Introduction Vaginal prostate-specific antigen (PSA) is a biomarker of recent semen exposure that may overcome potential fallacies of self-reported sexual behaviours, improving sexual exposure assessment in HIV/STI research.

Methods We conducted a cohort study to compare PSA detection and self-reported unprotected sex using data from a clinical trial in Malawi that randomised 73 HIV-infected and 24 HIV-uninfected women to depot medroxyprogesterone acetate injectable or levonorgestrel implant. Women were advised to abstain from vaginal intercourse or use a condom the day preceding study visits. We tested 539 vaginal swabs from 97 women for PSA using the ABAcard p30 rapid strip test (Abacus Diagnostics, Inc., West Hills, CA). Self-reported sexual behaviours and vaginal swabs were collected at six study visits. Log-binomial regression with generalised estimating equations were used to estimate associations of PSA detection and reported unprotected sex with demographic and behavioural factors, adjusting for study arm and pre- vs. post-contraception initiation.

Results Overall, 55 (57%) women tested positive for PSA and 54 (56%) women reported unprotected sex. Among PSA-positive samples, 62% (65/105) of instances reported no unprotected sex. The following were associated with PSA detection: HIV-negative status (Prevalence Ratio (PR): 1.69, 95% Confidence Interval (CI):1.09, 2.61), younger age (PR: 1.04, CI: 1.00, 1.08), reported unprotected sex (PR: 2.48, CI: 1.70, 3.60), and sex within past 48 hours (PR: 4.68, CI: 3.00, 7.30). The same factors were significantly associated with self-reported unprotected sex, as was PSA detection (PR: 2.45, CI: 1.7, 3.53).

Conclusions Self-reported sexual behaviours were significantly associated with PSA detection. However, PSA was detected among women reporting no unprotected sex, suggesting misreporting of condom use or condom failure. HIV-negative status and younger age were associated with unprotected sex, suggesting more frequent biases in research data for these groups based on social desirability and non-adherence to protocols.