Response to ‘High-risk sexual behaviour increases among London gay men between 1998 and 2001

Abstract

With interest we read the paper on HIV optimism and sexual behaviour among gay men by Elford et al. [1] in a previous issue of AIDS. As all studies concerning the association between HIV optimism and high-risk sexual behaviour are currently based on cross-sectional data at one timepoint [2–5], their study, using cross-sectional data at four timepoints, contributes to an understanding of the changes over time. The authors concluded that it is unlikely that HIV optimism can explain the increase in high-risk sexual behaviour in London gyms. Although this might be true, the study is hampered by some methodological issues that might compromise their findings and raise concerns about the validity of such a conclusion. First, the authors report that they have changed the answering categories on the two optimism items from a five-point linear scale (`not at all'; ‘a bit'; ‘somewhat'; ‘quite a lot'; and ‘a lot') in the cross-sectional studies in 1998 and 1999, to a four-point scale (`strongly disagree'; ‘disagree'; ‘agree'; ‘strongly agree') in 2000 and 2001 [1]. Consequently, the definition of being ‘optimistic’ changed. In the first 2 years men were considered optimistic when they answered the two optimism items with ‘a bit', ‘somewhat', ‘quite a lot', or ‘a lot', whereas in the last 2 years they were considered optimistic when they answered ‘agree’ or ‘strongly agree'. Results show that the level of optimism decreased remarkably after the year 2000. Although the authors mentioned that the level of optimism is possibly slightly overestimated in the period before the year 2000 because of discontinuity in the measurement of HIV optimism, they argued that it is unlikely that the change has introduced a major source of bias. However, we feel that the discontinuity could generate an important bias, not only in the optimism levels but also in the levels of unprotected anal intercourse among optimistic and non-optimistic men. Consequently, the validity of the univariate and multivariate results as presented in this study is arguable. Second, men who filled in the questionnaire more than once were excluded in the multivariate analyses but not in the univariate analyses [1], resulting in different sample sizes used throughout the paper. This creates difficulties in the interpretation of the multivariate results. Are the adjusted effects of ‘year of survey’ and ‘optimism 1’ among HIV-negative individuals in the multivariate results really higher than the unadjusted effects, or is it a result of bias introduced by using a smaller sample in the multivariate analysis? Reanalysing the univariate results using the same sample size as in the multivariate analyses would facilitate a better understanding of the multivariate effects and the confounding role of the different covariates in the model. Finally, the behavioural dynamics in homosexual populations are known to be very complicated [6]. Some men change to risky behaviour, some men to more preventative behaviour, whereas other men do not change their behaviours at all. Looking at the population level, as was done in the study by Elford et al. [1], cannot reveal the complicated behavioural dynamics at an individual level. At an individual level, treatment-related optimism appears to be a factor that increases high-risk sexual behaviour [7], whereas this effect might not be noticed at the population level. Future studies, examining changes over time at an individual level, might give more insight into the role of HIV optimism in the total increase in high-risk sexual behaviour. In conclusion, we are not convinced by the report of Elford et al. [1] that HIV optimism is unlikely to explain the increase in high-risk sexual behaviour. We would like to emphasize that HIV optimism should still be addressed by prevention efforts, although we do agree that it is important to keep searching for other important factors that help to explain the increase in sexual risk behaviour, enabling the development of more effective prevention strategies.