Pharmacometrician Pfizer Stonington, Connecticut, United States
Disclosure(s):
Tina M. Checchio: No relevant disclosure to display
Objectives: Osivelotor is an investigational small molecule hemoglobin (Hb) modifier in development for the treatment of sickle cell disease (SCD). A population pharmacokinetic- pharmacodynamic (PKPD) model was developed using pooled phase I/II data describing the relationship between PK exposure and the drug target, Hb. PKPD-based simulation generated in silico healthy control (HC) cohorts for assessment of organ impairment (OI) impact on osivelotor exposure, in lieu of recruiting healthy participants (HP), who derive no medical benefit from exposure to osivelotor.
Methods: Pooled data from study participants (N = 101) enabled estimation of the model parameters. Parameters for each in silico HC were sampled (n = 8 per cohort, N = 1000 cohorts) from a normal distribution, using the mean and variance estimated from the PKPD model. The HC in silico population were assigned random values for demographic characteristics, with age and weight sampled from a uniform distribution described by the minimum and maximum observations of the OI (test) group. Sex was sampled from binomial distribution based on the proportion of males to females in the OI (test) group. Baseline hemoglobin was sampled from a log -normal distribution with a mean and variance observed in HPs. Simulated sampling times were consistent with study sampling schemes and were derived using individual model parameters and a system of ordinary differential equations (R ver 3.4.1)1. For each in silico cohort (N=1000), an analysis of variance (ANOVA) was used to compare the log-transformed PK parameters (AUC and Cmax) between the participants with OI (test) and the in silico HC (reference), generating a distribution of geometric mean ratios (GMRs). The distribution was non-parametrically summarized by the median and percentiles. Pre-defined bounds (0.5 - 2.0) were used to assess the median GMR for no clinically relevant differences between OI and HC. Percentiles around the GMRs were used to provide additional context about the likelihood of simulated trials resulting in exposures that fell outside of the pre-specified range.
Results: Median GMRs (5th, 95th percentiles) for whole blood AUClast were 1.28 (0.92, 1.89) and 1.37 (1.04, 1.84) comparing renal impaired and hepatic impaired participants to simulated HC data, respectively. The median GMRs were contained within the prespecified bounds; moreover, of the 1000 clinical trial simulations, 1.2% and 1.9% GMRs were >2, respectively and none were less than 0.5%.
Conclusions: A PKPD-based simulation approach to the assessment of OI on osivelotor PK exposure leveraged the ability of random sampling to generate a distribution of possible outcomes. Relative to a traditional OI study design where a single healthy participant reference represents a random sample of 1, use of in silico HC reference cohorts was able to provide a robust assessment of the distribution of 1000 GMRs. Median GMR for whole blood AUC was used for adaptive decision-making for efficient OI study conduct. Of the 1000 virtual clinical trials, >98% produced a GMR < 2-fold, suggesting that OI does not have a meaningful impact on osivelotor exposure. Ultimately, the PKPD-based analysis supports the clinical development of osivelotor for patients with SCD and organ impairment.
Citations: 1. Baron K (2024). mrgsolve: Simulate from ODE-Based Models. R package version 1.5.2, https://github.com/metrumresearchgroup/mrgsolve, https://mrgsolve.org/docs/.
