita AI, Clancy RR. Sensitivity of amplitude-integrated electro-

encephalography for neonatal seizure detection. Pediatrics 2007; 120(4): 770–777.

10 Shellhaas RA, Clancy RR. Characterization of neonatal seizures by conventional EEG

and single-channel EEG. Clin Neurophysiol 2007; 118(10): 2156–2161.

11 Rennie JM, Chorley G, Boylan GB, Pressler R, Nguyen Y, Hooper R. Non-expert use of

the cerebral function monitor for neonatal seizure detection. Arch Dis Child Fetal

Neonatal Ed 2004; 89(1): F37–F40.

12 Hellstrom-Westas L, Vries LS, Rosen I. An Atlas of Amplitude-Integrated EEGs in the

Newborn. Parthenon: New York, 2003.

13 Acharya JN, Wyllie E, Luders HO, Kotagal P, Lancman M, Coelho M. Seizure

symptomatology in infants with localization-related epilepsy. Neurology 1997; 48(1):

189–196.

14 Tekgul H, Bourgeois BF, Gauvreau K, Bergin AM. Electroencephalography in neonatal

seizures: comparison of a reduced and a full 10/20 montage. Pediatr Neurol 2005;

32(3): 155–161.


242


ORIGINAL ARTICLE

Birth asphyxia survivors in a developing countryDR Halloran1, E McClure2, H Chakraborty2, E Chomba3, LL Wright4 and WA Carlo5

1Department of Pediatrics, Saint Louis University, Saint Louis, MI, USA; 2Department of Statistics and Epidemiology, RTIInternational, RTP, NC, USA; 3University Teaching Hospital, Lusaka, Zambia; 4Center for Research for Mothers and Children,National Institute of Child Health and Human Development, Bethesda, MD, USA and 5Department of Pediatrics, University ofAlabama at Birmingham, Birmingham, AL, USA

Objective: Determine the baseline incidence of birth asphyxia in neonatal

intensive care unit (NICU) survivors in a developing country and the early

neurodevelopmental outcomes of such infants.

Study Design: This cross-sectional, prospective study collected

diagnostic and examination findings on all infants seen in the University

of Zambia NICU follow-up clinic over a 4-week period.

Result: Of the 182 infants, 42 (23%) had a clinical diagnosis of birth

asphyxia. Of 42 infants with birth asphyxia, 13 (31%) had an abnormal

neurologic examination during the clinic visit; in contrast, 13 of 141

infants without birth asphyxia (9%) had an abnormal examination (odds

ratio 4.4, 95% confidence interval: 1.8, 10.4).

Conclusion: Birth asphyxia survivors account for almost a quarter of

NICU survivors in a developing country and half of those with an

abnormal neurologic examination. Studies are necessary to determine the

percent of birth asphyxia survivors who have permanent motor and

cognitive disabilities.


published online 27 November 2008

Keywords: asphyxia neonatorum; child development; neurologicmanifestations; developing countries; Zambia; infant

Introduction

Globally, 23% of neonatal deaths1 and 10% of all deaths inchildren less than 5 years of age2,3 are estimated to occur as aresult of birth asphyxia, defined as failure to initiate and sustainnormal breathing at birth,4 and account for one million deathseach year worldwide. A multicenter study involving hospitals ineight African countries including Zambia found birth asphyxia(defined as an Apgar score <7 at 5 min) to be the leading (23%)cause of neonatal mortality.4 In spite of the large impact that birth

asphyxia had on neonatal mortality, this study found that 87% ofinfants diagnosed with birth asphyxia survived to discharge home.

Survivors of birth asphyxia are at risk for neurodevelopmentalsequelae including motor and cognitive disabilities.5–9 One of thebest population studies of asphyxia in a developing countryreported that 18% of survivors of mild to moderate birth asphyxiahad neonatal encephalopathy and permanent severe neurologicimpairment.10 Birth asphyxia was a common cause of mentalretardation, cerebral palsy and other neurodevelopmental disorders.

Evaluation for long-term sequelae can best be performed whena child reaches an age at which established screening tools can beutilized, generally at least 18 months;11–13 however, many babiesmay have evidence of disabilities in early infancy. Therefore,neurodevelopmental examinations as early as 6 weeks of age mayprovide an indication of the magnitude of the problem.10

Although accurate estimates of neurodevelopmental sequelaefrom birth asphyxia in developing countries are not available,studies in developed countries have found that 15 to 18% of infantswho suffer moderate to severe asphyxia are disabled by 8 years ofage.7 Moderate to severe asphyxia was defined as alteredconsciousness with altered muscle tone or primitive reflexes after1 h of age along with intrapartum fetal distress, immediateneonatal distress, or immediate neonatal resuscitation. Givencurrent sociocultural factors in developing countries such asZambia, the burden of care is likely to be much greater than indeveloped countries.14 Poverty, illiteracy, low status of women, poorhygiene, lack of clean water and sanitation, family inability torecognize danger signs, and inadequate access to medical care allincrease the risk for morbidity following birth asphyxia indeveloping countries. Often rehabilitation resources are notavailable in these settings.

Birth asphyxia is likely a leading cause of neonatal death inmost developing countries such as Zambia;15,16 however,prospective studies to assess the early morbidity following birthasphyxia have not been performed in Sub-Saharan Africa.Although prevention and prompt treatment of birth asphyxia andreduction of its sequelae are among the developing world’s tophealth care priorities,17–19 studies are needed to determine the

Received 3 March 2008; revised 17 August 2008; accepted 15 September 2008; published online

27 November 2008

Correspondence: Dr DR Halloran, Department of Pediatrics, Saint Louis University, 1465

South Grand Blvd., St. Louis, MO 63104, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.192



frequency of such sequelae and the resulting burden on the healthsystem and society. This study was designed to determine thebaseline incidence of birth asphyxia in neonatal intensive care unit(NICU) survivors and the early neurodevelopmental outcomes ofthese infants.

MethodsSettingThe study was conducted in Lusaka, the capital of Zambia, adeveloping country in Sub-Saharan Africa. Health indicators inZambia are poor with actual per capita public expenditure on healthat $11 per year and total expenditure on health at 6.6% of the grossdomestic product.20 Life expectancy is 39 years, due in large part toeconomic deterioration and the HIV/AIDS epidemic.21,22 Infantmortality is estimated to be 95 per 1000 live births, and the under age5 years mortality rate is 182 per 1000.21,23

The public health system serves over 95% of neonates in itscatchment area and is divided into 25 primary health care centers,including 11 delivery centers and one referral hospital. High-riskpatients are routinely referred prenatally, during labor, or after birthto the University Teaching Hospital (UTH), the only teachinghospital which is affiliated with the University of Zambia. The NICUhas three ventilators, 34 incubators and 38 cots. The unit, whichhas approximately 1500 deliveries per year, is staffed by three to fivenurses per shift and a total of four to six physicians (a seniorconsultant, one or two registrars and two or three junior staff). TheNICU mortality rate is 41%; 60 to 65% of deaths occur in the first 3days following admission and 75 to 85% by the end of the first week(E Chomba, personal communication, February 27, 2006).

The infants cared for in the UTH NICU include all infants whosurvive following severe birth asphyxia and other significantdelivery complications in the Lusaka area (over 40 000 deliveriesper year) and are referred to the NICU follow-up clinic 4 to 6 weeksafter discharge. This clinic only follows infants cared for in theUTH NICU. Standard follow-up in the NICU follow-up clinic isevery 4 weeks but may be altered for clinical indications. Patientsare discharged from the NICU follow-up clinic to the local healthclinic when routine care is required or to the UTH neurologysubspecialty clinic when there is persistent evidence of delayedneurodevelopment beyond 3 months of age.

Study designA cross-sectional, prospective clinic data collection and retrospectivehospital chart review was completed on all infants who presented tothe NICU follow-up clinic for a 4-week period between March andApril 2005. The percent of infants who return for follow-up is notroutinely tracked but is estimated to be 75 to 85% based on hospitalrecords.

