February 8, 2006
When selecting an initial antiretroviral regimen, many providers use baseline resistance testing to help them identify which drugs are likely to be less effective. But conventional genotyping of circulating virus may underestimate the presence of resistance.1 Resistance mutations acquired during antiretroviral treatment tend to disappear gradually when drug pressure is withdrawn, as resistant virus is overgrown by fitter, archived wild-type clones. By contrast, when someone is primarily infected with drug-resistant virus, reversion to wild type must occur by mutation, so resistance mutations may persist for years.2
In someone who is treatment naive, detected mutations can usually be attributed to resistant virus transmitted at the time of primary infection. Because these pre-existing mutations will impair the response to antiretroviral therapy, baseline resistance testing is now recommended in areas where there is a high prevalence of primary or transmitted drug resistance, and is often performed on patients with chronic infection.
Unfortunately, conventional genotyping may not reliably detect mutations present in mixtures at less than 10% to 20% of circulating plasma virus. Previous studies, including some by this same group, have demonstrated that in women treated with single-dose nevirapine (NVP, Viramune) to prevent perinatal HIV transmission, newer, more sensitive resistance testing technologies may greatly increase the proportion of patients in whom mutations are identified.1
In this study, Jeffrey Johnson, of the U.S. Centers for Disease Control and Prevention (CDC), and colleagues asked: To what extent does conventional genotyping underestimate the presence of transmitted resistance? And do any newly identified mutations suggest resistance to additional drug classes?
The investigators, a team of researchers from the CDC and its Canadian counterpart, developed a sensitive real-time PCR assay (R-T PCR) using multiple mutation-specific primers to identify viral clones carrying the reverse transcriptase mutations M41L, K70R, K103N and M184V, as well as the protease mutation L90M. They validated their assay on a panel of 378 diluted samples of viruses cloned from clinical specimens, 301 of which carried only one of these 5 mutations, and the remaining 77 of which had wild type. Real-time PCR reliably detected mutations down to proportions of 0.2% to 0.5% of circulating virus -- a rate 20 to 50 times lower than the limit of commercial assays -- with sensitivities of 96% to 98% and a specificity of 100%.
This real-time PCR assay was then applied to 277 clinical specimens obtained from treatment-naive persons who had primary drug resistance identified by conventional genotyping; these individuals were participants in the CDC's sentinel surveillance program between 1997 and 2004, and in Canada's surveillance program between 1999 and 2003. The investigators state that "about 14%" of the study population represented recent infections, as determined by detuned ELISA, but no more precise breakdown is available.
Real-Time PCR Spots Many Previously Undetected Mutations
Key results are shown in the table below. For example, real-time PCR detected M41L in 9 (8.9%) of the 101 samples that had been deemed "negative" for that mutation by conventional genotyping, over and above the 19% of samples in which conventional genotyping had already detected M41L. The largest number of previously undetected mutations were found at K70R, for which the 15 picked up by real-time PCR more than doubled the proportion detected by conventional genotyping. The lowest rate of new detection was for K103N, in which real-time PCR added only 1% to the 20% found by conventional genotyping.
|Drug Resistance Mutation Frequency by Real-Time PCR Screening|
|Mutation||Samples negative by genotyping but positive by R-T PCR||Cases in which mutations are newly picked up by R-T PCR||Estimated frequency by conventional genotype vs. by R-T PCR||% Increase in frequency attributed to using R-T PCR vs. conventional genotype|
|L90M||4/222||1.8%||8 vs. 10%||21%|
|M41L||9/101||8.9%||19 vs. 26%||38%|
|K70R||15/252||6.0%||9 vs. 14%||60%|
|M184V||6/260||2.3%||9 vs. 11%||23%|
|K103N||2/202||1.0%||20 vs. 21%||4%|
Picking up new mutations by real-time PCR was indeed more likely in patients with recent HIV infection than in chronically infected patients: Although recent infection represented only 14% of specimens, seven (47%) of the 15 new class mutations detected were in this group, compared to only four in patients with known chronic infection and four for whom the duration of infection was unknown.
Finally, as expected, there was a rough trend in the data for mutations that substantially impair viral fitness (like M184V) to be detected at a higher rate by real-time PCR than mutations without much fitness impact (like K103N). This finding suggests that there is more pressure on fitness-impairing mutations to disappear from the majority circulating viral pool that is accessible to conventional genotyping.
Using Real-Time PCR to Detect New Class Mutations
In 15 (5%) of the 277 samples tested, real-time PCR demonstrated resistance mutations to a new class of drugs. As a result, the overall prevalence of drug resistance increased within the study population: While conventional genotype showed multidrug-resistant HIV (i.e., likely resistance to two or more classes) in 17% of the samples, when real-time PCR was used, the proportion of samples estimated to have multidrug-resistant HIV increased to 22%. Similarly, the proportion with three-class resistance increased from 1.4% to 2.2% (representing two new patients). Another 10 samples (4%) had mutations suggesting resistance to new drugs within classes to which they already showed some resistance.
Overall, these data mean that a physician's choice of antiretrovirals could have been affected by the use of real-time PCR in up to 9% of patient samples, although the study authors do not specifically calculate rates at which therapy would be likely to be changed. For example, finding a new protease mutation in a patient whom you'd already planned to treat with an NRTI + NNRTI regimen would not lead you to change that decision.
One limitation of this study, however, was that only a small number of mutations were tested for. So it is possible that many other important mutations would have been detected by a wider-ranging real-time PCR, thus increasing the proportion for whom this test would be therapy changing. Also, the study was limited by the fact that it tested only patients who were already known to have significant drug-resistance mutations. It's unknown what the frequency would be of finding new mutations with real-time PCR if the practice were applied to the entire pool of treatment-naive patients.
The Bottom Line
When conducting resistance tests on treatment-naive patients, "the harder you look, the more you find."3 More sensitive testing technologies will always reveal higher rates of mutations that exist in minority viral populations, and yes, sometimes these newly discovered mutations would change therapeutic choices. But the absolute number of patients for whom such changes would occur are small. And we do not know the clinical impact of mutations detected at very low levels -- does, for example, finding K103N in less than 1% of circulating virus mean that NNRTI-based therapy is more likely to fail? Clinical outcome studies are needed to answer this important question.
In addition, the technologically sophisticated, expensive real-time PCR testing used in this study is not available in clinical settings. As a result, clinicians are unlikely to have the choice to look any harder than we already do for low-frequency mutations.
Clinicians thus should always assume that there may be more resistance present than is detected by baseline genotyping, particularly in patients infected with HIV during the modern antiretroviral therapy era but who have not been infected long enough (i.e., more than three years) to have lost some key mutations. In those circumstances, it would be prudent to avoid drugs for which mutations detected by genotyping suggest partial but not complete resistance, and to err on the side of selecting drugs with higher genetic barriers to resistance.