Risk of HIV infection in a cohort of men who have sex with men attending CheckpointLX in Lisbon, Portugal

Abstract

Introduction To accurately measure incidence it is essential to define precisely the timing of infection. However, in presence of interval-censored survival data, in which the event date is not exactly known, but is known to lie within an interval, timing of infection often needs to be estimated. One common approach is to use the mid-point between last negative and first positive test assuming that event of interest has a uniform distribution. However, when testing intervals are not fixed the mid-point method may not be a good estimate of infection date. Alternative approaches taking into account this uncertainty regarding the date of infection must be used. These include non-parametric estimators such as the Turnbull's algorithm that we used to estimate the risk of HIV infection among MSM. Methods We used data from the Lisbon Cohort of MSM–a dynamic prospective cohort assembled since April 2011. The cohort is set up and conducted at CheckpointLX, a community-based HIV testing and counseling (CBVCT) center tailored for MSM run by trained peer counselors. The cohort recruits males reporting sex with other men, aged 18 or more, and HIV-negative at recruitment. For the survival analysis the nonparametric estimator of Turnbull was applied and the values of the estimator were obtained using the expectation-maximization algorithm. The analysis were conducted using the Icens R package. Results From April 2011 to June 2017, 5631 HIV-negative adult MSM were eligible for follow-up of whom 2387 had at least one follow-up visit. During follow-up 99 incident HIV infections occurred. From the survival analysis, the probability that the HIV infection occurred at 430 days of follow-up was equal to 0.03 and after 1827 days, approximately 5 years of follow-up, was 0.08. The risk of infection at the first two years decreased 0.05 while at the end of three years of follow-up the risk decreased 0.06 and standing at 0.92 of probability of survival at the end of the follow-up considered in this study. Conclusions We observed that the survival of HIV infection in the Portuguese cohort based in a CBVCT in Lisbon decreased slower as the person-time of follow-up increased. The major decrease on the survival function appeared at the first two years of follow-up (1 to 0.96). Three major causes can operate: the effect of risk reduction counseling and participation in a cohort study; the effect at community level of treatment as prevention and access to pre- and post-exposure prophylaxis; and differential losses-to follow-up.