Antepartum risk factors for newborn encephalopathy

Intrapartum risk factors are important indeveloping world

Editor—Badawi et al’s case-control study ofneonatal encephalopathy in Western Aus-tralia shows that clinical evidence ofdysfunction of the central nervous system innewborn infants is associated with a widerange of disorders.1 For most of their studypopulation these disorders had originsbefore the onset of labour. We have two con-cerns about generalising their findings toother populations: their definition of neo-natal encephalopathy, and the greaterimportance of intrapartum risk factors forneonatal encephalopathy in poorer popula-tions in the developing world.

The omission of intrapartum criteriafrom the case definition of neonatalencephalopathy has been advocated previ-ously and removes an important bias affect-ing other studies.2 The investigators’ broadclinical definition of neonatal encephalopa-thy, however, makes comparison withprevalence studies in other settings prob-lematic. For instance, isolated neonatalseizures are difficult to ascertain clinically.3

What proportion of the 109 infantsreported to have seizures had interictal evi-

dence of neurological dysfunction? Otherinvestigators have chosen to exclude estab-lished causes of encephalopathy such asovert congenital infection and hypoglycae-mia from prevalence studies, so it would behelpful for comparative purposes to knowwhat proportion of the authors’ cases hadevidence of these.

Finally, the inclusion of 37 infants withbirth defects (23% of the cases) clearly hasimplications for the subsequent analysis ofthe likely time of the insult. In a high incomesetting with almost universal antenatal careand a relatively low stillbirth rate, the livebirth of neurologically impaired fetuses willprobably be maximised. In many lowincome countries mothers are stunted, donot access antenatal care, have high stillbirthrates, and receive poor obstetric care. Underthese conditions intrapartum factors prob-ably remain more important in the causa-tion of neonatal encephalopathy.Matthew Ellis Lecturer in child healthInstitute of Child Health, Royal Hospital forChildren, Bristol BS2 8BJM.Ellis@bristol.ac.uk

Anthony M de L Costello Reader in internationalchild healthCentre for International Child Health, Institute ofChild Health, London WC1N 1EH

1 Badawi N, Kurinczuk JJ, Keogh JM, Alessandri LM,O’Sullivan F, Burton PR, et al. Antepartum risk factors fornewborn encephalopathy: the Western Australian case-control study. BMJ 1998;317:1549-53. (5 December.)

2 Birth asphyxia: a statement. Develop Med Child Neurol1993;35:1022-4.

3 Connel J, Oozer R, DeVries L, Dubowitz LMS, Dubowitz V.Continuous EEG monitoring of neonatal seizures:diagnostic and prognostic considerations. Arch Dis Child1989;64:452-8.

Inverse association of risk may be due toeasier delivery with elective caesareansection

Editor—Badawi et al highlight the complexinteraction between antenatal and intra-partum risk factors in the aetiology ofnewborn encephalopathy among terminfants.1 2 They have suggested that electivecaesarean section is a protective factor; thisrequires some clarification.

The number of cases managed by electivecaesarean section is small, and I suggest thatthe inverse association of risk of encepha-lopathy is in fact due purely to the absence ofdifficulty of delivery in this group. This wouldmake more sense both statistically and on thegrounds of biological plausibility. If one com-pares all babies in each group exposed to anapparently difficult delivery then the risk ofencephalopathy is greatly increased (table).The corollary is that all babies in each groupexposed to an apparently uncomplicateddelivery will have a reduced risk of encepha-lopathy. This does not necessarily mean thatthe delivery mode is protective.

This report of an apparently protectiveeffect of elective caesarean section on therisk of newborn encephalopathy might beused by those who wish to strengthen theirargument in favour of routine electivesection. The figures show that 32% of thestudy population had a complicated deliv-ery. Altogether, 174 additional electivesections would therefore be required to pre-vent one case of newborn encephalopathy;this is based on the assumption that one canpredict antenatally which women will havedifficulty in labour and that the encepha-lopathy is due to a difficult delivery alone.Badawi et al have shown that this is unlikelyto be the case.2 The authors have also omit-ted to report the apparently protective effectof spontaneous vaginal delivery on the riskof newborn encephalopathy (49 (30%) v 161(40%); unadjusted odds ratio 0.63 (95% con-fidence interval 0.43 to 0.93)).

Badawi et al have tried to unravel thecomplexity of the interaction of risk factorsin the aetiology of newborn encephalopa-thy, but it would be a shame if results basedon small numbers were taken out of context.Deirdre J Murphy Lecturer in obstetrics andgynaecologyDepartment of Obstetrics and Gynaecology,St Michael’s Hospital, Bristol BS2 8EG

1 Badawi N, Kurinczuk J, Keogh JM, Alessandri LM,O’Sullivan F, Burton PR, et al. Antepartum risk factors fornewborn encephalopathy: the Western Australian case-control study. BMJ 1998;317:1549-53. (5 December.)

2 Badawi N, Kurinczuk J, Keogh JM, Alessandri LM,O’Sullivan F, Burton PR, et al. Intrapartum risk factors fornewborn encephalopathy: the Western Australian case-control study. BMJ 1998;317:1554-8. (5 December.)

Advice to authorsWe prefer to receive all responses electronically,sent either directly to our website or to theeditorial office as email or on a disk. Processingyour letter will be delayed unless it arrives in anelectronic form.

We are now posting all direct submissions toour website within 72 hours of receipt and ourintention is to post all other electronicsubmissions there as well. All responses will beeligible for publication in the paper journal.

Responses should be under 400 words andrelate to articles published in the precedingmonth. They should include <5 references, in theVancouver style, including one to the BMJ articleto which they relate. We welcome illustrations.

Please supply each author’s currentappointment and full address, and a phone orfax number or email address for thecorresponding author. We ask authors to declareany competing interest. Please send a stampedaddressed envelope if you would like to knowwhether your letter has been accepted or rejected.

Letters will be edited and may be shortened.

www.bmj.comletters@bmj.com

Proportions of cases (babies with neonatal encephalopathy) and controls having complicated* anduncomplicated† deliveries in Badawi et al’s study

No (%) of cases (n=164) No (%) of controls (n=400) Unadjusted odds ratio (95% CI)

Complicated delivery 79 (48) 101 (25) 2.75 (1.88 to 4.02)

Uncomplicated delivery 85 (52) 299 (75) 0.36 (0.25 to 0.53)

*Complicated deliveries: instrumental deliveries, emergency caesarean sections, breech manoeuvres.†Uncomplicated deliveries: spontaneous vaginal deliveries, induced vaginal deliveries, elective caesarean sections.

Letters

Website: www.bmj.comEmail: letters@bmj.com

1414 BMJ VOLUME 318 22 MAY 1999 www.bmj.com

Authors’ reply

Editor—Ellis and Costello raise two crucialissues. Firstly, they point out that the risk fac-tor profile and the relative importance ofindividual risk factors in developing nationsmay differ from those in the developedworld. This is important, because the globalburden of perinatal morbidity and mortalityfalls mainly on the developing nations.

Secondly, they emphasise the need for astandard definition of newborn encepha-lopathy that does not assume an intrapartumaetiology and explicitly states exclusion crite-ria. Such criteria may include some birthdefects, but this requires debate. In our study,all infants meeting entry criteria wereincluded. This maximises generalisability.Only 11 case infants had birth defectssufficient, on their own, to explain theencephalopathy. Exclusion of these infantsmade no material difference to our conclu-sions. Of the 109 case infants with seizures,104 had evidence of interictal neurologicaldysfunction.

Murphy raises the role of elective caesar-ean section. As she points out, when uncom-plicated deliveries are compared with compli-cated deliveries, uncomplicated deliveries areassociated with a lower risk (crude odds ratio0.36 (95% confidence interval 0.25 to 0.53)).We would not choose this as a primary analy-sis because the complexity of delivery is inpart an outcome of the causal sequence lead-ing to the newborn encephalopathy andshould not therefore be treated as a simpleexplanatory variable. Nevertheless, the statedassociation clearly exists. But, crucially, itstrengthens our belief about the appropriateway to interpret our findings.

We believe four points to be true.(1) There is a strong negative associ-

ation between uncomplicated delivery andnewborn encephalopathy.

(2) There is a strong negative associ-ation between elective caesarean section andnewborn encephalopathy.

(3) If analysis is restricted to womenwho (when conventional criteria are used)might be expected to benefit from anelective section, a strong negative associationbetween elective section and newbornencephalopathy remains.

(4) A relatively large proportion of casemothers who might have been expected tobenefit from an elective section did notreceive one.

These observations may be interpreted intwo ways. Firstly, our study is observational,and the causal pathways leading to newbornencephalopathy are complex. Consequently,our findings may reasonably be interpreted asshowing no more than an interestingassociation. This conclusion would reflect thefact that we have provided absolutely nodefinitive evidence of a (negative) causal linkbetween elective section and newbornencephalopathy. Conversely, although thestudy is observational and the causal path-ways are complex, it may be argued that onesimple hypothesis that would be entirely con-sistent with the observed associations is thatan elective section can, in some at risk

pregnancies, mitigate the risk of newbornencephalopathy.

