and the Royal Marsden Hospital since GH treatment of such patients started there ..... unity with longer follow-up, as the initial selection wore off. This was the ...
0021-972X/00/$03.00/0 The Journal of Clinical Endocrinology & Metabolism Copyright © 2000 by The Endocrine Society
Vol. 85, No. 12 Printed in U.S.A.
Growth Hormone Treatment of Children with Brain Tumors and Risk of Tumor Recurrence* A. J. SWERDLOW, R. E. REDDINGIUS, C. D. HIGGINS, H. A. SPOUDEAS, K. PHIPPS, Z. QIAO, W. D. J. RYDER, M. BRADA, R. D. HAYWARD, C. G. D. BROOK, P. C. HINDMARSH, AND S. M. SHALET Section of Epidemiology, Institute of Cancer Research (A.J.S., C.D.H.), Sutton, Surrey SM2 5NG; Departments of Endocrinology (R.E.R., S.M.S.) and Medical Statistics (W.D.J.R.), Christie Hospital, Manchester, London; Centre of Paediatric Endocrinology and Metabolism, Middlesex Hospital (H.A.S., C.G.D.B., P.C.H.), London; Department of Neurosurgery, Great Ormond Street Hospital (K.P., R.D.H.), London; and Academic Department of Radiotherapy and Oncology, Royal Marsden Hospital National Health Service Trust (M.B.), Sutton, Surrey, United Kingdom ABSTRACT GH is increasingly used for treatment of children and adults. It is mitogenic, however, and there is therefore concern about its safety, especially when used to treat cancer patients who have become GH deficient after cranial radiotherapy. We followed 180 children with brain tumors attending three large hospitals in the United Kingdom and treated with GH during 1965–1996, and 891 children with brain tumors at these hospitals who received radiotherapy but not GH. Thirty-five first recurrences occurred in the GH-treated children and 434 in the untreated children. The relative risk of first recurrence in GH-treated compared with untreated patients, adjusted for poten-
G
H TREATMENT was introduced in 1959. Initially it was prescribed for childhood GH deficiency, but as availability has increased with the development of recombinant methods for synthesis its use has expanded greatly, with evidence of benefit for a growing range of indications in children and adults (1–7). GH, however, is mitogenic. In vitro and animal experiments suggest that it can raise the risk of hyperplasia and malignancy (8 –12), and there are indications, although not conclusive, that it may raise the risk of cancer incidence in humans (13–16). There is therefore concern that GH treatment of children who have undergone radiotherapy for brain tumor, with consequent GH deficiency, may increase their risk of tumor recurrence. There are insufficient data, however, to detemine whether this occurs. Brain tumor recurrence is a frequent cause of death in patients treated with GH (17, 18), but there is very limited information on whether recurrence rates are greater after this treatment than in comparable untreated patients. Three relatively small studies have been reported (19 –22), only one of which made allowance for confounding factors (22). Recurrence rates were not found to be increased in treated subjects (22), but confidence intervals (CIs) were wide and follow-up short, so that considerable uncertainty remains. The present Received June 20, 2000. Revision received August 29, 2000. Accepted September 6, 2000. Address all correspondence and requests for reprints to: Prof. A. J. Swerdlow, Institute of Cancer Research, Section of Epidemiology, D Block, Cotswold Road, Sutton, Surrey, United Kingdom SM2 5NG. * This work was supported by a grant from the United Kingdom Medical Research Council.
