W. David Hardy, MD Disclosures

Barcelona, Spain; Thursday, July 11, 2002 -- Andrew Zolopa and colleagues[1] today reported the findings of the GUESS study, which evaluated the interpretation of HIV-1 genotypic resistance tests performed on clinical specimens. The study tested the ability of an international panel of experts to predict the results of a phenotypic resistance assay performed on the same sample that was genotyped. Agreement between the experts with regard to predicting antiretroviral drug activity and making specific treatment recommendations also were evaluated.

A total of 50 randomly selected genotypes with matching phenotypes were selected for analysis from the VIRCO database. All viral isolates had to be resistant to at lease 1 drug (ie, > 4-fold increase in IC50 compared with wild-type HIV-1). The 12 experts were asked to predict each isolate's susceptibility to 16 antiretroviral agents by assigning it to a category of fold-change (FC) in IC50, namely < 2.5, 2.5 to < 4, 4 to < 7, 7 to < 10, 10 to < 20, and > 20. They were also asked to rate the expected degree of activity of each drug on a 6-point scale, and to state whether it would be recommended for use.

The experts' accuracy in predicting susceptibility to each antiretroviral varied by class. For nucleoside reverse transcriptase inhibitors (NRTIs), accuracy of predicting phenotype from genotype varied from 25% for abacavir susceptibility to 74% for lamivudine susceptibility. For protease inhibitors (PIs), it varied from 26% for nelfinavir to 40% for lopinavir/ritonavir. However, experts were able to correctly predict nonnucleoside reverse transcriptase inhibitor (NNRTI) susceptibility about two thirds of the time for all 3 NNRTI agents. Of note, the median difference between predicted FC and assayed FC category was found to be 0 (meaning there was no difference) for all drugs except zidovudine, stavudine, didanosine, zalcitabine, and abacavir; the median difference was +1 category for all 5 of these agents. Correlation of agreement (rho) among the experts for predicting phenotypes in each class ranged from 0.23 for tenofovir to 0.84 for lamivudine; from 0.89 for efavirenz to 0.93 for nevirapine; and from 0.68 for amprenavir to 0.79 each for indinavir and ritonavir.

Predictions for drug activity mirrored the same ranges as for predicting phenotype. Consensus among the experts on treatment recommendations for each antiretroviral agent ranged from 62% for didanosine to 90% for lamivudine, with a median of 79%.

In summary, although the experts were able to predict the phenotype fairly well, they frequently overestimated the FC (ie, they underestimated the susceptibility) of zidovudine, stavudine, didanosine, zalcitabine, and abacavir. Of note, while there was generally a high level of agreement regarding drug activity and treatment recommendations, there was significant variability for individual antiretroviral agents.

There are several important conclusions to draw from this study. First, it demonstrates that even among the experts, the "science" of genotypic interpretation remains variable and imprecise. Even for the NRTIs, the oldest and most heavily studied class of agents, there remains significant difficulty in reaching consensus for genotypic interpretation. Second, genotypic interpretation is easier and more consistent for agents such as the NNRTIs, where at least 2 mutations convey class-wide, high-level resistance. Third, interpretation of genotypic resistance to the PIs continues to evolve, as the list of available agents and the number of mutations increases, and their specific effects grow more complex to interpret. The use of boosted PIs may make genotyping less important by increasing the ratio of drug level to IC50, allowing the drugs to retain activity against virus with some degree of loss of susceptibility. Overall, this study appears to reinforce the notion that resistance testing will get much more complex, rather than less so, as we learn more.

Reference

1. Zolopa AR, Lazzeroni LC, Rinehart A, Kuritzkes, D. The GUESS study. An international evaluation of the accuracy, precision and consistency of expert interpretation of HIV-1 genotype interpretation. Program and abstracts of the XIV International AIDS Conference; July 7-12, 2002; Barcelona, Spain. Abstract ThOrB1385.

Copyright © 2002 Medscape http://www.medscape.com http://www.aids2002.com