1270
other larger series with more than 10 years of follow-up. It might nevertheless be prudent that all patients receiving adjuvant chemotherapy do so in the setting of carefully controlled long-term follow-up studies. R. T. D. OLIVER J. Y. H. ONG M. A. RAJA Department of Medical Oncology and Haematology, P. SPERANDIO Medical College of The Royal London, B. GIBBONS St Bartholomews, and Middlesex Hospitals, London E1 2AD, UK M. WALKER RTD, Dhaliwal HS, Hope Stone HF, Blandy JP. Short-course etoposide, bleomycin, and cisplatin in the treatment of metastatic germ cell tumours: appraisals of its potential as adjuvant chemotherapy for stage I testis tumours. Br J Urol 1988; 61: 53-58 2. Pont J, Holte W, Kosak D, et al. Risk adapted treatment choice in stage 1 non-seminomatous testicular germ cell cancer by regarding vascular invasion in the primary tumour, a prospective trial. J Clin Oncol 1990; 8: 16-20. 3 Oliver RTD Clues from natural history and results of treatment supporting the monoclonal origin of germ cell tumours. Cancer Surv 1990, 9: 338-68
with drugs targeting at DNA-topoisomerase II, we observed balanced aberrations at bands llq23 and 21q21 mainly in leukaemias developing a very short period after start of therapy and other chromosome abnormalities in later occurring cases, such as those described by Oliver et al.
JENS PEDERSEN-BJERGAARD Departments of Haematology and Oncology, Righospitalet, DK-2100 Copenhagen, Denmark, and Department of Biostatistics, State Seruminstitute, Copenhagen
GEDSKE DAUGAARD STEEN WERNER HANSEN PREBEN PHILIP SEVERIN OLESEN LARSEN MIKAEL RØRTH
1. Oliver
* These letters have been shown to Dr Pedersen-Bjergaard and colleagues, whose reply follows.-ED. L. SIR,-Dr Donatini and Dr Krupp conclude that the role of etoposide in the development of MDS-AML remains unproven. Referring to other studies, some discussed in our paper, they claim that the development of leukaemia simply reflects the natural history of germ-cell tumours, in which there is sometimes a yolk-sac component. T-cell phenotype is discussed (though the study cited does not mention this) and statistical procedures and methods different from those we used are recommended. Strong arguments for a causal relation between etoposide and MDS/AML are: (1) Single sporadic cases apart, leukaemia has not previously been reported in larger unselected series of patients with germ-cell tumours treated with cisplatin, bleomycin and vinblastine, despite long follow-up.1-3 (2) After the introduction of etoposide, MDS/AML is now being reported, in a dose-dependent fashion in our study. (3) Etoposide and teniposide have been shown to be as leukaemogenic in other tumour types.4’s (4) 3 patients with leukaemia in our study presented with chromosome aberrations characteristic of leukaemias related to therapy with etoposide and other cytostatic agents targeted at DNA-topoisomerase 11. It seems unlikely that our cases of leukaemia were of the type reported by Nichols et al/ deriving from a highly selected group of patients and related to the natural history of germ-cell tumours: none of the leukaemia patients presented with germ-cell tumour of the mediastinal type; none had yolk-sac elements in the initial histological preparation and only 1 (case 2) had increased serum a-fetoprotein; the latent period from start of chemotherapy, the clinical features, and the histological subtype of AML in general differ greatly from the pattern observed by Nichols et al; and none of our leukaemic patients had chromosome aberrations characteristic for germ-cell tumours.8 The statistical methods we used are standard. Statistical methods for evaluating the effect of therapy given over time in terms of survival analysis have not, to our knowledge, been developed. The normal Danish population had to be used for comparison in estimates of the relative risk of leukaemia to find out whether the leukaemias are merely fortuitous. Since a significant risk of MDS/AML has not been observed in larger cohorts of patients treated with cisplatin, vinblastine, and bleomycin, the estimate suggested by Donatini and Krupp would result in a relative risk of infinity, which is meaningless. Patients treated with high-dose etoposide had a poorer prognosis than those on standard doses but we allowed for that. Professor Oliver and colleagues confirm an increased risk of AML after inclusion of etoposide in the therapy of patients with germ-cell tumours. Their leukaemic patients had received lower cumulative doses of etoposide than ours. Their leukaemias appeared with a long latency and did not present balanced chromosome aberrations involving bands llq23 and 21q22. In a recent study (unpublished) also discussing leukaemia after therapy
1.
