IPD among Rural Bangladeshi Children • CID 2009:48 (Suppl 2) • S103

S U P P L E M E N T A R T I C L E

Invasive Pneumococcal Disease among Children inRural Bangladesh: Results from a Population-BasedSurveillance

Shams E. Arifeen,1 Samir K. Saha,2 Sayedur Rahman,1 Kazi Mizanur Rahman,1 Syed Moshfiqur Rahman,1

Sanwarul Bari,1 Aliya Naheed,1 Ishtiaq Mannan,1 M. Habibur R. Seraji,1 Nawshad U. Ahmed,3 M. Shameem Hassan,1

Nazmul Huda,1 Ashraf Uddin Siddik,1 Iftekhar Quasem,1 Maksuda Islam,2 Kaniz Fatima,2 Hassan Al-Emran,2

W. Abdullah Brooks,1 Abdullah H. Baqui,4 Robert F. Breiman,5 David Sack,1,4 and Stephen P. Luby1

1International Centre for Diarhoeal Disease Research, Bangladesh, and 2Dhaka Shishu, Children’s Hospital, Dhaka, and 3Kumudini Hospital,Mirzapur, Bangladesh; 4Department of International Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland; and 5Centersfor Disease Control and Prevention, Atlanta, Georgia

Background. Streptococcus pneumoniae infection is recognized as a global priority public health problem, andconjugate vaccines have been shown to prevent vaccine-type invasive pneumococcal disease (IPD) in children.However, better estimates of the disease burden and reliable population-based data on serotype composition areneeded for vaccine development and implementation in developing countries.

Methods. We initiated a population-based surveillance in the rural Bangladesh community of Mirzapur, cov-ering a population of ∼144,000. Village health care workers made weekly visits to ∼12,000 children 1–59 monthsof age in the study area. Children with reported fever, cough, or difficulty breathing were assessed by the villagehealth care workers using a clinical algorithm and were referred to the hospital if required. Children from thestudy area who were seen in the hospital underwent clinical examination and laboratory testing if they metstandardized case definitions. IPD was confirmed by blood and/or cerebrospinal fluid culture results. Isolates wereidentified, tested for susceptibility to antibiotics, and serotyped in accordance with standard laboratory methods.We present here the results from the first 3 years of the surveillance (July 2004–June 2007).

Results. Village health care workers identified 5020 cases of possible severe pneumonia and/or very severedisease (165 cases per 1000 child-years)and 9411 cases of possible pneumonia (310 cases per 1000 child-years) aswell as 2029 cases of suspected meningitis and/or very severe disease (67 cases per 1000 child-years) and 8967cases of high fever and/or possible bacteremia (295 cases per 1000 child-years). Pneumonia was the single mostcommon form of illness observed among 2596 hospitalizations (found in 977 [38%] of cases). We recovered 26S. pneumoniae isolates (25 isolates from 6925 blood cultures and 1 isolate from 41 cerebrospinal fluid cultures),which gave an overall IPD incidence of 86 cases per 100,000 child-years. Invasive pneumococcal infection wascommon during infancy (with infants accounting for 23 of the 26 cases), and 50% of the total isolates wereobtained from nonhospitalized patients who received a diagnosis of upper respiratory tract infection and fever.The most prevalent pneumococcal serotypes were serotypes 1, 5, 14, 18C, 19A, and 38. Ten of the 26 isolates werecompletely resistant to trimethoprim-sulfamethoxazole, and another 10 isolates had intermediate resistance.

Conclusions. IPD contributes substantially to childhood morbidity in rural Bangladesh. S. pneumoniae cancause invasive but nonsevere disease in children, and IPD incidence can be seriously under reported if such casesare overlooked. The emerging high resistance to trimethoprim-sulfamethoxazole should be addressed. Data onserotype distribution would help to guide appropriate pneumococcal conjugate vaccine formulation.

Acute respiratory infections are still among the leading

causes of childhood mortality in developing countries,

Reprints or correspondence: Dr. Shams E. Arifeen, Child Health Unit, PHSD,ICDDR,B, Mohakhali, Dhaka 1212, Bangladesh (shams@icddrb.org).

Clinical Infectious Diseases 2009; 48:S103–13� 2009 by the Infectious Diseases Society of America. All rights reserved.1058-4838/2009/4805S2-0010$15.00DOI: 10.1086/596543

including Bangladesh [1–5]. Recent data suggest that

pneumonia causes 19% of the ∼10 million annual

deaths among children worldwide, and 7 in 10 of these

deaths occur in Africa and Southeast Asia [1, 2]. Strep-

tococcus pneumoniae is identified consistently as the

leading cause of bacterial pneumonia in these popu-

lations [6–12]. Concerns for the persisting high burden

of invasive pneumococcal disease (IPD) and the rapid

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

S104 • CID 2009:48 (Suppl 2) • Arifeen et al.

Table 1. Case definitions used at study enrollment for possible pneumococcaldiseases.

Diagnosis Patient age Definition

Pneumonia 2–59 Months Tachypneaa and cough and/or difficult breathingSevere pneumonia 2–59 Months Tachypneaa and cough and/or difficult breathing

AND chest indrawing and no danger signsVery severe disease !5 Years Presence of danger signs with or without

pneumonia; if the child has a meningitis clin-ical syndrome, refer to the case definitionsfor meningitis; danger signs include inabilityto drink or breast-feed, convulsions, prostra-tion and/or lethargy (abnormally sleepy ordifficulty waking), severe malnutrition, stridorin a calm child, hypothermia (temperature,�35.5�C) or fever (temperature, �39�C), fastbreathing (if !2 months of age), chest in-drawing (if !2 months of age), and centralcyanosis

Meningitis !5 Years Sudden onset of fever and at least 1 of thefollowing: stiff neck, altered consciousnessand/or reduced level of consciousness, bulg-ing fontanelle (if !12 months of age), pros-tration and/or lethargy, convulsions, toxic ap-pearance, petechial or purpural rash, poorsucking, and irritability (if 12 months of age)

a Defined as �50 breaths/min for children aged �2 months but !12 months and as �40 breaths/min for children aged �12 months but !5 years.

emergence of multidrug-resistant S. pneumoniae [13–16] have

contributed to the recent increased interest in pneumococcal

vaccine research. Substantial progress has been made in the

development of pneumococcal polysaccharide and protein con-

jugate vaccines, and clinical trials have shown some impressive

results in protecting children against vaccine-type pneumo-

coccal disease, pneumonia, and meningitis in the United States

[17, 18] and South Africa [19, 20] and have led to reductions

in all-cause mortality among children in The Gambia [21].

