Drug resistance accumulates fast in people with low but detectable viral loads, Kenyan study finds

Finding poses challenge to current global treatment guidelines

A study measuring HIV drug resistance at two timepoints in Kenyan patients on second-line, protease inhibitor-based regimens has found very high levels of drug resistance in people with detectable viral loads.

Unexpectedly, it found higher rates of resistance in people with what it called ‘low level viraemia’ (defined as viral loads from 40 to 1000 copies/ml) than it did in people with viral loads over 1000 copies/ml. It also found continued accumulation of new HIV resistance mutations in these people.

The study suggests that current treatment failure guidelines may need to be revised – and poses the question of how to resource them. 


The study is one of the first to measure drug resistance in people on second-line regimens in a low-income setting, and the first to measure resistance at two time-points in the same individuals.


second-line treatment

The second preferred therapy for a particular condition, used after first-line treatment fails or if a person cannot tolerate first-line drugs.

drug resistance

A drug-resistant HIV strain is one which is less susceptible to the effects of one or more anti-HIV drugs because of an accumulation of HIV mutations in its genotype. Resistance can be the result of a poor adherence to treatment or of transmission of an already resistant virus.

first-line therapy

The regimen used when starting treatment for the first time.

boosting agent

Booster drugs are used to ‘boost’ the effects of protease inhibitors and some other antiretrovirals. Adding a small dose of a booster drug to an antiretroviral makes the liver break down the primary drug more slowly, which means that it stays in the body for longer times or at higher levels. Without the boosting agent, the prescribed dose of the primary drug would be ineffective.

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.

The study was conducted at the Moi Teaching and Referral Hospital (MTRH) clinic in the town of Eldoret, by AMPATH, a partnership between the hospital and other Kenyan clinics with the Kenyan government and American Universities.

At the MTRH clinic 18,282 adults with HIV are currently in care, of whom 76% are on antiretroviral therapy (ART). Ninety-one per cent of these are on first-line ART. 

When the study began in June 2011, first-line ART was most often based on a non-nucleoside reverse transcriptase inhibitor (NNRTI) (nevirapine) with the support of two of the nucleoside reverse transcriptase inhibitors (NRTIs) zidovudine, stavudine and lamivudine. Since 2013, people on stavudine have been switched to tenofovir, but many others remain on zidovudine.

The other 9% were on second-line therapy, which is generally based on the boosted protease inhibitor lopinavir (Aluvia, Kaletra) with boosted atazanavir also sometimes available. It was generally supported by two NRTIs from abacavir, didanosine, lamivudine, zidovudine and tenofovir. Although patients get an annual CD4 count, scarce resources mean that viral load and resistance tests are still not routinely used, so CD4 count or clinical progression are used to judge if people are failing treatment. 

At the time of the study no third-line regimen was on offer though regimens based on the second-line NNRTIs (rilpivirine and etravirine), and on boosted darunavir have been available since 2015, with dolutegravir available in special cases but still some way from routine provision. The AMPATH investigators wanted to determine if resistance acquired during failing first- and second-line regimens would rule out some of these options, and also to find out if allowing people to spend some time with low, but detectable, viral loads would be tolerable and not give rise to more resistance. 

The results indicate the opposite – that allowing people to spend time with low-level viraemia might lead to more drug resistance.

The study

The participants were recruited from the 1242 adults who had been taking second-line therapy for at least 24 weeks (5.5 months) and who had previously been on first-line therapy for at least six months. The researchers gave them two viral load tests and two resistance tests, an average of 55 days apart.

A total of 394 people were enrolled, which was under the target of 438, and may reflect the difficulty of patients attending regularly enough for a clinical trial. 

Their average age was 42, 60% were women, and their average CD4 count at enrolment was 282 cells/mm3. People had been on second-line ART for an average of 1.9 years and on first-line before that for 2.9 years. 

Boosted lopinavir was used in 94% of second-line regimens, with the most common accompanying NRTIs being abacavir and didanosine. This regimen, plus or minus lamivudine, was used in 52% of people. Tenofovir was used in 19%. The most common first-line regimen had been nevirapine, zidovudine or stavudine, and lamivudine.

Of the 394 people in the study, 203 (52%) had a viral load under 40 copies/ml, 109 (28%) had a viral load between 40 and 1000 copies/ml, and 82 (21%) had a viral load over 1000 copies/ml. 

