Concerns about a study on sexually transmitted infections after initiation of HIV preexposure prophylaxis

To the Editor

An article in AIDS by Nguyen and colleagues tests the hypothesis that use of HIV preexposure prophylaxis (PrEP) leads to risk compensation, and thus to increased rates of sexually transmitted infections (STIs).[1] The authors use a Canadian clinic-based sample of men who have sex with men (MSM) to make two comparisons: 1) STI rates in PrEP users in the 12 months after PrEP initiation compared with 12 months before PrEP initiation, and 2) STI rates in PrEP users in the 12 months after PrEP initiation compared with post-exposure prophylaxis (PEP) users in the 12 months after PEP use. They conclude that their data support the risk compensation hypothesis. However, we have four concerns about this study, particularly with respect to study design and the degree to which the results support the authors’ conclusions.

First, as the authors point out, gonorrhea and chlamydia incidence nearly doubled among males in Quebec during 2010–2015, so we would expect to see increases in STI rates over time regardless of PrEP initiation. Thus, any significant finding in the pre-/post-initiation comparison may be driven by these temporal changes rather than PrEP use. Analytic approaches exist to address secular changes over time; one is to compare PrEP users to a control group also experiencing the temporal change but not the exposure condition (i.e., PrEP), as in difference-in-differences analysis.[2]

Second, PEP users are not a valid control group for PrEP users. Although PEP users have had at least one potential HIV exposure, those who have ongoing HIV risk would ideally transition from PEP to PrEP, while those with only a brief period of risk or a single exposure would not be indicated for PrEP.[3] The authors did not present data on transitions from PEP to PrEP or discuss how these were handled in the analyses. If PEP patients with ongoing risk transitioned to PrEP, the higher rates of STIs observed would likely reflect the appropriate prescribing of PrEP rather than risk compensation. Furthermore, most PEP users report a decrease in condomless sex in the 12 months post-PEP;[4] indeed, a comparison to STI rates in the 12 months prior to PEP use may have been more appropriate.

Third, the authors do not acknowledge or discuss their null findings, and some conclusions are based on interpreting non-significant results as meaningful differences. For the post- vs. pre-PrEP comparison, there was no observed increase in rates of anal, oral, urethral, or any gonorrhea; syphilis; or anal, oral, or urethral chlamydia, and the conclusion that overall STI rates were higher post- vs. pre-PrEP was based on a result that was not statistically significant after adjustment for frequency of STI screening (adjusted incidence rate ratio [aIRR] 1.39, 95% CI: 0.98–1.96). For the post-PrEP vs. post-PEP comparison, there was no observed increase in rates of anal, oral, urethral, or any gonorrhea; syphilis; or oral chlamydia.

Finally, we do not understand the conclusion that the highest rate ratios were for anal and oral gonorrhea, nor the authors’ argument that increases in asymptomatic STIs indicate risk compensation. The rate ratios for anal and oral gonorrhea in the post-/pre-PrEP comparison were above and below the null value of 1, respectively, and neither approached statistical significance (anal: aIRR 1.29, 95% CI: 0.57–2.89; oral: aIRR 0.85, 95% CI: 0.38–1.90). Even if the authors had in fact observed increases in asymptomatic STIs, it is not clear why this would be consistent with risk compensation. The higher frequency of screening among PrEP users will detect asymptomatic STIs that would otherwise have gone undiagnosed, a bias that is unlikely to be fully removed by adjusting for number of STI tests. Thus, an increase in symptomatic STIs would be a better marker of risk compensation because it would be independent of differences in screening. Most urethral gonorrhea infections in males are symptomatic,[5] yet the authors observed no increased risk of urethral gonorrhea post- vs. pre-PrEP (aIRR 0.60, 95% CI: 0.22–1.64) or post-PrEP vs. post-PEP (aIRR 2.42, 95% CI: 0.47–12.55).

There is a pressing need for further research on clinical and behavioral PrEP outcomes, given the rapidly shifting dialogue about PrEP uptake, the optimal delivery of this health service, and its potential effects on individuals’ sexual behaviors and the sexual health of populations. That said, our concerns about the study by Nguyen et al., as well as similar concerns expressed about other recent studies on PrEP-associated risk compensation,[6, 7] highlight that there is an equally pressing need for rigorous research design and measured study interpretation that does not overreach the observed findings.