Gene-covariate interaction between dysplastic nevi and the CDKN2A gene in American melanoma-prone families.

The CDKN2A gene has been implicated in cutaneous malignant melanoma pathogenesis. Although CDKN2A mutations confer substantial risk for melanoma, clinicoepidemiological covariates including dysplastic nevi (DN), total nevi, and solar injury also enhance melanoma risk. To examine the relationship between CDKN2A and these three risk factors, we conducted combined segregation/linkage analysis using the class D regressive logistic model, as implemented in the computer program REGRESS. Genetic and covariate data were collected on 20 American melanoma-prone families, 13 of which had cosegregating CDKN2A mutations. Two types of analyses were conducted. The missing-indicator method used a missing-value indicator, set to 1 for unknown and 0 for known covariate status, and a second variable set to 1 for exposed and 0 for unexposed or unknown. The second method, complete-cases method, coded subjects with missing covariates as unknown for the affection status. The results for both analyses were very similar. Overall, there was a significant improvement in the likelihood when DN, total nevi or both covariates were added to the base model, which included dominant transmission of the CDKN2A gene and a linear increase of risk with the logarithm of age on the logit scale. In contrast, inclusion of solar injury did not significantly improve the likelihood for the base model. Significant evidence for a gene-covariate interaction was detected between DN and CDKN2A when DN was the only covariate in the model (missing-indicator method or complete-cases method) or when both DN and total nevi were in the model (complete-cases method only). Interestingly, in both methods, the odds ratio (OR) for DN was greater in subjects without mutations (OR, 20.1; 95% confidence interval, 4.8-92.8) versus those with CDKN2A mutations (OR, 3.3; 95% confidence interval, 1.1-10.0; complete-cases method). The CDKN2A-DN interaction illustrates the complex etiology of melanoma and needs to be confirmed in a larger sample of families.