Reproducibility of Needle Core Biopsy Diagnosis of Breast Papillary Lesions
LJ Elavathil, K Onuma, F Mothafar, N Akhtar-Danesh, A Lytwyn. Pathology & Molecular Medicine, Hamilton, Canada; Clinical Epidemiology & Biostatistics, McMaster University, Hamilton, ON, Canada
Background: Information on reproducibility of breast papillary lesion core biopsy diagnosis is limited. As part of a study assessing accuracy of core biopsies to identify breast papillary lesions requiring surgical excision, we evaluated interobserver agreement for classifying papillary lesions on needle core biopsy using a diagnostic algorithm.
Design: We designed a diagnostic algorithm based on WHO classification of papillary lesions and Page criteria for ADH and DCIS. Seven categories, capturing potential diagnostic thresholds for surgical excision, were created: (1) benign papilloma with no ductal proliferation; benign papilloma with (2) 10%, (3) >10% - <50%, (4) 50% UDH; (5) papilloma with ADH; (6) papilloma with DCIS; and (7) papillary carcinoma. Two senior and 2 junior breast pathologists piloted the algorithm by together reviewing a training set of 25 breast papillary core biopsies. We subsequently searched the pathology database from 1997-2006 for needle core biopsy diagnoses of papillary breast lesions. We excluded cores with concurrent invasive carcinoma, or with ADH or DCIS outside of the papillary lesion. Study cases, consisting of slides stained with H&E and for smooth muscle myosin heavy chain, were circulated to the 4 pathologists, for independent assessment using the diagnostic algorithm.
Results: The study consisted of 164 cases. There was substantial agreement for the 7 categories, with weighted k (kappa statistic) ranging from 0.69 to 0.80. Agreement was almost perfect for papillary carcinoma (k=0.85), moderate for papilloma with no ductal cell proliferation (k=0.51), slight for papilloma with DCIS (k=0.14), and fair for all other categories (k=0.23 to 0.36). Agreement was substantial when diagnoses were dichotomized into (a) ADH, DCIS or papillary carcinoma present vs benign papilloma with or without UDH (k=0.72; (b) benign papillomas with up to 10% or 50% UDH vs all other categories (k=0.77, 0.77, respectively), and moderate when threshold was (c) papilloma with no ductal proliferation vs all other categories (k=0.52). There was unanimous agreement for category assignment in 75% of cases for (a), 79% for (b) and 67% for (c).
Conclusions: Using our algorithm, we were able to achieve substantial agreement on needle core biopsies for potential diagnostically relevant categories of papillary breast lesions among pathologists with varied experience in breast pathology.
Monday, March 9, 2009 8:45 AM
Platform Session: Section B, Monday Morning