121
that the results of the drug consumption survey done in Emilia Romagna region can be extrapolated to the whole of Italy. With respect to calcitonin in elderly patients, most open or double-blind, controlled, randomised studies that confirm the drug’s effect on bone mass were done in subjects with established osteoporosis, with a mean age of about 65 years and proven vertebral ttacrures.2.3A large case-control study (MEDOS) on the incidence of risk factors for hip fracture was done in 1989 in Portugal, Spain, France, Italy, Greece, and Turkey, with the collaboration of WHO and the European Foundation for Osteoporosis. The results (presented at the llth international conference on calcium regulating hormones in Florence, in April, 1992) demonstrate that calcitonin combined with calcium significantly reduces the risk of hip fracture. Consequently, if the Emilia Romagna analysis, which indicates that calcitonin treatment lasts on average 45 days, were typical of all regions in Italy, then an increase in the duration of treatment could be recommended to further reduce fractures in elderly patients. The analgesic activity of calcitonin is well known.4,5 Finally, it has been proven that calcitonin inhibits oestrogendependent bone loss, both in physiological menopause6,’ and after oophorectomy,8 and provides an important alternative in the prevention of menopause-related bone loss in women who are unable or unwilling to take oestrogens.I Thus published data are certainly not consistent with Magrini and colleagues’ doubts that calcitonin is among the drugs on the Italian market that are devoid of proven clinical efficacy. Internal Medicine II, Medical Clinic, Università degli Studi di Roma "La
Sapienza",
00161, Rome, Italy,
G. F. MAZZUOLI C. GENNARI
and Istituto di Patologia Medica, Università degli Studí di Siena 1. Consensus
Development Conference: prophylaxis Osteoporosis Int 1991; 1: 114-17.
2. Mazzuoli
G, Passeri M, Gennari C,
et
and
treatment
osteoporosis.
al. Effects of salmon calcitonin in
postmenopausal osteoporosis: a controlled double blind clinical study. Calcif Tissue Int 1986; 38: 3-8. 3. Overgaard K, Riis BJ, Christiansen C, et al. Nasal calcitonin for treatment of established osteoporosis. Clin Endocrinol 1989; 30: 435-42. 4. Gennari C, Agnusdei D, Camporeale A. Use of calcitonin in the treatment of bone pain associated with osteoporosis. Calcif Tissue Int 1991; 49: S9-S13. 5. Lyritis GP, Tsakalakos N, Magiasis B, et al. Analgesic effect of salmon calcitonin in osteosporotic vertebral fractures: a double-blind placebo-controlled clinical study. Calcif Tissue Int 1991; 49: 369-72. 6. MacIntyre I, Stevenson JC, Whitehead MI, et al. Calcitonin for prevention of postmenopausal bone loss. Lancet 1988; i: 900-02. 7 Gennari C, Agnusdei D, Montagnani M, et al. An effective regimen of intranasal salmon calcitonin in early postmenopausal bone loss. Calcif Tissue Int 1992; 50: 381-83. 8. Mazzuoli GF, Tabolli S, Bigi F, et al. Effects of salmon calcitonin induced by ovariectomy. Calcif Tissue Int 1990; 47: 209-14.
on
the bone loss
Deaths from tobacco SIR,-Professor Peto and colleagues’ report (May 23, p 1268) provides indirect forecasts of mortality from smoking in the developed countries. Although more narrow in scope and using a different (and what is claimed to be a more robust) methodology than the study by some of the same authors reported at the Perth World Smoking and Health Conference in 1990, this is nevertheless one of the most ambitious exercises in quantitative epidemiology ever to be published. Since these and similar forecasts are likely to determine public policy, it is important that what will be regarded by decision-makers as expert advice is in fact well-founded. Forecasting is a notoriously fragile process and in view of the widespread scepticism about pronouncements in this field,l it is not unreasonable to look very carefully at the provenance of the latest set of forecasts.
