The presentation of a study at the recent Conference on Retroviruses and Opportunistic Infections (CROI 2026) in Denver caused consternation when it appeared to contradict data that have served as the basis for the ‘Undetectable equals Untransmittable’ (U=U) campaign, which informs people with HIV that they cannot pass on HIV during sex if they have an ‘undetectable’ viral load – defined in the foundational studies as below 200.
The study, presented by Dr Michael Martin of Johns Hopkins University calculated that someone testing with a low but detectable viral load ranging from 200-1000 did have a quantifiable risk of passing on HIV in the next year.
In couples where one partner was HIV positive and the other negative, its model suggested that someone whose viral load was 15,000 or more at a particular timepoint had a 14% risk of passing on HIV to their partner within a year, and 2.9% if their viral load was 1000 at that timepoint.
But what really concerned audience members was that the model said that if their viral load was 200 at the time of testing, then there was still 0.6% risk of their passing on HIV to their partner in the following year.
A new study from Rakai
The study is from the Rakai cohort in Uganda, which involves 15,000 heterosexual couples where, at the time they join the cohort, one partner has HIV and the other does not. This study’s conclusions gained piquancy because a previous study of HIV transmission in this cohort, published in March 2000, was the first to find that people with low viral loads rarely passed on HIV: there were 90 cases of HIV transmission between 415 couples in that study, but none where the HIV-positive partner had a viral load below 1500.
In the new Rakai study, according to the researchers, 674 couples were studied “that were in cohabitating partnerships… where the HIV seronegative partner did not report being in any other relationships to have had no more than one sexual partner over the past year.”
They found 111 cases of HIV transmission – a similar incidence to the 12% annual incidence in the first Rakai study. Among those 111, they found nine transmissions from someone whose viral load was between 200 and 1000 last time it was tested – 8% of all transmissions.
Background: the U=U studies
This doesn’t just go against the data from a 26-year-old study whose data was collected in the 1990s. Since then we have had three studies – PARTNER 1, PARTNER 2, and Opposites Attract.
Taken together, they found no occasions where a partner with HIV with a viral load below 200 passed on the virus to their HIV-negative partner in 152,888 occasions of condomless sex between 2013 heterosexual and gay male couples.
Importantly, no “phylogenetically linked” transmissions were seen. What this means is that there were a number of HIV acquisitions where the HIV negative person’s HIV came from an outside partner (with an unknown viral load). But there were no transmissions from a partner with an undetectable viral load, as established by genetic testing to see if it came from the primary partner.
It’s important to emphasise that in the new study from the Rakai cohort, phylogenetic testing was not done. So we do only have the participants’ assurance that they were monogamous.
Aims of the present study
The authors of the study, however, during the presentation and in correspondence with aidsmap, emphasised that their aim in the study was different from the foundational studies of U=U.
Professor Alison Hill of Johns Hopkins University, who led the study, said that it was not disputing the fact that someone who had an ‘undetectable’ viral load would not pass on HIV. Instead, it was asking the following question: if a person with HIV had a low-but-detectable viral load (on or off treatment) at one time point, what was the risk that they would pass on HIV during the following year? In other words, it was a modelling study that extrapolated a risk from observations, rather than directly measuring transmissions.
“We’re not questioning or providing any evidence against U=U,” she said at the conference. “That’s a policy for people on treatment who have a viral load under 200. Whatever it is, it’s most likely that their actual viral load is much lower – a few copies”.
“Our study is about the small minority of people whose viral load is suppressed, but detectable and quantifiable – between 200 and 1000.”
The study was prompted by a review by the World Health Organization that found a couple of cases where someone with a viral load between 200 and 1000 passed on HIV to a partner. The review was commissioned because the WHO was concerned that some viral load tests used in lower-income settings did not quantify viral loads below 1000 copies. Was it possible to say that transmissions from someone with a viral load below 1000 never happened?
Inconveniently, the review found a couple of occasions where someone with a viral load in that range (which the WHO called “suppressed but detectable”) did appear to transmit.
The time gap between viral load and HIV tests
One of the cases occurred in the HPTN 052 study, which compared the risk that a partner on antiretroviral therapy (ART) could transmit HIV with a partner not yet on ART. Among 43 occasions of a phylogenetically-linked virus being passed on, eight came from a partner on ART. While viral loads were generally over 10,000, in one case, the last viral load in the HIV positive partner was 617.
However there was a gap of 50 days between this viral load test and the estimated date they passed on HIV – more than enough time for their viral load to rebound to over 1000 copies.
