CDC researchers publish estimate of effectiveness of condom use in anal sex

This article is more than 9 years old. Click here for more recent articles on this topic

A paper this month in the Journal of Acquired Immune Deficiency Syndromes finally makes publicly available a study - originally reported by Aidsmap nearly two years ago from the 2013 CROI conference but until now unobtainable even as an abstract – which gives an estimate for the effectiveness of 100% condom use as the strategy of choice for the prevention of HIV infection in gay men. They estimate that condoms used consistently stop seven out of ten HIV infections acquired through anal sex between men.

This analysis, by Dawn K Smith and three other researchers from the US Centers for Disease Control and Prevention (CDC), analyses two historic studies of condom use and HIV incidence in gay men and other men who have sex with men (MSM) in the US – one, VAX004, from 1998-1999 and the other, EXPLORE, from 1999-2001.

It finds an overall effectiveness of 70.5% in both studies for all anal sex with a partner who has HIV. The effectiveness where the HIV-negative partner is the receptive partner is slightly higher at 72.3%. This contrasts with an approximately 80% effectiveness of consistent condom use as a prevention strategy in several meta-analyses of studies of heterosexuals.



Receptive anal intercourse refers to the act of being penetrated during anal intercourse. The receptive partner is the ‘bottom’.


Insertive anal intercourse refers to the act of penetration during anal intercourse. The insertive partner is the ‘top’. 


How well something works (in a research study). See also ‘effectiveness’.

statistical significance

Statistical tests are used to judge whether the results of a study could be due to chance and would not be confirmed if the study was repeated. If result is probably not due to chance, the results are ‘statistically significant’. 


The presence or absence of detectable antibodies against an infectious agent, such as HIV, in the blood. Often used as a synonym for HIV status: seronegative or seropositive.

The difference between the MSM and heterosexual studies is not actually statistically significant, but Smith and colleagues argue that “it is more appropriate to use the MSM-specific point estimate of 70% effectiveness for discussions and models involving anal sex among MSM than to continue use of the heterosexual 80% effectiveness point estimate for MSM”.

The studies analysed

The VAX004 study was the first large-scale phase III HIV vaccine study to be conducted in the USA and the EXPLORE study was the largest study ever conducted of a behavioural HIV prevention intervention in gay men. Taken together, 7725 men from both studies were eligible for this analysis. The average age of men in the two studies was 31, 82% were non-Hispanic white and 4.5% African American, and 76% had had some tertiary education.

These studies were chosen firstly because they are among the very few studies in gay men that report both self-reported condom use and HIV incidence in a large group of gay men longitudinally; secondly because they were large enough, and had enough seroconversions, for the figures computed for the risk of HIV infection to be, in the main, statistically significant.

According to national US surveillance, the proportion of gay men who report using condoms consistently over the last year has fallen from 52% in 2005 to 43% in 2011. In these two pre-2005 studies the proportion of gay men who reported ‘always’ using condoms in the six months before each study visit was 51% and the proportion who ‘never’ did  was 22%. The remaining 27% reported ‘sometimes’ using condoms. (As we report below, 100% condom use over the whole length of the studies was much less common.)

There was an element of ‘seropositioning’ in the behaviour of study participants. In terms of number of sex acts reported (rather than number of partners), study participants reported taking the insertive role 62% of the time. Whereas 30% of insertive sex acts involved participants who reported never using condoms during the previous six months, only 20% of acts of receptive anal sex did. The figures for ‘always’ using condoms in the previous six months was 36% for insertive sex and 46% for receptive sex. The figures for ‘sometimes’ using condoms were, however, the same regardless of sex role: 34%.

Study differences

Condom effectiveness differed between the two studies. The effectiveness seen in the EXPLORE study was 86.2% for all sex and 86.8% for receptive sex.  In contrast, the respective effectiveness figures seen in the VAX 004 study were only 61.3% and 64.1%.

Although participants in EXPLORE were on average younger, less likely to be white, better educated and more likely to be either west or east coast residents than in VAX 004, these factors are controlled for in this analysis and cannot explain the difference seen in condom effectiveness between the two studies. The difference may possibly be due to the fact that the earlier study was a trial that assessed the behaviour of gay men who were participating in a vaccine study while the second study was a trial of a behaviour-change intervention and although condom use only increased by 24.5% in this study and HIV infections decreased by 15.7%, it is possible that participants, especially in the intervention arm, acquired better condom-use skills during the four years of study follow-up.

The increase  in condom use in EXPLORE was with partners of known HIV status,  whether HIV positive (from about 60% to about 70% condom use) or HIV negative (from about 33% to 41%); condom use with unknown-status partners remained flat at about 67%. This could suggest that EXPLORE participants may have developed a bit more ease with disclosure and negotiation as they went along. In contrast, condom use declined in VAX004 in all partner HIV status categories (from about 67% to 53% with HIV positive partners and 47% to 32% with negative partners.)

Consistent condom use and HIV infections

The main comparison made by the researchers is between study participants who reported ‘never’ using condoms during discrete six-month periods of follow-up, and ones who reported ‘always’ using them during those periods. The researchers also determined how many study participants had partners in each six-month period that they knew to be HIV positive and looked specifically at condom usage rates with these partners.

Clearly the known HIV-positive partners would not be all the partners with HIV that study participants had contact with. However there is no way of knowing how many of their other partners whose HIV status was unknown or assumed to be negative in fact had HIV. Estimating this would not add clarity because it cannot take account of differences in individual participant behaviour.

