Chinese study sheds light on how well treatment may work as prevention in the real world

A study from the Henan province of China published recently shows that antiretroviral therapy (ART) may not be as effective at suppressing HIV and preventing onward transmission in ‘real world’ settings as it is in the best clinical practice.

The study found that ART given to the partner with HIV within monogamous heterosexual couples with different HIV status only reduced the number of HIV infections passed on between them by about 50% over the course of a study period – though its effectiveness of ART did increase over time and by the end of the study stood at 67%.


The study examined a cohort of people with HIV, in which research had already yielded two contradictory estimates of the effectiveness of ART in preventing transmission. Many of the original HIV-positive partners were infected through selling blood in non-sterile conditions between the early 1990s and 1998. In 1998, the practice was outlawed, but not before it had created a local HIV epidemic.

There appear to be low rates of extramarital sex and sexually transmitted infections (STIs) in this group, so HIV infections are usually directly transmitted between partners, allowing estimates of infectiousness to be made.


treatment failure

Inability of a medical therapy to achieve the desired results. 


Term used to indicate how well a particular drug is tolerated when taken by people at the usual dosage. Good tolerability means that drug side-effects do not cause people to stop using the drug.


A serodiscordant couple is one in which one partner has HIV and the other has not. Many people dislike this word as it implies disagreement or conflict. Alternative terms include mixed status, magnetic or serodifferent.

detectable viral load

When viral load is detectable, this indicates that HIV is replicating in the body. If the person is taking HIV treatment but their viral load is detectable, the treatment is not working properly. There may still be a risk of HIV transmission to sexual partners.

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’. 

In 2010, a study (Wang) reported on Aidsmap caused concern by finding no association between the HIV-positive partner being on ART and transmission to the HIV-negative partner. In fact, 5% of partners who were on ART transmitted HIV to their spouses during the study period but only 3% not on ART, though this difference was not statistically significant. When adjusted for baseline characteristics, 32% fewer infections came from partners on ART than from partners not on ART; this was still not statistically significant.

At the time the researchers were puzzled themselves, saying that ART status could have been misreported, though also pointing out that a previous Chinese study (Li) had found that only a third of patients in the same cohort were adherent to their therapy after six months. A clue was provided by the fact that people who had not changed their ART regimens were over 2.5 times more likely to transmit HIV, suggesting that a combination of poor drug tolerance, low adherence and resistance could be the reason.

Less well publicised was a follow-on study in 2013 (Wang 2013) by the same researchers that completely contradicted their first result. With three more years of data, the annual incidence of HIV within serodiscordant couples had halved over the whole time period of the study from 1.7% to 0.8%. It had fallen by two-thirds within three years from 2.14% at the end of 2008 to 0.9% by the end of 2010. And, most strikingly, the chance of a partner with HIV transmitting the virus to their spouse was 95% lower if they were on ART at the time of their most recent clinic visit. This result was in line with randomised controlled studies like HPTN 052 but the stark contrast with the previous findings was a puzzle.

The study and findings

The present study (Smith 2015) suggests some explanations. In this study, all available data from initially treatment-naive serodiscordant couples in Henan Province between 2006 and 2012 were analysed: a total of 4916 couples who contributed an average of 5.4 years’ follow-up. This is a larger group than in the previous studies.

Their average age was 44, and 54% of the HIV-positive partners were women. Forty-seven per cent of the positive partners had originally been infected through blood plasma donation. Nearly all (90%) were farmers, with low education levels. Condom use was reported as quite high – 63% said they ‘always’ used condoms. Only 0.7% (32 individuals) reported any extramarital sex and only 0.4% of the HIV-negative partners (13 people) had had an STI diagnosed in the previous year.

The important difference between this and previous studies is that it only looked at HIV-positive partners who were treatment-naive at the start of the study: unlike the previous studies, it excluded people who were already on ART at the start of the period studied. The reason for this is to eliminate two opposite sources of bias caused by the fact that there are fewer data on what happened to people on ART in the pre-study period.

On the one hand, ART in China before 2006 may have been far less tolerable and subject to interruption than it has become more recently. At this point, as the researchers point out, quite toxic drugs like didanosine (ddI, Videx) were still being used.

