This study was conducted in accordance with PRISMA 2020 guidelines and was prospectively registered in PROSPERO (CRD420251010606) (10). Two investigators (WPF and JHY) independently performed literature searching. Searching terms as following: “Colorectal, Neoplasm,” “Neoplasm, Colorectal,” “Colorectal Tumors,” “Colorectal Tumor,” “Tumor, Colorectal,” “Tumors, Colorectal,” “Neoplasms, Colorectal,” “Colorectal Cancer,” “Cancer, Colorectal,” “Cancers, Colorectal,” “Colorectal Cancers,” “Colorectal Carcinoma,” “Carcinoma, Colorectal,” “Carcinomas, Colorectal,” “Colorectal Carcinomas,” “Systemic Immune-Inflammation Index,” “systemic immune inflammation index.” The detailed literature searching strategy was depicted in Supplementary Table S1.

Inclusion criteria: 1) CRC confirmed by pathological diagnosis; 2) treatment involving surgery, radiochemotherapy, neoadjuvant radiochemotherapy, or a combination thereof; 3) evaluation of the prognostic significance of the SII for PFS or OS; 4) availability of HRs with 95% confidence intervals (CIs), either reported or derivable from data; 5) stratification into high and low SII groups based on predefined cut-off values; 6) publication as a full-text article.

Exclusion criteria were as follows: 1) reviews, commentaries, conference abstracts, case reports, or letters; 2) insufficient data to estimate HRs and CIs; 3) lack of survival outcomes; and 4) duplicate or overlapping datasets.

WPF and JHY screened titles and abstracts, reviewed full texts, and assessed study eligibility independently. Discrepancies were resolved through discussion.

WPF and JHY extracted the data independently, with discrepancies resolved through consensus. The first author, publication year, study location, study design, sample size, patient age, study duration, treatment modality, type of immune checkpoint inhibitor, timing of detection, cut-off value, follow-up period, and HRs with 95% CIs for OS and PFS were extracted. When studies simultaneously reported both univariate and multivariate analysis results, multivariate analysis results were preferentially extracted. For studies reporting “Low SII/High SII” comparisons, HRs and CIs were transformed by calculating reciprocals and inverting confidence limits to ensure consistent “High SII/Low SII” comparisons across analyses.

Study quality was evaluated by Newcastle-Ottawa Scale (NOS), which assesses selection, comparability, and outcome domains, with a maximum score of nine. High quality studies score 7-9. Detailed scoring criteria and reasons for each assessment are provided in Supplementary Table S2.

Pooled HRs with 95% CIs were computed to estimate the prognostic significance of the SII in CRC. Cochran’s Q test and Higgins’ I² statistic were used to evaluate the heterogeneity (11). Heterogeneity was deemed significant when the Q-test P-value was < 0.1 or I² exceeded 50%. Data synthesis was conducted using a random-effects model. Sensitivity and subgroup analyses were performed to assess result robustness and identify heterogeneity sources. Funnel plots, Egger’s tests were used to examine the publication bias. A two-sided P-value < 0.05 was considered statistically significant. All analyses were conducted using STATA 15.0 and Review Manager 5.4.

A total of 349 articles were retrieved: PubMed (n=142), Embase (n=105), Cochrane Library (n=8), and Web of Science (n=94). After removing 151 duplicates, 198 records remained. Screening of titles and abstracts excluded 131 records, comprising 117 unrelated diseases, 10 conference abstracts, book chapters, or letters, and 4 meta-analyses or systematic reviews. Sixty-seven articles were identified as potentially eligible and retrieved for full-text review. Twelve were unavailable, leaving 55 for further assessment. Twenty were excluded due to insufficient data or absence of outcome indicators. Ultimately, 35 studies comprising 26812 patients were included (Figure 1) (1–6, 8–40).

Studies were conducted across multiple countries, with the majority from Asia (China, Japan), followed by Europe (Italy, Hungary, Spain) and North America (USA). Four publications each included two cohorts, resulting in a total of 40 cohorts for analysis. Among these, 38 were retrospective and 2 were prospective. All cohort studies were published in English between 2016 and 2025. Study populations included patients with metastatic CRC, those undergoing surgical resection, and individuals receiving surgery, chemotherapy, targeted therapy, or immunotherapy. The SII was primarily measured at baseline or preoperatively, with some studies evaluating postoperative levels. Cut-off values varied substantially across studies, ranging from 340 to 1505. Detailed characteristics of studies are presented in Table 1.

