Doi:10.1016/s0140-6736(07)60313-4

Articles
Male circumcision for HIV prevention in men in Rakai,
Uganda: a randomised trial
Ronald H Gray, Godfrey Kigozi, David Serwadda, Frederick Makumbi, Stephen Watya, Fred Nalugoda, Noah Kiwanuka, Lawrence H Moulton, Mohammad A Chaudhary, Michael Z Chen, Nelson K Sewankambo, Fred Wabwire-Mangen, Melanie C Bacon, Carolyn F M Williams, Pius Opendi, Steven J Reynolds, Oliver Laeyendecker, Thomas C Quinn, Maria J Wawer Summary
Background Ecological and observational studies suggest that male circumcision reduces the risk of HIV acquisition Lancet 2007; 369: 657–66
in men. Our aim was to investigate the eff ect of male circumcision on HIV incidence in men.
See Editorial page 615
See Comment page 617
Methods 4996 uncircumcised, HIV-negative men aged 15–49 years who agreed to HIV testing and counselling were See Perspectives page 635
enrolled in this randomised trial in rural Rakai district, Uganda. Men were randomly assigned to receive immediate See Articles page 643
circumcision (n=2474) or circumcision delayed for 24 months (2522). HIV testing, physical examination, and See Viewpoint page 708
interviews were repeated at 6, 12, and 24 month follow-up visits. The primary outcome was HIV incidence. Analyses Johns Hopkins University,
were done on a modifi ed intention-to-treat basis. This trial is registered with ClinicalTrials.gov, with the number Bloomberg School of Public
NCT00425984.
Health, Baltimore, MD, USA
(Prof R H Gray MD,
Prof L H Moulton PhD,
Findings Baseline characteristics of the men in the intervention and control groups were much the same at enrolment. M A Chaudhary PhD,
Retention rates were much the same in the two groups, with 90–92% of participants retained at all time points. In the M Z Chen MSc,
modifi ed intention-to-treat analysis, HIV incidence over 24 months was 0·66 cases per 100 person-years in the Prof M J Wawer MD); Rakai
intervention group and 1·33 cases per 100 person-years in the control group (estimated effi
cacy of intervention 51%, Health Sciences Program,
Entebbe, Uganda
95% CI 16–72; p=0·006). The as-treated effi
cacy was 55% (95% CI 22–75; p=0·002); effi
cacy from the Kaplan-Meier (G Kigozi MBChB,
time-to-HIV-detection as-treated analysis was 60% (30–77; p=0·003). HIV incidence was lower in the intervention F Nalugoda MHS,
group than it was in the control group in all sociodemographic, behavioural, and sexually transmitted disease N Kiwanuka MBChB,
symptom subgroups. Moderate or severe adverse events occurred in 84 (3·6%) circumcisions; all resolved with P Opendi MBChB); Makerere
University, Institute of Public
treatment. Behaviours were much the same in both groups during follow-up.
Health, Kampala, Uganda
(D Serwadda MBChB,
Interpretation Male circumcision reduced HIV incidence in men without behavioural disinhibition. Circumcision F Makumbi PhD,
can be recommended for HIV prevention in men.
F Wabwire-Mangen PhD);
Makerere University, Mulago
Hospital, Department of
Introduction
Surgery, Urology Unit,
A number of ecological and observational studies, Patients
Kampala, Uganda
mainly from sub-Saharan Africa, have suggested that Our aim was to enrol 5000 HIV-negative, uncircumcised (S Watya MBChB); Makerere
University, Department of
male circumcision reduces the risk of HIV infection in men aged 15–49 years who agreed to receive their HIV Medicine, Kampala, Uganda
men.1–5 A meta-analysis of cross-sectional and results through voluntary counselling and HIV testing (N K Sewankambo MBChB);
prospective studies estimated that the adjusted provided by the study, and who consented to be randomly National Institute of Allergy
summary rate ratio of male HIV acquisition associated
assigned to receive circumcision within about 2 weeks of and Infectious Diseases,
National Institutes of Health,
with circumcision in general populations was 0·56 enrolment (intervention group), or to have circumcision Bethesda, MD, USA
(95% CI 0·44–0·70); in high-risk populations the delayed for 24 months (control group). Screening and (M C Bacon MPH,
adjusted summary rate ratio was 0·29 (0·20–0·41).1
enrolment was done in a central study facility and in C F M Williams PhD, However, observational fi ndings do not consistently mobile facilities in the rural communities. Before S J Reynolds MD, show protective associations in all studies, and to screening, participants were informed of study Prof T C Quinn MD); and Johns
exclude the possibility of confounding due to procedures and risks through verbal presentations, Hopkins Medical Institutions,
diff erences in sexual risk behaviours and cultural or written materials, and an information video. After Baltimore, MD, USA
religious practices associated with circumcision is providing written informed consent for screening, a (S J Reynolds, O Laeyendecker,
cacy of circumcision venous blood sample was obtained for HIV testing, and Correspondence to: for HIV prevention can be determined only by participants were given a physical examination. Men who Prof Ronald H Gray, Johns randomised trials. One randomised trial done in South had contraindications for surgery (eg, anaemia, active Hopkins University, Bloomberg Africa was ended early after an interim analysis showed genital infection, or other health risks) were treated, and School of Public Health, Suite that circumcision reduced HIV incidence by 60% if their medical condition resolved, they were re-screened E4132, 615 N Wolfe Street, (32–76).6 Two other randomised trials, one in Kisumu, and were enrolled into the trial if eligible. Those with [email protected]
Kenya and the other in Rakai, Uganda—the results of anatomical abnormalities (eg, hypospadias) were
which we report here—were also stopped early on excluded and referred to the urologist (SW) for
December 12, 2006, after interim analyses showed management. Men who had medical indications for
signifi cant effi
surgery (eg, severe phimosis) were excluded from the www.thelancet.com Vol 369 February 24, 2007
Articles
trial and were off ered circumcision as a service. Men who and the need to abstain from intercourse until complete were HIV positive or declined to receive their HIV results wound healing had been certifi ed by a clinical offi were enrolled in a complementary trial that will be (equivalent to a physician’s assistant). Participants were reported separately.
