The efficacy of male circumcision for HIV prevention in men has been established in three randomized trials conducted in South Africa, , Kenya , and Uganda , and following WHO/UNAIDS recommendations, male circumcision services for HIV prevention are being implemented in several sub-Saharan African countries . However, the long-term effectiveness of male circumcision for HIV prevention has not been clearly established. The Kenyan trial reported that HIV incidence was reduced by 64% in circumcised men over 54 months of observation suggesting prolonged effectiveness of male circumcision . Also, long-term effectiveness can be inferred from observational studies which show an association between male circumcision, largely performed during childhood, and subsequent reduced HIV prevalence and incidence among circumcised adult men [6–11].
There may be self-selection of men accepting circumcision and the public health impact on HIV incidence will depend on the proportions and risk profiles of men adopting male circumcision . Additionally, there is a concern that the belief that male circumcision reduces HIV risk may lead to behavioral disinhibition or risk compensation, whereby men and their sexual partners may become complacent and practice increased risky sexual behaviors. Evidence for risk compensation was observed in the South African trial, in which circumcised men reported more risky sex than uncircumcised men , but this was not observed in the Kenyan and Ugandan trials [2,3,13]. However, the intensive risk reduction strategies during the trials may have mitigated risk compensation in the latter studies [2,3]. After completion of the trials, male circumcision services were provided to control arm participants and health education was less intensive than during trial, so it is important to assess whether the circumcised control arm men showed riskier sexual behaviors compared to those who remained uncircumcised after trial completion.
Given these considerations, we report here the findings with regard to the posttrial effectiveness of circumcision for HIV prevention, the sociodemographic characteristics of men adopting or declining male circumcision, and sexual risk behaviors among circumcised and uncircumcised men after completion of the trial in Rakai, Uganda.
The Rakai trial has been reported in detail . In brief, 4996 HIV-negative uncircumcised men aged 15–49 years were consented, enrolled, and randomized to receive either immediate circumcision (intervention) or circumcision delayed for 24 months (controls). Participants were followed up at 6, 12, and 24 months to assess HIV incidence and sexual risk behaviors. The trial was stopped early on 12 December 2006 following an unplanned interim analysis, which demonstrated efficacy of male circumcision for HIV prevention. The characteristics of intervention and control arm men were comparable at enrollment and follow-up, and cumulative retention rates over 24 months were approximately 84% in both study arms (Fig. 1).
After closure of the trial, male circumcision was offered as a service to uncircumcised control arm participants and intervention arm crossovers, and surgeries were conducted as expeditiously as possible. In addition, all trial participants were consented and enrolled into a posttrial surveillance study, which was integrated into the schedule of an ongoing cohort study (the Rakai Community Cohort Study). By the data cut-off date for this analysis (15 December 2010), all participants had been contacted for their first posttrial visit, and further follow-up is ongoing.
All trial participants received voluntary counseling and testing (VCT) at enrollment. Seroconverters during or after the trial received repeat VCT and posttest counseling and were referred to HIV care provided by the Rakai program under Presidents Emergency Program for AIDS Relief funding. At each trial and posttrial visit, men provided venous blood for HIV testing and were interviewed in private by trained male interviewers to ascertain sexual behaviors. Circumcision status was confirmed by observation. HIV status was assessed by two enzyme immunoassays: Vironostika HV-1 (Organon Teknika, Charlotte, North Carolina, USA) and Cambridge Biotech (Worcester, Massachusetts, USA). Discordant enzyme immunoassay results and all seroconversions were confirmed by Western blot (bioMérieux VITEK, St Louis, Missouri, USA).
We conducted an as-treated analysis during the 2 years of trial follow-up and approximately 4.79 years posttrial surveillance. Men were classified by their actual circumcision status, irrespective of their initial arm of randomization. Thus, during the trial, intervention arm men who failed to come for circumcision within 6 months of randomization were classified as uncircumcised crossovers (n = 146), and control arm men who received circumcision from nontrial sources were classified as circumcised at the time they received surgery (n = 33). During the posttrial period, we continued to classify men by their actual circumcision status at each visit.
