Predictive role of systemic immune-inflammation index in the prognosis of patients with advanced left-sided colorectal cancer: a retrospective study

Study population

We retrospectively collected data from 231 patients with unresectable advanced (initially unresectable or postoperative recurrent stage IV) left-sided CRC who received CAPEOX ± bevacizumab as first-line chemotherapy at the Oncology Department of the Third Affiliated Hospital of Soochow University from October 2015 to December 2023. The reasons for selecting this patient population are as follows: firstly, a multicenter epidemiological survey from China showed that left-sided colon cancer and rectal cancer accounted for 19.8% and 58.7% of the total incidence, respectively (Shi et al., 2021). Additionally, multiple studies have indicated that, compared to right-sided colon cancer, left-sided colon and rectal cancer have a better prognosis, particularly in stage III–IV patients (Benedix et al., 2010; Price et al., 2015; Weiss et al., 2014). Therefore, these factors highlight the significance of prognostic analysis in patients with left-sided CRC. Secondly, CAPEOX ± bevacizumab, as a classic first-line chemotherapy regimen for advanced CRC recommended by the Chinese Society of Clinical Oncology (CSCO) guidelines, has been widely used in clinical practice (Diagnosis, Treatment Guidelines For Colorectal Cancer Working Group C, 2019; Yuan et al., 2019). Thirdly, although for patients with MSS or MSI-L/p-MMR and wild-type RAS and BRAF, the guidelines recommend first-line chemotherapy with FOLFOX/FOLFIRI ± cetuximab (a two-week regimen), however, for patients with limited financial resources, CAPEOX ± bevacizumab (a three-week regimen) can also be used (Diagnosis, Treatment Guidelines For Colorectal Cancer Working Group C, 2019; Yuan et al., 2019). Therefore, to ensure relative homogeneity in treatment regimens and treatment cycles among patients, we ultimately included patients who received CAPEOX ± bevacizumab as first-line chemotherapy and conducted the corresponding data collection and statistical analysis.

Patients eligible for inclusion met the following criteria: (1) a diagnosis of initially unresectable or postoperative recurrent stage IV metastatic CRC, as defined by the 2024 Chinese Society of Clinical Oncology (CSCO) Guidelines for CRC; (2) histopathological confirmation of adenocarcinoma of the left colon (distal to the splenic flexure) or rectum; (3) eligibility for palliative chemotherapy with CAPEOX ± bevacizumab, administered according to CSCO guidelines; and (4) availability of complete medical records. Patients were excluded if they had incomplete medical records, severe infections or acute inflammatory conditions affecting hematologic parameters, or concurrent primary malignancies at other sites. Ultimately, 231 patients with unresectable advanced (initially unresectable or postoperative recurrent stage IV) left-sided CRC were selected for the study.

This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committee of the Third Affiliated Hospital of Soochow University (Approval No.: 2023 J. No. 060, Approval Date: October 27, 2023, Valid until: June 15, 2026). Since some patients have passed away or are difficult to contact due to the passage of time, it is objectively impossible to obtain their informed consent. Therefore, the Ethics Committee approved the waiver of informed consent or the waiver of certain elements of informed consent.

Collection of clinicopathological data

Clinicopathological characteristics and laboratory findings obtained within one month prior to initiating first-line chemotherapy were retrospectively extracted from the electronic medical record system. The variables collected included patient demographic data (age, gender, height and weight), molecular profiles (KRAS/NRAS/BRAF mutation status), and clinical features (presence of liver-confined metastases, resection status of the primary tumor, history of adjuvant chemotherapy). Additionally, data on first-line chemotherapy regimens and hematologic parameters, including neutrophil, platelet, monocyte, and lymphocyte counts, were obtained.

Calculation formulas for systemic inflammatory indices and body mass index

NLR = Neutrophil count/Lymphocyte count, PLR = Platelet count/Lymphocyte count, MLR = Monocyte count/Lymphocyte count, SII = (Neutrophil count × Platelet count)/Lymphocyte count, BMI = weight (kg)/height² (m²).

