Indocyanine green fluorescence angiography for prevention of anastomotic leak in colorectal surgery: a graded systematic review and meta-analysis with trial sequential analysis of randomized controlled trials

Abstract

Background

Anastomotic leak (AL) is a major driver of morbidity after colorectal resection. Indocyanine-green fluorescence angiography (ICG-FA) provides real-time intraoperative perfusion assessment, but its impact across randomized trials remains debated. This review aims to evaluate the efficacy and safety of ICG-FA compared with conventional intraoperative assessment in preventing AL in colorectal surgery.

Methods

A systematic review and a meta-analysis was conducted in accordance with the PRISMA guidelines, including RCTs comparing ICG-FA with conventional assessment. The primary outcome was overall AL; secondary outcomes included leak grade, anatomic location, operative metrics, and change in surgical plan (CISP). Pooled risk ratios (RR) or mean differences (MD) with 95% confidence intervals (CI) were calculated using random- or fixed-effects models. Trial sequential analysis (TSA) was used to evaluate evidence sufficiency.

Results

Eight RCTs (n = 4712) met the inclusion criteria. ICG-FA reduced overall AL versus control (risk ratio [RR] 0.68, 95% CI 0.58–0.81; I² = 0%). The absolute risk reduction was 3.6%, yielding a number needed to treat of around 24 to prevent one leak. TSA crossed the superiority boundary for the primary endpoint, indicating sufficient evidence of the assumed control risk and effect size. Benefits were more apparent for left-sided anastomoses and for lower-grade leaks (A and A + B), whereas severe (Grade C) leaks and reoperation did not differ significantly. Overall complications showed a small relative reduction, but TSA indicated insufficient information size. Other safety outcomes (surgical site infection, ileus, mortality) were similar between groups.

Conclusions

High-certainty evidence demonstrates that ICG-FA reduces overall and low-grade AL, principally in left-sided resections, without prolonging operative time. ICG-FA should be considered an evidence-based adjunct to optimize perfusion assessment in colorectal surgery.

Registration

PROSPERO CRD420251139372.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s10151-026-03388-4.

Introduction

Anastomotic leak (AL) remains one of the most consequential complications after colorectal resection [1, 2]. Even at contemporary high-volume centers, AL continues to drive postoperative morbidity, reoperation, prolonged hospitalization, and stoma formation, and it may compromise long-term oncological outcomes in cancer surgery [3]. The multifactorial pathophysiology of AL includes impaired tissue perfusion, excessive tension, technical factors, and patient-related risks; among these, inadequate microperfusion at the anastomotic edges is a consistently implicated mechanism [4–7].

Indocyanine green fluorescence angiography (ICG-FA) offers a real-time, non-invasive means of assessing tissue perfusion intraoperatively using near-infrared imaging [7]. ICG rapidly binds plasma proteins and emits fluorescence proportional to regional blood flow, enabling surgeons to visualize perfusion at the planned transection and anastomotic sites within seconds after injection, thereby permitting intraoperative assessment of bowel perfusion and modification of the surgical plan when necessary [8].

Early observational studies suggested ICG-FA might reduce AL; however, generalizability was limited by variable dosing, timing, qualitative interpretation, and unblinded designs prone to co-intervention effects [9–11]. More recently, randomized controlled trials (RCTs) have sought to isolate the incremental value of ICG-FA over standard visual and tactile assessment [9]. These RCTs broadly agree on feasibility and safety but differ in baseline risk profiles, surgical indication (benign vs malignant), anastomotic level (right colon to low rectum), diversion practices, and imaging protocols, factors that could modulate the magnitude and clinical relevance of any observed effect [9].

Prior meta-analyses combining RCTs and non-randomized studies suggested benefit but were constrained by clinical heterogeneity and potential confounding [12, 13]. The first RCT-only meta-analysis, conducted by Meyer et al., synthesized three trials (n = 964) and reported a modest reduction in overall AL (RR 0.67, 95% CI 0.46–0.99; moderate certainty) while emphasizing the need for larger trials, subgroup definition, and cost-effectiveness evaluation [9]. Then, Elmajdub et al. [14], synthesized five trials (n = 1369) and reported that the ICG-FA was associated with a 45% reduction in the risk of overall anastomotic leaks (OR: 0.550, p = 0.012). Subgroup analysis showed a 47% reduction in leaks for low anastomoses (OR: 0.53, p = 0.143) and a 69% reduction in grade A leaks (OR: 0.31, p = 0.008), while emphasizing the need for further RCTs with larger sample sizes to confirm these findings [14]. Several additional RCTs have been reported, substantially enlarging the randomized evidence base and enabling more granular analyses (e.g., leak grade, anastomotic side, severity) and formal information size appraisal with more secondary outcomes.

