Utilizing Drug Combo to improve the design of phase 1 trials for anticancer drugs.

Abstract

1571 Background: Accurate prior knowledge of a single drug or drug combination on toxicity, pharmacokinetics (PK), maximum tolerated dose (MTD), and dose-limiting toxicity (DLT) can help oncologists and statisticians design an efficient phase 1 clinical trial. However, most of the relevant knowledge is shattered in various databases. More importantly, MTD and DLT are missing in any databases. This study constructed a comprehensive knowledge base with expert-curated MTD and DLT knowledge from published anticancer drug combination phase 1 clinical trial results. We also demonstrated the utility of the knowledgebase to design a more efficient phase 1 trial. Methods: Toxicity, pharmacology, and PK knowledge were integrated from publicly available data sources, including DrugBank, SIDER, and drug labels. An interdisciplinary team manually curated the MTD and DLT knowledge. Relevant papers were discovered by PubMed keyword search (drug name + MTD or RP2D). Details regarding patient eligibility, treatment information (dose, formula, and schedule), trial design method, dose levels, DLT definition, MTD or Recommended phase 2 dose, and DLT incident were curated in the structured data format and stored in a database. Rxnorm and MedDRA were used for normalizing drug names and DLT. Results: Currently, 2358 papers on cancer drugs phase 1 trial have been curated. Additionally, pharmacological and PK knowledge of 197 cancer drugs, ADE information from 1094 drug labels, and more than 500000 pieces of DDI evidence were integrated into the knowledge base. This knowledge base (drugcombo.info) could help researchers design an efficient phase 1 trial. For example, based on knowledge from Drug Combo, the combination of nivolumab and axitinib did not have potential PK DDI and severe overlapping toxicity. The reviewing physician considered a phase 1b study to start at a curated 5 mg maximum dose of axitinib with dose reduction to 3 and 2 mg and maintenance of a 3-mg/kg dose of nivolumab based on guidelines. With such additional preliminary knowledge, the reviewing statistician felt the adaptive statistical methods could be applied to redesign this phase I combination trial and minimize patient exposure to sub-therapeutic dose levels. Conclusions: To our knowledge, Drug Combo is the first knowledge that integrated curated MTD and DLT knowledge. It would promote applying curated knowledge to better design phase I trials to treat cancer.