Surgical safety of preoperative or perioperative immune checkpoint inhibitors in colorectal cancer surgery: protocol for a systematic review and meta-analysis focusing on anastomotic leakage

Abstract

Introduction

Immune checkpoint inhibitors (ICIs) are increasingly used before colorectal cancer surgery, particularly for mismatch repair-deficient or microsatellite instability-high tumours. Their surgery-specific safety, especially the risk of anastomotic leakage (AL), remains uncertain. We aim to synthesise AL and other postoperative outcomes after preoperative or perioperative ICI therapy.

Methods and analysis

We will search PubMed/MEDLINE, Embase, Web of Science, Scopus, CENTRAL, ClinicalTrials.gov, CNKI and the WHO International Clinical Trials Registry Platform from inception to July 2026 without language restrictions. We will include randomised and non-randomised comparative studies, cohort and case–control studies and single-arm studies reporting postoperative outcomes after preoperative or perioperative ICI therapy; case reports and case series with fewer than five surgically treated patients will be excluded. The primary outcome is AL among patients who underwent bowel resection with a constructed intestinal anastomosis. Two reviewers will independently screen records, extract data and assess risk of bias using design-appropriate validated tools. Comparative and non-comparative studies will be analysed separately: risk ratios from comparative studies will be pooled in random-effects models, and single-arm AL rates will be pooled as proportions using a binomial-logit generalised linear mixed model. Heterogeneity, prespecified subgroup analyses, sensitivity analyses and the certainty of the evidence (Grading of Recommendations Assessment, Development and Evaluation) will be reported.

Ethics and dissemination

Ethics approval is not required because only published aggregate data will be used. Results will be submitted to a peer-reviewed journal and presented at scientific meetings. Extraction materials, analytical code and the derived dataset will be shared in an open repository with the completed review.

PROSPERO registration number

CRD420261426134.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This review will provide a dedicated, surgery-focused synthesis of anastomotic leakage after preoperative or perioperative immune checkpoint inhibitor therapy in colorectal cancer.

• The primary outcome will use a denominator restricted to patients who underwent bowel resection with a constructed anastomosis, reducing denominator-related misclassification.

• Comparative and non-comparative studies will be synthesised separately, and dependent effect estimates will be handled using multilevel methods only when the data support them.

• A multilevel random-effects model will be considered only when a study contributes multiple statistically dependent effect estimates; otherwise, one independent estimate per study will be used in each analysis.

• The principal limitations are expected heterogeneity, incomplete reporting of surgery-specific outcomes and the likely predominance of small non-randomised or single-arm studies.

Introduction

Colorectal cancer (CRC) remains one of the most common malignant tumours worldwide and the second leading cause of cancer-related mortality, with approximately 1.9 million new cases and over 900 000 deaths annually.¹ Surgical resection is the cornerstone of curative treatment for localised colon and rectal cancer.² For patients undergoing CRC surgery—particularly those requiring low anterior resection, total mesorectal excision (TME), or other procedures involving intestinal reconstruction—the integrity and safety of the bowel anastomosis is a central surgical concern. Anastomotic leakage (AL) is among the most feared postoperative complications because it may result in pelvic sepsis, intra-abdominal infection, reoperation, prolonged hospitalisation, permanent stoma formation, delayed or omitted adjuvant therapy, impaired quality of life and potentially worse long-term oncological outcomes including increased local recurrence and reduced survival.^3–7 The reported incidence of AL after colorectal surgery ranges from 3% to 25% depending on tumour location, anastomotic level, patient risk factors and definition used.^{3 8}

In recent years, immune checkpoint inhibitors (ICIs) have fundamentally reshaped the treatment landscape of CRC, particularly for mismatch repair-deficient or microsatellite instability-high (dMMR/MSI-H) disease, where programmed cell death protein 1 (PD-1) blockade is now a standard treatment in the advanced setting.^9–11 Building on this success, neoadjuvant and perioperative ICI therapy—alone or combined with cytotoxic T-lymphocyte associated protein 4 (CTLA-4) blockade, chemotherapy or radiotherapy—has shown high rates of pathological response in resectable CRC across an expanding series of trials and cohort studies,^12–19 and is increasingly integrated into curative treatment pathways before surgery.

As these treatments are deployed earlier in the disease course, clinical attention has appropriately shifted from antitumour efficacy alone towards their safe integration into surgical treatment pathways. For colorectal surgeons and multidisciplinary teams, an unresolved question is whether preoperative exposure to ICIs affects postoperative inflammation, tissue healing and anastomotic safety.

