Validation of the novel indices of nuclear pleomorphism, polarity and spatial distribution in the grading of urothelial carcinoma.

Objective: To evaluate the role of the novel histomorphometric variables of pleomorphism, orientation and variability of spatial cytologic organization to objectively quantify histopathologic grade in urothelial carcinoma.

Study Design: As a validation model, we chose 51 papillary urothelial carcinomas. Thirty-six (70%) were low and 15 (29%) high grade tumors (World Health Organization grading system). Thirty-one cases (61%) were noninvasive, 12 (23%) exhibited lamina propria invasion, and 8 (16%) invaded the deep smooth muscle. Histomorphometric measurements were performed on noninvasive areas only. Pleomorphism was characterized by the anisokaryosis index (AIX), based on the autocorrelation function applied to classic nuclear shape descriptors. Additional indices of pleomorphism included standard deviations (sd) and coefficients of variation of these shape descriptors. Loss of polarity was assessed by the orientation index (ORX), based on Fourier transformation. Intraepithelial nuclear spatial distribution index (NSDX) was also computed.

Results: Low grade tumors exhibited less pleomorphism, higher orientation indices and a more homogeneous spatial distribution than did high grade tumors (area-AIX: P = .004, ORX: P < .0001, NSDX: P = .001). Multivariate analysis revealed the significant discriminators of grade to be nuclear size and orientation (size: max-diam, P < .0001; width: P < .0001; orientation: ORX: P < .0001). A discriminant score combining these independent variables distinguished between low and high grades in 98% of cases. The method was successfully validated using a testing sample of 40 new patients (accuracy, 94%). Lamina propria invasion was independently predicted by nuclear pleomorphism (width-sd: P = .0001, sensitivity = 64%, specificity = 91%). Muscle invasion was independently predicted by nuclear area (P < .0001) and pleomorphism (area-sd: P < .0001, max-diam-sd: P < .0001, width-AIX: P < .0001, sensitivity = 87%, specificity = 100%).

Conclusion: These novel morphometric methods may serve as more objective methods of histopathologic grading and may contribute to the development of automated systems for quantitative grading of stratified and transitional epithelial neoplasms.