Accordingly, we conducted a pre-registered, PRISMA-compliant systematic review and meta-analysis of large-sample-size randomized controlled trials (RCTs) that directly compare ICG-FA with conventional intraoperative assessment in colorectal surgery. Our primary objective was to update and refine evidence beyond Meyer et al. [9] and Elmajdub et al. [14] by definitively evaluating the impact of ICG-FA on overall AL rates across all available RCTs. More secondary outcomes, including leak grades, conversion rates, safety profile, overall complication rate and complications according to the Clavien-Dindo classification, reoperation rates, surgical site infections, ileus, mortality, and other perioperative metrics, were also rigorously assessed. To effectively counter the risks of random errors and repeated significance testing inherent in a limited RCT evidence base, we used trial sequential analysis (TSA) and certainty of evidence using grade assessment and conducted comprehensive analyses of prespecified subgroups, such as left- versus right-sided anastomoses and leak severity.

Methods

Study protocol and registration

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [15] and the Cochrane Handbook for Systematic Reviews and Meta-Analyses [16–18]. The study protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) under registration no. CRD420251139372.

The protocol was submitted on September 2, 2025, and approved on the same date. All prespecified objectives, eligibility criteria, and planned analyses outlined in the PROSPERO protocol were adhered to. The primary objective was to evaluate the effect of indocyanine green fluorescence angiography (ICG-FA) compared to conventional intraoperative assessment regarding the incidence of overall anastomotic leaks in colorectal surgery. Secondary objectives included evaluating the grade and location of leaks, overall postoperative complications, complications categorized by the Clavien-Dindo classification, rates of reoperation, occurrences of surgical site infections, instances of postoperative ileus, conversions to open surgery, operative time, intraoperative blood loss, length of hospital stay, and mortality rates.

Eligibility criteria

Studies were considered eligible if they enrolled adult patients undergoing colorectal resections with primary anastomosis. The intervention of interest was the intraoperative use of indocyanine green fluorescence angiography (ICG-FA) to assess bowel perfusion compared with conventional assessment of anastomotic viability. Conventional assessment primarily relied on the surgeon’s visual inspection and tactile judgment of bowel perfusion, including subjective indicators such as bowel color, bleeding from the bowel edge or marginal artery, mesenteric pulsation, and overall tissue viability. Some studies additionally incorporated adjunctive institutional methods, including air leak testing, underwater pneumatic testing, intraoperative endoscopy, sigmoidoscopy, and contrast evaluation, according to local practice. Only randomized controlled trials (RCTs) directly comparing these two approaches were included, while non-randomized studies, case series, case reports, and reviews were excluded. The primary outcome was overall anastomotic leak.

Secondary outcomes, including anastomotic leak by grade (A, B, and C), were also assessed in most of our included studies, in accordance with the International Study Group of Rectal Cancer (ISREC) classification [19], and included anastomotic leak by side (right vs left), overall complication rate and complications stratified according to the Clavien-Dindo classification, and reoperation, as well as surgical site infection, ileus, conversion to open surgery, operative time, intraoperative blood loss, length of hospital stay, and mortality. A 90-day assessment of overall complications and mortality rate was performed.

We also excluded trials conducted in children, studies with historical control groups, conference abstracts without complete data, and studies in languages other than English.

Information sources

A comprehensive search was conducted across MEDLINE, Embase, CENTRAL, Scopus, and Web of Science, covering the period from their inception to July 2, 2025. To ensure the review reflected the most current evidence, a secondary updated search was performed on July 20, 2025, which identified one newly published study that met the inclusion criteria. Additionally, the reference lists of the included studies and relevant reviews were meticulously hand-searched, along with completed trials listed on clinicaltrials.gov. This study exclusively includes English-language publications.

Search strategy

The detailed search strategies for each database, including controlled vocabulary and free-text terms, Boolean operators, and filters, are presented in Supplementary File 1; Table S1. Searches combined terms for “colorectal surgery,” “anastomotic leak,” and “indocyanine green fluorescence angiography.” Filters for human subjects and randomized controlled trials were applied. Duplicates were removed using EndNote 20, and records were screened in Rayyan. Titles and abstracts were reviewed for relevance by two independent reviewers, with a third reviewer resolving disagreements. Full texts of relevant articles were retrieved for screening against predefined inclusion and exclusion criteria.

Data extraction

Data were extracted using a structured Microsoft Excel form that was piloted before use. The extraction sheet was organized into three main sections. The first section captured key study characteristics, including study ID, country or region, total number of randomized patients, underlying pathology, operative approach, route of anastomosis, dose and timing of ICG administration, and duration of follow-up. The second section collected baseline demographic and anthropometric variables across all included trials, including patient distribution by treatment group, age, sex, body mass index, diabetes status, and smoking prevalence. The third section focused on prespecified efficacy and safety outcomes defined in the study protocol.

Two reviewers independently extracted data from each trial. Discrepancies between reviewers were resolved through discussion, and unresolved disagreements were adjudicated by a senior author. Where reported data were incomplete or unclear, a third reviewer reassessed the study before data were considered unavailable.