tially confounding prognostic variables, was decreased (0.6; 95% confidence interval, 0.4 – 0.9) as was the relative risk of mortality (0.5; 95% confidence interval, 0.3– 0.8). There was no significant trend in relative risk of recurrence with cumulative time for which GH treatment had been given or with time elapsed since this treatment started. The relative risk of mortality increased significantly with time since first GH treatment. The results, based on much larger numbers than previous studies, suggest that GH does not increase the risk of recurrence of childhood brain tumors, although the rising trend in mortality relative risks with longer follow-up indicates the need for continued surveillance. (J Clin Endocrinol Metab 85: 4444 – 4449, 2000)
study combined longer follow-up from the three largest treatment centers in the United Kingdom to give a much larger analysis, with greater power, on the risks of tumor recurrence in relation to GH replacement therapy. Subjects and Methods The study included all children treated with radiotherapy for brain tumors other than craniopharyngioma at Great Ormond Street Hospital and the Royal Marsden Hospital since GH treatment of such patients started there and all similar patients at the Christie Hospital who had survived for at least 2 yr; the latter criterion was employed because GH had not been used before 2-yr survival at the Christie. Patients with craniopharyngioma were excluded because these tumors have a different biology and more complex and less well defined risk factors for recurrence than other brain tumors, and treatment strategies vary considerably between surgeons. Data on diagnostic, prognostic, and demographic variables; treatment for malignancy; and follow-up, including tumor recurrence, were obtained from computerized databases at the study hospitals and the North West Childhood Cancer Registry, supplemented when necessary by examination of case notes. Recurrence was diagnosed on the basis of clinical symptomatology plus radiological appearances. Data on GH treatment were obtained from pediatric endocrinology files and case notes; the standard dosage was 15–20 IU/m2/week or 0.5 IU/kg/week. In all patients the diagnosis of GH deficiency was based on auxological and conventional provocative tests of GH status. Person-time at risk of recurrence of malignancy in GH-treated patients was calculated from the date of first GH treatment to the date of last clinical contact, first tumor recurrence, or death, whichever occurred first. This person-time was calculated separately by category of sex (i.e. separately for males and females), age at tumor diagnosis, histology, calendar period when the tumor was diagnosed, time since diagnosis, whether the patient received chemotherapy, and hospital of treatment. Person-time in patients not treated with GH was calculated similarly, starting from the date of first treatment for brain tumor (or at the Christie Hospital, the second
4444
GH AND TUMOR RECURRENCE anniversary of this), and including within the untreated category the experience before first GH treatment of children who subsequently received GH. The relapse rate among GH-treated patients was compared with that among untreated patients using Cox regression analysis (23), adjusting for the confounding variables detailed in the tables, with GH treatment as a time-dependent variable. In particular, it should be noted that this analysis allowed for the change in recurrence risk with time since tumor occurrence, and hence allowed for the tendency for follow-up of GHtreated subjects compared with untreated patients to be in years longer after the tumor occurred, when recurrence risks are comparatively low. Ninety-five percent CIs for the hazard ratios [relative risks (RRs)] were obtained assuming a normal distribution. Assessments of heterogeneity in RRs and trends in risk were conducted using the likelihood ratio test (24). As well as analyzing risks of relapse, we repeated the analyses using death instead of relapse as the outcome under analysis. For analysis of risk in relation to time since tumor diagnosis, the Cox model does not directly provide an estimate of the baseline relapse rate over time, and therefore to provide the analysis shown in Table 2, a Poisson rather than a Cox model was used, as it produces very similar RR estimates to the Cox model and, in addition, a simple estimate of the baseline rate. The use of the date of last clinical contact as an exit from risk in the analysis gave potential for bias, because recentness of follow-up could have been related to whether GH had been administered and whether an adverse outcome had occurred. To check whether such a bias might have affected the results, we recalculated all RRs using the date when the study dataset was compiled at each hospital, instead of the clinical follow-up date, as the date of exit from risk for patients who were not known to have died or relapsed, i.e. assuming that all patients not known to have done otherwise remained alive and relapse-free from their last clinical contact to the end of the study period.
