Male circumcision for HIV prevention: from evidence to action?

Introduction An estimated 2.5 million people were newly infected with HIV in 2007, of whom two-thirds live in sub-Saharan Africa 1. In the context of the urgent need for intensified and expanded HIV prevention efforts, the conclusive results of three randomized controlled trials (RCT) showing that male circumcision reduces the risk of HIV acquisition by approximately 60% 2–4 are both promising and challenging. Translation of these research findings into public health policy is complex and will be context specific. To guide this translation, we estimate the global prevalence and distribution of male circumcision, summarize the evidence of an impact on HIV incidence, and highlight the major public health opportunities and challenges raised by these findings. Male circumcision prevalence Male circumcision, one of the oldest and most common surgical procedures, is practised for religious, social and medical reasons. By reviewing nationally representative data sources and assuming that all Muslim and Jewish men are circumcised, we estimate that 30–34% of adult men are circumcised worldwide 5. Overall, an estimated 68% of circumcised men are Muslim and 1% are Jewish, with coverage almost universal in the Middle East, north Africa, Pakistan, Bangladesh and Indonesia (Fig. 1). Male circumcision is also practised for non-religious reasons either neonatally or as a rite-of-passage to manhood; and is very common in west Africa, parts of central and eastern Africa, the United States, Republic of Korea, and the Philippines 5. Within countries, prevalence can vary widely with religion, ethnicity and socioeconomic status 5,6.Fig. 1: Global map of male circumcision prevalence at country level, as of December 2006. No data; 80% prevalence. Source: World Health Organization. The boundaries and names shown and the designations used on this map do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement.Evidence that male circumcision reduces the risk of HIV infection Biological evidence Several plausible biological mechanisms could explain the increased risk of HIV and other sexually transmitted infections (STI) in uncircumcised men, including microtears and lesions in the mucosal surface of the inner foreskin and the longer survival of pathogens in the warm, moist subpreputial space. Most importantly, the inner foreskin is especially susceptible to HIV infection, as a result of a lack of keratinization and the high density of HIV target cells that are relatively accessible to infection compared with their deeper location under the keratinized surface of the outer foreskin and the glans 7,8. Observational evidence The hypothesis that male circumcision might protect against HIV infection was first suggested in 1986 9,10, and was subsequently supported by ecological descriptions of areas with low prevalence of male circumcision and high HIV prevalence in sub-Saharan Africa in the late 1980s 11,12, and later across 118 developing countries 13. Further evidence comes from two systematic reviews of observational studies comparing HIV risk between circumcised and uncircumcised men in the same populations 14,15. One, restricted to sub-Saharan Africa, included 27 studies 14, and the other was a global review including 37 studies 15. Circumcised men were consistently found to be at lower risk of HIV infection, and a meta-analysis of the 15 studies that adjusted for potential confounders showed this reduction to be large and highly statistically significant adjusted risk ratio (RR) 0.42, 95% confidence interval (CI) 0.34–0.54 14. Subsequent studies have found similar significantly reduced risks among circumcised men 16–18. Evidence from the randomized controlled trials Although compelling, the observational data do not prove causality, and three RCT of circumcision among consenting, healthy adult men in Uganda, Kenya and South Africa were initiated in 2002–2003. Each trial was halted early after recommendations by independent Data and Safety Monitoring Boards in 2005–2006, when interim analyses found a highly significant reduced risk of HIV seroconversion among the men randomly assigned to circumcision 2–4. In total, 10 908 uncircumcised, HIV-negative adult men were randomly assigned to intervention or control arms, and followed for up to 2 years (Table 1). Overall retention rates were high (86–92% at the end of follow-up, when men in the control arms were offered circumcision). HIV incidence was considerably lower in Uganda (1.33 per 100 person-years in the control arm) than in the other two sites (2.1 per 100 person-years; Table 1), possibly reflecting overall lower incidence in this population and the inclusion of older men in the trial.Table 1: Summary of the three randomized controlled trials of male circumcision on HIV acquisition in sub-Saharan Africa.Table 1 shows the cumulative risk among men who were HIV negative at enrolment, estimated using intention-to-treat Kaplan–Meier analysis. There have been no previous RCT of adult male circumcision 15, and to summarize the protective effects seen in the trials, we conducted a random-effects meta-analysis of results of these three trials, following the recommendations of the QUORUM statement for reporting trials as appropriate 19. There was no evidence of heterogeneity between the trials (P = 0.86), and the summary rate ratio was 0.42 (95% CI 0.31–0.57; Fig. 2), corresponding to a protective effect of 58% (95% CI 