Marion Bouillon-Pichault, Satyendra Suryawanshi, Lora Hamuro, Paul Statkevich, Amit Roy, Akintunde Bello, Tarek Leil
Clinical Pharmacology and Pharmacometrics, Bristol-Myers Squibb, Princeton, NJ
Objectives: Several immuno-oncology (IO) agents are currently approved for the treatment of melanoma. Phase 3 clinical trials to date have compared different treatments to one another, but between trial comparisons are not possible due to differences in trial design and population characteristics. Given these challenges, an evaluation of IO and non-IO therapies across clinical trials in melanoma would be informative.
Our objective was to develop a modeling framework to quantify the landscape of available treatments in melanoma to enable the positioning of new assets in clinical development. This framework would include both efficacy and safety and would quantify the impact of important covariate effects and population characteristics (eg, PD-L1 expression, performance score, line of treatment) across trials.
Methods: The analysis database was made of publicly available summary-level results from clinical trials investigating the efficacy (overall response rate – ORR) and safety (incidence of all grade 3+ adverse events [AEs]) of approved and investigational anti-PD1 and anti-CTLA4 agents in melanoma and their comparators. Four different monotherapy or combination treatments were included in the database. These data were used to quantify the efficacy, safety, and covariate effects of different therapeutic classes (anti-PD1, anti-CTLA4, anti-PD1 + anti-CTLA4 combination, and chemotherapy [no BRAF or MEK inhibitors]) using a non-linear mixed effects model-based meta-analysis approach (MBMA) (1).
Results: A general empirical MBMA was established linking ORR and incidence of AEs with received drug class; the residual variability was weighted by the square root of the number of patients in the trial arm to account for summary data precision (1). The model passed the goodness of fit plots diagnostics. The MBMA estimated that the efficacy of anti-PD1 compounds, both in monotherapy and in combination with anti-CTLA4, increases with the percentage of PD-L1–positive patients. The efficacy of all drug classes evaluated decreased with increasing line of therapy. Anti-PD1 was the drug class associated with the lower incidence of AEs and no covariate effect was quantified with this analysis.
Conclusions: We used an MBMA to estimate the efficacy and safety of IO and non-IO drug classes in melanoma and quantified the impact of PD-L1 expression and ECOG performance score on efficacy. This MBMA can be used to simulate head-to-head comparisons between compounds in development and currently available IO therapies in melanoma prior to conducting a clinical trial.
References:
[1] Ahn, J. E. & French, J. L. Longitudinal aggregate data model-based meta-analysis with NONMEM: approaches to handling within treatment arm correlation. Journal of Pharmacokinetics and Pharmacodynamics 37, 179–201 (2010).
Reference: PAGE 28 (2019) Abstr 9129 [www.page-meeting.org/?abstract=9129]
Poster: Methodology - New Modelling Approaches