Vaginal microbiome components as correlates of cervical human papillomavirus infection

Background The interplay between vaginal microbiome and human papillomavirus (HPV) remains unclear, partly due to heterogeneity of the microbiota. Methods We used data from 546 women enrolled in a cross-sectional study conducted in five Brazilian regions. We genotyped vaginal samples for HPV and sequenced V3-V4 region of 16S rRNA gene for vaginal microbiome analysis. We used stepwise logistic regression to construct two linear scores to predict high-risk HPV (hrHPV) positivity: one based exclusively on presence of individual bacterial taxa (microbiome-based [MB] score) and the other exclusively on participants' sociodemographic, behavioral and clinical (SBC) characteristics. The MB score combined coefficients of 30 (out of 116) species. The SBC score retained six out of 25 candidate variables. We constructed receiver operating characteristic curves for the scores as hrHPV correlates and compared areas under the curve (AUC) and 95% confidence intervals (CI). Results Overall, prevalence of hrHPV was 15.8%, and 26.2% had a Lactobacillus-depleted microbiome. The AUCs were 0.8022 (CI:0.7517-0.8527) for MB score and 0.7027 (CI:0.6419-0.7636) for SBC score (P=0.0163). Conclusions The proposed MB score is strongly correlated with hrHPV positivity - exceeding the predictive value of behavioral variables - suggesting its potential as an indicator of infection and possible value for clinical risk stratification.