Results
We identified 1438 children with brain tumors other than craniopharyngioma treated at the study hospitals from 1965 to
4445
1996, of whom 1084 had received radiotherapy. For 13 of these there was insufficient information for inclusion in the study. This left 1071 patients in the analysis, 180 who had been treated with GH and 891 not so treated. (The latter group included 11 patients who received GH solely after a first relapse, and who therefore contributed no follow-up after GH in the analysis, because the analysis was only concerned with the period before relapse.) Descriptive characteristics of the subjects are shown in Table 1. GH-treated patients were followed for an average of 6.4 yr after first receiving GH, with a maximum of 20 yr, and with 31 patients followed for 10 yr or more. The mean follow-up of patients who had not (yet) received GH, including the personyears accumulated before GH treatment by subjects who subsequently received this treatment, was 4.5 yr/patient. During follow-up, 469 patients relapsed, 81 died without relapse, and 521 were alive and relapse-free at last contact. Of the 35 relapses in GH-treated patients, 20 occurred during the treatment, and 15 after it. Risk of relapse in patients overall was significantly and independently related to histology of the brain tumor, calendar period of diagnosis, duration since first treatment of brain tumor, and whether chemotherapy was given (Table 2). Risk of relapse was significantly reduced in GH-treated patients compared with untreated patients before adjustment for prognostic variables (RR, 0.6; 95% CI, 0.4 – 0.9; not in table) and after adjustment (RR, 0.6; 95% CI, 0.4 – 0.9; Table 3). We examined the RR of relapse in GH-treated compared with nontreated patients in different prognostic subgroups; for instance, divided by age at diagnosis or by histology of
TABLE 1. Descriptive characteristics of patients treated with GH and those who did not receive this treatment GH-treated No. of patients
Sex Male Female Age at tumor diagnosis (yr) ⬍3.5 3.5– 6.4 6.5–9.9 10 –16 Histology Medulloblastoma Ependymoma Astrocytoma Other glioma Other Calendar period of tumor diagnosis ⬍1980 1980 – 4 ⱖ1985 Chemotherapyb None Any Hospital of tumor treatment Christie Great Ormond Street Royal Marsden Total a
Person-yr of follow-upa
Non GH-treated No. of first recurrences
No. of patients
Person-yr of follow-upa
No. of first recurrences
106 73
673 447
26 9
507 403
2567 2307
238 199
49 47 55 28
334 310 344 131
11 6 11 7
223 220 208 259
1134 1089 1086 1565
130 115 96 96
94 12 36 17 20
620 54 259 95 92
13 5 8 5 4
223 87 314 126 160
1357 477 2031 433 576
118 42 141 62 74
25 56 98
294 439 386
6 5 24
220 215 475
2386 1286 1203
82 111 244
119 56
725 384
20 14
495 349
3118 1502
197 222
70 96 13 179
572 496 52 1120
13 21 1 35
162 410 338 910
2217 1306 1352 4874
55 224 158 437
Person-years for treated patients are allocated to nontreated category before first treatment and to treated category thereafter, whereas number of patients refers to eventual treatment category. b Chemotherapy treatment status was unknown for 3 GH-treated patients and 65 non-GH-treated patients.
4446
JCE & M • 2000 Vol. 85 • No. 12
SWERDLOW ET AL.
