Smoking, lung cancer and hypothesis testing

Abstract

It has been shown previously that the precipitator hypothesis of causation is able to account for certain key features of the epidemiological evidence that conflict with initiator and promoter hypotheses. The precipitator hypothesis states that the carcinogenic risk depends linearly on the average rate of smoking at an interval of τ years before death and that the ‘doubling dose-rate’, of D cigarettes per yr, is constant with respect to age from 35yr and above. In an earlier paper this hypothesis was tested by analysing secular trends in sex- and age-specific mortality from lung cancer recorded in England and Wales in relation to secular trends in cigarette consumption. The periods considered extended from 1950 up to 1975. The great majority of data failed to support the precipitator hypothesis but the well known errors in death certification were undoubtedly responsible for many anomalies. There were some indications of improved diagnostic accuracy towards the end of the period and hence it was considered important to update the analysis. This paper considers the trends over the recent period, 1974-5 to 1977-8. Unfortunately these newer data also appear to contain anomalies and they fail to support the precipitator hypothesis. In view of the difficulties confronting initiator and promoter hypotheses we have little alternative but to adopt Oldham's agnostic attitude “…we still do not know how cigarettes cause lung cancer, nor even, if we are particularly rigorous in our use of scientific logic, whether they do”.