Growthhormone and tumour recurrence45 Hughes K, Yeo PPB, Lun KC, Sothy SP, Thai AC, Wang KW, et al. Ischaemic heart disease andits risk factors in Singapore in comparison with othercountries.

19 Xiang KS, Cox NJ, Sanz N, Huang P, Karam JH, Bell GI. Insulin-receptorand apolipoprotein genes contribute to development of NIDDM in ChineseAmericans. Diabetes 1989;38:17-23.

20 Yu YH. Prevalence and unawareness of hypertension in the petrochemicalindustrial population in China. Prev Med 1986;15:643-51.

21 Zhang HX, Wang ZY, Gou ZY, Jin LI, Hao XZ. Ten-year report on thecommunity control of hypertension and stroke in a rural district ofShijiazhuang, China. In: Yamori Y, Strasser T, eds. Ne;w horizons inpreventing cardiovascular diseases. Amsterdam: Elsevier, 1989:259-62.

22 Zimmet P, Dowse G, LaPorte R, Finch C, Moy C. Epidemiology-itscontribution to understanding of the etiology, pathogenesis, and preventionof diabetes mellitus. In: Creutzfeldt W, Lefebvre P, eds. Diabetes mellitus:pathophvsiology and therapv. Berlin: Springer-Verlag, 1989:5-26.

23 Linda MG. Diabetes mortality among Chinese migrants to New York city.Human Biol 1984;56:449-58.

24 McKeigue PM, M1iller GJ, Starmot MG. Coronary heart disease in SouthAsians overseas: a review. J Clin Epidemiol 1989;42:597-609.

25 D)owse GK, Gareeboo H, Zimmet PZ, Alberti KGMM, Tuomilehto J, FinchCF, et al. Abdominal obesity and physical inactivity as risk factors forNIDDMi and impaired glucose tolerance in Indian, Creole and ChineseMauritians. Diabetes Care 1991;14:271-8 1.

26 Nan L, Tuomilehto J, Dowse G, Zimmet P. Gareeboo H, Chitson P, et al.Prevalence and medical care of hypertension in four ethnic groups in thenewls-industrialised nation of Mauritius. J Hypertension 1991;9:859-66.

27 Reaven GM, Hoffman BB. A role for insulin in the aetiology and course ofhypertension? Lancet 1987;ii:435-7.

28 DeFronzo RA, Cooke C, Andres R, Faloona GR, Davis J. The effect of insulinin renal handling of sodium, potassium, calcium, and phosphate in man. JClin Invest 1975;55:845-55.

29 Tobev TA, Greenfield M, Kraemer F, Reaven GM. Relationship betweeninsulin resistance, insulin secretion, seery low density lipoprotein kineticsand plasma triglyceride levels in normotriglyceridemic man. Metabolism1981;30:165-71.

30 Modan M, Halkin H, Almog S, Lusky A, Esnkol A, Shefi M, et al.Hyperinsulinemia: a link between hypertension, obesity and glucoseintolerance. J Cltn Invest 1985;75:809-17.

31 Orchard TJ, Becker DJ, Bates M, Kuller LH. Plasma insulin and lipoproteinconcentrations: an atherogenic association? Am J Epidemiol 1983;118:326-37.

32 Zhong XL. Diabetes mellitus survey in China. Chinese Med J 1982;95:423-30.

33 SchwartzkopffW, Schleicher J, Pottins I, Yu SB, Han CZ, Du DY. Lipids,lipoproteins, apolipoproteins, and other risk factors in Chinese men andwomen with and without myocardial infarction. Atherosclerosis 1990;82:253-9.

34 ILuo CI, Lui M, Ping Y, Gou Z. Blood lipid analysis in coronarv heart diseaseand significancc. Tihan/in Medicaljournal 1986;1: 10.

35 Levv D, Kaninel WB. Cardiovascular risk: ncw insights from Framingham.Am Heart] 1988;116:266-72.

36 Stamler J, Wentworth 1), Neaton JD. Is the relationship between serumcholesterol and risk of prcmature death from coronary heart diseaseconttnuous and graded? _jAAA 1986;256:2823-8.

37 Yao CH, Hao EH, oXlu YY, Dong L, Wu YK. Dietary survey in Beijing in1983. Chinese Medj 1985;98:439-41.

38 Gong LS, Wang CX, Qian JA, Xu DH, Qian YS. Cardiovascular effects of'Faijiquan and Qigonig cxercises. In: Yamori Y, Strasser T, eds. Newz horizonsin pre-venting cardioVascular diseases. Amsterdam: Elsevier, 1989:163-7.

39 Rose G, Baxter PJ, Reid DD, McCartney 1'. Prevalence and prognosis ofelectrocardiographic finding in middle-aged men. Br Heart j 1978;40:636-43.