A structured data sheet was used to abstract information fromthe chart for each infant seen in the clinic during the study period.

Infant demographics, including date of birth, birth weight andgender, were collected. Delivery data included maternal parity,location (home, clinic or hospital) and method of delivery (vaginalor cesarean section), and Apgar score (1, 5, and 10 min) whenavailable.

All diagnoses documented during the hospital course wererecorded; therefore, infants may have had multiple diagnoses. Onlythose infants who received phototherapy were assigned a diagnosisof neonatal jaundice.

Birth asphyxia was a clinical diagnosis made during initialassessment by the UTH physician, based on delivery history and/orApgar score using the WHO (World Health Organization) definition offailure to initiate and sustain normal breathing at birth.24 For thisstudy the definition of clinical asphyxia was defined by hospitalprotocol as infants with a 5-min Apgar score <7; however, theclinical diagnosis may have been made in the absence of Apgar scoresfor home deliveries. This was similarly applied to both term andpreterm infants. Birth asphyxia was dichotomized as present or absentwithout assessment of severity. The rate of clinical diagnosis of birthasphyxia in this population was compared to rates of birth asphyxiacalculated with previously published criteria including the following:Bao criteria25 (5-min Apgar score <6), Ellis criteria10 (5-min Apgarscore p4) and Volpe criteria26 (5-min Apgar score <4).

Information recorded at the clinic follow-up visit included dateof visit, weight, skin color (normal or pale), neurologicexamination, problem list and plan of care. Neurologicexaminations were performed by three senior staff physicians,masked to clinical course, during the follow-up visit. The physicianwas aware of and considered the infant’s age and level ofprematurity when assessing tone. The neurologic examinationswere dichotomized as normal or abnormal. An abnormal examincluded any exam with increased or decreased head control and/or increased, decreased or mixed overall tone. Other examinationfindings, including any concerns with respect to hearing or visionimpairment, were recorded as open-ended text. The plan of carewas abstracted including when and where the infant was scheduledfor follow-up (NICU follow-up clinic, neurology clinic or ‘underfive’ clinic for routine care), what medications were prescribed, andwhether or not breastfeeding counseling was ordered. Other follow-up was recorded in an open-ended fashion. At the time the studywas conducted, no form of early intervention for infant stimulationwas present in the community. No information was obtained forthose infants who were not seen in the NICU follow-up clinic.

A few infants were seen more than once during the period ofchart abstraction. In these cases, the last follow-up visit for each ofthese infants was included in the analyses.

Data analysisOdds ratios and confidence intervals (CIs) were calculated for theentire population as well as for those infants born at (inborn) oroutside (outbornFclinic or home) the hospital. Data analyses

Birth asphyxia survivorsDR Halloran et al

244


included descriptive statistics, w2 and Fisher’s exact tests, andmultivariable backward regression analysis using SAS systems (SASInstitute Inc., Cary, NC, USA).

The study was approved by the Institutional Review Board at theUniversity of Alabama at Birmingham, the Research TriangleInstitute, as well as the Research Ethics Committee at the UTH inLusaka, Zambia.

Results

A total of 211 clinic visits were prospectively abstracted. A total of 19infants were seen twice and 5 infants were seen three times duringthe study period. A total of 182 infants remained for analyses. Themean age at the time of the visit was 67 days (s.d. ±58 days) witha median of 50 days.

HistoryThere was a slight predominance of males in the study population(99, 55%). The majority of infants (140, 79%) were born at UTHversus a clinic (21, 12%) or at home (17, 9%). Vaginal delivery wasthe most common method of delivery (137, 78%). Maternal parityof 1 to 2 was most common (114, 65%) versus parity X3 (48,27%) or nulliparity (14, 8%). The mean birth weight was 2243 g(s.d. ±781 g); the median was 2000 g. Suspected sepsis was theleading clinical diagnosis (74%) made during the initial NICU stay

among inborn and outborn survivors (Tables 1 and 2). A total of11 infants (6%) had a positive blood culture (Klebsiella (5),Proteus (2), Staphylococcus (1), Enterococcus (1), Enterobacter(1), Streptococcus (1) and Salmonella (1)).

A 1-min Apgar score was documented for 140 of 182 infants(77%); 114 infants (63%) had a 5-min and 34 infants (19%) had a10-min Apgar score documented. A total of 42 infants (23%) had aclinical diagnosis of birth asphyxia recorded on the hospital chart.Six of these infants did not have a 5-min Apgar score. Of the 36infants, 31 (86%) with Apgar scores met Bao criteria24 for birthasphyxia (Apgar score <7). Eight (22%) infants had birth asphyxiabased on Ellis criteria10 (Apgar score p4) whereas one infant hadbirth asphyxia based on Volpe criteria26 (Apgar score <4).

A total of 30% (35) of infants born in the hospital had a clinicaldiagnosis of birth asphyxia versus 20% (4) born in the clinic and12% (2) born at home. A total of 22% (30) of infants born byvaginal delivery had birth asphyxia versus 26% (10) born bycesarean section. A total of 24% (20) of female and 21% (21) ofmale infants had birth asphyxia. A total of 2 (14%) infants born tonulliparous mothers, 33 (29%) infants with maternal parity of 1 to2 and 4 (8%) infants with maternal parity X3 had birth asphyxia.These factors were not predictive of birth asphyxia.

Infants with birth weight <1500 g were 12.6 (95% CI 2.8, 57.2)times more likely to have birth asphyxia than infants >2500 gwhereas infants 1500 to 2500 g were 24.6 (95% CI 7.1, 85.3) times

Table 1 Study population characteristics by birth asphyxia and neurologic examination at follow-up for hospital birthsa

Birth asphyxia Neurologic examination

Characteristic N (%) Present (%)b OR (95% CI) Abnormal (%)b OR (95% CI)

Number of fetuses

Singleton 116 (83) 35 (30) F 21 (18) Referent

Twin/triplet 24 (17) 0 F F F 1 (4) 5.1 (0.65, 39.8)

Birth weight

<1500 21 (15) 1 (5) 20.7 (2.6, 164.3) 4 (19) 1.1 (0.3, 3.8)

1500–2500 59 (42) 3 (5) 19.3 (5.4, 68.7) 5 (8) 2.8 (0.9, 8.4)

>2500 59 (42) 30 (51) Referent 12 (20) Referent

Clinical diagnosesc

Sepsis 104 (74) 27 (26) 1.2 (0.5, 3.0) 18 (17) 1.7 (0.5, 5.3)

Prematurity 75 (54) 2 (3) 0.03 (0.0, 0.1) 9 (12) 0.5 (0.2, 1.4)

Respiratory distress 66 (47) 4 (6) 0.09 (0.03, 0.27) 9 (14) 0.7 (0.3, 1.9)

Birth asphyxia 35 (21) F F F F 11 (31) 3.9 (1.5, 10.1)

Jaundice 27 (19) 6 (22) 0.8 (0.3, 2.3) 5 (19) 1.3 (0.4, 3.9)

Seizures 12 (9) 8 (67) 7.5 (2.1, 26.7) 3 (25) 1.9 (0.5, 7.7)

Meningitis 7 (5) 4 (57) 4.4 (0.9, 20.7) 2 (29) 2.3 (0.4, 12.5)

aIndividual variable totals may vary due to missing data.bPercentage represents percent for a given characteristic (for example 35 of 116 singleton infants were born with birth asphyxia (30%)).cOdds ratios were calculated with the absence of a clinical diagnosis as the referent for each individual diagnosis (for example infants with birth asphyxia were 3.9 times more likely tohave an abnormal neurologic examination than infants without birth asphyxia).


245


more likely to have birth asphyxia. This pattern persisted in inbornbut not outborn infants (Tables 1 and 2). Infants with a history ofseizures (OR 7.4 (95% CI 2.0, 29.5)) and multiple births (OR 24.5(95% CI 1.5, 409)) were more likely to have a clinical diagnosis ofbirth asphyxia on bivariate analysis whereas infants withrespiratory distress were less likely to have birth asphyxia (OR 0.1(95% CI 0.02, 0.3)). Most of these relationships persisted in theinborn (Table 1) but not the outborn infants (Table 2).Multivariable regression analysis was not possible due tomulticolinearity and small sample size.