TABLE 2. Relative risks of relapse in relation to prognostic variables Prognostic variable
Sex Male Female Age at tumor diagnosis (yr) ⬍3.5 3.5– 6.4 6.5–9.9 ⱖ10
Unadjusted relative risk (95% CI)
Adjusted relative risk (95% CI)a
1.0 0.9 (0.8 –1.1)
1.0 0.9 (0.8 –1.1)
1.0 1.0 (0.8 –1.2) 0.8 (0.6 –1.0) 0.7 (0.5– 0.9)b
1.0 1.1 (0.8 –1.4) 1.0 (0.8 –1.3) 0.8 (0.6 –1.1) 12 trend ⫽ 2.1; P ⫽ 0.15
Histology Medulloblastoma Ependymoma Astrocytoma Other glioma Other
1.0 1.2 1.1 1.6 1.3
Calendar period of tumor diagnosis ⬍1980 1980 –1984 ⱖ1985
1.0 1.4 (1.1–1.8)c 2.4 (1.9 –3.0)d
Time since tumor diagnosis (yr)e ⬍2 2– 4 ⱖ5
1.0 1.0 (0.8 –1.4) 1.5 (1.1–2.1)b 12 trend ⫽ 12.8; P ⬍ 0.001
1.0 0.5 (0.4 – 0.6)d 0.1 (0.07– 0.13)d
1.0 0.7 (0.6 – 0.8)d 0.2 (0.1– 0.3)d 12 trend ⫽ 118.7; P ⬍ 0.001
Chemotherapy None Any Hospital of tumor treatmentf Christie Great Ormond Street Royal Marsden
(0.9 –1.7) (0.9 –1.4) (1.2–2.2)b (1.0 –1.8)
1.0 1.5 (1.1–2.1)c 1.7 (1.3–2.2)d 1.7 (1.3–2.3)b 1.2 (0.8 –1.6) 42 heterogeneity ⫽ 21.2; P ⬍ 0.001
1.0 1.9 (1.6 –2.3)d
1.0 2.3 (1.8 –2.8)d
1.0 1.6 (1.1–2.2)c 1.1 (0.7–1.6)
1.0 1.6 (1.0 –2.4)c 1.0 (0.6 –1.5) 22 heterogeneity ⫽ 5.2; P ⫽ 0.08
a
Risks for each category adjusted for all other variables in table. P ⬍ 0.01. P ⬍ 0.05. d P ⬍ 0.001. e Analyzed with a Poisson model rather than Cox regression. f Analysis of follow-up 2 yr or more only. b c
brain tumor (Table 3). In no subgroup was there a significantly raised risk of relapse in relation to GH and in no instance was there significant heterogeneity between prognostic subgroups in the relation of GH to risk of relapse. Table 4 shows risk in relation to the time elapsed since first treatment with GH. Risk was not raised in any time category, and although it was greater at 5 and more yr after treatment than earlier, there was no significant trend in risk by time. The table also shows risks in relation to the cumulative duration for which GH treatment was given. Again, there were no raised risks and no significant trend in risk with longer treatment, although risk was greatest for patients with 5 and more yr of treatment. We also considered risk in relation to current treatment status (not in table), although this was potentially biased because treatment may be discontinued because of illness that is subsequently diagnosed as a recurrence; the RR for recurrence while on treatment was 0.5 (95% CI, 0.3– 0.7), and that for recurrence while on treatment or within 100 days of its cessation was 0.5 (95% CI, 0.3– 0.8). The overall RR of mortality for patients treated with GH was 0.5 (95% CI, 0.3– 0.8; Table 5). In analyses by time since first treatment, although the RR of death was not signifi-
cantly increased in any individual period, there was a significant trend of increasing RR with longer time. Of the 12 deaths that occurred in GH-treated patients at least 5 yr after first treatment, 5 were due to relapse of the tumor and 1 to a second cancer (of the colon). When we repeated the above analyses with follow-up ending at a fixed recent date rather than at the date of last clinical contact (see Subjects and Methods), the RRs of recurrence and death in GH-treated patients were generally about 0.1 greater than those shown in Tables 3–5, but none became significantly raised (not in tables); the overall RR of relapse for GH-treated patients was 0.7 (95% CI, 0.5–1.1), and that of death was 0.6 (95% CI, 0.4 –1.0). Discussion
GH has been used as replacement therapy in children with GH deficiency for almost 40 yr and has brought considerable benefits (1). More recently, treatment of adults with hypopituitarism has also shown psychological, physical, and biochemical benefits (2), and the range of indications for this treatment is growing. There is some evidence of beneficial
