國立成功大學 統計與資料科學學系

Abstract

A collaboration experience between biostatisticians and clinicians will be shared in the context of a melanoma study. Over 60% of patients respond to immune checkpoint inhibitor therapy, but many subsequently progress on these therapies. Second-line targeted treatment is based on BRAF mutation status, but no agents are available yet for NRAS, NF1, CDKN2A, PTEN, and TP53 mutations. Patient-derived xenografts (PDX) were derived from 15 patients with melanoma for developing personalized drug in the future. In the meantime, we conducted a simulation study to evaluate different statistical approaches in assessing treatment effects across PDX models. The antitumor efficacy of different drugs or drug combinations was evaluated with each PDX model using a mixed-effect model to account for the correlation structure with the repeated measurements from the same mouse. The synergistic effect of two different drugs in PDX tumors was examined by estimating the individual effects and an interactive effect to assess treatment differences in tumor volume over days. RPPA data was analyzed using a mixed-effect model to identify potential biomarkers for treatment response. The results are graphically presented with a volcano plot and a heatmap. Overall, our results demonstrated the efficacy of the MDM2 antagonist as a single agent or as a part of a therapeutic combination with either MAPK pathway or Bcl-2–targeting agents in preclinical melanoma. Our simulation study showed the importance of choosing appropriate statistical analysis methods to best utilize the data collected and arrive at valid conclusions. We also hope to use this example to motivate our collaborators to get statisticians involved in a clinical study, even if it is an animal or PDX study, and at the early stage of the study.