The use of nuclear morphometry to predict response to therapy in Wilms' tumor.

Using nuclear morphometric analysis, a retrospective study was made of 27 patients with Wilms' tumor. The blinded group consisted of 17 patients with Stage I and II favorable-histology Wilms' tumors who did not respond to therapy and ten patients with Stage III and IV favorable-histology Wilms' tumors who did respond. In this complex group of patients, multivariate analysis (with several morphologic descriptors) was used to predict which patients responded to therapy. No single-shape descriptor predicted the response to therapy (P greater than 0.5). However, three shape descriptors: maximum ellipticity (ME), standard error of bending energy (SEBE), and the range of chain code-maximum peak (RCCM), with the multivariate formula of (1131 x SEBE) + (-7 x ME) + (50 x RCCM) + 7.18, separated the two groups (P less than 0.004). This result was statistically significant. Using this multivariate equation and a cutoff value of 0.6 units, this test yielded a sensitivity of 94% and a specificity of 70%. This result, in a complex group of patients, suggests that nuclear morphometry may be useful in the initial assessment of patients with Wilms' tumor and warrants further analysis.