Physical findingsA total of 26 infants (14%) had an abnormal neurologicexamination. Infants 1500 to 2500 g were 3.8 (95% CI 1.3, 11.1)more likely to have abnormal neurologic examinations thaninfants >2500 g. This relationship did not persist in thesubanalyses (Tables 1 and 2). A total of 22 infants (16%) born inthe hospital had an abnormal neurologic exam versus 2 (10%)born in the clinic and 1 (6%) born at home. A total of 19 infants(14%) born by vaginal delivery had an abnormal exam versus 5(13%) by cesarean delivery. A total of 12% (10) of female and 16%(16) of male infants had an abnormal neurologic exam. A total of1 (7%) infant born to nulliparous mothers, 18 (16%) infants withmaternal parity of 1 to 2 and 5 (10%) infants with maternal parityX3 had an abnormal exam. These factors were not predictive ofan abnormal neurologic examination.

Infants with clinical birth asphyxia were 4.4 times more likelyto have an abnormal neurologic examination than nonasphyxiatedinfants (95% CI: 1.8, 10.4). These findings persisted in the inbornanalysis only (Tables 1 and 2). A total of 50% (13 infants) of the26 infants with an abnormal neurologic examination had clinicalbirth asphyxia (Figure 1). Of the 26, 10 infants with an abnormalneurologic examination met Bao24 criteria for birth asphyxia

Table 2 Study population characteristics by birth asphyxia and neurologic examination at follow-up for non-hospital birthsa

Birth asphyxia Neurologic examination

Characteristic N (%) Present (%)b OR (95% CI) Abnormal (%)b OR (95% CI)

Number of fetuses

Singleton 32 (84) 6 (19) Referent 3 (9) Referent

Twin/triplet 6 (16) 0 (0) F F 0 (0) F F

Birth weight

<1500 7 (19) 1 (14) 3.8 (0.3, 41) 0 (0) F F

1500–2500 17 (46) 0 (0) F F F F 0 (0)

>2500 13 (35) 5 (38) Referent 2 (15) Referent

Clinical diagnosesc

Sepsis 28 (74) 5 (18) 2.0 (0.2, 19) 1 (4) 0.2 (0.01, 1.85)

Prematurity 18 (47) 0 (0) F F 0 (0) F F

Respiratory Distress 11 (29) 0 (0) F F 0 (0) F F

Birth Asphyxia 6 (16) F F F F 1 (17) 2.0 (0.2, 19.1)

Jaundice 7 (18) 0 (0) F F 0 (0) F F

Seizures 2 (5) 1 (50) 6.2 (0.3, 115.9) 1 (50) 6.2 (0.3, 115.9)

Meningitis 3 (8) 1 (33) 3.0 (0.2, 39.6) 1 (33) 3.0 (0.2, 39.6)

aIndividual variable totals may vary due to missing data.bPercentage represents percent for a given characteristic (for example 6 of 32 singleton infants were born with birth asphyxia (19%)).cOdds ratios were calculated with the absence of a clinical diagnosis as the referent for each individual diagnosis (for example infants with birth asphyxia were 2.0 times more likely tohave an abnormal neurologic examination than infants without birth asphyxia).

0

40

80

120

160

Normal Abnormal

Neurologic Examination

Nu

mb

er o

f C

hild

ren

Birth Asphyxia No Birth Asphyxia

50%16%

Figure 1 Neurologic examination findings for infants with and without aclinical diagnosis of birth asphyxia.


246


whereas only 1 infant with an abnormal neurologic examinationmet Ellis10 criteria for birth asphyxia; none met Volpe26 criteria forbirth asphyxia. On the basis of backward selection regressionanalysis, clinical birth asphyxia was the only statistically significantpredictor of an abnormal neurologic examination.

Treatment planOf the 26, 22 (85%) infants with an abnormal neurologic examwere scheduled for continued observation in the NICU follow-upclinic; the remaining 4 infants (15%) were referred for continuedfollow-up to the pediatric neurology clinic for developmental delay.A total of 53 infants (29%) seen in the clinic were referred forroutine care to the ‘under five’ regional clinics with normalneurologic exams; the remaining 103 infants (57%) werescheduled to return to the NICU follow-up clinic for continuedmonitoring. Of the 42 infants with a diagnosis of birth asphyxia, 2infants (5%) were referred to neurology clinic, 7 infants (17%)were referred to regional clinics for routine care and 33 infants(79%) were scheduled to return to the NICU follow-up clinic.

Although the majority (128, 71%) of infants in the study werescheduled for follow-up in 4 weeks, 5 (3%) were scheduled forfollow-up in 3 weeks, 17 (9%) at 2 weeks and 19 (11%) at 1 week.The remaining 12 infants were scheduled for follow-up in 6 to 8weeks. Additional specialty follow-up was scheduled for maternalHIV testing (3), breastfeeding counseling (32) and physical therapy(3). In addition, two children were referred to general surgery foringuinal hernias, one child was referred to general surgery for anumbilical hernia, one child was referred to orthopedics for bilateralclubfoot, one child was referred to plastic surgery for webbing ofthe fingers and toes, and one child was referred to plastic surgeryfor a cleft of the right cheek with lip and mandible deformities.Although breastfeeding was nearly universal in the population, 32children were referred for additional breastfeeding counseling andsupport. Six children were readmitted for further evaluation andmonitoring with a diagnosis of possible sepsis (4), jaundice (1)and pneumonia (1).

Discussion

Birth asphyxia survivors account for 23% of infants seen in a NICUfollow-up clinic in a developing country. In this study, assessmentof early neurodevelopmental outcomes revealed that 31% ofsurviving infants with clinical birth asphyxia had abnormalneurologic examinations at early follow-up. Half of the childrenseen in the NICU follow-up clinic with abnormal neurologicexaminations had birth asphyxia. To the best of our knowledge thishas not been shown before.

Although birth asphyxia is known to be a leading cause ofneonatal mortality, survivors of birth asphyxia may suffer fromhypoxic–ischemic brain injury affecting vulnerable areas of thebrain leading to neurodevelopmental sequelae including problems

with sensory-motor, auditory and language processing.6 Kinoti,4 ina multicenter study in Sub-Saharan Africa, found that 87% ofinfants with birth asphyxia, using the WHO definition, survived todischarge; however, he did not assess the neurologic outcome ofthese survivors. We assessed the early neurologic examinations ofall birth asphyxia survivors who presented for follow-up and foundbirth asphyxia to be the leading diagnosis associated with anabnormal exam in this population of NICU survivors.

Our findings are consistent with those of Ellis et al.,10 whofound that 20% of infants with neonatal encephalopathy hadneurologic impairments at 1 year follow-up versus 50% in ourstudy, though at an earlier age. The Ellis study was conducted at asingle hospital in Nepal, and serial follow-up was conducted at 6weeks, 6 months and 1 year. At 1 year, 22% of infants had been lostto follow-up including 3 of 7 infants who had grade 2 neonatalencephalopathy with impairment at 6 weeks of age which, inaddition to the extended length of follow-up in the study, mayexplain the discrepancy in the rate of abnormal neurologic examsbetween the studies.

