Statistical Methods for Selection and Evaluation of Biomarkers

生物标志物选择和评价的统计方法

• 批准号: 8483561
• 负责人: Ying Huang
• 金额: $ 32.47万
• 依托单位: FRED HUTCHINSON CANCER RESEARCH CENTER
• 依托单位国家: 美国
• 财政年份: 2013
• 资助国家: 美国
• 起止时间: 2013-05-10 至 2018-01-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8483561
• 关键词: Address; Algorithms; Area; Biological Markers; Case-Control Studies; Characteristics; Classification; Data; Detection; Development; Diagnosis; Diagnostic tests; Disease; Disease Outcome; Early Detection Research Network; Evaluation; General Population; Goals; Individual; Individual Differences; Laboratories; Linear Models; Machine Learning; Malignant Neoplasms; Malignant neoplasm of prostate; Measures; Medical; Methods; Modeling; Outcome; Pancreas; Patients; Performance; Population; Probability; ROC Curve; Research; Research Design; Risk Factors; Sampling; Scheme; Science; Selection for Treatments; Sensitivity and Specificity; Southwest Oncology Group; Specific qualifier value; Staging; Statistical Methods; Testing; Time; Treatment Cost; Women' s Health; base; burden of illness; cancer genome; case control; clinical practice; cohort; cost; design; disease classification; disease diagnosis; disorder risk; genome wide association study; improved; interest; malignant breast neoplasm; novel; patient population; programs; public health relevance; randomized trial; response; screening; tool; treatment effect

DESCRIPTION (provided by applicant): Recent advances in the laboratory sciences have led to the discovery of a large number of candidate biomarkers, which hold great potential for disease diagnosis and treatment. At this time, an important research bottleneck is the lack of well-developed statistical methods for effectively using these candidate biomarkers to enhance clinical practice. It is our goal to develop new tools to select, combine, and evaluate biomarkers for disease classification and treatment selection. Classification markers predict an individual's disease outcome and are useful for the detection of diseases at an early stage when a treatment is most effective. Research proposed in Aim 1 seeks to select and combine markers to improve the classification performance in disease screening and diagnosis. Treatment selection markers predict a patient's response to different therapies and allow for the selection of a therapy that has the best predicted outcome. Aim 2 seeks to develop marker-based treatment selection rules to maximize the benefit to the patient population. A biomarker that is useful for guiding treatment decision to the general population will have different values to different patients due to individual differences in their response to treatment and in their tolerance of the disease harm and treatment cost. Aim 3 seeks to develop a new graphical tool to customize the evaluation of a biomarker for aiding treatment decision based on personal characteristics. Our statistical methods will apply broadly to general medical fields. In particulr, we will apply these methods to analyze several cancer studies including (1) biomarker studies for prostate cancer and pan- creatic cancer from the Early Detection and Research Network; (2) the Women's Health Initiative breast cancer genome-wide association study; and (3) the Oncotype-Dx breast cancer study from the Southwest Oncology Group. Programs and algorithms developed in this proposal will be made available to public.

描述（由申请人提供）：实验室科学的最新进展导致大量候选生物标志物的发现，这些生物标志物在疾病诊断和治疗方面具有巨大的潜力。目前，一个重要的研究瓶颈是缺乏成熟的统计方法来有效地利用这些候选生物标志物来增强临床实践。我们的目标是开发新工具来选择、组合和评估生物标志物，以进行疾病分类和治疗选择。分类标记可以预测个体的疾病结果，并且有助于在治疗最有效的早期阶段检测疾病。目标 1 中提出的研究旨在选择和组合标记，以提高疾病筛查和诊断中的分类性能。治疗选择标记可以预测患者对不同治疗的反应，并允许选择具有最佳预测结果的治疗。目标 2 寻求制定基于标记的治疗选择规则，以最大限度地提高患者群体的利益。由于对治疗的反应以及对疾病危害和治疗费用的耐受性存在个体差异，可用于指导一般人群治疗决策的生物标志物对于不同的患者具有不同的价值。目标 3 寻求开发一种新的图形工具来定制生物标志物的评估，以帮助根据个人特征做出治疗决策。 我们的统计方法将广泛应用于一般医学领域。特别是，我们将应用这些方法来分析多项癌症研究，包括（1）来自早期检测和研究网络的前列腺癌和胰腺癌生物标志物研究； (2) 妇女健康倡议乳腺癌全基因组关联研究； (3) 西南肿瘤学组的 Oncotype-Dx 乳腺癌研究。本提案中开发的程序和算法将向公众开放。