Study quality was evaluated using the NOS, with scores of 7–9 reflecting high methodological rigor (Supplementary Table S2). The mean NOS score was 7.8, indicating overall good quality of included studies.

Sensitivity analysis confirmed the stability of the baseline SII findings, as effect sizes remained consistent when individual studies were sequentially excluded. No single study exerted a disproportionate influence on the results for OS (Figure 3A) or PFS (Figure 3B), supporting the overall reliability of the analysis.

Publication bias was assessed using funnel plots with pseudo 95% confidence limits and Egger’s regression test. Visual inspection of the funnel plots for OS and PFS suggested an approximately symmetric distribution of studies within the 95% confidence boundaries (Figures 4A, B), with no obvious small-study effects. Consistently, Egger’s test was not statistically significant for either OS (p = 0.669) or PFS (p = 0.261), indicating no evidence of publication bias.

A critical challenge hindering the clinical application of SII is the substantial heterogeneity in cut-off values across studies, ranging from 340 to 1505 in our analysis (1, 9).Notably, despite the wide variation in cut-off thresholds across studies, our post-hoc meta-regression did not identify the SII cut-off value as a significant source of heterogeneity for either OS or PFS, suggesting that threshold variability alone may not explain the observed between-study heterogeneity (Table 3). This variability stems from multiple factors: 1) Population differences: Genetic backgrounds, environmental exposures, and baseline inflammatory states vary across geographic regions and ethnic groups; 2) Statistical methodologies: Different approaches for determining optimal thresholds, including receiver operating characteristic (ROC) curve analysis, median values, quartiles, or tertiles; 3) Treatment contexts: SII thresholds may differ between neoadjuvant, adjuvant, and palliative settings; 4) Disease characteristics: Optimal cut-offs may vary between colon and rectal cancers, as evidenced by the different thresholds reported for colon (535) versus rectal (792) liver metastases (7).

This heterogeneity poses significant challenges for clinical implementation and cross-study comparisons. We recommend that future research should: 1) Conduct large-scale prospective multicenter studies using standardized ROC analysis to establish and validate universally applicable optimal cut-off values; 2) Report SII both as continuous variables and categorical variables based on specific thresholds; 3) Consider population-specific reference ranges accounting for age, ethnicity, and geographic factors; 4) Develop dynamic monitoring protocols with standardized time points for SII assessment throughout the treatment trajectory.

Elevated SII reflects a systemic inflammatory-immune imbalance characterized by: 1) Neutrophilia promoting tumor progression through VEGF and matrix metalloproteinases (MMPs) release, facilitating angiogenesis and metastasis; 2) Thrombocytosis enhancing tumor cell survival in circulation through platelet-tumor aggregate formation, protecting circulating tumor cells from immune surveillance and shear stress; 3) Lymphocytopenia impairing cytotoxic T cell-mediated antitumor immunity, reducing immunosurveillance capacity 4). This inflammatory-immune imbalance contributes to tumor progression by enhancing the infiltration of immunosuppressive cells, including myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), within the tumor microenvironment.

Our comprehensive subgroup analysis by TNM stage revealed important insights: 1) Stage III patients showed the strongest association between elevated SII and poor prognosis (OS: HR = 3.65), suggesting SII may be particularly valuable for risk stratification in this intermediate-stage population where treatment intensification decisions are most critical; 2) Stage I-II patients demonstrated significant associations (OS: HR = 2.31), indicating SII’s utility in identifying high-risk early-stage patients who might benefit from adjuvant therapy; 3) Stage IV/metastatic patients showed consistent prognostic value (OS: HR = 2.47), supporting SII’s role in treatment monitoring and prognosis prediction in advanced disease.

These findings suggest that SII’s prognostic value varies across disease stages, likely reflecting different biological contexts: in early-stage disease, SII may indicate occult micrometastatic burden and systemic inflammatory response to tumor presence; in advanced disease, SII may reflect tumor burden, treatment response, and overall host inflammatory status.

