Patient characteristics
This study included a total of 8813 patients from the eligible SEER database and 76 patients from the Chinese registry. Baseline characteristics are presented in Table 1. Among these patients, the rate of CEA positivity in the Chinese EOCRC cohort (84.2%) was higher than that in the SEER database EOCRC patients (43.2%), which may be related to a higher proportion of Chinese EOCRC patients with tumor diameter >5 cm (53.9%) compared to patients in the SEER database (35.9%)[31].
Incorporating 8813 ERCRC patients from the SEER database, we randomly allocated the 8813 EOCRC patients into a training cohort (n=6610) and a validation cohort (n=2203) at a 7:3 ratio using R software. There were no significant differences in baseline characteristics between the two groups. Detailed results are shown in Table 2. For the patients in the training cohort (n=6610), we first used univariate Cox regression analysis to screen for factors influencing CSM and OCM. Based on this, we conducted multivariate Cox regression analysis to identify independent risk factors.
Independent risk factors of Cause- different mortality
Univariate COX regression analysis showed that the risk factors affecting CSM in patients with EOCRC included age, gender, race, differentiation, TNM stage, number of lymph nodes cleared, CEA, size, marital status, pathological type, AJCC 6th edition staging, surgery or not, and information on radiation and chemotherapy. The results of multivariate COX regression analysis showed that race, tumour grade, CEA, marital status, pathology type, AJCC staging and surgery were independent risk factors affecting CSM, Among them, Black race had a higher risk of CSM compared to White race(HR=1.247,95CI=1.105-1.251,P<0.001);Poorly differentiated and Undifferentiated had a higher risk of CSM compared to Well differentiated (HR =2.006,95CI=1.590-2.531,P<0.001), (HR=2.279,95CI=1.705-3.048,P<0.001); CEA Positive had a higher risk of CSM compared to CEA Negative (HR=1.655,95CI=1.503-1.821,P <0.001); being married was a protective factor for CSM (HR=0.751, 95CI=0.688-0.820, P<0.001); Mucinous carcinoma and other pathological types had a higher risk of CSM compared to adenocarcinoma (HR=1.329, 95CI=1.132-1.561, P=0.001), ( HR=1.983, 95CI=1.689-2.329, P<0.001); AJCC higher Stage was a risk factor for CSM, with stageIIII (HR=14.834, 95CI=11.012-19.982, P<0.001), and finally, surgery was a protective factor for CSM (HR=0.511 , 95CI=0.385-0.678, p<0.001); in addition, age, gender, race, primary site of the tumour, M staging, number of lymph nodes cleared, chemotherapy, CEA, marital status and AJCC 6th edition staging were risk factors, and multivariate COX regression analysis showed that age, gender, chemotherapy information, CEA, marital status, and AJCC 6th edition information were independent risk factors for OCM in patients with EOCRC, Among them, the risk of OCM was higher for age ≥35 years (HR=2.217, 95CI=1.262-3.894, P=0.006); women had a lower risk of OCM compared to men (HR=0.761, 95CI=0.581-0.998, P=0.049); and patients who did not receive chemotherapy had a lower risk of OCM (HR= 0.451, 95CI=0.320-0.636, P<0.001); CEA-positive patients had a higher risk of OCM (HR=1.570, 95CI=1.179-2.088, P=0.002); in addition, marriage was also a protective factor for OCM (HR=0.466, 94CI=0.356-0.610, P< 0.001); for AJCC classification, stage IIII had a higher risk of developing OCM compared to stage I (HR=1.904, 95CI=1.110-3.265, P=0.019), yet there was no significant trend for stage III and stage II.. Detailed results are presented in Table 3 and Table 4.
Estimates of cumulative morbidity under Cause- different mortality
Subgroup Analysis using Gray's Test was conducted for CSM and OCM in EOCRC patients. The cumulative incidence function curves and P-values are shown in Fig.2. Among these curves, EOCRC patients with CEA positivity, low differentiation, other differentiation types, fewer than 4 lymph node dissections, chemotherapy, higher TNM staging, AJCC 6th edition staging, and those who did not undergo surgery exhibited higher cumulative death rates for CSM. On the other hand, EOCRC patients aged 35 years or older and those with low differentiation had higher cumulative death rates for OCM.
Construction and validation of the nomograms
We constructed nomograms predicting CSM in EOCRC patients based on independent risk factors . Factors for selection include degree of tumour differentiation, CEA, marital status, pathology type, AJCC staging and surgery. Among these factors, low differentiation grade, CEA positivity, and advanced AJCC staging were identified as risk factors for CSM, while pathological type as adenocarcinoma, being married, and undergoing surgery were protective factors for CSM.Based on the actual conditions of the patients, projecting each variable onto the upper 'Point' axis yields the corresponding score. Similarly, summing up the scores of each variable results in a total score. By projecting the total score onto the lower '3-year CSM ' and '5-year CSM ' axes, the 3-year and 5-year CSM occurrence rates can be estimated. For instance, a patient with Early-Onset Colorectal Cancer who has not undergone surgery (29 points), unmarried (11 points), in ACJJ stage III (46 points), CEA positive (19 points), and has moderately differentiated(4 points) adenocarcinoma (0 points), would have a total of 109 points. This total corresponds to a predicted 3-year CSM of 32.0% and a predicted 5-year CSM of 48.0%."Additionally, the model was tested through internal and external validation. The results demonstrated that the nomogram had excellent discriminatory capability. The corrected C-index for the training group and the internal and external validation groups were 0.806, 0.801 and 0.810, respectively. The AUC performance at 1 year, 3 years, and 5 years were "0.875, 0.864, 0.848," and in internal and external validation, they were "0.867, 0.849, 0.846," and "0.870, 0.868, 0.859," respectively, indicating the nomogram's strong predictive ability (Fig. 4) .Furthermore, calibration curves showed high concordance between predicted and observed values for CSM in the training and internal and external validation groups (Fig. 5).
Clinical Application of Nomograms.
The results of Decision Curve Analysis (DCA) demonstrate the valuable clinical utility of the nomograms (Fig. 6). This curve illustrates that at different time points, the nomograms provide favorable net benefits for most threshold probabilities. The nomograms constructed using the competing risk model offer greater value compared to traditional survival analysis methods.