Pharmacokinetics and statistical analysis
A sample size of 18 evaluable patients was selected to provide reasonable precision for the estimation of the effect of capmatinib on pharmacokinetics of digoxin and rosuvastatin, assuming the intra-patient coefficient of variation (CV) of digoxin (44.7%) and rosuvastatin (31.5%). Using these intra-patient CV% estimates and a sample size of 18, the half-width of the 90% confidence intervals (CIs) for treatment difference comparison (capmatinib + digoxin vs. digoxin alone; capmatinib + rosuvastatin vs. rosuvastatin alone) on the log scale was 0.247 for digoxin and 0.178 for rosuvastatin. These calculations were based on t-distribution with one-sided α-level of 0.05 and N-1 degrees of freedom. Due to the strict evaluability criteria and the requirement to conduct the DDI test at steady state, the total number of patients enrolled was expected to be approximately 32.
The pharmacokinetic analysis was based on patients in the pharmacokinetic analysis set. Three separate sets were considered; one for each of the probe drugs (digoxin and rosuvastatin) and one for capmatinib. For each of the probe drugs, all patients who provided an evaluable pharmacokinetic profile for all periods (first period after the cocktail administration and second period after administration of cocktail + capmatinib) were included for analysis. A profile was considered evaluable if patient received the planned dose of capmatinib on Day 22 and at least 3 consecutive days prior to co-administration with probe drugs; received planned dose of probe drugs, did not vomit within 4 hours after receiving capmatinib or probe drugs and provided at least one primary pharmacokinetic parameter (AUCinf, AUClast, or Cmax) for probe drugs. Pharmacokinetic analysis set for capmatinib included all patients who provided at least one evaluable concentration for capmatinib; a concentration was evaluable if patients received the same dose of capmatinib at least 3 consecutive days prior to sampling, did not vomit within 4 hours after receiving capmatinib, and had pre-dose samples collected before the next dose administration and 9 to 15 hours after the last dose administration.
Pharmacokinetic parameters were estimated by a non-compartmental method using Phoenix WinNonlin 6.4 (Pharsight, Mountain View, CA). The log-transformed pharmacokinetic parameters (AUClast, AUCinf and Cmax) were analyzed using a linear mixed model to assess the effect of multiple doses of capmatinib on the pharmacokinetics of a single oral dose of each of the probe drugs separately. The model included treatment (probe + capmatinib and probe alone) as a fixed effect and patient as a random effect. Point estimates of treatment differences and the corresponding 90% CIs were calculated and anti-logged to obtain the point estimates and 90% CI for the geometric means ratio of the probe + capmatinib versus probe alone on the original scale.