BHIVA: Current tests may miss a third of HIV resistance in treatment-naive

Standard HIV drug resistance tests may fail to detect HIV drug resistance in more than one-third of newly diagnosed patients, the 14th British HIV Association Conference heard last week. This is important clinically because if no more 0.4% of a person’s viral population is drug-resistant, this can triple the risk of treatment failure, the conference was told. Standard resistance tests can only detect resistance if 10-20% of a person’s viral population is drug-resistant.

Dr Jeffrey Johnson of the US Centers for Disease Control retested 205 samples of what was thought to be wild-type, non-resistant HIV from a group of people newly diagnosed with HIV between 2003 and 2005 in Los Angeles and Chicago, using hypersensitive resistance tests that could pick up resistant virus that comprised as little as 0.1% of the viral population .

The group was a very mixed population of newly-diagnosed people, with roughly a third each being white, Latino and black. Two-thirds were gay men, one in five were injecting drug users, and in the other 15% the HIV exposure risk was heterosexual sex.

The hypersensitive test found that 34 out of the 205 (17%) actually had resistance mutations. This would increase the proportion of people with transmitted drug resistance amongst the cohort from which the samples were drawn from the 20% picked up by standard tests to 33%.

In a similar retesting of 303 samples from people diagnosed between 1998 and 2005 who were already known to have at least one drug resistance mutation, hypersensitivity testing similarly increased the number of resistance mutations found by 60%.

The proportions of people with certain individual mutations were considerably elevated by hypersensitive resistance testing. This is because some mutations affect viral replication capacity more than others, so that resistance mutations that result in very unfit virus may exist at very low levels in people with them – until selective drug pressure brings them out.

For instance, while hypersensitive testing only found an additional 5% of people with the K103N NNRTI resistance mutation (which is known not to affect viral fitness), it doubled the number found to have the Y181C NNRTI mutation and more than quadrupled the number found to have the T215F thymidine analogue NRTI resistance mutation that (in conjunction with other mutations) confers resistance to AZT and d4T and to a lesser extent to most other nucleoside drugs.

It also picked up 20% more of the similar T215Y mutation, 23% more of the 3TC/FTC M184V mutation, 25% more of the L90M protease inhibitor resistance mutation, and 60% and 70% more of the other thymidine analogue resistance mutations M41L and K70R.

Using hypersensitive tests meant that 21 people out of the 303 (7%) were classified as resistant to a whole new class of drugs. Johnson commented that hypersensitive testing showed that in this population, 20-27% of patients with primary HIV resistance had multi-drug-resistant virus.

Did this matter clinically? Johnson asked. The answer was yes, most certainly; low-frequency resistance might be responsible for a high proportion of unexplained drug failures.

Johnson’s team re-tested patients participating in GSK-sponsored clinical trials of efavirenz, 3TC, abacavir and AZT. Samples were re-tested for low-frequency resistance to NNRTIs (K103N and Y181C) and to 3TC (M184V). Nine people out of 221 trial participants previously thought to have wild-type virus were found to have drug resistance. Of these, seven (78%) had experienced virological failure.

In another analysis of a sample of patients who had virologically failed treatment in drug trials with what appeared to be wild-type virus, ten per cent turned out to have the K103N NNRTI mutation at low frequency.

These studies allowed Johnson to calculate that all people with low-frequency mutations had eleven times the risk of virological failure compared with people with genuine wild-type virus. In people with K103N, if more than 0.9% of the viral population had the mutation, the risk of treatment failure was multiplied more than eight times; people with more than 0.4% resistant virus had three times the risk of failure; while people with more than 0.1% resistant virus had a 50% greater risk of failure.

Johnson pointed out that his were not the first studies to find that minority resistance was associated with failure. A previous study his team had conducted in 2005 had revealed that an additional 20% of women who had received single-dose nevirapine for the prevention of mother-to-child transmission turned out have resistance to the drug, in addition to the 40% picked up by standard resistance tests.

Similarly, Prof. John Mellors of Pittsburgh University had found, on retesting patients who had failed on efavirenz with apparently wild-type virus, that eight out of 12 patients in fact had minority resistance mutations, while Simen and Kozal, using a so called “ultra-deep-sequencing” technique, had found in a similar group of naïve patients that detecting patients with one to five per cent minority resistance variants doubled the total number with resistance to 25%, and that all patients with NNRTI resistance, whether majority or minority, experienced failure to this class of drugs.

A related question is whether standard tests can pick up resistance at lower viral loads than the ones they are designed for. If they could, this would be of considerable benefit, as one of the reasons doctors fail to order resistance tests for patients before switching them to other regimens is because of the time-lag necessary in waiting for a failing patient’s viral load to reach levels that can be resistance-tested (usually 1,000-2,000 copies/ml).

A survey of eight laboratories in the UK that do resistance testing found that seven regularly accepted requests for testing on samples with viral loads lower than 1,000 copies/ml, while two labs said they accepted requests for tests on ‘undetectable’ samples (viral load under 50 copies/ml).

Results at lower viral loads were highly variable. While labs were able to sequence (achieve a valid result) for 60-100% of samples with viral loads between 1,000 and 2,000 copies/ml, the success rate for viral loads between 50 and 100 copies/ml varied between zero and two-thirds.

Labs varied in the sensitivity and specificity of their techniques in a consistent way; the lab that was least sensitive at viral loads of over 1,000 copies/ml, with a 60% success rate where others claimed success in sequencing in between 80% and 100% of cases, was also the one that scored zero on attempts to sequence samples with viral loads between 50 and 100 copies/ml.

Conversely, the lab that scored 100% for samples with viral loads between 1,000 and 2,000 copies/ml also managed to sequence two-thirds of samples with viral loads between 50 and 100 copies/ml and even managed to sequence 38% of samples with viral loads under 50 copies/ml.

However the more sensitive techniques it was using paid the price with a lack of specificity, and this lab commented that the gene sequences it did obtain could have been picking up proviral DNA released from the nuclei of quiescent cells – which might give an inaccurate result as this may not represent the variant of HIV that is currently responsible for producing active virus.

This is the reason hypersensitive resistance tests are not carried out routinely; there is too much risk of contamination and false-positive results unless extraordinary care and therefore expense is taken to avoid it.

The researchers comment: “In conclusion, it is technically possible to carry out resistance testing at a viral load lower than currently recommended, though such tests require the most stringent quality control.”