In the United States, Levine et al. [22] have reinforced the

previous recommendation of the Centers for Disease Control

and Prevention Advisory Committee [23] for an immediate

and universal administration of the heptavalent pneumococcal

conjugate vaccine to children. However, in a developing area,

such as Bangladesh, where data are limited to only a few hos-

pital-based studies and where S. pneumoniae is not widely rec-

ognized as a priority public health problem, this knowledge

gap is a substantial barrier to any decision to incorporate a

pneumococcal vaccine into public health practice. The situation

is further complicated by the fact that available data from Ban-

gladesh on the serotype distribution of invasive pneumococcal

strains indicate a poor coverage with established vaccine for-

mulations, including the 11-valent conjugate vaccine [14, 24].

With assistance from the Pneumococcal Vaccines Accelerated

Development and Introduction Plan (PneumoADIP) [25], an

extensive hospital and population-based surveillance was ini-

tiated in 2004 to assess the burden associated with IPD. We

present here data from the first 3 years of a population-based

surveillance for IPD in a rural community of Bangladesh. We

provide information on IPD incidence, seasonal variations,

drug-resistance patterns, and serotype composition associated

with community-acquired disease in children in this rural Ban-

gladesh community.

PATIENTS, MATERIALS, AND METHODS

Study population, sample size and sampling procedure.

The surveillance was set-up in Mirzapur, a rural subdistrict of

Bangladesh, located 60 km north of the capital, Dhaka. Mir-

zapur has an estimated population of 400,000, distributed in

13 unions and 219 villages. The estimated annual birth cohort

is 11,000. The area is served by a 750-bed nonprofit private

hospital (Kumudini Hospital; Mirzapur, Bangladesh) and a 31-

bed government Upazilla (subdistrict) Health Complex, each

with laboratory and radiograph facilities.

Six of the 13 unions of Mirzapur were previously selected at

random for a neonatal health intervention study, and the same

6 unions also formed the population for this surveillance. The

study population was further divided into 72 clusters with

∼2000 population (175–190 children !5 years of age) each. All

children !5 years of age (excluding neonates 0–28 days of age)

who resided in those clusters were enrolled in the surveillance.

Informed consent was obtained from the parents or guardians

of the children. This was an open cohort, and children who

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

IPD among Rural Bangladeshi Children • CID 2009:48 (Suppl 2) • S105

Table 2. Revised case definitions for possible pneumococcal diseases used for analysis.

Diagnosis Patient age Case definition

Pneumonia 2–59 Months History of cough and/or difficult breathing and tachypneaa

Severe pneumonia (2–59 months) 2–59 Months History of cough and/or difficult breathing with or without tachypneaa

and chest indrawing; and no danger signsVery severe pneumonia !5 Years Child with cough and/or difficult breathing (with/without tachypneaa)

AND any of the following danger signs: inability to breast-feed ordrink, vomiting, convulsions, prostration and/or lethargy, central cy-anosis, severe respiratory distress, and tachypneab or chest in-drawing in children !2 months of age

Meningitis !5 Years Sudden onset of fever and at least 1 of the following: stiff neck, al-tered and/or reduced level of consciousness, bulging fontanelle (if!12 months of age), prostration and/or lethargy, convulsions,c toxicappearance, petechial or purpural rash, poor sucking and irritability(in children 12 months of age)

Very severe disease !5 Years Presence of at least 2 danger signs without pneumonia clinical syn-drome or lumbar puncture performed; if the child exhibits themeningitis or pneumonia clinical syndromes, refer to the case defi-nitions for meningitis and pneumonia; danger signs include inabilityto breast-feed or drink, vomiting, convulsions, prostrations and/orlethargy, severe malnutrition, and hypothermia (temperature,�36�C)

Bacteremia !5 Years Positive blood culture result but did not fulfill the case definitions forvery severe disease, meningitis, or pneumonia clinical syndrome

a Defined as �50 breaths/min for children aged �2 months but !12 months and as �40 breaths/min for children aged �12 months but !5years.

b Defined as �60 breaths/min.c Children !6 months or �6 years of age, any seizure; children �6 months but !6 years of age, any focal or prolonged seizure or 12 generalized

brief convulsions within a 24-h period.

were born or migrated into the study area were enrolled on a

continuing basis.

The surveillance sample size was based on an expected rate

of 0.23 episodes of IPD-like illness per year among a cohort

of ∼10,080 children under surveillance (i.e., 2318 episodes per

year). This figure was derived from the known !5-year acute

lower respiratory infection incidence in rural Matlab [26], an

area 105 km southeast of Mirzapur. With this estimate, we

expected ∼2300 blood cultures annually from the study site.

The surveillance started in July 2004. We here present the

findings from the surveillance for the period from July 2004

through June 2007.

Outcome variables, case definitions, and data collection.

A trained village health care worker (VHW) was assigned to

each of the 72 clusters for active surveillance of enrolled chil-

dren. The workers visited every household with an enrolled

child !5 years of age once per week and asked the mother (or

caretaker) of the child whether the child was experiencing any

illness at the time of the visit. If the child was reported to be

ill, the mother was then asked about the symptoms. If the

mother reported that the child had fever or cough or difficult

breathing, then the child was clinically assessed by the VHW.

The worker used an algorithm to classify children’s illnesses

that was adapted from the guidelines for the Integrated Man-

agement of Childhood Illness developed by the World Health

Organization [27 ]. Children classified as having possible severe

pneumonia (rapid breathing for children !2 months of age or

rapid breathing with at least 1 danger sign for children 2–59

months of age), suspected meningitis or very severe disease

(fever plus stiff neck, bulging fontanelle, inability to feed, con-

vulsion, being lethargic or abnormally sleepy or being difficult

to wake, and vomiting), or high fever and/or possible bacter-

emia (axillary temperature, �38.8�C for children �2 months

of age or �38.3�C for children !2 months of age, as measured

by the VHW) were referred to Kumudini Hospital. Rapid

breathing was defined on the basis of World Health Organi-

zation age-specific criteria [27 ] as �60 breaths/min for children

!2 months of age, �50 breaths/min for children 2–11 months

of age, and �40 breaths/min for children 12–59 months of

age. Danger signs included convulsion, lower chest indrawing,

being lethargic or abnormally sleepy or being difficult to wake,

being unconscious, inability to feed, and vomiting.