Our analysis by tumor location revealed that SII’s prognostic value was consistent across colon, rectal, and mixed colorectal cancers, with HRs of 2.18, 1.96, and 2.05 respectively for OS. However, we acknowledge an important limitation: most original studies did not provide hazard ratios stratified by tumor location, preventing more granular analysis of potential differences between right-sided, left-sided, and rectal cancers. This represents a significant knowledge gap, as tumor location is increasingly recognized as a critical prognostic and predictive factor in CRC.

Future research should prioritize providing location-stratified analyses, as biological differences between proximal and distal CRC may influence SII’s prognostic value. Right-sided tumors are associated with different molecular features (higher rates of microsatellite instability, BRAF mutations, and CpG island methylator phenotype), distinct microbiome profiles, and different patterns of systemic inflammation compared to left-sided tumors (4).

Our analysis revealed important differences based on treatment modality and timing: 1) Post-treatment SII measurements showed stronger associations with prognosis than baseline measurements, suggesting SII’s potential as a treatment response biomarker; 2) The attenuated association in patients receiving bevacizumab combination therapy may reflect this agent’s anti-inflammatory properties; 3) SII’s consistent prognostic value across different treatment modalities supports its broad applicability.

These findings suggest potential clinical applications: 1) Baseline SII assessment for initial risk stratification and treatment planning; 2) Serial SII monitoring during treatment to assess treatment response and detect early progression; 3) Integration with imaging and other biomarkers for comprehensive treatment monitoring. Recent studies suggest that dynamic SII changes (ΔSII) may offer superior prognostic precision compared to static measurements (5).

A significant limitation of our study is the geographic imbalance in included studies, with the majority originating from Asia (particularly China and Japan), limited representation from Europe, and minimal representation from North America, Africa, and other regions. This geographic bias raises important questions about the generalizability of our findings across different populations.

Several factors may contribute to geographic differences in SII prognostic value: 1) Genetic and ethnic variations in inflammatory responses and immune function; 2) Environmental factors including diet, microbiome composition, and exposure to infectious agents; 3) Healthcare system differences affecting cancer screening, diagnosis timing, and treatment protocols; 4) Population-specific baseline inflammatory marker ranges.

We strongly recommend that future research should prioritize inclusion of more diverse, globally representative populations. Multi-center international collaborations are essential to validate SII’s prognostic value across different ethnic groups and healthcare systems. Only through such efforts can SII be established as a truly universal prognostic biomarker suitable for global clinical implementation.

Our study provides significant incremental value compared to previous meta-analyses, particularly the recent analysis by Tan et al.: 1) Larger sample size: 35 studies with 26812 patients vs. 27 studies with 18,420 patients; 2) More recent literature: search updated to November 2025 vs. March 2024; 3) More comprehensive subgroup analyses: including TNM stage and tumor location analyses not performed in previous studies; 4) More detailed methodological reporting: comprehensive sensitivity analyses and quality assessment.

These enhancements provide stronger evidence for SII’s prognostic value and more nuanced understanding of its clinical applications across different patient subgroups.

This study has several limitations requiring cautious interpretation. First, the marked heterogeneity in SII cut-off values across studies may compromise result comparability and clinical applicability. Second, the predominance of retrospective designs (38 of 40 cohorts) introduces potential confounders and selection bias. Third, the geographic imbalance limits generalizability to non-Asian populations. Fourth, insufficient reporting of tumor location-specific data prevented more granular analysis of potential anatomical heterogeneity.

Future research should address these limitations through: 1) Large-scale prospective multicenter studies with standardized protocols; 2) Development and validation of universally applicable SII cut-off values using ROC analysis; 3) Integration of SII with molecular subtyping and other biomarkers for enhanced prognostic accuracy; 4) Investigation of SII’s role in treatment selection and monitoring; 5) Exploration of dynamic SII monitoring protocols; 6) Expansion to more geographically diverse populations.

Based on our findings, we propose the following clinical implementation framework: 1) Baseline SII assessment at diagnosis for initial risk stratification; 2) Serial SII monitoring during treatment (e.g., every 3 months) to assess treatment response; 3) Integration with TNM staging to refine prognostic assessment, particularly in Stage III disease; 4) Consideration of treatment intensification for patients with persistently elevated SII; 5) Development of nomograms incorporating SII with other clinical and molecular variables for personalized prognostication.