40 Uusitupa M, P6orala K, Raunio H, Rissanen V, Lampainen E. Sensitivity andspecificity of Minnesota code Q-QS abnormalities in the diagnosis ofmvocardial infarction serified at autopsy. Am Heartj 1983;106:753-7.

41 Liao Y, List K, Dyer A, Schoeniberger JA, Shekelle RB, Collette P, et al. Sexdifferential in the relationship of electrocardiographic ST-T abnormalitics torisk of coronary death: 11-5 year follow-up findings of the Chicago HeartAssociation detection pro-ject in industry. Circulation 1987;75:347-52.

42 Knutsen R, Knutsen S, Curb JD, Reed D, Kautz J, Yano K. The predictivevalue of resting electrocardiograms for 12-year incidence of coronary heartdisease in the Honolulu heart program. ] Clin Epidemiol 1988;41:293-7.

43 Kannel WB, Abbott RD. Incidence of prognosis of unrecognised myocardialinfarction: an update from the Framingham study. N Engl _j Med1984;311:1 144-7.

44 Brissonnette LGG, Fareed DS. Cardiovascular diseases as a cause of death inthe island of Mauritius, 1972-1980. World Health Stat Q 1985;38:163-75.

45 Hughes K, Yeo PPB, Lun KC, Sothy SP, Thai AC, Wang KW, et al.Ischaemic heart disease and its risk factors in Singapore in comparison withother countries. Ann Acad Med Singapore 1989;18:245-9.

46 Hughes K, Yeeo PPB, Lun KC, Thai AC, Sothy SP, Wang KW, et al.Cardiov-ascular diseases in Chinese, Mialays, and Indians in Singapore. II.Differences in risk factor lesels. ] Epidetniol Communutv Health 1990;44:29-35.

47 'I'ao SC, Huang ZD, WVu XG, Zhou BF, Xiao ZK, Hao JS, et al. CHI) and itsrisk factors in the P'eople's Republic of China. Int j Epidemiol 1989;18(suppl 1):S159-63.

48 Vartiainen E, Dianjun D, Marks JS, Korhonen H, Guanyi G, Ze-Yu G, et al.Mortality, cardio-ascular risk factors, and diet in China, Finland, and theUnited States. Public Health Rep 1991;106:41-6.

(Accepted 2 April 1992)

Growth hormone and tumour recurrence

A L Ogilvy-Stuart, W D J Ryder, H R Gattamaneni, P E Clayton, S M Shalet

Departments ofEndocrinology,Radiotherapy, andStatistics, Christie Hospitaland Holt Radium Institute,Manchester M20 9BXA L Ogilvy-Stuart, researchfellow in paediatricendocrinologyW D J Ryder, statisticianH R Gattamaneni, consultantradiotherapistP E Clayton, research fellowin paediatric endocrinologyS M Shalet, consultantendocrinologist

Correspondence to: DrShalet.

BMJ 1992;304:1601-5

AbstractObjective-To determine whether using growth

hormone to treat radiation induced growth hormonedeficiency causes tumour recurrence.Design-Comparison of tumour recurrence rates

in children treated with growth hormone for radiationinduced deficiency and an untreated population.Computed tomograms from children with braintumours were reviewed when starting growthhormone and subsequendly.Setting-North West region.Patients-207 children treated for brain tumour,

47 of whom received growth hormone and 161children with acute lymphoblastic leukaemia 15 ofwhom received growth hormone.Main outcome measures- Tumour recurrence and

changes in appearances on computed tomography.Results-Among children with brain tumour, five

(11%) who received growth hormone had recurrencescompared with 42 (26%) who did not receive growthhormone. Also adjusting for other variables thatmight affecttumour recurrence the estimated relativerisk of recurrence was 0-82 (95% confidence interval0*28 to 2 37). The only child with acute lymphoblasticleukaemia who relapsed while taking growth hormonehad relapsed previously before starting treatment.Two of the five children with brain tumours whorelapsed had abnormal appearances on computedtomography when growth hormone was started. 14other children who remained relapse free andhad follow up computed tomography showedno deterioration in radiological appearance duringtreatment.Conclusions-In this population growth hormone

did not increase the risk of tumour recurrence butcontinued surveillance is essential. Abnormal resultson computed tomography are not a contraindicationto treatment with growth hormone.

IntroductionAcute lymphoblastic leukaemia and brain tumours

are the two commonest childhood malignancies, thetreatment of which has consisted of cranial irradiationwith or without adjuvant cytotoxic chemotherapy.Long term management of the endocrine sequelae oftreatment, including growth failure, is fundamental tothe improved quality of life of these children. The useof growth hormone in children with radiation inducedgrowth hormone deficiency is now widely accepted,but questions still exist about the safety of thismitogenic hormone and whether it might cause arecurrence of a brain tumour or leukaemia.

