Engelberg Jesse A, Retallack Hanna, Balassanian Ronald, Dowsett Mitchell, Zabaglo Lila, Ram Arishneel A, Apple Sophia K, Bishop John W, Borowsky Alexander D, Carpenter Philip M, Chen Yunn-Yi, Datnow Brian, Elson Sarah, Hasteh Farnaz, Lin Fritz, Moatamed Neda A, Zhang Yanhong, Cardiff Robert D
Center for Comparative Medicine, University of California Davis, Davis, CA 95616.
School of Medicine, University of California San Francisco, San Francisco, CA 94143.
Hum Pathol. 2015 Nov;46(11):1694-704. doi: 10.1016/j.humpath.2015.07.008. Epub 2015 Jul 29.
Hormone receptor status is an integral component of decision-making in breast cancer management. IHC4 score is an algorithm that combines hormone receptor, HER2, and Ki-67 status to provide a semiquantitative prognostic score for breast cancer. High accuracy and low interobserver variance are important to ensure the score is accurately calculated; however, few previous efforts have been made to measure or decrease interobserver variance. We developed a Web-based training tool, called "Score the Core" (STC) using tissue microarrays to train pathologists to visually score estrogen receptor (using the 300-point H score), progesterone receptor (percent positive), and Ki-67 (percent positive). STC used a reference score calculated from a reproducible manual counting method. Pathologists in the Athena Breast Health Network and pathology residents at associated institutions completed the exercise. By using STC, pathologists improved their estrogen receptor H score and progesterone receptor and Ki-67 proportion assessment and demonstrated a good correlation between pathologist and reference scores. In addition, we collected information about pathologist performance that allowed us to compare individual pathologists and measures of agreement. Pathologists' assessment of the proportion of positive cells was closer to the reference than their assessment of the relative intensity of positive cells. Careful training and assessment should be used to ensure the accuracy of breast biomarkers. This is particularly important as breast cancer diagnostics become increasingly quantitative and reproducible. Our training tool is a novel approach for pathologist training that can serve as an important component of ongoing quality assessment and can improve the accuracy of breast cancer prognostic biomarkers.
激素受体状态是乳腺癌管理决策的一个重要组成部分。免疫组化4分法是一种综合激素受体、HER2和Ki-67状态的算法,可为乳腺癌提供半定量预后评分。高准确性和低观察者间差异对于确保准确计算该评分很重要;然而,此前很少有人致力于测量或减少观察者间差异。我们开发了一种基于网络的培训工具,称为“核心评分”(STC),使用组织微阵列来培训病理学家对雌激素受体(使用300分的H评分)、孕激素受体(阳性百分比)和Ki-67(阳性百分比)进行视觉评分。STC使用从可重复的手工计数方法计算得出的参考评分。雅典娜乳腺健康网络的病理学家和相关机构的病理住院医师完成了这项练习。通过使用STC,病理学家提高了他们对雌激素受体H评分、孕激素受体和Ki-67比例的评估,并证明了病理学家评分与参考评分之间具有良好的相关性。此外,我们收集了有关病理学家表现的信息,这使我们能够比较个体病理学家以及一致性测量结果。病理学家对阳性细胞比例的评估比他们对阳性细胞相对强度的评估更接近参考值。应采用仔细的培训和评估来确保乳腺生物标志物的准确性。随着乳腺癌诊断越来越趋于定量和可重复,这一点尤为重要。我们的培训工具是一种用于病理学家培训的新颖方法,可作为持续质量评估的重要组成部分,并可提高乳腺癌预后生物标志物的准确性。