The biological relationship between ICI therapy and anastomotic healing is complex and clinically uncertain. ICIs enhance T-cell activity and pro-inflammatory cytokine signalling, which could theoretically impair wound healing and collagen deposition at the anastomotic site.²⁰ ICI-related colitis, endocrinopathy, nutritional compromise and the frequent need for immunosuppressive therapy (particularly corticosteroids) for immune-related adverse events (irAEs) may further influence tissue perfusion, local inflammatory balance and intestinal healing capacity.^{21 22} Conversely, ICI therapy may avoid some deleterious local tissue effects associated with cytotoxic chemotherapy or pelvic radiotherapy—such as fibrosis, microvascular injury, tissue oedema and obliteration of surgical planes—that are well-established risk factors for anastomotic complications.^{23 24} Therefore, the surgical safety profile of preoperative ICI therapy cannot be extrapolated from the existing experience with conventional neoadjuvant chemotherapy or chemoradiotherapy.

Previous reviews have mainly examined pathological response, clinical response and general treatment-related safety after neoadjuvant immunotherapy in CRC.¹⁹ These outcomes do not establish surgery-specific safety: AL is often embedded within overall morbidity and may be reported without a denominator restricted to patients who underwent bowel anastomosis. Consequently, evidence on AL after perioperative ICI therapy remains fragmented and has not been synthesised as a dedicated outcome.

Objectives

Primary objective

To estimate the pooled incidence of AL after CRC surgery among patients who received preoperative or perioperative ICI therapy and underwent bowel resection with intestinal anastomosis.

Secondary objectives

1. To evaluate the pooled incidence of overall postoperative complications and severe postoperative complications (Clavien-Dindo grade ≥III).

2. To assess surgery-specific outcomes, including intra-abdominal or pelvic infection, surgical site infection, postoperative bleeding, ileus, reoperation, readmission, intensive care unit admission, postoperative mortality and length of postoperative hospital stay.

3. To evaluate stoma-related outcomes, including protective stoma formation, stoma reversal, delayed stoma reversal and permanent stoma.

4. To compare postoperative surgical safety outcomes between patients receiving ICIs and those receiving standard non-immunotherapy treatment (eg, surgery alone, neoadjuvant chemotherapy, chemoradiotherapy) where comparative data are available.

5. To explore whether surgical safety outcomes differ according to tumour location, mismatch repair status, ICI regimen, combination therapy, radiotherapy exposure, surgical procedure, protective stoma use and the interval between last ICI dose and surgery.

6. To assess risk of bias and certainty of evidence for key surgical safety outcomes using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.

Methods and analysis

Protocol and registration

This systematic review and meta-analysis will be conducted according to a prespecified protocol and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement²⁵ for systematic reviews and meta-analyses. The present protocol was developed following the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 guidance²⁶ (see online supplemental file 1 for the completed PRISMA-P checklist). This protocol has been registered in PROSPERO (registration number CRD420261426134; registration date: 18 June 2026; available from: https://www.crd.york.ac.uk/PROSPERO/view/CRD420261426134). The searches will be updated immediately before final analysis. Screening and data extraction are anticipated to be completed by December 2026. The full project timeline is provided in online supplemental file 4. Any important protocol amendments will be documented with the date, description and rationale in the final published review. The review will be conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions.²⁷

Eligibility criteria

Population

We will include adult patients (aged ≥18 years) with histologically confirmed CRC (colon cancer or rectal cancer) who have received preoperative or perioperative ICI therapy and subsequently undergone CRC surgery. Patients with dMMR/MSI-H CRC constitute the primary population of interest.^{9 14 15} Studies including patients with proficient mismatch repair/microsatellite-stable (pMMR/MSS) tumours, or mixed MMR status populations, will also be eligible if relevant postoperative surgical outcomes are reported separately or can be obtained.

For the primary outcome of AL, the analytical denominator will be restricted to patients who underwent bowel resection with a constructed intestinal, colorectal or coloanal anastomosis, whenever such information is available or can be derived. Patients who underwent abdominoperineal resection, Hartmann’s procedure, definitive stoma without anastomosis, local excision without bowel anastomosis or non-operative management (including watch-and-wait after clinical complete response) will not contribute to the AL denominator.

Intervention or exposure

The intervention of interest is preoperative or perioperative ICI therapy, including neoadjuvant or preoperative use of PD-1 inhibitors, programmed death-ligand 1 (PD-L1) inhibitors, CTLA-4 inhibitors or their combinations, administered before CRC surgery. Eligible agents include but are not limited to nivolumab, pembrolizumab, ipilimumab, dostarlimab, sintilimab, toripalimab, camrelizumab, tislelizumab, atezolizumab, durvalumab, envafolimab and other ICIs.

Studies evaluating ICI monotherapy or ICI therapy combined with chemotherapy, radiotherapy, chemoradiotherapy, targeted therapy or other systemic treatments will be considered. Different treatment strategies will be analysed separately where data permit.

