Multi-center evaluation of digital breast tomosynthesis with synthesized two-dimensional mammography for breast cancer screening

摘要 数字乳腺X射线摄影（DM）结合数字乳腺断层合成摄影（DBT）已成为一种有前途的 原发性乳腺癌筛查技术，与 单独筛查DM。然而，DM/DBT联合研究的总辐射剂量通常是 单独用DM进行筛查，引起了对筛查一生中累积剂量的关注。2013年 FDA批准的合成“类2D”（s2 D）图像，来源于DBT采集期间采集的数据，作为 在DBT筛查中替代传统的“剂量”DM成像。我们的初步单点数据和 早期的阅片者研究表明，s2 D/DBT筛查可能会进一步降低召回率和活检率， DM/DBT，虽然浸润性乳腺癌检出率高，但导管癌检出率降低 原位（DCIS）。此外，s2 D/DBT筛查的乳腺X线摄影密度分配发生了变化 认为“密集”的患者较少，这可能会影响风险评估和基于密度的临床 关于补充筛查的决定，因为乳房密度通知是法律规定的， 美国的一半。我们提出了一项多中心观察性研究，具有足够的把握度， 表征s2 D/DBT筛查性能的多个方面。我们将利用一个自然实验， 佛蒙特大学和宾夕法尼亚大学的早期采用放射学设施 已从DM/DBT过渡到s2 D/DM进行常规筛查的PROSPR研究中心。我们 具体目的是1）确定临床筛选性能（例如，敏感性和特异性） 使用s2 D/DBT进行癌症筛查; 2）评估通过检测特定恶性和良性病变的比率， 用s2 D/DBT进行乳腺癌筛查;和3）确定乳房X线摄影乳腺密度的分布 S2 D/DBT检查。对于所有目标，我们将比较s2 D/DBT筛查性能与 在采用s2 D/DBT之前，在同一机构进行DM/DBT筛查。殷切地盼望 样本量约为86，000例s2 D/DBT检查和98，000例DM/DBT检查。佛蒙特大学 和宾夕法尼亚大学的PROSPR研究中心前瞻性地收集了乳腺癌 自2011年以来，以标准化方式筛选数据，以便进行汇总分析。我们的注册表包含考试- 水平乳房成像公共数据元素，包括关于检查指示的高质量信息（即， 筛查与诊断）、乳房X线摄影视图、评估和乳腺密度，与病理学相关 数据库和全州范围的癌症登记处，从而允许良性和恶性肿瘤的调查 诊断。该项目的结果将为妇女评估筛查方案提供所需的证据， 提供者评估是否采用s2 D/DBT，医疗保健支付者考虑报销率，以及 政策制定者在这项新技术的背景下讨论乳腺癌筛查建议。

ABSTRACT Digital mammography (DM) combined with digital breast tomosynthesis (DBT) has emerged as a promising primary breast cancer screening technology with improvements in specificity and sensitivity compared to screening with DM alone. However, the total radiation dose of a combined DM/DBT study is often twice that of screening with DM alone, raising concerns about the cumulative dose over a lifetime of screening. In 2013, the FDA approved synthetic “2D-like” (s2D) images, derived from data acquired during the DBT acquisition, as a replacement for the conventional “dose” DM imaging in DBT screening. Our preliminary single site data and early reader studies suggest that s2D/DBT screening may further reduce recall and biopsy rates compared to DM/DBT, while providing high invasive breast cancer detection rates but reduced detection of ductal carcinoma in situ (DCIS). In addition, the mammographic density assignments from s2D/DBT screening changed significantly with less patients deemed “dense”, which could impact risk assessment and density-based clinical decisions regarding supplemental screening since breast density notification is mandated by laws in more than half of US states. We propose a multi-center observational study with adequate power to comprehensively characterize multiple aspects of s2D/DBT screening performance. We will leverage a natural experiment at early adopter radiology facilities participating in the University of Vermont and University of Pennsylvania PROSPR Research Centers that have transitioned from DM/DBT to s2D/DM for routine screening. Our specific aims are 1) To determine the clinical screening performance (e.g., sensitivity and specificity) of breast cancer screening with s2D/DBT; 2) To evaluate the rates of specific malignant and benign lesions detected by breast cancer screening with s2D/DBT; and 3) To determine the distribution of mammographic breast density assessments on s2D/DBT exams. For all aims, we will compare s2D/DBT screening performance to the performance of DM/DBT screening at the same facilities prior to their adoption of s2D/DBT. We anticipate a sample size of approximately 86,000 s2D/DBT exams and 98,000 DM/DBT exams. The University of Vermont and University of Pennsylvania PROSPR Research Centers have prospectively collected breast cancer screening data since 2011 in a standardized fashion to permit pooled analyses. Our registries contain exam- level breast imaging common data elements, including high quality information on indication of exam (i.e., screening vs. diagnostic), mammographic views, assessments, and breast density, linked to pathology databases and statewide cancer registries thereby permitting investigation of both benign and malignant diagnoses. The results of this project will provide needed evidence for women evaluating screening options, providers assessing whether to adopt s2D/DBT, health care payers considering reimbursement rates, and policy makers debating breast cancer screening recommendations in the context of this new technology.