Commentary: Schairer and Schöniger's forgotten tobacco epidemiology and the Nazi quest for racial purity

Robert N Proctor
2001 International Journal of Epidemiology  
Judged by modern standards of epidemiology Schairer and Schöniger's work fails on several grounds: (1) the small number of cases of lung cancer (93 men and 16 women), (2) the use of surrogate informants for the patients who had died from cancer and living informants for information about themselves for the controls, (3) the low response rate to the questionnaires about the cancer patients (53% for those with lung cancer, 40-60% for those with cancers of the tongue, oesophagus, stomach, colon,
more » ... , stomach, colon, and prostate, and an even lower rate [39%] for the controls) and (4) the use of controls in only one narrow age group (53-54 years) selected to correspond with the average age of the lung cancer patients (53.9 years). The epidemiology of non-communicable diseases was, however, in its infancy at the time and it would be unfair to the authors to judge their work in this way. However, it should be noted that much higher standards had already been set in a few studies, notably in the UK by Lane-Claypon 8 in her study of 500 cases of breast cancer and 500 controls and by Stocks and Karn 9 in their study of the diet of 462 patients with a variety of cancers and 435 controls, and in the US by English et al. 10 in their study of smoking and myocardial infarction. The strength of Schairer and Schöniger's study lay in the fact that they compared their findings for people who had died from lung cancer with those for people who had died from other types of cancer, principally stomach cancer (then the most prevalent type of fatal cancer in Germany, as it was in most of the developed world), as well as with their findings for healthy controls. Its value lies principally in their percipient discussion of the meaning of their findings. They noted the great differences they recorded in the proportions of heavy smokers and of non-smokers in their lung cancer patients compared with those recorded for gastric cancer and control subjects and the similarity of their results with those obtained earlier by Müller 3 (whose findings had stimulated their enquiry). They also highlighted the increase in the frequency with which lung cancer had occurred over the previous 20 years, its correspondence with the increased use of tobacco, the greater frequency of the disease in men than in women (corresponding to the sex difference in smoking), the successful induction of cancer on the skin of animals by the application of tobacco tar by Roffo 11 and the presence in such tar of 'the strongly carcinogenic benzpyrene'. When all these pieces of evidence were combined Schairer and Schöniger thought the total evidence made a strong case for causality, particularly as they were able to cite several reasons (q.v.) for thinking that occupation, combustion engines and city life could not account for the findings. Their list of evidence, it may be noted, includes most of the types described later by Hill 12 as guidelines that could be used to determine causality. One piece of the evidence would, however, have been disputed in the UK, for the temperature at which Roffo burnt his tobacco was greater than the temperature at which tobacco is burnt in normal smoking and, although Kennaway's colleagues eventually showed that benzo(a)pyrene was present in normal tobacco smoke in small amounts 13 and Wynder et al. 14 showed that cancers could be produced on the skin of animals by prolonged painting with tars produced at the appropriate temperature, Roffo's experiments should not have been cited as biological evidence of the plausibility of a causal relationship. Schairer and Schöniger realized the weaknesses in their study and noted that the proportion of non-smokers in their comparison groups were suspiciously high and that the low response rate in the controls might have biased the results; heavy smokers in particular having refrained from replying. Although these considerations should not, they thought, have applied to the other cancer comparison groups. Stomach cancer patients, however, might have suffered from a 'weak stomach' for some time and consequently refrained from heavy smoking. They recognized that their comparison material was 'less than satisfactory' and reached the cautious conclusion that 'the association between heavy
doi:10.1093/ije/30.1.31 pmid:11171846 fatcat:gmxwqhughjg7jlsuhpk6blpblm