Comparative study of computer-assisted image analysis and light-microscopically determined estrogen receptor status of breast carcinomas.

Abstract

Efficient tamoxifen treatment of breast cancer patients depends on the degree to which pathologists are consistently able to use the so-called "modified quickscore" method to differentiate between negative and low-grade positive scores based on the current four-grade scale of the estrogen receptor (ER) status of the tumors. To quantitatively test pathologists' ability to adequately render correct readings of the ER status of breast tumors. The ER status of breast carcinomas was estimated by two pathologists and measured by computer-supported analysis in sections stained with ER antibody by the immunoperoxidase technique. Levels of agreement between the examiners' semiquantification and histomorphometrically gauged measurements were compared statistically. The kappa coefficients were 0.28 on a case-by-case collation of the grades of nuclear staining, 0.52 on applying binary categories of positively versus negatively stained cell nuclei, and 0.89 using binary categories with a weighted score of 1.2 to separate ER-negative from ER-positive breast carcinomas. The findings suggest that an optimum rating system is achieved by assigning tumors with a weighted score of 1.2 or less to the ER-negative and those with a weighted score of 1.3 or more to the ER-positive category of breast carcinomas.