ABSTRACT
The prognostic value of nuclear morphometry in addition to clinical and pathologic features was retrospectively studied in 64 cases of colorectal carcinoma resected for cure with a minimum of five years of follow-up. By univariate analysis, patient outcome was found to correlate with the presence of serosal involvement (P = .003), the presence of lymph node involvement (P = .01), the number of involved lymph nodes (P = .0001) and the mean nuclear area (P = .02). With multivariate analysis, only the number of involved lymph nodes significantly correlated with the survival (P = .0001). In a subsequent multivariate model expressing lymph node status as the presence or absence of metastasis, the presence or absence of serosal involvement and the mean nuclear area were both found to independently correlate with the outcome (P = .003 and P = .02, respectively). Linear regression analysis revealed significant colinearity between the mean nuclear area and the number of involved lymph nodes (P = .03). Accelerated failure time models based on determination of serosal involvement and then either specification of the number of involved lymph nodes or calculation of the mean nuclear area were of comparable predictive value to the determination of the number of involved lymph nodes alone. The former appeared to be better at identifying a subgroup of patients with good prognosis. This study demonstrates that two or more models based on pathologic features may be of comparable predictive value in colorectal carcinoma resected for cure, including models that incorporate mean nuclear area.