GH AND TUMOR RECURRENCE TABLE 3. Relative risks of relapse in relation to GH treatment within prognostic subgroups of patients
Prognostic variable
Sex Male Female Age at tumor diagnosis (yr) ⬍3.5 3.5– 6.4 6.5–9.9 ⱖ10 Histology Medulloblastoma Ependymoma Astrocytoma Other glioma Other Calendar period of tumor diagnosis ⬍1980 1980 –1984 ⱖ1985 Time since tumor diagnosis (yr) ⬍2 2– 4 ⱖ5 Chemotherapy None Any
Relative risk of relapse for GH-treated compared with non-GH-treated patients (95% CI)a
0.8 (0.5–1.3) 0.4 (0.2– 0.8)b 12 heterogeneity ⫽ 3.1; P ⫽ 0.08 0.5 (0.1–2.0) 0.7 (0.3–1.7) 0.7 (0.3–1.4) 0.7 (0.4 –1.4) 12 trend ⫽ 0.01; P ⫽ 0.94 0.5 (0.2– 0.9)c 1.1 (0.3–3.6) 0.7 (0.3–1.4) 1.0 (0.3–3.0) 0.8 (0.2–3.3) 42 heterogeneity ⫽ 3.8; P ⫽ 0.44 1.3 (0.5–3.5) 0.3 (0.1– 0.8)c 0.7 (0.4 –1.1) 12 trend ⫽ 0.01; P ⫽ 0.91 0.8 (0.4 –1.9) 0.5 (0.3– 0.9)c 0.9 (0.4 –2.0) 12 trend ⫽ 0.51; P ⫽ 0.47 0.7 (0.4 –1.2) 0.5 (0.3–1.0)c heterogeneity ⫽ 0.02; P ⫽ 0.88 2 1
Hospital of tumor treatment Christie Great Ormond Street Royal Marsden
Total
0.8 (0.4 –1.6) 0.6 (0.4 –1.0) 0.4 (0.1–2.8)
22 heterogeneity ⫽ 7.7; P ⫽ 0.02 0.6 (0.4 – 0.9)c
a Relative risks for each category adjusted for all other variables in table. b P ⬍ 0.01. c P ⬍ 0.05.
effects with larger, supraphysiological doses in short normal children (7), girls with Turner’s syndrome (6), and healthy older people (3), and there has been pressure for more widespread prescription (5). In this context it is important to ascertain whether the treatment is safe. There are several reasons for concern that GH might increase the risk of cancer recurrence or incidence. The hormone is mitogenic. In animal experiments it increases the incidence of leukemia (9) and solid tumors (8, 11), and in vitro it can promote replication of leukemic blast cells from human marrow (10). Increased chromosome fragility has been found in lymphocytes of treated children (25), and there have been over 40 case reports of leukemia occurring during or after treatment (13). An increased risk of leukemia has been shown in a cohort study in the U.S. of GH-treated patients (14),
4447
although many of the cohort members had other risk factors for malignancy. In Japanese and U.S. cohorts of GH recipients without such risk factors there was no apparent increased risk of leukemia (26, 27), and in the U.S. there was no increased risk of extracranial nonleukemic neoplasms overall, although risks were not analyzed by cancer site (28). Evidence that the GH/insulin-like growth factor I axis may be important in cancer etiology in man has come from recent reports of a strong positive association between insulin-like growth factor I levels and the risk of certain malignancies subsequently (15, 16). In follow-up of GH-treated patients, brain tumor recurrence has been one of the most common causes of death (17, 18), but without data on comparable patients who did not receive GH, it is unclear whether this indicates a raised risk. Similarly, reports of recurrence rates in GH-treated patients without comparison data on similar, but untreated, patients (29 –31) are difficult to interpret. Assessment of whether recurrence rates are affected by GH treatment requires comparison of recurrence rates in GH-treated and untreated subjects from the same centers; three studies have made such a comparison (19, 20, 22). Two of these made no allowance for likely differences between the two groups in confounding factors, such as duration of follow-up (19, 20), which, again, makes interpretation unclear. The one study that allowed for confounding variables (22) found no raised risk of recurrence in relation to treatment, but