HIV incidence was estimated per 100 person-years, assuming that infection occurred at the midpoint of the follow-up interval between the last negative and first positive HIV serologic test. For the posttrial period, person time was accrued from the last trial visit to the last available posttrial follow-up visit. Incidence rate ratios (IRRs) and 95% confidence intervals (CIs) were estimated by Poisson regression. Additionally, Kaplan–Meier survival analysis and Cox regression modeling was used to estimate the adjusted hazard ratios (adjHRs) of HIV acquisition, and the effectiveness of male circumcision during posttrial surveillance was estimated as 1 – adjHR. The date of the last trial visit was considered the time origin for the Cox regression. For incident cases, the time from last trial visit to the midpoint between the last negative and first positive serologic test was used to estimate the time to HIV infection during posttrial follow-up, and for nonincident cases, observation time was censored at the last available follow-up visit. The Cox model was adjusted for potential confounders identified from earlier Rakai studies. Fixed covariates included age, marital status, and education at the 24-month trial visit. Time-varying posttrial covariates included sexual behaviors reported during a 12-month referent period prior to interview and included any sexual activity (yes/no), number of sex partners (none/single/multiple), nonmarital sexual relationships (yes/no), condom use (never/inconsistent/consistent), and alcohol consumption before sex (yes/no). For the posttrial period, we assessed HIV incidence among all circumcised and uncircumcised men (including intervention arm participants circumcised during the trial), and separately among control arm participants who accepted or declined male circumcision after completion of the trial.
To assess possible self-selection of men accepting circumcision after the trial, we tabulated sociodemographic characteristics and sexual behaviors of uncircumcised HIV-negative men reported at the last (24 month) trial visit, stratified by whether the men subsequently accepted or declined male circumcision by the time of their first posttrial visit. Additionally, to assess potential behavioral disinhibition among this group of men, we tabulated their sexual risk behaviors reported at the first completed posttrial visit stratified by their circumcision status at this visit. Sexual risk behaviors reported at the first completed posttrial visit were also tabulated for men circumcised during the trial (i.e., mainly intervention arm men). Chi-squared tests were used for statistical inference on dichotomous and categorical behavioral variables, and Wilcoxon-rank test was used for comparison of number of sex partners.
The studies were reviewed and approved by four Institutional Review Boards in Uganda and the USA. The trial was monitored by an independent Data Safety Monitoring Board, and the trial was registered with Clinical Trials.Gov number NCT 00425984.
A total of 4145 HIV-negative men were followed up during the posttrial period of whom 3566 (86.0%) provided information at at least one follow-up visit (Table 1 and Fig. 1). By the data cut-off date, these men provided 10 715.8 person-years observation over approximately 4.75 years.
There were 2134 randomized control arm participants available at the last trial visit, of whom 33 (1.5%) were circumcised control arm trial crossovers (i.e., circumcised by other sources), and 51 (2.4%) had seroconverted during the trial (Fig. 1). Including the intervention arm crossovers, there were 2196 (2144 HIV-negative) uncircumcised men at the last trial visit. Among the 1602 initially HIV-negative and uncircumcised men followed at the first posttrial visit, 1261 (78.7%) had received circumcision and 341 (21.3%) had chosen not to be circumcised.
Table 1 shows HIV incidence by circumcision status during the trial and posttrial surveillance for all men and separately for control participants. For all participants, HIV incidence over the 2-year trial follow-up was 0.47/100 person-years in circumcised men and 1.14/100 person-years in uncircumcised men, corresponding to an as-treated circumcision effectiveness of 59% (95% CI 32–75%, P = 0.0003). During posttrial surveillance, overall HIV incidence was 0.50/100 person-years in circumcised men and 1.93/100 person-years in the remaining uncircumcised men, corresponding to an effectiveness of 74% (95% CI 57–85%, P < 0.0001). The difference between these estimates was not statistically significant. Figure 2a shows the Kaplan–Meier survival curves for time from the last trial visit to estimated HIV infection according to time-dependent circumcision status during the posttrial period. For all trial participants, the hazard ratio between circumcised and uncircumcised men was 0.27 (95% CI 0.16–0.44, P < 0.0001). After adjusting for sociodemographic characteristics at the last trial visit and time-dependent sexual behaviors during posttrial follow-up, the adjHR was 0.27 (95% 0.16–0.45), and the adjusted effectiveness was 73% (95% CI 55–84%).
Among control arm participants (Table 1, lower panel), there were no incident events in the 33 control arm crossovers during the trial. During posttrial follow-up, HIV incidence was lower in the controls who opted for circumcision (0.54/100 person-years) than in the remaining uncircumcised controls (1.71/100 person-years), corresponding to an IRR of 0.32, and the effectiveness was 68% (95% CI 40–83%, P = 0.0054). Figure 2b shows the survival curves for time-to-HIV infection by time-dependent circumcision status among these control participants. The unadjusted HR between circumcised and uncircumcised controls was 0.32 (95% CI 0.17–0.61), and the adjusted HR was 0.33 (0.17–0.62) with an effectiveness estimate of 67% (95% CI 38–83%) after adjustment for sociodemographic characteristics and time-dependent sexual behaviors.