In conclusion, our study does not provethat an elective section protects againstnewborn encephalopathy. Nevertheless, ourdata are consistent with such a protectiveeffect. Furthermore, it is difficult to come upwith an alternative explanation that is ascoherent. In this situation we believe thatone must take a decision-theoreticalapproach to inference by considering thecosts and benefits of accepting oneinterpretation over the other. The formerinterpretation implies that nothing needs tobe done. The latter suggests that furtherresearch needs to be undertaken, with studydesigns that will allow more to be said aboutthe role of elective section in the aetiologyof newborn encephalopathy. Given thatelective section may genuinely be protec-tive, we believe that to assume the formerinterpretation is unacceptable.Nadia Badawi Consultant neonatologistJennifer J Kurinczuk EpidemiologistFiona J Stanley DirectorTVW Telethon Institute for Child Health Research,PO Box 855, West Perth, Western Australia 6872,Australia

John M Keogh Consultant obstetrician andgynaecologistDepartment of Obstetrics and Gynaecology,Hornsby Ku-ring-Gai Hospital, HornsbyNew South Wales 2077, Australia

Paul R Burton Senior biostatisticianDepartment of Paediatrics, University of WesternAustralia, Western Australia 6907, Australia

Backdoor euthanasia

Withholding food and fluids is justifiableonly for terminally ill

Editor—Recent articles in the general1 andmedical press2 on “backdoor euthanasia”illustrate the confusion that surrounds thesubject. This is especially true for decisionsto withdraw or withhold food and fluidsfrom ill patients. Focusing on the issue of theintention behind such actions may help toachieve some clarity.

Current research, although limited,suggests that the desire for food and drinklessens in terminally ill patients and thatartificial hydration neither prolongs survivalnor alleviates symptoms.3 4 Furthermore,drips and nasogastric feeding tubes cancause unnecessary distress to patients andtheir relatives. In such cases the withholdingof these treatments is entirely appropriate asthe intention is to save the patient from atreatment that has no medical benefit.

But what about those who are not in theterminal stages of an illness—for example,patients who have had a recent stroke orwith acute infection? In most instancesdeath is not imminent. However, not provid-ing food and fluids would certainly lead todeath. In this case the intention behind theomission is to let the patient die. It fulfils thedefinition of an act of euthanasia and assuch is morally unjustifiable.

This leaves one further question: when isa patient’s condition terminal? That is, when

is a patient at a stage where recovery ishighly improbable and provision of artificialnutrition and hydration is of no benefit?This is not always easy to answer. Clinicalsigns and biochemical tests act as a goodindicator, as does experience in dealing withsuch cases. But when there is doubt, as thereoften is, it seems common sense to err onthe side of caution and provide allreasonable means to aid recovery, includingthe basic human needs for food and drink.James Paul Specialist registrar in palliative medicineKilburn, London NW6 7HHJamesPaul@compuserve.com

1 Various articles on euthanasia Times 6 January 1999:1, 2,9, 17.

2 Dyer C. Police investigate “euthanasia” deaths. BMJ1999;318:143. (16 January.)

3 Ellershaw JE, Sutcliffe JM, Saunders CM. Dehydrationand the dying patient. J Pain Symptom Management 1995;10:192-7.

4 Dunlop RJ, Ellershaw JE, Baines MJ, Saunders CM. Onwithholding nutrition and hydration in the terminally ill:has palliative care gone too far? A reply. J Med Ethics1995;21:141-3.

Intravenous and enteral fluids may causerather than alleviate suffering

Editor—I am worried by the recent report-ing and police investigation of “euthanasia”deaths due to dehydration.1 I was distressedby articles in the Times which seemed toexploit bereaved families who had supportedthe decline and death of their relatives overmany years, by suggesting that they mayhave suffered unnecessary pain and distressbecause of poor, or even criminal, mis-management.2

Not all doctors and nurses agree thatgiving intravenous or enteral fluids to dyingpatients who can no longer swallow or feelthirst eases suffering. Many experts in pallia-tive care believe the contrary. Zerwekh, aclinical coordinator of the Hospice ofSeattle, observed that giving fluids and inter-fering with the natural course of dehydra-tion can cause acute discomfort to thepatient near death and emotional distress tothe family.3 He commented that if thekidneys have not shut down, the fluids cansharply increase the flow of urine. If patientsare extremely weak, have lost bladdercontrol, or are in a coma, this increase maynecessitate insertion of a catheter. The fluidsalso significantly increase gastrointestinalfluids, which is a major problem for patientswhose vomiting is difficult to control,especially when there is bowel obstructionrequiring a nasogastric tube. Intravenousfluids also tend to increase respiratory secre-tions, making it more difficult for patients tocatch their breath or cough, and suction maybe required. Fluids can also cause a flare upof oedema and ascites and expand theoedema layer around tumours, aggravatingsymptoms, particularly pain.3

With increasing reliance on technologyrather than nursing and family support fordying patients, we seem to be followingNorth American medicine into the hole thatit is currently trying to climb out of. A studyof long term patients found that tubefeeding did not prevent aspiration pneu-monia. The authors concluded “for almost

Letters

1415BMJ VOLUME 318 22 MAY 1999 www.bmj.com

all conscious patients we suggest a dedicatedattempt at feeding by hand.”4 As the nursesat Kingsway Hospital observed,2 over zeal-ous nil by mouth orders can also causeunnecessary dehydration and suffering. Weneed to remember the real meaning of theword euthanasia—a quiet and easy death.This is what we all want for ourselves andneed to strive to achieve for patients by indi-vidual assessment of their needs, not byadherence to prescriptive protocols.Mary Bliss Consultant geriatricianBryning Day Hospital, Homerton Hospital,London E9 6SR

1 Dyer C. Police investigate “euthanasia” deaths. BMJ1999;318:143. (16 January.)

2 Various articles on euthanasia Times 6 January 1999:1, 2,9, 17.

3 Zerwekh JV. The dehydration question. Nursing 1983;13:47-51

4 Finetune TE, Bynum JPW. Use of tube feeding to preventaspiration pneumonia. Lancet 1996;348:1421-4

Widening screening to detectChlamydia trachomatis is moreimportant than using DNAmethodsEditor—The letter from Taylor-Robinsonand Robinson on the use of DNA amplifica-tion methods for diagnosing chlamydialinfections raises several difficult issues.1 Thecost of changing to such methods issubstantial: we estimate that an additional£5m a year would be required for thecurrent workload in genitourinary medicine.In addition, the substantial costs of training,additional space, and gaining formalaccreditation would have to be met beforethe methods could be widely applied. Manylaboratories are finding it hard to accept thehigh costs of amplification tests, even inaffluent California.2

Any expenditure of this magnitude mustbe based on the best evidence. A recentFrench multicentre survey of 28 commercialkits for the diagnosis of chlamydial infectionfound that enzyme linked immunoassaydetection limits varied over a 10-fold range;only two kits attained the same values asdirect immunofluorescence assays.3 A previ-ous study in Bristol also found variation inperformance.4 Clearly, the most sensitiveenzyme linked immunoassays must be usedin trials to establish the true difference insensitivities. In some reports the differencein sensitivity observed has been much lessthan that quoted by the authors when themost sensitive enzyme linked immunoassaysare used. A recent example is the paper byTanaka et al on detection in vaginal and cer-vical swabs; in both cases the best enzymelinked immunoassay gave sensitivities closeto those of the polymerase chain reaction.5

We are saddened by the suggestion thatwomen who have had a diagnosis by enzymelinked immunoassay might pursue litiga-tion; given all the imponderables of chlamy-dial infection, we are not convinced that theywould be successful. No current test has100% sensitivity. Would a woman who had adiagnosis by nucleic acid amplification assay

be in a position to sue if the laboratory hadnot checked her specimen for inhibitors?

We understand the authors’ desire tooffer patients the best test but are concernedthat the high costs of amplification assayswill lead to restrictions on testing. Thiswould be disastrous when we could bewidening chlamydial screening to reach thesubstantial number of infected women whocurrently receive no service at all.Alan Herring Head, PHLS Genitourinary InfectionsReference LaboratoryA.Herring@PHLSBRISTOL.btinternet.com

Owen Caul Consultant virologistIan Paul Medical laboratory scientific officerBristol Public Health Laboratory, Bristol BS2 8EL

Patrick Horner Consultant physicianMilne Centre, Bristol Royal Infirmary, BristolBS2 8HW

1 Taylor-Robinson D, Robinson AJ. DNA methods should beused to detect Chlamydia trachomatis. BMJ 1998;317:1525. (28 November.)