All children from the surveillance area who sought care for

illness from Kumudini Hospital, including those children who

were referred by the VHWs, are included in this analysis. Chil-

dren were examined by study physicians in the outpatient de-

partment of the hospital and, from March 2006, were also

examined in the emergency department. Blood samples were

collected for culture by a study nurse from children who were

admitted to the hospital with suspected pneumonia, severe

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

S106 • CID 2009:48 (Suppl 2) • Arifeen et al.

Table 3. Study flow chart.

VariableJul–Dec

2004Jan–Jun

2005Jul–Dec

2005Jan–Jun

2006Jul–Dec

2006Jan–Jun

2007 Total

No. of registered children 14,443 1416 1557 1660 1437 1865 22,378Mean no. of children seen per week 8158 10,767 11,190 10,805 10,299 10,087 10,231Child-years of observation 3909 5365 5576 5384 5132 5026 30,392No. of episodes assessed by village health care worker 6631 11,201 13,972 12,262 11,116 9206 64,388No. of episodes referred to Kumudini Hospital 1588 3200 2432 2621 2437 1617 13,895No. of child visits at hospital 1589 2857 2323 3875 3811 2898 17,353No. of hospitalizations 290 580 469 446 458 353 2596No. of blood cultures performed 91 577 787 1251 1952 2267 6925No. of CSF cultures performed 7 7 12 4 6 5 41

pneumonia, very severe disease, high fever (temperature,

138�C), or meningitis. A CSF sample was also collected by a

study physician for culture in case of meningitis or very severe

disease. However, on-duty clinicians also exercised their judg-

ment when deciding whether or not blood and/or CSF samples

would be collected. Study physicians enrolled children who met

standardized surveillance case definitions (table 1) and used

standardized forms to obtain and record clinical information

from these children. From July 2005, children who were not

admitted to the hospital but who otherwise met the case def-

initions had a blood sample collected for culture if consent was

given.

IPD was confirmed on the basis of isolation of the organism

from blood and/or CSF. Conventional blood culture bottles

with trypticase soy broth supplemented with 0.025% SPS and

1% isovitalex were inoculated with 2–3 mL of blood, incubated

at 37�C, and subcultured on days 1, 3, and 5. Isolates were

identified, tested for susceptibility to antibiotics, and serotyped

in accordance with the standard procedures [14, 24, 28, 29].

Further details of the methods are provided in Saha et al . [30].

Although a set of standardized surveillance case definitions

(table 1) were used by the study physicians for case ascertain-

ment and enrollment, a revised set of case definitions (table 2)

were used during analysis to ensure consistency with other

published data. The major differences between the 2 sets of

case definitions include the addition of “very severe pneu-

monia” and “bacteremia” classifications in the revised case def-

initions. Most individuals who were classified as having “very

severe disease” (with pneumonia) according to the previous

case definitions were classified as having “very severe pneu-

monia” according to the new case definitions. The new case

definitions also did not require the presence of rapid breathing

to classify a patient as having severe pneumonia. History of

cough and/or difficulty breathing, accompanied by chest in-

drawing alone, was sufficient to classify a case as severe pneu-

monia. The new case definitions addressed the issue of a de-

crease in respiratory rate in individuals with severe cases who

may present with only chest-indrawing. To exclude febrile con-

vulsion, the new case definition of meningitis also now con-

siders the type of seizure and the age of the child. Finally, the

new case definitions have eliminated temperature �39�C as an

entry criteria for the application of the various case definitions.

Data analysis. Statistical analysis was performed using

Stata, version 8.0 (Stata). A child started to contribute to per-

son-time as soon as consent was given. Children stopped con-

tributing to person-time if they died, moved out of the study

area, or reached the age of 60 months. Person-time was cal-

culated on the basis of the assumption that each weekly visit

was equivalent to 7-day follow-up of every child. For calculating

incidence rates, the number of isolates was divided by the child-

years of observation. The incidence for each of the months was

averaged over the 3 years of follow-up and was plotted to

indicate seasonality pattern.

The protocol was approved by the Ethical Review Committee

of the International Centre for Diarhoeal Disease Research,

Bangladesh.

RESULTS

From July 2004 through June 2007, a total of 22,378 children

were enrolled in the surveillance (table 3). During this period,

the mean number of children seen each week by the VHWs

was 10,201, which contributed to an estimated 30,392 person-

years of follow-up. VHWs assessed 64,388 episodes of reported

illness in these children and referred 13,895 episodes in a total

of 7259 children to Kumudini Hospital. An additional 5398

cases were referred by VHWs for moderate fever. At the hos-

pital, there were 17,353 visits at the outpatient and emergency

departments of the hospital by children enrolled in the sur-

veillance. Of these visits, 15,860 were referrals by VHWs (in-

cluding 4491 in which the child had only moderate fever), and

1493 children came to the hospital without a referral. Among

the children seen at the hospital, 2596 were admitted. A total

of 6925 blood samples and 41 CSF samples were obtained from

these children at outpatient and emergency department visits

and by inpatient departments.

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

Tabl

e4.

Ass

essm

ent

and

clas

sific

atio

nof

sick

child

ren

byhe

alth

care

wor

kers

,by

patie

ntag

e.

