Statistical Methods for Evaluation and Validation of Diagnostic Tests

生物统计学家、放射学家和医学物理学家的多学科努力旨在开发 创新的统计方法，用于评估和验证新的和低成本的诊断模式。 具体目的是（1）确定诊断技术非劣效性的统计方法;（2） 用于快速准确评估时间关联的统计设计和分析方法 新的放射学标志物与疾病结局的治疗效果;（3）应用统计学方法， 临床研究方法和开发用户友好的软件。具体目标1研究评价非 基于各种条件下受试者工作特征曲线的新诊断测试的劣性 条件（缺失数据、混合数据类型和聚类数据）、协变量的影响（ 调整协变量并识别新检验劣效、上级或 贝叶斯效用模型（Bayesian utility model）具体目标2研究了一种新的横断面病例对照短， 长期随访（CSCCSTFU）设计和分析方法，以估计新的 诊断测试罕见的结果。它还将提出估计治疗下限的方法 对通过新测试诊断的患者有效。数学理论和模拟研究将用于 了解CSCCSTFU设计的估计比值比可以推广的条件。 具体目标3将具体目标1的结果应用于骨质疏松性骨折研究的数据， 骨质疏松症和超声研究，以测试是否和当定量超声，外周骨 骨密度（BMD）和生化骨标记物可作为非劣效或等效技术， 髋部和脊柱骨密度评估腰椎骨折风险它还将利用具体目标2的结果， 确定乳腺体积密度和结构参数是否为新的单X线 在一项对15，000名妇女进行的研究中，吸收测量技术可以前瞻性地预测乳腺癌。 一个由国际公认的骨质疏松症和乳腺癌研究领导人组成的咨询委员会将 向研究团队提供关于与医生联系和辅助研究机会的建议。研究成果 将通过出版临床和医学领域的软件、研究和辅导论文进行传播， 统计期刊和影响专业组织的立场文件。

This multi-disciplinary effort by biostatisticians, radiologists, and medical physicists aims to develop innovative statistical methods for evaluation and validation of new and low cost diagnostic modalities. Specific Aims are (1) statistical methods for determining non-inferiority of diagnostic techniques; (2) statistical design and analysis methods for rapid and accurate evaluation of the chronological association and treatment efficacy of new radiological markers with disease outcome; (3) applying the statistical methods to clinical research and developing user-friendly software. Specific Aim 1 studies evaluation of non- inferiority for a new diagnostic test based on the receiver operating characteristics curve under various conditions (missing data, mixed data types, and clustered data), effects of covariates (non-inferiority after adjustment of covariates and identification of subgroups within which the new test is inferior, superior or equivalent), and Bayesian utility models. Specific Aim 2 studies a new cross-sectional case-control short- term follow-up (CSCCSTFU) design and analysis methods to estimate the chronological odds ratio of a new diagnostic test for a rare outcome. It will also propose methods to estimate the lower boundary of treatment efficacy for patients diagnosed by the new test. Mathematical theory and simulation studies will be used to understand the conditions in which the estimated odds ratio of the CSCCSTFU design can be generalized. Specific Aim 3 will apply results from Specific Aim 1to data from the Study of Osteoporotic Fractures and the Osteoporosis and Ultrasound Study to test whether and when quantitative ultrasound, peripheral bone mineral density (BMD), and biochemical bone markerscan serve as non-inferior or equivalent techniques to BMD of hip and spine for assessing osteoporotic fracture risks. It will also use results of Specific Aim 2 to determine whether the volumetric breast density and structural parameters of the new single X-ray absorptiometry technique can prospectively predict breast cancer in an on-going study of 15,000 women. An advisory committee of internationally recognized leaders in osteoporosis and breast cancer research will advise the research team on outreach to physicians and opportunities for ancillary studies. Research results will be disseminated through publication of software, research and tutorial papers in both clinical and statistical journals and by influencing position papers of professional organizations.