Studies in which ICIs were administered only as postoperative adjuvant therapy (ie, no preoperative or perioperative exposure) will not be included in the primary analysis of AL because such exposure does not precede the early postoperative period in which AL typically occurs. These studies may be narratively summarised only if they report relevant postoperative safety data.

Comparator

Both comparative and non-comparative studies will be included. Eligible comparators include surgery alone, standard care, conventional neoadjuvant chemotherapy, neoadjuvant chemoradiotherapy, total neoadjuvant therapy, postoperative adjuvant therapy without preoperative ICI or other non-immunotherapy treatment strategies. Studies comparing different ICI regimens will also be eligible if postoperative surgical outcomes are reported.

Study designs

We will include randomised controlled trials, non-randomised clinical trials, prospective cohort studies, retrospective cohort studies, case–control studies, single-arm clinical studies and real-world studies. Conference abstracts and trial registry records will be considered if they provide sufficient extractable postoperative surgical outcome data.

Exclusion criteria

We will exclude case reports, case series with fewer than five surgically treated patients, animal studies, in vitro studies, narrative reviews, editorials, commentaries and letters without original data. Studies focusing exclusively on metastatic or unresectable CRC without subsequent surgical resection will be excluded. Studies in which the denominator for patients undergoing intestinal anastomosis cannot be determined from the available information will be excluded from the primary meta-analysis of AL and summarised narratively instead.

Setting

Studies conducted in any hospital setting in which CRC surgery is performed will be eligible, including tertiary academic centres, comprehensive cancer centres, community and general hospitals and mixed clinical practice, regardless of whether operations are performed by colorectal surgeons, surgical oncologists or general surgeons. Hospital type and the specialty of the operating surgeon will be recorded during data extraction and considered as potential sources of heterogeneity. No restriction will be placed on country, region, healthcare system or surgical approach (open, laparoscopic, robotic or transanal surgery). The complete inclusion and exclusion criteria are summarised in table 1.

Table 1. Inclusion and exclusion criteria.

Domain	Inclusion criteria	Exclusion criteria
Population	Adults (≥18 years) with histologically confirmed colon or rectal cancer who received preoperative or perioperative ICI therapy and subsequently underwent colorectal surgery.	Paediatric populations; non-colorectal malignancies without separately extractable CRC data; exclusively metastatic or unresectable CRC without subsequent surgery.
Intervention/exposure	Preoperative, neoadjuvant or perioperative PD-1, PD-L1 or CTLA-4 inhibitor therapy, alone or combined with chemotherapy, radiotherapy, chemoradiotherapy, targeted therapy or another ICI.	ICI administered only after surgery, with no preoperative or perioperative exposure, for the primary analysis.
Comparator	Surgery alone, standard care, non-ICI neoadjuvant therapy or another ICI regimen; no comparator required for eligible single-arm studies.	No exclusion solely because a comparator is absent.
Study design	Randomised trials, non-randomised clinical trials, prospective or retrospective cohorts, case–control studies, single-arm clinical studies, case series with ≥5 surgically treated patients and real-world studies.	Case reports; case series with <5 surgically treated patients; reviews, editorials, commentaries and letters without original data; animal or in vitro studies.
Setting	Any hospital setting in which colorectal cancer surgery is performed: tertiary academic centres, cancer centres, community or general hospitals; any operating surgeon specialty (colorectal surgery, surgical oncology or general surgery); any country, region, healthcare system and surgical approach (open, laparoscopic, robotic or transanal).	No exclusion based on hospital type, surgeon specialty, country, region, healthcare system or surgical approach.
Outcomes	At least one extractable postoperative surgical outcome after CRC surgery. For the primary AL meta-analysis, both the number of AL events and the denominator of patients with a constructed bowel anastomosis must be available or derivable.	No extractable postoperative surgical outcome. Studies with an indeterminate anastomosis denominator are excluded from the primary AL meta-analysis but may contribute to other outcomes or narrative synthesis.
Report characteristics	Full articles, conference abstracts, preprints and registry results with sufficient extractable data; all languages; no publication-date restriction.	Duplicate or overlapping reports that provide no additional data; reports for which relevant data remain unobtainable after reasonable clarification attempts.

AL, anastomotic leakage; CRC, colorectal cancer; CTLA-4, cytotoxic T-lymphocyte associated protein 4; ICIs, immune checkpoint inhibitors; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1.

Information sources and search strategy

A comprehensive literature search will be conducted in the following electronic databases from inception to July 2026 (the final search date): PubMed/MEDLINE, Embase, Web of Science, Scopus, the Cochrane Central Register of Controlled Trials (CENTRAL), ClinicalTrials.gov, China National Knowledge Infrastructure (CNKI) and the WHO International Clinical Trials Registry Platform (ICTRP). Additional sources will include reference lists of included studies and relevant reviews, major oncology and colorectal surgery conference abstract repositories (American Society of Clinical Oncology (ASCO), European Society for Medical Oncology (ESMO), American Society of Colon and Rectal Surgeons (ASCRS) and European Society of Coloproctology (ESCP)), and preprint servers (medRxiv) when sufficient postoperative outcome data are available. The rationale for each bibliographic database and registry is presented in table 2.

