Exploring the optimal follow-up time for resectable colorectal cancer patients: a multicenter, five-year longitudinal cohort study.

Follow-up after colorectal cancer surgery is crucial for the early detection of recurrence and metastasis. We aimed to develop a comprehensive mixed-effects model using multidimensional clinical data to explore the optimal postoperative follow-up time, thereby improving patient survival and economic benefits. This study was a cohort study that collected data from patients who underwent radical resection for rectal cancer at Xijing Hospital from 2011 to 2023, as well as data from PLA General Hospital and Shaanxi Provincial People's Hospital. The development cohort and internal validation were from Xijing Hospital, with the remaining two centers serving as external validation sites. Based on both conventional and dynamic data, a multivariate mixed-effects model was constructed, and the model performance was evaluated using the area under the receiver operating characteristic curve, F1 score, absolute difference parameters, Brier score and calibration curves. This study included 5,848 patients with a total of 78,885 patient-years of data. Multivariate Cox regression analysis indicated that age, BMI, gender, operation, number of lymph nodes dissected and positive lymph nodes, S100, CD34, D2-40, MMR, TNM stage, chemotherapy, neoadjuvant chemotherapy, surgery time, obstruction, differentiation and tumor site were risk factors for colorectal cancer patients. The China Colorectal Cancer - Dynamic and Integrated System for Predicting Outcomes (CCC-DISPO) model demonstrated high performance, with the high values reached in the external validation at Shaanxi Provincial People's Hospital and PLA General Hospital, with AUC values of 0.87 and 0.78. Subgroup analysis showed that in stage I patients, stable high values were maintained throughout the first 12 months, while for stage II, III and IV occurred after the third month. As a dynamic prediction model, CCC-DISPO continuously updates and achieves high precision and stability. Based on the model performance over time, we believe that follow-up at 6 months postoperatively for all patients, 3 months for stage II, III, IV and within 12 months for stage I are of significant importance.