Any forecasting procedure should include assumptions about the underlying relations, which are most conveniently considered in terms of a mathematical model. The resulting forecasts will be determined by the content and structure of the underlying model. However, the biological and physical mechanisms that determine death in man are poorly understood, especially with respect to possible agents such as tobacco. All types of smoking-related disease also arise in people who have never been exposed to tobacco smoke, and it is clear that other influences
or
confounding factors may be
involved. Neither the identity nor the relative importance of such factors in relation to any particular cause of death has been established on a definitive basis, although many associations (which may or may not be causal) have been reported. Nor is it known how the confounding factors may combine with any effects of tobacco to produce changes in the health status of the individual. On general grounds, it is very likely that the set of relevant confounding factors will vary from disease to disease as well as from society to society. Potential confounding factors may themselves be associated with smoking habits. For example, in societies where the smoking habit is well-established (which in general are the only societies for which evidence can be obtained), smoking is associated with low social class. The innate characterisics of the individual must also play a part, but the only available quantitative evidence about these matters is derived from uncontrolled epidemiological studies in which smokers and non-smokers are self-selected groups. There is no reason to suppose that the smoking habit is determined independently of personal characteristics or confounding factors. On all these grounds, it is clear that any model that is devised as a basis for generating forecasts on the basis of available knowledge is tentative. None of the models used by Peto and his colleages in recent publications seem to take account of these complex issues, and no evidence is presented that the simple models that are used do provide a realistic representation of the factors involved. Having summarised the experience of a "survey" population in terms of a mathematical model, Peto et al apply the results with the same model to a different target population, often in a different era and possibly in a different part of the world. This implies that the extent of any hazard associated with smoking is exactly the same in both survey and target populations, as are the confounding factors and the associations between the elements of the model. In view of the likely presence of differences in the properties and amounts of tobacco smoked, in the way the tobacco is smoked, in the confounding variables, and in the self-selection criteria for smoking, this is indeed a major act of faith. Peto’s indirect method of forecasting is thus subject to these general criticisms. In addition, the use of lung cancer rates as a yardstick involves further unsupported and arguably implausible hypotheses, including the assumptions: that lung cancer rates in non-smokers as reported in national statistical returns must be the same in the survey and target populations; that any relation between smoking, confounding factors, and mortality are of the same form for all relevant diseases; and that different types of product affect specific diseases in the same way. Furthermore, the arbitrary halving of the calculated excess risk associated with smoking may be generous to the tobacco industry, but does not seem to have any scientific foundation. A detailed analysis of the difficulties of assessing smoking-related mortality is being published elsewhere2 and will not be repeated here. Nevertheless, with the evident uncertainties of this process, it is impossible to avoid the question as to whether the calculation by any method of forecasts of mortality due to smoking on the basis of available information can ever be a proper foundation for public policy or anything more than an exercise in speculative arithmetic. Ide House, Ide, Exeter EX2 9RB, UK
JOHN R. ASHFORD
1. Levin B. Rise of the cigarette police. The Times, June 1, 1992. 2. Ashford JR. Problems in assessing smoking-related mortality. Disorders (in press).
J Smoking-Related
Serological evidence for congenital transmission of human
herpesvirus 6
SIR,-Human herpesvirus 6 (HPV-6) is the causative agent of exanthem subitum.1 Seroprevalence surveys indicate that more than 90% of children have evidence of HHV-6 infection before age 2.2 The virus has been detected in oropharynx and salivary glands of healthy adults, leading to speculation that HHV-6 is transmitted through contact with infected oral secretions.3 The possibility of congenital transmission of HHV-6, in analogy to other human herpesviruses, has been considered.’ We have retrospectively evaluated stored cord-blood specimens for evidence of congenital HHV-6 infection. 799 cord-blood sera were chosen randomly from
that the results of the drug consumption survey done in Emilia Romagna region can be extrapolated to the whole of Italy. With respect to calcitonin in elderly patients, most open or double-blind, controlled, randomised studies that confirm the drug’s effect on bone mass were done in subjects with established osteoporosis, with a mean age of about 65 years and proven vertebral ttacrures.2.3A large case-control study (MEDOS) on the incidence of risk factors for hip fracture was done in 1989 in Portugal, Spain, France, Italy, Greece, and Turkey, with the collaboration of WHO and the European Foundation for Osteoporosis. The results (presented at the llth international conference on calcium regulating hormones in Florence, in April, 1992) demonstrate that calcitonin combined with calcium significantly reduces the risk of hip fracture. Consequently, if the Emilia Romagna analysis, which indicates that calcitonin treatment lasts on average 45 days, were typical of all regions in Italy, then an increase in the duration of treatment could be recommended to further reduce fractures in elderly patients. The analgesic activity of calcitonin is well known.4,5 Finally, it has been proven that calcitonin inhibits oestrogendependent bone loss, both in physiological menopause6,’ and after oophorectomy,8 and provides an important alternative in the prevention of menopause-related bone loss in women who are unable or unwilling to take oestrogens.I Thus published data are certainly not consistent with Magrini and colleagues’ doubts that calcitonin is among the drugs on the Italian market that are devoid of proven clinical efficacy. Internal Medicine II, Medical Clinic, Università degli Studi di Roma "La
Sapienza",
00161, Rome, Italy,
G. F. MAZZUOLI C. GENNARI
and Istituto di Patologia Medica, Università degli Studí di Siena 1. Consensus
Development Conference: prophylaxis Osteoporosis Int 1991; 1: 114-17.
2. Mazzuoli
G, Passeri M, Gennari C,
et
and
treatment
osteoporosis.
al. Effects of salmon calcitonin in
postmenopausal osteoporosis: a controlled double blind clinical study. Calcif Tissue Int 1986; 38: 3-8. 3. Overgaard K, Riis BJ, Christiansen C, et al. Nasal calcitonin for treatment of established osteoporosis. Clin Endocrinol 1989; 30: 435-42. 4. Gennari C, Agnusdei D, Camporeale A. Use of calcitonin in the treatment of bone pain associated with osteoporosis. Calcif Tissue Int 1991; 49: S9-S13. 5. Lyritis GP, Tsakalakos N, Magiasis B, et al. Analgesic effect of salmon calcitonin in osteosporotic vertebral fractures: a double-blind placebo-controlled clinical study. Calcif Tissue Int 1991; 49: 369-72. 6. MacIntyre I, Stevenson JC, Whitehead MI, et al. Calcitonin for prevention of postmenopausal bone loss. Lancet 1988; i: 900-02. 7 Gennari C, Agnusdei D, Montagnani M, et al. An effective regimen of intranasal salmon calcitonin in early postmenopausal bone loss. Calcif Tissue Int 1992; 50: 381-83. 8. Mazzuoli GF, Tabolli S, Bigi F, et al. Effects of salmon calcitonin induced by ovariectomy. Calcif Tissue Int 1990; 47: 209-14.
