A Predictive Model for Nodal Metastases in Patients With Appendiceal Cancers.

Abstract

Histologic subtypes of appendiceal cancer vary in their propensity for metastases to regional lymph nodes (LN). A predictive model would help direct subsequent surgical therapy. The National Cancer Database was queried for patients with appendiceal cancer undergoing surgery between 1998 and 2012. Multivariable logistic regression was used to develop a predictive model of LN metastases which was internally validated using Brier score and Area under the Curve (AUC). A total of 21,647 patients were identified, of whom 9079 (41.9%) had node negative disease, 4575 (21.1%) node positive disease, and 7993 (36.9%) unknown LN status. The strongest predictors of LN positivity were histology (carcinoid tumors OR 12.78, 95% CI 9.01-18.12), increasing T Stage (T3 OR 3.36, 95% CI 2.52-4.50, T4 OR 6.30, 95% CI 4.71-8.42), and tumor grade (G3 OR 5.55, 95% CI 4.78-6.45, G4 OR 5.98, 95% CI 4.30-8.31). The coefficients from the regression analysis were used to construct a calculator that generated predicted probabilities of LN metastases given certain inputs. Internal validation of the overall model showed an AUC of 0.75 (95% CI 0.74-0.76) and Brier score of 0.188. Histology-specific predictive models were also constructed with an AUC that varied from 0.669 for signet cell to 0.75 for goblet cell tumors. The risk for nodal metastases in patients with appendiceal cancers can be quantified with reasonable accuracy using a predictive model incorporating patient age, sex, tumor histology, T-stage, and grade. This can help inform clinical decision making regarding the need for a right hemicolectomy following appendectomy.