Table 2. Rationale for bibliographic databases and other search sources.

Source	Coverage	Rationale for inclusion
PubMed/MEDLINE	Biomedical and clinical literature	Core source for indexed medical, oncological and surgical studies; combines MeSH with title/abstract terms.
Embase	Biomedical, pharmacological and conference literature	Provides broad drug and device indexing through Emtree and extensive European and conference coverage.
Web of Science Core Collection	Multidisciplinary citation database	Captures multidisciplinary and emerging studies and enables forward citation tracking.
Scopus	Multidisciplinary abstract and citation database	Adds broad international journal and conference coverage and complements Web of Science citation searching.
CENTRAL	Controlled trials and trial records	High-sensitivity source for randomised and quasi-randomised trials assembled from multiple databases and handsearching.
CNKI	Chinese academic literature	Captures Chinese-language studies and locally reported clinical evidence that may not be indexed internationally.
ClinicalTrials.gov	Clinical trial registry	Identifies ongoing, completed and unpublished trials and supports detection of selective outcome reporting.
WHO ICTRP	International trial-registry portal	Aggregates records from multiple national registries and improves international coverage of registered studies.
Conference repositories and medRxiv	Grey literature and preprints	Reduces time-lag and publication bias by identifying recent or incompletely published evidence; eligible reports must contain extractable postoperative data.

For conference abstracts and grey literature, the following criteria will be applied: (1) Only abstracts that provide extractable postoperative outcome data (eg, number of patients with AL, denominator for anastomosis) will be considered for inclusion; (2) if the same study is reported in both an abstract and a full-text publication, the full-text publication will be used as the primary source, with complementary data extracted from the abstract if needed; (3) authors of conference abstracts will not be contacted for additional data beyond what is publicly available in the abstract; (4) preprints will be screened but will be replaced by peer-reviewed versions when available, with version dates documented.

No language restrictions will be applied. Studies in all languages will be screened, and translation assistance will be sought where necessary. No date restrictions will be applied. The complete search strategy for each database will be reported in the final review. Although ICIs have been in clinical use for only about a decade, searching from inception follows Cochrane Handbook guidance and avoids arbitrary date cut-offs that could miss early-phase immunotherapy trials reporting surgical outcomes.²⁷ Because the search strategy is built around immunotherapy-specific concepts (ICIs, PD-1/PD-L1, CTLA-4 and named agents), records from the pre-immunotherapy era will contain few, if any, of these terms and will be excluded rapidly at title and abstract screening at negligible cost.

Complete reproducible search strategies for all sources are provided in online supplemental file 2.

The strategy will be adapted for each database using database-specific controlled vocabulary, fields and proximity operators. To maximise sensitivity, postoperative outcomes will not be used as a mandatory search concept because such outcomes may be reported only in the full text or the online supplements of eligible reports. Complete reproducible strategies are provided in online supplemental file 2 and will undergo Peer Review of Electronic Search Strategies (PRESS) 2015 peer review before the final searches are run.

Study selection

All retrieved records will be imported into a reference management software (eg, EndNote, Zotero or Covidence), and duplicate records will be removed. Before formal screening, a pilot test will be conducted where two reviewers will independently screen a random sample of 50 records to calibrate eligibility criteria and ensure consistent application. Any discrepancies will be discussed and resolved, with refinements made to the screening criteria as needed.

Following the pilot phase, two reviewers will independently screen titles and abstracts against the predefined eligibility criteria. Studies considered potentially eligible by either reviewer will proceed to full-text review. Full-text articles will be independently assessed for eligibility by two reviewers, with reasons for exclusion documented. Disagreements at any stage will be resolved through discussion; if consensus cannot be reached, a third reviewer will be consulted. Inter-rater reliability (Cohen’s κ) will be calculated at both title/abstract and full-text screening stages. The study selection process will be documented using a²⁵ PRISMA 2020 flow diagram.²⁵ If multiple reports of the same study are identified (eg, trial registration, conference abstract and full-text publication), the most complete and recent report will be used as the primary source, with complementary data extracted from other reports where needed.

Data extraction

Two reviewers will independently extract data using a standardised, pre-piloted data extraction form developed in Microsoft Excel (or equivalent electronic format). Before formal data extraction, a pilot test will be conducted on three to five included studies to ensure consistent understanding and application of the extraction categories. The data extraction form will be refined based on pilot feedback.