Dia

gnos

is

Chi

ldre

n1–

11m

onth

sof

age

(528

2ch

ild-y

ears

ofob

serv

atio

n)

Chi

ldre

n12

–23

mon

ths

ofag

e(6

141

child

-yea

rsof

obse

rvat

ion)

Chi

ldre

n24

–35

mon

ths

ofag

e(6

396

child

-yea

rsof

obse

rvat

ion)

Chi

ldre

n36

–47

mon

ths

ofag

e(6

461

child

-yea

rsof

obse

rvat

ion)

Chi

ldre

n48

–59

mon

ths

ofag

e(6

112

child

-yea

rsof

obse

rvat

ion)

Ove

rall

(30,

392

child

-yea

rsof

obse

rvat

ion)

No.

ofca

ses

No.

ofca

ses

per

1000

child

-yea

rsN

o.of

case

s

No.

ofca

ses

per

1000

child

-yea

rsN

o.of

case

s

No.

ofca

ses

per

1000

child

-yea

rsN

o.of

case

s

No.

ofca

ses

per

1000

child

-yea

rsN

o.of

case

s

No.

ofca

ses

per

1000

child

-yea

rsN

o.of

case

s

No.

ofca

ses

per

1000

child

-yea

rs

Sus

pect

edm

enin

gitis

and/

orve

ryse

vere

dise

ase

668

126

616

100

371

5823

336

141

2320

2967

Pos

sibl

ese

vere

pneu

mon

iaan

d/or

very

seve

redi

seas

e27

4552

012

1819

858

792

304

4716

627

5020

165

Hig

hfe

ver

and/

orpo

ssib

leba

cter

emia

2176

412

2470

402

1834

287

1420

220

1067

175

8967

295

Pos

sibl

epn

eum

onia

3433

650

2963

482

1575

246

902

140

538

8894

1131

0

Low

feve

r32

5061

528

9847

221

4433

516

9526

213

1421

511

,301

372

Cou

ghan

dco

ld14

,200

2688

9473

1543

7098

1110

5677

879

4206

688

40,6

5413

38

Oth

erill

ness

(mod

erat

efe

ver)

1229

233

949

155

712

111

652

101

556

9140

9813

5

Ove

rall

22,5

4142

6816

,052

2614

11,1

0417

3684

5713

0962

3410

2064

,388

2119

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

S108 • CID 2009:48 (Suppl 2) • Arifeen et al.

Table 5. Distribution of isolates from blood and CSF cultures, by patient age.

Pathogen

No. of isolates, by age in months

All ages,no. (%) of

isolates(n p 93)1–11 12–23 24–35 36–47 48–59

Escherichia coli 1 0 1 0 0 2 (2.2)Staphylococcus aureus 3 0 0 0 0 3 (3.2)Streptococcus species 4 0 0 2 0 6 (6.5)Salmonella typhi 0 6 7 15 8 36 (38.7)Haemophilus influenzae type b 8 1 2 0 0 11 (11.8)Streptococcus pneumoniae 14 9 1 1 1 26 (28.0)Other 5 1 3 0 0 9 (9.7)

Total 35 17 14 18 9 93 (100)

Table 6. Cause-specific distribution of hospitalized patients, no. of hospitalizations per 1000 child-years of observation, and Strep-tococcus pneumoniae isolation.

Variable PneumoniaaSevere

pneumoniaaVery severepneumoniaa Meningitisa

Very severediseasea Bacteremiaa Fever

Othercondition Total

Total no. of patients admitted to the hospital(no. of hospitalizations per 1000 child-years of observationb) 817 (27) 78 (3) 82 (3) 175 (6) 13 (0.4) 20 (0.7) 711 (23) 700 (23) 2596

No. of blood cultures performed 414 52 74 159 13 20 487 144 1363

No. of CSF cultures performed 1 0 0 39 0 0 0 1 41

No. of patients with a sample obtained 414 52 74 159 13 20 487 144 1363

No. of patients with S. pneumoniae isolated

S. pneumoniae isolated from blood 2 0 1 4 0 5 0 0 12

S. pneumoniae isolated from CSF 0 0 0 1 0 0 0 0 1

Overall 2 0 1 5 0 5 0 0 13c

a As defined in table 1.b 30,392 child-years of observation.c Thirteen additional S. pneumoniae isolates were obtained from nonhospitalized patients.

In table 4, we present the illness classification of the children

as assessed by the VHWs. The VHWs identified 2029 cases of

suspected meningitis or very severe disease, 5020 possible cases

of severe pneumonia, and 8967 cases of high fever and/or pos-

sible bacteremia. The VHWs were trained to refer these cases

to the hospital, and this resulted in 13,895 referrals, because

many of the children had 11 illness classification. The incidence

rate for all classifications decreased with age, with the highest

rates among children !1 year of age.

From the total of 6925 blood cultures and 41 CSF cultures

of samples from hospitalized patients, outpatients, and patients

seen at emergency department visits, 93 organisms were isolated

(isolation rate, 1.34%) (tables 3 and 5). More than two-thirds

of the isolates were Salmonella typhi (39% of isolates) and S.

pneumoniae (28%). Haemophilus influenzae type b (Hib) ac-

counted for 12% of the isolates. S. pneumoniae was most com-

mon among children !2 years of age, whereas Hib was found

mostly in children !1 year of age. On the other hand, S. typhi

was found only among children 1–4 years of age and was par-

ticularly common among children 12 years of age.

Of the 2596 total hospital admissions, 817 were for pneu-

monia, 78 were for severe pneumonia, and 82 were for very

severe pneumonia. There were 175 hospital admissions for

meningitis, 13 hospital admissions in which the illness was

classified as very severe disease, and 20 cases of bacteremia as

defined in table 1 (table 6). In addition, 74 hospitalized children

with fever and another 700 children with conditions defined

as “other” are also included in this analysis. The condition

“other” included those children who did not meet the case

definitions of pneumonia, meningitis, very severe disease, or

bacteremia. Table 6 also provides estimates of cause-specific

hospitalization rates. S. pneumoniae was isolated from 13 of the

children who were admitted to the hospital with pneumonia,

very severe pneumonia, meningitis, or bacteremia. Thirteen S.

pneumoniae isolates were obtained from patients seen at out-

patient or emergency department visits. This included 8 isolates

obtained from patients who received a diagnosis of upper re-

spiratory tract infection and another 5 isolates from patients

with fever. Five of these 13 patients received amoxicillin, 5

received trimethoprim-sulfamethoxazole, and 1 received ce-

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

IPD among Rural Bangladeshi Children • CID 2009:48 (Suppl 2) • S109

Table 7. Rates of invasive pneumococcal and Haemophilus influenzae type b (Hib) disease, by period and culture type.