43–69%), identical to that found in the observational studies (58%, 95% CI 46–66%) 14.Fig. 2: Random-effects meta-analysis for the randomized controlled trials intention-to-treat analysis, with summary risk ratio for the observational data. CI, Confidence interval. aRisk ratio based on 15 studies that adjusted for potential confounders 14.The true biological protective effect of male circumcision, however, may be better estimated by an ‘as-treated’ analysis 20, which assigns person-time according to the actual circumcision status of participants. In each trial, not all men adhered to the arm they were randomly assigned to. For example, in the South African trial, 10.3% of men randomly assigned to the control arm had been circumcised outside the trial at month 21. This was greater than in the other trials (1.1–1.3%) perhaps because of the greater local availability of male circumcision services. In each trial, approximately 5–6% of men randomly assigned to be circumcised declined surgery. An ‘as-treated’ meta-analysis of the three trials shows a stronger effect than the intention-to-treat analysis (summary RR 0.35, 95% CI 0.24–0.54). The Ugandan trial reported efficacy in subgroups, and in general found greatest efficacy among men at higher risk (those with two or more partners during follow up, non-marital sexual partners, reporting transactional sex or having a history of genital ulcers); i.e. approximately 70% risk reduction. These results agree with previous observational data suggesting a stronger protective effect in high-risk populations (summary RR 0.29, 95% CI 0.20–0.41) compared with general populations (RR 0.56, 95% CI 0.44–0.70) 14. The Ugandan and Kenyan trials found that circumcised men were at approximately half the risk of self-reported or clinically diagnosed genital ulcer disease (GUD) during the trial. This suggests that the stronger protective effect in high-risk groups may be caused partly by circumcision protecting against other STI, especially GUD 21, thus providing additional indirect protection against HIV 22. Models based on the Kisumu data estimate that approximately 10–20% of the HIV infections prevented by male circumcision were caused by efficacy against STI 23. In the Kenyan and Ugandan trials there was little evidence of a protective effect until 6–12 months after randomization. In contrast, in South Africa, a protective effect was seen within 1–3 months (RR 0.23, 95% CI 0.05–1.04). These differences may be due to chance, or differences in behaviour such as the resumption of sex before complete wound healing (which can take up to 6 weeks). Also, in the Kenyan trial, four men in the circumcision arm seroconverted within a month of randomization, and assuming a short period of abstinence after surgery, are likely to have been already infected at baseline. Pooled analyses of the trial data focusing on early seroconvertors would help determine when and how protection begins. What are the implications of the trials stopping early? Larger than expected treatment effects that result in trial termination may be due to chance, but the risk of overestimating the treatment effect decreases when the number of events is over approximately 200 24. After the South African trial was published, it was suggested that inferences from the trial may be weak because the study was stopped early 25. There are, however, several reasons why early termination is unlikely to bias the trial results. First, all three trials had conservative predetermined stopping rules that were met. Second, the consistency of the results and indication of a somewhat stronger effect of the intervention over time in two of the trials argues that, if anything, the early stopping may have underestimated the effect. Third, the overall number of events is greater than the suggested threshold of 200 24. Finally, the observed effect in each of the male circumcision trials is not larger than expected, but is identical to that seen in previous observational studies. The findings of the male circumcision trials are in contrast to the recent disappointing results of other trials of HIV prevention tools, including the cellulose sulfate microbicide, the female diaphragm and gel, herpes simplex virus suppressive therapy and, most recently, an adenovirus-5-based HIV vaccine 26–29. These results highlight the need to expand services for confirmed HIV prevention strategies including safe adult male circumcision. Public health relevance of the trial results Responding to the conclusive evidence that male circumcision offers significant protection for men from HIV infection, several countries are planning to introduce or expand safe male circumcision programmes, including Kenya, Zambia, Swaziland and Rwanda. International funding agencies are also backing this strategy, with programmes such as the US President's Emergency Plan For AIDS Research (PEPFAR) providing funds to complement domestic funding for expanded circumcision services. Furthermore, the Agence Nationale de Recherche sur la SIDA (ANRS), the Bill and Melinda Gates Foundation and the US National Institutes of Health are supporting operational and related research. Among the major concerns about the expansion of male circumcision services for HIV prevention are surgical complications, the potential for men to increase their risky sexual behaviour if they believe themselves to be fully protected, the optimal messages to relay about offering male circumcision services to men who are