Background:  The effect of delavirdine (DLV) hypersusceptibility has not been demonstrated and appropriate cut-points that best define hypersusceptibilit are needed to make optimal use of phenotype assays. 

Methods:  Phenotypic drug susceptibility (fold change in IC₅₀, ViroLogic) was determined for 96 randomly selected NNRTI-naïve patients who received DLV during ACTG 359. Indinavir-experienced patients (= 6 months) were randomized to SQV and either RTV or NFV together with DLV, adefovir, or both. DLV hypersusceptibility was analyzed as a dichotomous score (fold change cut-point at 0.4) and a continuous phenotypic sensitivity score (CPSS:  0 for fold change values >2.5, intermediate for fold change values 0.1 to 2.5, and 1 for fold change values <0.1). A CPSS was used to account for susceptibility to other regimen agents. Regression methods were used to model baseline DLV hypersusceptibilit and week 4 and 16 continuous (log₁₀ value) or categorical (= 500 copies/mL) virologic outcome after accounting for important covariates. Regression and CART were used to define the best fold change cut-point for DLV hypersusceptibility.

Results:  The median baseline log₁₀ HIV RNA was 4.5 and CD4 was 203. DLV hypersusceptibility virus was present in 36% and was associated with resistance to NRTI. Overall, 33% of subjects achieved HIV RNA = 500 at week 16. Both continuous and categorical week 16 HIV RNA responses were significantly associated with baseline HIV RNA and CPSS, but not DLV hypersusceptibility. DLV hypersusceptibility was significantly associated with continuous and categorical week 4 HIV RNA responses after accounting for CPSS and baseline viral load. For example, patients with hypersusceptibilit virus were 3.3 times more likely to achieve HIV RNA = 500 at week 4 than those without hypersusceptibilit (adjusted p = 0.03). Linear and logistic regression models explored hypersusceptibility cut-points from 0.2 to 1.0 using the week 4 virologic response. Using a binary endpoint, a cut-point of 0.3 was most predictive while a cut-point of 0.4 was the best predictor of continuous outcome. Interestingly, the CART analysis identified DLV fold change <0.44 as a second predictor of week 4 viral load (HIV RNA was the first predictor). The DLV hypersusceptibility cut-point of 0.44 separated patients with a mean week 4 viral load of 2.5 and 3.2.

Conclusions:  DLV hypersusceptibility was an independent predictor of week 4, but not later, virologic responses. This might have been due to lack of NRTI pressure and the early rebound seen in this study with a low response rate. Several analytic methods identified a hypersusceptibility fold change cut-point of 0.3 to 0.4. In refining phenotypic cut-points, early virologic responses (not confounded by rebound) may be better metrics than later responses.

Keywords: phenotype; hypersusceptibility; virologic response