A competing-risks nomogram for predicting probability of death from CRC in Chinese Han patients with Stage I-III CRC.

Abstract

Many patients with colorectal cancer are elderly with competing comorbidities. When constructing nomogram for assessing survival, we should consider the influence of competing risk. A competing-risks nomogram was developed to estimate the probability of death due to colorectal cancer for patients after curative surgery. A total of 2442 patients with non-metastatic colorectal cancer were included to develop competing-risks nomogram. Competing-risks nomogram were established based on the results of Fine and Gray competing-risks proportional hazards model. To maximize the accuracy of prediction, model selection was not carried out, and non-linear continuous variables were flexibly modeled with restricted cubic splines. The nomogram was internal-validated by bootstrapping, and externally validated with a separate database of 299 patients from The Cancer Genome Atlas. The performance of this model was assessed by concordance index and a calibration curve. There were 332 patients died of colorectal cancer and 46 died of other causes during the follow-up period. Age, T stage, N stage, histological type, tumor location, adjuvant chemotherapy, preoperative carcinoembryonic antigen, lymph vascular invasion, lymph node ratio and sample lymph nodes were integrated into competing-risks nomogram. The competing-risks nomogram for predicting probability of death due to colorectal cancer with a concordance index of 0.768, ameliorating the stratification provided by the seventh edition tumor-node-metastasis staging system of the American Joint Committee on Cancer (AJCC). The concordance index for validation dataset was 0.783. We developed and externally validated a competing-risks nomogram for Chinese Han patients with non-metastatic colorectal cancer, which could provide probability of death from colorectal cancer in the presence of competing risks.