Two different patterns of mutations are involved in the genotypic resistance score for atazanavir boosted versus unboosted by ritonavir in multiple failing patients

The protease inhibitor atazanavir (ATV) can be used either boosted by ritonavir (ATV300/r) or unboosted (ATV400). To date, however, genotypic resistance scores (GRSs) have been developed only for boosted-ATV. We have determined GRS associated with virologic response (VR) for both ATV300/r and ATV400 in highly pre-treated HIV-1 infected patients. We analyzed the results of genotypic tests available 0-3 months before the initiation of an ATV-containing regimen in 159 patients with HIV-RNA a parts per thousand yen 500 copies/ml (ATV300/r group: 74; ATV400 group: 85) who were enrolled in the CARe study through an Early Access Program. The impact of baseline protease mutations on VR (a parts per thousand yen 1 log(10)copies/ml HIV-RNA decrease at 12-24 weeks) was analyzed using Fisher's exact test. Mutated protease amino acid positions (MPP) with p < 0.20 were retained for further analysis. The GRSs were determined by a step-by-step analysis using the chi(2) test for trend. The GRSs for ATV300/r and ATV400 revealed differing sets of mutations. For ATV300/r, 12 MPPs (10C/I/V + 32I + 34Q + 46I/L + 53L + 54A/M/V + 82A/F/I/T + 84V + 90M - 15E/G/L/V - 69K/M/N/Q/R/T/Y - 72M/ T/V; p = 1.38 x 10(-9)) were the most strongly associated with VR (VR: 100%, 78.3%, 83.3%, 75% and 0% of patients with a score of -2/-1, 0, 1, 2, and a parts per thousand yen 3, respectively); the last three MPPs (I15/H69/I72) were associated with a better VR. For ATV400, nine MPPs (16E + 20I/M/R/T/V + 32I + 33F/I/V + 53L/Y + 64L/M/ V + 71I/T/V + 85V + 93L/M; p = 9.42 x 10(-8)) were most strongly associated with VR (VR: 83.3%, 66.7%, 5.9%, 0% of patients with 0, 1/2, 3, and a parts per thousand yen 4 MPP, respectively). Differences between GRSs for ATV300/r and ATV400 may be due to different ATV drug levels (boosted vs unboosted), favoring different pathways of escape from antiviral pressure. Both GRSs were independent predictors of response in a multivariable logistic regression model. Nevertheless, cross-validation of these GRSs on different patient databases is required before their implementation in clinical practice.