Statistical Methods for Cancer Absolute Risk Prediction

描述（由申请人提供）：美国国家癌症研究所（NCI）发布的《癌症风险和预后预测模型的开发、应用和评估》支持创新统计方法的研究，以开发和评估新的和现有的癌症风险预测模型。预测准确性是决定癌症风险预测模型在识别高危人群进行早期预防方面临床适用性的关键因素。随着癌症危险因素的不断出现，迫切需要及时评估其预测的附加价值。但是，平衡成本和统计效率的研究设计以及相应的统计方法大多缺乏。本提案通过开发和评估具有成本效益的两期分层病例对照研究设计以及用于开发和评估癌症绝对风险预测模型的统计推断方法来响应这一PA。我们提出的工作是建立在一种流行的方法上，该方法将病例对照数据和癌症的外部危险率和死亡率结合起来，以预测癌症的绝对风险。作为我们的统计方法和配套软件的展示示例，我们探讨了乳腺体积密度在预测乳腺癌风险方面的附加价值。

DESCRIPTION (provided by applicant): Development, Application, and Evaluation of Prediction Models for Cancer Risk and Prognosis issued by the National Cancer Institute (NCI) supports research on innovative statistical methods for developing and evaluating new and existing cancer risk prediction models. Predictive accuracy is a key factor for determining clinical applicability of cancer risk prediction models in identifying high-risk individuals for ealy prevention. With continuously emerging risk factors for cancer, there is a pressing need for timely assessment of their added values for prediction. But study designs that balance cost and statistical efficiency and accompanying statistical methods are mostly lacking. This proposal responds to this PA by developing and evaluating cost-effective two-phase stratified case-control study designs and statistical inference methods for developing and evaluating absolute risk prediction models for cancer. Our proposed work is built on a popular method that integrates case-control data and external hazard rates of cancer and mortality to predict cancer absolute risk. As a showcase example for our statistical methods and accompanying software, we explore the added value of volumetric breast density for predicting breast cancer risk.