Patient characteristics
After the selection, a total of 708 stage II CRC patients were included in the study and they were divided into chemotherapy group and non-chemotherapy group. The clinicopathological characteristics of patients were summarized and compared using the Chi-square test in Table 1 and Mann-Whitney U test in Table 2. 447 (63.1%) patients received chemotherapy and 261 (36.9%) patients did not; 431 (60.9%) patients were male and 277 (39.1%) were female; 354 (50.0%) colon cancer patients and 354 (50.0%) rectal cancer patients. The median age of the patients was 63 (range 23-88). According to the Chi-square test and Mann-Whitney U test, there were significant differences in age, differentiation, T category, CCVD, LMR level and PNI level between the chemotherapy group and the non-chemotherapy group. Higher proportions of advanced age (46.7% vs. 11.0%, p<0.001), poor differentiation (9.6% vs. 6.0%, p=0.038) and positive cardio-cerebrovascular comorbidities (16.9% vs. 10.7%, p=0.019) were seen in the non-chemotherapy group, while the chemotherapy group were more often diagnosed with T4 category (67.3% vs. 54.0%, p<0.001).
Overall survival analysis
For all the 708 patients with stage II CRC, Kaplan-Meier method with the log-rank test and univariate Cox proportional hazard regression analysis were used to calculate RMST and HR and compare OS between the chemotherapy group and the non-chemotherapy group. The result showed that the chemotherapy group had a better OS than the non-chemotherapy group (Figure 1A, RMST: 56.2 months vs 53.7 months, HR: 0.580, p=0.007).
Optimal inflammatory marker
STEPP analysis was performed, taking respectively the level of the four concerned inflammatory markers, including NLR, PLR, LMR and PNI, as x-axis and the cumulate mortality at 60 months measured by Kaplan-Meier method as y-axis to compare the OS between the chemotherapy group and the non-chemotherapy group in different subgroups divided by the level of the inflammatory markers.
In PLR-related STEPP, when PLR>130, the cumulate mortality of the non-chemotherapy group was significantly and constantly higher than the cumulate mortality of the chemotherapy group, while the tendency was just on the contrary when PLR<130 (Figure 2A). Besides, no such tendency was found in the analyses related with the other inflammatory markers (Figure 2B, 2C and 2D). This result indicated that PLR was closely associated with the survival benefit of chemotherapy. We could further regard the PLR level of 130 as the cut-off value to distinguish chemotherapy-effective patients. Therefore, we divided the patients into high-PLR subgroup (PLR≥130) and low-PLR subgroup (PLR<130) and performed the Chi-square test to compare the PLR level between the chemotherapy group and the non-chemotherapy group (Table 1).
Subgroup survival analysis
Besides the PLR level, we also divided the patients into subgroups by other clinicopathological characteristics, including gender, age, tumor localization, tumor size, differentiation, T category, numbers of examined lymph nodes and the CEA level for further stratified analysis. In each subgroup, Kaplan-Meier method with the log-rank test and univariate Cox proportional hazard regression analysis were used to calculate RMST and HR and compare OS between chemotherapy and the non-chemotherapy patients. Besides, multivariate Cox proportional hazard regression analysis was used to investigate the interaction between the effect of chemotherapy and the clinicopathological characteristics (Table 3).
The result of Kaplan-Meier survival analysis indicated that the chemotherapy patients with these following conditions had a better OS than the non-chemotherapy patients: male (55.9 months vs.53.2 months, HR: 0.592, p=0.035), rectal cancer (55.6 months vs. 52.2 months, HR: 0.539, p=0.016), well or moderate differentiation (56.3 months vs. 53.9 months, HR: 0.569, p=0.008), T4 category (55.6 months vs. 51.8 months, HR: 0.501, p=0.002), number of examined lymph nodes≥12 (57.1 months vs.54.5 months, HR: 0.500, p=0.007), and the PLR level>130 (Figure 1B and 1C, 56.5 months vs.51.3 months, HR: 0.371, p<0.001). However, according to multivariate Cox survival analysis, PLR was the only characteristic associated with the effect of chemotherapy (p interaction =0.027).
PSM for chemotherapy patients and non-chemotherapy patients
To minimize the differences in the clinicopathological characteristics, we performed PSM. In total, 166 chemotherapy patients and 166 non-chemotherapy patients were paired with 1:1 on the propensity scores using the nearest-neighbor method with a caliber of 0.02. The clinicopathological characteristics were summarized and compared using the Chi-square test in Table 4. After PSM, the differences in the characteristics between the chemotherapy group and the non-chemotherapy group were acceptable (p value of the Chi-square test>0.05 in all subgroups).
Survival analysis after PSM
The 332 matched patients were also divided into subgroups by the clinicopathological characteristics. In each subgroup, we performed Kaplan-Meier method with the log-rank test and univariate Cox proportional hazard regression analysis to calculate RMST and HR and compare OS between chemotherapy and the non-chemotherapy patients. We also used multivariate Cox survival analysis to assess the interaction between the effect of chemotherapy and the clinicopathological characteristics (Table 5).
However, different from the result before PSM, there was no significant difference in RMST between the chemotherapy patients and the non-chemotherapy patients (Figure 3A, 56.0 months vs. 53.3 months, HR: 0.584, p=0.058). Besides, the result of subgroup analysis was also not quite the same as the result before PSM. Kaplan-Meier survival analysis and univariate Cox proportional hazard regression analysis showed that the chemotherapy patients who were male (55.9 months vs.52.0 months, HR: 0.487, p=0.042), T4 category (55.2 months vs. 50.9 months, HR: 0.501, p=0.027), number of examined lymph nodes≥12 (57.5 months vs.54.1 months, HR: 0.404, p=0.018), and the PLR level>130 (Figure 3B and 3C, 57.6 months vs.51.3 months, HR: 0.272, p=0.005) had a longer RMST than the non-chemotherapy patients. However, in accordance with the result before PSM, PLR was still the only characteristic associated with the effect of chemotherapy (p interaction=0.038) according to multivariate Cox proportional hazard regression analysis.