Studies with small numbers of patients at our centre'and others23 suggested that growth hormone is notresponsible for recurrence of brain tumours, but noneof these studies applied statistical analysis. An analysisof deaths in recipients of pituitary growth hormoneshowed brain tumour recurrence to be one of the mostcommon causes, but a comparative group who hadnot received growth hormone was not available foranalysis.4 Therefore, it could not be establishedwhether growth hormone contributed to tumourrecurrence.

In children treated with growth hormone aftertreatment for a brain tumour radiographs of the centralnervous system often appear abnormal at the start oftreatment. No information is available to determine if

BMJ VOLUME 304 20 JUNE 1992 1601

on 22 March 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.304.6842.1601 on 20 June 1992. Dow

nloaded from

such children are at special risk of a clinically apparentrelapse after receiving growth hormone. We thereforecompared tumour recurrence rates in a large number ofchildren with radiation induced growth hormonedeficiency treated with growth hormone with rates inan untreated population.

Patients and methodsWe studied all children aged less than 14-4 years who

had brain tumour diagnosed between 1965 and 1989 oracute lymphoblastic leukaemia between 1970 and 1989in the North West region who were clinically relapsefree at least two years after diagnosis. Sixty eightchildren, 53 with a brain tumour distant from thehypothalamic-pituitary axis (36 boys) and 15 withacute lymphoblastic leukaemia (nine boys) were treatedwith growth hormone for radiation induced growthhormone deficiency. Six of the 53 children with braintumours were excluded from the statistical analysis.

In the early part of this study only children withobvious growth failure were referred to the endo-crinologist and because of the scarcity of pituitaryderived growth hormone only those with the worstgrowth prognosis received growth hormone. Now thatthe natural course ofradiation induced growth hormonedeficiency is better understood all children who arerelapse free at two years are assessed for growthhormone deficiency and, with the wide availability ofsynthetic human growth hormone, all children withgrowth hormone deficiency are considered for treat-ment. The tumour prognosis did not knowingly affectpatient selection for growth hormone treatment inthose with a diagnosis of a brain tumour.The peak growth hormone concentration on pro-

vocative testing with either insulin hypoglycaemia(0 2 U/kg intravenously) or glucagon (15 Itg/kgintramuscularly) in all patients treated with growthhormone was less than 15 mU/l. Growth hormone wasstarted at least two years after completion of radio-therapy-that is, after the time when tumour recur-rence is most likely to occur. The dose of growthhormone was 12 IU/week before 1989 and 0 5 IU/kg/week after 1989. Twenty three children with braintumours and six with acute lymphoblastic leukaemiareached final height and discontinued growth hormone.Each child received cranial irradiation. The median

dose to the head was assessed in those who receivedgrowth hormone. In children treated for a braintumour the median dose was 3000 cGy (range 1500-4750) in 20 (8-28) fractions over 27 (9-36) days. Inaddition, 36 children received a boost to the tumoursite (median dose 1500 cGy (range 1000-2000) in 10(4-11) fractions over 13 (3-22) days). In childrentreated for acute lymphoblastic leukaemia the mediancranial dose was 2400 (1800-4200) cGy in 16 (11-50)fractions over 15 (11-35) days.

Each child with acute lymphoblastic leukaemiareceived conventional chemotherapy and radiotherapyexcept one who had already had a haematological andcentral nervous system relapse. This child was addition-ally treated with intrathecal radioiodine targetedmonoclonal antibodies. Most of the children with braintumours had had surgery and insertion of a ventriculo-peritoneal shunt before radiotherapy. The dose andtechniques remained standard throughout the study.The children with brain tumours were randomlyassigned to receive chemotherapy, which consisted ofvincristine alone or in combination with a nitrosourea,with or without procarbazine over 12 to 18 months.The records of all children registered with the North

West children's cancer registry were reviewed.Diagnosis, treatment details (use of radiotherapy andchemotherapy), relapse-free survival, age at diagnosis,and sex were noted.

Relapse was defined as clinical recurrence of theoriginal tumour either at the primary site, or elsewhere.The first child treated with growth hormone after abrain tumour received initial irradiation treatment in1965, and the first child treated after acute lympho-blastic leukaemia received irradiation in 1970.Each child with a brain tumour was regularly

reviewed by both the radiotherapist and the neuro-surgeon. Computed tomograms taken at the time ofstarting growth hormone were reviewed and comparedwith scans taken during follow up.

