Phillips A N, Smith G D
Academic Department of Genito-urinary Medicine, University College and Middlesex School of Medicine, London, UK.
Int J Epidemiol. 1994 Feb;23(1):42-9. doi: 10.1093/ije/23.1.42.
It is widely believed that some sexually transmitted pathogen plays a key role in the aetiology of cervical cancer. However, although certain human papilloma viruses are strongly implicated, the pathogen responsible and its mechanism of action remain to be finally characterized. The correlation between cigarette smoking and sexual activity that exists in most cultures therefore makes evaluation of the potential additional role of smoking difficult, due to confounding with the presence of the aetiological pathogen. Epidemiological studies of the association between smoking and cervical cancer have adjusted for the lifetime number of sexual partners as a proxy measure of the presence of the pathogen and, in most cases, the association has diminished but remained statistically significant. Since the use of a proxy will tend to result in underestimation of the effect of the aetiological pathogen on risk of cervical cancer, however, the adjustment is likely to be insufficient, thus resulting in overestimation of the adjusted or 'independent' effect of smoking. In an attempt to address this concern we used a simulation approach to investigate whether a substantial 'independent' association between smoking and cervical cancer might be expected as a result of the use of a poor proxy for the aetiological pathogen, even if there is no true effect of smoking. Using realistic estimates of the association between the presence of the aetiological pathogen and both smoking and risk of cervical cancer, 'independent' relative risks for cigarette smoking of two and above were generated. It is therefore plausible that the observed 'independent' effect of cigarette smoking on cervical cancer is due to residual confounding.
