This Cochrane review is an excellent

This Cochrane review is an excellent attempt to estimate the effectiveness of condom use in reducing heterosexual transmission of HIV. While several studies have looked at this issue, many of these have been small, and the reviews done to date have had a number of methodological problems associated with them.

The authors of the Cochrane review identified and selected 14 studies involving discordant couples (i.e. couples in which one of the partners is HIV-positive and the other free from HIV). A total of 587 people who reported "always" using condoms during sexual intercourse, and 276 people who reported "never" using condoms were included in these studies.

The incidence of HIV infection among those who reported always using condoms was 1.14 per 100 person-years (95% confidence interval 0.56-2.04), while it was 5.75 per 100 person-years (95% CI 3.16-9.66) among those who never used them. This gave an 80% reduction in the incidence of infection with condom use.

It is important to note, however, two important limitations: (a) the meta-analysis was done using data from observational studies; and (b) the authors did not provide confidence intervals for their estimated effect of 80%.

This is a well done review. As the authors correctly point out, it is logistically and ethically impossible to do randomized controlled trials to find out whether condom use reduces the risk of HIV transmission. As such, we have to rely on observational studies. The studies used in this review typically compare rates of HIV infection in groups of people who report always or never using condoms. Observational studies inherently carry a risk of bias as people make choices for reasons, and if the choice about using condoms or not is related to other risk factors for HIV, then the estimate of effect that we get will be inaccurate.

So, in this study, in spite of the best efforts of the authors, we can’t be sure that condoms really reduce HIV incidence by 80%. It may be, for example, that people who use condoms always also practice safe sex and have fewer partners, thereby reducing their HIV risk irrespective of condom use. It might also be that some people who report always using condoms do not actually do so but tell the researchers that they do in order to “look good”. Similar points could be made about people reporting not using condoms. Self-reported data always have the risk of being unreliable; and in this case there is no other means of confirming the findings.

If we take the extreme values from the 95% confidence intervals of the HIV incidence rates reported in the two cohorts, we find that condom use may be associated with reduced HIV incidence of between 94% and 35%. The wide range of values is explained by the small numbers of people (863) and HIV infections (51) in the studies. Concurrent use of other measures could also contribute to this variation. As a result, the estimate of the incidence of HIV in the two groups is fairly imprecise (between 0.56 and 2.04 per 100 person-years for users, and between 3.16 and 9.66 per 100 person-years for non-users).

This lack of precision is not the fault of the authors and their review. It is an inherent feature of the available studies. As such, it is not really appropriate to estimate condom effectiveness at 80%. While 80% is the best single estimate of effectiveness, it is also fair to say that the true measure of the effect could be as low as 35% or as high as 94%, as the authors state in the review.