Characterizing the temporal dynamics of human papillomavirus DNA detectability using short-interval sampling

Background: Variable detection of human papillomavirus (HPV) DNA can result in misclassification of infection status, but the extent of misclassification has not been quantitatively evaluated. Methods: In 2005-2007, 33 women of ages 22 to 53 years self-collected vaginal swabs twice per week for 16 consecutive weeks. Each of the 955 swabs collected was tested for 37 HPV types/subtypes. Assuming that a woman's underlying infection status did not change over the short study period, biases in prevalence estimates obtained from single versus multiple swabs were calculated. Using event history analysis methods, time to recurrent gain and loss of at least one HPV type was determined, separately. Baseline any-type and high risk- type HPV prevalence was 60.6% and 24.2%, respectively. Cumulative any-HPV and high-risk HPV prevalence over the 16-week period was 84.8% and 60.6%, separately. Results: Overall, there were 319 events of detection and 313 events of loss of detection. Median times to a recurrent detection and loss of detection were 11 and seven days, respectively. Neither vaginal sex nor condom use during follow-up was associated with recurrent viral detection or loss of detection. Assuming the cumulative 16-week prevalence reflects the true prevalence of infection, the baseline any-HPV prevalence underestimated infection status by 24.2%, with a bootstrapped mean of 20.2% [95% confidence interval (CI), 8.9%-29.6%]. Conclusions: These findings suggest that a substantial proportion of HPV-infected women are misclassified as being uninfected when using a single-time DNA measurement. Impact: Short-term variation in detectable HPV DNA needs to be considered while interpreting the natural history of infections using single samples collected at long intervals. Cancer Epidemiol Biomarkers Prev; 23(1); 200-8.