Statewide study of diagnostic agreement in breast pathology.

BACKGROUND This study assessed the degree of diagnostic agreement among community-based general pathologists reading slides of representative breast tissue specimens and tested whether diagnostic variability is associated with type of breast specimen (e.g., core needle or excisional biopsy) or slide quality. METHODS Twenty-six of the 44 eligible pathologists working at community-based pathology practices in New Hampshire participated. Each pathologist evaluated slides of breast tissue obtained from 30 case subjects randomly selected from a statewide breast pathology database. The diagnostic categories used were benign, benign with atypia, noninvasive malignant, and invasive malignant. The levels of agreement (i.e., kappa coefficients) for the diagnoses were assessed. RESULTS Agreement was high among pathologists for assignment of diagnostic category (kappa coefficient = 0.71) and was nearly perfect for their selection of benign versus malignant categories (kappa coefficient = 0.95). There was less agreement for the categories of noninvasive malignant and benign with atypia (kappa coefficients of 0.59 and 0.22, respectively). There was no apparent relationship between levels of diagnostic agreement and specimen type or perceived slide quality. CONCLUSIONS Diagnostic agreement for breast tissue specimens is high overall among community-based pathologists, but clinically relevant disagreements may occur in the assessment of noninvasive malignant diagnoses. The establishment of reread policies for certain diagnostic categories may reduce the possibility that diagnostic misclassification will lead to overtreatment or undertreatment. The high degree of diagnostic reproducibility for invasive cancerous lesions of the breast suggests that it is unnecessary for a central review of these lesions in national cancer trials.