A WLW model of incident oncogenic HPV detection by type (Model(1.8)) showed that women with a CD4+ count <200 cells/mm3 and HIV RNA > 100,000 copies/mL had significantly increased risk of incident oncogenic HPV detection (HR = eβ̄ = 4.9; 95% CI,3.8–6.3) relative to HIV-negative women (Table 2). Further analysis involving measurement of type-specific associations (Model(1.9)), using HPV16 as the referent HPV type, showed that most oncogenic HPV types, with the exception of HPV18 and HPV56, had stronger associations with host immune status than HPV16 (Table 3). While on an individual basis these two-fold or greater differences did not reach significance, the power to study incident HPV detection is less than to study HPV prevalence; in our WLW model of incident HPV detection each person contributed one time-to-event result per HPV type, whereas in our GEE logistic regression model on HPV prevalence the dataset involved repeated observations of each HPV type at each visit. That said, when assessed as a group, the incident detection of all other oncogenic HPV types excluding HPV18 and HPV56 (model (1.9)) had a three-fold (HR = eβ̄otheronc / eβ̂16 = 3.1) greater (95% CI,1.4–7.2) association with CD4+ count than HPV16.