Follow-up and outcome assessment

Patients were monitored through a combination of outpatient and inpatient data from the electronic medical record system and telephone follow-ups. OS was defined as the duration from the diagnosis of initially unresectable or postoperatively recurrent stage IV metastatic CRC (as confirmed through imaging or pathological evaluation) to either the date of death or the last follow-up. The study follow-up concluded on October 11, 2024, with durations ranging from 1.0 to 103.2 months and a median follow-up time of 23.6 months. Among the cohort, 26 patients were lost to follow-up, corresponding to a loss-to-follow-up rate of 11.3%.

Statistical methods

Data were organized using Microsoft Excel 2019 (Microsoft, Redmond, WA, USA), while statistical analyses and visualizations were performed using SPSS version 26 (IBM Corp., Armonk, NY, USA) and R version 4.2.3 (R Core Team, 2022). X-Tile version 3.6.1 was used to establish cutoff values for age, body mass index (BMI), NLR, PLR, MLR, and SII levels in predicting OS in advanced left-sided CRC. Based on these thresholds, the study cohort was stratified into high-value and low-value groups.

To minimize the risk of overfitting in small sample sizes when using X-Tile to determine cutoff values based on the log-rank test, we adopted a 2-fold cross-validation approach proposed by Faraggi & Simon (1996) to validate the reliability of the cutoff selection. This method has been applied in recent prospective studies (Chen et al., 2021; Porter et al., 2024; Syeda et al., 2021). The specific steps of this method are as follows: for each variable requiring a cutoff, the entire sample was randomly divided into two subsets, Subset I and Subset II, ensuring each subset contains a similar number of observations. Using Subset I, X-Tile was employed to determine the cutoff value; this cutoff was then applied to Subset II to categorize patients into Group A (above the cutoff) and Group B (below the cutoff). Similarly, a cutoff was determined using Subset II and applied to Subset I. After this process, each patient was assigned to either Group A or Group B based on the cutoff derived from the other subset. A log-rank test stratified by subset was then performed to assess the significance of survival differences between the two groups. If the result was significant, it indicated that the risk of overfitting when directly applying X-Tile to the entire sample was acceptable. Following this analysis, significant survival differences were observed between the two groups for all variables except BMI (Table S1). Therefore, for BMI, we adopted the WHO standard classification for Asian populations. Given the small number of patients in the underweight and obese categories, we referred to the study by Liu et al. (2013) and divided all patients into two groups based on a BMI threshold of 23.0.

Chi-square (χ²) tests were conducted to compare the clinicopathological characteristics of patients across varying levels of NLR, PLR, MLR, and SII. Kaplan–Meier methods were employed for survival analysis, with survival curves generated and compared using log-rank tests. Univariable analysis was performed using the Cox proportional hazards model to identify prognostic factors. To assess the correlations between systemic inflammatory indices (SII, NLR, PLR, MLR), the Shapiro–Wilk test for normality and Spearman’s correlation analysis were employed. To avoid potential collinearity among the systemic inflammatory indices, the indices that were statistically significant in the univariable Cox analysis were separately included in multivariable Cox regression analyses along with the clinicopathological factors that were also statistically significant in the univariable Cox analysis.

Using R version 4.2.3, five-fold cross-validation was conducted by randomly partitioning the dataset into five approximately equal subsets. A fixed random seed was applied to ensure reproducibility. Each subset was alternately designated as the validation set, with the remaining subsets serving as the training set. This process was repeated five times. The average C-index for systemic inflammatory indices identified in multivariable analyses was calculated across the five training and validation sets to evaluate prognostic value. The stability and generalizability of the C-index across different data partitions were further evaluated by calculating the mean squared error (MSE) between the training and validation sets during cross-validation. The formula for MSE is as follows: ${\displaystyle MSE=\frac{1}{n}\sum _{i=1}^n{\left(y_i-{\hat{y}}_1\right)}^2}$ where n is the number of cross-validation iterations, and y_i and ${\hat{y}}_1$ represent the C-index of the training and validation sets in the i-th cross-validation iteration.