small numbers of subjects and relatively short follow-up left considerable uncertainty. The present analysis, which included almost 4 times as many GH-treated subjects, did not find increased risk associated with GH treatment, and the CI of the overall finding excluded substantial risk. Several potential biases in the study require consideration. The most important is that patients may have been selected for GH treatment on grounds associated with good prognosis. Adjustment for major known prognostic variables did not reveal increased recurrence rates in GH-treated patients, however, and indeed, this adjustment caused the RRs to decrease. If the low RRs in the treated patients were due to residual confounding by unmeasured or unknown prognostic variables, then these risks would be expected to rise to unity with longer follow-up, as the initial selection wore off. This was the pattern seen for both recurrences and deaths, and it is notable that the RRs for the longest follow-up period did not increase appreciably above unity. Thus, although it is impossible to be certain that confounding by unknown prognostic factors has not disguised an increased risk from GH therapy, the evidence does not suggest that this is so. Bias might also potentially have occurred if there had been incomplete ascertainment of recurrences, and this had differed between the GH-treated and untreated patients. This is extremely unlikely, however, because all clinical or radiological recurrences would have been dealt with by the primary treatment centers, and although more intensive surveillance might have accelerated diagnosis of recurrences by a few weeks or months, it would not have revealed lesions that would otherwise have been undiscovered. Furthermore, the mortality results, to which such potential biases would not apply, were similar to those for recurrences. A final potential source of bias was the use of the date
4448
JCE & M • 2000 Vol. 85 • No. 12
SWERDLOW ET AL.
TABLE 4. Relative risks of recurrence in relation to time since first GH treatment and duration of GH treatment Person-yr of follow-up
No. of relapses
Relative risk of recurrence (95% CI)a
Time since first GH treatment No treatment ⬍2 2– 4 ⱖ5
4793 324 365 465
434 19 8 8
1.0 0.6 (0.4 –1.0)b 0.5 (0.3–1.2) 0.9 (0.4 –2.2) 2 1 trend ⫽ 0.43; P ⫽ 0.51c
Duration of GH treatment No treatment ⬍2 2– 4 ⱖ5
4793 392 535 226
434 23 7 5
1.0 0.7 (0.4 –1.1) 0.4 (0.2– 0.9)b 0.9 (0.3–2.6) 12 trend ⫽ 0.02; P ⫽ 0.88c
Duration (yr)
a Adjusted for sex, age at tumor diagnosis, histology of tumor, time since tumor diagnosis, calendar period of tumor diagnosis, whether chemotherapy included in the primary treatment, and hospital of treatment. b P ⬍ 0.05. c Trend tests calculated excluding the no treatment category. (Comparisons between untreated and treated are covered in Table 3.)
TABLE 5. Relative risks of death in relation to time since first GH treatment and duration of GH treatment Person-yr of follow-up
No. of deaths
Time since first GH treatment No treatment ⬍2 2– 4 ⱖ5
5488 354 405 504
378 10 6 12
1.0 0.4 (0.2– 0.7)b 0.3 (0.1– 0.8)c 1.1 (0.5–2.1) 12 trend ⫽ 4.2; P ⫽ 0.04d
Duration of GH treatment No treatment ⬍2 2– 4 ⱖ5
5488 440 584 239
378 15 9 4
1.0 0.5 (0.3– 0.9)c 0.5 (0.2– 0.9)c 0.5 (0.2–1.5) 12 trend ⫽ 0.02; P ⫽ 0.88d
Duration (yr)
Relative risk of death (95% CI)a
a Adjusted for sex, age at tumor diagnosis, histology of tumor, time since tumor diagnosis, calendar period of tumor diagnosis, whether chemotherapy included in the primary treatment, and hospital of treatment. b P ⬍ 0.01. c P ⬍ 0.05. d Trend tests calculated excluding the no treatment category. (Comparisons between untreated and treated are covered in Table 3.)