2 Dean D, Ferrero D, McCarthy M. Comparison of perform-ance and cost-effectiveness of direct fluorescent-antibody,ligase chain reaction, and PCR assays for verification ofchlamydial enzyme immunoassay results for populationswith a low to moderate prevalence of Chlamydiatrachomatis infection. J Clin Microbiol 1998;36:94-9.

3 Biachi A, De Barbeyrac B, Bebear C, Buffet-Janvresse C, EbF, Janot C, et al. Multi-laboratory comparison of 28commercially available Chlamydia trachomatis diagnostictests. In: Stephens RS, Byrne GI, Chrisiansen G, Clarke IN,Grayson JT, Rank RG, et al, eds. Chlamydial infections.Proceedings of the ninth international symposium on humanchlamydial infection. San Francisco: International Chlamy-dia Symposium, 1998. (ISBN 0-9664383-0-2.)

4 Paul ID, Caul EO. Evaluation of three Chlamydiatrachomatis immunoassays with an unbiased, non-invasiveclinical sample. J Clin Microbiol 1990;28:220-2.

5 Tanaka M, Nakayama H, Yoshida H, Takahashi K, NagafujiT, Hagiwara T, et al. Detection of Chlamydia trachomatis invaginal specimens from female commercial sex workersusing a new improved immunoassay. Sex Transm Infect1998;74:435-8.

Audit of implantablecardioverter-defibrillators in asingle UK centreEditor—We support Causer and Connolly’seditorial emphasising the value of implant-able cardioverter-defibrillators in patientswith life threatening ventricular arrhythmias.1

A recent audit conducted in our cardiac unithighlights two additional issues: psychosocialconcerns of patients and cost implications.We reviewed clinical records and psycho-social structured questionnaires of the 42patients receiving their first implantablecardioverter-defibrillator (through pectoralroutes) between January 1995 and February1998 (table). To our knowledge, this is the firstreported series of such patients in the UnitedKingdom.

At audit five patients had died (threefrom progressive heart failure, two fromcancer), three of them within six months ofimplantation. Twenty five patients receivedappropriate treatment and thereby survivedan otherwise potentially fatal arrhythmia.

Sixteen patients (38%) reported somedegree of anxiety or depression after implan-tation of the cardioverter-defibrillator, apercentage similar to that in a larger series ofsuch patients in the United States.2 Most ofthese patients had received multiple shocksfrom the device. Twenty six patients wereunable to drive, and only 12 were in full time

employment. Overall, however, most patientsreported a positive attitude to health and hadan improved quality of life score afterimplantation.

The total cost of implantation in the 42patients was about £850 000 (not includingthe costs of hospital readmission and followup care). In comparison with proceduressuch as renal dialysis or coronary arterybypass surgery, implantation of cardioverter-defibrillators is comparatively cost effectivewhen used in selected high risk groups.3

However, to base practice in the UnitedKingdom on evidence from recent trials4

would require at least a threefold increase inimplantation rate (to about 1000 patients/year), with a projected total cost to the NHSof £24.1m/year.5

Patients surviving a first cardiac arrestrelated to sustained ventricular tachycardiaor ventricular fibrillation are at high risk of afurther, potentially fatal, event.1 An implant-able cardioverter-defibrillator can be life

Summary of audit in patients receiving implantablecardioverter-defibrillators. Values are numbers ofpatients unless stated otherwise

Deviceactivated(n=25)

Device notactivated(n=17)

Men, women 21, 4 12, 5

Age (years):

Range 31-79 32-74

Mean (SD) 62 (11) 53 (13)

Aetiology:

Ischaemic heart disease 21 12

Coronary artery spasm 0 1

Hypertrophic cardiomyopathy 1 4

Right ventricular dysplasia 3 0

Rhythm:

Ventricular tachycardia 7 5

Ventricular defibrillation 18 12

Ejection fraction (%):

Range 15-70 20-80

Mean 37 46

Local complications:

Related to the lead 2 0

Haematoma 3 3

Pneumothorax 0 1

No of shocks delivered during follow up (antitachycardiapacing,* defibrillation):

1 6, 5 0

2-5 6, 5 0

6-9 2, 2 0

>10 5, 4 0

Shocks deliveredinappropriately†

6 0

Regular amiodarone treatment 19 12

Readmitted for heart failure ormultiple shocks

10 0

Quality of life score (maximum 10):

Before implantation 5.6 7.2

After implantation 8.3 8.4

Able to drive 4 12

Able to work 5 7

Anxiety present 8 3

Depression present 4 1

Regular antidepressanttreatment

3 0

*For ventricular tachycardia.†During sinus rhythm or supraventricular tachycardia.

Letters

1416 BMJ VOLUME 318 22 MAY 1999 www.bmj.com

saving in such people. Psychosocial factors,before and after implantation, need carefulevaluation (and appropriate further man-agement) and are as important as thecardiological and cost aspects of treating themalignant arrhythmias. We hope that theresults of a study based in the United King-dom may provide further insight and finallyhelp to identify the subgroup of patientsmost likely to benefit from having animplantable cardioverter-defibrillator.Sandeep Gupta Specialist registrarHasan Hasan Senior house officerOswaldo Valencia Research fellowSue Jones Senior pacing coordinatorEdward Rowland Consultant cardiologistA John Camm Professor of cardiologyCharles W Pumphrey Consultant cardiologistDavid E Ward Consultant cardiologistDepartment of Cardiology, St George’s Hospital,London SW17 OQT

1 Causer JP, Connolly DT. Implantable defibrillators forlife threatening arrhythmias. BMJ 1998;317:762-3.(19 September.)

2 Heller SS, Ormont MA, Lidagoster L, Sciacca RR,Steinberg JS. Psychosocial outcome after ICD implanta-tion: a current perspective. PACE 1998;21:1207-15.

3 Kupperman M, Luce BR, McGovern B, Podrid PJ, BiggerJT, Ruskin JN. An analysis of the cost-effectiveness of theimplantable-defibrillator. Circulation 1990;81:91-100.

4 Antiarrhythmics Versus Implantable Defibrillators (AVID)Investigators. A comparison of antiarrhythmic drug therpywith implantable defibrillators in patients resuscitatedfrom near fatal ventricular arrhythmias. N Engl J Med 1997;337:1576-83.

5 Pathmanathan RK, Lau EW, Cooper J, Newton L, SkehanJD, Garratt CJ, Griffith MJ. Potential impact of antiarrhyth-mic drugs versus implantable defibrillators on themanagement of ventricular arrhythmias: the Midlands trialof empirical amiodarone versus electrophysiologicallyguided intervention and cardioverter implant registrydata. Heart 1998;80:68-70.

Vitamin D concentrations inAsian children living inEngland

Limited vitamin D intake and use ofsunscreens may lead to rickets

Editor—I would like to add a sixthsuggestion to Wharton’s five approaches toimproving children’s vitamin D status toprevent rickets.1 A case of rickets wasrecently diagnosed in a 12 month old whiteinfant at the Hospital for Sick Children inToronto. He had clinical, radiological, andbiochemical evidence of rickets, whichresponded to vitamin D treatment. VitaminD intake seemed adequate until six monthsafter treatment began but was sporadic afterthat. Although the infant spent timeoutdoors in the summer, his skin wasprotected with potent sunscreens (SPF 30).

Rickets presumably developed becauseof limited vitamin D intake combined withdiminished skin penetration of ultravioletradiation. I believe that this is the first case tobe diagnosed in a white infant. Althoughnumerous previous reports have docu-mented vitamin D deficiency rickets frominadequate oral intake of vitamin D, in mostcases the infants had dark skin and had lim-ited exposure to ultraviolet B.2

The ultraviolet B rays that synthesise vita-min D3 in the skin from 7-dehydrocholesterolare also associated with adverse effects on theskin. Public health advice has emphasised

that skin should be shielded from the sunfrom birth onwards, especially in lightskinned individuals. Both Health Canada andthe Canadian Dermatology Associationadvise that infants under 1 year should bekept out of direct sunlight. Infants over 6months are advised to wear a sunscreen ofSPF 15 or higher on all areas that areexposed to the sun.3

With the increasing use of sunscreenone may justifiably ask whether infants andchildren are at an increased risk of vitamin Ddeficiency. This question is biologically plau-sible since sunscreens block the absorptionof ultraviolet B radiation. A sunscreen withan SPF of 30, like the one used in this case,allows only a thirtieth of the ultraviolet raysto penetrate the skin. The photoisomerisa-tion of 7-dehydrocholesterol to previtaminD3 is prevented by sunscreens.4

Although regular use of sunscreens hasbeen associated with low body stores of vita-min D,5 the case presented is the first todocument an association between use ofsunscreens and vitamin D deficiency rickets.Infants who ingest a suboptimal amount ofvitamin D and whose skin is protected bypotent sunscreens should receive a vitaminD supplement.Stanley Zlotkin ProfessorDepartment of Paediatrics and of NutritionalSciences, Hospital for Sick Children, Faculty ofMedicine, University of Toronto, 555 UniversityAvenue, Toronto, Ontario, Canada M5G 1X8szlotkin@sickkids.on.ca

Competing interests: Stanley Zlotkin occasionallygives opinions, for which he receives a fee, to compa-nies making nutritional products for children.

