Randomized, Controlled Intervention Trial of Male Circumcision for Reduction of HIV Infection RiskThe ANRS 1265 Trial

If decades of observational data suggest that circumcised men get HIV at lower rates, and if that pattern holds across cultures and continents, then the logical next move is to design a trial rigorous enough to prove it or kill the idea for good. That chain of reasoning landed a research team outside Johannesburg with three thousand men, sealed randomization envelopes, and one of the most consequential questions in global HIV prevention. The observational signal had been building since 1986. A systematic review of sub-Saharan African studies found an adjusted relative risk of zero point four two for HIV among circumcised men — roughly half the infection rate of uncircumcised men — across dozens of datasets. But the signal came tangled with confounders. Male circumcision is woven into culture and religion: nearly universal in Muslim-majority regions of West Africa, practiced by almost every Kenyan tribe except the Luo, shaped by colonial history and tribal rites of passage elsewhere. Those same cultural differences correlate with sexual behavior, health-seeking habits, and a dozen other factors that drive HIV transmission. Correlation wasn't causation, and policymakers knew it. Without experimental data, no one could say whether the foreskin itself was the issue or whether circumcised men simply lived differently. So Bertran Auvert and colleagues designed the ANRS 1265 trial — a randomized, controlled intervention study in Orange Farm, a semi-urban area near Johannesburg chosen for its high HIV burden and mixed circumcision practices. HIV prevalence among pregnant women in Gauteng province stood at twenty-nine point six percent in 2003. The setting made the question both urgent and answerable. Between July 2002 and February 2004, the team recruited three thousand two hundred seventy-four uncircumcised men aged eighteen to twenty-four. After a pre-screening visit, at least three days to consider participation, and written consent, men were randomized using sealed envelopes drawn from a basket — five labeled intervention and five control per set of ten. Those randomized to intervention were offered medical circumcision within a week; the median wait was one day. Those randomized to control were asked to wait until the end of the trial, at which point they would be offered the procedure. Follow-up visits were scheduled at months three, twelve, and twenty-one. At each visit: a face-to-face questionnaire, a blood draw for HIV and syphilis testing, a genital examination, and a counseling session. Sexual behavior data were collected in detail — number of partners, contact frequency, condom use, and partnership type. Crucially, the laboratory database was kept separate from the clinical database; investigators at the site couldn't see HIV results during the study. The analysis used an intention-to-treat framework, meaning everyone was analyzed in their assigned group regardless of whether they actually received circumcision. This preserves the benefits of randomization. Because infections were assessed at discrete visits with unequal intervals, the team modeled rates using a piecewise exponential proportional hazards approach — in plain language, they assumed the baseline infection rate was constant within each of three time windows but could change between them, and they ran the whole thing as a Poisson log-linear model. The result is an incidence rate ratio comparing the two groups, and protection is simply one minus that ratio. An independent Data and Safety Monitoring Board reviewed adverse events and conducted a prespecified interim analysis after all month-twelve visits were complete. What they found stopped the trial. Among three thousand one hundred twenty-eight HIV-negative men with a mean follow-up of eighteen point one months, there were twenty HIV infections in the circumcision group and forty-nine in the control group. That translates to incidence rates of zero point eight five per one hundred person-years in the intervention arm versus two point one in the control arm. The rate ratio was zero point four zero, with a ninety-five percent confidence interval of zero point twenty-four to zero point sixty-eight and a p-value of zero point zero zero zero fifty-nine. Protection: sixty percent, with a confidence interval running from thirty-two to seventy-six percent. To put the rate ratio in plain speech — for every infection in the circumcised group, there were roughly two and a half in the control group. The interim result cleared the stopping threshold, and the board advised ending the trial and offering circumcision to the control group. The most important challenge to address immediately is behavioral confounding. Sexual activity increased slightly in the intervention group — a plausible disinhibition effect, men perhaps feeling protected. The team measured this and built it into multivariate models along with condom use and health-seeking behavior. After those adjustments, estimated protection held at sixty-one percent, with a confidence interval of thirty-four to seventy-seven percent. Essentially unchanged. Two sensitivity analyses reinforce the point: restricting to participants who completed the month-twenty-one visit gave a rate ratio of zero point three eight, and removing a forty-two-day window after surgery — to test whether the required abstinence period was artificially inflating the result — gave a rate ratio of zero point four three. The effect is not an artifact of behavior or methodology. It is real and it is large. On safety, the picture is reassuring but not spotless. Among one thousand five hundred sixty-eight circumcisions performed in the intervention arm, sixty adverse events were recorded in the first month — a rate of three point eight percent. These included pain, excessive bleeding, swelling, dissatisfaction with appearance, infection, and a small number of cases involving damage to the penis. No deaths related to circumcision were reported. Men who were HIV-positive at randomization had a higher adverse event rate — eight point two percent versus three point six percent in HIV-negative men. At the three-month visit, ninety-eight point five percent of circumcised participants reported being very satisfied with the result. At month twenty-one, only about one percent reported ongoing problems, mostly minor urinary issues or aesthetic concerns. Why would removing the foreskin reduce HIV transmission? Auvert and colleagues point to several converging factors. The foreskin's inner surface is a thin, moist mucosa with a high density of Langerhans cells and other CD4-positive target cells — prime HIV entry points. Removing it eliminates that tissue. After circumcision, the glans undergoes keratinization, essentially thickening its outer layer. The total penile surface exposed during intercourse decreases. And faster post-coital drying may reduce the window during which the virus can survive on the skin. The paper is careful to say the trial doesn't isolate a single mechanism — but multiple biological pathways point in the same direction. The caveats matter, and Auvert and colleagues state them plainly. This trial measured male acquisition of HIV through heterosexual contact in one region of South Africa. It says nothing, directly, about male-to-female transmission — that would require separate investigation. Sixty percent protection is not immunity; circumcised men remained susceptible. And the brief window immediately after surgery, before healing is complete, may carry elevated risk — which is why the trial required six weeks of abstinence and why removing those forty-two days from the analysis changed the early-period rate ratio only modestly. The framing Auvert and colleagues chose for their conclusion is deliberately provocative: male circumcision, they write, provides protection equivalent to what a vaccine of high efficacy would have achieved. In a field where HIV vaccine trials had failed for decades, that comparison lands hard. A one-time surgical procedure, inexpensive and widely deliverable, cutting infection risk by sixty percent over a year and a half of follow-up. They are also careful about what circumcision is not. It is not a replacement for condoms — the intervention group's slight uptick in sexual activity is exactly the kind of behavioral shift that can erode a prevention effect at the population level if the message is misread as full protection. Circumcision belongs in a toolkit alongside condoms and counseling, not instead of them. Auvert and colleagues recommend it as an important public health intervention in settings where most men are uncircumcised and HIV transmission is predominantly heterosexual — with the stipulation that scale-up must be medicalized to avoid the very adverse events their trial documented. What the ANRS 1265 trial did was transform an observational pattern into experimental proof. It answered the causal question that had been open since 1986. Thirty-six years of correlation gave way to three thousand two hundred seventy-four men, sealed envelopes, and a result clear enough to stop a trial early. The questions that follow — about female partners, long-term durability of protection, and what safe large-scale implementation actually looks like — are real and still open. But the foundational claim is no longer in doubt. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.