Background:  Drug susceptibility scoring systems can predict ART outcomes in individual patients. A modified phenotypic susceptibility score (PSS) and incorporation of replicative capacity (RC) data may allow improved predictive power of both virological and immunological outcomes.

Methods:  Baseline HIV-1 isolates from patients in Argenta were tested for phenotypic drug susceptibility and RC in a recombinant assay. Lower and upper clinical susceptibility cut-offs were determined for each drug using established criteria and correlates from published datasets to define the lower clinical susceptibility cut-offs (FC above which viral load response starts to diminish) and the upper clinical susceptibility cut-offs (where viral load response is completely lost). The ability of an incremental PSS (iPSS, FC values intermediate between clinical susceptibility cut-offs classified as 0.5 susceptibility) and a continuous PSS (cPSS, intermediate FC values given a continuous susceptibility value by linear interpolation) and of the traditional, dichotomous PSS (dPSS) to predict viral load and CD4 outcomes were compared. Correlation of each PSS with log change from baseline viral load and change from baseline CD4 for up to 36 months was analyzed by multiple linear regression models.

Results:  We analyzed 139 patients:  72% male, 33% injecting drug users, median age 38 years, CD4 284 cells/mL, viral load =4.31 log₁₀ copies/mL, 37% CDC class C, median prior HAART regimen = 2 (range 1 to 5). Baseline virus was susceptible to a median 10 of 17 drugs (0 to 17), median RC was 59% (IQR 31 to 89%). The first salvage regimen after baseline had a mean dPSS of 1.66 (SD 1.07), a mean iPSS of 1.92 (1.01), and a mean cPSS of 2.00 (1.04) (each paired comparison p <0.001). After adjusting for baseline viral load, dPSS predicted changes from baseline viral load at months 3 (R = –0.27, p = 0.004), 6 (R = –0.28, p = 0.003), and 12 (R = –0.27, p = 0.012); iPSS predicted viral load changes at month 3 (R = –0.33, p <0.001), 6 (R = –0.34, p <0.001), 9 (R = –0.20, p = 0.05), and 12 (R = –0.26, p = 0.016); cPSS also yielded more accurate predictions than dPSS (R = –0.32 at 3 and 6 months), but was not a more accurate predictor than iPSS. iPSS, but not dPSS and cPSS, also predicted changes from baseline CD4 counts at month 3 (R = 0.21, p = 0.029) and 12 (R = 0.22, p = 0.033). After adjusting for baseline RC, all 3 susceptibility scores were more strongly and significantly associated with CD4 changes at month 3 and 12.

Conclusions:  Both iPSS and cPSS performed comparably better than dPSS in predicting treatment outcomes. Furthermore, RC may enhance prediction of the CD4 outcomes by the PSS. These observations merit further study.