off ered an information sheet to share with their wives or Eligible participants were asked to provide an additional partners, explaining wound care, hygiene, and the need written informed consent for enrolment. The consent to abstain from intercourse until wound healing was forms described the risks and benefi ts of participation, complete. Surgery was provided within 2 weeks of randomisation, and other trial procedures, and provided enrolment to 2255 (91%) of the men in the intervention information on HIV prevention (sexual abstinence, group; the median interval from enrolment to surgery monogamous relationships with an uninfected partner, was 2 days and the maximum delay was 149 days.
or consistent condom use). At enrolment, participants Circumcisions were done by trained and certifi ed completed a detailed questionnaire administered by a physicians in well-equipped operating theatres with trained interviewer on sociodemographic characteristics, careful attention to asepsis. All instruments, drapes, and sexual risk behaviours, genital hygiene, and health. other materials were autoclaved and sterility was assured Participants were asked to provide a urine sample for by use of thermologues (Comply, 3M Healthcare, St Paul, future testing of sexually transmitted infections. Two MN, USA) and biological indicators (BT Sure, Barnsead/subpreputial and shaft swabs were also obtained for Thermolyne, Dubuque, IA, USA). Participants showered future testing for human papillomavirus infection and preoperatively to clean the genital area. The skin was other sexually transmitted infections.
prepared with povidone-iodine before administration of local anesthesia via a dorsal penile nerve block with a Procedures
mixture of lidocaine and bupivacaine. Circumcision was Participants were randomly assigned to the intervention done with the sleeve procedure, in which the foreskin or control groups as follows. Treatment assignment was was retracted and a distal incision made 0·5–1·0 cm randomly generated in blocks of 20, stratifi ed on proximal to the coronal sulcus, followed by a proximal community, with each community receiving four blocks incision on the unretracted prepuce at the corona. The of 20 assignment envelopes. Because enrolment occurred superfi cial lamina of Bucks fascia was exposed and a concurrently at more than one community site, this sleeve of foreskin was freed from the underlying Bucks procedure ensured balance within sites. 20 assignments fascia and removed.7 Bleeding was controlled with bipolar in opaque envelopes were placed in batches, and electrocautery and skin edges apposed with 4-0 absorbable participants were asked to select one envelope from the sutures. Men were kept under observation for box. After an assignment envelope was selected, it was 30–60 minutes before discharge. Men who lived close to replaced by the next envelope from the next batch the surgical facility returned home, whereas those men designated for that community. This procedure could who lived distant from the facility were off ered free and did result in some temporary imbalance between overnight accommodation in a study facility to ensure study groups, with a maximum potential run of 20 instead access to care should short-term complications arise.
of the standard ten same-group assignments, but it Postoperative follow-up visits were scheduled at ensured that all participants had the opportunity to select 24–48 hours, 5–9 days, and 4–6 weeks. The fi rst visit was one of 20 envelopes. An alternative procedure was done at the surgical clinic site; subsequent visits occurred considered in which participants would select from each in mobile clinics in the communities. Care was available block of 20 envelopes without replacement, which would for participants at any time between scheduled visits. ensure that every 20 assignments within a site was Follow-up was done by clinical offi perfectly balanced. However, this method was rejected by the urologist to diagnose and treat complications or to because it would progressively reduce a participant’s refer patients as needed. Potential adverse events related to options for envelope selection.
surgery were predefi ned and graded as mild (requiring no HIV status at screening was assessed by two enzyme treatment), moderate (requiring treatment), or severe immunoassays: Vironostika HIV-1 (Organon Teknika, complications (requiring surgical intervention [eg, wound Charlotte, NC, USA) and Welcozyme HIV 1+2 (Murex exploration for active bleeding, repair of wound dehiscence], Diagnostics, Dartford, UK). Men with concordant negative hospitalisation, or referral for specialised care). At each results were enrolled into the trial. Discordant results postoperative follow-up visit, participants were questioned were confi rmed by western blot (Cambridge Biotech about symptoms suggestive of complications, and the HIV-1 western blot, Caltype Biomedical Corp, Rockville, wound was inspected. Participants were asked about MD, USA); men who were negative by western blot were resumption of sexual intercourse, and those who had resumed such activity were asked about condom use.
