Aims: Study protocols for antidiabetic drugs often include withdrawing study participants due to elevated fasting plasma glucose (FPG) and glycosylated hemoglobin (HbA1c) above clinically acceptable limits, or withdrawing participants for safety reasons. Furthermore, patients can voluntarily dropout from the study for any reason at any point during the study. We sought to quantify the exposure-response relationships for efficacy (FPG and HbA1c) and safety (alanine aminotransaminase [ALT]) for an investigational glucagon receptor antagonist (LY2409021) in patients with type 2 diabetes mellitus (T2DM). In addition, we included information on dropout and evaluated the impact of this information on the estimated exposure-response relationship. Lastly, we carried out clinical trial simulations to investigate the impact of dropout on study power and tested alternative study designs.
Methods: Data from a total of 373 healthy subjects and patients with T2DM (7 studies) were combined to develop the population pharmacokinetics (PK) model and investigate the effect of predefined covariates. Pharmacodynamic data (FPG, HbA1c and ALT) were available from a 24 week phase 2 dose ranging study in which patients with T2DM were equally randomized to 3 LY2409021 dose levels and placebo (N = 254). PK and exposure-response modeling were performed using NONMEM 7.3.0. Clinical trial simulations were carried out using R 3.2.2.
Results: A one-compartment model with first order absorption and elimination adequately described the PK of LY2409021. Body weight was the only significant predictor of clearance and volume. A linked indirect response model was used to describe the time course of FPG and HbA1c simultaneously. Time course of ALT was also modeled simultaneously with the FPG and HbA1c using an indirect response model. Treatment with LY2409021 decreased FPG, consequently leading to a long term reduction in HbA1c. Through an Emax model, the glucose-lowering effect of LY2409021 was constrained such that the HbA1c could not be reduced below a physiological threshold. A time-dependent placebo effect on FPG and HbA1c was included in the model. There was a trend towards higher ALT concentrations with increased LY2409021 exposure, which was adequately captured through a non-linear relationship. A time to event modeling approach was used to describe the dropout, and FPG was a significant predictor of the hazard. Patients with higher FPG values (those in the placebo arm and those with lower LY2409021 exposure) were at higher risk of dropping out of the study, resulting in patients randomized to higher LY2409021 doses being more likely to complete the study, potentially biasing the estimation of drug effect.
Conclusion: The exposure-response relationships for LY2409021 were described through a simultaneous modeling approach that included a model for the dropout. Simulations revealed that daily doses of 10mg to 20mg of LY2409021 would be effective in achieving clinically significant FPG and HbA1c reductions in patients with T2DM. The model incorporating dropout was also used to determine the study power for various study designs taking into account the impact that dropout would have on the number of patients completing a 24 week study period. Adaptive designs that increase drug exposure in patients with elevated FPG may result in a more efficient and optimized clinical trial designs.