To ensure comparability, predefined hierarchies were applied when handling multiple outcome definitions or time points. When definitions of anastomotic leak varied, outcomes classified according to the ISREC grading system were preferentially extracted when available, while alternative definitions were mapped to the closest ISREC-equivalent grade where necessary [19].

To identify potential overlapping publications from the same randomized study, study characteristics were systematically compared across all screened records. Trial names or acronyms, registration numbers, authorship, institutions, recruitment periods, sample sizes, and methodological details were cross-checked. When two or more reports demonstrated highly similar characteristics suggestive of the same cohort, only the report containing the most complete and updated dataset—typically the final publication—was retained for quantitative synthesis. This approach ensured that each patient population contributed only once to the meta-analysis and that pooled estimates reflected independent clinical data.

The primary outcome was overall anastomotic leak. Perfusion-related secondary outcomes included leak grade, leak side, and reoperation related to anastomotic complications. Additional exploratory secondary outcomes included overall complications, Clavien-Dindo complications [20], surgical site infection, ileus, conversion to open surgery, operative time, blood loss, length of hospital stay, and mortality.

Risk of bias and certainty of evidence

Two reviewers assessed the risk of bias independently using the Cochrane Risk of Bias 2 (RoB 2) tool [21]. Five key domains were evaluated: outcome measurement, selection of reported results, deviations from intended interventions, missing outcome data, and bias related to the randomization process. Risk-of-bias summary figures were generated using the robvis tool [22]. Disagreements were resolved by a senior author.

Using the GRADE approach, two authors independently assessed the evidence's indirectness, imprecision, inconsistency, and publication bias. It was further classified into four categories based on its quality of evidence: very low, low, moderate, and high [23, 24]. Then, a summary table was created using the GRADE profiler (version 3.6.1).

Statistical analysis

All statistical analyses were conducted using Review Manager (RevMan), version 5.4. Continuous variables were analyzed by calculating the mean difference (MD) with a 95% confidence interval (CI), while dichotomous outcomes were analyzed using the risk ratio (RR) with a 95% CI, and the Mantel-Haenszel method was used. Risk ratios were selected a priori because they provide clinically interpretable effect estimates in randomized controlled trials. The I² statistic and the chi-square test were used to evaluate statistical inconsistency across the included studies. An I² value > 50% suggested considerable statistical inconsistency, prompting the use of a random-effects model; if it was lower, a fixed-effects model was used instead. Statistical inconsistency was evaluated in accordance with the Cochrane Handbook (Chapter 9) [18]. A chi-square p-value < 0.1 was regarded as statistically significant for assessing statistical inconsistency. Additional sensitivity analyses using risk difference (RD) models were performed for sparse-event outcomes and outcomes containing zero-event studies. A p-value < 0.05 was deemed statistically significant for all analyses. Type 1 and type 2 errors may arise from limited reported data and the cumulative aggregation of trials.

To mitigate this, we employed trial sequential analysis (TSA), TSA 0.9.5.10 (http://www.ctu.dk; Copenhagen Trial Unit, Denmark) to assess whether the evidence from the aggregated trials was definitive and trustworthy. A confidence level is deemed adequate when the z-line on the curve intersects both the conventional boundary and the sequence monitoring boundary, suggesting that additional studies are unnecessary. Conversely, if the z-line of a curve fails to cross any boundary, the evidence is deemed inadequate, and further investigation is required. We applied an alpha error of 0.05 and a beta error of 80% power for this meta-analysis. Throughout the meta-analysis, the risk ratio was computed to ascertain the sample size necessary for TSA. Detailed instructions for the TSA software are provided in the user manual.

Bibliometric and network analyses were performed using R software (version 4.4.2, via RStudio) with the Bibliometrix and Biblioshiny packages. Trial publication metadata were imported in RIS format. All network visualizations and thematic maps were generated interactively in Biblioshiny and exported for reporting.

Results

Study selection

The initial search identified 968 articles of potential relevance. After removing 299 duplicates, 669 unique records remained. Title and abstract screening excluded 435 articles for being unrelated to the research questions and because of the study designs or patient populations. A total of 234 full-text articles were assessed for eligibility, of which eight randomized controlled trials (RCTs) met the inclusion criteria and were synthesized in this review [25–32]. The study selection process is depicted in the PRISMA flow diagram (Fig. 1).

Fig. 1.

PRISMA flow diagram for study selection. The diagram outlines the systematic review process used to identify, screen, and include randomized controlled trials (RCTs) comparing ICG fluorescence angiography-guided and conventional methods. A total of 968 records were identified from five databases, and after screening and duplicate removal, 7 RCTs were included in the final review

Study characteristics and demographics

Across the 8 RCTs, 4712 patients were randomized: 2348 (49.84%) to indocyanine green (ICG) fluorescence angiography and 2364 (50.16%) to conventional perfusion assessment. The mean age ranged from 57 to 70 years, and 58.13% were male (Table 1).

Table 1.

