Generating Control Groups for Organ Impairment Studies: A Case‐Study Comparing Statistical and Population Pharmacokinetic‐Based Matching Approaches

A common challenge in conducting phase 1 studies that assess the impact of organ impairment on the pharmacokinetics of a drug is the recruitment of a demographically matched control group. The work presented here evaluated alternative approaches for generating control groups in these studies. Available phase 1 data from the upadacitinib and elagolix clinical programs were leveraged as case studies. A statistical matching approach and a population pharmacokinetic model-based approach were evaluated retrospectively for these programs' hepatic and renal impairment clinical studies. Geometric mean ratios of logarithmically transformed C_max and AUC_inf were used to compare exposure in organ impairment groups to respective matched or virtual control groups. In the statistical matching approach, the genetic matching algorithm using Mahalanobis distance showed that external control groups were adequately demographically balanced across all impairment groups of the study except for age. A 3:1 k-match approach minimized the prediction error between matched and reference in-study results for both case studies, resulting in differences in geometric mean ratios ranging from -19% to 3% and -27% to 40% for upadacitinib and elagolix, respectively, compared to in-study controls. Similarly, the population pharmacokinetic approach used models developed from phase 1 data in healthy participants and found that the results were generally comparable to the in-study results, with differences in geometric mean ratios ranging from -30% to 17% and -24% to 41% for upadacitinib and elagolix, respectively. These analyses demonstrate that both approaches may be viable alternatives to assess the impact of organ impairment on pharmacokinetics.