Men randomly assigned to the intervention group were All participants in both groups were followed up at asked to provide written consent for surgery on the day of 4–6 weeks, and at 6, 12, and 24 months post-enrolment. the procedure, and were again provided with detailed At each follow-up visit, participants answered questions information on the procedure, postoperative wound care, on sexual risk behaviours (marital and non-marital www.thelancet.com Vol 369 February 24, 2007
Articles
partners, condom use, alcohol consumption with sexual intercourse, and transactional sexual intercourse [ie, sexual intercourse in exchange for money or gifts]) and symptoms of sexually transmitted diseases (genital ulcer disease, urethral discharge, or dysuria) since their previous visit. Men were questioned about illnesses or hospitalisations to record all adverse events that occurred during trial participation. Additionally, men were examined to assess circumcision status and to diagnose any penile pathology. Samples of venous blood and urine and two penile swabs were collected, and repeat HIV counselling and testing and health education were provided. Free condoms were off ered to all sexually active participants at all study visits, and were also available through community-based condom depots stocked by the Rakai programme.
The procedure for HIV testing at each follow-up visit was the same as at enrolment. All seroconversions or discordant enzyme immunoassay results were further assessed by western blot. For participants who had under- gone seroconversion during follow-up, the previous serologically negative sample and in selected cases the fi rst positive sample were tested by reverse transcriptase (RT) PCR (Amplicor HIV-1 Monitor version 1.5, Roche Molecular Systems, Branchburg, NJ, USA).
The Rakai Health Sciences Program has an HIV treatment programme that is funded by the Presidential Emergency Fund for AIDS Relief. Participants found to be HIV positive at trial screening and those who subsequently became infected with HIV during the trial were referred to the HIV treatment programme. All individuals enrolled into the HIV treatment programme were provided
with prophylaxis with sulfamethoxazole-trimethoprim, Figure 1: Trial profi le
insecticide-impregnated bednets, and water purifi cation. Those who were eligible for antiretroviral therapy (CD4 cell trial. Participants were compensated for their time, travel count less than 250 cells per µL or WHO advanced stage costs, and absence from work. Men received US$5 at III or stage IV disease) and who agreed to receive care screening and enrolment, $5 at the time of surgery, and were provided with antiretrovirals. None of the HIV- $5 on completion of postoperative follow-up. Control infected participants from the trial were eligible for participants who were circumcised at completion of their antiretroviral therapy at the time of going to press.
24 months of follow-up received identical compensation. The protocol was reviewed and approved by the The amount of compensation for routine follow-up visits Prevention Sciences Research Committee of the Division at 6, 12, and 24 months was $3 per visit. The community of AIDS, National Institute of Allergy and Infectious advisory board and institutional review boards approved Diseases (NIAID), in the US National Institutes of Health (NIH), and by the Rakai community advisory board. The
study was approved by three institutional review boards: Statistical analysis
the Science and Ethics Committee of the Uganda Virus For incidence rate and Poisson regression calculations,
Research Institute (Entebbe, Uganda), the Committee HIV seroconversion was estimated assuming that
for Human Research at Johns Hopkins University,
Bloomberg School of Public Health (Baltimore, MD,
Intervention group
Control group
USA), and the Western Institutional Review Board (Olympia, WA, USA). The trial was done in accordance with the Good Clinical Practices and International Clinical Harmonisation guidelines with clinical trial monitoring done by Westat Corporation under a Division Data are n/N (%). Percentages have been rounded.
of AIDS, NIAID, NIH contract. The NIH Vaccine and Table 1: Trial retention rates
Prevention Data Safety Monitoring Board oversaw the www.thelancet.com Vol 369 February 24, 2007
Articles
infection occurred at the mid-time point between the last Exploratory analyses assessed the comparability of the negative and fi rst positive serological tests, or at the time two study groups at enrolment. HIV incidence during the of the fi rst positive RT-PCR for those participants seen trial was assessed by fi xed covariates such as age, marital during the period before HIV antibody seroconversion. status, and education at enrolment, and by time-varying For participants who were positive by PCR but who were covariates such as sexual risk behaviours (eg, number of negative for HIV antibody, the date of the positive PCR partners, non-marital relationships, condom use, and was used as the date of infection. In both groups, time alcohol use), and symptoms of sexually transmitted from enrolment was accumulated up to the 24 month diseases reported at follow-up visits. Men who were follow-up visit and HIV incidence was estimated originally allocated to circumcision but who did not present per 100 person-years. for surgery within 6 months of enrolment were assessed as crossovers, as were individuals in the control group who opted to have circumcisions done outside the study.
Intervention group Control group
We used a modifi ed intention-to-treat approach for the (n=2474)
(n=2522)
cacy analysis, which included all participants who were serologically or PCR negative at enrolment. Three participants who were PCR-positive but antibody negative at enrolment were deemed to have been infected before randomisation and were excluded from this modifi ed intention-to-treat analysis. The primary modifi ed intention-to-treat population included crossovers and participants who reported periods of sexual abstinence during the 24 months of follow-up. Incidence rate ratios (IRR) and 95% CI of HIV acquisition in the intervention versus the control group were estimated via exact methods, with Poisson multiple regression used for the adjusted analyses, including trend assessments. Because the trial was ended early, the Poisson analysis for the 0–24 month interval is weighted by the preponderance of person-time accrued during the fi rst 12 months, and thus is a conservative estimate. Primary analyses adjusted for postulated potential confounders identifi ed in previous studies in Rakai8 and included baseline values of age, marital status, and sexual risk behaviours. Time varying Number of sexual partners in the past year covariates (eg, self-reported genital ulcer disease) could be in the causal pathway, so were not adjusted for during follow-up. We did an as-treated analysis that included control crossover participants who had received circumcision from outside sources, with person-time in the circumcised state ascribed to the beginning of the follow-up interval in which the surgery occurred. For crossovers in the intervention group who did not receive surgery, person-time was ascribed to the uncircumcised state from time of enrolment. Poisson multiple regression models were fi t for the whole population and for strata of particular interest (eg, self-reported genital ulcer disease).
