Background:  Optimal adherence to therapy resulting in adequate plasma drug concentrations is essential for successful antiretroviral therapy. Drug concentrations should be interpreted in relation to cut-off values for increased risk of virologic failure. We analyzed possible cut-off values for patients in the 2NN study, which compared the efficacy of neverapine (NVP) and efavirenz (EFV) in addition to stavudine (d4T) and lamivudine (2TC) in antiretroviral-naïve patients.

Methods:  NVP and EFV concentrations were assessed at day 3 and week 1, 2, 4, 12, 24, and 48. The C_min was estimated using a non-linear mixed effects model, for non-steady state (< week 4) and steady state (≥ week 4). Cut-off values of steady state C_min were evaluated for risk of virologic failure with proportional hazard analyses. Data were censored at end of study or change of allocated treatment. Virologic failure was defined as never a plasma HIV-1 viral load < 50 copies/mL or a rebound to 2 consecutive plasma viral load readings of >50 copies/mL. Sensitivity, specificity, and predictive values were calculated for the C_min at which risk of virologic failure was increased.

Results:  The proportion of patients with virologic failure was similar in the group of patients with a C_min below the 50th, 25th, and 10th percentile in both the NVP and the EFV group (± 25% for NVP, and ± 19% for EFV). The risk of virologic failure with NVP was slightly increased at a C_min < 3.1 mg/L (hazard ratio: 1.34; 95% CI 0.90 to 2.00), and < 2.3 mg/L (hazard ratio:  1.38; 0.79 to 2.41), but there was no cut-off value below which a significant increased risk of virologic failure occurred. The cut-off value for a significant increased risk of virologic failure with EFV was a Cmin < 1.1 mg/L (hazard ratio: 1.95; 1.08 to 3.24), but the sensitivity of this value was just 59%. The receiver operator curves for the C_min parameter were very flat for both NVP and EFV, indicating poor predictor properties for virologic failure. The negative predictive value of the parameter was more appropriate, indicating a reasonable probability of therapy success when the drug concentration as above 2.3 mg/L (77%), or 3.1 mg/L (78%) for NVP, and above 1.1 mg/L (88%) for EFV.

Conclusions:  There was no clear C_min of NVP below which the risk of virologic failure was significantly increased. A reasonable probability of therapy success (77%) already existed when the NVP C_min was above 2.3 mg/L. For EFV, this value was 88% for a cut-off > 1.1 mg/L. Therapeutic drug monitoring should target drug concentrations above these values.

Keywords: pharmacokinetics; non-nucleosides; 2NN