There is not full agreement on the research definition for birthasphyxia in the developing world. As described above, criteria byBao, Volpe and Ellis have been developed based on Apgar scores. Toallow comparison with previously published work on birth asphyxiain developing countries, we determined the rate of birth asphyxiain our population using these criteria. A total of 86% of our cohortof infants with a clinical diagnosis of birth asphyxia and an Apgarscore met Bao’s criteria for birth asphyxia leading to conclusionssimilar to ours about the impact of birth asphyxia onneurodevelopmental delay. However, future studies which utilizethe strict definition of birth asphyxia offered by Volpe and/or Ellismay, in fact, miss children who suffer from birth asphyxia and theresulting neurologic sequelae. Infants born without assignment ofan Apgar score, whether at home or in a clinic, are at risk for birthasphyxia and should be included in studies to identify birthasphyxia morbidity, particularly in light of previous studies whichhave questioned the validity of the Apgar score as a predictor ofbirth asphyxia.27 Although the World Federation of NeurologyGroup defined birth asphyxia as ‘a condition of impaired gasexchange leading y to progressive hypoxemia and hypercapnia,’obtaining a blood gas in developing countries for each infant whopresents with a concerning history is impossible.28

This study utilized a clinical diagnosis based on the WHOdefinition of birth asphyxia as failure to initiate and sustainbreathing at birth, the only feasible definition in this setting. Giventhe small sample size for certain outcomes, for example birthasphyxia in infants born at home, we are unable to make definitiveconclusions with respect to risk for these outcomes. Neurologicexaminations were performed at the time of the visit, given theaverage age of 67 days at the time of the visit, neurologic findingsmay be unreliable and may improve with time. However, data fromdeveloped countries indicate that early intervention is critical to


247


improved motor and cognitive impairments and such interventionsare currently unavailable in Zambia.29

All infants who presented for follow-up were included in theanalyses; however, the characteristics of the infants who failed tokeep the appointment are not known. Although previous studies indeveloped countries have found a bias in the outcomes of infantswho present for long-term follow-up,30–32 it is unclear if this istrue in developing countries. Transportation is difficult to afford formany families, regardless of the health of the infant, and likelycontributed significantly to the lack of follow-up for certain infants.Ultimately, this study examined the morbidity associated withsurvivors of birth asphyxia who attended the follow-up; ourfindings, therefore, represent only a portion of the true burden ofbirth asphyxia. Although subanalyses for inborn and outborninfants were presented, the number of infants born at home or inthe clinics was small limiting conclusions about the health ofinfants born outside the hospital.

The study was conducted in a single center which is the referralcenter for all the high-risk deliveries and newborn infants in Lusaka. Asdescribed in the ‘Methods’ section, we anticipate the patient populationis representative of other large municipalities in Sub-Saharan Africabecause of its similarities in demography, life expectancy and infantmortality.20,21 Formal neurological testing was not performed duringthe initial hospitalization; however, later neurologic examinations canbe considered reliable and provide an early assessment of infants whenperformed by experienced physicians.6 In future studies, long-termsequelae will be measured using established screening tools.

Conclusions

Birth asphyxia survivors accounted for 23% of infants seen in aNICU follow-up clinic in a developing country and 50% of infantswith early, abnormal neurologic examinations. Further research todetermine the long-term developmental consequences of birthasphyxia is needed. Birth asphyxia survivors are likely to be a largecontributor to early deaths and long-term disability in developingcountries. Cost-effective strategies to reduce the high burden ofbirth asphyxia are urgently required.

Acknowledgments

We thank all of the patients, physicians, nurses and medical records personnel at

The University Teaching Hospital in Lusaka, Zambia, who helped make this study

a success. In addition, thank you to Carolyn S Ashworth, MD, MPH at the

University of Alabama at Birmingham whose understanding and flexibility made

the work in Zambia possible.

This study was supported by Global Network for Women’s and Children’s

Health Research (HD 434646-03)FNational Institute of Health, The Bill and

Melinda Gates Foundation; BRAIN Planning Grant for the Developing World

(TW006703-02)FNational Institute of Child Health and Development; UAB

Perinatal Health and Human Development Pilot Grant; KL2 RR024994-ICTS

Multidisciplinary Clinical Research Career Development Program.

Conflicts of interest/Disclosure

The authors have no conflicts of interest to report.

References

1 Lawn JE, Cousens S, Zupan J, Lancet Neonatal Survival Steering Team. 4 million

neonatal deaths: when? Where? Why? Lancet 2005; 365: 891–900.

2 Jones G, Steketee RW, Black RE, Zulfigar AB, Morris SS, Bellagio Child Survival

Study Group. How many child deaths can we prevent this year? Lancet 2003; 362:

65–71.

3 Bryce J, Boschi-Pinto C, Shibuya K, Black RE, WHO Child Health Epidemiology

Reference Group. WHO estimates of the causes of death in children. Lancet 2005; 365:

1147–1152.

4 Kinoti SS. Asphyxia of the newborn in East, Central, and Southern Africa. East Af Med J

1993; 70: 422–433.

5 Simon NP. Long-term neurodevelopmental outcome of asphyxiated newborns.

Resuscitation Fetus Newborn 1999; 26: 767–778.

6 Robertson CMT, Finer NN. Long-term follow-up of term neonates with perinatal

asphyxia. Clin Perinatol 1993; 20: 483–499.

7 Robertson CMT, Finer NN, Grace MGA. School performance of survivors of

neonatal encephalopathy associated with birth asphyxia at term. J Pediatr 1989; 114:

753–760.

8 Cioni G, Prechtl HFR, Ferrari F, Paolicelli PB, Einspieler C, Roversi MF. Which better

predicts later outcome in fullterm infants: quality of general movements or

neurological examination? Early Hum Dev 1997; 50: 71–85.

9 Brookhouser PE. Sensorineural hearing loss in children. Pediatr Clin North Am 1996;

43(6): 1195–1216.

10 Ellis M, Manandhar N, Shrestha PS, Shrestha L, Manadhar DS, de L Costello AM.

Outcome at 1 year of neonatal encephalopathy in Kathmandu, Nepal. Dev Med Child

Neurol 1999; 41: 689–695.

11 Durkin MS, Wang W, Shrout PE, Zaman SS, Hasan ZM, Desai P et al. Evaluating a ten

question screen for childhood disability: reliability and internal structure in different

cultures. J Clin Epidemiol 1995; 48: 657–666.

12 Palisano R, Rosenbaum P, Walter S, Russell D, Wood E, Galuppi B. Development and

reliability of a system to classify gross motor function in children with cerebral palsy.

Dev Med Child Neurol 1997; 39: 214–223.

13 Chopra G, Verma IC, Seetharaman P. Development and assessment of a screening test

for detecting childhood disabilities. Indian J Pediatr 1999; 66: 331–335.

14 Stoll BJ, Measham AR. Children can’t wait: improving the future for the world’s poorest

infants. J Pediatr 2001; 139: 729–733.

15 Lawn JE, Cousens SN, Wilczyneska K. Estimating the Causes of Four Million Neonatal

Deaths in the Year 2000; Statistical AnnexFThe World Health Report 2005. World

Health Organization: Geneva, 2005.

16 English M, Muhoro A, Aluda M, Were S, Ross A, Peshu N. Outcome of delivery and

cause-specific mortality and severe morbidity in early infancy: a Kenyan district

hospital birth cohort. Am J Trop Med Hyg 2003; 69: 228–232.

17 Sall M, Sylla J. African Prime Ministers take lead in child survival. Lancet 2005; 366:

198–199.

18 Martines J, Paul VK, Bhutta ZA, Koblinsky M, Soucat A, Walker N et al. Neonatal

survival: a call for action. Lancet 2005; 365: 1189–1197.

19 Lawn JE, Manandhar A, Haws RA, Dramstadt GL. Reducing one million child deaths

from birth asphyxiaFa survey of health systems gaps and priorities. Health Res Policy

Syst 2007; 5: 4.

20 The World Health Report 2005FMake Every Mother and Child Count World Health

Organization: Geneva 2005, pp 200–204. Accessed at http://www.who.int/whr/2005/en/

on November 18, 2008.

21 The World Health Report 2005FMake Every Mother and Child Count World Health

Organization: Geneva 2005, pp 182–186. Accessed at http://www.who.int/whr/2005/en/

on November 18, 2008.


248


http://www.who.int/whr/2005/en/

http://www.who.int/whr/2005/en/

22 Mathers Cd, Sadana R, Salomon JA, Murray CJL, Lopez AD. Healthy life expectancy in

191 countries, 1999. Lancet 2001; 357: 1685–1691.