Baseline demographic and anthropometric data at baseline across all included randomized controlled trials comparing ICG fluorescence angiography-guided and conventional methods, including age, BMI, sex, prevalence of diabetes, and smoking

Study	Study arms	Number in each group	Age (Mean ± SD) (Years)	BMI (Mean ± SD) (kg/m2)	Male N (%)	Diabetes N (%)	Smoking N (%)	Neoadjuvant radiotherapy
Nardi et al., 2019 [25]	ICG	118	62.25 ± 17.1	25.2	60(50.8)	17 (14.4)	N/A	28(23.7)
	Control	122	62.25 ± 17.1	25.6	66 (54)	19 (15.5)	N/A	21(17.21)
Alekseev et al., 2020 [26]	ICG	187	63 ± 10.8	N/A	92(49.1)	15 (8)	N/A	20(10.7)
	Control	190	63 ± 3.1	N/A	92(48.4)	17 (9)	N/A	19(10.0)
Jafari et al., 2021 [27]	ICG	178	57.2 ± 11.4	27.8 ± 5.6	109(61.2)	22 (12.4)	49 (27.5)	113(63.5)
	Control	169	57 ± 11.4	28.2 ± 5.9	99 (58.6)	23 (13.6)	31 (18.3)	111(65.7)
Watanabe et al., 2023 [28]	ICG	422	65 ± 12.64	22.6 ± 3.7	266 (63)	62 (14.7)	N/A	21 (5.0)
	Control	417	66.33 ± 11.9	22.9 ± 3.6	274(65.7)	72 (17.3)	N/A	30 (7.2)
Faber et al., 2024 [29]	ICG	463	67.6 ± 11.9	26.2 ± 4.2	249 (54)	60 (13)	49 (11)	10 (2)
	Control	468	67.5 ± 12.6	26.1 ± 4.1	236 (50)	62 (13)	59 (13)	14 (3)
Rinne et al., 2025 [31]	ICG	567	70 ± 11	27.6 ± 5	297(55.4)	N/A	55 (10.4)	NR
	Control	569	70 ± 11	27.6 ± 5	284(52.5)	N/A	63 (11.7)	NR
Jayne et al., 2025 [30]	ICG	383	62.9 ± 11.6	26.7 ± 4.7	253 (66)	N/A	45 (12)	NR
	Control	383	63.3 ± 11.5	26.5 ± 4.8	248 (65)	N/A	49 (13)	NR
Gach et al., 2023 [32]	ICG	41	64.7 ± 10.6	26.5 ± 4.7	26 (63.4)	N/A	N/A	29 (70.7)
	Control	35	64.8 ± 10.6	26.7 ± 4.2	21 (60.0)	N/A	N/A	30 (85.7)

The included trials spanned a broad geographical spectrum, reflecting both single-center expertise and multinational collaboration (Table 2). These comprised a Russian single-center trial (FLAG) [26], multicenter trials in Italy [25], Finland (ICG-COLORAL) [31], The Netherlands (AVOID) [29], and Japan (EssentiAL) [28], a North American multicenter trial (PILLAR III) [27], a preliminary trial in Poland [32], and a large pan-European collaboration (IntAct) [30]. Collectively, these RCTs provide a genuinely global evidence base, capturing surgical practices across Eastern and Southern Europe, Scandinavia, North America, and East Asia.

Table 2.

Summary of study region, design, sample size, pathology, operative approach, route, time, dose of ICG, and follow-up intervals for each included RCT comparing ICG fluorescence angiography-guided and conventional methods

Study	Country	Design	Total sample size	Pathology (benign or malignant)	Operative approach	Route of ICG administration	Time of ICG administration	Dose of ICG administration	Conventional/control assessment	Duration of follow-up
Nardi et al., 2019 [25]	Italy	RCT	240	ICG = malignant 86, benign 32 Control = malignant 85, benign 37	Laparoscopic and open	Intravenous (IV)	Before performing anastomosis	0.3 mg/kg	Visual/tactile assessment including bowel appearance, bleeding, and mesenteric pulsation	30 days
Alekseev et al., 2020 [26]	Russia	RCT	377	ICG = malignant 187, benign 4 Control = balignant 184, benign 6	Laparoscopic and open	Intravenous (IV)	Before performing anastomosis	0.2 mg/kg	Visual assessment of bowel color, pulsation, and marginal artery bleeding	30 days
Jafari et al., 2021 [27]	US	RCT	347	Malignant	Laparoscopic, open and robotic	Intravenous (IV)	Before performing anastomosis	2.5 mg/ml	Macroscopic assessment of bowel color and bleeding	56 days
Watanabe et al., 2023 [28]	Japan	RCT	839	Malignant	Robotic, laparoscopic and transanal endoscopic	Intravenous (IV)	Before performing anastomosis	12.5 mg	Assessment of bleeding, mesenteric pulsation, and bowel discoloration	30 days
Faber et al., 2024 [29]	The Netherlands	RCT	931	ICG = malignant 405, benign 58 Control = malignant 421, benign 47	Laparoscopic or robotic	Intravenous (IV)	Before bowel transection	5 mg	Surgeon-dependent visual assessment including tissue color and mesenteric pulsation	90 days
Rinne et al., 2025 [31]	Finland	RCT	1136	ICG = malignant 380, benign 156 Control = malignant 383, benign 158	Laparoscopic	Intravenous (IV)	Before and after anastomosis formation	5 mg	Routine clinical assessment using bowel color and bleeding from bowel ends	30 days
Jayne et al., 2025 [30]	Belgium, Germany, Ireland, Italy, The Netherlands, Slovenia, Sweden, and UK	RCT	766	Malignant	Laparoscopic or robotic	Intravenous (IV)	1st one: After mobilization of the colon and division of the rectum 2nd one: After anastomosis creation	0.1 mg/kg	White-light assessment using bowel color, pulsation, and bleeding	90 days
Gach et al., 2023 [32]	Poland	RCT	76	Malignant	Laparoscopic	Intravenous (IV)	1st one: After intestinal dissection 2nd one: After the anastomosis	50 mg	Visible-light macroscopic assessment of bowel perfusion	30 days