We did a Kaplan-Meier estimation based on analyses of time-to-detection of HIV infection at the visit at which positive serology or PCR was fi rst identifi ed. Due to the Prior receipt of voluntary counselling and testing discrete nature of the timing of follow-up, data from Self-reported symptoms of sexually transmitted diseases in past year visits were ascribed to the time of scheduled follow-up visits. An overall risk diff erence and risk ratios were calculated at the end of follow-up, with CI based on standard Greenwood formula variance estimates. The Kaplan-Meier risk ratios are not aff ected by the early trial Data are n (%). Percentages have been rounded. *Sexual intercourse for money or gifts. closure, and this method was used in both other trials of Table 2: Enrolment characteristics, risk behaviours, and symptoms of sexually transmitted diseases by
male circumcision. Therefore, we present Kaplan-Meier study group
risk ratios for comparative purposes.
www.thelancet.com Vol 369 February 24, 2007
Articles
To assess possible behavioural disinhibition, risk Intervention
Incidence rate
behaviours were tabulated by follow-up visit, and ratio (95% CI)
diff erences between study groups were assessed by χ² 0–6 months follow-up interval
and Fisher exact tests. Symptoms of sexually transmitted diseases reported at each visit were cumulated over the 24 months of follow-up to estimate the prevalence of symptoms per 100 visits in intervention and control participants. Prevalence risk ratios (PRR) were estimated with log-binomial regression with a robust variance 6–12 months follow-up interval
adjustment to account for within-person correlation. We also examined possible associations between reported symptoms of sexually transmitted diseases and incident HIV infection, by use of subgroup-specifi c models to determine whether any eff ects of circumcision on HIV 12–24 months follow-up interval
incidence might be mediated by symptomatic sexually The frequencies of adverse events both related and unrelated to study participation were assessed in both study groups. Multiple adverse events diagnosed at a Total 0–24 months follow-up
single visit were counted as separate events despite the fact that they could have been causally related (eg, wound dehiscence and infection), to provide an estimate of the maximum frequency of adverse events without making Cumulative incidence per 100 person-years Table 3: HIV incidence by study group and follow-up interval, and cumulative HIV incidence over 2 years
The study had 80% power to detect a rate ratio of 0·5 for incident HIV in the intervention group relative to the control group, with a projected total person-time of 8993 person-years, assuming a 15% annual loss to follow-up and 10% crossover over 24 months. Formal statistical monitoring used the Lan-DeMets group sequential approach9 with an O’Brien-Fleming type α spending function10 to minimise the chance of in- appropriate premature trial termination. Two interim analyses were done, the fi rst with a data cutoff date of April 30, 2006, when about 43% of projected person-time had been accrued, and the second interim analysis with a
data cutoff date of Oct 31, 2006, when about 72% of Cases of HIV/total participants
projected person-time had been accrued. The second Intervention 0/2474
interim analysis showed a signifi cant diff erence in HIV inci dence between the two study groups Figure 2: Kaplan-Meier cumulative probabilities of HIV detection by study
(nominal α=0·0215); as a result, NIAID terminated the group
Actual visits grouped by the three scheduled visits at 6 months, 12 months, and cacy on Dec 12, 2006. The analyses presented 24 months after enrolment. The cumulative probabilities of HIV infection were here are based on all data accrued up to the time of trial 1·1% in the intervention group and 2·6% in the control group over 24 months.
closure in December, 2006, and encompass about 73% of total anticipated person-time. Results were deemed to be Johns Hopkins University and Columbia University. statistically signifi cant at the α=0·05 level. All data were FM, LHM, and MAC had full access to all the data until double entered. East was used for spending function the trial closed. Thereafter, the principal investigator calculations and Stata version 8 was used for analysis.
and co-investigators (RHG, GK, DS, MJW, FN, NKS, This trial is registered with ClinicalTrials.gov, with the FWM, AND SJR) had access to all the data. Staff at the Division of AIDS maintained oversight of progress and reporting, and participated in study conduct and data Role of the funding source
interpretation as members of the study executive This trial was funded through a cooperative agreement committee. Data analyses was done by the research with the Division of AIDS, NIAID/NIH. The study was teams at John Hopkins University and the Rakai Health done by the Rakai Health Sciences Program, a research Sciences Program. The corresponding author had fi nal collaboration between the Uganda Virus Research responsibility for preparing and submitting results for Institute, and researchers at Makerere University and publication.
www.thelancet.com Vol 369 February 24, 2007
Articles
enrolment record was retained in the dataset for the Figure 1 shows the trial profi le. 5000 eligible men were primary intent-to-treat analysis and the second enrolment initially enrolled. However, during follow-up we dis covered was deleted, leaving 4996 enrolled participants. that four men (two in each study group) had re-enrolled 146 (6%) participants in the intervention group did not under assumed names. For these individuals, the fi rst come for surgery within 6 months of randomisation and Intervention group
Control group
Incidence rate ratio
(95% CI)
Characteristics at enrolment
Behaviour and symptoms of sexually transmitted infections during follow-up
*Among those sexually active in the follow-up interval.