1 Wharton BA. Low plasma vitamin D in Asian toddlers inBritain. BMJ 1999;318:2-3. (2 January.)

2 Mughal MZ, Salama H, Greenaway T, Laing I, Mawer EB.Florid rickets associated with prolonged breast feedingwithout vitamin D supplementation. BMJ 1999;318:39-40.(2 January.)

3 Health and Welfare Canada. The sun, your baby and you: aparent’s guide to sun protection. Ottawa: Canadian Govern-ment Publishing Centre, Supply and Services, 1996.

4 Matsuoka LY, Ide L, Wortsman J, MacLaughlin JA, HolickMF. Sunscreens suppress cutaneous vitamin D3 synthesis. JClin Endocrinol Metab 1987;64:1165-8.

5 Matsuoka LY, Wortsman J, Hanifan N, Holick MF. Chronicsunscreen use decreases circulating concentrations of25(OH)D. Arch Dermatol 1988;124:1802-4.

Concentrations found may be function ofanalytical methodology used

Editor—Lawson and Thomas suggestthat the serum concentration of 25-hydroxycholecalciferol is substantially lowerin Asian infants than in the general popula-tion.1 Although this finding is highly plausi-ble, their paper does not provide reliableevidence regarding the presence or magni-tude of this difference. Plasma samples inAsian subjects were collected in a differentyear and in a different geographical areaand were apparently assayed in a differentlaboratory from those in the externalcontrol group. Analytical methods areunspecified.

Year to year differences in plasma25-hydroxycholecalciferol are likely. Moreimportantly, the North West Thames externalquality assurance survey showed that serummeasurements of 25-hydroxycholecalciferoldiffered considerably both between and

within laboratories over the period of thestudy.2 3 The number of subjects classified ashaving 25-hydroxycholecalciferol below 25ìmol/l may be as much a function of analyti-cal methodology as of ethnic group. A similarexplanation may account for the apparentabsence of seasonal variation in the controldata.Aubrey Blumsohn Senior registrar in biochemicalmedicineDirectorate of Biochemical Medicine, NinewellsHospital, Dundee DD1 9SYablumsohn@btinternet.com

Competing interests: None declared.

1 Lawson M, Thomas M. Vitamin D concentrations in Asianchildren aged 2 years living in England: population survey.BMJ 1999;318:28. (2 January.)

2 Carter GD, Hewitt J. The 25-hydroxyvitamin D EQAS: anupdate. Proc UK NEQAS Meeting 1996;2:157. (ISSN 1357-4558.)

3 Veith R. Vitamin D supplementation, 25-hydroxyvitaminD concentrations, and safety. Am J Clin Nutr 1999;69:842-56.

Income distribution,socioeconomic status, andself rated health in US

Authors ignored data in their study

Editor—I take issue with Kennedy et al’sconclusion that the frequently observedassociation between income inequality andhealth is not an artefact of an omitted relationbetween individual income and individualhealth.1 They had only limited informationon individual incomes in their study. Further-more, they ignored some of the income andinequality data that they did have available.Consequently, their conclusion that incomeinequality exerts an independent effect onindividual health is inappropriate.

The authors controlled for individualincome by using survey data on householdincome and controlling for household size.Their income data were categorical and sowere necessarily measured with error. Thetop category was for households withincomes of >$50 000 (17.3% of the sample).Without explanation the authors combinedthis category with the category of house-holds with an income of $35 000 to< $50 000 (another 15% of the sample). Theauthors also retained observations onpeople for whom income data were missing(12.6% of the sample). Thus 44.9% of theirsample had missing data or the highestincomes in the study.

The authors also analysed subsamples oftheir data stratified by income. However,they combined several income categories ineach subsample and failed to control forthis. Finally, they ignored additional infor-mation by using only a four category meas-ure of income inequality even though theyhad available a continuous measure of it (theGini coefficient).

Despite these measurement issues, theauthors show that the relative risk associatedwith residence in a state with high incomeinequality falls by about one third whencontrols for individual income are added tothe logistic regression. This suggests that if

Letters

1417BMJ VOLUME 318 22 MAY 1999 www.bmj.com

the authors had more complete data onindividual incomes the association betweeninequality and health might be further miti-gated; the apparent association might evendisappear completely, which is the oppositeof what the authors conclude from theiranalysis.

By choosing to ignore relevant infor-mation Kennedy et al have added to theambiguity in their findings and unnecessar-ily undermined the value of their analysis.Jeffrey Milyo Postdoctoral fellowHealth Policy Program, Yale University, New Haven,CT 06511, USAjmilyo@tufts.edu

1 Kennedy BP, Kawachi I, Glass R, Prothrow-Stith D. Incomedistribution, socioeconomic status, and self-rated health inthe United States: multilevel analysis. BMJ 1998;317:917-21. (3 October.)

Authors’ reply

Editor—Milyo’s main criticism is that wedid not adequately address the statisticalartefact issue as our estimates of the effectsof individual income on health used house-hold income categories rather than acontinuous measure of household income.Our study was limited to the income dataavailable to us. Measurement of householdincome in a telephone survey requires a bal-ance between precision and loss of infor-mation to non-response. It is now standardsurvey practice to ask for income bycategory to minimise non-response, as theloss of precision is not considered to haveany substantive impact on estimates. We didnot ignore information; it was simplyunavailable.

We acknowledge that using categoricalinstead of continuous individual householdincome information may have hindered ourability to rule out completely the residualconfounding effect of individual householdincome on health. This, though, is unlikelyto have affected our results.

Milyo expresses concerns about ourhaving collapsed the top two incomecategories and included cases for whichincome data were missing, suggesting againthat we ignored these issues. Although wedid not clarify this in the paper, we includedthe observations of people who refused toanswer the income question to avoid losingthe information that this group couldcontribute to the estimates of the odds ratiosfor the demographic, risk factor, and house-hold composition control variables. Therelation between self reported health andhousehold income flattened out afterr35 000, and for ease of interpretation wecombined the r35 000 to < r50 and>r50 000 income categories as they did notdiffer substantively. We made these decisionsafter checking whether dropping peoplewith missing income and combining the toptwo income categories affected the results;they did not.

A final concern of Milyo is that use of theGini coefficient as a categorical rather than acontinuous variable ignores importantinformation. After determining that the lossof information made little difference to our

conclusion we presented our findings interms of Gini categories rather than in termsof a continuous Gini measure to make theresults understandable to a broad audience.For most readers, the change in an oddsratio associated with a 1 unit change in theGini coefficient has no intuitive meaning.

We do not see a need to qualify ouroriginal conclusion. The relation betweenincome inequality and self reported health isrobust. Even after we controlled for factorssuch as individual household income(though not as perfectly measured as mightbe), race, smoking, obesity, and education—all of which may be in the causal pathwaybetween inequality and poor health—therelation remained significant.Bruce P Kennedy Deputy directorIchiro Kawachi Associate professorRoberta Glass Research specialistDeborah Prothrow-Stith ProfessorHarvard School of Public Health, Boston,MA 02115, USA

Prediction of cardiovascularrisk

Program is not suitable for diabeticpatients

Editor—Hingorani and Vallance describe aclinic based computer program for themanagement of cardiovascular risk factors.1

However, the program has limitations fordiabetic patients.

This program and previous methods,together with the recent joint British recom-mendations,2 are based on algorithmsderived from the Framingham heart studyof 5573 people aged 30-74 years originallyscreened in 1968. However, only 4% (237subjects) of this cohort had diabetes, asdefined by a random glucose concentration> 9 mmol/l or the use of diabetic treatment.Importantly, diabetes was not classifiedaccording to type, and the presence ofproteinuria or microalbuminuria (impor-tant cardiovascular risk factors in diabetes)were not included.

The prevalence and characteristics ofother risk factors further highlight that peo-ple with diabetes are not average patients3;“diabetic dyslipidaemia,” increased preva-lence of hypertension, and ethnic origin (forexample, south Asian) contribute to anincreased atherogenic potential. Despite thestrong link between triglyceride concentra-tion and coronary heart disease in diabetes,this important risk factor was omitted fromthe program.

The glycaemic state was included in theprogram as a risk factor both as “diabetes”and as a blood glucose concentration. Theimpact of glycaemia on cardiovascularevents is not dichotomous, but, like bloodpressure, has a continuous relation.4 Theinclusion of blood glucose as a variable is animportant addition to risk assessment.

