Population pharmacokinetic models for lamivudine and nevirapine to assess drug concentrations obtained during therapeutic drug monitoring
Jan-Stefan van der Walt (1), Phumla Sinxadi (1), Karen Cohen (1), Helen McIlleron (1), Gary Maartens (1). Mats O Karlsson (2)
(1) Division of Clinical Pharmacology, University of Cape Town, Cape Town, South Africa (2) Division of Pharmacokinetics and Drug Therapy, Uppsala University, Uppsala, Sweden
Background and aims: Nucleoside reverse transcriptase inhibitors (NRTIs), the backbone of combined antiretroviral therapy (cART) in resource-limited countries, are not considered candidates for routine antiretroviral therapeutic drug monitoring (TDM). We are investigating lamivudine (3TC) and nevirapine (NVP) TDM as additional adherence tools. Routine sampling at antiretroviral (ARV)-outpatient clinic is seldom pre-dose/trough samples. We developed population pharmacokinetic models of 3TC and NVP to assess drug concentrations obtained during TDM to measure adherence.
Methods: The initial 3TC and NVP models were developed using rich healthy volunteer (HV) 3TC data (n=25, 15 samples per HV) and rich patient NVP data (N=25, 9 samples per patient). Sparse 3TC and NVP data were obtained from blood samples collected during oral glucose tolerance tests in HIV-infected patients on cART (N=128 3TC, N=48 NVP). Rich and sparse data were analyzed separately, simultaneously (combined) and using rich data as prior information for sparse data analysis (prior using the TNPRI functionality in NONMEM VI).
Results: One-compartment models with first-order absorption and elimination, and an absorption lag time best described the 3TC and NVP log-transformed concentration-time data. NVP sparse patient data were inadequate to reliably estimate population parameters. However, there were no marked differences in parameters between rich and sparse patient data and a combined analysis of sparse and rich data (CL/F = 2.52 L/h [4.7% RSE]; V/F 115 L [14%]; ka 3.21 /h [41%]) or an analysis of sparse data with prior from rich data (CL/F = 2.52 L/h [4.56% RSE]; V/F 104 L [8.8%]; ka 2.69 /h [35%]) gave similar population parameter estimates. For 3TC, higher CL/F and ka in HV than patients were indicated in both the combined analysis and when using the prior functionality. Once this was accounted for, the different analyses provided similar patient parameter estimates: combined analysis (CL/F = 14.6 L/h [4.2%]; V/F 89 L [5.7%]; ka 1.81 /h [8.9%]), prior (CL/F = 14.3 L/h [2.5%]; V/F 86 L [2.5%]; ka 1.80 /h [0.2%]).
Conclusions: Population pharmacokinetic models for 3TC and NVP were developed from rich and sparse data. Combined analyses and sequential rich-sparse analyses using the prior functionality led to the same model decisions and very similar parameter estimates.