Table 4: Cumulative HIV incidence over 24 months by sociodemographic characteristics at enrolment, and behavioural characteristics and symptoms of
sexually transmitted infections during follow-up
www.thelancet.com Vol 369 February 24, 2007
Articles
Intervention group
Control group
Prevalence risk ratio (95% CI)*
*Based on robust variance estimates adjusting for multiple observations on the same individuals Table 5: Prevalence of self-reported symptoms of sexually transmitted infections per visit, cumulatively over 24 months follow-up
were classifi ed as crossovers. Among the controls, 33 men between the cumulative probabilities of HIV detection were circumcised from other sources, a crossover rate was signifi of 1·3%. There were 15 deaths among participants in the 0·43 (0·24–0·75). The as-treated Poisson analysis, which intervention group over 3352·4 person-years and 17 deaths assigned person-time according to the actual circumcision in the control group over 3391·8 person-years (4·5 deaths status of participants, showed an incidence of 0·61 cases per 1000 person-years vs 5·0 deaths per 1000 person-years, per 100 person-years in the intervention group (20 events p=0·8). None of the deaths were related to trial in 3268·1 person-years), and 1·35 cases per 100 person-participation.
years in the control group (47 events in 3481·6 person-years) Trial retention rates are shown in table 1. All 1 year with an IRR of 0·45 (95% CI 0·25–0·78; p=0·0022). The follow-up visits had been completed at time of trial as-treated Kaplan-Meier risk ratio was 0·40 (0·23–0·70,
termination, and retention rates at 12 months were p=0·003).
equivalent in both groups. By December 12, 2006, the
Table 4 shows cumulative HIV incidence over 24 months date of trial termination, 44% of men in both groups had by sociodemographic characteristics at enrolment, and by reached their 24 month follow-up time point; retention self-reported sexual risk behaviours and symptoms of rates for these men were much the same in both groups.
sexually transmitted infections during follow-up. The rates The baseline characteristics of the enrolled participants of HIV acquisition were lower among circumcised men in are shown in table 2. The two arms were much the same all strata of characteristics, risk behaviours and symptoms in terms of sociodemographic characteristics (age, of sexually transmitted infections examined, with the marital status, religion, and education) and in sexual risk exception of those men who reported no sexual activity behaviours (number or partners, condom use, alcohol within the follow-up interval of seroconversion. HIV consumption with sex, and sex for money or gifts). At incidence was highest in the 25–29 year age-group, but in enrolment, previous receipt of voluntary counselling and all age-groups, incidence was lower in the intervention testing was slightly higher in the intervention group than than in the control group. Similarly, HIV incidence was in the control group. The two groups reported similar lower in circumcised than in uncircumcised men in all rates of symptoms of sexually transmitted infections.
categories of marital status and education. Among sexually Table 3 shows HIV incidence by study arm and follow-up active men, circumcision reduced HIV acquisition visit intervals, together with cumulative incidence over irrespective of the number of partners, non-marital 2 years. The intention-to-treat analysis showed a relationships, condom use, consumption of alcohol before progressive decrease in incidence in the intervention sexual intercourse, and transactional sexual intercourse. group over the entire follow-up period (p for trend 0·014). Men reporting symptoms of sexually transmitted diseases Incidence fell in the control group between the time of during a follow-up interval had higher rates of HIV fi rst follow-up and the time of second follow-up, and acquisition than did asymptomatic participants, but the remained stable thereafter; however, the trend was not protective eff ects of circumcision were observed irrespective signifi cant (p=0·6). The IRR of HIV acquisition associated of the presence of such symptoms. However, circumcision with circumcision also fell over time; this increase in was not protective against HIV acquisition in the few men effi cacy was of borderline signifi cance (p=0·054 for the who reported no sexual activity in a given follow-up time-by-study arm interaction). The 24 month cumulative interval. There were six incident cases (three in each group) HIV incidence was 0·66 cases per 100 person-years in the during periods of reported abstinence. None of these six intervention group, compared with 1·33 cases participants reported receipt of injections or transfusions per 100 person-years in the control group. The unadjusted during the follow-up interval of HIV seroconversion; these IRR was 0·49 (95% CI 0·28–0·84; p=0·0057). After participants probably under-reported their sexual activity.
