To date, therapeutic drug monitoring (TDM) is carried out with antiretrovirals and is usually based on total concentrations (Ct ). However, for some patients, TDM does not reflect efficacy or the avoidance of toxicity as is the case for atazanavir (ATV), a HIV protease inhibitor. As the unbound concentration (Cu ) is the pharmacological active form, the aim of the study was to evaluate the value of Cu and the unbound fraction (fu , fu = Cu /Ct ) for the TDM of ATV. The variability of Cu and the corresponding fu of ATV was explored in 43 patients treated with ATV for an average of 13.5 months. Cu was determined by coupling ultrafiltration and liquid chromatography. As ATV is highly bound to alpha-1 acid glycoprotein (AAG), the correlation between fu and AAG was also evaluated. The viral load was monitored to evaluate the patients' virologic response, while total plasma bilirubin and unconjugated plasma bilirubin were used as biomarkers of ATV toxicity. Median trough Cu and Ct were 37.9 μg/L (Interquartile range (IQR) 20.6-94.9 μg/L) and 628.6 μg/L (IQR 362.7-1078.1 μg/L), respectively. fu , Cu and Ct showed high variability, but the fu variability was not correlated with the AAG level. The unbound concentration and fraction were unrelated to the virologic response (P = 0.21 and P = 0.65 for Cu and fu , respectively) nor to the unconjugated bilirubin (Pearson correlation coefficient (ρ), ρ = 0.22; P = 0.17 for Cu ). Neither total nor unbound concentrations of ATV fully explained hyperbilirubinaemia or virologic failure. From this study, we conclude that unbound ATV did not appear to be more relevant than Ct .