VariableJul–Dec

2004Jan–Jun

2005Jul–Dec

2005Jan–Jun

2006Jul–Dec

2006Jan–Jun

2007 Overall

No. of child-years 3909 5365 5576 5384 5132 5026 30,392

No. of Streptococcus pneumoniae isolates obtained by culture

Blood 0 4 5 6 4 6 25

CSF 0 1 0 0 0 0 1

No. of cases of invasive S. pneumoniae disease per 100,000 child-years 0 93 90 111 78 119 86

No. of S. pneumoniae isolates obtained from inpatients, by culture

Blood 0 4 4 2 1 1 12

CSF 0 1 0 0 0 0 1

No. of cases of invasive S. pneumoniae disease per 100,000 child-yearsamong inpatients

0 93 72 37 19 20 43

No. of H. influenzae type b isolates obtained, by culture

Blood 2 1 3 2 0 3 11

CSF 2 0 1 0 0 0 3

No. of cases of invasive Hib disease per 100,000 child-years 51 19 54 37 0 60 36

Figure 1. Seasonality in incidence of invasive pneumococcal disease

phradine. Four of these outpatients still had symptoms when

they were assessed by VHWs at their routine weekly follow-up

visits. Two of these patients were referred again to Kumudini

Hospital, but only 1 of them complied with the referral; that

patient was treated again with trimethoprim-sulfamethoxazole

and recovered.

Twenty-five S. pneumoniae isolates were obtained from blood

cultures, and 1 was obtained from a CSF culture, giving an IPD

rate of 86 cases per 100,000 child-years observed (table 7). This

rate would have been only 43 cases per 100,000 child-years if

only hospitalized patients were included. This estimate would

be 98 cases per 100,000 child-years if data for 2004 are ignored,

when the newly established laboratory was not fully functional.

Also, if only the period from April 2006 (when the outpatient

surveillance became fully functional) through June 2007 is con-

sidered, the estimate increases to 108.1 cases per 100,000 child-

years. Eight Hib isolates were obtained from blood cultures,

and 3 were obtained from CSF cultures (from a total of 8

children), giving a Hib disease rate of 36 cases per 100,000

child-years observed. A seasonal variation of IPD incidence was

seen, with a prominent peak in March and another moderate

peak in September. This apparent seasonality could be greatly

influenced by random variation, given the small number of

isolates obtained (figure 1).

Serotypes 1, 5, 14, 18C, 19A, and 38 comprised more than

three-fourth of the S. pneumoniae isolates; serotypes 5, 14, and

19A were the most common serotypes (table 8). The S. pneu-

moniae isolates were found to be completely susceptible to

chloramphenicol, ampicillin, and ceftriaxone (table 9). Suscep-

tibility to ciprofloxacin was also high (96%). In contrast, only

23% of the S. pneumoniae isolates were susceptible to trimeth-

oprim-sulfamethoxazole; 38% were resistant, and 38% had re-

duced susceptibility. Susceptibility to gentamicin was even

lower (11% of isolates).

DISCUSSION

This population-based active case surveillance study in rural

Bangladesh provides valuable estimates of the burden of IPD.

Individuals with possible cases were actively detected in the

community through weekly home visits and were referred to

a hospital serving the population, and individuals seen at the

hospital had their cases investigated. We have a better under-

standing of the antimicrobial susceptibility and the serotype

distribution of invasive S. pneumoniae strains, the latter having

a direct relationship with possible vaccines. Almost all available

evidence of IPD burden is from surveillance studies performed

in hospitals and health care facilities, with some population-

based data from the control arms of large vaccine trials [7, 9,

21, 31–33]. Seeking care from health care facilities is uncom-

mon in Bangladesh; therefore, community-based active sur-

veillance is the only means for providing good estimates of

disease burden [3].

The surveillance in the Mirzapur community was designed

to identify and refer children with possible severe pneumonia,

meningitis and/or very severe disease, and high fever (possible

bacteremia). Although the incidence of possible severe pneu-

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

S110 • CID 2009:48 (Suppl 2) • Arifeen et al.

Table 8. Pneumococcal serotype distribution of invasive isolates, by age.

Serotype

No. (%) of isolates, by patient age in months

1–11 12–23 24–35 36–47 48–59 Overall

All serotypes 14 (100) 9 (100) 1 (100) 1 (100) 1 (100) 26 (100)1 0 (0) 1 (11) 1 (100) 0 (0) 0 (0) 2 (8)5 4 (29) 1 (11) 0 (0) 0 (0) 0 (0) 5 (19)10 F 0 (0) 1 (11) 0 (0) 0 (0) 0 (0) 1 (4)12 A 0 (0) 0 (0) 0 (0) 0 (0) 1 (100) 1 (4)14 2 (14) 3 (33) 0 (0) 1 (100) 0 (0) 6 (23)18 B 0 (0) 1 (11) 0 (0) 0 (0) 0 (0) 1 (4)18 C 1 (7) 1 (11) 0 (0) 0 (0) 0 (0) 2 (8)19 A 2 (14) 1 (11) 0 (0) 0 (0) 0 (0) 3 (11)35 B 1 (7) 0 (0) 0 (0) 0 (0) 0 (0) 1 (4)38 2 (14) 0 (0) 0 (0) 0 (0) 0 (0) 2 (8)45 1 (7) 0 (0) 0 (0) 0 (0) 0 (0) 1 (4)Not available 1 (7) 0 (0) 0 (0) 0 (0) 0 (0) 1 (4)

Table 9. Antimicrobial susceptibility pattern of Streptococcus pneumoniae isolatesfrom blood and CSF samples.

Antimicrobial agent

No. (%) of isolates

Tested forsusceptibility Susceptible

Reducedsusceptibility Resistant

Chloramphenicol 26 26 (100) 0 (0) 0 (0)Trimethoprim-sulfamethoxazole 26 6 (23) 10 (38) 10 (38)Ampicillin 26 26 (100) 0 (0) 0 (0)Ceftriaxone 26 26 (100) 0 (0) 0 (0)Ciprofloxacin 26 25 (96) 0 (0) 1 (4)Gentamicin 26 3 (11) 2 (8) 21 (81)Penicillin 26 22 (85) 4 (15) 0 (0)

monia and meningitis and/or very severe disease was highest

in the first year of life and decreased with age, high fever was

most common among children 12–23 months old. The overall

incidence of pneumonia (including both severe and nonsevere

pneumonia) was 0.47 episodes per child-year, which is sub-

stantially higher than the incidence of 0.23 episodes per child-

year reported from another Bangladesh rural area (Matlab) for

the period 1988–1999 [26]. One basic difference between the

2 sites is that, in Matlab, episodes were defined on the basis of

a minimum of 7 illness-free days between episodes with use of

data collected at twice-weekly home visits, whereas in Mirzapur,

illness seen at each of the weekly home visits was considered

to be a separate episode.