STATISTICAL METHODS

The primary end point was chosen to be relapserather than survival; relapse is a marker for survivaland the management policy was to consider onlychildren with growth hormone deficiency who wereclinically relapse free as eligible for growth hormonetreatment. Hence the decision not to start growthhormone in a relapsed child is highly associated withprognosis and results in the self selection of childrenwith a better chance of survival for growth hormonetreatment.

Results for children with brain tumours wereanalysed by Cox's regression model' with a timedependent indicator variable,'6 taking the values oneor zero at time t (measured from irradiation) for eachchild at risk according to whether the child had or hadnot received growth hormone by that time. Thisvariable was recalculated for each child still at risk ateach relapse time, so a child was considered to be in theno growth hormone group up until the time growthhormone was actually given when he or she switched tothe growth hormone group. Allowance was made forother covariates thought to influence relapse freesurvival-namely, diagnosis, sex, age at diagnosis, andwhether or not chemotherapy was included in theprimary treatment. As the cases spanned several yearsand there seemed to be some improvement in prognosisover time, the analysis was stratified by quinquennia ofinitial treatment-that is, 1965-9, 1970-4 . . ., 1985-9,which permitted different baseline risks of relapsebetween strata. The Cox models were fitted to the databy using program 2L of the biomedical programs datapackage.We excluded from the analysis all children who had

relapsed within two years after diagnosis, all those whohad not received radiotherapy during primary treat-ment (and who would therefore not be at risk ofradiation induced growth hormone deficiency), sixchildren who received growth hormone therapy withintwo years after diagnosis, and two other children inwhom treatment details were incomplete. Thus 207children (123 boys) with brain tumours aged between0 5 and 14 4 years (median 6-7 years) were included inthe analysis, 47 of whom (29 boys) received growthhormone. The median length of time from diagnosis tostarting growth hormone was 4-5 (range 2-02-10-8)years, and the median duration of growth hormonetreatment was 3-2 years.The Cox model could not be used for the children

with a primary diagnosis of acute lymphoblasticleukaemia because only small numbers were treatedwith growth hormone and there were no first relapsesin the treated group. In addition, although selection ofpatients to receive growth hormone was not knowinglybased on prognosis in the brain tumour group, theremay have been selection in those children with acutelymphoblastic leukaemia who subsequently receivedgrowth hormone. As only one child with a primarydiagnosis ofacute lymphoblastic leukaemia was treatedwith growth hormone before at least five years afterdiagnosis we have reported the relapse numbers ofthose who were not treated with growth hormone afterfive years relapse free survival.

BMJ VOLUME 304 20 JUNE 19921602

on 22 March 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.304.6842.1601 on 20 June 1992. Dow

nloaded from

ResultsBRAIN TUMOURS

Table I shows the characteristics of the children withbrain tumours who did and did not receive growthhormone. Five of the 47 children (11%) treated withgrowth hormone had a clinical relapse associated withrecurrence of brain tumour. In two this occurred 1-8and 4-4 years after completion of growth hormonetreatment. In the remainder relapse occurred whilereceiving growth hormone at 0-5, 0-7, and 3-3 yearsafter starting treatment. One child with an astrocytomasurvived the recurrence. Forty two of the 160 children(26%) who did not receive growth hormone relapsed.None ofthe six children who received growth hormone

TABLE I- Characteristics ofchildren with brain tumours

No (%) not treated No (%) treated withwith growth hormone growth hormone

(n= 160) (n=47)

Diagnosis:Medulloblastoma 43 (27) 26 (55)Ependymoma 20 (13) 6 (13)Juvenile astrocytoma 74 (46) 7 (15)Adult astrocytoma 16 (10) 4 (9)Other glioma 7 (4) 4 (9)

Year of diagnosis:1965-9 41 (26) 4 (9)1970-4 41 (26) 7(15)1975-9 22 (14) 10 (21)1980-4 29 (18) 18 (38)1985-9 27 (17) 8 (17)

Sex:Male 94 (59) 29 (62)Female 66 (41) 18 (38)

Age (years):<5 50 (31) 22 (47)5-10 61 (38) 22 (47)>10 49 (31) 3 (6)

Median age (years) 7 2 5 1Chemotherapy:No 133 (83) 24 (51)Yes 27 (17) 23 (49)

but were subsequently removed from the analysisrelapsed.

