In search for the best qualitative microscopical or morphometrical predictor of oestrogen receptor in breast cancer.

The correlation between oestrogen receptor (OR), qualitative and quantitative microscopical features has been studied in 198 breast cancers, 60 patients were younger than 50 years of age, 138 were older than 50. This age distinction was made because of the obvious age dependency of the OR-content, with 50 years as the cut-off point. Compared with elastosis grade, the morphometric features have a much lower percentage of doubtful cases (22.2 vs. 40.4%) with a higher percentage overall correct classifications (55.6 vs. 45.4%). In terms of sensitivity and specificity, selection of the best single predictor depends on the age of the patient. Mean nuclear area has the highest combination of sensitivity and specificity (88.9 and 80.0%) in woman younger than 51, while elastosis grade is the best single predictor in patients older than 50 (sensitivity: 91.2%; specificity: 70.6%). Using multivariate analysis, a combination of mean and standard deviation of the nuclear area results in 83.3% correct classifications with only 5% doubtfuls in the younger age group. With more features in the analysis, no false negatives and only 13.3% false positives is the most optimal result. In the older age group, a decision tree consisting of elastosis grade and mean nuclear area gives the best results. Subsequent investigation of mean nuclear area in elastosis grade 0 cancers gives a considerable reduction of the false negatives, thus increasing the specificity to 94.6% and the correct negative predictions to 72.7%. It is concluded that selective morphometry gives a considerable enhancement of the histopathological prediction capacities of oestrogen receptor in breast cancer.