Statistical methods to develop and evaluate risk prediction models

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K15956
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Statistical science
• Research Institution: National Cancer Center Japan
• Principal Investigator: Kuchiba Aya 国立研究開発法人国立がん研究センター, 研究支援センター, 室長 (40510699)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 予測モデル / がん罹患 / Calibration / Discrimination / 疫学研究

Outline of Final Research Achievements

Recent cancer epidemiology has provided much evidence for cancer risk factors, including environmental, lifestyle and genetic factors. Prediction model to provide the risk of cancer based on the established risk factors, would contribute to developing risk-based prevention or health management strategies.
In the research project, the methods to estimate prediction ability of the prediction models has been examined and applied to real epidemiology studies, in particular for data from often-used study designs with biomarker data. The statistical methods for assessing etiologic heterogeneity of a cancer have been developed and applied to the cohort study data as a relevant topic. In addition, new index to evaluate prediction ability for outcome with more than two categories has been proposed and the characteristics of this index have been examined by simulation studies.

Free Research Field

生物統計学/疫学