Table II shows the age at irradiation, time and lengthof growth hormone treatment, and years of survivalsince completion of growth hormone treatment in alltreated children with and without a tumour recurrence.Medulloblastoma-In all, 124 children had medullo-

blastoma diagnosed between 1965 and 1989 in theNorth West region, 69 ofwhom were clinically relapsefree after two years. Of the 26 who received growthhormone, two (8%) relapsed. One further child died asa result of an accident. Fifteen of the 43 children (35%)who did not receive growth hormone relapsed, four ofwhom survived. One other child died of a treatmentrelated cause.Ependymoma-Seventy cases of this poor prognosis

tumour were diagnosed between 1965 and 1989.Twenty six of the children were relapse free two yearsafter diagnosis. Two of the six children (33%) whoreceived growth hormone relapsed compared withseven of20 (35%) who did not receive growth hormone.Juvenile astrocytoma-142 cases were diagnosed of

this relatively good prognosis brain tumour which isnot always treated with radiotherapy. Eighty one of thechildren treated with radiotherapy were relapse freetwo years after diagnosis. Fourteen of the 74 children(19%) who did not receive growth hormone subse-quently relapsed. Two children died of other causes.One of the seven children (14%) who received growthhormone relapsed. She survived the relapse but had afurther relapse, although she remained alive despitethe presence of disease.

Adult astrocytoma-Twenty of the 86 children wererelapse free two years after diagnosis. Four receivedgrowth hormone and remained disease free. Four ofthe 16 children (25%) who did not receive growthhormone relapsed.

TABLE II-Clinical details of children treated with growth hormone (individual data on children who relapsed and mean data on those who didnot)

Time from radiotherapy Length of growth Time to relapse fromAge at to starting growth hormone hormone treatment starting growth Years since growth

Patients Tumour radiotherapy (years) (years) hormone (years) hormone stopped

With relapse:I Medulloblastoma 13-1 2-9 3 0 3-4 1-82 Medulloblastoma 8-3 3-4 0 7 07 *3 Ependymoma 10-9 2-4 0-4 0-4 *4 Ependymoma 11-3 2-5 1 0 5-3 4-45 Astrocytoma 2-1 9-1 3-3 3-3 12-4*6 Acute lymphoblastic

leukaemia 3-0 8-4 1 7 1 7 2*4*tWithout relapse:Growth hormone

completed (n= 27) 5-4 5-9 4-1 5-1Growth hormone

ongoing(n=29) 4 0 5 0 2-8

*Relapse while taking growth hormone.tPrevious relapse before starting growth hormone.

TABLE III -Cox regression analysis on results in children with brain tumour: parameter estimates from fullmodel* stratified by quinquennia ofdiagnosis

ParameterVariable estimate Standard error Relative risk p Valuet

Diagnosis (reference= medulloblastoma): 0-24Z, (=1 ifependymoma,0 otherwise) 0-49 0-43 1-63Z2 (= 1 if juvenile astrocytoma,0 otherwise) -0-52 0-38 0 60Z3 ( 1 if adult astrocytoma, 0 otherwise) 0 17 0 57 1 19Z4 ( 1 if other glioma, 0 otherwise) -0-39 0-78 0-68

Sex (reference= male): <0-01Zs (= I if female, 0 otherwise) -0-87 0-35 0-42

Age (reference= 5 years): 0-18Z6 I if >5, and l IO years) 0-21 0-37 1 23Z7 I if >IO years) 0-76 0-42 2 15

Chemotherapy (reference= none): 0 11Z8 (= 1 if chemotherapy, 0 otherwise) 0 7 0-44 2-02

Growth hormonet: 0-71Z9 (t) (= 1 if given by time t, 0 otherwise) -0-2 0-54 0-82

*None of the two factor interactions contributed significantly to the model.tP Value from a likelihood ratio test of omitting each factor from the full model.tNote the growth hormone status is recalculated for each patient still at risk at each relapse time.

Other glioma-Twenty seven children had otherbrain tumours diagnosed that were not classified intoone of the previous four categories. Eleven died andfive others had relapsed within two years. Of the 11who remained clinically relapse free at two years, fourreceived growth hormone and none relapsed. Two ofseven (29%) who did not receive growth hormonerelapsed, one of whom survived. One other child diedfrom a treatment related cause.The impact on relapse of diagnosis, sex, age, use of

chemotherapy in the initial treatment, and growthhormone treatment is shown in table III. The childrenwho received chemotherapy (which was randomlyallocated as part of a multicentre trial), had a worseprognosis two years after diagnosis than those who didnot receive chemotherapy. Proportionally morechildren who received growth hormone therapy hadreceived chemotherapy in their initial managementthan those who did not receive growth hormone. The

BMJ VOLUME 304 20 JUNE 1992 1603

on 22 March 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.304.6842.1601 on 20 June 1992. Dow

nloaded from

reason for this is not clear, but in these cases there is anapparent protective effett of growth hormone againstthe risk of relapse.

