We can also assess HPV type-specific relationships with the exposure variables using GEE models. Compared with conducting separate standard logistic regression models, GEE models still obtain greater efficiency since we reduce the number of covariate coefficients by assuming a common effect for “nuisance variables”. For example, only one “life-time number of sexual partners” coefficient is required if all individual oncogenic HPV are studied in the same GEE model, whereas the sexual partners variable would require a separate coefficient in each separate standard logistic regression for each HPV type. Further, GEE models allow us to statistically evaluate if the exposure-disease association is common across types.
