Smoking and Cervical Cancer: Cause or Coincidence

To the Editor. — In an editorial¹ accompanying the recently published article on cigarette smoking and cervical cancer,² Dr Layde raises two important issues. First, he correctly indicates that adjusting an observed association (in this case smoking with cervical cancer) by a confounding variable (number of sexual partners) reduces the strength of the association, and, because the confounder is a surrogate measure, the resulting adjusted estimate may overestimate the true risk. Second, he states that resolving the issue of the causal nature of the smoking—cervical cancer association is not urgent because smoking cessation already is justified by a variety of established harmful consequences. Both of these points deserve comment.

A confounding variable is, by definition, a factor that is associated with both the variable under study and the outcome variable and that, therefore, accounts for all or part of the observed association between the study variable and the