23 UNICEF. Information by country. Issues facing children in Zambia. Accessed at http://

www.unicef.org/infobycountry/zambia_1391.html on February 27, 2008.

24 Maternal and Newborn Health/Safe Motherhood Unit. Division of Reproductive Health

(Technical Support). Basic Newborn Resuscitation: A Practical Guide. World Health

Organization: Geneva, 1997; 4. Accessed at http://www.who.int/reproductive-health/

publications/newborn_resuscitation/basic_newborn_resuscitation.pdf on February 27, 2008.

25 Bao XL, Yu RJ, Li ZS. 20-item neonatal behavioral neurological assessment used in

predicting prognosis of asphyxiated newborn. Chin Med J 1993; 106: 211–215.

26 Volpe J. Hypoxic-ischemic encephalopathy: clinical aspects. In: Volpe J (ed). Neurology

of the Newborn, 3rd edn. WB Saunders: Philadelphia, 1995, pp 279–344.

27 American Academy of Pediatrics, American College of Obstetricians and Gynecologistis.

Use and abuse of the Apgar score. Pediatrics 1996; 98: 141–142.

28 Bax M, Nelson KB. Birth asphyxia: a statement. World Federation Neurology Group.

Dev Med Child Neurol 1993; 35: 1022–1024.

29 Spittle AJ, Orton J, Doyle LW, Boyd R. Early developmental intervention programs post

hospital discharge to prevent motor and cognitive impairments in preterm infants.

Cochrane Database Syst Rev 2007; 2: CD005495.

30 Tin W, Fritz S, Wariyar W, Hey E. Outcome of very preterm birth: children reviewed

with ease at 2 years differ from those followed up with difficulty. Arch Dis Child Fetal

Neonatal Ed 1998; 79: F83–F87.

31 Castro L, Yolton K, Haberman B, Roberto N, Hansen NI, Ambalavanan N et al. Bias in

reported neurodevelopmental outcomes among extremely low birth weight survivors.

Pediatrics 2004; 114: 404–410.

32 Callanan C, Doyle L, Rickards A, Kelly E, Ford G, Davis N. Children followed

with difficulty: how do they differ? J Paediatr Child Health 2001; 37:

152–156.


249


http://www.unicef.org/infobycountry/zambia_1391.html

http://www.unicef.org/infobycountry/zambia_1391.html

http://www.who.int/reproductive-health/publications/newborn_resuscitation/basic_newborn_resuscitation.pdf

http://www.who.int/reproductive-health/publications/newborn_resuscitation/basic_newborn_resuscitation.pdf

PERINATAL/NEONATAL CASE PRESENTATION

An embolic complication in a donor twin with severe twin–twintransfusion syndrome after fetoscopic interventionHO Ballard1, L Shook1, KY Lain2, L Burns1 and TM Crombleholme3

1Department of Pediatrics, University of Kentucky Medical Center, Lexington, KY, USA; 2Department of Obstetrics and Gynecology,University of Kentucky, Lexington, KY, USA and 3Division of Pediatric General, Thoracic, and Fetal Surgery, Cincinnati Children’sHospital Medical Center, University of Cincinnati College of Medicine, Cincinnati, OH, USA

Necrotic injury of an extremity in a donor twin is a rare complication of

twin–twin transfusion syndrome after selective fetoscopic laser

photocoagulation. We present the case of a 20-year-old gravida 2, para 1

with a twin gestation with severe twin–twin transfusion syndrome

(Quintero Stage 3B) who had treatment with selective fetoscopic laser

photocoagulation. Selective fetoscopic laser photocoagulation may be

associated with extremity necrosis in a donor twin.

Journal of Perinatology (2009) 29, 250–251; doi:10.1038/jp.2008.182

Keywords: necrotic injury; twin–twin transfusion syndrome; selectivefetoscopic laser photocoagulation

Introduction

Twin–twin transfusion syndrome (TTTS) is a potentially lethalcomplication occurring in 10 to 15% of monochorionic twinpregnancies, with the most severely affected neonates having amortality of 60 to 100%. The current therapies for TTTS includeserial amnioreduction, amniotic septostomy and fetoscopicapproaches. Selective fetoscopic laser photocoagulation (SFLP) ofplacental arteriovenous anastamosis is used in an effort to decreaseshunting from the donor fetus to the recipient fetus.1 We report acase of TTTS treated with SFLP at 20 weeks gestational age. Viabletwins were delivered at 28 weeks gestation with the donor twinfound to have severe ischemic injury to the left distal forearm.

Case

A 20-year-old gravida 2, para 1 with a twin gestation presented toan outside institution at 12 weeks gestation with severe TTTS.Following amnioreduction at 17 weeks, she was referred to a fetaltherapy center at 19 5/7 weeks for further evaluation andmanagement of severe TTTS (Quintero Stage 3B). Subsequenttherapy included SFLP of five vessels, amniotic fluid reduction and

laser microseptostomy of the intertwin membrane, all of whichwere performed at 19 5/7 weeks gestation. Cardiac function wasunchanged at the end of the procedure reflecting the baselinecardiomyopathy. A follow-up ultrasound demonstrated upperextremity asymmetry in the donor twin consistent with edema fromthe left mid-forearm distally. The edema resolved by 23 weeksgestation. The twins were delivered by cesarean section at 28 weeksat a referring institution due to placental abruption. The recipienttwin weighed 1135 g and had Apgars of 7 and 9 at 1 and 5 min.The initial hemogram had a white blood cell count of13.6� 103 ml�1, hemoglobin of 19.4 g (100 ml)�1 and plateletcount was 354� 103 ml�1. The donor twin had Apgars of 4 at1 min and 8 at 5 min. Birth weight was 789 g. The initialhemogram showed a white blood cell count of 14.7� 103 ml�1,hemoglobin of 8.6 g (100 ml)�1 and platelet count of203� 103 ml�1. Both twins required supplemental oxygen atresuscitation before transport and a brief course of CPAPafter admission to the neonatal intensive care unit. A vasculardisruption sequence injury to the donor twin’s left upper extremitywas noted at the time of birth (Figure 1). The physicalexamination of the donor twin did not demonstrate any evidence ofamniotic band sequence. The infant subsequently had distalamputation of the necrotic arm and recovered withoutcomplication. Pathology on the amputated extremity was consistentwith coagulative necrosis.

Discussion

Vascular disruption sequence injuries have been reported inmonochorionic twins, particularly in the setting of TTTS. Thesetypes of ischemic limb defects more commonly occur in therecipient twin and have a predilection for the lower extremity. Thecurrent treatment options of TTTS include SFLP of thearteriovenous anastamosis. SFLP has been demonstrated to have animprovement in survival of at least one twin compared toamnioreduction (76 vs 56%) and improve gestational age atdelivery,2 but more recent studies have not demonstrated a benefitof SFLP over amnioreduction.3 SFLP is associated with lessReceived 20 May 2008; revised 26 August 2008; accepted 29 September 2008

Correspondence: Dr HO Ballard, Department of Pediatrics, University of Kentucky Medical

Center, MS 467, 800 Rose Street, Lexington, KY 40536-0298, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.182



cystic periventricular leukomalacia compared to amnioreduction(6 vs 14%),2 although there has been no difference inneurodevelopmental outcome between survivors treated with SFLPvs those treated with amnioreduction. Previously describedneonatal complications of SFLP include extremity necrosis(recipient),4 ileal atresia (recipient) and thromboembolic liver

calcifications (donor).5 Other complications of SFLP includeintraoperative injuries from the trocar and/or laser. Althoughpossible in this case these causes of injury are less likely as theprocedure is guided by ultrasonography throughout and wasperformed at an experienced fetal care center. Our case is uniquebecause necrotic injury is well described in recipient twins with orwithout SFLP, but is rarely seen in the donor twin and usuallyaffects the lower extremity. The mechanism of necrosis in recipienttwins is thought to be secondary to polycythemia/hyperviscosity.The mechanism of injury in this case is unclear, but could beinflammatory and/or embolic in nature, or related to systemichypoperfusion at the time of the ablation. Knowing that this infanthas survived to 9 months without neurologic sequela, thelikelihood of a systemic process (inflammatory or hypoperfusion)appears less likely. Prior reports have demonstrated a case ofthromboembolic calcifications noted at autopsy on a donor twindelivered at 21 weeks gestation who had undergone SFLP at 19weeks.5 Although the mechanism of necrotic injury is not knownin this case, the most plausible explanation is an embolic event atthe time of the SFLP.

