Genotypic resistance algorithms interpret drug-resistance mutations, but are mainly developed for HIV-1 subtype B, meanwhile non-B subtypes cause 90% of worldwide infections. They include clade-specific amino acid at drug-resistance positions different than subtype B. This study explores: (i) the variability at resistance-related positions in 128 non-B and 226 B sequences from 354 treatment-naïve patients diagnosed in Spain (1999-2007); (ii) the discordances between five resistance interpretation algorithms (ANRS, Stanford, Rega, Geno2pheno, RIS); and (iii) the reliability of five subtyping tools (Stanford, Geno2pheno, Rega, NCBI, EuResist) for each HIV-1 variant. Primary drug-resistance prevalence was 13.6%, although higher in non-B vs. B subtypes (18.7% vs. 10.6%), due to a twofold higher NNRTI-resistance prevalence (15.7% vs. 7.6%). Most secondary PI-resistances, more frequent in non-B, were in fact clade-specific residues. Most sequences were interpreted as susceptible to all antiretrovirals by the five resistance algorithms, except for tipranavir by ANRS in non-B clades. Interalgorithm discordances were significantly higher in non-B variants for specific drugs. The agreement with phylogenetic analysis differed among subtyping tools testing non-B variants. We found a higher prevalence of NNRTI-resistance mutations in non-B subtypes. Certain algorithms overestimate the resistance in non-B subtypes due to natural patterns of mutations. Subtyping tools should be optimised for non-B variants.
Copyright 2009 Elsevier B.V. All rights reserved.