Sample sizes varied considerably, ranging from 76 patients in the preliminary trial [32] to 1136 in the Finnish ICG-COLORAL trial [31] (Table 2). Smaller studies provide exploratory signals, while larger multicenter trials yield more robust estimates with greater statistical precision.

The study populations were heterogeneous. Some trials enrolled patients undergoing both benign and malignant colorectal resections [25, 26], whereas others focused exclusively on rectal cancer cohorts [28, 30]. This variability enhances external validity but introduces challenges for direct comparability across trials, particularly regarding tumor location and baseline risk of anastomotic complications.

Protocol heterogeneity and geographic scope

The included trials differed substantially in their operative approaches, dosing regimens, and ICG administration protocols. Early single-center studies [25, 26] used higher weight-based doses (0.2–0.3 mg/kg) given once before anastomosis, whereas later large multicentric trials, such as AVOID [29] and IntAct [30], standardized fixed doses of 5 mg or 0.1 mg/kg with dual injections before and after anastomosis (Table 2).

Operative approach and pathology type were also diverse: most recent trials [27, 28, 30, 32] were fully laparoscopic or robotic and restricted to malignant disease, whereas earlier trials [25, 26, 29, 31] included open or mixed procedures and benign pathology (Table 2). This progression reflects the temporal evolution of ICG adoption from feasibility studies to oncological standardization. The follow-up periods ranged from 30 to 90 days, aligning with ISREC definitions of leak assessment and ensuring consistency across datasets.

Collectively, Table 2 shows that while the evidence base is geographically and methodologically heterogeneous, its internal consistency across continents and protocols reinforces the external validity of the pooled findings (below).

Risk of bias

All eight RCTs were assessed using the Cochrane Risk of Bias 2.0 tool and visualized with the robvis package (Fig. 2). Most trials were rated as having “some concerns,” primarily due to the lack of blinding of surgeons and outcome assessors, which is inherent in intraoperative interventions such as ICG. Randomization processes and allocation concealment were generally adequate, and attrition bias was minimal given the low dropout rates. Selective outcome reporting was not evident. Importantly, the two largest trials, ICG-COLORAL [31] and AVOID [29], demonstrated strong methodological rigor, thereby mitigating the impact of bias from smaller studies. Overall, the risk of bias was judged as low to moderate across the evidence base.

Fig. 2.

Risk-of-bias (RoB 2) summary and traffic-light plots for all included RCTs, generated with robvis. Domain-level judgments (randomization process; deviations from intended interventions; missing outcome data; measurement of the outcome; selection of the reported result) are displayed per study and in aggregate, showing predominantly low risk or some concerns, largely reflecting the practical inability to blind surgeons to fluorescence imaging in operative trials

Certainty of evidence (GRADE)

The certainty of evidence across outcomes was appraised using the GRADE framework. For the primary endpoint, anastomotic leakage, the certainty was rated as high, supported by consistent effect estimates, narrow confidence intervals, and adequate information size. Certainty was also rated highly for reoperation and mortality, both of which are classified as critical outcomes.

Among secondary outcomes, certainty was high to moderate for overall complications, conversion rates, and surgical site infections, all of which were rated as important for clinical decision-making. Certainty was moderate for intraoperative blood loss, an outcome deemed of limited importance, and low for length of hospital stay, reflecting imprecision and statistical inconsistency across trials.

Detailed GRADE approach judgments, including evidence profiles for each outcome, can be found in Supplementary File 2 Table S2.

