ABSTRACT:
The complexity of the viral response to long-term therapy renders prediction of clinical response from genotype data difficult. Approaches based on the number of mutations at primary resistance-associated sites do not explain the variation in long-term response. To determine whether amino acid identity in baseline sequences influenced response to subsequent therapy received during ACTG 241, a trial of zidovudine/didanosine versus zidovudine/didanosine/nevirapine, we have used multiple regression techniques on all variable sites to analyze virological response. Amino acid sequences in the RT domain from baseline isolates were obtained from 55 patients. Twenty amino acid sites in RT showed a variant in more than five individuals.
Univariate regression of amino acid variation on virological response at week 8 [response=log10 (baseline plasma viral RNA/week 8 plasma viral RNA)] identified position 215 (r2=0.197/ P<0.001). Stepwise multiple regression to incorporate additional sites included 44 and 215 with an r2 of 0.291 (P<0.0001). In contrast, virological response at week 48 was not strongly predicted by amino acid 215: r2 for position 214 was highest (0.136; P<0.01). Stepwise regression identified a model incorporating baseline sequence data at 214, 215 and two others: 60 and 202 (r2=0.35; P<0.0002). Mutants at these three positions were associated with better virological response. V60I is a polymorphism which is particularly common in some non-B subtypes: in ACTG 241 patients, all 60I variants were found with 215F/Y, but their median virological response after 48 weeks was threefold better than for 60V 215F/Y individuals.
The mechanism underlying this effect requires investigation. From these results it is clear that a more thorough understanding of the genotype basis of resistance to antiretrovirals may require both the inclusion of data on previously unidentified sites, and the adoption of appropriate methods of analysis.