The present program and other meth-ods derived from Framingham data haveinconsistencies for the primary preventionof cardiovascular disease in diabetes. This is

especially true for patients with type 2diabetes, who have a 2 to 5-fold increase inage adjusted mortality from cardiovasculardisease. The emerging evidence is thatpatients with diabetes should be treated asrecommended for secondary prevention.5

Andrew Zambanini Senior registrar in clinicalpharmacologyMartin R Smith Specialist registrar in diabetes andendocrinologyMichael D Feher Senior lecturer in clinicalpharmacologySection of Clinical Pharmacology, Imperial CollegeSchool of Medicine and Beta Cell Diabetes Centre,Chelsea and Westminster Hospital, LondonSW10 9NHrnkk003@cxwms.ac.uk

1 Hingorani A, Vallance P. A simple computer program forguiding management of cardiovascular risk factors andprescribing. BMJ 1998;318:101-5. (9 January.)

2 Wood D, Durrington P, Poulter N, McInnes G, Rees A,Wray R. Joint British recommendations on prevention ofcoronary heart disease in clinical practice. Heart 1998;80(suppl 2): S4-5.

3 Feher MD, Elkeles RS. Lipid modification and coronaryheart disease in type 2 diabetes: different from the generalpopulation? Heart 1999;81:10-1.

4 Gerstein HC, Yusuf S. Dysglycaemia and risk of cardio-vascular disease. Lancet 1996;347:949-50.

5 Haffner SM, Lehto S, Ronnemaa T, Pyorala K, Markku L.Mortality from coronary heart disease in subjects with type2 diabetes and in nondiabetic subjects with and withoutprior myocardial infarction. N Engl J Med 1998;339:229-34.

Hypothesis of program is flawed

Editor—Hingorani and Vallance advocateusing a computer program to predict thecardiovascular risk of individuals and esti-mate the benefit expected from treatment.1

They assume that the risk ratio correspond-ing to a decrease of a risk factor in a givenindividual is the same as the risk ratiobetween two people with the same differ-ence in risk factors.We call this assumptionthe isotropy hypothesis.2 The isotropyhypothesis is unlikely to be true for severalreasons. Firstly, for a given individual, thereduction in risk will not be proportional tothe reduction in blood pressure for therange of blood pressures that treatmentcould produce. In other words, we do notbelieve that a person with an initial systolicblood pressure of 170 mm Hg would obtaina relative risk reduction of stroke of about50% if his blood pressure reached 155 mmHg with treatment, about 75% if it fell to 140mm Hg, and 88% if it fell to 115 mm Hg.Secondly, for coronary events, randomisedcontrolled trials have reported differencesbetween relative risks observed andexpected from a same blood pressurereduction,3 which raises doubts about isot-ropy. Thirdly, a recent trial showed thatusing drugs to reduce blood pressurefurther than with conventional treatmenthas no effect on cardiovascular risk.4 Webelieve this approach may be dangerous andmisleading. For example, it is simple to showa smoker what his risk would have been if hedid not smoke However, this lower risk maynot be the value he would achieve by givingup smoking, and is certainly not what adoctor can expect, on average, frominterventions against smoking.

Hingorani and Vallance’s proposal todetermine patients’ targets for blood pres-sure or cholesterol concentration based on

Letters

1418 BMJ VOLUME 318 22 MAY 1999 www.bmj.com

the desired risk reduction may lead to un-realistic expectations and in some cases dan-gerous behaviour by prescribers or patients.

We favour the approach proposed byJackson.5 A patient’s absolute risk reductionis derived by multiplying the predicted riskby the relative risk reduction estimated froma systematic overview. This estimate ofrelative risk reduction could be modulatedaccording to the results of a search formodifiers of treatment effect, ideally per-formed on an individual patient databasefrom all available clinical trials. Such modifi-ers may include the intensity of reduction ofthe risk factor.François Gueyffier Research fellowJean-Pierre Boissel HeadClinical Pharmacology Unit, EA 643,Claude Bernard University, BP 3041, 69394 LyonsCedex 03, France

1 Hingorani A, Vallance P. A simple computer program forguiding management of cardiovascular risk factors andprescribing. BMJ 1998;318:101-5. (9 January.)

2 Boissel JP, Gueyffier F. Is the isotropy hypothesis true inhypertension? Findings from the INDANA data base.Controlled Clin Trials 1998;19(suppl 3S):85-6S.

3 Colins R, Peto R, MacMahon SW, Herbert P, Fiebach N,Eberlein K, et al. Blood pressure, stroke, and coronaryheart disease. 2. Lancet 1990;335:827-38.

4 Hansson I, Zanchetti A, Carruthers SG, Dahlöf B, ElmfeldtD, Julius S, et al. Effects of intensive blood pressure lower-ing and low dose aspirin in patients with hypertension:principal results of the hypertension optimal treatment(HOT) randomised trial. Lancet 1998;351:1755-62.

5 Jackson R, Barham P, Bills J, Birch T, McLennon L,MacMahon S, et al. The management of raised blood pres-sure in New Zealand. BMJ 1993;307:107-10.

Authors’ reply

Editor—Any program designed to calculatecardiovascular risk is inevitably limited bythe original data gathering process used toderive the risk equations. Despite Zambaniniet al’s concerns about the limited number ofdiabetic patients included in the originalFramingham cohort, the Framingham riskequation is proving to be a robust tool forestimating risk in the United Kingdom.1

One advantage of a computer basedapproach is that as new risk factors are iden-tified and estimates of risk further refined,the appropriate changes can be included inthe predictions. We hope that further dataon both diabetic and non-diabetic patientswill be included in the program as theybecome available. All risk calculations onpatients stored in the database will then berefined automatically.

We also considered it important todesign a system which not only provided anassessment of risk but also indicated thelikely maximum benefits of intervention.Gueyffier and Boissel raise doubts about thevalidity of such calculations, but ourcomparison of computer predictions withdata from the West of Scotland coronaryprevention study and recently publishedmeta-analyses2 support the view that, forcholesterol lowering at least, individualsadopt a new level of risk predicted from theFramingham cohort study. As we explain inour paper, this is also likely to be true forstopping smoking, although the effects ofblood pressure reduction on risk of coron-ary heart disease may be overestimated.Gueyffier and Boissel’s proposal that thebenefits of risk factor intervention be

estimated from a systematic overview ofintervention trials is a reasonable alterna-tive, and our program could be adapted toperform such a task. Either way, thepredictions for blood pressure interventionwould be tested against trial data.

In designing this program, our aim wasto take forward theoretical discussions aboutmultiple risk factors, risk calculations, andthe implementation of thresholds for statinand antihypertensive treatment. In particu-lar, we felt there was a need to develop a toolwhich would aid the practising doctor and“operationalise” the extensive available evi-dence on risk factors and intervention.3 Thelevel of interest which our paper has gener-ated, particularly inquiries about the avail-ability of the program, seems to endorse ourview that there is a great clinical need for aprogram such as ours.Aroon D Hingorani British Heart Foundationintermediate fellowPatrick Vallance ProfessorCentre for Clinical Pharmacology, Department ofMedicine, University College London, RayneInstitute, London WC1E 6JJ

1 Haq IU, Ramsay LE, Yeo WW, Jackson PR, Wallis EJ. Is theFramingham risk function valid for Northern Europeanpopulations? A comparison of methods for estimatingabsolute coronary risk in high risk men. Heart 1999;81:40-6.

2 Gould AL, Rossouw JE, Santanello NC, Heyse JF, FurbergCD. Cholesterol reduction yields clinical benefit. Impact ofstatin trials. Circulation 1998;97:946-52.

3 Grover S. Gambling with cardiovascular risk: picking thewinners and the losers. Lancet 1999;353:254-5.

Secondary prevention incoronary heart disease

Effects of statins have been in addition tothose of aspirin and â blockers

Editor—Ferner acknowledges the apparentsimilarity between the abilities of continuingaspirin and simvastatin to prevent “badthings.”1 He overlooks the fact that the posi-tive outcomes of the secondary preventionstatin studies were additional to the use ofboth aspirin and â blockers according to theinvestigators’ cardiological practices at thetime (table).2–4 Since the inception of theScandinavian simvastatin survival study overa decade ago, subsequent trials have testedlipid modification with statins against abackground of increasingly optimal prac-tice, comfortably exceeding that described inthe survey by Campbell et al.5

Ferner’s argument is focused on costeffectiveness. The cost implications ofaspirin, however, are negligible—indeed,many patients buy it themselves. The costeffectiveness of statins in secondary preven-

tion is established, equates to that of otheraccepted healthcare interventions, and willimprove further as patents expire.

