Evolution of Genotypic Resistance Algorithms and Their Impact on the Interpretation of Clinical Trials: An OPTIMA Trial Substudy

Abstract

Purpose: The outdated rules of older HIV genotypic resistance algorithms can affect virologic responses. This study was designed to determine how often these incorrect resistance interpretations affect analyses of long–term clinical trials, antiretroviral (ARV) choices, and HIV disease progression rates. Method: Baseline VIRCO virtual phenotypes (VVP) from patients screened in 2001–2002 for OPTIMA were compared to 2005 Stanford HIV resistance database algorithm (HIVDB–10/05, version 4.1.4) interpretations of the HIV–1 pol sequences. Drugs were called discordant if resistant by one algorithm and sensitive by the other. Results: Of 2,341 drug comparisons, 501 (21.4%) were discordant, affecting 140 (86.4%) of 162 screened patients. NRTI/NtRTIs were more discordant than NNRTIs and PIs (38.6% vs. 4.3% vs. 12.8%; p < .0001). Sixty-nine (53%) patients were placed on ≥2 drugs reported as sensitive by VVP but resistant by HIVDB–10/05; they had higher than expected rates of disease progression and a similar time to first event or death as patients on ARVs classified as resistant by both algorithms (p = .61). Conclusions: Underestimation of drug resistance by older genotypic algorithms resulted in using ARVs incorrectly thought to be sensitive and in higher than expected rates of HIV disease progression. The use of older genotypes to interpret long–term clinical trials should account for this underestimation, because results may be different if viral sequences are interpreted with newer algorithms.