Response to Brewer et al. regarding our article ‘Association between medical injections and HIV infection in a community based study in India’

Abstract

We thank Dr Brewer and colleagues for their interest in our study. As they point out, we did observe an association between the number of medical injections reported to have been received in the past year and the risk of HIV infection. On univariate analysis, this was significant for women reporting five or more injections in the past year, and for men, 10 or more injections. On multivariate analysis, the association was statistically significant only among women who reported receiving five or more injections, although there was a trend towards an association for men reporting 10 or more injections. As Brewer and colleagues indicate, causal inferences cannot be drawn from a cross-sectional study, and it is possible that individuals may either have acquired HIV infection on account of receiving unsafe injections, or have first acquired HIV and then received injections related to illness. Only prospective studies or other studies that measure incidence can tease out causality in these relationships. To address the issue raised by Brewer and colleagues of sexual behaviour as a risk factor for HIV infection, we examined the relationship between HIV infection and individuals reporting two or more lifetime sex partners. On univariate analysis, this relationship was statistically significant for men [odds ratio (OR) 2.7, 95% confidence interval (CI) 1.6–4.7] and for women (OR 3.4, 95% CI 1.0–12.3). We also developed a logistic regression model with HIV infection as the outcome, and the independent variable of interest being individuals reporting two or more lifetime partners. We controlled for injections received in the previous year in addition to the other variables controlled for in our initial analysis [1]. On multivariate analysis, there was a trend towards an association among men, with men who reported multiple partners being 1.8 times more likely to be infected with HIV (P = 0.07). However, there was no significant association observed among women. A potential limitation of these data is the underreporting of risk behaviours, particularly among women. Other studies that we have conducted in the same area in India have shown that more anonymous forms of collecting behavioural information yield a much greater response with respect to reported risky sexual behaviours than face to face interviews [2]. In any case, population-based longitudinal studies are required, using more detailed and accurate measures of sexual and parenteral exposures, to assess their relationships with the risk of HIV infection in these settings.