HPTN 083: Injectable PrEP can make breakthrough infections hard to detect

Some cases of resistance, but not in people with falling cabotegravir levels after injections are stopped
Professor Raphael Landovitz presenting to CROI 2021.

A detailed analysis of results from the HPTN 083 PrEP study, which compared the efficacy of a two-monthly injection of the drug cabotegravir against daily oral tenofovir disoproxil fumarate and emtricitabine (TDF/FTC) pills, confirmed that there were only a third as many HIV infections in people given the injections as in people taking the pills.

Professor Raphael Landovitz told the virtual Conference on Retroviruses and Opportunistic Infections (CROI 2021) that final analysis of the 4566-person study showed that there were 39 HIV infections in participants given TDF/FTC versus 12 in participants given the cabotegravir injections (plus a lead-in period of four weeks taking cabotegravir pills, to rule out toxicity). This presented an annual incidence of 1.22% in the TDF/FTC arm and 0.37% in the cabotegravir arm.

This represents an incidence rate ratio (IRR) of 0.32, in other words that there were 68% fewer infections in people given the injections. This is virtually the same as the efficacy presented at the AIDS 2020 conference.

Late-detected baseline HIV infections

Behind this slight adjustment, however, is a wealth of interesting data. The reason for the slight change from the IRR of 0.34 given last year was that two more infections thought to have occurred while people were receiving cabotegravir injections turned out to be in people who actually had HIV when they entered the study. 

The first person was diagnosed (with a standard HIV antibody test) nine weeks after entering the study, two weeks after their second cabotegravir injection (the first two injections were only four weeks apart; after this they were every eight weeks). 

The second person was diagnosed with a standard HIV antibody test no less than 41 weeks into the study and after receiving five cabotegravir injections.

These two infections were revealed to be baseline ones because every participant who acquired HIV despite being randomised to cabotegravir had their stored blood samples subjected to a battery of retrospective, hypersensitive antibody, antigen and viral load tests. This showed that both participants had had HIV at baseline, with viral loads of 1360 and 44,180 respectively.

Importantly, the reason these had not been detected at the time is that the cabotegravir they took – first as pills, then injections – quickly suppressed their viral load to undetectable. They also became ‘HIV negative’ (had no detectable antibodies to HIV) for periods of time in confirmatory antibody/antigen tests.

"The evidence that injected cabotegravir could delay the detection of HIV infections is of concern."

One of the participants already known to have had HIV at baseline (they became HIV-antibody-positive at week eight but retrospective clinic-site testing showed a detectable viral load from baseline onwards) developed cabotegravir resistance mutations during their time on PrEP.

Three of the four participants with baseline infections (the other was lost to follow-up) started antiretroviral therapy (ART) and achieved viral suppression. Two did so as soon as diagnosed (including the person with cabotegravir resistance, who took a boosted protease inhibitor regimen), while the third started ART some considerable time afterwards. This person was therefore viraemic for a long time as cabotegravir levels tailed off in their body, and it was feared this would be an ideal way to set up drug resistance, but it did not happen.

These four baseline infections, classed as group A infections, were not even the most notable ones.

Infections in people who stopped or switched

In group B were five participants who tested HIV-positive a long time after they had stopped taking cabotegravir. In two cases they had basically never taken more than a pill or two of oral cabotegravir before leaving the study, while another tested positive 31 weeks after their last cabotegravir injection.

Group B also included two participants who were among the handful who switched to open-label TDF/FTC because of adverse reactions to the cabotegravir injections. One was lost to follow-up and was found to be HIV-positive when they re-engaged more than a year after their last TDF/FTC dispensation, and the other tested positive nine weeks after their last dispensation. So these two were in reality on oral TDF/FTC PrEP preceding their infection but, because they were randomised to the cabotegravir arm, count as infections on cabotegravir.

None of the three who had ever taken significant amounts of cabotegravir developed resistance to it, despite spending periods of 121, 31 and 40 weeks respectively with a detectable HIV viral load while levels of drug fell in their body. 

Infections during the oral lead-in phase

In group C were three participants who were infected during the period they were taking oral cabotegravir.

One of these participants had not taken their pills and re-engaged with the trial at week 33 (testing positive for HIV). It turned out they had actually acquired HIV at week five. The other two, however, had what should have been adequate levels of cabotegravir in their bodies at the time they acquired HIV, at weeks four and three of the study. These two may represent people who did not start taking their pills immediately.  Alternatively, these cases raise the possibility that oral cabotegravir might need a second drug to be fully effective. One had a low viral load (120) in their retrospective test while the other was undetectable on a standard test but detectable on a qualitative (yes/no) hypersensitive RNA test.

‘Breakthrough’ infections despite adequate cabotegravir levels

This left the most interesting cases: the four participants in group D who tested positive to HIV antibodies while receiving cabotegravir injections and appearing to have good drug levels.

Two turned out, by retrospective testing, to have been infected 17 and 6 weeks before their positive antibody test respectively, in both cases 8 weeks after their second cabotegravir injection.

The first had a detectable viral load (860) at their retrospective test but this had become undetectable between then and their diagnostic antibody test 17 weeks later. The second was only detectable with a hypersensitive qualitative RNA test in their retrospective test.

