Evaluating risk prediction models for use in lung cancer screening in diverse populations around the world

点击翻译按钮获取中文摘要

PROJECT SUMMARY Screening by low-dose computed tomography (CT) can reduce lung cancer mortality among individuals with a heavy smoking history. Screening is more efficient when eligibility is based on individual risk, predicted using a model including demographics, smoking, and health information. Individuals whose risk exceeds a minimum threshold can be offered screening. Current lung cancer risk models and thresholds were developed and validated in healthy, largely non-Hispanic white, US populations, and their portability outside that context is unknown. This evidence gap represents a major and growing concern, because risk-based lung screening efforts are expanding rapidly worldwide without the needed evidence to support effective implementation. Therefore, the proposed project will (1) harmonize data across 21 cohorts worldwide to produce a unique consortium database on lung cancer risk; (2) evaluate the performance of lung cancer risk models across US races/ethnicities and different countries; (3) build and validate a risk model specifically for Asians; and (4) calculate the sensitivity of proposed risk thresholds across US and international cohorts. To achieve these aims, the research team will (1) capitalize on its existing involvement in the Lung Cancer Cohort Consortium, NCI Cohort Consortium, and Asia Cohort Consortium. Statistical analyses will (2) quantify calibration (E/O) and discrimination (AUC) of risk models among US whites, blacks, Hispanics, and Asians, and in countries across Europe, Asia, and Australia; (3) build and validate a flexible parametric survival model for 5-year lung cancer risk among Asians; and (4) calculate the proportion of ever-smoking lung cancer cases that would be screening eligible based on proposed risk thresholds. This work will provide critical data to accelerate implementation of evidence-based, risk-tailored lung cancer screening programs worldwide, thus maximizing the reductions in lung cancer mortality that can be achieved through screening with finite resources.