on
the bone loss
Deaths from tobacco SIR,-Professor Peto and colleagues’ report (May 23, p 1268) provides indirect forecasts of mortality from smoking in the developed countries. Although more narrow in scope and using a different (and what is claimed to be a more robust) methodology than the study by some of the same authors reported at the Perth World Smoking and Health Conference in 1990, this is nevertheless one of the most ambitious exercises in quantitative epidemiology ever to be published. Since these and similar forecasts are likely to determine public policy, it is important that what will be regarded by decision-makers as expert advice is in fact well-founded. Forecasting is a notoriously fragile process and in view of the widespread scepticism about pronouncements in this field,l it is not unreasonable to look very carefully at the provenance of the latest set of forecasts.
Any forecasting procedure should include assumptions about the underlying relations, which are most conveniently considered in terms of a mathematical model. The resulting forecasts will be determined by the content and structure of the underlying model. However, the biological and physical mechanisms that determine death in man are poorly understood, especially with respect to possible agents such as tobacco. All types of smoking-related disease also arise in people who have never been exposed to tobacco smoke, and it is clear that other influences
or
confounding factors may be
involved. Neither the identity nor the relative importance of such factors in relation to any particular cause of death has been established on a definitive basis, although many associations (which may or may not be causal) have been reported. Nor is it known how the confounding factors may combine with any effects of tobacco to produce changes in the health status of the individual. On general grounds, it is very likely that the set of relevant confounding factors will vary from disease to disease as well as from society to society. Potential confounding factors may themselves be associated with smoking habits. For example, in societies where the smoking habit is well-established (which in general are the only societies for which evidence can be obtained), smoking is associated with low social class. The innate characterisics of the individual must also play a part, but the only available quantitative evidence about these matters is derived from uncontrolled epidemiological studies in which smokers and non-smokers are self-selected groups. There is no reason to suppose that the smoking habit is determined independently of personal characteristics or confounding factors. On all these grounds, it is clear that any model that is devised as a basis for generating forecasts on the basis of available knowledge is tentative. None of the models used by Peto and his colleages in recent publications seem to take account of these complex issues, and no evidence is presented that the simple models that are used do provide a realistic representation of the factors involved. Having summarised the experience of a "survey" population in terms of a mathematical model, Peto et al apply the results with the same model to a different target population, often in a different era and possibly in a different part of the world. This implies that the extent of any hazard associated with smoking is exactly the same in both survey and target populations, as are the confounding factors and the associations between the elements of the model. In view of the likely presence of differences in the properties and amounts of tobacco smoked, in the way the tobacco is smoked, in the confounding variables, and in the self-selection criteria for smoking, this is indeed a major act of faith. Peto’s indirect method of forecasting is thus subject to these general criticisms. In addition, the use of lung cancer rates as a yardstick involves further unsupported and arguably implausible hypotheses, including the assumptions: that lung cancer rates in non-smokers as reported in national statistical returns must be the same in the survey and target populations; that any relation between smoking, confounding factors, and mortality are of the same form for all relevant diseases; and that different types of product affect specific diseases in the same way. Furthermore, the arbitrary halving of the calculated excess risk associated with smoking may be generous to the tobacco industry, but does not seem to have any scientific foundation. A detailed analysis of the difficulties of assessing smoking-related mortality is being published elsewhere2 and will not be repeated here. Nevertheless, with the evident uncertainties of this process, it is impossible to avoid the question as to whether the calculation by any method of forecasts of mortality due to smoking on the basis of available information can ever be a proper foundation for public policy or anything more than an exercise in speculative arithmetic. Ide House, Ide, Exeter EX2 9RB, UK
JOHN R. ASHFORD
1. Levin B. Rise of the cigarette police. The Times, June 1, 1992. 2. Ashford JR. Problems in assessing smoking-related mortality. Disorders (in press).
J Smoking-Related
Serological evidence for congenital transmission of human
herpesvirus 6
SIR,-Human herpesvirus 6 (HPV-6) is the causative agent of exanthem subitum.1 Seroprevalence surveys indicate that more than 90% of children have evidence of HHV-6 infection before age 2.2 The virus has been detected in oropharynx and salivary glands of healthy adults, leading to speculation that HHV-6 is transmitted through contact with infected oral secretions.3 The possibility of congenital transmission of HHV-6, in analogy to other human herpesviruses, has been considered.’ We have retrospectively evaluated stored cord-blood specimens for evidence of congenital HHV-6 infection. 799 cord-blood sera were chosen randomly from
