PO.PS01.07 · 人群科学
整合多基因风险评分与非遗传因素的多癌症风险的验证与人群投射
Validation and population projection of multi-cancer risk incorporating polygenic risk scores and non-genetic factors
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
背景:不同癌症之间共享多种风险因素,然而风险评估与预防策略在很大程度上仍是特定部位的。与此同时,多癌症检测试验正作为对特定部位筛查策略(如乳腺X线摄影、结肠镜检查和低剂量CT肺部扫描)的补充方法而涌现。多癌症风险预测可以更全面地了解个体的癌症风险,有可能改善跨多种癌症类型的风险分层预防和筛查策略。
方法:我们使用已发表的、整合了26个风险因素的非遗传风险模型,连同各癌症部位已建立的多基因风险评分(PRS),为14个癌症部位开发了绝对风险模型,并使用个体化一致绝对风险估计器(iCARE)工具整合相对风险参数、风险因素分布、人群发病率和死亡率。我们扩展了iCARE,以在指定时间区间内估计跨多个癌症部位的风险。在前列腺癌、肺癌、结直肠癌和卵巢癌(PLCO)癌症筛查试验的非西班牙裔白人人群中,前瞻性地评估了特定部位和联合多癌症风险预测的校准度(预期/观察(E/O)风险)和区分度(曲线下面积(AUC))。通过将我们的模型与年龄特异性发病率和死亡率应用于源自国家健康与营养调查、国家健康访谈调查和乳腺癌监测联盟的参考人群,投射了美国非西班牙裔白人人群的多癌症风险分层。
结果:在PLCO中的验证显示两性之间的风险分层相当,但女性的校准度(多癌症10年风险E/O=1.05(95%CI:1.03-1.07),AUC=0.60)优于男性(E/O=0.76(95%CI:0.75-0.77),AUC=0.61)。在预测10年风险的第一个和最后一个十分位数之间,相应的观察风险在女性中为3-20%,在男性中为7-30%。针对非西班牙裔白人美国参考人群的多癌症风险投射显示,18-74岁的男性中有24%、女性中有23%处于中至高风险(10年绝对风险≥10%)。
结论:结合非遗传风险因素和PRS的多癌症风险预测模型展现出有意义的风险分层。未来研究将包括这些模型在美国非白人人群中的校准、验证和投射。此类模型可能有助于就影响多种癌症风险的生活方式干预进行咨询,并帮助识别高风险个体以采用新兴的多癌症筛查方法,如液体活检检测。然而,仍需进一步的模型开发和评估,以确保在各人群群体中的最佳表现。
查看英文原文 English abstract
Background: Multiple risk factors are shared across different cancers, yet risk assessment and prevention strategies remain largely site-specific. Meanwhile, multi-cancer detection tests are emerging as complementary approaches to site-specific screening strategies such as mammography, colonoscopy, and low-dose CT lung scans. Multi-cancer risk prediction could provide a more holistic understanding of an individual's cancer risk, with the potential to improve risk-stratified prevention and screening strategies across multiple cancer types.
Methods: We developed absolute risk models for 14 cancer sites using published non-genetic risk models incorporating 26 risk factors, together with established polygenic risk score (PRS) for each cancer site, using the Individualized Coherent Absolute Risk Estimator (iCARE) tool to integrate relative risk parameters, risk-factor distributions, population incidence rates, and mortality rates. We extended iCARE to estimate risk across multiple cancer sites over specified time intervals. Calibration (expected/observed (E/O) risk) and discrimination (area under the curve (AUC)) of site-specific and combined multi-cancer risk predictions were prospectively evaluated in the non-Hispanic White population in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. Multi-cancer risk stratification for the US non-Hispanic White population was projected by applying our model with age specific incidence and mortality rates to a reference population derived from the National Health and Nutrition Examination Survey, the National Health Interview Survey, and the Breast Cancer Surveillance Consortium.
Results : Validation in PLCO showed comparable risk stratification between the sexes but better calibration for females (multi-cancer 10-yr risk E/O = 1.05 (95% CI: 1.03-1.07), AUC = 0.60) than males (E/O = 0.76 (95% CI: 0.75-0.77), AUC = 0.61). Between the first and last decile of predicted 10-year risk, the corresponding observed risk ranges 3-20% in females and 7-30% in males. Multi-cancer risk projections for the non-Hispanic White US reference population showed that 24% of males and 23% of females aged 18-74 are at moderate to high risk (10-year absolute risk ≥10%).
Conclusion: Multi-cancer risk prediction models combining non-genetic risk factors and PRS demonstrate meaningful risk stratification. Future studies will include calibration, validation, and projection of these models in US non-White populations. Such models may support counseling on lifestyle interventions that influence multiple cancer risks and help identify high-risk individuals for emerging multi-cancer screening approaches such as liquid biopsy tests. However, further model development and evaluation is needed to ensure optimal performance across population groups.
利益披露 Disclosure
E. L. Norton, None..
S. Mukhopadhyay, None..
J. Balasubramanian, None..
E. Kim, None..
S. Li, None..
N. Chatterjee, None.