Abstract Body

Background: HIV resistance testing in HAART patients experiencing low-level viremia (LLV, 50-999 HIV-RNA copies/mL) predicts the risk of subsequent virologic failure (VF, ≥1000 copies/ml). Suboptimal plasma drug levels can arise as a result of poor adherence and/or pharmacokinetic issues. Here, our aim was to evaluate whether retrospective analysis of plasma levels also provides insight into patient outcomes after experiencing LLV.

Methods: The first documented LLV episode of 2176 patients was analyzed. A total of 328 consenting patients with drug levels, genotypic resistance data and with follow-up clinical data while on constant therapy available were eligible. Untimed plasma drug levels (UDL) of PIs and NNRTIs in the sample corresponding to the first LLV episode were measured by HPLC-tandem MS. Drug levels were categorized as ‘therapeutic’ or ‘suboptimal’ based on target trough concentrations from DHHS guidelines. Resistance was assessed using the Stanford algorithm (GSS, corresponding to the number of ‘active’ drugs prescribed). Time to VF after LLV was evaluated by Kaplan–Meier analysis and Cox proportional hazards regression.

Results: 78 of 328 patients (24%) had suboptimal drug levels at LLV, compared with 63 (19%) having GSS<3. Both suboptimal UDL and GSS<3 independently increased the risk of future pVL >1000. Within a year, 56/78 (72%) patients with suboptimal UDL had failed, compared to 45/63 (71%) with GSS <3 and to 103/206 (50%) with both optimal GSS and UDL. Of those with suboptimal UDL, 43/78 (55%) had undetectable levels of PI/NNRTI, with most (81%) failing by one year. Only 18 patients had both suboptimal UDL and GSS<3.
In the adjusted multivariable model, variables associated with VF were suboptimal UDL (Adjusted Hazard Ratio, AHR 2.9, 95%CI 2.0-4.2, p<0.001), female gender (AHR 1.9, 95%CI 1.3-2.9, p=0.001), GSS<3 (AHR 1.5, 95%CI 0.9-2.3, p=0.098) and being HAART-naïve (AHR 1.7, 95%CI 1.1-2.7, p=0.031). When UDL and GSS categories were combined, the lowest VF was found for optimal UDL &GSS≥3 (P<0.001). Numerous sensitivity analyses confirmed these findings.

Conclusions: A single untimed drug level (UDL) and GSS independently predict subsequent VF, with suboptimal UDL having a greater hazard than GSS<3, especially if drug levels are undetectable. Together, UDL and GSS can explain a higher proportion of treatment failures than either measure alone. These results could justify the potential investigation of UDL in prospective of management of LLV.