Stochastic Assessment of Adaptive Volunteer Infection Study (aVIS) Designs in Malaria – A Case Study with Artefenomel
Background: Optimal dose selection for Phase 2 malaria clinical trials requires a well described quantitative dose-response (DR) relationship. We propose the probability of recrudescence (pR) versus dose as a functional structure for dose optimization. Fisher Information Matrix based optimization was used to select doses for subsequent cohorts in a reduced aVIS study design. Artefenomel published results1 were used for a case study stochastic assessment of aVIS trial designs and the impact of a 2nd optimized cohort on pR vs dose definition.
Methods: Artefenomel aVIS trials were simulated over 3 dose ranges: literature reported doses (100, 250, 500 mg), doses at the ow end of the DR curve (50, 100, 200 mg), and high doses (1000, 1500, 2000 mg). 100 trials of 2-2-4 (n=8 subjects) and 3-3-6 (n=12) designs2 were evaluated for cohort 2 using a Hill function for pR vs dose. Optimal cohort 2 doses were selected for each cohort 1 dose range. Maximum slope of parasite descent was evaluated as a predictor of pR. Impact of definition of cure (parasite level = 0.0002, 0.0003, 0.003, 0.01 parasites/mL at 336 h post drug) in model simulations was evaluated.
Results: Literature (2% bias) and low dose (14 and 10% bias); range results were well estimated ED50; bias was larger for high dose group (79 and 89%). A 2nd cohort markedly improved pR predictions when original dose range did not contain EC50 (bias = 0.4 for both low dose designs and 22 and 31% for high dose designs). Definition of cure had small impact on ED50 (ED50 = 491, 490, 494, 490 mg, for the cure definitions, respectively).
Conclusions: Effective methods for dose selection of 1st cohort aVIS studies are needed. Inclusion of a 2nd cohort with optimized dose selection has greatest benefit when the cohort 1 doses do not include a dose near the true ED50. Slope of parasite decline is correlated with pR, but lacks separation around dose = ED50.
By James Clary, Andrew Castleman, Thaddeus Grasela, Joel Owen