A New Prediction Model for Local Recurrence After Curative Rectal Cancer Surgery: Development and Validation as an Asian Collaborative Study.

Abstract

Local recurrence is one of the remaining problems in rectal and rectosigmoid cancer, and it is sometimes difficult to treat. This study aimed to explore various factors that are highly related to local recurrence and to develop a new prediction model for local recurrence after curative resection. This is a retrospective cohort study SETTINGS:: This study was conducted at 2 academic hospitals in Japan and Korea. A total of 2237 patients with stage I to III rectal and rectosigmoid cancer who underwent a curative operation with a negative circumferential margin were selected. Surgical treatment was the intervention. Local recurrence was the primary outcome measure. A total of 1232 patients were selected, and rectosigmoid cancer with rare local recurrence (2/221) was excluded. A different set of 792 patients with rectal cancer were chosen for validation. Multivariate analysis showed the following factors as significant for local recurrence: poorly differentiated tumor (HR, 11.2; 95% CI, 4.5-28.0), tumor depth (HR, 5.0), lymph node metastasis (HR, 4.1), operative procedure (HR, 3.2), postoperative complications (HR, 2.9), tumor location (HR, 2.6), and CEA level (HR, 2.4); a new prediction score was created by using these factors. A poorly differentiated tumor was assigned 2 points, and all other factors were assigned 1 point each. Patients who scored more than 5 points (n = 21) were judged as "high risk," with a 2-year local recurrence rate of 66.5%. The new predictive model could also separate the patients into different risk groups in the validation set. The high-risk group had higher recurrence rates than medium- and low-risk groups (2-year local recurrence rate: 41%, 15%, and 2.1%). This study was limited by its retrospective nature and potential for selection bias. Seven factors were shown to be significantly correlated with the local recurrence of rectal cancer, and the usefulness of this new prediction model was demonstrated. See Video Abstract at http://links.lww.com/DCR/A429.