Dr Myron Cohen, HPTN 052’s principal investigator, wrote at the time that “In all eight cases, the data indicated that the index participant was most likely viraemic at the time of HIV-1 transmission.”
This could be what happened in the nine cases in the second Rakai study. Professor Hill told aidsmap the following facts about the individual transmissions:
- In one couple, the HIV-positive partner had a viral load between 200 and 1,000 at the time of the HIV-negative partner’s last negative test, and a viral load over 1,000 at the follow-up visit.
- In two of the couples, the HIV-positive partner had a viral load over 1,000 at the time of the HIV-negative partner’s last negative test, and a viral load under 1,000 at the follow-up visit.
- In four couples, the HIV-positive partner had a viral load between 200 and 1,000 at the time of the HIV-negative partner’s last negative test and was undetectable at follow-up.
- And in two couples, the HIV-positive partner had a viral load between 200 and 1,000 at the time of the HIV-negative partner’s last negative test, but here was no viral load result available at the time they first tested HIV positive.
So, in these cases, what was the time gap between the previous viral load test and the HIV-negative partner testing positive? Although gaps of as long as two years between the HIV positive partner’s last viral load and the initially negative partner testing positive were allowed in the study, the average time gap was 141 days – 4.6 months.
This doesn’t provide a full answer to why transmissions were seen – because in the PARTNER studies, the time gap between the positive partner’s last viral load test and the negative partner’s positive HIV test was similar – four to five months.
The instability of viral loads
Presenting the new study at CROI, Michael Martin explained the motivation behind it more fully. Their main concern was to quantify the instability of low-level viral loads, he said.
“We don’t think low-level viraemia (LLV) is a significant driver of population-level HIV incidence right now, but that could change if treatment and test access is threatened,” he said.
“If someone has a viral load between 200 and 1000, it’s usually on the way somewhere. In future a higher proportion of people with LLV could be on their way to a rebound.” In other words he agreed that, in the HIV-positive partner, their viral load may have been high and now on the way to undetectability, or it might have previously been low but now be increasing.
To try to rule out these cases, they further restricted their mathematical model to cases where the HIV positive partner was not on ART. The reason for doing this, they said, was that viral load in people on ART tended to vary more than viral load not on ART. People on ART might rebound due to ART access restrictions or adherence problems. But people off ART who had LLV were usually people with some degree of innate control of HIV, and this tended to vary less.
There were 415 people in the study who were either not on ART, or weren’t at the time of the initially-negative partner’s positive test (they excluded people who started ART during the time gap between the two tests).
In the study presentation, Martin said that this model produced an estimate of “1.3 transmissions per 100 person-years of monogamous heterosexual contact from people with low-level viremia.”
In other words, if someone had a viral load between 200 and 1000, there was a 1.3% chance that their partner would test HIV positive within a year. The model – which was described as a “mechanistic relationship” – predicted that the statistical chance of transmission to the partner during the following year if the person’s viral load was exactly 200 was 0.6%. If viral load was 1000, the chance was 2.9% – compared with about 13% if their viral load was over 15,000.
Despite disclaimers that the model aimed to quantify the rate of instability in low-level viraemia rather than its absolute risk, Michael Martin did recommend that people presenting with a viral load between 200 and 1000 should be counselled about the risks.
After its review in 2023, the WHO said that “People living with HIV who have a suppressed viral load [of 200 to 1000] and are taking medication as prescribed have almost zero or negligible risk of transmitting HIV to their sexual partner(s).”
“I wouldn’t call this risk negligible”, he said.
Reactions and comments
This caused some consternation in the audience. One sexual health worker asked if the 0.6% risk of transmission at a viral load of 200 was based on an actual observation of a transmission at that figure. “What am I supposed to say to people who come in?” he asked.
There were also questions about whether sexually transmitted infections might have influenced the transmission risk, and whether transmissions were seen more often from man to woman or the other way round.
At a wrap-up panel on the last day of CROI, Professor Chloe Orkin of the Barts and Royal London Health Trust in London picked up on lack of phylogenetic testing as the possible explanation for the unexpectedly high estimates of transmission rate between people “considered to be monogamous”. She criticised the researchers’ recommendation that people with LLV should be counselled on the risk as “very bold”.
She said: “In this context, a single case report of a single transmission from someone with a viral load below 200 would have a huge impact – and so far, there’s been no such case report.”
The session’s moderator, Professor Joseph Eron of the University of North Carolina, said: “I’m more likely to get patients coming in with undetectable viral loads, below 20, who ask ‘Am I still U=U?’, and my answer to them will still be, absolutely, yes.”
