Read the full EBM Focus and earn CME credit.
Reference: JAMA 2015 Mar 17;313(11):1122
Breast cancer is the most common form of cancer and second leading cause of cancer deaths in women in the United States (J Natl Cancer Inst 2011 May 4;103(9):714). Increased mammography screening has increased the diagnosis of early stage breast cancer over the last 30 years, but overdiagnosis is common (N Engl J Med 2012 Nov 22;367(21):1998). While approximately 25% of breast biopsies have been reported to diagnose invasive carcinoma, most biopsy specimens result in a diagnosis of benign or precancerous lesions (Cancer 2006 Feb 15;106(4):732). A recent study evaluated the rate of diagnostic inaccuracies of 240 breast biopsy samples evaluated by 126 pathologists from across the United States.
Excisional or core needle biopsy samples were randomly selected from a registry of 19,498 cases based on the original diagnosis, patient age, breast density, and biopsy type. The consensus diagnosis from a panel of 3 expert pathologists served as the reference standard, with a consensus diagnosis of invasive breast cancer in 10%, ductal carcinoma in situ (DCIS) in 30%, atypical hyperplasia in 30%, and benign without atypia in 30%. All 126 individual pathologists participating in the study had ≥ 1 year of experience interpreting breast biopsy specimens and planned to continue for ≥ 1 additional year. Each participating pathologist was randomized to 1 of 4 test sets containing 60 samples each, with a single slide for each case. Of the 126 pathologists randomized, 115 (91%) completed the assessment of all 60 slides in their assigned test set and these 6,900 interpretations were compared to the expert consensus for each specimen.
Overall concordance between the expert panel diagnosis and the individual pathologist interpretation was 75.3%. The rate of concordance as well as the rate of overinterpretation and underinterpretation can be found in the table below.
| Reference Diagnosis | Rate of Concordance | Rate of Underinterpretation | Rate of Overinterpretation |
| Invasive Carcinoma | 96% | 4% | —- |
| DCIS | 84% | 13% | 3% |
| Atypical hyperplasia | 48% | 35% | 17% |
| Benign without atypia | 87% | —- | 13% |
While the rate of concordance was very high for invasive carcinoma, with only 4% of cases misinterpreted as DCIS, the concordance rate drops considerably for atypical hyperplasia, with most misinterpretations resulting in an underinterpretation of the disease as a benign biopsy. Misinterpretation was not limited to a few specific pathologists or specific cases, but was found to be more widely distributed. Higher rates of disagreement were significantly associated with increased breast density and pathologist characteristics including lower weekly case volume, smaller practices, and nonacademic settings.
The results of this study suggest that while pathologists generally agree upon interpretation of invasive carcinoma samples, their interpretations of precancerous lesions may be more variable. Pathologists based their diagnosis on a single specimen, however, and in practice pathologists usually review multiple specimens before establishing a diagnosis. This limitation may have contributed to the variability observed, especially for diagnoses with less standardized criteria. Decreasing variability in biopsy diagnosis is important as overinterpretation may lead to unnecessary treatment and underinterpretation may withhold necessary treatments or decrease surveillance in patients with precancerous lesions. While further studies may be necessary to determine ways to better identify atypical histological features, a second opinion may increase the reliability of diagnosis and help prevent incorrect treatment.
For more information, see the Breast cancer in women topic in DynaMed.