The benefits of aspirin and â blockersare beyond doubt, but Ferner’s messageserves to inhibit further the implementationof treatment with statins in secondaryprevention and the potential for furtherreductions in mortality and morbidity.Surely it is better to offer patients requiringsecondary prevention the complete range oflifestyle and therapeutic options and, forcost savings, to concentrate on areas of lesssubstantiated prescribing.Jonathan Morrell Committee member, health caresection of British Hyperlipidaemia AssociationFitznells Manor Surgery, Ewell, Surrey KT17 1TF

Competing interests: Dr Morrell has received feesfrom pharmaceutical companies marketing lipidlowering drugs for speaking at and organising edu-cational meetings.

1 Ferner RF. Secondary prevention in coronary heartdisease. BMJ 1998;317:1592. (5 December.)

2 Scandinavian Simvastatin Survival Study Group. Ran-domised trial of cholesterol lowering in 4444 patients withcoronary heart disease: the Scandinavian simvastatinsurvival study (4S). Lancet 1994;344:1383-9.

3 Sacks FM, Pfeffer MA, Moye LA, Rouleau JL, RutherfordJD, Cole TG, et al. The effect of pravastatin on coronaryevents after myocardial infarction in patients with averagecholesterol levels. N Engl J Med 1996;335:1001-9.

4 Long-term Intervention with Pravastatin in Ischaemic Dis-ease (LIPID) Study Group. Prevention of cardiovascularevents and death with pravastatin in patients with coronaryheart disease and a broad range of initial cholesterol levels.N Engl J Med 1998;339:1349-57.

5 Campbell NC, Thain J, Deans HG, Ritchie LD, Rawles JM,Squair JL. Secondary prevention in coronary heartdisease: baseline survey of provision in general practice.BMJ 1998;316:1430-4.

Author’s reply

Editor—The proportion of patients receiv-ing aspirin in the trials to which Morrellrefers is encouraging. Evidence from Camp-bell et al’s study suggests that many morepatients could benefit from aspirin thanreceive it.1 Even in the trials, few patientswere treated with â adrenoceptor antago-nists, which are underused.2 An importantsecondary prevention study from South-ampton, which quoted high rates of use ofaspirin, did not provide information about âblockers.3

Chen et al highlighted the differencebetween America’s “best hospitals” and therest by examining the outcomes for nearly150 000 elderly patients who had had heartattacks.4 One patient in 25 admitted to thebest hospitals survived who would have diedin the other hospitals. The authors ascribedthis to two straightforward facts: 15% morepatients were given aspirin and over 25%more patients were given â blockers in thebest hospitals.

Morrell rightly points out that thebenefits of aspirin and statins may beadditive. However, the benefits of statins,even in this high risk population, fall shortof miraculous. In one study (the long termintervention with pravastatin in ischaemicdisease study) the absolute benefit of pravas-tatin 40 mg daily was to save 1.9 lives per610 patient years of treatment.5 That is onelife saved for every 320 patient years oftreatment, at a direct drug cost of nearly

Percentages of patients in each study who weretaking aspirin or â blockers

Study Treatment

% taking:

Aspirin â Blockers

4S2 Simvastatin 37 57

CARE3 Pravastatin 83 40

LIPID4 Pravastatin 82 47

Campbell et al5 63 32

Letters

1419BMJ VOLUME 318 22 MAY 1999 www.bmj.com

£180 000—over 10 years’ pay for an averagewage earner.

When all patients who might benefit aretaking aspirin and â blockers, which costlittle, we can worry more about theexpensive treatments, whose ability toreduce relative risk is heavily promoted butwhose absolute benefit is quite small.R E Ferner DirectorWest Midlands Centre for Adverse Drug ReactionReporting, City Hospital, Birmingham B18 7QHr.e.ferner@bham.ac.uk

Competing interests: Dr Ferner advises BirminghamHealth Authority on prescribing matters.

1 Campbell NC, Thain J, Deans HG, Ritchie LD, Rawles JM,Squair JL. Secondary prevention in coronary heartdisease: baseline survey of provision in general practice.BMJ 1998;316:1430-4.

2 Gottlieb SS, McCarter RJ, Vogel RA. Effect of beta-blockade on mortality among high-risk and low-riskpatients after myocardial infarction. N Engl J Med1998;339:489-97.

3 Jolly K, Bradley F, Sharp S, Smith H, Thompson S,Kinmonth A-L, et al. Randomised controlled trial of followup care in general practice of patients with myocardial inf-arction and angina: final results of the Southampton heartintegrated care project (SHIP). BMJ 1999;318:706-11.(13 March.)

4 Chen J, Radford MJ, Wang Y, Marciniak TA, Krumholz HM.Do “America’s best hospitals” perform better for acute myo-cardial infarction? N Engl J Med 1999;340:286-92.

5 Long-term Intervention with Pravastatin in Ischaemic Dis-ease (LIPID) Study Group. Prevention of cardiovascularevents and death with pravastatin in patients with coronaryheart disease and a broad range of initial cholesterol levels.N Engl J Med 1998;339:1349-57.

Minimisation is much betterthan the randomised blockdesign in certain casesEditor—In commenting on our editorial1

Ross suggests that the randomised blockdesign is similar to minimisation but haseven more power.2 We are unaware of anypublication which shows that this is the case.He also states that these two methods havethe same disadvantage—that assignment to ablock becomes a major undertaking. In fact,minimisation does not have the problem ofassignment to a block that the randomisedblock design has, and this is precisely whyminimisation was invented.

In the context of clinical trials therandomised block design is referred to asstratified randomisation—for example, themen and women each have their ownrandom allocation series; if stratified ran-domisation is feasible it is indeed anexcellent method for obtaining balancedtreatment groups. However, stratified alloca-tion becomes unwieldy and eventuallyimpossible as the number of relevant patientcharacteristics increases. Ross mentions age,sex, or number of pack years smoked aspatient characteristics that could be used instratified allocation, but he does not considerwhat would happen if one wished to balancethe treatment assignment for all three ofthese characteristics.

Suppose we have three age groups andfour groupings of pack years smoked, plus,of course, two sexes. This gives 24 strata, andthe randomised block design would there-fore have to have 24 separate randomallocation series. This might just be manage-able in a large trial, but many trials have

considerably more than three patient char-acteristics that relate to prognosis. Forexample, Lee et al compared their patientgroups at baseline on 14 “selected” dichot-omised characteristics, which in combina-tion give 214 ( = 16 384) strata.3 Minimisationcould handle treatment allocation in such asituation with ease, but the randomisedblock design would be in some difficulty.Tom Treasure Professor of cardiothoracic surgerySt George’s Hospital Medical School, LondonSW17 0QT

Kenneth D MacRae Medical statistician5 Northcroft Terrace, London W13 9SP

1 Treasure T, MacRae KD. Minimisation: the platinum stand-ard for trials? BMJ 1998;317:362-3.

2 Ross N. Randomised block design is more powerful thanminimisation. BMJ 1999;318:263. (23 January.)

3 Lee KL, McNeer F, Starmer F, Harris PJ, Rosati RA. Clini-cal judgment and statistics: lessons from a simulated rand-omized trial in coronary artery disease. Circulation1980;61:508-15.

Is there a clinical future for thegeneral practitioner?Editor—At the recent “Clinical Futures”conference, and in the book,1 a subthemeemerged about the perceived redundancy ofthe general practitioner. Diagnosis would begenotypic rather than phenotypic, treatmentin the primary care setting would mostly begiven by nurses or paramedics, and anythingcomplicated would require secondary care,which could be accessed directly.

It is already clear that a large proportionof current general practitioners’ workloadcould be carried out by nurses (but this isnothing new—Tudor Hart pointed it out in19852). Moreover the “corner shop” mental-ity of general practitioners, their central rolein the primary healthcare team, and theirstatus as employers (and often owners of thepremises) is anachronistic and unsustain-able. So the 1972 model, Future GeneralPractitioner,3 has reached its sell-by date, andwe must design a “new model.” New modelgeneral practitioners will be employees oflocality sized primary care trusts, and therewill be fewer of them. Branch surgeries(which often will be located in shoppingcentres) will be staffed entirely by staff suchas nurse practitioners, with networked,district-wide electronic records and tele-medicine links to provide medical back upand access for patients.

It is unrealistic, however, to suppose thatpatients will no longer require personaladvice from a doctor in primary care. On thecontrary, the exponential growth of infor-mation and technical capability will needexpert management to ensure minimisationof harm to, and maximisation of clinicaleffectiveness for, individual patients. Generalpractitioners will have a key role in helpingpatients to make complex decisions aboutdiagnosis and treatment. They will still needwell developed consultation and other clini-cal skills, but in addition they must beexperts in evidence based medicine andinformation technology. Thus they will beable to help patients to identify and

prioritise their problems; ask structured,answerable questions based on those prob-lems; search for, find, and appraise infor-mation to answer these questions; anddiscuss the findings and implications in ameaningful and useful way with patients.They will also need to be trained aseducators and managers, in order tomaximise the clinical and organisationaleffectiveness of their teams.