Combining figures from both trials, and excluding some participants such as those reporting sex with women or injecting drug use, the researchers included 7725 men, of whom 3490 (45%) reported sex with a known HIV-positive partner at any time during the studies. In total there were 42451 six-monthly appointments attended, and sex with a known HIV-positive partner was reported by participants at 22% of those visits. Among those men 225 (6.4%) acquired HIV infection, with no significant difference between VAX004 (7%) and EXPLORE (6%), and 2.5% of men per six-monthly visit turned out to have acquired HIV since their last visit.

In seroconverters, 14% in EXPLORE (19% in receptive sex) and 29% in VAX004 (38% in receptive sex) maintained they had always used condoms with HIV-positive partners in the six-month period during which they acquired HIV; 42% and 30% respectively said they never had; and 46% and 33% respectively said they sometimes had. This of course says nothing about condom usage in the previous six months with partners assumed to be negative or of unknown status.

In terms of the absolute likelihood of catching HIV, the researchers calculated that the 100% consistent use of a condom transformed the per-act likelihood of HIV infection in EXPLORE from one infection in 286 sex acts in men who never used condoms to one in 1429 in men who always did, and in VAX004 from one in 556 acts to one in 1000 acts. When only sex as the receptive partner was considered, the observed per-act transmission risk in EXPLORE was one in 90 acts without any condom use and one in 667 acts with 100% use, and in VAX004 one per 244 and one per 454 acts respectively.

Intermittent condom use

One of the other headline findings of this study however was that ‘sometimes’ using condoms was at best only marginally effective as a prevention strategy. While, as reported above, the effectiveness of ‘always’ versus ‘never’ condom use in the two studies overall was 70.5%, there was no significant difference in HIV acquisition rates between men reporting they ‘never’ used condoms in the previous six months and men reporting they ‘sometimes’ used them. There was a slight reduction in HIV incidence in men who reported receptive anal sex but this approximately 25% reduction in HIV infections in men who ‘sometimes’ versus ‘never’ used condoms was not statistically significant, in either or both studies.

It is important to say here that intermittent condom use was not stratified by frequency of use: so men reporting they ‘sometimes’ used condoms could have been using them 90% of the time, or 10% of the time. Therefore the above effectiveness figures for intermittent use do not mean that intermittently using condoms is never effective: it means that there is some point, as yet undetermined, between always and never using condoms, where they become ineffective as a strategy.

Long-term consistency of condom use

The other headline figures in the study was that always using condoms proved to be very difficult for trial participants over the long term. Over a  three-year period (the length of the VAX004 study: the extra year’s data in EXPLORE was cut off at the three-year limit) just 13% of participants maintained 100% condom use throughout in any anal sex, and 20% maintained condom use throughout as the receptive partner.

In fact inconsistency of condom use was the most consistent behaviour: 95.6% of participants used a condom at least once during the three years, so consistently never using condoms was rare. It is however notable that when only taking the receptive role is counted, 9% of participants never used a condom once throughout the study period when they were the receptive partner in anal sex.

Limitations and conclusions

This study has a number of limitations. Its findings will no doubt continue to be debated and it is notable that the study heading includes a disclaimer saying that “The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC”.

Firstly all reports of condom use, sexual activity and partner serostatus are self-reported.  We know that participants in research trials are likely to under-report behaviour that they think will be disapproved of (so-called ‘social desirability bias’). This would imply that condom use may have been lower than stated, which in turn would imply that condoms, where they were used, were more efficacious on an individual level than this study would imply.

Condoms used absolutely perfectly – i.e. with only mechanical failure, not human error, contributing to failure – are up to 99.5% effective, according to laboratory tests. What this study shows, though, is that even consistent use is hard to maintain over the long term, let alone perfect use. The CDC, in their own condom fact sheet, say that it is almost impossible to provide an estimate for the real-life effectiveness of consistent and correct use, “because of numerous methodological challenges inherent in studying private behaviors that cannot be directly observed or measured.”

This study is also based on findings from only two, very different, HIV prevention studies. However its findings do agree with the only previous study done of the effectiveness of condoms in preventing HIV in gay men (Detels), which found the same figure for effectiveness – 70%.

Partner serostatus was as reported by participants, and clearly some may have been infected by partners believed or assumed to be negative. But since the comparison in this study is between always and never using condoms with partners known to have HIV, this should not make a strong difference to the condom efficacy seen.

The publication of this study in a journal may give additional impetus to the need to research and roll out methods of HIV prevention for gay men that are additional to condoms.

“These data,” say the researchers, “suggest the need to further intensify efforts to educate HIV-uninfected MSM and improve their ability to accurately assess both their risk of HIV acquisition and the effectiveness of their current use of condoms (consistently or inconsistently) during anal sex.”

In addition, they add, their data also suggest “a need to provide access to additional highly-effective HIV prevention tools and strategies, including more frequent HIV testing (including testing as couples) and daily oral antiretroviral pre-exposure prophylaxis (PrEP) when indicated.”


Smith DK, Herbst JH, Zhang XJ, Rose CE. Condom effectiveness for HIV prevention by consistency of use among men who have sex with men (MSM) in the US. J Acquir Immune Defic Syndr, e-publication ahead of print. Doi: 10.1097/QAI.0000000000000461. December 2014.

Detels R et al. Seroconversion, sexual activity, and condom use among 2915 HIV seronegative men followed for up to 2 years. J Acquir Immune Defic Syndr 2:77–83, 1989.