The East African AIDS epidemic and the absence of male circumcision: what is the link? Using circumcision to prevent HIV infection in sub-Saharan Africa: the view of an African

There are now two schools of thought about the link between lack of circumcision and HIV infection in Africa. One school is that of Bongaarts et al. (1989), Moses et al. (n.d.) and Caldwell and Caldwell (1994) who use geographical distribution evidence to argue that the association between lack of circumcision and a high level of HIV infection in Africa is so convincing that the likelihood of a link should be recognized and taken into account where possible in the battle against AIDS. Moses et al. (n.d.) have gone further to recommend circumcision interventions for Africa. In contrast, De Vincenzi and Mertens (1994) argue that the evidence for an association, at least from small-scale surveys, is doubtful and hence not conclusive enough to qualify circumcision as an intervention. My view is twofold. First, as scientists we should look at the existing evidence for and against the hypothesis that lack of circumcision is positively related to HIV infection. Although many studies have shown that there is a significant association between the two variables (Bongaarts et al. 1989; Moses et al. 1990; Caldwell and Caldwell 1994), there is no evidence of a causal relationship between the two. Caldwell and Caldwell (1994) pointed out that 'Positive association does not necessarily show a direct causal relationship'. De Vincenzi and Mertens (1994) make the salient point that most of the quoted studies were not designed to test a hypothesis about circumcision and did not report their methodological details for evaluation. They claim there is no evidence that enough statistical rigour was applied to the data and where, in small-scale studies, this was done by controlling for several variables like age, number of partners, contact with prostitutes, ethnic origin and birthplace, the results lost statistical significance (Greenblatt et al. 1988). Either those findings that did not fit into the conventional thinking of the hypothesis were rejected by the journals as not interesting, or the researchers did not submit them for publication for fear of rejection. As in the study by Serwadda et al. (1992), recent work on data from the 1987/88 National Survey of Uganda assumed Muslims to be circumcised and non-Muslim Ugandans not to be circumcised. This was a fair assumption since the cultural groups that circumcise were not included in the National Survey. The study shows an association at the bivariate analysis which disappears at the multivariate level (Tusingwire 1995). Perhaps the fact that …