3.1 Characteristics of patients
As presented in Table 1, a total of 235 patients with stage I-III colorectal cancer were enrolled in this study. In terms of nutritional assessment, according to the NRS2002 score, 77 patients (32.8%) were identified as malnourished preoperatively (Table 2). Based on the CONUT score, 121 patients (51.4%) were classified as normal, 83 patients (35.4%) had mild malnutrition, 28 patients (11.9%) had moderate malnutrition, and 3 patients (1.3%) had severe malnutrition (Table 3).
Regarding tumor characteristics, 101 patients (43.0%) had rectal cancer, and 134 patients (57.0%) had colon cancer. Based on TNM staging, there were 34 patients (14.5%) in stage I, 88 patients (37.4%) in stage II, and 113 patients (48.1%) in stage III. Postoperative pathological results revealed that 72 patients (30.6%) had a tumor diameter of ≥5 cm, 50 patients (21.3%) exhibited vascular invasion, and 13 patients (5.5%) showed neural invasion.
Postoperative complications were recorded within the hospitalization period or within 30 days after surgery, and only complications with Clavien-Dindo classification grade ≥II were considered. Among all the enrolled patients, a total of 43 individuals (18.3%) experienced postoperative complications. During the follow-up period, out of the 235 patients, 41 cases (17.4%) experienced tumor recurrence. These cases included 1 case in stage I, 11 cases in stage II, and 29 cases in stage III. The recurrence rates were 8.9% at 1 year, 15.7% at 2 years, and 17.4% at 3 years.
Table 1
Clinical characteristics of the 235 patients with colorectal cancer
Features
|
Total (n=235)
|
Gender (Male)
|
101(43.0%)
|
Tumor location (Colon)
|
134(57.0%)
|
Macroscopic type
|
|
Protrude type
|
103(43.8%)
|
Ulcerative type
|
128(54.5%)
|
Infiltrating type
|
4(1.7%)
|
Tumor size (≧5 cm)
|
72(30.6%)
|
Tumor differentiation
|
|
Poor
|
17(7.2%)
|
Medium
|
205(87.2%)
|
high
|
13(5.5%)
|
pT stage
|
|
T1
|
7(3.0%)
|
T2
|
32(13.6%)
|
T3
|
98(41.7%)
|
T4
|
98(41.7%)
|
pN stage
|
|
N0
|
124(52.8%)
|
N1
|
66(28.1%)
|
N2
|
45(19.1%)
|
Vascular invasion (Positive)
|
50(21.3%)
|
Perineural invasion (Positive)
|
13(5.5%)
|
CEA (≧5 ng/ml)
|
148(63.0%)
|
CA125(>35U/ml)
|
25(10.6%)
|
CA19-9(>34U/ml)
|
32(13.6%)
|
HGB(g/l)
|
121.98±27.10
|
LDH(U/l)
|
174.58±35.76
|
Bowel obstruction(yes)
|
34(14.5%)
|
Postoperative complications(yes)
|
43(18.3%)
|
Diabetes (Yes)
|
31(13.2%)
|
Hypertension (Yes)
|
53(22.6%)
|
Recurrence and metastasis (Yes)
|
41(17.4%)
|
Table 2
Characteristics of NRS2002 scores in 235 patients with colorectal cancer
Features
|
Total (n=235)
|
Age (≧70y)
|
180(76.6%)
|
BMI≦18.5(kg/m2)
|
16(6.8%)
|
Duration of Weight Loss >5%
|
|
No weight reduction
|
180(76.6%)
|
Within three months
|
19(8.1%)
|
Within two months
|
18(7.7%)
|
Within one month
|
18(7.7%)
|
Reduced food intake within one week
|
|
0
|
159(67.7%)
|
25-50%
|
54(23.0%)
|
51-75%
|
18(7.7%)
|
76-100%
|
4(1.7%)
|
NRS2002 ≥3
|
77(32.8%)
|
Table 3
Characteristics of CONUT scores in 235 patients with colorectal cancer
Features
|
Total (n=235)
|
Serum albumin(g/l)
|
|
≧35
|
175(74.5%)
|
30-34.9
|
39(16.6%)
|
25-29.9
|
18(7.7%)
|
<25
|
3(1.3%)
|
Total lymphocyte count (*10^9/L)
|
|
≥1.6
|
43(18.3%)
|
1.2-1.599
|
81(34.5%)
|
0.8-1.199
|
32(13.6%)
|
<0.8
|
9(3.8%)
|
Total cholesterol (mg/dl)
|
|
≥180
|
129(55.9%)
|
140-179
|
77(32.8%)
|
100-139
|
23(9.8%)
|
<100
|
6(2.5%)
|
CONUT score
|
|
Normal
|
121(51.4%)
|
Mild
|
83(35.4%)
|
Moderate
|
28(11.9%)
|
Severe
|
3(1.3%)
|
3.2 BMI, preoperative bowel obstruction, neural invasion, tumor differentiation degree, serum albumin, total lymphocyte count, and pN stage are risk factors for postoperative recurrence and progression in colorectal cancer.