Publication bias

Publication bias was evaluated through funnel plot asymmetry and sensitivity analyses. Visual inspection of funnel plots did not reveal significant asymmetry, although the limited number of included RCTs restricts the interpretability of such assessments. Trial sequential analysis (TSA) was also performed to account for the risk of random errors and repeated significance testing. For anastomotic leakage, the cumulative Z-curve crossed both the conventional and trial sequential monitoring boundaries, suggesting that the current evidence base is sufficient to support the benefit of ICG for overall leaks. By contrast, outcomes such as reoperation, overall complications, and length of stay remain underpowered in cumulative analysis, indicating a potential risk of false positives and the need for further large-scale trials.

Operative time

Six RCTs reported data on operative duration [25, 26, 28, 29, 31, 32]. Pooled analysis using a random-effects model demonstrated no significant difference in mean operative time (OT) between the ICG and control groups (MD = 0.80 min, 95% CI –6.22 to 7.83; p = 0.82) (Fig. 3A). However, substantial statistical inconsistency was observed (I² = 65%, p = 0.01).

Fig. 3.

Perioperative outcomes and intraoperative decision changes with ICG fluorescence guidance versus conventional assessment. A Operative time (min) (6 studies; n = 3540; random effects, inverse variance). B Intraoperative blood loss (ml) (3 studies; n = 2293; fixed-effect, inverse variance). C Length of hospital stay (days) (7 studies; n = 4287; random effects, inverse variance). D Change in surgical plan (CISP) (3 RCTs; n = 1881; fixed effect, Mantel-Haenszel). Squares show study-specific estimates weighted by inverse variance; diamonds show pooled effects; whiskers are 95% CIs. For panels A–C, negative mean differences favor ICG; for panel D, RR > 1 indicates more CISP with ICG

To investigate the source of statistical inconsistency, a sensitivity analysis was performed. Excluding the trial by Alekseev et al. (2020) [26], which reported disproportionately prolonged OT and was identified through leave-one-out analysis as the primary contributor to inconsistency, reduced statistical inconsistency (I² = 46%, p = 0.12). Despite this adjustment, the pooled effect remained non-significant (MD = 3.70 min, 95% CI –2.53 to 9.92; p = 0.38) (Supplementary File 3 Fig. S3.1).

Taken together, these findings indicate that ICG fluorescence angiography does not significantly prolong OT compared with conventional techniques. The conclusion remained consistent after sensitivity testing, supporting the robustness of this outcome despite inter-study variability.

Blood loss

Three RCTs, comprising 2293 patients (ICG group: n = 1145; control group: n = 1148), reported intraoperative blood loss as a prespecified outcome [28, 29, 31]. Pooled analysis demonstrated a statistically significant reduction in blood loss with ICG guidance compared with conventional assessment (MD = – 4.77 ml, 95% CI –8.97 to –0.56; p = 0.03). Importantly, there was no evidence of inter-study heterogeneity (I² = 0%, p = 0.77) (Fig. 3B).

Although the absolute reduction is numerically modest, the consistency across trials strengthens confidence in the finding. These results suggest that ICG use may improve intraoperative hemostasis, even if the effect size is unlikely to be of major clinical consequence in standard-risk populations.

Length of hospital stay

Seven RCTs [25, 26, 28–32] reported postoperative length of hospital stay (LOS). Pooled analysis revealed no significant difference between ICG and control groups (MD = 0.12 days, 95% CI –0.52 to 0.27; p = 0.54), with moderate between-study inconsistency observed (I² = 48%, p = 0.08) (Fig. 3C).

To explore statistical inconsistency, a sensitivity analysis was conducted. Excluding Faber et al. (2024), which reported comparatively longer and more variable lengths of stay reflective of Japanese perioperative practice, reduced statistical inconsistency to zero (I² = 0%). Under this model, the pooled estimate did not reach statistical significance, though the direction of effect favored ICG (MD = –0.30 days, 95% CI –0.65 to 0.06; p = 0.10) (Supplementary File 3 Fig. S3.2).

These findings suggest that, while ICG use does not consistently shorten hospitalization across all healthcare systems, modest reductions may be observed in settings with standardized early recovery protocols. The variation in discharge practices between regions (e.g., Japan vs. Europe/Scandinavia) likely explains the statistical inconsistency observed in the primary analysis.

Change in surgical plan (CISP)

Three RCTs (n = 1881) reported intraoperative change in the surgical plan (CISP) [30–32]. Pooled analysis showed a higher likelihood of CISP with ICG versus control (RR 1.77, 95% CI 1.27–2.48; p = 0.0008; I² = 0%). Across trials, CISP typically comprised a more proximal transection and/or diversion prompted by perfusion concerns. This aligns with the hypothesized pathway by which ICG may reduce low-grade leaks; however, definitions varied, and the estimate was sensitive to small study effects. Sensitivity analyses (continuity correction; exclusion of zero-event trials) showed similar directionality (Fig. 3D). Sensitivity analysis using a risk difference (RD) model demonstrated findings consistent with the primary analysis, confirming a significantly higher likelihood of intraoperative change in surgical plan with ICG guidance (RD = 0.04, 95% CI 0.02–0.06; p = 0.0008), with no statistical inconsistency observed (I² = 0%) (Supplementary File 3 Fig. S3.3).

