Condom efficacy may have been underestimated, new CDC study suggests

Image: kittiwat chaitoep/

Summary – what's different about this analysis

A new meta-analysis (Johnson) of the efficacy of condoms in preventing HIV transmission via anal sex between gay men has found a considerably higher estimate of their efficacy than two previous analyses.

The new estimate by the US Centers for Disease Control and Prevention (CDC) finds that condoms, used 100% of the time, stop more than nine out of ten HIV infections. Two previous analyses, one published back in 1989 (Detels) and the other in 2015 (Smith) found that they only stopped seven out of ten infections.

This new estimate for condom efficacy in sex between men is much closer to the estimated efficacy for sex between men and women; a synthesis of studies finds an efficacy in the region of 80-85%.

Why is this new estimate of efficacy (91.6%, in the case where the HIV-negative partner is the receptive one) so much higher than previous ones (72.3% for the same risk in the 2015 analysis)?



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


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


Having sex without condoms, which used to be called ‘unprotected’ or ‘unsafe’ sex. However, it is now recognised that PrEP and U=U are effective HIV prevention tools, without condoms being required. Nonethless, PrEP and U=U do not protect against other STIs. 


A pill or liquid which looks and tastes exactly like a real drug, but contains no active substance.


When the statistical data from all studies which relate to a particular research question and conform to a pre-determined selection criteria are pooled and analysed together.

There are several possible reasons, the CDC researchers think. One is that the new analysis examines condom efficacy in four different studies, whereas the previous ones only looked at a single study (Detels) or two (Smith).

However, they think that the crucial difference is that they look at condom efficacy per number of partners instead of per sex act.

Counting the number of partners may be a more reliable guide to risk than counting sex acts. This is because in cases where there are multiple sex acts between one couple, the risk tends to go down with time: at least one early study showed an extremely high risk (about a 25% chance of infection) during the first year of exposure but then a considerable falling-off of risk; transmissions became infrequent after a few years. This may be due to variation in viral load: partners with high viral loads transmit in the first year, while those with low ones may never do.

Because there is less risk of infection as time goes on, the risk of not using condoms also diminishes over time – and so, therefore, does their apparent efficacy.

If on the other hand, someone continues having sex with multiple partners, their infection risk does not diminish over time because their chances of encountering someone with a high viral load stays constant – and therefore so does the efficacy of condoms.

The studies examined

This analysis examines condom efficacy in gay men in four different studies, all conducted before 2005, three of them studies of gay men’s eligibility and preparedness to participate in vaccine studies. The primary reason later studies were not examined is because after that the proportion of partners known to have HIV who were on fully suppressive antiretroviral therapy and therefore not infectious became too large.

The studies were:

  • 1996: JUMP-START, a US study of gay men’s readiness to participate in a vaccine study, conducted in 1993-94.
  • 2004: Project Explore, a large study of a behavioural intervention to reduce risky sex in gay men in two US cities. It produced modest reductions in sexual risk-taking.

All of these studies collected data on the difference in the number of HIV infections in HIV-negative gay men who had anal sex with partners known or perceived to have HIV and either always used a condom or used them never or sometimes. HIVNET and VAX004 did not only include men who have sex with men (MSM), but only MSM were included in this analysis. Studies varied in length but all lasted at least 18 months. Participants reported on sexual risk behaviour in the previous six months in all studies.

Condom efficacy across and within studies

The present (2018) analysis looked at all four studies. The analysis published in 2015 only looked at the latter two of the above studies. Interestingly, there was a big difference in condom efficacy for receptive anal sex between these two studies (it was 87% in Project Explore, but only 64% in the VAX004 study).

During their work on the present analysis, the researchers looked at another analysis of the four studies that was published in 2014 (Scott). This did not present estimates of condom efficacy, but the authors of the present analysis were able to calculate them from infection rates.

In that 2014 study, the authors found a pooled condom efficacy of 89% (for receptive sex) over the HIVNET, Explore and VAX004 studies. The risk of infection per sex act was 0.73% (one infection per 136 acts) in people who sometimes or never used condoms, and 0.08% (one in 1250 acts) in people who always did. A separate analysis of the JUMP-START study found an efficacy of 93%.

