Predicting Breast Cancer Risk after Benign Percutaneous Biopsy

项目摘要 超过100万妇女通过经皮活检被诊断为良性乳腺疾病（BBD） 每年在美国，并将受益于改善乳腺癌（BC）的风险信息，因为他们面临 筛查和预防决策。BBD与BC风险增加相关，范围为1.5-2.0倍 从最轻微的类别到最严重类型的四倍。然而，这些风险适用于妇女群体， 而不是个人，在BBD类别中，个人风险差异很大。此外，我们还表明， 乳腺癌（BC）风险预测模型，例如“Gail模型”，在患有BBD的妇女中表现不佳。 以前，我们开发了用于手术活检的BBD-BC模型，该模型提供了个体风险估计 基于自我报告的因素，BBD范围和严重程度的详细特征，以及 周围组织学结构（终末导管小叶单位）的退化（收缩和消失 （TDLU）），其中大多数BC前体产生。BBD-BC在预测BC风险方面优于Gail模型。 然而，鉴于放射学引导的小（经皮）活检已在很大程度上取代了手术 为了诊断，需要基于这种活检方法的新模型。此外， 乳腺摄影密度作为一个重要的乳腺癌风险因素，开发新的方法来评估TDLU 在常规处理的临床样品中生物标志物的退化和增加的使用提供了机会， 为经皮活检诊断为BBD的女性开发改进的BC风险预测工具。的 本项目的目标是为经皮穿刺诊断为BBD的女性建立BC风险预测工具 可以在不同人群中验证并在临床上实施的活检。我们建议发展一个 马约的队列包括> 7，000名通过经皮活检被诊断患有BBD的女性，其中 公元前400年后发展起来的。我们将开发一个模型来预测BC，其中包括BBD-BC模型中的因素， 手术活检。我们还将评估乳腺摄影密度，测量体积和面积，使用 验证方法。我们将确定可应用于BBD活检的免疫组化标记物， 预测未来发生BC的风险，并评估新的NanoString RNA检测方法， 相关基因作为反映癌样特征、增殖和突变样特征的复合“标记”， TP 53肿瘤抑制基因的重要性。最后，我们将建立一个流行病学“病例队列” 这包括来自整个队列的随机女性子集（n~500）和所有发展的女性 侵袭性BC（n~250）。我们将在750名妇女的病例队列中评估BC风险预测， 使用新型机器学习方法，在没有生物标志物和有生物标志物的情况下，风险模型的性能 与更典型的统计模型相比，它提供了优势。利用这些数据，我们将建立一个绝对的 风险预测模型，可在其他人群中进行测试。

Project Abstract More than one million women are diagnosed with benign breast disease (BBD) by percutaneous biopsy annually in the U.S. and would benefit from improved breast cancer (BC) risk information as they face screening and prevention decisions. BBD is associated with increases in BC risk, ranging from 1.5-2.0 times for least severe categories to fourfold for most severe types. However, these risks apply to groups of women, not individuals, and individual risk varies considerably within BBD categories. Further, we have shown that breast cancer (BC) risk prediction models, such as the “Gail Model”, perform poorly among women with BBD. Previously, we developed the BBD-BC model for surgical biopsies, which provides individual risk estimates based on self-reported factors, detailed characteristics of BBD extent and severity, and assessment of involution (shrinkage and disappearance) of surrounding histologic structures (terminal duct lobular units (TDLUs)) from which most BC precursors arise. BBD-BC outperforms the Gail Model in predicting BC risk. However, given that radiologically-guided small (percutaneous) biopsies have largely replaced surgical biopsies for diagnosis, a new model based on this biopsy approach is needed. Further, the emergence of mammographic density as an important BC risk factor, development of novel methods to assess TDLU involution and increased use of biomarkers in routinely processed clinical samples offer an opportunity to develop an improved BC risk prediction tool for women with percutaneous biopsy diagnoses of BBD. The goal of this project is to build a BC risk prediction tool for women with BBD diagnosed on percutaneous needle biopsy that could be validated in diverse populations and implemented clinically. We propose to develop a cohort at Mayo that includes >7,000 women who were diagnosed with BBD on a percutaneous biopsy of whom >400 later developed BC. We will develop a model to predict BC that includes factors in the BBD-BC model for surgical biopsies. We will also assess mammographic density, measured as a volume and area, using validated methods. We will identify immunohistochemical markers that can be applied to BBD biopsies to predict future risk of developing BC and evaluate novel NanoString RNA assays, which measure expression of related genes as composite “signatures” reflecting cancer-like characteristics, proliferation, and a mutation-like score for the important TP53 tumor suppressor gene. Finally, we will develop an epidemiologic “case-cohort” that includes a random subset of women from the full cohort (n~500) and all the women that developed invasive BC (n~250). We will evaluate BC risk prediction in this case-cohort of 750 women to evaluate performance of risk models without biomarkers and with biomarkers using novel machine learning approaches that offer strengths compared with more typical statistical models. Using these data, we will build an absolute risk prediction model for the full cohort that can be tested in other populations.