Background:  The utility of plasma ARV concentrations for therapeutic drug monitoring in HIV is limited by day-to-day (intra-individual) variability in levels. Adherence patterns prior to clinic visits can be atypical, and fluctuations in pharmacokinetic parameters influence single measurements. Levels of ARV in hair provide a measure of average exposure over weeks to months and may contribute to models of treatment response.

Methods:  We developed methods of extracting efavirenz (EFV) from small samples of finely cut human hair (10 to 20 strands, ~2 mg) from patients on therapy using organic solvent extraction under weakly alkaline conditions. The concentration of extracted EFV was then measured by liquid chromatography/tandem mass spectrometry and validated from 0.01 to 4.0 ng/mg of hair. We then analyzed hair levels of EFV in 68 participants of the Women’s Interagency HIV Study (WIHS) approximately 6 months after initiating EFV-based regimens. Multivariate logistic regression models estimated the effect of drug levels in hair and important covariates (including self-reported adherence) on viral load decline to below limits of detection.

Results:  Concentrations of EFV in hair demonstrated the strongest independent association with virologic suppression in multivariate models. The odds ratio (OR) for viral suppression was 3.6 (95%CI 1.4 to 9.3, p =0.008) for every 2-fold increase of EFV level in hair when controlling for age, race, pre-treatment viral load and CD4, self-reported adherence, and prior ARV experience. African American women trended toward poorer outcomes (OR for achieving virologic success 0.065 (95%CI 0.003 to 1.68, p = 0.10)). No other covariate showed a strong association with virologic suppression in multivariate models. Women who reported ≥95% adherence over the preceding 6 months showed a trend towards a negative association with outcome (OR 0.086, 95%CI 0.005 to 1.45, p = 0.09) for ≥95% adherence versus <95%, perhaps indicating reports were of variable accuracy.

Conclusions:  In models examining predictors for virologic suppression after initiation of EFV-based regimens, levels of EFV in hair were strongly and independently associated with viral load undetectability. Self-reported adherence is a poor indicator of ARV exposure, adding little to models of outcome. Levels of ARV measured in small samples of hair estimate long-term exposure to a drug and may serve as noninvasive monitoring tools for treatment outcomes.