In the 2015 (Smith) study, the risk per act of receptive sex was similar in people who never or sometimes used condoms (0.54% or one infection per 185 acts) but was considerably higher (0.2% or one per 500 acts) in people who always did. The risk of infection over a whole year was 13.2% in people who sometimes or never used condoms and 3.8% in people who always did. This led to this study’s condom efficacy estimate of 72%.

Why the differences? The present study explains why. Using a different methodology, they calculated the risk per extra partner of HIV infection in people sometimes or never using condoms for receptive anal sex versus people who always used them. The per-partner risk in people who sometimes or never used condoms was 83%. This means that for every additional partner with HIV they had condomless receptive sex with, their risk of HIV infection rose by 83%. In people who always used condoms, it only rose by 7.3%. This leads to the headline condom efficacy of 91%.

There was a considerable difference in efficacy between the studies, however. In the HIVNET study, the extra risk per partner was 164% in people who sometimes or never used condoms and only 0.7% in people who always did – indicating near-100% condom efficacy.

In contrast in VAX004 the extra risk per partner was 40% in those who sometimes or never used condoms – and 11% in people who did. This led to a more modest per-partner condom efficacy of 73%.

It’s VAX004 that is the outlier, though. The per-partner condom efficacies calculated for JUMP-START and Project Explore were 95% and 96%.

The difference in condom efficacy was largely driven by men who had many partners. In men who had up to seven HIV-positive partners, the per-partner reduction in infections in men who always used condoms was in the order of 50-70%. However, in men with ten or more partners, the risk in men who sometimes or never used condoms was 80% per extra partner – meaning a ninefold increase in risk for ten partners – and virtually zero in people who always used condoms.

Discussion and comments

Why the differences between studies, and between partner numbers? The researchers told that they had not been able to confirm any speculative cause from the data. The VAX004 study was the longest, at three years, and the only one where people received an actual vaccine or placebo. Did these factors influence people’s condom use over time?

It’s also important to note that the studies dealt with perceived as well as known HIV status in partners. Were men’s perceptions of partners’ status very different from study to study – especially in the one study, VAX004, conducted internationally?

And in the case of men with lots of partners, it’s interesting that the per-partner risk in this analysis actually goes down in men who always use condoms the more partners they have. Did men who had lots of partners but were still using condoms 100% of the time become more skilled and consistent at using them than men who only had one or two?

The researchers in their last paragraph remind us that it depends on what question you ask when determining condom efficacy. They find that over the four studies, only 17.3% of participants said that they had ever had condomless receptive anal sex with a partner who had or was seen to have HIV. But when the same participants were asked a different question – had they always used condoms with all partners, and for both insertive and receptive sex? – the answer is that 83.7% had condomless anal sex during the study period at least once.

Finally, the last data in this analysis was collected 15 years ago. Much has changed since then – in terms of gay men’s condom use, their perception of risk, the likely infectiousness of partners with HIV, and whether they are likely to be using other prevention methods such as pre-exposure prophylaxis (PrEP). For these reasons, studies conducted recently might come to different conclusions in estimating both the absolute efficacy of condoms – and their effectiveness in contributing to the prevention of HIV within the gay population.


Main reference

Johnson WD et al. Per-partner condom effectiveness against HIV for men who have sex with men. AIDS 32(11):1499-1505. 2018.

Other condom efficacy analyses

Detels R et al. Seroconversion, sexual activity, and condom use among 2915 HIV seronegative men followed for up to 2 years. JAIDS 2(1):77–83, 1989.

Smith DK et al. Condom effectiveness for HIV prevention by consistency of use among men who have sex with men in the United States. JAIDS 68(3):337-344. 2015.

Scott HM et al. Age, race/ethnicity and behavioural risk factors associated with per-contact risk of HIV infection among men who have sex with men in the United States. JAIDS 65(1): 115-121. 2014.