Developing new statistical methods and designs for clinical research involving categorical variables

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K11790
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 60030:Statistical science-related
• Research Institution: Yokohama City University
• Principal Investigator: YAMAMOTO Kouji 横浜市立大学, 医学研究科, 教授 (10548176)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: 臨床試験 / F1スコア / 選択デザイン

Outline of Final Research Achievements

In clinical research,categorical variables are often selected as the primary endpoint.In this study,we studied two problems that were derived from real-world problems: (1) the proposal of an F1 score-based analysis method for comparisons of methods with three or more categories of results, and (2) the development of a flexible treatment candidate selection design with multiple binary variables as the co-primary endpoints.
For (1), we proposed the F1 score, which is a commonly used performance measure in the machine learning field, and clarified the statistical properties of F1 score measures used in the case of three or more categories. For the second, we proposed a new design that simultaneously evaluates two binary variables, efficacy and safety, and selects the best treatment.

Free Research Field

医学統計

Academic Significance and Societal Importance of the Research Achievements

本研究は実際の医学研究における課題から着想を得たものであり，これらに対して新たな解析手法等を提案した．これは新たな統計的手法開発にとどまらない．課題(1)に対してはより科学的に新たに開発された検査法や診断法の有効性を述べることができ，課題(2)に対してはより多面的な角度から最善の治療法を選択できる可能性が高まる．本研究手法を今後の医学分野へ応用することにより，より効率的な研究遂行が可能となり，最終的には疾患で苦しむ患者さんへのよりよい医療の提供につながるものと期待される．