2
Pedersen-Bjergaard J, Daugaard G, Hansen SW, et al. Increased risk of myelodysplasia and leukaemia after etoposide, cisplatin, and bleomycin for germ-cell tumours Lancet 1991; 338: 359-63. Roth BJ, Greist A, Kubihs PS, et al. Cisplatin-based combination chemotherapy for disseminated germ cell tumors long-term follow-up J Clin Oncol 1988; 6:
1239-47. 3. Roth BJ, Einhom LH, Greist A. Long-term complications of cisplatin-based chemotherapy for testis cancer. Semin Oncol 1988; 15: 345-50. 4 Ratain MJ, Kaminer LS, Bitran JD, et al. Acute nonlymphocytic leukemia following etoposide and cisplatin combination chemotherapy for advanced non-small-cell carcinoma of the lung Blood 1987; 70: 1412-17 5. Pui C-H, Behm FG, Raimondi SC, et al Secondary acute myeloid leukemia in children treated for acute lymphoid leukemia. N Engl J Med 1989; 321: 136-42. 6 Pedersen-Bjergaard J, Philip P Balanced translocations involving chromosome bands 11q23 and 21q22 are highly characteristic of myelodysplasia and leukemia following therapy with cytostatic agents targeting at DNA-topoisomerase II Blood
1991; 78: 1147-48. CR, Roth BJ, Heerema N, et al. Hematologic neoplasia associated with primary mediastinal germ-cell tumors N Engl J Med 1990; 322: 1425-29. 8. Bosl GJ, Dmitrovsky E, Reuter VE, et al Isochromosome of chromosome 12. clinically useful marker for male germ cell tumors J Nat Cancer Inst 1989; 81: 7 Nichols
1874-78.
Development after in-utero exposure to mifepristone SjR,—Dr Pons and colleagues (Sept 21, p 763) describe a fetus malformed after exposure to mifepristone in early pregnancy. This is the only such case. It happened in 1987 and was reported in Nature in 1989 (338: 110) and to the health authorities concerned. The patient received 600 mg mifepristone, not 400 mg as reported by Pons et al. At that time, the use of mifepristone alone was being investigated for termination of early pregnancy and the patient was, therefore, not due to receive a prostaglandin analogue. In this case several of the deformities seen seem consistent with development before the administration of mifepristone. There have been no abnormal babies born to women who have taken mifepristone for termination of pregnancy and who subsequently carried their pregnancies to term.1 Roussel Santé R &D, 93230 Romainville, France, and Roussel Uclaf, Paris
ANDRÉ ULMANN
Roussel Laboratories, Denham, UK
JOAN BARNARD
IAN RUBIN
BH, Lees DAR, Bjornsson S, et al. Normal development after mifepristone m early pregnancy Lancet 1990; 336: 257-58.