References

1 De Lia JE, Kuhlmann RS, Harstad TW, Cruikshank DP. Fetoscopic laser ablation of

placental vessels in severe previable twin–twin transfusion syndrome. Am J Obstet

Gynecol 1995; 172: 1202–1208; discussion 1208–11.

2 Senat MV, Deprest J, Boulvain M, Paupe A, Winer N, Ville Y. Endoscopic laser surgery

versus serial amnioreduction for severe twin-to-twin transfusion syndrome. N Engl J

Med 2004; 351: 136–144.

3 Crombleholme TM, Shera D, Lee H, Johnson M, D’Alton M, Porter F et al. A prospective,

randomized, multicenter trial of amnioreduction vs selective fetoscopic laser

photocoagulation for the treatment of severe twin–twin transfusion syndrome. Am J

Obstet Gynecol 2007; 197: 396.e1–396.e9.

4 Lopriore E, Sueters M, Middeldorp JM, Oepkes D, Vandenbussche FP, Walther FJ.

Neonatal outcome in twin-to-twin transfusion syndrome treated with fetoscopic laser

occlusion of vascular anastomoses. J Pediatr 2005; 147: 597–602.

5 Schnater JM, van Zalen-Sprock RM, Schaap AH, Festen S, Aronson DC. Ileal atresia and

thrombo-embolic liver calcifications diagnosed after treatment with intrauterine laser

coagulation therapy for twin-to-twin transfusion syndrome: report of 2 cases. J Pediatr

Surg 2005; 40: 875–876.

Figure 1 The photograph was taken immediately after delivery anddemonstrates necrosis of the distal left forearm with obvious protrusion of theradius and ulna.

Embolic complication in a donor twinHO Ballard et al

251


PERINATAL/NEONATAL CASE PRESENTATION

Epileptic status refractory to conventional treatment caused byvitamin B6 deficiencyL Valle-Morales1, E Cortes-Cros1, A Santana2,3, M Barber1, T Figueras1, JA Garcıa-Hernandez1

1Obstetrics and Gynecology Department, Hospital Universitario Materno Infantil de Canarias, Las Palmas, Spain; 2Genetic Unit,Hospital Universitario Materno Infantil de Canarias, Las Palmas, Spain and 3Research Unit, Hospital de Gran Canaria Dr Negrin,Las Palmas, Spain

Epileptic seizures due to pyridoxine deficiency, although rare, are known to

occur. They are characterized by untreatable convulsions during childhood.

We present the case of an epileptic status resistant to conventional treatment

in a 29-year-old woman, with no previous history of epileptic seizures, in

the 29th week of pregnancy, who responded to the intravenous

administration of pyridoxine.


Keywords: epileptic seizures; vitamin B6 deficit; pregnancy

Case study

We report the case of a 29-year-old healthy, pregnant woman with anobstetric history of one miscarriage and one premature delivery. Shewas treated daily with 400 mg of folic acid and 2 mg of vitamin B12during the first trimester and 80 mg of ferrum sulphate since the 24thweek. No other drugs were administrated. She reported a normal dietand she had not smoked or drunk alcohol during the pregnancy.

The patient was referred to the Emergency Service in the 29thweek of pregnancy, after suffering several self-limited episodes,1 min in duration, of intense headache accompanied by absenceseizures. In the Emergency Service, she presented three complexpartial seizures on the right side of the body, with secondarygeneralization. Vital signs remained within normal levels and theanalysis showed normal kidney function, negative proteinuria,normal hepatic enzymes and a discrete microcytic anemia withnormal platelet levels. In the obstetric examination, nopathological findings were noted; the fetus presented normalcardiac activity and adequate biometric parameters for its age.

A computerized cranial tomography was performed, with nosigns of acute ischemia, a ventricular system of normal size andmorphology, and no signs of expansion suggesting the ausence of

intracranial mass. A sample of cerebrospinal liquid was taken,presenting non-xanthochromia, glucose levels of72 mg (100 ml)�1, protein levels of 15.3 mg (100 ml)�1

(15–40 mg (100 ml)�1), 1 white blood cell per mm3, and lactatelevels of 1.6 mM (1.1–2.4 mM). A magnetic resonance imaging(MRI) of the cranium showed no clinically significant alterations.

Despite treatment with carbamazepine, phenytoin and valproate,she began to have generalized, subintrant seizures, failing toregain consciousness, for which she was transferred to the intensivecare unit. She showed vital signs within normal levels, with 96%arterial oxygen saturation.

She continued having generalized seizures, required sedation,orotraqueal intubation and artificial respiration. Despite administrationof a high dosage of midazolam by infusion (up to 40 mg (100 ml)�1)the seizures continued, so that propofol was added to the treatment ata dosage of 2 mg kg�1 h�1. Continuous electroencephalographicmonitoring (EEG) was prescribed, which proved persistent seizures.Magnesium sulfate was also administered in initial dose (6 g) andposterior permanent infusion of up to 54 mEq in 24 h.

Owing to the persistence of the seizures, plasma levels ofvitamin B6 were measured with results of <9 mM (14–100 mM).After the administration of 300 mg of pyridoxine intravenously, theseizures disappeared. An initial dose of 300 mg per 12 h wasadministered intravenously for the first 48 h, and was reduced to300 mg per day in the following days. At 7 days after the initialdose of vitamin B6, its concentration was 79 mM. The valproate andthe carbamazepine were withdrawn, and the magnesium infusionwas gradually reduced, with no seizures observed in the EEG. At 20days after the admission, the patient was conscious, orientated andcollaborative, with no apparent neurological focality, and theplasmatic concentration of vitamin B6 was 137 mM. The last EEG,previous to discharge, showed an acute activity in the anterior lefttemporal lobe with no clinical seizures. That is why, even theadequate response to vitamin B6 treatment, she continuedtreatment with phenytoin and vitamin B6 on discharge.

The patient had a posterior normal control of her pregnancy inthe high-risk obstetric unit. Finally, at 34þ 4 weeks of pregnancyReceived 11 March 2008; revised 18 August 2008; accepted 31 August 2008

Correspondence: Dr L Valle-Morales, Obstetrics and Gynecology Department, Hospital

Universitario Materno Infantil de Canarias, Avda. Marıtima del Sur S/N, 35016 Las Palmas

de GC, Canary Islands, Spain.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.180



she had a premature rupture of membranes. Labor was inductedwith a female newborn of 2160 g, with an Apgar test of 7/9 andumbilical artery and vein pH of 7.25 and 7.20, respectively. Thecourse of the puerperium was normal.

She had several controls in the Neurology Service after thepregnancy, with normal levels of vitamin B6, so that treatment wasstopped 3 months after the initial diagnosis of vitamin B6 deficit.She never had more seizures.

Discussion

Convulsions during pregnancy are an amazing problem, especiallyin cases like that of our patient without a background ofneurological pathology. Differential diagnosis is very important topromptly start the adequate treatment with the best results formother and fetus. The possibility of eclampsia must be ruled out.Our patient did not show signs of high blood pressure and/orproteinuria, or other disorders associated with preeclampsia such asthrombocytopoenia, or the alteration of hepatic enzymes andcoagulation factors.

The presence of intracerebral neoplasias was ruled out by thecranial scan and thrombosis of venous cerebral sinuses by cranialMRI. In the differential diagnosis, we included acute viral hepatitis,which was ruled out by the negative serology and the fact thatthere were no hepatic enzyme disorders. The normal urine levelsdetected of d-aminolevulinic acid and porphobilinogen excludedthe possibility of intermittent acute porphyria.