HIV seropositive, and the costs and opportunity costs of expanding services in often overstretched health systems. The trials provide initial insights into these issues; however, further operational research is needed to evaluate these concerns in the ‘real world’. Complications of male circumcision Adolescent or adult circumcision requires suturing and can cause bleeding and, more rarely, haematoma or sepsis. Comparing the adverse event rates in the three trials is complex, as different definitions and criteria were used. In the Kenyan trial, adverse events possibly, probably or definitely related to circumcision occurred in 23 of 1334 circumcised participants (1.7%). All adverse events were mild or moderate and resolved with treatment within hours or days. In the South African trial, the adverse event rate was 54 per 1495 (3.6%) in HIV-negative men. In Uganda, the risk of an adverse event related to surgery was higher, at 7.6% (178/2328). This may be attributable to differences in adverse event case management. The risk of moderate adverse events related to surgery was 3% and there were five severe adverse events (0.2%). All of these events were successfully managed and resolved. These trial data indicate that adult male circumcision can be safely undertaken in limited-resource settings when performed in a clinical setting by experienced, well-trained providers. Similar conclusions were found from a recent review of complications of male circumcision in Anglophone Africa 30. When male circumcision is undertaken in un-antiseptic conditions, however, by inexperienced providers with inadequate instruments, or with poor aftercare, serious complications or even death can result 31. It is possible that unmet demand for male circumcision may result in an increase in non-medical circumcision services offered by untrained individuals as a means of income generation, with a heightened risk of significant harm. To assist in preventing these problems, WHO/UNAIDS/JHPIEGO have produced a manual for performing adult male circumcision under local anaesthesia 32. National policies, however, are needed to maximize the safety, efficiency, and availability of male circumcision service provision. Behaviour change after male circumcision The adoption of, or increase in, unsafe sex practices (‘risk compensation’) after adult circumcision could potentially offset the protective effect of male circumcision 33. The Rakai trial found no differences in sexual behaviour during the trial by circumcision status. The South African trial showed a significantly increased mean number of sex acts between 4 and 21 months among men in the circumcision arm, but not an increase in the number of sexual partners or a change in condom use. In the Kenyan trial there was a decline in reported risk-taking behaviour during the 24 months of follow-up in both arms. At 24 months, however, significantly fewer men in the control arm reported unprotected sexual intercourse (46 versus 51%) and these men were also more likely to report consistent condom use (41 versus 36%). There was also a tendency for a greater proportion of the uncircumcised men to report practising sexual abstinence at 24 months (18 versus 14%). Although reassuring, these trial data may not be generalizable. The trials provided the highest standards of preventive care, with men receiving intensive, individual counselling and without knowing that circumcision reduced their risk of HIV. The challenges of expanding services within already overstretched health systems include the need to provide adequate counselling to convey the message that male circumcision is a risk-reduction strategy that provides partial protection only. The only data on sexual behaviour after adult male circumcision outside a clinical trial setting the RCT findings. In a study of men in Kenya of whom half had to circumcised, circumcised men were no more likely to report risky sexual behaviour of unprotected sex number of partners, condom during the month period than uncircumcised men. were when the period was This study suggests that, within the context of adequate counselling on risk circumcised men not increase their risky but this study was conducted before of the RCT findings. Further follow-up studies of men to be circumcised are needed as male circumcision services are expanded and by individuals and In further strategies to counselling and including among men already in settings is of male circumcision in African about the of male circumcision in Africa studies from countries in sub-Saharan Africa have shown that of of uncircumcised men to circumcised, of circumcision for their partners, and of men and would their The of among uncircumcised men was from a study in eastern Uganda in before male circumcision was more widely as possibly with HIV recently, the data from Rakai that 60% of men were to be circumcised to HIV more than half of uncircumcised men in the were to The to were of and with lower risk of STI and other health the These data that and ethnicity are not major to the of male circumcision in most of sub-Saharan of expanding male circumcision services a having and an history in parts of the the expansion of male circumcision services including and legal The protection and of is to all of HIV prevention and care, and all male circumcision services that the is under of and without Further research will be needed in different settings to a better of the and the of circumcision among different and to