Confusing association with causation?, a response to Brewer et al., ‘Valid evaluation of iatrogenic and sexual HIV transmission requires proof’

Our report of an association between injections and the incidence of HIV need not necessarily be interpreted as causal, as suggested by Brewer and colleagues. The main point of our paper [1] was to explore the most likely causative pathways for this association, which we concluded could be caused by contaminated injections, but were more likely to be due to injections for seroconversion illness or other consequences of unprotected sex. That study, and others from Uganda and elsewhere in Africa [2–4], provide very strong evidence that the majority of infections in adults in this mature epidemic are caused by transmission between spouses or regular sexual partners. It is very probable that all, or almost all, of the 22 seroconvertors with HIV-infected partners in the study were caused by transmission from those partners. This leaves only four seroconvertors (out of 36) who reported injections and who had partners with negative or unknown HIV status (and three of these were of unknown serostatus so may also have been seropositive). Brewer and colleagues mention that many households in Uganda have been reported to keep injection equipment. Such equipment would only be used occasionally by family members, so the chance of the virus surviving between injections is likely to be small [5]. Of more realistic concern from our report is that participants reported receiving injections from travelling itinerant injectionists. These, however, only accounted for 7% of all injections. Brewer and colleagues mention a previous letter of ours [6], in which we reported that primary (seroconversion) illness was not distinguishable among the many episodes of 'flu-like illnesses from other causes in this community. In that study HIV-seropositive individuals tended to have more symptoms than seronegative controls, however, which may reflect more severe illnesses, which in turn could plausibly have led to a higher frequency of injections as part of their treatment. The association of sexual variables with HIV infection was essentially unchanged after adjusting for injections, but because of small numbers was not of statistical significance. It is, worth noting, however, that six out of 36 HIV incident cases reported two or more sexual partners in the previous 3 months, compared with one out of 35 controls. Brewer and colleagues state that ‘the burden of proof rests on those who assert that penile–vaginal sex is the dominant mode of transmission (in sub-Saharan Africa)’. May we direct our readers to the vast body of literature accumulated over the past 20 years, which shows that the great majority of HIV in African adults is acquired through heterosexual contact? By contrast, the small minority of scientists who believe that needles play an important role need to give a plausible explanation for the extremely low prevalence and incidence of infection in children aged 5–14 years confirmed in studies from several parts of Africa [7–13]. A comprehensive re-analysis of data from the same study site in Uganda as our report has recently been published [14]. That study uses Bayesian methods to examine how previous beliefs about the role of injections should be modified when confronted with the field data, and shows that, although important uncertainty remains, only a small proportion of incident infections are likely to be caused by injections.