adjustment for age, marital status, and sexual risk The prevalence rates of self-reported symptoms of behaviours at enrolment, the IRR was 0·49 (0·29–0·81; sexually transmitted diseases reported at each follow-up p=0·003). Figure 2 shows the Kaplan-Meier survival visit, cumulated over 24 months, are shown in table 5. curves for time-to-detection of HIV infection for the Over all study visits, the prevalence of self-reported genital modifi ed intention-to-treat analysis. The diff erence ulcers during the preceding interval was lower in the www.thelancet.com Vol 369 February 24, 2007
Articles
it was in the control group (table 6; p=0·11). Similarly, Intervention group
Control group
inconsistent condom use was higher in the intervention 6 months follow-up (reference period 6 months since enrolment)
group than it was in the control group (table 6; p=0·0004). At the 12 and 24 months follow-up visits, the number of sexual partners, non-marital relationships, and condom use were much the same in the two groups. However, participants in the control group reported slightly higher rates of alcohol use with sexual intercourse in all follow-up intervals than did those in the intervention group; this was signifi cant at the 6 month (p=0·001) and 24 month (p=0·02) visits (table 6). Transactional sexual intercourse was infrequent and did not diff er between study groups. There is, therefore, no consistent or substantial evidence of behavioural disinhibition after circumcision in the 12 months follow-up (reference period 6 months)
Adverse events unrelated to trial participation were frequent. 1391 adverse events were reported in the intervention group, compared with 1320 in the control group (56% vs 52%; p=0·083). Of these adverse events, 1213 (87%) in the intervention group were unrelated to the trial; all adverse events in the control group were unrelated to the trial. Almost half of the unrelated adverse events were mild grade 1 events (46% [n=558] of those in the intervention group and 50% [n=660] of those in the control group). The rate of all adverse events related to surgery in the intervention group was about 8% (178 events in 2328 surgeries); most of these events were mild (94 of 178 events). The rate of moderate adverse events related to surgery was about 3% (79 events in 2328 surgeries), and 24 months follow up (reference period 12 months)
there were fi ve severe adverse events, with a rate of 0·2 events per 100 surgeries. The severe adverse events included one wound infection, two haematomas that required re-exploration and ligation of active bleeding vessels, one wound disruption due to external cause, and one case of severe postoperative herpetic ulceration not involving the surgical wound requiring hospitalisation in the programme’s facility. All moderate and severe adverse events were successfully managed and resolved.
Discussion
This large, randomised trial of adult male circumcision in a rural Ugandan population shows that such a surgical intervention reduces the risk of the acquisition of HIV in Date are n (%). *Among those who reported sexual activity in the follow-up interval.
men. We noted a signifi cant reduction in HIV incidence Table 6: Sexual risk behaviours by study group and follow-up visit
among circumcised men compared with uncircumcised control participants. The effi intervention group than in the control group (3·1% prevention of incident HIV was 51% in the Poisson vs 5·8%; PRR 0·53, 95% CI 0·43–0·64; p<0·0001). intention-to-treat analysis; adjustment for enrolment However, circumcision had little eff ect on the prevalence characteristics, behaviours, and symptoms of sexually of urethral discharge or dysuria.
transmitted infections did not aff ect this estimate. In the To assess possible behavioural disinhibition, sexual risk behaviours were assessed at each follow-up visit (table 6). During the fi rst 6 month follow-up interval, sexual activity fi ndings are compatible with observational data,1–5 as well was reported by 1801 (79%) participants in the intervention as data from a randomised trial in South Africa group, compared with 1787 (77%) of those in the control (60% intention-to-treat effi group (p=0·049). Consistent condom use during this effi cacy in a semi-urban population aged 18–24 years),6 interval was slightly higher in the intervention group than and a trial in Kenya (53% intention-to-treat effi www.thelancet.com Vol 369 February 24, 2007
Articles
biologically plausible since it suggests that circumcision 18–24 years),11 suggesting similar effi cacy in widely could be protective against cutaneously acquired infections divergent populations. Thus, circumcision must now be harboured in the moist subpreputial space, but the deemed to be a proven intervention for reducing the risk procedure does not seem to be protective against urethral of heterosexually acquired HIV infection in adult men.
infections, which presumably are unaff ected by the HIV incidence in the intervention group fell signifi cantly removal of the foreskin.
over time, whereas it remained fairly constant in the That circumcision reduces the risk of male HIV infection cacy of circumcision is biologically plausible. The foreskin is rich in HIV target increased progressively during later follow-up intervals cells (Langerhans’ and dendritic cells, CD4+ T cells, and (eg, 75% effi cacy during the 12–24 month follow-up macrophages),18–21 and the inner preputial mucosa is interval, table 3). The Kaplan-Meier curves for time to unkeratinised, making it vulnerable to HIV infection.20,22 detection of HIV infection did not diverge until the The foreskin is retracted over the shaft during intercourse, twelfth month of follow-up, meaning that the diff erence which exposes the inner mucosa to vaginal and cervical in HIV acquisition began during the 6–12 month fl uids.22 Also, breaches in the mucosa can occur due to follow-up interval (fi gure 2). The HIV incidence in the microtears during intercourse, especially at the frenulum,22 control group (1·3 cases per 100 person-years), is identical and uncircumcised men are more susceptible to genital to that seen in uncircumcised men in the Rakai population ulcer disease, which could increase HIV entry.13,22 at the time the trial was done.12 Also, 45% of HIV-negative The 24 month transmission risks were 2·6% in the uncircumcised men in the Rakai cohort volunteered to control group and 1·11% in the intervention group, giving enroll in the trial, which suggests that the trial results are a risk diff erence of 1·49%. Thus, assuming completion of probably generalisable to the Rakai population as a whole. 24 months of follow-up, we estimate that about At the time of trial closure, 80% of eligible control 67 circumcisions are needed to prevent one HIV infection participants who had completed 24 months follow-up in the 2-year postoperative interval. However, this estimate agreed to be circumcised, suggesting high acceptability.
