High Levels of Prevention-effective Adherence to HIV PrEP

An Analysis of Substudy Data From the EPIC-NSW Trial

Benjamin R. Bavinton, PhD; Stefanie Vaccher, PhD; Fengyi Jin, PhD; Garrett P. Prestage, PhD; Martin Holt, PhD; Iryna B. Zablotska-Manos, PhD; Rebecca Guy, PhD; Janaki Amin, PhD; David J. Templeton, PhD; Barbara Yeung, MPH; Mohamed A. Hammoud, PhD; David Lewis, PhD; David Baker, MBBS; Nila Dharan, MD; Anna M. McNulty, MBBS; Andrew E. Grulich, PhD


J Acquir Immune Defic Syndr. 2021;87(4):1040-1047. 

In This Article


In this large-scale PrEP implementation study, we observed prevention-effective adherence in more than 99% of observed follow-up weeks among substudy participants. A small proportion of weeks involving CLS were inadequately protected by PrEP (6.6% of CLS weeks, where inadequate PrEP protection was defined as having taken <4 pills or <6 pills in the last full week in those assigned male or female at birth, respectively), and only 1.4% of all CLS weeks were inadequately protected and involved sexual activity with higher-risk partners (ie, HIV-positive partners with detectable/unknown viral load or unknown HIV status partners). We have shown that measuring both pill taking and sexual behavior during the same period more fully captures the context of adherence. Studies that measure pill taking or drug concentrations only cannot explain apparent mismatches between level of PrEP adherence or persistence and HIV incidence. Our findings help to explain why the long-term analysis of EPIC-NSW participants found very low HIV incidence despite MPR at the end of the study decreasing to 0.6 (thus suggesting PrEP discontinuation or suboptimal adherence).[13,15] Thus, although MPR is commonly used to estimate PrEP adherence in research studies,[20–22] its limitations must be acknowledged when interpreting HIV outcomes data.

Our data indicated that few participants had suboptimal adherence around times of risk, as opposed to having stopped or paused PrEP. Of the higher risk CLS weeks with inadequate PrEP protection, 88.4% were in participants who took no pills at all in that week. This suggests that among our substudy participants, the main issue was safely stopping and restarting PrEP rather than continued, suboptimal adherence. The small number of seroconversions in EPIC-NSW[13] demonstrates that some individuals struggled with maintaining prevention-effective adherence, although it must be acknowledged that poor adherence appears to be less of a systemic problem in Australia than in many other settings internationally.[23] There are 2 main issues of concern: individuals may stop or pause PrEP and then be exposed to HIV or they may continue using PrEP with suboptimal adherence.[24] Internationally, discontinuation of PrEP has been associated with younger age,[25–27] minority ethnic status,[27,28] financial barriers,[29] and drug use.[25,26] However, as many people do appear to stop or pause their PrEP use in line with periods of lower risk,[30,31] more research is needed to determine ways to identify patients who are more likely to discontinue PrEP and subsequently engage in risk behavior. This issue has become especially pertinent in the context of the COVID-19 pandemic, during which many GBM ceased taking PrEP while physical distancing restrictions were in effect.[32]

In our analysis, weeks involving CLS with higher risk partners and having inadequate PrEP protection were more likely in participants with a preference for nondaily PrEP. It is not surprising that individuals with a preference for nondaily PrEP took fewer pills than those who preferred daily dosing. Nonetheless, it is a concern that this association occurred in weeks involving higher risk CLS. Daily PrEP is not suitable for all people engaging in higher risk sex, and these findings underscore the continuing need for new PrEP modalities not reliant on daily pill taking, such as long-acting injectable PrEP or long-acting PrEP implants.[33] It is important to note that there had not been any widespread community education about nondaily PrEP options in NSW during this substudy's period of observation. Other research has found relatively low levels of event-driven dosing among PrEP-using GBM in Australia (ranging from 6% among former EPIC-NSW participants in late 2019/early 2020[8] to 13% in gay community HIV behavioral surveillance in 2020[34]), although 43% of PrEP-experienced men showed interest in using it.[8] Event-driven PrEP may help reduce the occurrence of these episodes of higher risk CLS by providing an option more acceptable to people who do not want to take daily PrEP. Community education about event-driven PrEP is a critical priority.

Other factors were also associated with higher risk CLS weeks and having inadequate PrEP protection. As mentioned above, younger age has been associated with discontinuation of PrEP and also poorer adherence,[35] including in EPIC-NSW,[15] and there is evidence from multiple Australian sources that younger men tend to have lower rates of PrEP uptake.[36–38] In our data, older age (ie, 45 years and older) was indeed associated with better prevention-effective adherence, and participants younger than 25 years had the highest proportion of higher risk CLS weeks not adequately protected by PrEP. Living in a suburb with a lower concentration of gay-identified resident men (ie, less than 20% of the men) was associated with inadequate PrEP protection in higher risk CLS weeks. Before PrEP, evidence suggested that living in suburbs with high concentrations of gay men was associated with higher HIV risk.[36] However, PrEP uptake and the population-level declines observed in HIV diagnoses have been greatest in these suburbs,[14] along with a greater sense of connection to the gay community, higher formal education and employment, more frequent HIV testing, and lower levels of "unprotected" CLS (ie, anal intercourse not protected by condoms, PrEP, or undetectable viral load).[36,39,40] Men living outside the traditional "gaybourhoods"[16] require targeted PrEP adherence education and support appropriate to their circumstances, the sex they have, and where they live. Consideration should also be given to the support needs of PrEP providers outside of inner-city locations.[41]

Our analysis had some limitations. Adherence was measured by self-report only and may have been subject to social desirability bias. However, in our data, self-reported adherence in the last full week was related to MPR from the quarter that week fell in, and previous research in NSW has demonstrated a strong correlation between self-reported adherence and direct measurement of tenofovir levels in blood.[17] The short recall period of the last full week may have ameliorated problems with longer-term recall. Some weeks may have been misclassified as having inadequate PrEP protection in users of event-driven regimen: if the loading dose occurred in the days immediately before the first day of the last full week, only 1–2 pills may have been reported in the last full week. However, this situation is likely to have been rare given that only 2.6% of last full weeks were in participants reporting event-driven dosing. The EPIC-NSW trial was designed to be as "real-world" as possible, and the surveys were optional; there was incomplete follow-up survey data (although, as the primary unit of analysis was the "last full week," the goal was not to have an uninterrupted period of follow-up as is the case in most longitudinal cohort analyses).

Additionally, the results highlight the challenging problem of selection bias in the context of an optional survey-based substudy implemented as part of a larger trial (which at the time was the only way to access PrEP in NSW). Participants who completed fewer surveys were more likely to have inadequate PrEP protection in higher risk CLS weeks (Table 4), and they had slightly lower MPR. Based on the number of seroconversions (ie, 4 in the substudy participants versus 30 in the parent study), it is likely that those who could be included in this analysis had lower HIV risk, even considering that there were few differences between those included and the full EPIC-NSW cohort as shown in Table 1. Given that EPIC-NSW and this substudy were primarily in cisgender GBM in Australia, the findings may not be generalizable to other populations or GBM in other countries.