A nonlinear mixed effects model to characterise lamivudine absorption and distribution

Background: Lamivudine (3TC), a potent selective inhibitor of HIV reverse transcriptase, is used in combination antiretroviral therapy (ART). 3TC pharmacokinetics is characterised by highly variable absorption and is rapidly distributed to a slow equilibrating peripheral compartment. Although not considered a candidate for routine therapeutic drug monitoring, 3TC pharmacokinetics is unaltered by concomitant anti-tuberculosis treatment, a common opportunistic infection in our population.

Methods: Rich 3TC concentration-time data from healthy volunteers (N=25) and HIV-1 infected patients (N=25) were analysed using nonlinear mixed effects modelling. Absorption was described using either no absorption lag time, absorption lag time, or a transit compartment model where the mean transit time (MTT) and the number of transit compartments (NN) were estimated. Concentrations below the limit of quantification (BLQ) were handled using the F_flag (M3) method. Typical population values were estimated using the FOCE LAPLACIAN method in NONMEM 6.2. The stability of the final model was evaluated using the nonparametric bootstrap method. Visual predictive checks (VPCs) were used to evaluated if a published model (Moore et al) and our final model could predict 3TC concentrations in HIV-1 infected patients on ART (N=25) during and after anti-tuberculosis treatment.

Results: 350 3TC concentration-time points were available. A two compartment model with a transit compartment absorption model best described the data. Typical population values were CL/F = 22.4 L/h (CV 15%), V2/F = 2.8 L (CV 84%), Q/F = 6.9 (CV 40%), V3 = 69.3 L (CV 79%), Ka = 0.356/h (CV 19%), MTT = 0.463 h (CV 34%), NN = 10.6 (CV 72%). VPC indicated that the published model predicted faster absorption with a lower Cmax and overestimated the between subject variability during the elimination phase in both healthy volunteers and HIV-1 infected. Handling BLQ data using the F_FLAG method resulted in obtaining acceptable estimates of the peripheral volume of distribution, irrespective of the method of characterizing absorption.

Conclusion: We developed a population pharmacokinetic model of 3TC to characterise its variable absorption and distribution. The model was stable and predicted 3TC concentration-time profiles in HIV-1 infected patients during and after rifampicin-based antituberculosis therapy.