1 Lim
exposure to
Insulin clearance and diabetes SIR,-Dr Cruickshank and colleagues (Oct 5, p 842) suggest that ethnic differences in the secretion and hepatic clearance of insulin underlie differences in diabetes prevalence between European, South Asian, and Afro-Caribbean peoples. They use logistic regression models to show that newly detected diabetes is associated with raised fasting plasma concentrations of insulin and C-peptide. This is not surprising. Most diabetics have fasting hyperglycaemia that stimulates insulin secretion; under these conditions a rise in insulin and C-peptide is an effect of diabetes not a cause. When testing hypotheses on the aetiology of diabetes in a cross-sectional study, it is as inappropriate to include insulin or C-peptide as it would be to include glucose in the model as an explanatory variable. In such a model, fasting glucose would "explain" associations with diabetes (and the 2 h glucose would be even better) but the absurdity would be at once apparent. Cruickshank and colleagues’ use of
other larger series with more than 10 years of follow-up. It might nevertheless be prudent that all patients receiving adjuvant chemotherapy do so in the setting of carefully controlled long-term follow-up studies. R. T. D. OLIVER J. Y. H. ONG M. A. RAJA Department of Medical Oncology and Haematology, P. SPERANDIO Medical College of The Royal London, B. GIBBONS St Bartholomews, and Middlesex Hospitals, London E1 2AD, UK M. WALKER RTD, Dhaliwal HS, Hope Stone HF, Blandy JP. Short-course etoposide, bleomycin, and cisplatin in the treatment of metastatic germ cell tumours: appraisals of its potential as adjuvant chemotherapy for stage I testis tumours. Br J Urol 1988; 61: 53-58 2. Pont J, Holte W, Kosak D, et al. Risk adapted treatment choice in stage 1 non-seminomatous testicular germ cell cancer by regarding vascular invasion in the primary tumour, a prospective trial. J Clin Oncol 1990; 8: 16-20. 3 Oliver RTD Clues from natural history and results of treatment supporting the monoclonal origin of germ cell tumours. Cancer Surv 1990, 9: 338-68
with drugs targeting at DNA-topoisomerase II, we observed balanced aberrations at bands llq23 and 21q21 mainly in leukaemias developing a very short period after start of therapy and other chromosome abnormalities in later occurring cases, such as those described by Oliver et al.
JENS PEDERSEN-BJERGAARD Departments of Haematology and Oncology, Righospitalet, DK-2100 Copenhagen, Denmark, and Department of Biostatistics, State Seruminstitute, Copenhagen
GEDSKE DAUGAARD STEEN WERNER HANSEN PREBEN PHILIP SEVERIN OLESEN LARSEN MIKAEL RØRTH
1. Oliver
* These letters have been shown to Dr Pedersen-Bjergaard and colleagues, whose reply follows.-ED. L. SIR,-Dr Donatini and Dr Krupp conclude that the role of etoposide in the development of MDS-AML remains unproven. Referring to other studies, some discussed in our paper, they claim that the development of leukaemia simply reflects the natural history of germ-cell tumours, in which there is sometimes a yolk-sac component. T-cell phenotype is discussed (though the study cited does not mention this) and statistical procedures and methods different from those we used are recommended. Strong arguments for a causal relation between etoposide and MDS/AML are: (1) Single sporadic cases apart, leukaemia has not previously been reported in larger unselected series of patients with germ-cell tumours treated with cisplatin, bleomycin and vinblastine, despite long follow-up.1-3 (2) After the introduction of etoposide, MDS/AML is now being reported, in a dose-dependent fashion in our study. (3) Etoposide and teniposide have been shown to be as leukaemogenic in other tumour types.4’s (4) 3 patients with leukaemia in our study presented with chromosome aberrations characteristic of leukaemias related to therapy with etoposide and other cytostatic agents targeted at DNA-topoisomerase 11. It seems unlikely that our cases of leukaemia were of the type reported by Nichols et al/ deriving from a highly selected group of patients and related to the natural history of germ-cell tumours: none of the leukaemia patients presented with germ-cell tumour of the mediastinal type; none had yolk-sac elements in the initial histological preparation and only 1 (case 2) had increased serum a-fetoprotein; the latent period from start of chemotherapy, the clinical features, and the histological subtype of AML in general differ greatly from the pattern observed by Nichols et al; and none of our leukaemic patients had chromosome aberrations characteristic for germ-cell tumours.8 The statistical methods we used are standard. Statistical methods for evaluating the effect of therapy given over time in terms of survival analysis have not, to our knowledge, been developed. The normal Danish population had to be used for comparison in estimates of the relative risk of leukaemia to find out whether the leukaemias are merely fortuitous. Since a significant risk of MDS/AML has not been observed in larger cohorts of patients treated with cisplatin, vinblastine, and bleomycin, the estimate suggested by Donatini and Krupp would result in a relative risk of infinity, which is meaningless. Patients treated with high-dose etoposide had a poorer prognosis than those on standard doses but we allowed for that. Professor Oliver and colleagues confirm an increased risk of AML after inclusion of etoposide in the therapy of patients with germ-cell tumours. Their leukaemic patients had received lower cumulative doses of etoposide than ours. Their leukaemias appeared with a long latency and did not present balanced chromosome aberrations involving bands llq23 and 21q22. In a recent study (unpublished) also discussing leukaemia after therapy
1.
