Comment on: Population pharmacokinetics of the rilpivirine long-acting formulation after intramuscular dosing in healthy subjects and people living with HIV

We read with great interest the article by Neyens et al.¹ published in your journal. This study describes the population pharmacokinetics (PK) modelling strategy of the rilpivirine long-acting (LA) formulation after intramuscular (IM) dosing in healthy and HIV-infected subjects. Data from seven clinical trials were analysed. Briefly, a one-compartment model with linear elimination and two parallel absorption pathways with sequential zero first-order processes adequately captured rilpivirine PK.

As stated by the authors, the aim of the study was to quantify the inter- and intra-individual variability and to provide a quantitative assessment of the potential effect of intrinsic and extrinsic factors on rilpivirine exposure. Nevertheless, it is noteworthy that the influence of physiologically plausible covariates (i.e. age, body weight, BMI, sex, race, health status and needle length) was evaluated on absorption parameters only (it is not clear on which parameters). It was stated that a previously developed population PK model for rilpivirine after oral administration in Phase III studies did not identify statistically significant covariates.² The disposition and elimination being independent of the administration mode, the authors deemed it unnecessary to assess the covariate effects on all pharmacokinetic parameters. However, reading the results of the Crauwels et al. study,² gender and race were identified as statistically significant covariates on rilpivirine apparent clearance although the authors considered these effects as not clinically significant. Furthermore, in another ‘real-life’ observational population PK study, Néant et al.³ pointed out a rilpivirine half-life two times shorter compared with the one estimated by Crauwels et al.² from the ECHO/THRIVE Phase III trials. The authors ascribed this discrepancy on half-life to the difference between the ‘real-life’ population coming from therapeutic drug monitoring (TDM) and the population included in the clinical trials.

In the modelling strategy, four Phase I studies with intensive PK sampling in healthy adults were used to develop the reference population PK model (dataset A). Sparse PK data from one Phase IIb study and two Phase III studies in adults living with HIV were used for external evaluation (dataset B). The reference model developed with dataset A was adapted to account for the presence of systemic rilpivirine before IM administration of the rilpivirine LA formulation, resulting from the 4 week oral lead-in period (OLI) included in the Phase IIb and III studies. The adapted model consisted of a combination of the reference model and the previously available model for rilpivirine after oral administration obtained from Phase III trials.²

This model failed to capture the observed time course of rilpivirine plasma concentrations in people living with HIV from the Phase IIb and III studies (dataset B). Datasets A and B were, therefore, pooled to update the population PK model (dataset C). The updates in the model consisted of: (a) addition of a study phase effect to account for the observed difference in exposure between Phase I and IIb/III studies; this effect was included as a relative bioavailability factor; (b) fixing KA2 and the effect of age on KA2 to the estimates obtained from the healthy volunteer dataset; and (c) fixing the oral PK parameters and addition of a bioavailability-correcting factor for the OLI phase model as the model was not fitting the actual observed data from the OLI phase of Phase IIb/III studies.

While these updates are proposed to help in fitting the data, they appear as a ‘statistical’ strategy rather than an appropriate approach for explaining the actual differences of rilpivirine exposure with clinically significant covariates. The global effect of covariates on rilpivirine PK was illustrated by the geometric mean ratio of the AUC. However, it is not clear which AUC was used. Furthermore, it would also have been interesting to have some insights on the simulated trough concentrations (i.e. C_trough ≥ 17.3 ng/mL, as presented in the analysis of the same data presented in a poster⁴) at different timepoints after the LA formulation initiation (with or without the OLI phase, which is no longer mandatory⁵).

This raises concerns about the adequacy of the model in a real-world setting and its ability to predict adequate rilpivirine exposure without any dose adjustment or TDM strategy. Thoueille et al.⁶ reported the concentrations observed in a small cohort of patients (n = 46) who were relatively healthy and, therefore, similar to the population observed in clinical trials. They already observed a significant PK variability and cases of suboptimal rilpivirine concentrations with the standard dosing scheme.

Lastly, parameters of the model were not presented. Knowing them would have been helpful to understand the LA rilpivirine PK and its variability, and would provide clinical pharmacologists the tool to perform model-based TDM to optimize the dosing scheme early and for every sampling time (by shortening or extending the IM dosing interval, optimizing the oral bridging when necessary), avoiding suboptimal rilpivirine concentrations. It is noteworthy that a cabotegravir LA PK model has been described and the parameters are presented in the Han et al. study.⁷

Funding

None.

Transparency declarations

None.