of last clinical contact as the date of exit from risk for patients who had not died or relapsed. This was unavoidable because it would have been infeasible to update follow-up individually by personal contact for the entire cohort. The use of this end date gave potential for bias, because routine follow-up tended to occur more frequently for GH-treated than for untreated patients, and thus their date of last clinical contact tended to be more recent. The person-years since last contact are likely to have been virtually free of recurrences or deaths, because these events would have brought the patient to clinical attention. Thus, loss from the study of relapse-free personyears would tend to increase apparent relapse rates, and if this occurred more for untreated than treated patients, it would lead to underestimation of the RRs in treated patients. We found, however, that when the analysis was repeated with follow-up until the end of the study unless patients were known to have died or relapsed (i.e. probably overcompensating for any bias), there were still no substantially increased risks in relation to GH treatment. If there were an effect of GH on tumor recurrence, it is unknown how soon it would occur after exposure, and how it would relate to the extent of exposure; with regard to the
latter, although the dosage of GH has been fairly standard, the cumulative duration for which treatment has been given has varied greatly between patients. We therefore analyzed risks in relation to time since first treatment and cumulative duration of treatment. The latter showed no relation to risk, and although the former showed a rising trend with time, this started from a RR below unity and did not rise significantly above it, so it may have been due to the wearing off of initial selective effects, as discussed above. In summary, the results provide evidence, based on far more person-years of treatment and follow-up than previously, that recurrence rates of brain tumor are not substantially increased after GH replacement therapy. The data for the period 5 and more yr after first treatment leave a need for further surveillance, however, both because the CIs for risks in this period were wider, and because the significant rising trend in mortality RRs in the GH-treated subjects, although not in itself a cause for alarm at present, leaves open the possibility of an increased risk. In addition, the possibility that GH treatment may affect de novo cancer incidence, particularly when used in larger than replacement doses, needs to be clarified in much larger cohort studies.
GH AND TUMOR RECURRENCE Acknowledgments We thank Ms. S. Ashley, Dr. G. Birch and the North West Childhood Cancer Registry, Dr. B. Brennan, Dr. B. L. de Stavola, Dr. H. R. Gattamaneni, and Dr. A. Ogilvy-Stuart for their assistance and advice.
References 1. Bridges NA, Brook CGD. 1995 The management of short stature. In: Brook CGD, ed. Clinical paediatric endocrinology, 3rd Ed. Oxford: Blackwell; 173–186. 2. Salomon F, Cuneo RC, Hesp R, So¨nksen PH. 1989 The effects of treatment with recombinant human growth hormone on body composition and metabolism in adults with growth hormone deficiency. N Engl J Med. 321:1797–1803. 3. Vance ML. 1990 Growth hormone for the elderly? N Engl J Med. 323:52–54. 4. Jørgensen JOL, Christiansen JS. 1993 Growth hormone therapy. Brave new senescence: GH in adults. Lancet. 341:1247–1248. 5. Lancet. 1994 Little bigger, little better. Lancet. 344:627– 628. 6. Taback SP, Collu R, Deal CL, et al. 1996 Does growth-hormone supplementation affect adult height in Turner’s syndrome? Lancet. 348:25–27. 7. McCaughey ES, Mulligan J, Voss LD, Betts PR. 1998 Randomised trial of growth hormone in short normal girls. Lancet. 351:940 –944. 8. Moon HD, Simpson ME, Li CH, Evans HM. 1950 Neoplasms in rats treated with pituitary growth hormone. I. Pulmonary and lymphatic tissues. Cancer Res. 10:297–308. 9. Rogers PC, Komp D, Rogol A, Sabio H. 1977 Possible effect of growth hormone on development of acute lymphoblastic leukaemia. Lancet. 2:.434 – 445. 10. Zadik Z, Estrov Z, Karov Y, Hahn T, Barak Y. 1993 The effect of growth hormone and IGF-I on clonogenic growth of hematopoietic cells in leukemic patients during active disease and during remission–a preliminary report. J Pediatr Endocrinol. 6:79 – 83. 11. Ogilvy-Stuart AL. 1995 Safety of growth hormone after treatment of a childhood malignancy. Horm Res. 44(Suppl 3):73–79. 12. Ng ST, Zhou J, Adesanya OO, Wang J, LeRoith D, Bondy CA. 1997 Growth hormone treatment induces mammary gland hyperplasia in aging primates. Nat Med. 3:1141–1144. 13. Shalet SM, Brennan BM, Reddingius RE. 1997 Growth hormone therapy and malignancy. Horm Res. 48(Suppl 4):29 –32. 14. Fradkin JE, Mills JL, Schonberger LB, et al. 1993 Risk of leukemia after treatment with pituitary growth hormone. JAMA. 270:2829 –2832. 15. Chan JM, Stampfer MJ, Giovanucci E, et al. 1998 Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study. Science. 279:563–566.