Convulsive seizures associated with vitamin B6 are a specificform of epilepsy first described in 1954. Symptoms are generalizedconvulsions a few hours after birth that are resistant to standardanticonvulsive treatment and are controlled by the administrationof pyridoxine. In 1978, Lott and et al.1 studied the brain of apatient with this form of epilepsy, finding a reduction in theneurotransmitter, g-aminobutyric acid, in association with adecrease in white matter.

Its prevalence is estimated at 1 in 400,000–700,000 and anautosomal recessive inheritable gene2 has been associated.Recently,3 it has been possible to identify the gene ALDH7A, in the5q31 zone, associated with many cases of pyridoxine-dependentepilepsy. Numerous mutations of ALDH7A have been described insufferers, although it is not currently possible to establishgenotype–phenotype correlations. The consequence of enzymedeficiency, due to mutations of the gene, is the accumulation ofthe metabolite P6C, which deactivates pyridoxal phosphate,essential cofactor in the metabolism of many neurotransmitters.The administration of pyridoxine in these patients not only correctssecondary pyridoxal phosphate deficiency but also induces thereduction of P6C and other toxic metabolites.

Some cases have been published4 in children and in adultspresenting hardly controlled convulsions since infancy. Schulze-Bonhage et al.5 published the case of a pregnant woman with

pyridoxine-associated convulsions in her infancy, free of seizuresfor 23 years under treatment with vitamin B6 supplements, whostarted new epileptic seizures resistant to standard anticonvulsivetreatment during pregnancy.

Some authors6 describe two different syndromes related withvitamin B6. First is characterized by pyridoxine-dependent seizures(PDS) and the other with pyridoxine-responsive seizures (PRS).Patients with PDS need a lifelong vitamin B6 supplement of15 mg kg�1 per day; in patients with PRS only the immediaterepletion of pyridoxine is necessary with no need for this to besupplemented for life.

Ours is a PRS case because seizures never appeared again with B6withdrawal. A deficiency of vitamin B6 alone is uncommon. There are,however, instances where diets deficient in the vitamin orchemotherapeutic/fortuitous ingestion of antagonists have led tohypovitaminosis B6 including the tuberculostatic isoniazid orpenicillamine used in treatment of patients with Wilson disease.7 Inabsence of other causes of vitamin B6 depletion, in our patient thepregnancy seems to be the most probable cause of vitamin decreasebecause of increased demand.8 Moreover, the anticonvulsivesthemselves may contribute to a greater decrease in vitamin B6 levels.9

Probably it had not been necessary to maintain the treatmentwith phenytoin in our patient, as we had the cause and theadequate treatment for her problem. However, as our first case, andwith the last EEG showing abnormal electric activity (withoutclinic effects), our neurologist decided to maintain it.

In summary, we recommend to test plasmatic vitamin B6 levelsin all patients with convulsions resistant to conventionalantiepileptic treatment, even where there is no background ofseizures during infancy.

References

1 Lott I, Coulombe T, Di Paolo R, Richardson E, Levy H. Vitamin B6-dependent seizures:

pathology and chemical findings in brain. Neurology 1978; 28: 47–54.

2 Bennett C, Huynh H, Chance P, Glass I, Gospe S. Genetic heterogeneity for autosomal

recessive pyridoxine-dependent seizures. Neurogenetics 2005; 6: 143–149.

3 Mills P, Struys E, Jakobs C, Plecko B, Baxter P, Baumgartner M et al. Mutations

in antiquin in individuals with pyridoxine-dependent seizures. Nat Med 2006; 12:

307–309.

4 Nakagawa E, Tanaka T, Ohno M, Yamano T, Shimada M. Efficacy of pyridoxal

phosphate in treating an adult with intractable status epilepticus. Neurology 1997; 48:

1468–1469.

5 Schulze-Bonhage A, Kurthen M, Walger P, Elger CE. Pharmacorefractory status

epilepticus due to low vitamin B6 levels during pregnancy. Epilepsia 2004; 45: 81–84.

6 Mishra D, Gupta VK. Pyridoxine dependent and pyridoxine responsive seizures. Indian

Pediatr 2005; 42: 291–292.

7 Jaffe I. The antivitamin B6 effect of penicillamine. Clinical and immunological

implication. Adv Biochem Psycopharmacol 1972; 4: 217–226.

8 Bruinse HW, Van den Berg H. Changes of some vitamin levels during and after normal

pregnancy. Eur J Obstet Reprod Biol 1995; 61: 31–37.

9 Sener U, Zorlu Y, Karaguzel O, Ozdamar O, Coker I, Topbas M. Effects of common anti-

epileptic drug monotherapy on serum levels of homocysteine, vitamin B12, folic acid

and vitamin B6. Seizure 2006; 15: 79–85.

Epileptic seizures and vitamin B6 deficiencyL Valle-Morales et al

253


IMAGING CASE REPORT

Osteopenia of prematurity, staphylococcal rib osteomyelitisTE Herman and MJ Siegel

Mallinckrodt Institute of Radiology, Washington University School of Medicine, St Louis, MO, USA


Case presentation

A 770 g infant boy was born to a Gravida 1, para 0 mother bycaesarean section following premature rupture of membranes. Theinfant had initial Apgars of 4 at 1 min, 6 at 5 min and 8 at 10 min.He was intubated immediately after birth and initially treated in anoutside hospital intensive care nursery until day 114 of life, whenhe was transferred to our neonatal intensive care unit because ofpersistent respiratory failure, bronchopulmonary dysplasia andretinopathy of prematurity (Figure 1). At the time of transfer theinfant was found to have staphylococcal sepsis, which was treatedwith vancomycin. He subsequently developed and was treated forCandidal and Klebsiella urinary tract infections while in thehospital. At 2 months of life, diffuse and focal soft tissue swelling ofthe chest was appreciated and chest and rib radiographs wereobtained. These were repeated 1 week later when swelling was morefocal (Figure 2). Diffuse osteopenia, healing rib fractures and afocal expansile lytic lesion of the left ninth rib and right seventhwere present. A chest wall sonogram was obtained (Figure 3) overthe area of short tissue swelling. On the day of the sonogram thewhite cell count was 24 000, having been 15 000 3 weeks before,

the erythrocyte sedimentation rate was 10 mm h�1 (normal 0 to20), and the C-reactive protein level was 42.7 mg l�1 (normal 0 to10). The alkaline phosphatase was 510 IU l�1 (normal 100 to310 IU l�1). Plasma calcium and phosphorus levels were normal.

Denouement and discussion

The patient underwent incision and drainage of the left chest walllesion adjacent to the ninth rib. The fluid obtained from this

Figure 1 Frontal film of chest at admission. The ribs and scapula are wellmineralized. Minimal right upper lobe infiltrate is present.

Figure 2 (a) Frontal chest at 2 months of life, (b) Frontal chest 1 week later.The skeleton has become demineralized. On the chest radiograph at 2 months oflife the posterior rib margins and scapula are difficult to visualize. The end of theright seventh rib (black arrow) is expanded and there is a destructive lesion seenin the left ninth rib (white arrow). On the radiograph 1 week later, the ends of theright seventh rib (R7) and left ninth rib (L9) are more expanded (white arrows).The cortex of both ribs is destroyed posteriorly. There is marked overlying softtissue swelling (curved white arrow). Splenomegaly is also present.Received 23 July 2008; accepted 1 August 2008

Correspondence: Dr TE Herman, Mallinckrodt Institute of Radiology, St Louis Children’s

Hospital, Washington University School of Medicine, 510 South Kingshighway Blvd., St Louis,

MO 63110, USA.




www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.204



subperiosteal abscess grew a pure culture of oxacillin-sensitiveStaphylococcus aureus, indicating acute osteomyelitis of the rib.The osteopenia and rib fractures with slightly elevated alkalinephosphatase led to a diagnosis of osteopenia of the premature. Theinfant was treated with Vitamin D supplementation, high mineralformula (EPF 24) and 42 days of intravenous oxacillin.Mineralization returned to normal as did the alkaline phosphatase.Cardiac sonography was repeated at intervals and remained normalwithout endocarditis. The rib lesions and osteopenia resolved(Figure 4), and the infant was discharged in good condition.

