Assessing the performance of international pre‐exposure prophylaxis (PrEP) eligibility guidelines in a cohort of Chinese MSM, Beijing, China 2009 to 2016

INTRODUCTION: PrEP is a powerful HIV prevention tool, and locally relevant eligibility criteria are necessary to optimize the prevention impact of PrEP. We assessed performance of existing national and international PrEP eligibility criteria to predict future HIV seroconversion among MSM in Beijing, China. METHODS: Participants were MSM aged ≥18 years who enrolled in a cohort study between July 2009 and March 2016. Participants completed HIV testing, syphilis testing, and a questionnaire on recent sexual health behaviours at each follow‐up visit and were followed until HIV seroconversion or dropout. We assessed PrEP eligibility at the most recent follow‐up visit prior to the final study visit. Participants were classified as indicated for PrEP (or not) based on criteria from guidelines from Europe, Korea, South Africa, Taiwan, the United Kingdom, United States and the World Health Organization. To compare guideline performance, we calculated sensitivity, specificity, Youden’s Index (YI), Matthew’s Correlation Coefficient (MCC), F1 scores and diagnostic odds ratios. For each guideline, performance measures were compared to random allocation of PrEP by randomly selecting a proportion of participants equal to the proportion indicated. RESULTS: There were 287 (17∙3%) incident HIV seroconversions among 1663 MSM. The number of men indicated for PrEP from different guidelines ranged from 556 (33∙4%) to 1569 (94∙2%). Compared to random allocation, sensitivity of algorithms to predict seroconversion ranged from slightly worse (−4∙7%) to 30∙2% better than random. However, in absolute terms, none of the sensitivity values increased by more than 11% when compared to random allocation. For all guidelines, specificity was not meaningfully better than random allocation. No guidelines had high binary classification performance measures. CONCLUSIONS: The performance of international indication guidelines in this sample was only slightly better than random allocation. Using such guidelines to screen out MSM self‐identifying as interested in PrEP could lead to misallocation of resources and to good candidates for PrEP being denied access. For settings in which international guidelines perform poorly, alternative indication approaches should be considered.