HIV is the only STI for which formal meta-analyses of condom efficacy have been published, though only for heterosexual couples2 3 and NIAID evaluated what was then the most recent analysis, by Davis and Weller.2 This analysis, in which only longitudinal or cohort studies were included, used the following criteria to select studies related to condom use and HIV/AIDS prevention:
- The sample only included serodiscordant, sexually active, heterosexual couples.
- HIV status was determined by serology (so that exposure to HIV was known).
- Data collection included self-reporting about condom use.
- The study design afforded longitudinal follow-up of HIV uninfected partners.
Davis and Weller found twelve studies that met these criteria. The meta-analysis noted the direction of transmission (male-to-female, female-to-male, and unstated) and date of study enrolment. Condom usage was classified into the following three categories: always (100% use), sometimes, and never.
Among participants who reported always using condoms, the summary estimate of annual HIV/AIDS incidence from the twelve studies was 0.9 seroconversions per 100 person years (0.9%). Among those who reported never using condoms, the summary estimate of annual HIV/AIDS incidence from the seven studies was 6.7%. Overall, Davis and Weller estimated that condoms provided an 85% reduction in HIV/AIDS transmission risk when infection rates were compared in ‘always’ versus ‘never’ users, though the confidence intervals meant that the ‘true’ efficacy could be as low as 60% or as high as 96%.
To cite one of the twelve studies in more detail,4 researchers looked at Italian serodiscordant couples in which the male partner was HIV positive. Among 171 couples who used condoms consistently, three women (2%) became infected with HIV whereas among 134 couples who used condoms inconsistently or not at all, 16 women (12%) became infected, a relative risk of 6.6 indicating a condom efficacy of about 84%. Annual HIV incidence was 7.2% among women not using condoms and 1.1% among consistent condom users. This study also found that reported anal sex raised the risk to women by another 40%, but only if they did not use condoms.
Davis and Weller subsequently published another meta-analysis in 2003,3 this time of 14 studies. These included 13 cohorts of ‘always’ users that yielded an annual HIV incidence estimate of 1.14%, with a confidence interval ranging from 0.56 to 2.04%. There were ten cohorts of ‘never’ users. The studies with the longest follow-up time, consisting mainly of studies of partners of haemophiliac and transfusion patients, yielded an HIV incidence estimate of 5.75% with a confidence interval of 3.16 to 9.66%. Overall effectiveness, the proportionate reduction in HIV seroconversion with condom use, was approximately 80%, a downward revision from 1999.
The only other meta-analysis of condom efficacy that satisfied NIAID’s criteria was in 1997.5 In this analysis by Pinkerton, two sets of studies were evaluated employing two different standards of ‘consistent’ use. For the less rigorous standard, in which ten studies compared ‘users’ with ‘non-users’, the estimated condom efficacy rate was 79%. For the more rigorous standard, nine studies compared ‘always’ users with ‘never or inconsistent’ users (note this is different from Davis and Weller, who compared ‘always’ with ‘never’ users).
In these nine studies there were only four seroconversions reported among 277 (1.4%) steady partners of HIV-positive men and women who consistently used condoms. In contrast, when condoms were used inconsistently or not at all, 171 of 867 (19.7%) sexual partners seroconverted. This led to an efficacy estimate of 93% effective when condoms are used consistently as opposed to inconsistently.
This is about the highest standard of proof we can expect from studies of condom efficacy. Taking Davis and Weller and Pinkerton together, one can say that the best efficacy estimates we have for the use of condoms in preventing HIV are:
- “Almost always” versus “never”: 79%.
- “Always” versus “never”: 80 to 85%.
- “Always” versus “sometimes”: 93%.