The complete data extraction form template, including all fields and coding instructions, is provided in online supplemental file 3. Disagreements will be resolved through discussion or consultation with a third reviewer. If important outcome data are missing or unclear, study authors will be contacted by email (up to two attempts at 2-week intervals). Individual participant data will not be sought.

The following data will be extracted:

Study characteristics

First author, publication year, country/countries, study design, study period, number of participating centres, hospital type (tertiary academic centre, cancer centre or community/general hospital), specialty of the operating surgeon (colorectal surgery, surgical oncology or general surgery), total sample size, funding source and trial registration number.

Patient characteristics

Age, sex, body mass index, American Society of Anesthesiologists classification, comorbidities (including diabetes, hypertension, cardiovascular disease, smoking status), preoperative nutritional status, serum albumin level, preoperative corticosteroid or immunosuppressant use, tumour location (colon vs rectum), clinical and pathological tumour, node, metastases stage, MMR/MSI status, history of prior abdominal or pelvic surgery and prior treatment history.

Treatment details

ICI agent(s), dose regimen, number of cycles, treatment duration, combination therapy details (chemotherapy agents, radiotherapy dose and fractionation, targeted therapy), treatment setting (neoadjuvant vs perioperative vs preoperative plus adjuvant), occurrence and grade of irAEs (any grade and grade ≥III, per the Common Terminology Criteria for Adverse Events (CTCAE)), requirement for corticosteroid or other immunosuppressive therapy for irAEs and interval between the last dose of ICI and surgery.

Surgical details

Surgical approach (open, laparoscopic, robotic, transanal), type of resection (right hemicolectomy, left hemicolectomy, sigmoidectomy, anterior resection, low anterior resection, abdominoperineal resection, total colectomy, proctocolectomy), elective versus emergency surgery, presence and type of intestinal anastomosis (ileocolic, colocolic, colorectal, coloanal, ileorectal, ileoanal), anastomotic technique (handsewn vs stapled vs hybrid), anastomotic level (distance from anal verge, where available), intraoperative perfusion assessment (indocyanine green fluorescence angiography or other methods), intraoperative air leak test, protective stoma creation (ileostomy or colostomy), operative time, intraoperative blood loss and conversion from minimally invasive to open surgery.

Outcome data

AL (with definition used, diagnostic method, severity classification and time window), overall postoperative complications, severe postoperative complications (Clavien-Dindo grade ≥III), specific complications (intra-abdominal/pelvic infection, surgical site infection, organ/space surgical site infection, postoperative bleeding, ileus, anastomotic stricture), reoperation rate, readmission rate, intensive care unit admission, postoperative mortality (30-day and/or 90-day), length of postoperative hospital stay, protective stoma formation and reversal (including time to reversal, delayed reversal and permanent stoma), delayed surgery due to irAEs, R0 resection rate, pathological complete response (pCR; ypT0N0), major pathological response (MPR; ≤10% residual viable tumour) and treatment-related adverse events (any grade and grade ≥III).

Outcomes

Primary outcome

The primary outcome is AL after CRC surgery among patients who underwent bowel resection with a constructed intestinal, colorectal or coloanal anastomosis. AL will be defined according to the criteria used in each included study. Where possible, reported definitions will be mapped to the International Study Group of Rectal Cancer (ISGRC) definition: ‘communication between the intra- and extraluminal compartments owing to a defect of the intestinal wall at the anastomosis’,⁸ and classified using the ISGRC grading system (Grade A: requiring no active therapeutic intervention; Grade B: requiring active therapeutic intervention but manageable without reoperation; Grade C: requiring reoperation). Where this system is not used by individual studies, AL will be classified into clinically relevant categories:

• Any AL: including both radiologically detected and clinically apparent leaks.

• Clinically relevant AL: AL requiring active therapeutic intervention (antibiotics, percutaneous drainage, endoscopic or radiological intervention, reoperation or stoma creation)—corresponding approximately to ISGRC Grade B and C.

• Severe AL: AL requiring reoperation or associated with postoperative mortality—corresponding approximately to ISGRC Grade C.

• The primary time window will be within 30 days after surgery or during the index postoperative hospital stay. Where available, 90-day outcomes will be extracted and analysed separately.

Secondary outcomes

Secondary outcomes include: (1) Postoperative surgical outcomes: overall postoperative complications (any Clavien-Dindo grade), severe postoperative complications (Clavien-Dindo grade ≥III), intra-abdominal or pelvic infection, organ/space surgical site infection, superficial surgical site infection, postoperative bleeding requiring transfusion or intervention, ileus, reoperation for any cause, readmission within 30 days, intensive care unit admission, postoperative mortality (30-day and 90-day) and length of postoperative hospital stay. (2) Stoma-related outcomes: protective (defunctioning) stoma formation rate, stoma reversal rate, time to stoma reversal, delayed stoma reversal (defined as >6 months from creation if data permit) and permanent stoma (defined as stoma not reversed within the study follow-up period or at final follow-up). (3) Perioperative process outcomes: delayed surgery due to irAEs (defined as surgery postponed beyond the planned date because of ICI-related toxicity), interval between last ICI dose and surgery (days), operative time (minutes), intraoperative blood loss (mL) and conversion from minimally invasive to open surgery. (4) Oncological and pathological outcomes: R0 resection rate, pCR (ypT0N0), MPR (≤10% residual viable tumour) and treatment-related adverse events (any grade and grade ≥III, graded per CTCAE V.4.0 or V.5.0).