These skills are neither trivial nordispensable. Far from becoming redundant,general practitioners, as highly skilled medi-cal generalists and information specialists,will be perhaps the only essential doctors inthe whole of healthcare practice. Everythingelse will be done by nurses, technicians, androbots.Toby Lipman General practitioner and Northern andYorkshire research training fellowWesterhope Medical Group, Westerhope, Newcastleupon Tyne NE5 2LHtoby@tobylipm.demon.co.uk

1 Marinker M, Peckham M, eds. Clinical futures. London: BMJBooks, 1998.

2 Hart JT. The world turned upside down: proposals forcommunity-based undergraduate medical training(George Swift Lecture 1985). J R Coll Gen Pract 1985;35:63-8.

3 Royal College of General Practitioners. The future generalpractitioner: learning and teaching. London: RCGP, 1972.

Role of general practitioners inNHS must not be undervalued

Author underestimated figures

Editor—We should be grateful to Majeedfor his re-estimation of the proportion ofmedical care that takes place in generalpractice.1 His calculation that the proportionis about 70% is, however, too low, as he hasextracted the wrong figure from the fourthnational survey of morbidity statistics fromgeneral practice. The stated figure of 2.9consultations per person year is the doctorcontact rate. The calculated consultationrate, which allows for the fact that there maybe consultations for a range of separate con-ditions during the same visit, was 3.5 perperson year, or 3.6 per person year allowingfor underrecording in the national survey.2

For an annual estimate we must startwith the total English population in 1991minus prisoners and service personnel—48million. This should be inflated by 3.5% forduplicate registrations (in the fourthnational survey of morbidity in generalpractice) and then multiplied by 3.6 to give aprojection of the total number of face to faceconsultations; this can be further increasedby 8% to allow for telephone consultations.3

The resulting total is 193 million, or 76% ofNHS medical care.

A better comparison would be to look atestimates of complete episodes of care. Herethe estimated general practice rate from thefourth national survey of morbidity ingeneral practice was 2.0 per person year—“New and first ever episodes.”2 Adjusted asbefore, this projects 111 million generalpractice episodes, compared with a hospitalestimated total for 1991-2 of 23.5 million—

Letters

1420 BMJ VOLUME 318 22 MAY 1999 www.bmj.com

3.5 million emergency inpatients,4 11 mil-lion first attenders at accident and emer-gency,5 and 9 million referral outpatientattendances.5 This suggests that the pro-portion of medical episodes that take placein general practice may be close to 83%,which I believe corresponds to Chisholm’sreference to "90% of all episodes of healthcare” which Majeed mentions in the firstparagraph of his letter.Tom Hennell Strategic analystNHS Executive North West, Warrington WA3 7QNthennell@doh.gov.uk

1 Majeed A. Role of hospitals in NHS must not be under-valued. BMJ 1998;317:1653. (12 December.)

2 Royal College of General Practitioners, Office ofPopulation Censuses and Surveys, Department of Health.Morbidity statistics from general practice: fourth national study,1991-92. London: HMSO, 1995.

3 Office of Population Census and Surveys. General householdsurvey 1991. London: HMSO, 1993.

4 Department of Health. Hospital episode statistics 1991-92.London: DoH, 1994.

5 Department of Health. Outpatients and ward attenders forEngland 1991-92. London: DoH, 1993.

Author’s reply

Editor—Hennell tries to shed more light onthe proportion of episodes of medical carethat take place in primary care. I think, how-ever, that his statement that 76-83% ofmedical contacts take place in primary careis based on several invalid assumptions andconsequently is incorrect.

Firstly, although it is true that patientsoften present to general practitioners withmore than one problem at a time, the sameis also true of patients presenting to hospitaldoctors. Secondly, it is inappropriate toinclude telephone consultations in the totalfor general practice contacts without alsoincluding them in the total for hospital con-tacts. Finally, in his calculation of completedepisodes of care Hennell assumes thatpatients attending for repeat outpatientappointments do not present with newproblems or with exacerbations of old prob-lems. This assumption is also incorrect.

Until proved otherwise, the most accu-rate figure for the proportion of totalmedical contacts taking place in primarycare remains the 70% quoted in myprevious letter.Azeem Majeed Senior lecturer in general practiceSchool of Public Policy, University College London,London WC1H 9EZa.majeed@ucl.ac.uk

Noise trauma after inflation ofair bags in low speed carcrashes

Article was unclear and possiblymisleading

Editor—Buckley et al draw attention topossible noise trauma and sensorineuralhearing loss and tinnitus as a result of theinflation of air bags in low speed car crashes.1

Their report is unclear on several issues andmay be misleading.

In neither of the two cases reported ismention made of how soon after the

accident the audiometry was carried out.The accident in case 1 occurred in theUnited States, and audiometry was carriedout in Leeds by one of the authors. There isno record of any hearing assessment in theUnited States, which is surprising given thatthe patient experienced bilateral hearingloss, tinnitus, and unsteadiness (theunsteadiness lasting two weeks). Patients’perception of hearing loss does not implythat such a loss was simply sensorineural;indeed, conductive loss may have occurredand been transient. The small high fre-quency hearing loss depicted on the left sidein case 1 is not one that most patients willnotice and may have been present beforethe accident. There is the added factor of aflight from the United States.

Sensorineural hearing loss has beenreported after minor trauma to the ear2; itusually involves high frequencies and istransient. The hearing loss depicted in case 2mainly involved low frequencies, and read-ers are not told if it was reversible. Thispatient had had high frequency hearing lossnoted 18 months before the accident. It isunclear if the small additional loss in thehigh frequencies was in fact related to thetrauma. This patient was receiving treatmentfor hyperlipidaemia and was 68.

It would be useful to know if the associ-ated tinnitus in both cases was bilateral andwhether, with the undocumented improve-ment in hearing in the left ear in case 1, tin-nitus on that side stopped.S S M Hussain Consultant otologistDirectorate of Otolaryngology, Ninewells Hospitaland Medical School, Dundee DD1 9SYMusheerhussain@btinternet.com

1 Buckley G, Setchfield N, Frampton R. Two case reports ofpossible noise trauma after inflation of air bags in lowspeed car crashes. BMJ 1999;318:499-500. (20 February.)

2 Hussain SSM. Hearing loss in the 4 to 8 kHz range follow-ing tympanic membrane perforation from minor trauma.Clin Otolaryngol 1995;20:211-2.

Author’s reply

Editor—In both cases audiometry wascarried out more than three weeks after theinjury; it was not carried out in the UnitedStates in case 1. The suggestion of atransient conductive hearing loss is purespeculation and is largely irrelevant sincethe audiometrically measured loss is clearlysensorineural. The high frequency loss inthe left ear may have been present beforethe accident, but this is also speculation. Wedisagree that it is a hearing loss that mostpatients would not notice. In a 38 year oldpatient it is significant and probably sympto-matic. The damage to cochlear outer haircells implied by such a loss affects frequencyresolution and speech perception as well assimply volume of sound. Hussain’s pointabout the flight from the United States beingan added factor is a curious one. Flying maycause barotrauma to the middle ear with atemporary conductive hearing loss, but theloss here is sensorineural. Is he suggestingthat it is due to flying?

Hussain’s comments about case 2 repeatthe points we discussed in the article.Deterioration due to age and other factors is

always a possibility, but the deteriorationseen is greater than would be likely in thetime frame. Hussain also seems to miss thepoint in the article that the hearing lossoccurred at both 1 kHz and 4 kHz. Since theaudiogram was obtained three weeks afterthe event the hearing loss is unlikely to be atransient effect. The “small” additional lossat high frequencies amounted to 25 dB at3 kHz and 4 kHz. We suspect that Hussainwould not think this small if it affected him.

The initial tinnitus in the left ear in case1 did indeed settle, but that in the right didnot. The patient in case 2 had bilateral tinni-tus, which was not present before the injury.

One final point should be made. Bothsubjects presented because they noticed apersistent hearing loss and tinnitus afterinflation of the air bags. Other potentialcontributory factors always exist, but the evi-dence in the cases that we reported iscompelling. The important issue is whetherthese cases are isolated occurrences or indi-cate a more widespread problem. Althoughthe danger of overinterpretation alwaysexists, we fail to see how the paper could beregarded as misleading.Graham Buckley Consultant otolaryngologistDepartment of Otolaryngology (Head and NeckSurgery), St James’s University Hospital, LeedsLS9 2TFJGrahamBuckley@compuserve.com

Every hospital must improvein-hospital resuscitationEditor—Timely and properly performedcardiopulmonary resuscitation and defibril-lation save many lives.1 The first fewmoments and minutes are all-important.

