Reader Comments

Post a new comment on this article

Trial of circumcision for HIV prevention should be interpreted with caution

Posted by plosmedicine on 30 Mar 2009 at 23:48 GMT

Author: Edward Mills
Position: Fellow
Institution: McMaster University
Additional Authors: Joel Menard, P.J. Devereaux, Victor M. Montori, Gordon H. Guyatt
Submitted Date: December 05, 2005
Published Date: December 6, 2005
This comment was originally posted as a “Reader Response” on the publication date indicated above. All Reader Responses are now available as comments.

With the publication of a positive trial of adult circumcision for the prevention of HIV in South African men, some will argue that the evidence is now irrefutable. Are they correct? We think not.
Randomized control trials (RCTs) allow stronger inferences than observational studies, but the frequency with which early trials provide overly optimistic results suggest caution in acting on the results of initial RCTs. This is particularly true for trials with methodological limitations.

The investigators of this randomized trial of circumcision clearly strove for methodological rigor, yet difficulties remain. The most serious concern arises from the investigators' decision, prompted by a recommendation from their data monitoring committee, to stop the trial early. As we have pointed out in a recent systematic review of 143 RCTs stopped early for benefit[1], such truncated RCTs will, on average, overestimate treatment effects. The fewer events that have accrued at the time investigators terminate their trial, the higher the risk of overestimation. When comparing the trials with events fewer than the median number (66) to those above the median, we found the odds ratio for a magnitude of effect greater than the median (relative risk reduction 47%) was 28 (95% CI 11 - 73%). The investigators only observed a total of 69 events (patients newly HIV positive), placing the trial in a category at high risk of serious effect overestimation.

In this trial, the Data Safety and Monitoring Board recommended early study termination at the 2nd interim analysis, having shown no statistical difference at the first. The results reached the defined statistical threshold for stopping, but the number of events per group was low (20 in active compared to 49 in the control). As demonstrated in our systematic review, it is possible a random high in the number of outcomes occurs at the time of an interim analysis and longer follow-up may have reduced or eliminated the differences between groups. While Auvert and colleagues have provided exciting results that appropriately alert us to the potential benefit of circumcision, it is unlikely the results of their trial are accurate.

Implementing circumcision as public policy following this trial highlights the need for investigators and data monitoring committees to strike the appropriate ethical balance between responsibilities to trial participants and responsibilities to a wider community that is poorly served by unnecessary uncertainty and overestimates of intervention effects. The media attention that this trial received warrants efforts to ensure that circumcised men do not increase risky behavior due to a perceived sense of protection.

Stopping trials prior to planned termination requires caution beyond statistical and biological rationale. The ongoing trials of circumcision will add greater insights into the protective role of circumcision and an individual patient data meta-analysis would also help. The tremendous commitment and impressive accomplishment of the Auvert team leave those in charge of the fight against the AIDS pandemic with the responsibility to fund a definitive large randomized trial of adult male circumcision for HIV prevention. Such a trial should include foolproof concealment of randomization and not stop before at least 200 events have accumulated.


1. Montori VM, Devereaux PJ, Adhikari NK, Burns KE, Eggert CH, Briel M, Lacchetti C, Leung TW, Darling E, Bryant DM, Bucher HC, Schunemann HJ, Meade MO, Cook DJ, Erwin PJ, Sood A, Sood R, Lo B, Thompson CA, Zhou Q, Mills E, Guyatt GH. Randomized trials stopped early for benefit: a systematic review. JAMA 2005;294:2203-9.

No competing interests declared.