Risk of bias and study quality assessment

Risk of bias and study quality will be assessed independently by two reviewers using the following validated tools:

• Randomised controlled trials: The Cochrane Risk of Bias 2 (RoB 2) tool²⁸ will be applied to each relevant outcome. Domain and overall judgements will follow the RoB 2 algorithms and will be categorised as low risk of bias, some concerns or high risk of bias.

• Non-randomised comparative studies of interventions: The Risk Of Bias In Non-randomised Studies-of Interventions (ROBINS-I) tool,²⁹ with judgements across seven domains (confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, selection of reported results) categorised as low, moderate, serious or critical risk of bias.

• Non-comparative studies: the design-specific Joanna Briggs Institute (JBI) critical appraisal checklist will be used; the JBI checklist for case series³⁰ will be applied to case series, and the relevant JBI checklist will be selected for other non-comparative designs. Tool selection will be prespecified by study design and reported transparently.

• A single design-appropriate tool will be applied to each study; overlapping quality scales will not be used in parallel with outcome-level risk-of-bias tools.

Particular attention will be paid to confounding by tumour location (colon vs rectum), surgical procedure (TME vs non-TME), radiotherapy or chemoradiotherapy exposure, protective stoma use, anastomotic level, preoperative nutritional status, corticosteroid or immunosuppressant use and severity of irAEs. Established risk factors for AL include low anastomotic level, male sex, obesity, smoking, preoperative radiotherapy and prolonged operative time, all of which will be considered during data synthesis.^31–33

Risk of bias due to selective non-reporting will be assessed by comparing published reports with trial registry records (ClinicalTrials.gov, WHO ICTRP, Chinese Clinical Trial Register), published protocols, statistical analysis plans, conference abstracts and supplementary materials of the published reports where available. Selective non-reporting of surgical outcomes—especially AL and postoperative complications—will be specifically considered and documented.

Disagreements in risk of bias assessment will be resolved through discussion or consultation with a third reviewer. Results of risk of bias assessments will be presented using summary tables and visualisations (traffic-light plots and weighted bar plots).

Data synthesis

Narrative synthesis

A narrative synthesis will be conducted for all included studies, summarising study characteristics, patient populations, treatment protocols, surgical details and reported outcomes. The narrative synthesis will be structured around the primary and secondary outcomes, with particular attention to clinical and methodological heterogeneity across studies.

Quantitative synthesis (meta-analysis)

Where clinically and methodologically comparable data are available, comparative and non-comparative studies will be synthesised separately. One independent effect estimate per study will be selected for each analysis. If a study contributes multiple eligible, statistically dependent estimates that cannot be combined, a three-level random-effects model or robust variance estimation will be used, provided that the number of studies is adequate.

Comparative studies: For dichotomous outcomes (AL, postoperative complications, severe complications, infection, reoperation, readmission, mortality, stoma-related outcomes), effect estimates will be pooled using risk ratios (RRs) with 95% CIs. ORs will be used when RRs cannot be calculated (eg, when adjusted ORs are reported from multivariable analyses). Adjusted and unadjusted effect estimates will not be pooled together. For continuous outcomes (length of hospital stay, operative time, intraoperative blood loss), mean differences (MDs) will be used when the same measurement scale is used across studies and standardised MDs with Hedges’ g correction will be used when different measurement approaches are employed. If median and IQR or range are reported instead of mean and SD, these will be converted to approximate means and SDs using validated methods described by Wan et al³⁴ or Hozo et al,³⁵ as appropriate.

Non-comparative studies: pooled proportions with 95% CIs will be estimated using a generalised linear mixed model (GLMM) with binomial likelihood and logit link. Zero-event studies will be retained. Freeman-Tukey and continuity-corrected logit analyses will be used only as sensitivity analyses when estimable.

The GLMM will be the primary model for pooling proportions because it accommodates zero-event studies without an ad hoc continuity correction. Sensitivity analyses will assess whether conclusions change under alternative transformations.

Statistical model and heterogeneity assessment

Before performing meta-analysis, clinical and methodological heterogeneity across studies will be assessed qualitatively based on study designs, patient populations (eg, tumour location, MMR status), interventions (ICI agents, combination therapies), surgical techniques, outcome definitions and follow-up durations. This assessment will inform decisions about the appropriateness of quantitative synthesis.