"No resistance arose in people with low-level cabotegravir in the post-injection ‘long tail’ phase, not even in the case of the participant who developed a very high viral load."

Both of these patients had had big dips in the cabotegravir levels in their body after their first injection, to about a quarter of the target concentration, before reaching target levels after their second injection. This suggests that in some people it may take longer than in others for injected cabotegravir to diffuse into the tissues where it needs to be in the rectal, vaginal and penile mucous membranes. However, if they did acquire HIV after their first injection this still leaves an unusually long period before signs of HIV infection became detectable.

The third case was even more puzzling. In this case, the participant tested positive for HIV with a standard test 72 weeks into the study, at the appointment for their tenth cabotegravir injection. Retrospective testing pushed the date of infection back to week 56, at their eighth cabotegravir injection, at which point they had a viral load of 120. They did have a brief dip in their cabotegravir levels after their first injection, and a very slight dip after their sixth injection at week 42. But even if they acquired HIV then, this would leave the unusually long time of 14 weeks between infection and the first sign of HIV appearing.

This was a participant in which cabotegravir levels were only just at the target concentration of 1.33 micrograms per millilitre (mcg/ml) throughout. However this target concentration is eight times the IC90 of cabotegravir, which is the concentration needed to inhibit viral replication by 90%. This should be enough, but does the concentration need to be higher in some people?

In this participant, delays in arranging ART for 13 weeks led to integrase inhibitor resistance developing and their viral load increasing from 158 to 153,000. However, it responded successfully to TDF/FTC/efavirenz and their next viral load 13 weeks later was below 40.

In the last case, the signs indicating HIV infection were so ambiguous that the study participant did not accept at first that he actually had HIV. This person tested HIV antibody positive at week 41 (his sixth injection appointment). At this point he was positive on a standard HIV antibody test but was negative for viral RNA and for a more specific antibody/antigen test. Retrospective testing revealed positive results for a qualitative viral RNA test at week 27, 14 weeks before his antibody test and shortly after his fourth injection.



Refers to the mouth, for example a medicine taken by mouth.

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.

retrospective study

A type of longitudinal study in which information is collected on what has previously happened to people - for example, by reviewing their medical notes or by interviewing them about past events. 

concentration (of a drug)

The level of a drug in the blood or other body fluid or tissue.

ribonucleic acid (RNA)

The chemical structure that carries genetic instructions for protein synthesis. Although DNA is the primary genetic material of cells, RNA is the genetic material for some viruses like HIV.


After diagnosis, there were two completely negative test panel results at weeks 44 and 55, interspersed with some that gave positive antibody/antigen tests and indeterminate confirmatory antibody tests. Viral RNA tests were negative between weeks 34 and 65, when one was positive, though on a quantitative test his viral load was still below 40.

The participant was reluctant to take ART as he was not convinced the results showed he had HIV. He did take a four-week course of a protease inhibitor-based regimen between weeks 47 and 51, which he regarded as PEP (emergency treatment to prevent HIV infection). His viral load stayed below 40 for the rest of the time he was in the trial, out to week 75, even though he received no more cabotegravir injections after his positive antibody result. He did not develop resistant virus.

Anecdotally, after he left the trial, his viral load eventually increased to over 1000 and he started a boosted darunavir-based regimen, which was fully suppressive.

Infections in people on TDF/FTC

There were 42 HIV infections in people randomised to TDF/FTC, but three of these were in people who turned out to have acute HIV at baseline. Two of these three individuals developed the M184V/I emtricitabine resistance mutation.

In all but 2 of the 39 who acquired HIV while taking TDF/FTC PrEP, the last drug level test taken before their positive test showed no or very low levels of drug. One, with moderate levels, had the M184V/I mutation too. The other participant was infected despite having levels indicative of daily dosing, but this person turned out to have the K65R resistance mutation to tenofovir. They also had three resistance mutations to non-nucleoside reverse transcriptase inhibitors (NNRTIs), which suggested they had acquired a PrEP-resistant virus, rather than developed the resistance while on PrEP.


Landovitz said that the occurrence of infection in four trial participants despite on-target cabotegravir concentrations was of concern and that further investigations of, for instance, how cabotegravir concentrations varied from one body compartment to another was ongoing. This was particularly important because the oral lead-in of cabotegravir pills was not going to be obligatory in the open-label extension of HPTN 083.

The evidence that injected cabotegravir could delay the detection of HIV infections was also of concern, he added. The open-label extension would use regular viral load as well as antibody tests to screen for HIV infections. The development of cabotegravir resistance mutations when both viral and drug levels were high showed that it was important to detect HIV infections and start ART as soon as possible.

On the other hand, no resistance had arisen in people with low-level cabotegravir in the post-injection ‘long tail’ phase, not even in the case of the participant who developed a very high viral load. This may indicate that it is safe to stop injectable PrEP without covering the ‘tail’ with oral PrEP, though we'd need more data before recommending this.


Marzinke M (presenter Landovitz RJ) Laboratory analysis of HIV infections in HPTN 083: injectable CAB for PrEP. Conference on Retroviruses and Opportunistic Infections, abstract 183, 2021.

Download the abstract eBook from the conference website.