Replication Data for: Stopping and re-starting PrEP and loss to follow-up among PrEP-taking MSM at risk of HIV-1 participating in a prospective cohort study in Kenya

Understanding predictors of stopping and re-starting pre-exposure prophylaxis (PrEP) is important to improve PrEP programming. We assessed frequency and predictors of stopping/re-starting PrEP and loss to follow-up (LTFU), and assessed whether risk behaviour was different while having stopped and restarted PrEP among men who have sex with men (MSM) with access to PrEP services in coastal Kenya. From June 2017 to June 2019, we included MSM who had initiated daily oral PrEP and were followed monthly for HIV-1 testing, PrEP refill, risk assessment, and risk reduction counselling. We estimated transition intensities (TI) and predictors of switching (i) between having stopped and having (re-)started PrEP and (ii) from either PrEP state and LTFU (i.e. not returning to the clinic for >90 days) using a multi-state Markov model. We compared the proportion with any receptive anal intercourse (RAI), any insertive anal intercourse (IAI) and 134 MSM ever starting PrEP were followed for a median 20.3 months [interquartile range (IQR) 7.7-22.1]. Of them, 49 (36.6%) stopped PrEP a total 73 times [TI=0.6/PY, 95% confidence interval (CI)=0.5-0.7] and re-started PrEP 38 times (TI=1.2/PY, 95%CI=0.9-1.7). In multivariable analysis, stopping PrEP was related to anal sex in the past 3 months, substance use disorder and travelling. Re-starting PrEP was related to non-Christian or non-Muslim religion and travelling. 54 participants were LTFU: having (re-)started-PrEP (n=47, TI=0.3/PY, 95%CI=0.3-0.5) and having stopped-PrEP (n=7, TI=0.2/PY, 95%CI=0.1-0.4). In multivariable analysis, becoming LTFU while having (re-)started-PrEP was associated with secondary education or higher, years lived in the area (≤1 years), residence outside the immediate clinic catchment area and alcohol use disorder. No factors were associated with becoming LTFU while having stopped-PrEP. Participants reported higher proportions of IAI and lower proportions