Overall anastomotic leak

All seven RCTs reported anastomotic leak as a primary or secondary endpoint, encompassing a total of 4636 patients. The pooled analysis demonstrated a significant reduction in overall anastomotic leak with ICG compared with conventional assessment (RR = 0.69, 95% CI 0.58–0.82; p < 0.0001), and no important statistical inconsistency was observed (I² = 0%) (Fig. 4C). TSA supported the robustness and sufficiency of the evidence (Fig. 5A). Sensitivity analysis using a risk difference (RD) model demonstrated findings consistent with the primary RR analysis, with ICG remaining associated with a significant reduction in overall anastomotic leak risk (Supplementary File 3 Fig. S3.4).

Fig. 4.

Forest plots for anastomotic leak (AL) outcomes structured by clinically meaningful strata. Subpanels display A AL by anatomic location (right- vs left-sided resections), B AL severity by ISREC grades A–C, and C overall AL incidence. Squares are study weights (inverse variance); diamonds are pooled effects; whiskers are 95% CIs. RR < 1 favors ICG. Subgroup panels highlight greater relative reduction for left-sided and low-grade leaks

Fig. 5.

Trial sequential analysis (TSA) of anastomotic leak (AL) endpoints. Cumulative Z-curves are plotted for overall AL and for left-sided AL against conventional significance and trial-sequential monitoring boundaries, with the required information size indicated. Crossing the superiority boundary signals that the accrued randomized evidence is sufficient to support a leak-reducing effect of ICG without an undue risk of random-error-driven false positivity

Leak by location

Two RCTs [29, 31] reported anastomotic leak rates stratified by location (right- vs. left-sided resections), including a combined total of 1968 patients. For right-sided anastomoses, the pooled analysis of 505 patients in the ICG group and 460 in the control group demonstrated no significant difference in leak rates (RR = 0.84, 95% CI 0.50–1.43; p = 0.52), with no statistical inconsistency observed (I² = 0%). For left-sided anastomoses, the pooled dataset included 509 patients in the ICG arm and 494 in the control arm. Here, ICG use was associated with a statistically significant reduction in leak risk (RR = 0.56, 95% CI 0.37–0.85; p = 0.006), again with no statistical inconsistency observed (I² = 0%). When pooling both anatomical locations, the analysis indicated an overall reduction in leaks with ICG compared with conventional assessment (RR = 0.65, 95% CI 0.47–0.90; p = 0.01), with no important statistical inconsistency observed (I² = 0%) (Fig. 4A). TSA supported the robustness and sufficiency of the evidence for left-sided anastomoses (Fig. 5B).

Leak by severity

Three RCTs [28, 30, 31] reported anastomotic leak rates stratified by ISREC severity grading (Grades A–C). Subgroup meta-analysis demonstrated that ICG use was associated with a significant reduction in Grade A leaks (RR = 0.55, 95% CI 0.35–0.85; p = 0.007; I² = 0%). For Grade B leaks, there was a non-significant trend favoring ICG (RR = 0.57, 95% CI 0.31–1.07; p = 0.08; I² = 29%). In contrast, Grade C leaks, which are most strongly linked to technical or non-perfusion-related factors, showed no difference between groups (RR = 1.04, 95% CI 0.65–1.66; p = 0.89; I² = 0%).

When pooling all grades, ICG use was associated with a significant overall reduction in leaks (RR = 0.66, 95% CI 0.49–0.91; p = 0.01; I² = 19%) (Fig. 4B). An exploratory analysis restricted to Grades A and B only revealed a stronger protective effect (RR = 0.54, 95% CI 0.39–0.75; p < 0.0003) and eliminated statistical inconsistency (I² = ^{0%) (}Supplementary File 3 Fig. S3.5).

Overall complication rate

All eight included RCTs reported overall postoperative complications. Pooled analysis demonstrated a significant reduction in complications with ICG guidance compared with conventional assessment (RR = 0.90, 95% CI 0.83–0.98; p = 0.02), with no evidence of inconsistency (I² = 0%) (Fig. 6A). TSA did not demonstrate sufficient evidence to confirm the robustness of this outcome (Supplementary File 3 Fig. S3.6).

Fig. 6.

Forest plots for major postoperative morbidity. Panels summarize pooled risk ratios (95% CIs) for A overall complication rate, B complications stratified by Clavien-Dindo grade, C reoperation, and D conversion to open surgery. Model choice and statistical inconsistency are shown within each panel. Squares = study weights; diamonds = pooled effects; whiskers = 95% CIs. Direction of effect generally favors ICG for complications and conversions while acknowledging variability across trials that tempers certainty for some secondary outcomes. Reoperation signals are null

In a risk-of-bias sensitivity analysis excluding Jayne et al., the pooled estimate for overall complications was attenuated and no longer reached statistical significance (RR = 0.92, 95% CI 0.83–1.02; p = 0.10), although the direction and magnitude of effect remained largely consistent with the primary analysis. This finding aligns with the TSA, which suggested that evidence for this outcome remains inconclusive (Supplementary File 3 Fig. S3.7).

