Purpose: This study was designed to evaluate the ability of a previously published nuclear morphometry discriminant function to predict disease-free survival in patients with Wilms' tumor. Patients and Methods: We identified 218 patients with stage I-IV Wilms' tumor of favorable histology who were entered onto the National Wilms' Tumor Study (NWTS) between January 1, 1990 and April 15, 1994. The nuclear morphometry score was calculated for each patient as follows: MV(f) = (0.02 x AGE) + (1.17 x SNRF) + (90.6 x LEFD) - 94, with AGE denoting age at diagnosis in months, SNRF the skewness of the nuclear roundness factor, and LEFD the lowest value of nuclear ellipticity as measured by the feret diameter method. Relative risks of relapse were estimated for the total score and for each of its components. Sensitivity and specificity were determined for the criterion of 'MV(f) is greater than - 0.35' as a predictor of relapse. Results: By contrast with previously published results, neither the SNRF nor the LEFD made any contribution to the prediction of disease-free survival. Sensitivity and specificity of the criterion of 'MV(f) is greater than -0.35' were 71% and 56%, respectively. Conclusion: Re-evaluation of a published nuclear morphometry score showed that it did not predict disease-free survival in patients with Wilms' tumor. The earlier study very likely overestimated the predictive power of nuclear morphometry by using the same data set both to develop the score and to evaluate its properties. Because of the huge number of combinations of nuclear morphometry measurements that may enter into the multivariate discriminant function, use of appropriate statistical methods is essential to estimate accurately the sensitivity and specificity.
ASJC Scopus subject areas
- Cancer Research