Table IV shows the estimated relative risk of growthhormone influencing tumour recurrence unadjusted,adjusted for the various parameters above, and adjustedbut excluding the effect of chemotherapy. In each casethe analysis has been stratified by quinquennia ofdiagnosis to allow for changing practices in tumourmanagement that may have affected prognosis. In our

population the use ofgrowth hormone has not increasedthe risk of tumour recurrence (adjusted relative risk,excluding chemotherapy 1 01). The size of the confi-dence intervals, however, suggests that larger numberswill need to be studied for complete reassurance on thisissue.

ACUTE LYMPHOBLASTIC LEUKAEMIA

Acute lymphoblastic leukaemia was diagnosed in484 children between 1970 and 1989 in the North Westregion. In all, 161 children were relapse free at fiveyears, and 14 had received growth hormone. One otherchild who received growth hormone relapsed (table II).He had had two previous relapses before startinggrowth hormone but was thought to have a reasonableprognosis at the time treatment was started. No childwho had been relapse free when growth hormone wasstarted subsequently relapsed. Eleven of 147 (7%)children who did not receive growth hormone relapsedand a further three died of other related causes.No child who received growth hormone developed a

second primary tumour and no child treated forbrain tumour and subsequently with growth hormonedeveloped leukaemia.

COMPUTED TOMOGRAPHY

Forty four children with brain tumours hadcomputed tomography around the time of startinggrowth hormone (mean 1-4 (SD 1 3) years beforegrowth hormone started). Apart from changes resultingfrom surgery and radiotherapy in 19 children computedtomography showed other abnormalities. In 10 childrenthe scans showed residual tumour and in nine non-

enhancing low attenuation or cystic lesions in the areaof the original tumour.The five children with brain tumours who had

relapsed had had computed tomography when growthhormone was started. One had evidence of residualtumour and one had a low density non-enhancinglesion. In four children subsequent relapse wasconfirmed by finding an enhancing tumour on com-

puted tomography. In the fifth child clinical relapsewas confirmed by cerebrospinal fluid cytology before alesion became apparent on computed tomography.Of the 39 children who had scans around the time of

starting growth hormone but who had not clinicallyrelapsed, 14 had subsequent scans up to 14-9 yearsafter starting growth hormone (mean 3 7 (SD 4-4)years). Four children with follow up scans had residualtumour at the time of starting growth hormone. In twothe follow up scan showed no residual tumour and inthe other two children there was no notable change inscan appearance. In three other children computedtomography showed an area of low density at the timeof starting growth hormone. In one of these children a

subsequent scan appeared normal and in the other twothere was no change on follow up. None had adeterioration in appearance of the residual lesion. Theother scans, which showed no residual tumour whengrowth hormone was started remained unchanged on

follow up. One child whose scan showed residualtumour when growth hormone was started subse-quently died as a result ofan accident 1-3 years later. Atpostmortem examination there was no evidence ofresidual tumour.

DiscussionMost children receiving cranial irradiation with a

dose in excess of 24 Gy for treatment of brain tumoursor leukaemia will become growth hormone deficientafter five years; the speed of onset of growth hormonedeficiency is dose dependent.7 It is now recognised thatthe ultimate height in these children is further compro-mised by early puberty and, in those receiving cranio-spinal irradiation, poor spinal growth. We thereforenow aim to start growth hormone treatment in childrenwho are growth hormone deficient at a time at whichrelapse is less likely but before there has been an

appreciable decrease in stature. In our unit the dose ofgrowth hormone is now based on weight. Therefore wewould expect that the length of time on growthhormone and the dose per kg (except when the weightis less than 24 kg) for an individual child has increasedin recent years.

Growth hormone is mitogenic, and there is evidencein animals of a cause and effect relation betweensupraphysiological doses of growth hormone anddevelopment of acute leukaemia.89 In addition, in vitrodata have shown that growth hormone and insulin-likegrowth factor 1 promote leukaemic blast cell replicationfrom human marrow.'0 In animals the incidence ofsolid tumours is increased after giving growth hormone.In female rats lymphosarcomas of the lung, adreno-cortical and adrenomedullary carcinomas, ovariantumours, and breast tumours have been described aftergiving growth hormone." In patients with acromegaly,in which a high growth hormone concentration ispathologically sustained, there is a significantlyincreased incidence of cancer in general.12 Develop-ment of acute lymphoblastic leukaemia has beenreported de novo in children receiving growth hormoneeither for idiopathic growth hormone deficiency orafter treatment of a brain tumour. 3 14 But although theincidence of leukaemia may be slightly increased aftergrowth hormone treatment in growth hormonedeficient patients, it is not clear that this can beattributed to growth hormone.'5 In our study theonly child with acute lymphoblastic leukaemia whorelapsed while taking growth hormone was at highrisk because of the two previous relapses of leukaemia;furthermore, no child with a primary diagnosis ofbraintumour developed acute lymphoblastic leukaemia afterreceiving growth hormone.

