A Simulation and Estimation Platform for Malaria Model Evaluation
Background: Accelerating clinical development of new compounds demands efficient systems for evaluation and interpretation of trial results. Systematizing trial evaluation methods yields efficiency and confidence in results. A simulation/estimation (S/E) platform was employed for definitive assessment of parasite models used for analysis of volunteer infection studies (VIS). Using rich data, parasite models were evaluated for identifiability and performance.
Methods: Simulated hourly parasite counts (mrgsolve; 500 replications) were analyzed (NONMEM 7.3; KIWI 2) with 4 structural models with various random effects (RE). Three empirical models (traditional first-order growth and drug effect [TFGDE], indirect response [IDR], and Gompertz [GOMP]) and a semi-mechanistic model (Gordi) were evaluated. Recrudescence, limit of quantification (LOQ of 10 or 111 parasites/mL), growth phase, and drug effects were considered.
Results: The TFGDE with RE on 2 and 5 parameters and Gordi with RE on 3 parameters were most stable with respect to identifiability and precision of parameter estimates. For TFGDE and Gordi models with LOQ = 10, drug effect was well estimated with EC50 (0.0123 mcg/mL; 95% confidence interval [CI] = 0.0122 – 0.0123) and Kpinj (0.329 mcg/mL x h; 95% CI = 0.328 – 0.331), respectively.
Conclusion: Traditional and Gordi models perform well. Further work on LOQ and limited data scenarios is needed. The S/E platform allows assessment of relative model performance to guide efficient model selection and refinement.
By Kayla Ann Andrews, Joel S. Owen, Luann Phillips, Nathalie Gobeau, Jorg J Mohrle, Thaddeus H Grasela
ASCPT 2019 Annual Meeting, March 13-16, 2019, Washington, DC