The use of logistic regression for modelling risk factors: with application to non-melanoma skin cancer.

Logistic regression was used to estimate the "relative" risk of basal and squamous skin cancer for such factors as cumulative lifetime solar exposure, age, complexion, and tannability. In previous reports, a subject's exposure was estimated indirectly, by latitude, or by the number of sun days in a subject's habitat. In contrast, these results are based on interview data gathered for each subject. A relatively new technique was used to estimate "relative" risk by controlling for confounding and testing for effect modification. A linear effect for the "relative" risk of cancer versus exposure was found. Tannability was shown to be a more important risk factor than complexion. This result is consistent with the work of Silverstone and Searle.