2
Pedersen-Bjergaard J, Daugaard G, Hansen SW, et al. Increased risk of myelodysplasia and leukaemia after etoposide, cisplatin, and bleomycin for germ-cell tumours Lancet 1991; 338: 359-63. Roth BJ, Greist A, Kubihs PS, et al. Cisplatin-based combination chemotherapy for disseminated germ cell tumors long-term follow-up J Clin Oncol 1988; 6:
1239-47. 3. Roth BJ, Einhom LH, Greist A. Long-term complications of cisplatin-based chemotherapy for testis cancer. Semin Oncol 1988; 15: 345-50. 4 Ratain MJ, Kaminer LS, Bitran JD, et al. Acute nonlymphocytic leukemia following etoposide and cisplatin combination chemotherapy for advanced non-small-cell carcinoma of the lung Blood 1987; 70: 1412-17 5. Pui C-H, Behm FG, Raimondi SC, et al Secondary acute myeloid leukemia in children treated for acute lymphoid leukemia. N Engl J Med 1989; 321: 136-42. 6 Pedersen-Bjergaard J, Philip P Balanced translocations involving chromosome bands 11q23 and 21q22 are highly characteristic of myelodysplasia and leukemia following therapy with cytostatic agents targeting at DNA-topoisomerase II Blood
1991; 78: 1147-48. CR, Roth BJ, Heerema N, et al. Hematologic neoplasia associated with primary mediastinal germ-cell tumors N Engl J Med 1990; 322: 1425-29. 8. Bosl GJ, Dmitrovsky E, Reuter VE, et al Isochromosome of chromosome 12. clinically useful marker for male germ cell tumors J Nat Cancer Inst 1989; 81: 7 Nichols
1874-78.
Development after in-utero exposure to mifepristone SjR,—Dr Pons and colleagues (Sept 21, p 763) describe a fetus malformed after exposure to mifepristone in early pregnancy. This is the only such case. It happened in 1987 and was reported in Nature in 1989 (338: 110) and to the health authorities concerned. The patient received 600 mg mifepristone, not 400 mg as reported by Pons et al. At that time, the use of mifepristone alone was being investigated for termination of early pregnancy and the patient was, therefore, not due to receive a prostaglandin analogue. In this case several of the deformities seen seem consistent with development before the administration of mifepristone. There have been no abnormal babies born to women who have taken mifepristone for termination of pregnancy and who subsequently carried their pregnancies to term.1 Roussel Santé R &D, 93230 Romainville, France, and Roussel Uclaf, Paris
ANDRÉ ULMANN
Roussel Laboratories, Denham, UK
JOAN BARNARD
IAN RUBIN
BH, Lees DAR, Bjornsson S, et al. Normal development after mifepristone m early pregnancy Lancet 1990; 336: 257-58.
1 Lim
exposure to
Insulin clearance and diabetes SIR,-Dr Cruickshank and colleagues (Oct 5, p 842) suggest that ethnic differences in the secretion and hepatic clearance of insulin underlie differences in diabetes prevalence between European, South Asian, and Afro-Caribbean peoples. They use logistic regression models to show that newly detected diabetes is associated with raised fasting plasma concentrations of insulin and C-peptide. This is not surprising. Most diabetics have fasting hyperglycaemia that stimulates insulin secretion; under these conditions a rise in insulin and C-peptide is an effect of diabetes not a cause. When testing hypotheses on the aetiology of diabetes in a cross-sectional study, it is as inappropriate to include insulin or C-peptide as it would be to include glucose in the model as an explanatory variable. In such a model, fasting glucose would "explain" associations with diabetes (and the 2 h glucose would be even better) but the absurdity would be at once apparent. Cruickshank and colleagues’ use of