INTERPRETATION OF STUDY

The North West tumour registry has allowed us toobtain accurate data on children with the same tumourdiagnosis in whom growth hormone was not used. Aswell as documenting non-fatal tumour recurrence or

persistent disease, the registry differentiates deathfrom causes other than tumour recurrence. For eachchild there are accurate records of treatment. It istherefore encouraging that in each tumour category, inwhich the prognosis differs, there was no associationbetween growth hormone therapy and tumourrecurrence.

Several children had abnormal appearances on

computed tomography when growth hormone wasstarted. The reasons for the abnormalities includepostoperative features, radiation changes, or residualor recurrent tumour. Even those reported as showingresidual tumour were not more likely to show evidenceof progression of the lesion on follow up scanning. Wefound no evidence that abnormal results on computedtomography should be a contraindication to startinggrowth hormone treatment.The results of an earlier study from this centre

(the patients representing a subset of the currentpopulation) suggested growth hormone was notresponsible for tumour recurrence.' In that studyfewer children with radiation induced growth hormone

BMJ VOLUME 304 20 JUNE 1992

TABLE IV-Estimated relative nisksand 95% confidence intervals forimpact ofgrowth hormone on tumourrecurrence unadjusted; adjusted forage, sex, diagnosis, and use ofchemotherapy; and adjusted butexcluding the effect ofchemotherapy

Estimated 95%relative Confidencerisk interval

Unadjusted 1-35 0O49to3-73Adjusted 0 82 0 28to2-37Adjusted minuschemotherapy 1 01 0-36 to 2-83

1604

on 22 March 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.304.6842.1601 on 20 June 1992. Dow

nloaded from

deficiency were treated with growth hormone, at alower individual dosage, and for a shorter time. Withthe more widespread use of growth hormone, andparticularly with its current use in immunosuppressedpatients, review of its safety is essential. Childrentreated with cranial irradiation account for the largestgroup of children with growth hormone deficiency dueto an organic lesion, and growth hormone has provedbenefits in this group. It is encouraging that there is noincreased risk of tumour recurrence in our children inwhom growth hormone therapy has been used. Cautionmust still exist, however, and continuing surveillanceis crucial.

We thank Serono Laboratories (UK) for financial supportand the Manchester Children's tumour registry, which issupported by the Cancer Research Campaign.

1 Clayton PE, Shalet SM, Gattamaneni HR, Price DA. Does growth hormonecause relapse of brain tumours? Lancet 1987;i:71 1-3.

2 Arslanian SA, Becker DJ, Lee PA, Drash LA, Foley TP. Growth hormone andtumor recurrence. Findings in children with brain neoplasms and hvpo-pituitarism. AmJ7 Dis Child 1985;139:347-50.

3 Rodens KP, Kaplan SL, Grumbach MM, Teller WM. Does growth hormone

therapy increase the frequency of tumor recurrence in children with braintumors? Acta Endocnrnol (Copenh) 1987;283 (suppl): 188-9.

4 Buchanan CR, Preece MA, Milner RDG. Mortality, neoplasia, and Creutz-feldt-Jakob disease in patients treated with human pituitary growthhormone in the United Kingdom. B,J 1991;302:824-8.

5 Cox DR. Regression models and life-tables (with discussion). Journal of theRoyal Statistical Society 1972;34[B]: 187-220.

6 Cox DR, Oakes D. Analysis of surinval data. London: Chapman and Hall,1984:112-41.

7 Clayton PE, Shalet SM. Dose dependency of onset of radiation-inducedgrowth hormone deficiency.I Pediatr 1991;118:226-8.

8 Bentley HB, Hughes ER, Peterson RDA. Effect of hypophysectomy on avirus-induced T-cell leukaemia. Nature 1974;252:747-8.

9 Rogers PC, Komp D, Rogol A, Sabio H. Possible effect of growth hormone ondevelopment of acute lymphoblastic leukaemia. Lancet 1977;ii:434-5.

10 Estrov Z, Meir R, Barak Y, Zaizov R, Zadik Z. Human growth hormone andinsulin-like growth factor-l enhance the proliferation of human leukemiccells. J Clin Oncol 1991;9:394-9.

11 Moon HD, Simpson ME, Li CH, Evans HM. Neoplasms in rats treated withpituitary growth hormone. I. Pulmonary and lymphatic tissues. Cancer Res1950;10:297-308.

12 Ezzat S, Melmed S. Are patients with acromegaly at increased risk forneoplasia?J7 Clin Endocrinol Metab 1991;72:245-9.

13 Watanabe S, Tsunematsu Y, Fujimoto J, Komiyama A. Leukaemia in patientstreated with growth hormone. Lancet 1988;i: 1159.

