Cancer Diagnosis Program | Biorepositories and Biospecimen Research Branch

Search BRD Papers

Diagnostic concordance among pathologists interpreting breast biopsy specimens.

Author(s): Elmore JG, Longton GM, Carney PA, Geller BM, Onega T, Tosteson AN, Nelson HD, Pepe MS, Allison KH, Schnitt SJ, O'Malley FP, Weaver DL

Publication: JAMA, 2015, Vol. 313, Page 1122-32

PubMed ID: 25781441 PubMed Review Paper? No

Purpose of Paper

Conclusion of Paper

Studies

1. Study Purpose

   This study assessed how patient age and breast density, pathologist workload, and practice size and academic affiliation influences the diagnostic concordance of breast cancer among pathologists. A single H&E stained slide was chosen for each of the 240 breast biopsies evaluated, and an initial diagnosis of benign without atypia (30% of biopsies, including 10% non-proliferative and 20% proliferative without atypia), atypia (30% of biopsies), DCIS (30% of biopsies), and invasive carcinoma (10% of biopsies) was established based upon the consensus of three pathologists. A 60 case subset was then re-evaluated in a blind study by 126 different pathologists with more than 1 year of experience diagnosing breast biopsies. 

   Summary of Findings:

   The initial three pathologists agreed on a diagnosis for 75% of the 240 cases examined, and for 90.3% of the 720 interpretations (650 of 720). Diagnostic agreement for the same cases was poorer among the 126 additional participting pathologists, which produced a consensus diagnosis in 75.3% of interpretations. Diagnostic agreement among pathologists was stronger for certain diagnoses; for example,  agreement was 96% among specimens diagnosed as invasive breast cancer, 87% for DCIS, and 82% for those diagnosed as benign without atypia, but only 54% for those with atypia. Further, for atypia the underestimation rate (where the actual diagnosis was benign without atypia) was 19% and the overestimation rate (where the actual diagnosis was DCIS) was 27%. Diagnostic agreement rates were lower among specimens with greater  breast density (P<0.001), specifically when pathologists ranked the case as difficult or borderline between diagnostic categories or when the pathologist stated they were less confident in their interpretation or would seek a second opinion. Concordance with the reference diagnosis occurred most frequently when the pathologists worked in an academic setting (OR=0.799, P=0.006), worked in a practice with 10 or more pathologists (OR=0.849, P=0.0399), or evaluated 10 or more breast cases per week (OR=0.799, P=0.006). When the patient was 40-49 years old rather than over 50 years old, the rate of over-interpretation (i.e., atypia as DCIS) was higher and the under-interpretation (i.e., DCIS as atypia) rate was lower.

   Biospecimens

   • Tissue - Breast

   Preservative Types

   • Formalin

   Diagnoses:

   • Neoplastic - Benign
   • Neoplastic - Carcinoma

   Platform:

   Analyte	Technology Platform
   Morphology	H-and-E microscopy

   Pre-analytical Factors:

   Classification	Pre-analytical Factor	Value(s)
   Preaquisition	Diagnosis/ patient condition	Benign without atypia Atypia DCIS Invasive carcinoma
   Preaquisition	Patient age	40-49 years ≥50 years
   Preaquisition	Prognostic factor	Varying breast density
   H-and-E microscopy Specific	Data handling	126 different Pathologists Pathologist in academic setting Pathologist in non-academic setting Pathologist in practice with fewer than 10 pathologists Pathologist who evaluates 10 or more breast cases a week Pathologist who evaluates fewer than 10 breast cases a week

   View more