does not include possible reductions in secondary We did not fi nd evidence that men in the intervention transmissions to women or the probable long-term group adopted higher sexual risk behaviours than did eff ectiveness of circumcision in men. Mathematical those in the control group (table 6). This could have been models have been used to estimate the number of due to the intensive health education provided during surgeries required per HIV infection averted in both men the trial to minimise risk compensation. These fi ndings and women over varying periods of time. In Rakai, a diff er from those from the South African trial, which stochastic simulation model suggested that, with a reported an increase in the mean number of sexual circumcision effi contacts in men in the intervention group.6 Future 1·3 per 100 person-years in uncircumcised men, the circumcision programmes must emphasise that circum- number of surgeries per HIV infection averted over cision provides only part protection, and that there is a 10 years was about 35, assuming all uncircumcised men critical need to practise safer sex after circumcision (eg, accept the procedure.12 In South Africa, with a circumcision partner limitation and consistent condom use).
cacy of 60% and HIV incidence among uncircumcised Circumcision also reduced the rate of self-reported men of 3·8 per 100 person-years, the number of surgeries symptoms of genital ulcer disease with a cumulative per infection averted over 20 years is much lower.23 Thus, effi cacy of 48% over all follow-up visits (table 5), which is the number of surgeries needed to prevent one HIV comparable with the protective eff ects of circumcision on infection will vary depending on background HIV genital ulcer disease in observational studies.13 At this time, incidence, the level of acceptance, and the duration of we cannot determine whether the procedure reduced the projected protection. Policymakers will have to determine incidence of ulcerative infections due to syphilis, herpes whether adult male circumcision is likely to be an simplex virus 2, and Haemophilus ducreyi, or whether appropriate and cost-eff ective intervention in specifi c removal of the prepuce reduced the severity, duration, or settings. In the longer term, neonatal circumcision or recurrence of ulceration, leading to lower recognition of circumcision of younger boys will provide a simpler, safer, symptoms. Since genital ulcer disease is a risk factor for and cheaper option, although the HIV benefi ts will be the acquisition of HIV,14–16 and symptomatic genital ulcer delayed until these boys reach sexual maturity.
disease was associated with higher rates of HIV acquisition Adult male circumcision is not without risk. In this trial in this trial (table 4), it is plausible that the protective eff ect the rate of moderate and severe adverse events related to of circumcision on HIV could be mediated in part by the surgery was almost 4%, which is comparable with rates in protective eff ects of the procedure on self-reported genital the South African and Kenyan trials.6,9 One should note ulcer disease. By contrast, there was no eff ect of that there were cases in which appropriate follow-up circumcision on symptoms of discharge or dysuria management was required to prevent more serious (table 5), which is consistent with data from observational sequelae. Furthermore, substantially higher complication studies that indicate a lack of an eff ect of circumcision on rates have been reported when surgery is done in rural gonorrhoea or chlamydia prevalence.3,17 The fi nding is clinics or by traditional circumcisers.24 The scale-up of www.thelancet.com Vol 369 February 24, 2007
Articles
circumcision services will require careful attention to 4 Lavreys L, Rakwar JP, Thompson ML, et al. Eff ect of circumcision training of personnel, provision of facilities, equipment on incidence of human immunodefi ciency virus type 1 and other sexually transmitted diseases: a prospective cohort study of trucking and supplies, postoperative care to minimise and manage company employees in Kenya. J Infect Dis 1999; 180: 330–36.
complications, and monitoring of the quality of services 5 Baeten JM, Richardson BA, Lavreys L, et al. Female-to-male infectivity of HIV-1 among circumcised and uncircumcised Kenyan
men. J Infect Dis 2005; 191: 546–53.
The use of surgery for disease prevention is an unusual Auvert B, Taljaard D, Lagarde E, Sobngwi-Tambekou J, Sitta R, public-health intervention. One precedent is the mass Puren A. Randomized, controlled intervention trial of male sterilisation camps in India during the 1970s, which were circumcision for reduction of HIV infection risk: the ANRS 1265
Trial. PLoS Med 2005; 2: e298.
poorly implemented and resulted in serious surgical 7 Walsh PC, Retik AB, Stamey TA, Vaughan ED, eds. Campbell’s complications, deaths, and ultimately the collapse of the urology, 6th edn. WB Saunders Co, 1992: 2972–73. programmes.25,26 Thus, future provision of circumcision 8 Zablotska IB, Gray RH, Serwadda D, et al. Alcohol use before sex for HIV prevention must maintain the highest achievable and HIV acquisition: a longitudinal study in Rakai, Uganda. AIDS
2006; 20: 1–6.
levels of safety to be acceptable and sustainable.
Lan KK, DeMets DL. Discrete sequential boundaries for clinical The consistency of epidemiological evidence from three trials. Biometrika 1983; 70: 659–63.
randomised trials and multiple observational studies 10 O’Brien PC, Fleming TR. A multiple testing procedure for clinical presents a compelling case for the promotion of male trials. Biometrics 1979; 35: 549–56.
