Update of Population Pharmacokinetic Model for Tenofovir (TFV) in HIV-1-Uninfected Members of Serodiscordant Couples

Purpose: Oral tenofovir disproxil fumarate (TDF) has demonstrated success in HIV pre-exposure prophylaxis (PrEP) among high risk groups. A population pharmacokinetic model (PoP PK) was reported by us using the Partners PrEP trial data in serodiscordant couples. The objective of current work was to update, the prior population pharmacokinetic model of tenofovir with pharmacokinetic and adherence data from the Demonstration project. Methods: Two plasma samples were collected from study subjects and their dosing data was extracted from medication event monitoring system (MEMS®) records. Data from the Demonstration project and the Partners PrEP trial were combined in the analysis. The PoP PK model developed was a two compartment model parameterized with first order absorption rate constant (Ka) and absorption lag-time (Alag), clearance (CL), central and peripheral volumes (Vc & Vp) and inter-compartmental clearance (Q). Creatinine clearance was included as covariate on CL. Exponential error was used for between-subject-variability (BSV) on parameters. Residual error was modeled as combined additive and proportional error model. Gibiansky’s correction for bio-availability was included in the model to adjust for dosing errors. Model was qualified with visual predictive check (VPC, n=1000 samples) and bootstrap procedure (n=500 iterations). NONMEM software (version 7.3) was used for modeling and R software package (version 3.4.2) was used for data management and plots. Results: A total of 1,592 TFV levels from 565 subjects were used for model development. The final fixed effect parameter estimates were: CL-47.8 L/h; Vc – 214 L; Vp – 512 L; Ka – 1.7; Q – 300 L/h and lag time was 0.69 hr. Relative standard Error (RSE) of estimates were in the range of 2 to 32% for all parameters. Final random effects parameters were BSV on clearance (23%), on Vc (76%), on Ka (79%) and on additive error (150%). RSE of random effect parameters ranged from 12 to 22%. Additive error (SD) was 19 ng/mL and proportional variability was 21% (CV). Goodness of fit plots showed that the model did not have major bias. VPC showed that the distribution of model simulated data agreed with observed data. Conclusion: Tenofovir PoP PK model was updated with the new data from demonstration project. The model will be utilized for interpreting concentration based threshold of protection using MEMS® adherence patterns and text messaging on sexual activity information available in the trial.