Regression-based risk scores using sociodemographic and sexual behaviour data do not predict asymptomatic sexually transmitted infections among HIV PrEP users

OBJECTIVES: Among users of oral HIV pre-exposure prophylaxis (PrEP), condom use is low and incidence of sexually transmitted infections (STIs) is high, hence guidelines recommend STI screening every 3-6 months. Identifying individuals with higher asymptomatic STI risk for targeted screening may offer an opportunity to reduce the burden of STI screening. METHODS: In the Netherlands, PrEP has been offered through the National PrEP Pilot Program since 2019, which includes screening every 3 months. We included data of all individuals who received care through the PrEP programme between July 2019 and June 2022 and attended at least one PrEP care visit. STI-related symptoms and notification of possible STI exposure by sexual partners are recorded during each visit. We assessed the predictors of any chlamydia, gonorrhoea or syphilis infection diagnosed during routine asymptomatic STI screening (ie, no reported symptoms or partner notification) using logistic regression and calculated risk scores from coefficients of the multivariable logistic regression model. We estimated the sensitivity and specificity for the optimal prediction score cut-off. RESULTS: Among the 11 035 included individuals (97% men who have sex with men), 14 926 bacterial STIs (9114 diagnosed during routine asymptomatic screening) were diagnosed during a median of 24 months (IQR 15-30) of follow-up. We found that PrEP users who engaged in sex work, had condomless anal sex, participated in group sex or chemsex (ie, use of gamma-hydroxybutyrate/gamma-butyrolactone, mephedrone or crystallised methamphetamine during sex), injected drugs or used alcohol or non-chemsex-related drugs during sex had an increased risk of STIs diagnosed during routine asymptomatic screening. PrEP users born in the Netherlands and those who attended college or university had a lower STI risk. A risk score using these covariates resulted in a sensitivity of 0.55 (95% CI 0.54 to 0.56) and specificity of 0.55 (95% CI 0.54 to 0.55). Individuals eligible for STI screening accounted for 54% of STIs diagnosed during follow-up. CONCLUSIONS: Using routinely available demographic and behavioural data, it was not possible to construct a well-performing risk score to identify individuals at high risk of STIs diagnosed during routine asymptomatic screening. Other factors, methods or ways to analyse data may be needed to increase predictive capacity for STI risk scores.