11 Bailey RC, Moses S, Parker CB, et al. Male circumcision for HIV circumcision for HIV prevention in populations where prevention in young men in Kisumu, Kenya: a randomised circumcision is infrequently practiced and where HIV controlled trial. Lancet 2007; 369: 643–56.
transmission is mainly due to heterosexual intercourse. 12 Gray RH, Li X, Kigozi G, et al. The impact of male circumcision on HIV incidence, and cost-per infection prevented: a stochastic Such practice is especially relevant in east and southern simulation model from Rakai, Uganda. AIDS (in press).
Africa, where circumcision rates are low in many 13 Weiss HA, Thomas SL, Munabi SK, Hayes RJ. Male circumcision populations and the HIV epidemic is most severe. and risk of syphilis, chancroid, and genital herpes: a systematic
review and meta-analysis. Sex Transm Infect 2006; 82: 101–09.
However, trials that are stopped early could overestimate 14 Freeman EE, Weiss HA, Glynn JR, Cross PL, Whitworth JA, cacy when compared with subsequent studies27 and to Hayes RJ. Herpes simplex virus 2 infection increases HIV undertake long-term post-circumcision trial surveillance is acquisition in men and women: systematic review and essential to determine the eff ectiveness of circumcision in meta-analysis of longitudinal studies. AIDS 2006; 20: 73–83
15 Reynolds SJ, Risbud AR, Shepherd ME, et al. Recent herpes populations with varying HIV prevalence, and to assess simplex virus type 2 infection and the risk of human the durability of any observed benefi ts. Furthermore, to immunodefi ciency virus type 1 acquisition in India. J Infect Dis assess whether perceptions of circumcision effi 2003; 187: 1513–21.
16 Rottingen JA, Cameron DW, Garnett GP. A systematic review of the an exaggerated belief in the protective eff ects of the epidemiologic interactions between classic sexually transmitted procedure, thus engendering increases in HIV risk diseases and HIV. Sex Transm Dis 2001; 28: 579–97.
17 Gray RH, Azire J, Serwadda D, et al. Male circumcision and the risk of sexually transmitted infections and HIV in Rakai, Uganda. AIDS Contributors
2004; 18: 2428–30.
All authors took part in the design, implementation, and analysis of this 18 Hussain LA, Lehner T. Comparative investigation of Langerhans’ study and saw and approved the fi nal version.
cells and potential receptors for HIV in oral, genitourinary and
rectal epithelia. Immunology 1995; 85: 475–84.
Confl ict of interest statement
19 Donoval BA, Landay AL, Moses S, et al. HIV-1 target cells in We declare that we have no confl ict of interest.
foreskins of African men with varying histories of sexually Acknowledgments
transmitted infections. Am J Clin Pathol 2006; 125: 386–91.
The study was supported by a grant (UO1 AI11171-01-02) from the 20 McCoombe SG, Short RV. Potential HIV-1 target cells in the human National Institutes of Allergy and Infectious Disease (NIAID), Division penis. AIDS 2006; 20: 1491–95.
of AIDS, National Institutes of Health (NIH), and in part by the Division 21 Patterson BK, Landay A, Siegel JN, et al. Susceptibility to human of Intramural Research, NIAID, NIH. This publication was supported, immunodefi ciency virus-1 infection of human foreskin and cervical in part, by a fellowship/grant from the Fogarty International tissue grown in explant culture. Am J Pathol 2002; 161: 867–73.
Center/USNIH: grant number 2 D 43 TW000010-19-AITRP. We thank 22 Szabo R, Short RV. How does male circumcision protect against the members of the NIH data safety monitoring board who monitored HIV infection? BMJ 2000; 320: 1592–94.
this trial, as well as the institutional review boards that provided 23 Kahn JG, Marseille E, Auvert B. Cost-eff ectiveness of male oversight (the scientifi c and ethics committee of the Uganda Virus circumcision for HIV prevention in a South African setting. Research Institute, the committee for human research at Johns Hopkins, PLoS Med 2006; 3: e517.
and Western Institutional Review Board). We are also grateful for the 24 Bailey RC, Egesah O. Assessment of clinical and traditional male advice provided by the Rakai community advisory board. Finally, we wish circumcision services in Bungoma district, Kenya. Complication rates and operational needs. Special report. Washington, DC: to express our gratitude to study participants whose commitment and cooperation made the study possible.
25 Kabra SG, Narayanan R. Sterilisation camps in India. Lancet 1990; References
335: 224–25.
Weiss HA, Quigley MA, Hayes RJ. Male circumcision and risk of 26 Kumar S. Health-care camps for the poor provide mass sterilisation HIV infection in sub-Saharan Africa: a systematic review and quota. Lancet 1999; 353: 1251.
meta-analysis. AIDS 2000; 14: 2361–70.
27 Montori VM, Devereaux PJ, Adhikari NK, et al. Randomized trials Gray RH, Kiwanuka N, Quinn TC, et al. Male circumcision and stopped early for benefi t: a systematic review. JAMA 2005; 294:
HIV acquisition and transmission: cohort studies in Rakai, Uganda. Rakai Project Team. AIDS 2000; 14: 2371–81.
Reynolds SJ, Shepherd ME, Risbud AR, et al. Male circumcision
and risk of HIV-1 and other sexually transmitted infections in India.
Lancet 2004; 363: 1039–40.
www.thelancet.com Vol 369 February 24, 2007

Source: http://www.shangring.cn/downloadcenter/Reprint_RHGray_20070224.pdf