To assess self-selection of posttrial circumcision service acceptance among men who were uncircumcised and HIV-negative during the trial, we tabulated their characteristics and behaviors reported at the last trial visit, stratified by whether they subsequently accepted or declined male circumcision by the first posttrial visit (Table 2). At trial closure, there were no statistically significant differences in sociodemographic characteristics or sexual risk behaviors among men who opted for male circumcision compared to those who declined male circumcision, except that men who opted for male circumcision were slightly more likely to report no sexual activity in the prior 12 months (14.3%) than those declining male circumcision (10.6%, P = 0.08). Among sexually active men, those opting for circumcision more frequently reported multiple sex partners in the prior 12 months (39.7%) than men who remained uncircumcised (34.1%, P = 0.08). The median number of sex partners did not differ significantly between men opting for male circumcision [1 partner, interquartile range (IQR) 1–2] and men remaining uncircumcised (1 partner, IQR 1–2, P = 0.75).
A total of 4233 men were seen at the last trial visit, of whom 3115 (73.6%) completed the first posttrial visit and were included in the analysis of risk behaviors (Fig. 1). The men who were observed at the last trial visit but who missed the first posttrial visit were younger and less likely to be sexually active than those followed up (age under 25 years, lost to follow-up 48.2%, followed up 42.0%, P < 0.001. not sexually active lost to follow-up 15.2%, followed up 12.8%, P = 0.04). The median interval and IQR from the last (24 month) trial visit to the first posttrial surveillance visit was 1.67 years (IQR 1.30–2.30) for the intervention arm and 1.67 years (IQR 1.30–2.27) for the control arm participants To assess possible behavioral risk compensation, we tabulated risk behaviors reported at the first posttrial visit among men who accepted or declined circumcision after the trial (Table 3). There were no statistically significant differences in sexual behaviors reported by circumcised and uncircumcised men at their first posttrial visit. The proportion of nonsexually active men declined over follow-up time irrespective whether they opted for or declined male circumcision (Tables 2 and 3). The posttrial risk behaviors of men circumcised during the trial are also shown in Table 3 and were not significantly different from men who self-selected for male circumcision after the trial or men who remained uncircumcised.
The as-treated effectiveness of male circumcision for HIV prevention was 59% over 24 months during a randomized trial, and the adjusted effectiveness of circumcision over approximately 4.8 years of posttrial observation when men could opt for or decline free circumcision services was approximately 73% (Table 1, Fig. 2). Our findings are comparable to those from a Kenyan investigation in which efficacy during the trial was 53%  and effectiveness of circumcision was 64% over 4.5 years of observation . It is possible that the higher effectiveness after than during the trial, although not statistically significant, reflects some unmeasured self-selection of men choosing not to accept male circumcision. The increase in HIV incidence among uncircumcised controls from 1.14/100 person-years during the trial to 1.71/100 person-years during the posttrial period (Table 1) may suggest that the men who declined male circumcision had un-measured higher risk behaviors; or that the reduced intensity of preventive education after trial closure may have resulted in riskier behaviors, whereas men opting for male circumcision after the trial received additional education and counseling during their surgery visit. Therefore, we conclude that the protection from HIV afforded by circumcision is likely to be sustained for a protracted period, as possibly suggested by observational studies showing reduced HIV incidence among adult men who were circumcised during infancy or childhood [8,9].
A high proportion of uncircumcised participants accepted circumcision after completion of the trial and, among men followed up at the first posttrial visit, circumcision uptake was 78.3%. There were no statistically significant differences in the characteristics or sexual risk behaviors between men subsequently choosing to be circumcised or to remain uncircumcised during posttrial surveillance (Table 2). This suggests minimal self-selection of men opting for male circumcision. Moreover, there were no statistically significant differences in sexual risk behaviors among circumcised and uncircumcised men at the first posttrial surveillance visit (Table 3) when men had been informed of the efficacy of circumcision for HIV prevention and the health education provided was, of necessity, less intensive than during the trial. This suggests that the procedure did not lead to significant risk compensation during 1–2 years after circumcision.