14 Delemarre-Van de Waal HA, Ondink JH, de Grauw TJ, de Waal FC.Leukaemia in patients treated with growth hormone. Lancet 1988;i: 1159.

15 Fisher DA, Job J-C, Preece MA, Underwood LE. Leukaemia in patientstreated with growth hormone. Lancet 1988;i: 1159-60.

(Accepted 27 April 1992)

Effect of serotesting with counselling on condom use andseroconversion among HIV discordant couples in Africa

Susan Allen, Jeffrey Tice, Philippe Van de Perre, Antoine Serufilira, Esther Hudes,Francois Nsengumuremyi, Joseph Bogaerts, Christina Lindan, Stephen Hulley

Department of Pathology,University of California,San FranciscoSusan Allen

Division of ClinicalEpidemiology, Departmentof Epidemiology andBiostatistics, and Centerfor AIDS PreventionStudies, University ofCalifornia, San FranciscoJeffrey TiceEsther HudesChristina LindanStephen Hulley

Projet San Francisco,Kigali, and Ministry ofHealth, RwandaAntoine SerufiliraFrancois Nsengumuremyi

National AIDS ControlProgram, Kigali, RwandaPhilippe Van de Perre

Centre Hospitalier deKigaliJoseph Bogaerts

Correspondence andrequests for reprints to:Dr Susan Allen, 74 NewMontgomery St, Suite 600,San Francisco, CA 94105,USA.

BMJ7 1992;304:1605-9

AbstractObjective-To determine whether HIV testing

and counselling increased condom use and decreasedheterosexual transmission of HIV in discordantcouples.Design-Prospective study.Setting-Kigali, the capital of Rwanda.Subjects-Cohabiting couples with discordant

HIV serology results.Main outcome measures-Condom use in the

couple and HIV seroconversion in the negativepartners.Results-60 HIV discordant couples were identi-

fied, ofwhom 53 were followed for an average of 2-2years. The proportion of discordant couples usingcondoms increased from 4% to 57% after one year offollow up. During follow up two of the 23 HIVnegative men and six of the 30 HIV negative womenseroconverted (seroconversion rates of 4 and 9 per100 person years). The rate among women was lessthan half that estimated for similar women in dis-cordant couples whose partners had not been sero-tested. Condom use was less common among thosewho seroconverted (100% v 5%, p=0-01 in men;67% v 25%, p=0-14 in women).Conclusions-Roughly one in seven cohabiting

couples in Kigali have discordant HIV serologicalresults. Confidential HIV serotesting with counsel-ling caused a large increase in condom use and wasassociated with a lower rate of new HIV infections.HIV testing is a promising intervention for prevent-ing the spread of HIV in African cities.

IntroductionThe major route of HIV transmission in Africa is

heterosexual intercourse.`'3 Genital ulcerations, use ofhormonal contraceptives, lack of male circumcision,and the clinical status of the index case have beenimplicated as risk factors in transmission of HIV.2 4-8 Aswith gonorrhoea, gender is also thought to have a role,

with male to female transmission being more efficientthan female to male transmission.9 However, theimportance of these factors in explaining the highprevalence of infection in Africa and their utilityin designing effective prevention programmes areunknown.

Cohort studies of HIV discordant couples (onepartner HIV positive, the other HIV negative) thatmonitor the risk factors, sexual behaviour, and HIVstatus of both partners over time provide data forexamining the determinants of heterosexual trans-mission. Studies of discordant couples in the UnitedStates have found a wide variation in reported rates ofinfection among the heterosexual partners of peopleinfected with HIV6 10; the lack of a detectable associa-tion between the number of acts of sexual intercourseand the risk of infection911"13 suggests the presenceof other factors affecting HIV infectivity and suscepti-bility.The aim of this study is to evaluate the effectiveness

of a prevention programme linking confidential HIVserotesting with a programme for counselling andproviding condoms to urban couples. We report herethe factors associated with risk reduction (particularly,the use of latex condoms) and seroconversion among 53HIV discordant couples followed prospectively for twoyears in the capital of Rwanda in Africa.

Subjects and methodsSUBJECTS

In 1986 the Projet San Francisco was established inKigali, Rwanda, by the University of California in SanFrancisco in collaboration with the National AIDSProgram of Rwanda. During 1986 and 1987 a consecu-tive, population based sample of 3702 women aged18-35 was screened for HIV at the prenatal andpaediatric outpatient clinics of the Centre Hospitalierde Kigali.14 Stratified random samples of 460 HIVpositive and 998 HIV negative women were recruitedin 1988'5 for longitudinal studies of the predictors of

BMJ VOLUME 304 20 JUNE 1992 1605

on 22 March 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.304.6842.1601 on 20 June 1992. Dow

nloaded from