The age distribution of patients with S. pneumoniae and Hib

isolates was consistent with reports from other countries, with

Hib isolates being obtained almost exclusively from children

!1 year of age and S. pneumoniae isolates being obtained most

often from children !2 years of age. This suggests that admin-

istration of Hib and S. pneumoniae vaccines in early infancy

should be very effective in reducing much of the disease burden

caused by S. pneumoniae and Hib [34]. We did not observe

any obvious seasonal patterns of IPD, although a larger sample

of isolates might reveal such a pattern.

We note that 13 of the 26 S. pneumoniae isolates were ob-

tained from children who were seen as outpatients or in emer-

gency departments but were not hospitalized, which raises the

question of the severity of these cases. Eleven of these non-

hospitalized patients were given antibiotics, and 2 of them had

to be rereferred to Kumudini Hospital. It is possible that these

cases may have had the potential to become severe illnesses but

were identified and treated early, probably because of the active

population surveillance. The original plans for the surveillance

in Mirzapur did not include performing blood cultures on

samples obtained from nonhospitalized patients. The fact that

50% of the S. pneumoniae isolates were obtained from non-

hospitalized patients confirms that S. pneumoniae infection can

result in invasive but apparently nonsevere disease, and if such

conditions are overlooked, the total disease burden will be se-

riously underestimated. The overall incidence of IPD is 86 cases

per 100,000 child-years of observation, whereas if outpatient

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

IPD among Rural Bangladeshi Children • CID 2009:48 (Suppl 2) • S111

cases are excluded, the incidence decreases to 43 cases per

100,000 child-years of observation. This finding supports the

hypothesis that the variation in rates of IPD in socioecon-

omically similar regions, such as Western Europe and North

America, can be attributed to different blood culture practices,

especially in the outpatient departments [35]. This can result

in underdiagnosis and underreporting of mild IPD in some

populations. We note here that, of the outpatient cases yielding

S. pneumoniae, approximately one-half were diagnosed as upper

respiratory tract infection, and the rest were diagnosed as viral

fever. On the other hand, S. pneumoniae was isolated in 0.2%

of the outpatient cases, which is much lower than the 1% S.

pneumoniae isolation rate among the hospitalized children.

Hospitalized children yielded 13 of the 26 S. pneumoniae

isolates (12 from blood cultures and 1 from a CSF culture).

Although 2 of the 13 S. pneumoniae isolates were from patients

with pneumonia, none of the patients with severe pneumonia

yielded an isolate. We note that only 17% of the hospitalized

children with pneumonia or severe pneumonia had a blood

culture performed. This may partly explain the lack of any S.

pneumoniae isolates from these patients, because on the basis

of the 1% isolation rate, the 79 cultures performed could be

expected to yield !1 isolate. The decision as to whether a blood

sample would be obtained from a particular patient was made

by the hospital pediatrics team, and in many instances, they

did not feel that the clinical situation warranted a blood culture,

despite the suggestion of the PneumoADIP-agreed case defi-

nition. We may also hypothesize that the children admitted to

the hospital with pneumonia and/or severe pneumonia more

closely reflected the cases seen in the outpatient department.

These children yielded invasive S. pneumoniae isolates at a lower

rate than did other children hospitalized with more-severe ill-

ness and may have proceeded to more-severe illness with a

greater likelihood of yielding an S. pneumoniae isolate if they

had not been actively traced in the community, referred to the

hospital, and admitted for treatment.

We report an overall incidence of IPD of 86 cases per 100,000

child-years. This estimate of invasive pneumoccal disease bur-

den is consistent with findings from other populations in Africa

and Latin America, where rates ranging from 15 cases per

100,000 child-years in Mali [7] to 34 cases per 100,000 child-

years in Chile [31], 171 cases per 100,000 child-years in The

Gambia [8], 436 cases per 100,000 child-years in Kenya [36],

and 447 cases per 100,000 child-years in Kamalapur, Bangladesh

[37], have been reported. It is not clear what explains this

variation in the reported incidence of IPD. There are differences

in case detection procedures, eligibility criteria, case definitions,

and laboratory techniques in these studies. However, there are

also likely to be true differences in disease incidence.

We observed a predominance of serotypes 1, 5, 14, 18C, 19A,

and 38. Of these, serotype 38 has not, to our knowledge, been

reported as a common serotype in previous studies from Ban-

gladesh [14]. Of the serotypes covered in the 7-valent pneu-

mococcal conjugate vaccine available on the market (containing

pneumococcal capsular polysaccharide serotypes 4, 6B, 9V, 14,

18C, 19F, and 23F), only serotypes 14 and 18C were found in

this community, and they accounted for only 31% of all the

isolates [17]. Serotype coverage for the newer 10-valent (7-

valent pneumococcal conjugate vaccine antigens plus serotypes

1, 5, and 7F) and 13-valent (10-valent pneumococcal conjugate

vaccine antigens plus serotypes 3, 6A, and 19A) vaccines are

58% and 69% of all IPDs, respectively. Currently available vac-

cines do not provide good coverage of invasive pneumococcal

strains in this community, although the coverage is substantially

higher with vaccines that are in development. Even if the cur-

rent vaccine is introduced into the large birth cohort (∼4 mil-

lion) in Bangladesh, this would result in the prevention of a

large number of IPD cases. However, because a number of

strains that are prevalent in Bangladesh are not included in the

current vaccines or in upcoming vaccine candidates, we expect

that a substantial number of cases of IPD would continue to

occur in Bangladesh. Therefore, a vaccine formulation that

would provide species-wide coverage or include substantially

more serotypes would provide the greatest public health benefit.

