Male circumcision substantially decreased the risk of HIV infection in two recent analyses of data from Sub-Saharan Africa. According to a meta-analysis that pooled data from 28 studies in eight countries, circumcised men are only about half as likely to contract HIV as men who have not been circumcised, and the protective effect of circumcision is even greater for men at high risk of HIV infection.¹ In addition, a population-based cohort study of HIV-negative men and HIV-discordant couples in rural southwestern Uganda found a lower incidence of HIV infection among circumcised men than among uncircumcised men.² Data from the HIV-discordant couples in that study suggest that male circumcision results in lower HIV infection rates among males, and may help to prevent male-to-female transmission when the HIV-positive male partner has a low viral load.

The Meta-Analysis

The authors identified 28 studies, reported in 22 papers published through April 1999, that examined the association between male circumcision and female-to-male transmission of HIV in Sub-Saharan Africa. These studies were conducted in eight countries and included 19 cross-sectional, five case-control, three cohort and one partner study.

Because the association between circumcision and HIV transmission could vary according to the background prevalence of HIV and other sexually transmitted diseases, the authors grouped studies using the following categories: population-based studies, studies of men at high risk of sexually transmitted infections, and other studies (e.g., factory workers, volunteers). They used the relative risk as the standard measure of risk; the odds ratio was substituted for studies where the relative risk was unavailable. A random effects meta-analysis was used to calculate all pooled results.

Crude (i.e., unadjusted) relative risks were available for 27 of the 28 studies. Fourteen of these studies showed a statistically significant protective effect of circumcision on risk for HIV. When the results of the 27 studies were pooled, circumcised men were only half as likely as uncircumcised men to be infected (relative risk of 0.52).

Fifteen studies reported relative risks adjusted for one or more confounders, such as age, ethnic group, sexual risk behaviors and presence of sexually transmitted diseases; circumcision was found to have a statistically significant protective effect in 10 of them. The pooled adjusted relative risk (0.42) showed a slightly stronger protective effect than the pooled crude estimate.

The authors also calculated pooled adjusted relative risks for the population-based studies and the high-risk studies. The pooled adjusted relative risk calculated from six population-based studies showed circumcision to be protective (0.56). For the 12 studies that calculated crude estimates for men at high risk for HIV, the pooled adjusted analysis using seven high-risk studies found a greatly reduced risk of HIV associated with circumcision (0.29). Sensitivity analyses suggested that the meta-analysis was not unduly influenced by either the study with the largest sample size or by publication bias.

The investigators argue that the significant variation in the relative risk estimates across studies might be expected, given the differences in factors that affect HIV transmission, particularly the prevalence of certain sexually transmitted diseases. They point to a number of weaknesses in the studies used in the meta-analysis, including the limitations of observational studies in establishing causality and the potential inaccuracy of circumcision status determined through self-report rather than through clinical examination, but assert that the evidence is nevertheless compelling. The investigators conclude that consideration should be given to the acceptability of providing safe male circumcision services as an additional HIV prevention strategy in areas of Africa where men traditionally are not circumcised.

The Cohort Study

The data for analysis came from 5,507 HIV-negative men who participated in a trial of antibiotic treatment for sexually transmitted diseases conducted in 10 community clusters in rural southwestern Uganda between November 1994 and October 1998. These men were observed for a total of 10,231 person-years; HIV incidence per 100 person-years was used to calculate the rate ratio of HIV infection associated with circumcision. The investigators identified 187 HIV-negative men whose female partner was HIV-positive, in order to compare female-to-male transmission rates by circumcision status. In addition, male-to-female transmission rates were calculated based on 223 discordant couples in which the man was the HIV-positive partner.

More than four-fifths of the circumcised men were Muslim, compared with only 1% of uncircumcised men. Circumcised men were older and were more likely to have more than one wife and to abstain from alcohol. In the bivariate analysis, overall HIV incidence was lower for circumcised men than for uncircumcised men (rate ratio of 0.61). However, this association was not consistent across all subgroups. Circumcision at or before the age of 12 years was significantly associated with a decreased risk of HIV (0.54), but circumcision at 13 years of age or older was not. Similarly, circumcision was protective for selected subgroups such as ever-married men, those reporting no extramarital partners and those who abstained from alcohol consumption, but had no effect on other subgroups.

The authors used a multivariate Poisson regression model to obtain a rate ratio adjusted for age, marital status, age at time of circumcision, number of sex partners in the past year, exchange of sex for money or gifts, condom use and syphilis. Compared with uncircumcised men, circumcised men in general and men who had been circumcised at or before the age of 12 years were only half as likely to become infected with HIV (adjusted rate ratios of 0.53 and 0.49, respectively). There was no significant association between HIV incidence and circumcision performed at or after 13 years of age. Having three or more sexual partners, testing positive for syphilis, and being 20-39 years old were associated with an increased risk of HIV, independent of circumcision.

The authors were unable to assess how religion affected the association between circumcision and HIV because more than 99% of Muslims reported being circumcised. Among the circumcised men in the study, Muslim men aged 20-29 years had a lower risk of HIV infection than their non-Muslim counterparts (rate ratio of 0.24). Compared with circumcised non-Muslims, circumcised Muslims were significantly younger, were less likely to drink alcohol or to have ever married, and were more likely to have been circumcised before puberty and for religious rather than health reasons.

In the 187 discordant couples in which the woman was the HIV-infected partner, circumcision appeared to be extremely protective: No circumcised men became infected with HIV, while the HIV incidence among uncircumcised men was 16.7 per 100 person-years. This association was present regardless of the female partner's viral load. Among the 223 discordant couples in which the male was the HIV-positive partner, circumcision was significantly associated with lower HIV transmission to female partners only in the subgroup in which the male's viral load was less than 50,000 copies per milliliter. Male-to-female transmission rates were similar for circumcised and uncircumcised men with viral loads above 50,000 copies per milliliter.

While noting that their results suggest that circumcision has a protective effect against HIV infection for men, the authors caution that this effect was not observed in all subgroups and that "the interpretation of these observational data on circumcision [is] complex." They point out that because Muslim affiliation is highly correlated with circumcision and is also associated with behaviors likely to lower the risk of HIV infection, "it may be impossible to determine the effects on reduced HIV incidence caused by Islamic religion and culture from the separate biological effects of circumcision per se." They cited reasons for circumcision and misclassification of reported circumcision status as other potential sources of bias.

Conclusion

According to the authors, these studies break new ground in the investigation of a possible association between male circumcision and HIV incidence: The meta-analysis improves on the methodology and scope of its most recent predecessor, and the cohort study is the first to be conducted using a "representative, population-based cohort."

The authors of the meta-analysis conclude that it provides "compelling evidence of a substantial protective effect of male circumcision against HIV infection in Sub-Saharan Africa," especially among men at high risk for HIV, and recommend that feasibility studies and randomized trials be done to explore the potential of male circumcision as an HIV prevention strategy. The authors of the cohort study are more cautious in interpreting their data, saying only that "male circumcision may protect HIV-negative men from acquiring HIV infection to varying degrees." They suggest that clinical trials are needed before circumcision can be promoted as a means of HIV prevention, but also acknowledge the "major difficulties in design and execution" and "ethical obstacles" that such trials present. --A. Hirozawa

REFERENCES

1. Weiss HA et al., Male circumcision and risk of HIV infection in sub-Saharan Africa: a systematic review and meta-analysis, AIDS, 2000, 14(15):2361-2370.

2. Gray RH et al., Male circumcision and HIV acquisition and transmission: cohort studies in Rakai, Uganda, AIDS, 2000, 14(15):2371-2381.