People with viral loads over 1000 copies/ml (defined as viral failure or VF) were more likely than those who were virally suppressed to be younger, to be taking tuberculosis treatment, to have spent less time on second-line therapy, and to have had more days of consecutive missed therapy (treatment interruption) on first line. Female participants were more likely to have been pregnant. 

People with viral loads between 40 and 1000 copies/ml (defined as low-level viraemia or LLV) were similar to those who were virally suppressed except that, if women, they were less likely to have been pregnant, but more likely to have first taken ART to prevent mother-to-child transmission. 

Drug resistance at the first test

Of the 191 people with detectable viral load, 105 (55%) had viable resistance tests done, 37% of them with viral loads over 1000 copies/ml and 18% (35 people) with viral loads under 1000 copies/ml. 

Of the people with resistance test results, 67% had HIV resistance to one or more NRTI drugs, 74% to NNRTI drugs and 55% to both classes. Resistance to just one class was quite uncommon, with only 3% having resistance to NRTI drugs alone. Nine per cent had resistance to protease inhibitors and 7% resistance to all three drug classes.

People with LLV were no more likely to have drug resistance in general than other people, but seemed more likely than those with VF to have combined NRTI + NNRTI resistance (66% vs 50%), although was not statistically significant.

What was significant was that people with LLV had a larger average number of individual drug resistance mutations (4.57) than people with VF (3.01). They also had a larger number of individual mutations to NRTI drugs (3.03 vs 1.54) but not to NNRTIs (1.37 to 1.23).

People with LLV appeared more likely to have resistance to the drugs they were currently using than people with VF (71% vs 57%) and to any of the future drugs they could use (69% vs 59%), but these differences did not reach statistical significance.

New drug resistance at the second test

Of more concern were the results of the second resistance test, an average of two months later. Only 48 individuals or 46% of those with initial resistance tests had second resistance tests taken: 17 (49% of those with a first resistance test) with LLV and 31 (44%) with VF.

Nineteen people altogether had new resistance mutations at the second visit, eight with LLV and eleven with VF. People with LLV had acquired significantly more additional drug mutations than people with VF (1.41 vs 0.65 on average), and more individual NRTI mutations (0.76 vs 0.26).

The proportion of people with LLV who had resistance to any of the current drugs they were taking had increased, from 71% to 82%. But in people with VF it had decreased, from 57% to 48%. This was a statistically significant difference between the two groups. Resistance to future drugs did not change, but two people with LLV acquired PI mutations between the first and second viral load tests. 

There were also people who lost mutations between visits; 28 people had a drug resistance mutation at the first test that wasn’t there at the second. But the problem with drug mutations is that they never completely disappear. Because drug-resistant viruses are often less fit, they become an ‘archived’ minority but will reappear again if future treatments are based on the drugs they are resistant to.

Implications and conclusions

We already know, from two previous studies conducted in Kenya – one in people on first-line therapy and one in people on second-line – that keeping people on failing drug regimens is giving rise to high levels of HIV drug resistance that may considerably limit future treatment options.

What this study adds is to make it clear that tolerating low, but detectable viral loads in the hope of prolonging the time people can stay on second-line therapy is not an option. The question it does not answer is what the threshold for switching therapy should be.

It is less puzzling than it might first appear that people with low, but detectable, viral loads acquire more drug resistance than people failing treatment. Drug resistance happens if people take some, but not enough, medication. This creates the conditions in which viruses with resistance mutations can out-compete ones that don’t have them. 

People with complete viral failure are more likely to have taken little or none of their ART, so their virus is not exposed to the drugs and does not so often become resistant. 

The study also found that resistance mutations to the NNRTI drugs continued to accumulate even in people on second-line therapy who were not actually taking NNRTIs; some of these mutations were to second-generation NNRTI drugs such as rilpivirine, ruling them out as future options. Their continued accumulation in the absence of drug pressure implies that NNRTI mutations – which do not lead to reduced replicative capacity in HIV and may even enhance it – may increase in any population where a high proportion of people have low, but not undetectable, viral loads.

The big question is how to afford routine viral load monitoring in low-income countries in a world in which global funding to treat HIV infection is shrinking.


Kantor R et al. HIV second-line failure and drug resistance at high- and low-level viremia in western Kenya. AIDS, e-pub ahead of print. doi: 10.1097/QAD.0000000000001964. August 2018. See abstract here.