Acute osteomyelitis is uncommon in neonates.1 The mostcommon organism isolated is S. aureus.1,2 Factors which maypredispose to the development of acute osteomyelitis in neonatesare prematurity, omphalitis, umbilical catheterization, prolongedhospitalization, pneumonia or meningitis.1–3 Common clinical

features include soft tissue swelling, tenderness and decreasedmotion. Neonatal osteomyelitis most frequently involves thehumerus and femur.2 In both sites septic arthritis oftenaccompanies the osteomyelitis and the infant frequently hasswelling and marked decreased motion, referred to aspseudoparalysis in 95% of cases.3

Radiographic findings of osteomyelitis usually require 7 to14 days to appear. The findings of osteomyelitis are focal lyticlesion, particularly metaphyseal in long bones, periosteal reaction,soft tissue swelling. Treatment of acute osteomyelitis is drainage ofabscess cavities and intravenous antibiotics.3

Osteopathy of prematurity is a multietiologic metabolic bonedisease resulting in reduced bone mineralization and fractures. Themost common risk factors are inadequate supply of calcium andphosphorus, immobility, parenteral nutrition, diuretic or steroidtherapy.4 Treatment is aimed at beginning enteral feedings as soonas possible to allow efficient absorption of phosphate, and adequatedoses of Vitamin D.

References

1 Korakaki E, Aligizakis A, Manoura A, Hatzidaki E, Saitakis E, Anatoliotaki M et al.

Methicillin-resistant Staphylococcus aureus osteomyelitis and septic arthritis in

neonates: diagnosis and management. Jpn J Infect Dis 2007; 60: 129–131.

2 Weissberg ED, Smith AL, Smith DH. Clinical features of neonatal osteomyelitis.

Pediatrics 1974; 53: 505–510.

3 Knudsen CJM, Hoffman EB. Neonatal osteomyelitis. J Bone Joint Surg 1990; 72B:

846–851.

4 Harrison CM, Johnson K, McKechnie E. Osteopenia of prematurity: a national survey and

review of practice. Acta Paediatr 2008; 97: 407–413.

Figure 3 (a) Transverse and (b) longitudinal sonographic images of the chestwall over the area of soft tissue swelling. A well defined heterogeneous, hypoechoicmass (M) is present immediately anterior to the echogenic wall of the adjacent(ninth) rib (R), consistent with a subperiosteal abscess.

Figure 4 Frontal chest radiograph. After completion of therapy there is minimalresidual deformity of the right seventh and left ninth ribs (arrows), but theosseous structures of the chest are now otherwise normal and well mineralized.

Osteopenia of prematurityTE Herman and MJ Siegel

255


LETTERS TO THE EDITOR

Early administration of surfactant in spontaneous breathing withnCPAP through a thin endotracheal catheterFAn option in thetreatment of RDS in ELBW infants?

Journal of Perinatology (2009) 29, 256; doi:10.1038/jp.2008.245

With great interest we read the article ‘Surfactants: past, present andfuture’ by HL Halliday.1 This article describes very well the uniquesuccess story of this wonderful substance that has not yet beenfinished. We were surprised to read about ‘surfactant administrationin spontaneously breathing infants using a fine gastric tube’ as asingle report at a meeting. In our German literature, this method hasbeen more extensively reported. At our center, we developed andintroduced this method for the first time. This method combinesthe positive effects of nCPAP and surfactant. The underlying idea isthat inspiring surfactant is more physiologically appropriate thanreceiving it by positive pressure inflations as with the INSUREprocedure. We reported the results in 2003 at the annual meetingof the GNPI (Gesellschaft fur Neonatologie und PadiatrischeIntensivmedizinFSociety for Neonatology and Pediatric IntensiveCare) in Cologne.2 This method was a topic of discussion at theannual meetings of the GNPI in the following years too.3–7 Therefore,we performed a pilot study in infants with a gestational age below27 0=7 weeks. In this study, we showed the feasibility and animprovement of outcomes of the study group compared with ahistorical control group.8 The need for mechanical ventilationbecause of RDS was reduced from 77% in the historical control groupto 48% in the study group. Mortality decreased from 35 to 12% andthe rate of severe IVH in survivors from 32 to 5%. These results,combined with subsequent experiences, were published this year.9 Thedata confirmed the results of the original feasibility study. Ourongoing randomized controlled trials utilizing this technique areregistered (AMV trial ISRCTN 05025922, NINSAPP trial ISRCTN64011614), but the results are not yet available.

A KribsDepartment of Neonatology, Childrens’ Hospital University of

Cologne, Children’s Hospital, Cologne, GermanyE-mail: [email protected]

References

1 Halliday HL. Surfactants: past, present and future. J Perinatol 2008; 28: S47–S56.

2 Kribs A, Pillekamp F, Hunseler C, Bauerfeld C, Vierzig A, Roth B. 29th annual meeting of

the Society for Neonatology and Pediatric Intensive Care 2003 jul 3–5 Cologne,

Germany. Z Geburtsh Neonatol 2003; 207(Suppl 1): S4.

3 Teig N, Weitkamper A, Lilienthal E, Rossler L, Hohendahl J, Rieger C. 31th annual

meeting of the Society for Neonatology and Pediatric Intensive Care 2005 jun 16–18

Magdeburg, Germany. Z Geburtsh Neonatol 2005; 209(Suppl 1): S24.

4 Kribs A, Gopel W, Frankenbusch K, Gebauer C, Wieg C, Kattner E et al. 33 th annual

meeting of the Society for Neonatology and Pediatric Intensive Care 2005 jun 14–16

Hamburg, Germany. Z Geburtsh Neonatol 2007; 211(Suppl 1): S4.

5 Gopel W, Kribs A, Spiegler J, Hartel C, Gebauer C, Wieg C et al. 33 th annual meeting of

the Society for Neonatology and Pediatric Intensive Care 2005 jun 14–16 Hamburg,

Germany. Z Geburtsh Neonatol 2007; 211(Suppl 1): S5.

6 Kutschera J, Reiterer F, Muller W, Urlesberger B. 33 th annual meeting of the Society for

Neonatology and Pediatric Intensive Care 2005 jun 14–16 Hamburg, Germany.

Z Geburtsh Neonatol 2007; 211(Suppl 1): S5.

7 Prescher J, Moorthi C, Raedler E, von der Wense A. 33 th annual meeting of the Society

for Neonatology and Pediatric Intensive Care 2005 jun 14–16 Hamburg, Germany.

Z Geburtsh Neonatol 2007; 211(Suppl 1): S5.

8 Kribs A, Pillekamp F, Hunseler C, Vierzig A, Roth B. Early administration of surfactant in

spontaneous breathing with nCPAP: feasibility and outcome in extremely premature

infants (postmenstrual age <27 weeks). Paediatr Anaesth 2007; 17: 364–369.

9 Kribs A, Vierzig A, Hunseler C, Eifinger F, Welzing L, Stutzer H et al. Early surfactant in

spontaneously breathing with nCPAP in ELBW infantsFa single centre four year

experience. Acta Paediatr 2008; 97(3): 293–298.

Potential confounder of NEC clinical trials


We would like to congratulate Moss et al.1 for their recentmulti-center cohort study looking for clinical parametersthat predict the outcome in necrotizing enterocolitis (NEC).

Enrolling over 450 patients with NEC is a feat. However, wewrite to offer a different interpretation of their findings,particularly the association of NEC with progression to severedisease (i.e. surgery) within infants having never receivedenteral feeds.



www.nature.com/jp

http://dx.doi.org/10.1038/jp.2008.245



Ascertaining causes of neonatal deaths using verbal autopsy: current methods and challenges

Documents

Transcript of Ascertaining causes of neonatal deaths using verbal autopsy: current methods and challenges