4449
16. Hankinson SE, Willett WC, Colditz GA, et al. 1998 Circulating concentrations of insulin-like growth factor-I and risk of breast cancer. Lancet. 351:1393–1396. 17. Buchanan CR, Preece MA, Milner RDG. 1991 Mortality, neoplasia, and Creutzfeldt-Jakob disease in patients treated with human pituitary growth hormone in the United Kingdom. Br Med J. 302:824 – 828. 18. Taback SP, Dean HJ, members of the Canadian Growth Hormone Advisory Committee. 1996 Mortality in Canadian children with growth hormone (GH) deficiency receiving GH therapy 1967–1992. J Clin Endocrinol Metab. 81:1693–1696. 19. Arslanian SA, Becker DJ, Lee PA, Drash AL, Foley Jr TP. 1985 Growth hormone therapy and tumor recurrence. Findings in children with brain neoplasms and hypopituitarism. Am J Dis Child. 139:347–350. 20. Rodens KP, Kaplan SL, Grumbach MM, Teller WM. 1987 Does growth hormone therapy increase the frequency of tumor recurrence in children with brain tumors? Acta Endocrinol (Copenh). 283(Suppl):188 –189. 21. Clayton PE, Shalet SM, Gattamaneni HR, Price DA. 1987 Does growth hormone cause relapse of brain tumours? Lancet. 1:711–713. 22. Ogilvy-Stuart AL, Ryder WDJ, Gattamaneni HR, Clayton PE, Shalet SM. 1992 Growth hormone and tumour recurrence. Br Med J. 304:1601–1605. 23. Cox DR. 1972 Regression models and life-tables. J R Stat Soc (B). 34:187–220. 24. Clayton D, Hills M. 1993 Statistical models in epidemiology. Oxford: Oxford University Press. 25. Tedeschi B, Spadoni GL, Sanna ML, et al. 1993 Increased chromosome fragility in lymphocytes of short normal children treated with recombinant human growth hormone. Hum Genet. 91:459 – 463. 26. Nishi Y, Tanaka T, Takano K, et al. 1999 Recent status in the occurrence of leukemia in growth hormone-treated patients in Japan. J Clin Endocrinol Metab. 84:1961–1965. 27. Allen DB, Rundle AC, Graves DA, Blethen SL. 1997 Risk of leukemia in children treated with human growth hormone: review and reanalysis. J Pediatr. 131:S32–S36. 28. Tuffli GA, Johanson A, Rundle AC, Allen DB. 1995 Lack of increased risk for extracranial, nonleukemic neoplasms in recipients of recombinant deoxyribonucleic acid growth hormone. J Clin Endocrinol Metab. 80:1416 –1422. 29. Chatelain PG, Wilton P. 1992 KABI Pharmacia International Growth Study: adverse events report. KIGS Biannu Rep No. 8:42– 47. 30. Wilton P, Price DA. 1992 Rate of relapse of craniopharyngioma in children treated with growth hormone. KIGS Biannu Rep No. 8:48 –54. 31. Moshang T Jr, Rundle AC, Graves DA, Nickas J, Johanson A, Meadows A. 1996 Brain tumor recurrence in children treated with growth hormone: the National Cooperative Growth Study experience. J Pediatr. 128:S4 –S7.