Random-effects models will be used because clinical and methodological heterogeneity is anticipated. For conventional comparative meta-analyses, restricted maximum likelihood will estimate between-study variance (τ²) and Hartung-Knapp CIs will be used when appropriate. Heterogeneity will be described using τ², I² with its uncertainty interval and Cochran’s Q, while recognising that these statistics do not replace clinical assessment of comparability.

Prediction intervals will be reported when they are estimable and sufficiently informative, generally when at least five studies contribute to an analysis.³⁶ Analyses will be conducted in R (V.4.3 or later) using validated packages such as metafor and meta; package versions and analysis code will be reported with the completed review.

Meta-analysis will not be performed when populations, interventions, surgical procedures, outcome definitions or follow-up periods are too clinically or methodologically diverse, or when data are insufficient. This decision will not be based on an I² threshold alone. Such results will be synthesised narratively, with reasons for not pooling documented.

Assessment of publication bias and small-study effects

Funnel plots (with the effect measure on the x-axis and its SE on the y-axis) and Egger’s regression asymmetry test will be used to assess publication bias and small-study effects when at least 10 studies are available for a given outcome.³⁷ If fewer than 10 studies are available, formal statistical tests will not be performed and potential reporting bias will be discussed qualitatively. Asymmetry in funnel plots may also be explored by contour-enhanced funnel plots to distinguish publication bias from other sources of asymmetry (eg, true heterogeneity, selective outcome reporting). Selective non-reporting of AL and other surgery-specific complications will be specifically considered and discussed.

Subgroup analyses

Subgroup analyses will be undertaken only when categories are clinically coherent and each category contains sufficient independent studies for a meaningful comparison. As a general rule, at least five studies per subgroup category will be sought; analyses based on fewer studies will be labelled exploratory or presented narratively.

Core subgroup analyses (pre-planned and will be reported regardless of statistical significance): (1) Tumour location: colon cancer versus rectal cancer, (2) surgical procedure: TME-related rectal surgery versus non-TME colorectal surgery, (3) treatment strategy: ICI monotherapy versus ICI plus chemotherapy versus ICI plus radiotherapy/chemoradiotherapy versus ICI plus targeted therapy, (4) radiotherapy exposure: regimens with pelvic radiotherapy or chemoradiotherapy versus regimens without radiotherapy, (5) protective stoma use: protective (defunctioning) stoma versus no protective stoma and (6) study design: comparative studies versus single-arm studies.

Exploratory subgroup analyses (conducted only when ample data permit and interpreted cautiously): (7) Mismatch repair status: dMMR/MSI-H versus pMMR/MSS or mixed populations, (8) ICI regimen: PD-1/PD-L1 monotherapy versus combination immunotherapy (eg, PD-1+CTLA-4), (9) interval between last ICI dose and surgery: <4 weeks versus 4–8 weeks versus >8 weeks, (10) surgical approach: open versus laparoscopic/robotic surgery, (11) anastomotic level: high colorectal anastomosis versus low colorectal anastomosis versus coloanal anastomosis (where data permit) and (12) geographical region: Asia versus Europe versus North America versus other regions.

Categorical subgroup differences will be evaluated with an interaction test. Meta-regression will be reserved for analyses with adequate numbers of independent studies and will not be used to compare single-arm studies directly with comparative effect estimates. All subgroup findings will be interpreted as exploratory because of multiplicity, ecological bias and residual confounding.

Meta-regression

Random-effects meta-regression will be considered only when at least 10 independent studies are available per modelled covariate. The number of covariates will therefore be restricted according to the available evidence. Candidate moderators, selected a priori and assessed one at a time before any justified multivariable model, are year of publication, proportion of rectal cancer, proportion with dMMR/MSI-H disease, interval from last ICI dose to surgery, radiotherapy exposure and protective stoma use.

Sensitivity analyses

Sensitivity analyses, where applicable, will exclude studies at high risk of bias (RoB 2), serious or critical risk of bias (ROBINS-I), studies with fewer than 10 surgically treated patients, studies without a clear AL definition, studies combining ICI with pelvic radiotherapy or chemoradiotherapy and reports available only as abstracts, preprints or registry results. Alternative choices for outcome definition, effect measure, zero-event handling and statistical model will also be examined when sufficient studies are available.

Influence analysis will be performed using leave-one-out meta-analysis (iteratively omitting one study at a time) to assess whether any single study disproportionately influences the pooled estimate.

For comparative analyses of AL, alternative estimators and CI methods will be examined when sparse events make the primary random-effects analysis unstable. Fixed-effect results, if presented, will be explicitly labelled as sensitivity analyses and will not supersede clinical assessment of heterogeneity.