Prehospital cardiac care is increasinglybeing used, and improvements must now bemade to in-hospital cardiopulmonary resus-citation. Health care in hospitals is beingincreasingly compartmentalised, but provi-sion of cardiopulmonary resuscitation shouldnot be limited to a few individuals in any hos-pital. Various studies have shown the short-comings of physicians.2 Now, Fielden andBradbury show that surgeons and anaesthet-ists in theatre lack defibrillation skills.3

The resuscitation councils and thenational health services should enforce apragmatic policy to provide basic lifesupport resuscitation in every area andadvanced life support in some priority areasof our hospitals. Theatres and emergencyrooms should be recruited into this processfirst. At least one person trained in advancedlife support resuscitation should be availablein every active operating theatre, and at alltimes in emergency rooms.

One possibility is to require surgeons,anaesthetists, and physicians working inoperating theatres and emergency rooms tohold advanced life support certificationbefore they can renew their medical licence.Only with such mandatory policies wouldhospitals be in a position to provide properhealth care. No individuals or their familiesshould be left to think that surgical

Letters

1421BMJ VOLUME 318 22 MAY 1999 www.bmj.com

procedures were carried out in less thanideal settings or that resuscitation was left toinadequately trained staff.

I presume that the will for this exists; letus find a way to restore and maintain thepublic’s trust in our healthcare systems.V M K Bhaskarabhatla Research associate, internalmedicine associatesMount Sinai Medical Center, New York, USABaskarbhat@aol.com

1 Advanced Life Support Working Party of the EuropeanResuscitation Council. Guidelines for advanced lifesupport. Resuscitation 1992;24:111-22.

2 Chin D, Morphet J, Coady E, Davidson C. Assessment ofcardiopulmonary resuscitation in the membership exam-ination of the Royal College of Physicians. J R CollPhysicians Lond 1997;31:198-201.

3 Fielden JM, Bradbury NS. Observational study of defibril-lation in theatre. BMJ 1999;318:232-3. (23 January.)

Head lice can be controlledwithout application ofinsecticide lotionsEditor—Dawes et al’s evidence based casereport about treatment for head lice wasinteresting as an exercise in evidencesearching, but the authors failed to ask theright question in the first place.1 A more use-ful research question would have been, “Inchildren attending primary care with headlice, what is the most effective interventionthat will keep the problem under control?”

Head lice are most prevalent in primaryschool children, and the average number ofadult lice per child is only 8-10.2 This doesnot constitute a major health hazard, andthere is little evidence of an association withimpetigo as quoted in the article. Insecticidelotions are effective in killing head lice ifused correctly, but what is the point of usingrepeated applications of chemicals onschoolchildren? The children may becleared of head lice by insecticide lotion oneday, only for them to be reinfected on theirreturn to school the next day. Extensive andexpensive campaigns in the past, with entirecommunities being treated with insecticidelotion, have shown that reinfection fromoutside sources occurs quickly.

Wet combing (combing through wet,well conditioned hair with a fine toothed nitcomb), if done correctly and repeated everythree days, will control the problem byphysical removal of the lice.3 A plastic nitcomb, a short teaching session with theCommunity Hygiene Concern’s trainingvideo (160 Inderwick Rd, London N8 9JT),and support from health professionals willenable parents to take control of theproblem of head lice in their children. In thelong term, surely this is more effective andenvironmentally friendly than repeat pre-scribing of insecticide lotion.Gill Lewendon Senior clinical medical officerSouth and West Devon Health Authority,Dartington TQ9 6JEGill.Lewendon@sw-devon-ha.swest.nhs.uk

1 Dawes M, Hicks NR, Fleminger M. Evidence based casereport: Treatment for head lice. BMJ 1999;318:385-6. (6February.)

2 Mellanby K. Natural population of the head louse oninfected children in England. Parasitology 1942;34:180-4.

3 Elizabeth S. Can purchasers adopt the innovations offeredby the voluntary sector? Kings Fund News 1994;17:8.

WHO should undertake fullinquiry into Gulf war illnessEditor—Coker et al found a high prevalenceof various illnesses among British Gulf warveterans, confirming other recent reportsfrom the United Kingdom and the UnitedStates.1 They say that they found no evidencefor the existence of a unitary “Gulf warsyndrome” and think that several possiblecauses are responsible for the symptomsreported. There have also been numerousreports of increasing ill health among Iraqicivilians, over and above deaths ascribed tomalnutrition and lack of adequate medicalcare. These are widely blamed on thesanctions.2 Sikora describes both a break-down in cancer services and a risingincidence of some cancers,3 and there are dis-turbing, if anecdotal, reports of an increase inmajor congenital malformations.4 Some pos-sible causes (toxic chemicals from oil fires,chemical warfare agents, depleted uranium)could be common to both troops and Iraqis,but others (insecticides, multiple immunisa-tions) could not; it is hard to blame cancerand congenital deformities on stress.

As Sikora notes,3 the World HealthOrganisation has already worked in Iraq,but to a limited extent. We propose thatthe WHO should be authorised—andsupported—to undertake a full inquiry intoboth effects and causes of illness following theGulf war, not least as we are told that thiscould be a foretaste of future high-intensitywars. In your accompanying editorial,5 Mur-phy writes that proactive prevention must bedeveloped to reduce the burden of post-war illnesses. A comprehensive study such aswe propose might go further and say that waras a way of settling disputes must be avoideduntil diplomacy as an alternative really hasbeen ruled out. This was clearly not the casein 1990-1, nor, we believe, in the December1998 bombing, in which hospitals seem tohave been “collateral damage.”Robin Stott Chairmedact@gn.apc.orgDouglas Holdstock SecretaryMedact, London N19 4DJ

1 Coker WJ, Bhatt BM, Blatchley NF, Graham JT. Clinical find-ings for the first 1000 Gulf war veterans in the Ministry ofDefence’s medical assessment programme. BMJ 1999;318:290-4. (30 January.)

2 Garfield R, Zaidi S, Lennock J. Medical care in Iraq after sixyears of sanctions. BMJ 1997;315:1474-5.

3 Sikora K. Cancer services are suffering in Iraq. BMJ1999;318:203. (16 January.)

4 O’Kane M. Did we do this? Guardian 1998;21 December:G2 1-4.

5 Murphy FM. Gulf war syndrome. BMJ 1999;318:274-5.(30 January.)

GP surgeries were hard tocontact over ChristmasEditor—Accident and emergency depart-ments were overwhelmed with patientsduring the Christmas period. The healthminister, Frank Dobson, believes that a fluepidemic precipitated this problem. A morelikely explanation is that the general practiceservice in England and Wales broke down;people could not contact their generalpractitioners during the holiday period.

As part of a pharmacological researchproject, I was involved with a telephonesurvey of general practitioners. We selecteda simple random sample of general practi-tioners and hired a researcher to makephone calls to 200 of them. She attemptedto contact 91 surgeries by phone beforeChristmas and 93 after Christmas. The tableshows the number of times that she phonedeach general practitioner’s surgery beforethe phone was answered.

The results show that more phone callshad to be made to get through to a surgerybefore Christmas than after Christmas. Ittook up to nine attempts to get through to the89 of 91 surgeries that were reached beforeChristmas, but after Christmas all calls to the90 of 93 surgeries that were reached wereanswered within six attempts. These findingssuggest that the general practice system inEngland responded poorly to patients’ needsbefore Christmas, and patients would havehad extreme difficulty in getting advice from ageneral practitioner. Many patients wouldprobably therefore have attended a hospitalcasualty department.

The general practice system may havefailed because the general practitioners onduty were overstretched because of a flu epi-demic, adverse weather, or other reasons. Amore likely explanation, however, is that thegeneral practice service was understaffedover the Christmas period.

To prevent another fiasco it would beprudent for Mr Dobson to ensure thatgeneral practitioners keep their surgerydoors open over the Christmas period. Itmakes little sense that, each year, even thougha mid-winter hospital bed crisis is inevitable,general practitioners provide only a skeletonservice at Christmas time. The effects of a fail-ure of the general practice system cannot beimmediately corrected and have a dramaticeffect on the provision of beds, nurses, andmedical staff for the rest of the year.Andrew Rouse Senior lecturerDepartment of Public Health, University ofBirmingham, Birmingham BT15 2TT

Number of attempts that had to be made tocontact general practices by phone before andafter Christmas

BeforeChristmas(n=91)*

AfterChristmas(n=93)†

Answered at 1st attempt 20 54

Answered before or at 3rd attempt 62 81

Answered before or at 6th attempt 84 90

Answered before or at 9th attempt 89

Surgeries were telephoned on Tuesdays and Thursdaysbetween 0900 and 1130 or between 1300 and 1600.*Calls were made on 15, 17, and 22 Dec. Two practicescould not be reached (three and six attempts made).†Calls were made on 29 Dec and 5, 7, and 12 Jan. Threepractices could not be reached (three, four, and fourattempts made).

Letters

1422 BMJ VOLUME 318 22 MAY 1999 www.bmj.com