Computerized image analysis of p53 and proliferating cell nuclear antigen expression in benign, hyperplastic, and malignant endometrium

Context-The endometrium is an intrinsically dynamic tissue with great capab ility for regeneration and proliferation; consequently, there is some overl ap between features seen in benign, premalignant, and malignant lesions. Th is leads to marked intrabiopsy, interbiopsy, and interobserver variability. Objective.-We studied the specificity and sensitivity of computerized image analysis of molecular markers to evaluate its potential use as a diagnosti c tool. Design.-Specimens from 100 patients were examined and the following histolo gic diagnoses were assigned: proliferative endometrium (n = 10), secretory endometrium (n = 10), endometrial hyperplasia (n = 40; 30 with no atypia, 1 0 with atypia), and carcinoma (n = 40; 20 endometrioid, 10 serous, and 10 c lear cell). All cases were evaluated immunohistochemically for p53 and prol iferating cell nuclear antigen (PCNA) expression. Computerized image analys is was performed with a CAS 200 digital analyzer. Results.-Expression of p53 was found only in carcinomas (65%) and endometri al hyperplasia with atypia (30%). Expression of p53 was higher in the poor prognostic categories (serous carcinoma and clear cell carcinoma) than in e ndometrioid carcinoma. In endometrioid carcinoma, p53 expression correlated with grade. Proliferating cell nuclear antigen showed a similar pattern of results to p53 in the various carcinoma subtypes and endometrioid carcinom a grades. Endometrial hyperplasia PCNA values were the lowest among all the groups. Both carcinomas and proliferative endometrium showed higher glandu lar and stromal PCNA values, significantly different from endometrial hyper plasia with atypia. In proliferative endometrium, stromal PCNA was the high est among all of the groups. The p53 and PCNA results correlated with each other for carcinoma. Conclusions.-Computerized image analysis correlates well with the establish ed morphologic groups of endometrial pathology and yields results consisten t with previous studies. Owing to its higher degree of sensitivity, compute rized image analysis is of potential use in cases of diagnostic dilemmas an d can help objectively allocate the case in the correct category (eg, proli ferative endometrium vs endometrial hyperplasia, endometrial hyperplasia wi th atypia vs endometrioid carcinoma). It is particularly useful in the eval uation of stromal changes.