This research study involves a meticulous evaluation of various clinical indicators from enrolled patients. These encompass tumor pathology features, biomarkers, clinical characteristics, NRS2002 score components, and CONUT score components, as presented in Tables 1, 2, and 3. Lasso-Cox regression parameter selection is performed on these indicators, and the changing characteristics of variable coefficients are illustrated in Fig. 2A. Iterative analysis is conducted using a 10-fold cross-validation approach. Ultimately, a high-performance model with fewer variables is obtained at the point of minimum error, as depicted in Fig. 2B.
The selected variables include BMI, preoperative bowel obstruction, neural invasion, tumor differentiation degree, serum albumin, total lymphocyte count, and pN stage. The relationship between the selected variables and postoperative recurrence in colorectal cancer patients was explored using K-M log-rank test analysis. The results, as shown in Fig. 3, reveal significant differences in survival curves among various groups.
3.3 BMI ≤ 18.5 kg/m², neural invasion, and pN stage are independent risk factors for recurrence in colorectal cancer patients.
Based on the results of Lasso-Cox regression analysis, a Cox proportional hazards regression was performed on BMI, preoperative bowel obstruction, neural invasion, tumor differentiation degree, serum albumin, total lymphocyte count, and pN stage to predict postoperative DFS in colorectal cancer patients. The findings were visualized using a forest plot, as shown in Fig. 4. BMI ≤ 18.5 kg/m² (HR = 3.39, P < 0.005), perineural invasion(HR = 4.51, P<0.005), and pN stage(HR = 4.19, P<0.005) were identified as independent risk factors for recurrence in colorectal cancer patients.
3.4 Nomogram for colorectal cancer patients.
To ensure a wider clinical applicability of the model, encompassing preoperative nutritional status and a broader range of indicators, we employed the minimum error point from Lasso-Cox regression to construct the predictive model. The selected variables included BMI, preoperative bowel obstruction, neural invasion, tumor differentiation degree, serum albumin, total lymphocyte count, and pN stage. Using R software, we generated a Nomogram depicting the disease-free survival (DFS) rates at 1 year, 2 years, and 3 years following curative resection surgery for colorectal cancer patients, as illustrated in Fig. 5
3.5 Internal validation demonstrates that the model exhibits favorable discriminative ability, calibration, and clinical applicability.
The model's performance was assessed and validated through bootstrap resampling (1,000 iterations) to compute the C-index and construct the Receiver Operating Characteristic (ROC) curve (Fig. 6A), as well as the calibration curve (Fig. 6B). The calculated C-index values are as follows: 1-year DFS C-index of 0.8 (95%CI: 0.77-0.84), 2-years DFS C-index of 0.8 (95%CI: 0.78-0.83), and 3-years DFS C-index of 0.78 (95%CI: 0.75-0.80). The Area Under the ROC Curve (AUC) values are: 1-year DFS AUC of 0.82 (95%CI: 0.78-0.86), 2-years DFS AUC of 0.82 (95%CI: 0.80-0.85), and 3-years DFS AUC of 0.80 (95%CI: 0.76-0.83). These results indicate that the model possesses reliable discriminative ability
Furthermore, the calibration curves for 1-year DFS, 2-years DFS, and 3-years DFS demonstrate good fit with the standard curve, indicating the model's strong calibration.
In some scenarios, when the AUC is high and there is good consistency between predictions and observations, column charts may not be as informative. To further validate the clinical effectiveness of the model, a Decision Curve Analysis (DCA) was performed using R software and compared with the TNM staging (AJCC) (Figure 6C). The model's curve consistently remains in the upper-right corner, indicating better clinical applicability in predicting 1-year, 2-years, and 3-years DFS compared to AJCC staging. The threshold probability is more extensive and practical.
In conclusion, the developed model exhibits strong predictive capability for 1-year, 2-years, and 3-years DFS after curative resection surgery in colorectal cancer patients. The model highlights that patients with lower serum albumin levels, BMI ≤ 18.5 kg/m², lower preoperative total lymphocyte counts, preoperative bowel obstruction symptoms, pathological neural invasion, lower tumor differentiation degree, and higher pN stage have an increased probability of postoperative recurrence.