Population pharmacokinetics of lopinavir/ritonavir (Kaletra) in HIV-infected patients.

Background: A relationship between plasma concentrations and viral suppression in patients receiving lopinavir (LPV)/ritonavir (RTV) has been observed. Therefore, it is important to increase our knowledge about factors that determine interpatient variability in LPV pharmacokinetics (PK).

Methods: The study, designed to develop and validate population PK models for LPV and RTV, involved 263 ambulatory patients treated with 400/100 mg of LPV/RTV twice daily. A database of 1110 concentrations of LPV and RTV (647 from a single time-point and 463 from 73 full PK profiles) was available. Concentrations were determined at steady state using high-performance liquid chromatography with ultraviolet detection. PK analysis was performed with NONMEM software. Age, gender, height, total body weight, body mass index, RTV trough concentration (RTC), hepatitis C virus coinfection, total bilirubin, hospital of origin, formulation and concomitant administration of efavirenz (EFV), saquinavir (SQV), atazanavir (ATV), and tenofovir were analyzed as possible covariates influencing LPV/RTV kinetic behavior.

Results: Population models were developed with 954 drug plasma concentrations from 201 patients, and the validation was conducted in the remaining 62 patients (156 concentrations). A 1-compartment model with first-order absorption (including lag-time) and elimination best described the PK. Proportional error models for interindividual and residual variability were used. The final models for the drugs oral clearance (CL/F) were as follows: CL/F(LPV)(L/h)=0.216·BMI·0.81(RTC)·1.25(EFV)·0.84(ATV); CL/F(RTV)(L/h) = 8.00·1.34(SQV)·1.77(EFV)·1.35(ATV). The predictive performance of the final population PK models was tested using standardized mean prediction errors, showing values of 0.03 ± 0.74 and 0.05 ± 0.91 for LPV and RTV, and normalized prediction distribution error, confirming the suitability of both models.

Conclusions: These validated models could be implemented in clinical PK software and applied to dose individualization using a Bayesian approach for both drugs.