appropriate and counselling circumcision is a and than adult circumcision and for reasons of and countries may to include circumcision, under as a HIV prevention This would have the additional of the risk of infections in the first of impact of male circumcision in sub-Saharan Africa that expanded services can have a effect on HIV incidence in a very The impact could be greater than the efficacy if a large proportion of men circumcised, and, assuming full male circumcision could (95% CI million HIV infections and (95% CI million over the 10 years in sub-Saharan Africa, and million (95% CI HIV infections and (95% CI million in the following 10 years study based on in in Kenya, and also found that male circumcision programmes in large and in HIV prevalence over time An to the impact of the coverage of male circumcision is by at the ecological which represent a with very high or universal coverage of male circumcision without relatively low and HIV prevalence which approximately The of male circumcision in the HIV was in the which found higher of reported risk-taking behaviour in the city of prevalence been at for compared with Kenya and HIV is more The that biological for HIV male circumcision and herpes simplex virus 2 infection, were likely to be in the HIV in sub-Saharan Africa and this been confirmed in of the data of male circumcision for HIV prevention in sub-Saharan Africa data from the Ugandan and South African trials indicate that male circumcision is likely to be a even intervention The South African estimate was for HIV prevalence is and the of men are In this assuming full coverage of male circumcision, and using data from the trial, the per HIV infection was CI with of million over years per Similar findings were seen in it is estimated to per HIV infection In contrast, in Uganda, with an HIV incidence of per 100 assuming 60% efficacy against but not and would be needed to one HIV infection over 10 at a of per HIV infection over 10 years costs of HIV infections were not included in that study and thus potential were not the of circumcision are likely to be and of male circumcision is very likely to be a of findings for female partners do not male circumcision reduces the risk of HIV Observational data from Uganda had suggested that circumcision reduced the risk of HIV but recent data from Uganda and little protective effect An RCT of the impact of male circumcision in protecting against HIV in Uganda, was stopped at an interim analysis in December as there was little of a statistically significant impact a protective effect be and the reduced risk of GUD among circumcised men would be expected to the risk of in high HIV prevalence settings with low male circumcision prevalence will if there is a reduction in HIV incidence among men who are circumcised, especially in programmes coverage This is because would have a lower of a sexual with HIV In as for men, there are to female partners of circumcised men, most a lower risk of infection and and also possibly a lower risk of male circumcision services are expanded without appropriate and individual however, there is the potential for increased for to sex behaviour if their male partners believe themselves to be from infection female partners be included in the and counselling of men circumcision to provide for to to the risk of the early resumption of sexual There are also concerns the potential for between messages for male circumcision and female genital female circumcision, in which parts of the sexual of are There are no health with Although in all areas of Africa men are not circumcised there is also no of it is that the of male circumcision it from of findings for men who have sex with men sex between men was in the three African trials, and the implications of the trial findings for men who have sex with men are HIV from intercourse is to the a risk unlikely to be by circumcision status. It however, plausible that male circumcision provides partial protection against HIV as it for There are observational studies and no RCT of the impact of circumcision on HIV among and results are evidence may only from an RCT among the message for all men, of sexual or circumcision be that practising sex including and consistent condom is the to In randomized controlled trials have provided conclusive evidence that male circumcision provides approximately 60% protection against the acquisition of HIV in men. Male circumcision provides only partial protection against HIV and it is that services are within HIV prevention programmes, with counselling messages and the message that the will but not HIV the increased of accessible safe adult male circumcision services could also increase opportunities to men about and sexual health including and the need for HIV prevention The of male circumcision for HIV prevention by the World Health Organization and will probably increase the demand for safe adult male circumcision services in settings with high rates of HIV Several agencies have to the expansion of and individual countries are this would be an appropriate to their HIV prevention programmes, the local of and of the and of these programmes will be needed to that counselling and follow-up and that service expansion to other health For reasons of safety, and countries may also to circumcision as a In male circumcision provides a needed to the HIV prevention It is not a or but possibly the and the most surgical The evidence from the trials is and the challenges to be of