Finally, we employed the Cox proportional hazards model to assess the interaction effect between the SII and treatment regimen in advanced left-sided CRC. We constructed a Cox proportional hazards model that included SII, treatment regimen (CAPEOX versus CAPEOX plus bevacizumab), and their interaction term. By examining the interaction term, we determined whether the prognostic value of SII varied by treatment regimen or resection status. The significance level was set at α = 0.05.

Association between SII and clinicopathological characteristics in patients with advanced left-sided CRC

This study encompassed 231 patients, with a median age of 65 years. Of these patients, 162 (70.1%) were male, and 69 (29.9%) were female. Primary tumor resection had been performed in 137 patients, while 94 patients did not undergo surgical intervention. Regarding first-line chemotherapy, 97 patients were treated with the CAPEOX regimen (86 with initially unresectable disease and 11 with recurrent disease), and 134 received CAPEOX combined with bevacizumab (93 with initially unresectable disease and 41 with recurrent disease). Using the X-Tile 3.6.1 software, optimal cutoff values for predicting OS were established for various parameters, including age (67 years), NLR (2.8), PLR (184.4), MLR (0.2), and SII (1,424.8). Based on the SII cutoff, patients were stratified into low-SII (SII ≤ 1,424.8) and high-SII (SII > 1,424.8) groups. Statistically significant differences were observed in primary tumor resection status between these groups (P < 0.05). Further details regarding the distribution of SII and clinicopathological variables between the two groups are provided in Table 1. Other systemic inflammatory indices are presented in Table S2.

Association between systemic inflammatory indices and prognosis of patients with advanced left-sided CRC

Kaplan–Meier survival analysis demonstrated that elevated SII, NLR, PLR, and MLR were each significantly associated with reduced OS (all P < 0.05) (Fig. 1). Univariable Cox regression analysis identified age, primary tumor resection, SII, NLR, PLR, and MLR as significant prognostic factors influencing OS (P < 0.05; Table 2).

To assess the correlations between the SII, NLR, PLR, and MLR, we employed the Shapiro–Wilk test to evaluate the normality of these systemic inflammatory markers. The results demonstrated that all indices deviated from a normal distribution (P < 0.05) (Table S3). Subsequently, a Spearman rank correlation heatmap was constructed, which revealed a moderate degree of correlation among the systemic inflammatory indices, with correlation coefficients ranging from 0.234 to 0.530 (Fig. S1). Therefore, to avoid potential collinearity issues, we conducted multivariable analysis using four separate models. Each multivariable model included only one systemic inflammatory index. Multivariable Cox regression analysis confirmed SII (HR = 3.534, 95% CI [1.988–6.282], P < 0.001), NLR (HR = 1.844, 95% CI [1.243–2.737], P = 0.002), PLR (HR = 1.707, 95% CI [1.156–2.520], P = 0.007), and MLR (HR = 1.832, 95% CI [1.191–2.817], P = 0.006) as independent prognostic factors for OS (Table 2).

Comparison of prognostic value of systemic inflammatory indices

The prognostic value of SII was evaluated in comparison with NLR, PLR, and MLR, utilizing a 5-fold cross-validation approach. The mean C-index for both the training and validation datasets was calculated across the five validation folds, alongside MSE. The findings revealed that SII demonstrated a higher mean C-index relative to NLR, PLR, and MLR in both the training and validation sets. Additionally, SII exhibited lower MSE values, underscoring its superior prognostic value, robustness, and generalizability across different data partitions (Fig. 2; Table 3).

Interaction analysis between SII and treatment regimen

Additionally, the interaction effect between treatment regimens (CAPEOX ± bevacizumab) and SII categories (high-value group vs. low-value group) was assessed. Multivariable Cox proportional hazards analysis revealed no statistically significant interaction between the treatment regimens and SII (P = 0.935). Similarly, Kaplan–Meier survival analysis and log-rank tests demonstrated no significant differences in survival outcomes between the CAPEOX and CAPEOX + bevacizumab treatment groups (P = 0.09). These findings suggest that variations in treatment regimens do not significantly modify the prognostic value of SII.