There are limitations to this study which preclude generalization of findings to programmatic male circumcision settings. Trial participants were self-selected because they volunteered to enroll and be randomized, and all received VCT as a criterion for trial eligibility. Also, trial participants received intensive health education at each trial visit, and the effect of this education may have modified their subsequent risk behaviors during the posttrial surveillance period. In addition, male circumcision uptake among men randomized to the control arm of the trial (78.3%) was higher than male circumcision uptake in the general Rakai cohort population of nontrial participants in which circumcision was offered as a service (∼25%, R.G., personal communication). There are also inherent limitations to the interpretation of the posttrial data. Men who were circumcised during the posttrial period spent part of each follow-up interval in an uncircumcised state. Although we know the dates of circumcision, we do not know when HIV infection occurred within a follow-up interval, so incident infections cannot be definitively ascribed to an uncircumcised or circumcised state. Thus, if men were circumcised during a follow-up interval when they seroconverted, we conservatively assumed the HIV infection occurred after circumcision, although the HIV infection may have preceded the procedure. This may have misclassified circumcision status for some incident HIV infections and biased our estimates toward the null. Retention rates were approximately 74% during the first posttrial follow-up interval (∼1.7 years after trial closure), and we could not assess behaviors among men lost to follow-up. It is also possible that differential behavior change may emerge after longer time intervals. Follow-up is ongoing and will not be completed until 2013, so behaviors at the second or subsequent incomplete posttrial visits could not be estimated for the majority of men (Fig. 1). Of necessity, men in the posttrial study were self-selected by their decision to accept or reject circumcision, and this could lead to a social desirability bias in reporting sexual behaviors. However, there were no significant differences in sociodemographic or sexual risk behaviors between men accepting or declining posttrial circumcision suggesting minimal evidence of self-selection bias.
In summary, the effectiveness of male circumcision for HIV prevention in a posttrial observational study was comparable to or higher than the as-treated effectiveness of circumcision observed during a randomized trial, and there was no evidence of significant self-selection or behavioral risk compensation. This suggests that male circumcision confers long-term protection from HIV infection in men.
The present study was supported by grants #U1AI51171 and 1UO1AI075115-O1A1 from the National Institutes of Health, Division of Allergy and Infectious Diseases.
The trial was registered with Clinical Trials.Gov number NCT 00425984.
Conflicts of interest
There are no conflicts of interest.
1. Auvert B, Taljaard D, Lagarde E, Sobngwi-Tambekou J, Sitta R, Puren A. Randomized, controlled intervention trial of male circumcision for reduction of HIV infection risk: the ANRS 1265 Trial
. PLoS Med
2. Bailey RC, Moses S, Parker CB, Agot K, MacLean I, Kieger JN, et al. Male circumcision for HIV prevention in young men in Kisumu, Kenya: a randomised controlled trial
3. Gray RH, Kigozi G, Serwadda D, Makumbi F, Watya S, Nalugoda F, et al. Male circumcision for HIV prevention in men in Rakai, Uganda: a randomised trial
4. UNAIDS. New data on male circumcision and HIV prevention: policy and programme implications.
Montreux: UNAIDS; 2007.
5. Bailey R, Moses S, Parker CB, Agot K, MacLean I, Krieger JN, et al
. The protective effect of adult male circumcision against HIV acquisition is sustained for at least 54 months: results from the Kisumu, Kenya trial.
In: International AIDS Society
; Vienna; 2010.
6. Reynolds SJ, Shepherd ME, Risbud AR, Gangakhedkar RR, Brookmeyer RS, Divekar AD, et al. Male circumcision and risk of HIV-1 and other sexually transmitted infections in India
7. Gray RH, Kiwanuka N, Quinn TC, Sewankambo NK, Serwadda D, Mangen FW, et al. Male circumcision and HIV acquisition and transmission: cohort studies in Rakai, Uganda. Rakai Project Team
8. Weiss HA, Quigley MA, Hayes RJ. Male circumcision and risk of HIV infection in sub-Saharan Africa: a systematic review and meta-analysis
9. Kelly R, Kiwanuka N, Wawer MJ, Serwadda D, Sewankambo NK, Wabwire-Mangen F, et al. Age of male circumcision and risk of prevalent HIV infection in rural Uganda
10. Lavreys L, Rakwar JP, Thompson ML, Jackson DJ, Mandaliya K, Chohan BH, et al. Effect of circumcision on incidence of human immunodeficiency virus type 1 and other sexually transmitted diseases: a prospective cohort study of trucking company employees in Kenya
. J Infect Dis
11. Siegfried N, Muller M, Deeks JJ, Volmink J. Male circumcision for prevention of heterosexual acquisition of HIV in men
. Cochrane Database Syst Rev
12. UNAIDS/WHO/SACEMA. Male circumcision for HIV prevention in high HIV prevalence settings: what can mathematical modelling contribute to informed decision making? PLoS Med
13. Mattson CL, Campbell RT, Bailey RC, Agot K, Ndinya-Achola JO, Moses S. Risk compensation is not associated with male circumcision in Kisumu, Kenya: a multifaceted assessment of men enrolled in a randomized controlled trial
. PLoS One