On the other hand, with regimens where toxicity is less of an issue, people who have been on ART for a long time are often more settled on it, have better adherence, and are less likely to switch and a lot less likely to experience treatment failure, than people in their first year on therapy. By including people who had been on ART since before the start of the study, researchers could have been biasing their study group in the direction of stability and underestimating the potential for transmission during the HIV-positive partner’s first year on ART when they are more likely to have detectable viral load and to experience treatment failure or switch their therapy.

In the event, over 80% of the initially HIV-positive partners had started ART by 2012. Establishing the start date of therapy also enabled the researchers to see whether ART became more or less efficient at restricting HIV transmissions over time in a more sensitive way than the 2013 study.

There were 157 HIV infections noted in over 26,000 couple-years of follow up, an incidence of 0.59% a year. Incidence in couples before the HIV-positive partner started ART was considerably higher: 5.87% a year. However, because the cohort spent the majority of time with the initially HIV-positive partner on ART, a higher number of actual infections (84 versus 73) came from a positive partner on ART. This mean that, crudely, the actual reduction in risk conferred by the HIV-positive partner taking ART was only 29%.

However, people are more likely to start treatment if they are sicker, which also means they are more infectious but have less sex. After adjusting for this, the reduction in risk conferred by putting the HIV-positive partner on ART was 48%.

More sex or lower self-reported condom use was also associated with higher HIV transmission rates. The direction of transmission in terms of gender – male-to-female or female-to-male – did not influence transmission rates.

The efficacy of ART in reducing transmission increased as time went on. Between 2006 and 2008 it only had a modest 32% efficacy in reducing the likelihood of HIV infection, in agreement with the 2010 study; but between 2009 and 2012 its efficacy was 67%. This efficacy is in accordance with some other cohorts, though nothing like that seen in randomised controlled trials or prospective cohorts like PARTNER.

One other striking finding was that the efficacy of ART in reducing transmission did not apply at all when the initially HIV-positive partner had a CD4 count below 250 cells/mm3. Below this, there was no difference in transmission rate; above it, the efficacy of ART in reducing risk was 59%.

Questions and comments

Why was efficacy lower than in these studies, and why did it improve over time? The researchers cannot be certain but speculate it was a mix of improved and less toxic regimens arriving, and better drug delivery systems with fewer stockouts and times off therapy. Similarly they cannot establish why the effect of ART did not extend to partners with low CD4 counts. Was it because people with low CD4 counts were more infectious? Or did they have lower CD4 counts because they were non-adherent and therefore more infectious?

The researchers comment that further light will not be shed on these questions until there is ‘specialised monitoring’ – i.e. viral load testing – so that drug failures can be picked up on and adherence supported. It’s also the case that all transmissions were assumed to come from the spouse. If extramarital sex was more common than participants admitted, then a fair proportion of infections could have come from non-regular partners not on ART, though the low rate of STI diagnosis seems to put some limit on this.

In practice, these transmission rates and reductions in transmission due to ART may be closer to what is currently achievable in practice in lower-income settings. In large parts of the world, viral load testing is still not feasible and specialised adherence support unavailable. These patients were, in the main, poor rural folk receiving basic-level HIV care from a network of village clinics: the type that may be necessary if we are to expand global HIV treatment availability further, especially to levels where it could start to significantly impact HIV transmission rates. Clearly high adherence, tolerable regimens and ease of reliable monitoring remain crucial for the success of such programmes.


Smith MK et al. Treatment to prevent HIV transmission in serodiscordant couples in Henan, China, 2006 to 2012. Clinical Infectious Diseases, epub ahead of print, pii: civ200. 2015.

Wang L et al. HIV transmission among serodiscordant couples: a retrospective study of former plasma donors in Henan, China. J Acquir Immune Defic Syndr, 55: 232-38, 2010.

Wang L et al. Heterosexual transmission of HIV and related risk factors among serodiscordant couples in Henan province, China. Chin Med J (Engl 126(19):3694-700. 2013.

Li JY et al. Prevalence and evolution of drug resistance HIV-1 variants in Henan, China. Cell Res 5(11-12):843-9, 2005.