Objectives: Lopinavir/ritonavir (LPV/r) is a co-formulated protease inhibitor (PI) used to treat HIV-1 infection. Lopinavir (LPV) inhibits HIV-1 protease and ritonavir (RTV) inhibits CYP3A-mediated LPV metabolism thereby increasing LPV plasma concentrations [1]. ACTG 5230 was a multicenter, open-label, single arm pilot trial that evaluated LPV/r monotherapy (400/100 mg twice daily) as second-line antiretroviral therapy (ART) in PI-naïve HIV-infected patients from resource-limited settings (RLS) experiencing virologic failure on a first-line nonnucleoside reverse transcriptase inhibitor-based regimen [2]. Given this novel treatment strategy and unique patient population, the objectives of the present work were to characterize LPV plasma pharmacokinetics (PK) and identify patient-specific characteristics associated with PK variability.

Methods: A single, random, steady-state plasma sample was drawn from each subject during study visits at weeks 8, 12, 16, and 24 and analyzed for LPV concentrations by HPLC-UV. RTV concentrations were not measured. LPV concentrations below the assay’s limit of quantification (BLQ) were excluded from the analysis. Several structural models were evaluated including one- and two-compartment models with and without absorption lag time. Model parameter estimates were obtained using the first order conditional estimation with interaction (FOCE-I) method in NONMEM version 7.2. Covariate data consisted of the continuous variables (mean ± sd) weight (63 ± 12.3 kg) and age (39.3 ± 8.4 yrs), and the categorical variables sex and race (Thai or Black African). Concomitant medications were not tested because drugs known or suspected of interacting with LPV/r were prohibited in the study. Covariate models were developed based on clinical relevance of the covariates and likelihood ratio testing during stepwise forward inclusion and backward elimination. In the stepwise procedure, a change in the objective function value of 7.88 units was defined as significant (p<0.005, df=1). A bootstrap re-sampling technique with 1000 runs was used to evaluate stability of the final model and precision of the parameter estimates.  

Results: 427 LPV concentrations from 111 subjects were used in the population PK analysis. 7 concentrations (6 BLQ; 1 near BLQ) were excluded. A one-compartment model with first-order absorption and elimination best described the data. Sex and race modeled by a power function were significant covariates on apparent oral clearance (CL/F) and were included in the final model. Males had 32% faster CL/F than females. Thai subjects had 24% faster CL/F than Black African subjects from Malawi, Tanzania, and South Africa. Eta-shrinkage for CL/F was 12.2%. All parameter estimates, relative standard errors (RSE) and 95% confidence intervals (CIs), as well as bootstrap estimates and 95% CIs are below. Successful estimation and covariance steps were achieved in 99.8% of the bootstrap runs suggesting a robust model.

Parameter

Final Estimate

RSE (%)

95% CI

Bootstrap Estimate

Bootstrap 95% CI

CL/F (L/h)

SEX (M/F) effect

RACE (Thai/African) effect

V/F (L)

Ka

Intersubject variability in CL/F (CV%)

Residual variability by proportional error model (CV%)

2.83

1.32

1.24

45.4

0.364

0.071 (26.7%)

0.089 (29.8%)

4.28

5.77

7.15

26.2

34.6

18.7

10.9

2.59-3.07

1.17-1.47

1.07-1.41

22.1-68.7

0.117-0.611

0.045-0.097

0.070-0.108

2.82

1.33

1.25

44.6

0.375

0.069

0.089

2.57-3.06

1.18-1.50

1.06-1.43

20.4-74.5

0.127-0.699

0.044-0.097

0.071-0.108

RSE (%) = 100% x SE/estimate; V/F = apparent volume of distribution; Ka = absorption rate constant

Conclusions: The PK of LPV in PI-naïve HIV-infected subjects from RLS receiving LPV/r monotherapy as second-line ART were adequately described by a one-compartment model. LPV CL/F was influenced by sex and race with males and Thai subjects exhibiting faster clearances. After accounting for sex and race, the remaining interindividual variability in CL/F is small and contributions from other sources such as host genetics are as yet unknown.