The Government of Bangladesh continues to recommend

trimethoprim-sulfamethoxazole as the first-choice antibiotic

for the presumptive treatment of pneumonia. Previous reports

(from the urban area of Dhaka, India, and largely involving

meningitis cases) indicated that almost two-thirds of strains

were resistant to trimethoprim-sulfamethoxazole [14]. In the

Mirzapur rural community, nearly 80% of the S. pneumoniae

isolates were resistant to or exhibited reduced susceptibility to

trimethoprim-sulfamethoxazole. The high level of resistance of

S. pneumoniae to trimethoprim-sulfamethoxazole is an issue to

be addressed. There are reports that indicate that in vitro an-

timicrobial resistance and clinical effectiveness are sometimes

different for trimethoprim-sulfamethoxazole [38], and this ar-

gument has been used to justify the continued use of trimeth-

oprim-sulfamethoxazole for the treatment of community-ac-

quired pneumonia. In this population, more of the inpatient

isolates were completely resistant to trimethoprim-sulfameth-

oxazole (53% of inpatient isolates vs. 23% of outpatient iso-

lates) (data not shown). This emphasizes the need to revisit

the current World Health Organization recommendations of

trimethoprim-sulfamethoxazole as the antibiotic of choice for

treating nonsevere pneumonia in the community.

There are 4 main factors that may have contributed to an

underestimation of the true magnitude of IPD in Mirzapur.

First, as mentioned earlier, the once per week home visits may

have missed cases in children whose parents sought care from

other health care providers, although at this time, we do not

have data on the number of such children. A proportion of

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

S112 • CID 2009:48 (Suppl 2) • Arifeen et al.

children also visit government and other health care facilities.

Unfortunately, we did not collect any data from these facilities.

Second, because our VHWs were not offering any treatment

but only referral, there was very little incentive for the com-

munity to actively seek their help in case of childhood illness.

Third, blood cultures were not performed for all eligible cases

and would have contributed some cases of IPD that were

missed. Fourth, current blood culture techniques remain of low

sensitivity and do not identify all IPD cases. Blood cultures may

be augmented by other techniques, such as PCR and latex

agglutination, that substantially improve the yield of cases of

IPD.

When a 9-valent pneumococcal vaccine was evaluated in The

Gambia, the reduction in the prevalence of radiological pneu-

monia and IPD was much larger than was predicted from the

previously known distribution of pneumococcal serotypes in

The Gambia [21]. In Bangladesh, we have observed shifts in

the serotype distribution over time [14, 16, 24]. This cannot

be attributed to vaccine-induced serotype replacement, because

pneumococcal conjugate vaccine is not widely used in this com-

munity. Consequently, a pneumococcal conjugate vaccine

probe study may provide the best possible disease burden es-

timate for a developing country, such as Bangladesh. However,

caution must be used in interpreting any shift of serotype in

the future, after the introduction of the pneumococcal con-

jugate vaccine [30].

We believe that the findings from this study have provided

confirmation that IPD contributes substantially to childhood

illness in Bangladesh. One-half of the IPD cases were in children

who were seen as outpatients (in particular, children with symp-

toms of upper respiratory infection or fever). Prevention

through the administration of appropriate vaccines in the pub-

lic health system and high coverage of treatment of possible

cases with appropriate antibiotics is imperative.

Acknowledgments

Financial support. International Centre for Diarhoeal Disease Re-search, Bangladesh (ICDDR,B), and Pneumococcal Vaccines AcceleratedDevelopment and Introduction Plan (PneumoADIP), based at Johns Hop-kins Bloomberg School of Public Health and funded by GAVI Alliance;Australian Agency for International Development; Government of the Peo-ple’s Republic of Bangladesh; Canadian International Development Agency;Embassy of the Kingdom of The Netherlands; Swedish International De-velopment Cooperation Agency; Swiss Agency for Development and Co-operation; and Department for International Development, UK.

Supplement sponsorship. This article was published as part of a sup-plement entitled “Coordinated Surveillance and Detection of Pneumococ-cal and Hib Disease in Developing Countries,” sponsored by GAVI Alli-ance’s PneumoADIP of Johns Hopkins Bloomberg School of Public Health,Baltimore, Maryland.

Potential conflicts of interest. All authors: no conflicts.

References

1. Bryce J, Boschi-Pinto C, Shibuya K, Black RE. WHO estimates of thecauses of death in children. Lancet 2005; 365:1147–52.

2. Williams BG, Gouws E, Boschi-Pinto C, Bryce J, Dye C. Estimates ofworld-wide distribution of child deaths from acute respiratory infec-tions. Lancet Infect Dis 2002; 2:25–32.

3. Arifeen SE, Akter T, Chowdhury HR, et al. Causes of death in childrenunder five years of age. In: Bangladesh Demographic and Health Survey2004. Dhaka, Bangladesh: NIPORT, Mitra and Associates, ORC Macro,2005.

4. Baqui AH, Sabir AA, Begum N, Arifeen SE, Mitra SN, Black RE. Causesof childhood deaths in Bangladesh: an update. Acta Paediatr 2001; 90:682–90.

5. Rudan I, Tomaskovic L, Boschi-Pinto C, Campbell H. Global estimateof the incidence of clinical pneumonia among children under five yearsof age. Bull World Health Organ 2004; 82:895–903.

6. Enwere G, Biney E, Cheung YB, et al. Epidemiologic and clinical char-acteristics of community-acquired invasive bacterial infections in chil-dren aged 2–29 months in the Gambia. Pediatr Infect Dis J 2006; 25:700–5.

7. Campbell JD, Kotloff KL, Sow So, et al. Invasive pneumococcal in-fections among hospitalized children in Bamako, Mali. Pediatr InfectDis J 2004; 23:642–9.

8. Usen S, Adegbola R, Mulholland K, et al. Epidemiology of invasivepneumococcal disease in the Western Region, The Gambia. PediatrInfect Dis J 1998; 17:23–8.

9. O’Dempsey TJD, Mcardle TF, Lloyd-Evans N, et al. Pneumococcaldisease among children in a rural area of West Africa. Pediatr InfectDis J 1996; 15:431–7.

10. Forgie IM, O’Neill KP, Lloyd-Evans N, et al. Etiology of acute lowerrespiratory tract infections in Gambian children. I. Acute lower re-spiratory tract infection in infants presenting at the hospital. PediatrInfect Dis J 1991; 10:33–41.

11. Forgie IM, O’Neill KP, Lloyd-Evans N, et al. Etiology of acute lowerrespiratory tract infections in Gambian children. II. Acute lower re-spiratory tract infection in children ages one to nine years presentingat the hospital. Pediatr Infect Dis J 1991; 10:42–7.

