Development and validation of nomograms for predicting overall survival and cancer-specific survival in unresected colorectal cancer patients undergoing chemotherapy

Abstract

This study aims to develop nomograms for predicting overall survival (OS) and cancer-specific survival (CSS) in colorectal cancer (CRC) patients who did not receive primary site surgery but underwent chemotherapy. We analyzed data from 3,050 patients treated with chemotherapy without primary site surgery from 2010 to 2015, sourced from the Surveillance, Epidemiology, and End Results (SEER) database. The data were randomly divided into training and validation sets. Initial variable selection was performed using the least absolute shrinkage and selection operator (LASSO), and multivariate Cox regression analysis was used to identify independent prognostic factors. Two nomograms were subsequently constructed based on these factors. Survival analysis was conducted using Kaplan-Meier plots and the log-rank test. We identified nine significant predictors of OS and CSS: age, marital status, primary site, grade, histology, T stage, M stage, tumor size, and CEA levels. The models for OS and CSS exhibited excellent predictability, with time-dependent area under the receiver operating characteristic curves (AUCs) exceeding 0.7. Calibration curves confirmed the accuracy of these predictions in the training and validation sets. Additionally, decision curve analysis (DCA) indicated that our models provide greater clinical benefit than traditional TNM staging. Notably, survival outcomes varied significantly across risk categories, affirming the models' effective discrimination. For CRC patients who did not receive primary site surgery but underwent chemotherapy, this validated nomogram enables precision prognostication fundamentally shifting the paradigm from population-level TNM estimates to individualized risk-adaptive management.

Keywords: Chemotherapy; Colorectal cancer; Nomogram; Prognosis; SEER database.

Fig. 1

Flowchart of the data selection and analysis process.

Fig. 2

LASSO coefficient profiles of the 17 candidate predictors for OS (A) and CSS (C), and plots of partial likelihood deviance for OS (B) and CSS (D).

Fig. 3

Nomograms for predicting the OS (A) and CSS (B) of unresected CRC patients receiving chemotherapy.

Fig. 4

Calibration curves of the OS prognostic nomogram for 2, 3, and 5 years in the training cohort (A) and validation cohort (B); CSS prognostic nomogram for 2, 3, and 5 years in the training cohort (C) and validation cohort (D).

Fig. 5

ROC curves of the OS prognostic nomogram for 2, 3, and 5 years in the training cohort (A) and validation cohort (B); ROC curves of the CSS prognostic nomogram for 2, 3, and 5 years in the training cohort (C) and validation cohort (D).

Fig. 6

DCA curves of the OS prognostic nomogram in the training cohort (A) and validation cohort (B); CSS prognostic nomogram in the training cohort (C) and validation cohort (D).

Fig. 7

Kaplan-Meier survival curves of the OS in three risk subgroups for the training cohort (A) and validation cohort (B); Kaplan-Meier survival curves of the CSS in three risk subgroups for the training cohort (C) and validation cohort (D).