Certainty of evidence

The certainty of evidence for key outcomes (AL, overall postoperative complications, severe postoperative complications (Clavien-Dindo ≥III), infection, reoperation, postoperative mortality, length of hospital stay and stoma-related outcomes) will be assessed using the GRADE approach.³⁸ The certainty of evidence will be rated as high, moderate, low or very low based on assessments of risk of bias, inconsistency, indirectness, imprecision and publication bias.

For each GRADE domain, the following considerations will apply: (1) Risk of bias: downgraded if the majority of evidence comes from studies at high or serious risk of bias. (2) Inconsistency: downgraded if substantial unexplained heterogeneity is present (I² >50% or overlapping prediction intervals spanning opposite conclusions). (3) Indirectness: downgraded if there are important differences between the study populations, interventions or outcomes and those specified in the review question. (4) Imprecision: downgraded if the 95% CI for the pooled estimate includes both appreciable benefit and appreciable harm (for AL, a relative risk increase or decrease crossing 1.25 or 0.75, or an absolute incidence difference crossing a clinically important threshold). (5) Publication bias: downgraded if evidence of publication bias or selective outcome reporting is identified.

GRADE assessments will be performed independently by two reviewers, and disagreements will be resolved through discussion or consultation with a third reviewer. Summary of findings tables will be generated for key outcomes using GRADEpro GDT software (or equivalent).

Patient and public involvement

No patient or public involvement is planned for this systematic review and meta-analysis protocol. Patients and the public were not involved in the design, conduct or reporting of this protocol.

Ethics and dissemination

Ethical approval is not required for this systematic review and meta-analysis because it will use data from previously published studies and publicly available sources. The results of this systematic review and meta-analysis will be submitted for publication in a peer-reviewed journal in the fields of colorectal surgery, surgical oncology, gastrointestinal oncology or perioperative medicine. The findings may also be presented at relevant national and international academic conferences (eg, ASCRS, ESCP, ASCO, ESMO). The review will use published aggregate data. The extraction form, analytical code and derived dataset will be deposited in an open repository when the review is published.

Discussion

This systematic review and meta-analysis will synthesise available evidence on AL and postoperative surgical complications after preoperative or perioperative ICI therapy in CRC. The review is designed to address an important and clinically relevant evidence gap in the current literature. Although ICIs have demonstrated promising and often remarkable efficacy in selected CRC populations—particularly dMMR/MSI-H disease—their surgery-specific safety profile remains incompletely and inconsistently defined.

A key strength of this review is its dedicated focus on AL as a prespecified primary outcome. AL is not merely a general postoperative complication; it is a clinically meaningful surgical endpoint that directly affects postoperative recovery, reoperation risk, stoma formation, timing of adjuvant treatment, quality of life and long-term oncological outcomes.^{3–7 39–41} By separating AL from general treatment-related adverse events and overall postoperative morbidity, this review may provide more clinically relevant evidence for colorectal surgeons, surgical oncologists and multidisciplinary cancer teams.

Several limitations merit careful consideration. First, the definitions of AL and postoperative complications are expected to vary across studies. While standardised classification systems exist (eg, the ISGRC definition for AL⁸), they may not be uniformly applied, particularly in oncology-focused trials where surgical outcomes are not a primary endpoint. Second, many ICI efficacy trials and case series may not report surgery-specific outcomes in sufficient detail, and the denominator for patients who actually underwent bowel resection with intestinal anastomosis may be unclear or inconsistently reported. Third, clinical heterogeneity is anticipated because of differences in tumour location and stage, ICI regimen and duration, type and intensity of combination therapy (chemotherapy, radiotherapy, targeted therapy), surgical approach and technique, timing of surgery relative to ICI therapy, anastomotic level and protective stoma practices. Fourth, patients managed non-operatively after achieving a clinical complete response to ICI therapy—particularly relevant in rectal cancer¹⁴—will not contribute to the quantitative analysis of AL, which may introduce selection considerations when interpreting surgical safety among patients who ultimately proceed to surgery. Fifth, the inclusion of single-arm studies, while necessary given the emerging evidence base, limits the ability to derive robust comparative effect estimates. To address this, we will conduct separate analyses for comparative and non-comparative studies and will transparently report the strengths and limitations of each approach.

Despite these limitations, this review will provide a focused, systematic and clinically relevant synthesis of the surgical safety of perioperative ICIs in CRC. The findings may inform perioperative treatment planning, preoperative patient counselling, multidisciplinary tumour board decision-making, informed surgical consent discussions and the design of future clinical trials that should prospectively and systematically report AL and other surgery-specific complications using standardised definitions and with clear specification of the denominator. Prospective collaborative registries that capture both oncological and surgical outcomes in patients receiving perioperative ICI therapy are needed to generate higher-quality evidence in this rapidly evolving field.