12. Shann F. Etiology of severe pneumonia in children in developing coun-tries. Pediatr Infect Dis 1986; 5:247–52.

13. Levine S, Dejsirilert S, Sangsuk L, et al. Serotypes and antimicrobialresistance of Streptococcus pneumoniae in Thailand 2002–2004. PediatrInfect Dis J 2006; 25:176–8.

14. Saha SK, Baqui AH, Darmstadt GL, et al. Comparison of antibioticresistance and serotype composition of carriage and invasive pneu-mococci among Bangladeshi children: implications for treatment policyand vaccine formulation. J Clin Microbiol 2003; 41:5582–7.

15. Song JH, Lee NY, Ichiyama S, et al. Spread of drug-resistant Strepto-coccus pneumoniae in Asian countries: Asian Network for Surveillanceof Resistant Pathogens (ANSORP) Study. Clin Infect Dis 1999; 28:1206–11.

16. Saha Sk, Rikitomi N, Ruhulamin M, et al. Antimicrobial resistance andserotype distribution of Streptococcus pneumoniae strains causing child-hood infections in Bangladesh, 1993 to 1997. J Clin Microbiol 1999;37:798–800.

17. O’Brien KL, Moulton LH, Reid R, et al. Efficacy and safety of seven-valent conjugate pneumococcal vaccine in American Indian children:group randomised trial. Lancet 2003; 362:355–61.

18. Haddy RI, Perry K, Chacko CE, et al. Comparison of incidence ofinvasive Streptococcus pneumoniae disease among children before andafter introduction of conjugated pneumococcal vaccine. Pediatr InfectDis J 2005; 24:320–3.

19. Klugman KP, Madhi SA, Huebner RE, et al. A trial of a 9-valentpneumococcal conjugate vaccine in children with and those withoutHIV infection. N Engl J Med 2003; 349:1341–8.

20. Madhi SA, Kuwanda L, Cutland C, Klugman KP. The impact of a 9-valent pneumococcal conjugate vaccine on the public health burdenof pneumonia in HIV-infected and uninfected children. Clin InfectDis 2005; 40:1511–8.

21. Cutts FT, Zaman SMA, Enwere G, et al. Efficacy of nine-valent pneu-mococcal conjugate vaccine against pneumonia and invasive pneu-

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from

IPD among Rural Bangladeshi Children • CID 2009:48 (Suppl 2) • S113

mococcal disease in The Gambia: randomized, double-blind, placebo-controlled trial. Lancet 2005; 365:1139–46.

22. Levine OS, O’Brien KL, Knoll M, et al. Pneumococcal vaccination indeveloping countries. Lancet 2006; 367:1880–2.

23. Centers for Disease Control and Prevention. Preventing pneumococcaldisease among infants and children: recommendations of the AdvisoryCommittee on Immunization Practices (ACIP). MMWR Recomm Rep2000; 49(RR-9):1–35.

24. Saha SK, Rikitomi N, Biswas D, et al. Serotypes of Streptococcus pneu-moniae causing invasive childhood infections in Bangladesh, 1992 to1995. J Clin Microbiol 1997; 35:785–7.

25. The Pneumococcal Vaccines Accelerated Development and Introduc-tion Plan (pneumoADIP) Web site. Available at: http://www.preventpneumo.org. Accessed 31 December 2008.

26. Zaman K, Baqui AH, Yunus M, et al. Acute respiratory infections inchildren: a community-based longitudinal study in rural Bangladesh.J Trop Pediatr 1997; 43:133–7.

27. World Health Organization, Department of Child and AdolescentHealth and Development UNICEF. Management of the child with aserious infection or severe malnutrition: guidelines for care at the first-referral level in developing countries. Geneva, Switzerland: Dept. ofChild and Adolescent Health and Development UNICEF, 2000.

28. Jorgensen JH, Howell AW, Maher LA. Quantitative antimicrobial sus-ceptibility testing of Haemophilus influenzae and Streptococcus pneu-moniae by using the E-test. J Clin Microbiol 1991; 29:109–14.

29. National Committee for Clinical Laboratory Standards (NCCLS). Per-formance of standards for antimicrobial susceptibility testing: 12th in-ternational supplement. NCCLS document M100-S12. Wayne, PA:NCCLS, 2002.

30. Saha SK, Naheed A, Arifeen SE, et al.; Pneumococcal Study Group.

Surveillance for invasive Streptococcus pneumoniae among hospitalizedchildren in Bangladesh: antimicrobial susceptibility and serotype dis-tribution. Clin Infect Dis 2009; 48(Suppl 2):S75–81 (in this issue).

31. Lagos RL, Munoz A, Valenzuela MT, Heitmann I, Levine MM. Pop-ulation-based surveillance for hospitalized and ambulatory pediatricinvasive pneumococcal disease in Santiago, Chile. Pediatr Infect Dis J2002; 21:1115–23.

32. Ulloa-Gutierrez R, Avilla-Aguero ML, Herrera ML, Herrera JF, Ar-guedas A. Invasive pneumococcal disease in Costa Rican children: aseven year survey. Pediatr Infect Dis J 2003; 22:1069–74.

33. INCLEN. Prospective multicentre hospital surveillance of Streptococcuspneumoniae disease in India. Lancet 1999; 353:1216–21.

34. Baqui AH, El Arifeen S, Saha SK, et al. Effectiveness of Haemophilusinfluenzae type B conjugate vaccine on prevention of pneumonia andmeningitis in Bangladeshi children: a case-control study. Pediatr InfectDis J 2007; 26:565–71.

35. Hausdorff WP, Siber G, Paradiso PR. Geographical differences in in-vasive pneumococcal disease rates and serotype frequency in youngchildren. Lancet 2001; 357:950–2.

36. Brent AJ, Ahmed I, Ndiritu M, et al. Incidence of clinically significantbacteraemia in children who present to hospital in Kenya: community-based observational study. Lancet 2006; 367:482–8.

37. Brooks WA, Breiman RF, Goswami D, et al. Invasive pneumococcaldisease burden and implications for vaccine policy in urban Bangla-desh. Am J Trop Med Hyg 2007; 77:795–801.

38. Sazawal S, Black RE; Pneumonia Case Management Trials Group. Effectof pneumonia case management on mortality in neonates, infants, andpreschool children: a meta-analysis of community-based trials. LancetInfect Dis 2003; 3:547–56.

by guest on May 18, 2011

cid.oxfordjournals.orgD

ownloaded from