ICAAC 2003 coverage
HIV resistance patterns are shifting, study shows
Regression model could aid clinicians
Investigators studying genotypes of 64,000 clinical samples submitted for genotyping to Virco based in Mechelen, Belgium, have found that HIV-1 drug resistance remains extensive, but that trends have shifted in recent years.
With about six years of genotyping data, researchers analyzed predicted phenotypes, based on a large database of matching genotypes and previously determined phenotypes, says Lee Bacheler, PhD, senior director of clinical virology at Virco Lab Inc. in Durham, NC, a secondary site of the corporation.
The study was presented at the 43rd annual Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), held recently in Chicago.
"There were almost 30,000 samples with both genotype and drug susceptibility phenotype that are the basis of the virtual phenotype prediction," Bacheler says. "So it’s that prediction that we used on this larger set of 64,000 genotypes that we analyzed between 1998 and 2003."
Generally, the findings are that resistance is extensive and has remained extensive during the entire period, but that the kinds of resistance observed has been changing over time, she explains. "We’ve seen a very dramatic increase in the number of samples with K65-R mutation in reverse transcriptase," Bacheler says. "We can speculate that the rise in K65-R may be associated with increased use of tenofovir, which can select for that mutation."
However, the study was limited to routine clinical samples, and investigators did not have available any treatment data about particular patients, she notes. "A similar story is an increase in another mutation, 225-H, which is associated with virologic failure on Sustiva," Bacheler says.
When researchers looked at a 2002 snapshot of a phenotype analysis of samples, they found that 60% had resistance to nucleosides, 45% to non-nucleosides, 36% to protease inhibitors; and 30% had no detectable phenotypic resistance, she says.
Comparing this 2002 snapshot of resistance to previous years, the study found that some patterns of resistance increased while others decreased. Here were the key observations, Bacheler explains:
- Isolates that have resistance to all three classes of drugs are declining.
- Isolates that have resistance to only nucleosides and protease inhibitors are declining.
- Isolates that have resistance to one or more nucleosides and one or more non-nucleosides are increasing.
- Samples sent to Virco Lab for testing that have no predicted resistance are increasing.
"We speculated that treatment practices are changing over this time period, and that may change the kind of resistance we’re seeing when a regimen fails," she says.
The study concludes that the change in resistance patterns also may be due to increased use of resistance testing and evolution of the patient population.1
In another study presented at ICAAC, Virco Lab investigators discussed a new methodology for predicting quantitative level of drug resistance from genotyping.
"It’s based on linear regression modeling, and the data set we use to develop the model is the same large data set of viral genotype and matching drug susceptibility phenotype," Bacheler says. "So basically, resistance is modeled as the contributions from individual mutations which have different effects on resistance."
Specifically, the model was tested through models that had been built for ritonavir resistance and was based on 28,000 matched viral genotypes and ritonavir drug susceptibility phenotypes, she says.
"The model identified 53 different mutations in the protease gene and 96 pairs of mutations that either had a positive effect or a negative effect on susceptibility to ritonavir," Bacheler explains. "And among those list of mutations are these mutations that people have heard of before."
For instance, of the 22 mutations associated with ritonavir resistance as listed by the International AIDS Society-USA (IAS), the linear model picked up 20 out of the 22, she says.
"It also picked up additional mutations that may affect resistance to ritonavir but are not on the IAS list," Bacheler adds. "And it also picked up other mutations that had not been previously identified as affecting resistance to ritonavir, and we’re further evaluating whether the model is telling us something new."
The study concludes linear regression modeling holds promise as a technique for analyzing HIV-1 drug resistance and identifying resistance-associated mutations.2
The model will be further evaluated and tested before it can be considered for use as a commercial product, she says.
But if it does become available, it would be a cost-effective and relatively easy way for clinicians to quantify how resistant their patients are to individual drugs, Bacheler notes.
"Our research into this linear regression model is an effort to find an even more accurate way to make the prediction, even for complex viral genotypes," she says.
Virco Lab investigators are continuing their research and have made it a priority to understand what levels of resistance are going to predict different levels of response, Bacheler adds.
1. Bacheler L, Vermeiren H, Mckenna P, et al. Trends in genotypic and phenotypic HIV-1 drug resistance among recent clinical samples submitted for resistance testing. Poster presented at 43rd Interscience Conference on Antimicrobial Agents and Chemotherapy. Chicago; Sept. 14-17, 2003. H-917.
2. Van Marck H, Van Den Bulcke T, Van Houtte M, et al. Quantitative prediction of HIV drug susceptibility from viral genotype through linear regression modeling. Poster presented at 43rd Interscience Conference on Antimicrobial Agents and Chemotherapy. Chicago; Sept. 14-17, 2003. H-918.