The Lancet, February 2007, vol. 369, iss. 9562, pp. 657-666. Available From:Link to Source
This study rigorously evaluates the effectiveness of male circumcision as a tool to prevent human immunodeficiency virus (HIV) infection in rural Uganda. Enrolling men aged 15 to 49 years residing in Rakai district, the researchers randomly assign half of the 4,996 participants to the intervention group (immediate circumcision) and the remaining into the control group (delayed circumcision). Both groups are then followed for 24 months, with three rounds of HIV testing, medical examinations and behavioural interviews at 6, 12 and 24 months after intervention delivery, respectively.
Theoretically, male circumcision can lead to a reduction in the risk for HIV infection, because the inner surface of the human foreskin has a nine times greater density of HIV target cells compared with cervical tissue. The present study, together with similar studies in Kenya and South Africa, helps quantify the magnitude of these prevention effects.
The primary statistical analysis compares rates of HIV incidence across treatment groups. For each follow-up round, the authors report average rates of HIV infection amongst control and treatment individuals, measured as incidence per 100 person-years. In addition, they report p values associated with a null hypothesis of equality between the two groups. Beyond comparing overall group averages, the authors also examine possible heterogeneous treatment effects by disaggregating the results by age group, marital status, education and reported sexual behaviour. Secondary outcomes include the prevalence of self-reported symptoms of sexually transmitted infections such as genital ulcer disease. Finally, they examine the evidence for possible behavioural responses to circumcision along the dimensions of condom use and number of sexual partners.
The main finding of the study is that male circumcision reduces the risk for HIV infection in men by half. Specifically, the intervention group had a cumulative incidence rate of 0.66 per 100 person-years, compared with 1.33 for the control group. This corresponds to a risk ratio of 0.49 (95 per cent confidence interval = 0.28 to 0.84), and a p value of0.0057, indicating statistical significance at the 1 per cent level. It should be noted that the intervention appears to be effective only after 6 months, because results at 6-month follow-up did not show a significant difference between treatment and control individuals (risk ratio = 0.76, p = .439).
In terms of heterogeneous treatment effects, the intervention did not differentially affect participants according to their age, marital status or education. There is some weak evidence that people with multiple sexual partners may have experienced greater gains from the intervention than those with a single sexual partner (risk ratios of 0.30 and 0.55, respectively). Similarly, people with non-marital sexual partners appear to have benefited more than people with no such partners (risk ratios of 0.34 and 0.64, respectively). These ratios are not precisely estimated, however.
There were some significant programme impacts for self-reported symptoms of sexually transmitted infections. In particular, the intervention group experienced many fewer episodes of genital ulcer disease (168 vs. 322 cases, statistically significant at the 1 per cent level). Other symptoms, such as genital discharge and dysuria, did not show significant effects. Finally, there was no evidence of any behavioural change as a result of the intervention.