Complications according to Clavien-Dindo classification

Six RCTs reported postoperative complications stratified by Clavien-Dindo grade [25, 28–32]. Subgroup meta-analysis demonstrated a significant reduction in Grade I complications with ICG use (RR = 0.75, 95% CI 0.59–0.96; p = 0.02; I² = 0%). No significant differences were observed for Grade II (RR = 1.00, 95% CI 0.84–1.19; p = 1.00), Grade III (RR = 0.90, 95% CI 0.73–1.10; p = 0.29), Grade IV (RR = 1.18, 95% CI 0.54–2.59; p = 0.68), or Grade V complications (RR = 0.54, 95% CI 0.17–1.73; p = 0.29) (Fig. 6B). When all grades were pooled, there was a non-significant trend toward fewer complications with ICG (RR = 0.90, 95% CI 0.80–1.01; p = 0.08; I² = 0%).

Sensitivity analysis using a risk difference (RD) model demonstrated consistent findings. ICG use remained associated with a significant reduction in Grade I complications (RD =  − 0.02, 95% CI − 0.04 to − 0.01; p = 0.009; I² = 0%), while no significant differences were observed for Grade II (RD = 0.00, 95% CI − 0.02 to 0.02; p = 0.72; I² = 6%), Grade III (RD =  − 0.01, 95% CI − 0.03 to 0.01; p = 0.39; I² = 0%), Grade IV (RD = 0.00, 95% CI − 0.00 to 0.01; p = 0.46; I² = 0%), or Grade V complications (RD =  − 0.00, 95% CI − 0.01 to 0.01; p = 0.61; I² = 48%). When all grades were pooled, there was no significant difference in overall postoperative complications between groups (RD =  − 0.00, 95% CI − 0.01 to 0.00; p = 0.26; I² = 34%) (Supplementary File 3, Fig. S3.5).

Reoperation rate

All eight RCTs reported reoperation as a postoperative outcome. Pooled analysis using a random-effects model showed no significant difference between the two groups (RR = 0.97, 95% CI 0.72–1.31; p = 0.85), with moderate but non-significant statistical inconsistency observed (I² = 44%, p = 0.10) (Fig. 6C).

To assess robustness, a leave-one-out sensitivity analysis was performed. Excluding Watanabe et al. (2023, EssentiAL) [28], which contributed disproportionately to between-study variability, reduced statistical inconsistency to I² = 22% (p = 0.27). However, the pooled effect remained unchanged and non-significant (RR = 1.01, 95% CI 0.80–1.27; p = 0.92) (Supplementary File 3 Fig. S3.4).

Conversion to open surgery

Four RCTs [25, 28, 29, 31] reported conversion rates from laparoscopic to open surgery, encompassing a total of 3194 patients (ICG: n = 1599; control: n = 1595).

Pooled analysis using a random-effects model demonstrated no significant difference in conversion rates between ICG-guided and conventional approaches (RR = 1.15, 95% CI 0.84–1.57; p = 0.40). No important statistical inconsistency was observed (I² = 0%, p = 0.71) (Fig. 6D).

Surgical site infection

Three RCTs [25–27] reported surgical site infection (SSI) as a postoperative outcome, including a total of 964 patients (ICG: n = 483; control: n = 481).

Pooled analysis using a random-effects model showed no significant difference in SSI risk between the ICG and control groups (RR = 0.73, 95% CI 0.19–2.86; p = 0.66). Statistical inconsistency was low to moderate (I² = 30%, p = 0.24) (Fig. 7A). Sensitivity analysis using a risk difference (RD) model demonstrated consistent findings for surgical site infection (SSI), with no significant difference observed between the ICG and conventional groups (RD =  − 0.00, 95% CI − 0.03 to 0.02; p = 0.72; I² = 60%) (Supplementary File 3, Fig. S3.10).

Fig. 7.

Forest plots for additional postoperative outcomes: A surgical site infection (SSI), B postoperative ileus, and C mortality. Each analysis presents pooled risk ratios with 95% CIs, I², and the corresponding meta-analytic model. Mortality and ileus show no statistically significant differences between groups, and SSI trends toward reduction with ICG but remains underpowered, together indicating that fluorescence guidance does not introduce new safety risks

Postoperative Ileus

Three RCTs [25–27] reported the incidence of postoperative ileus, including a total of 964 patients. Pooled analysis with a random-effects model showed no significant association between ICG use and postoperative ileus (RR = 1.07, 95% CI 0.36–3.23; p = 0.90), with moderate statistical inconsistency observed (I² = 53%) (Fig. 7B).

To investigate this statistical inconsistency, a leave-one-out sensitivity analysis was performed. Exclusion of Jafari et al. (2021) [27], which reported outlier results and contributed disproportionately to variability, reduced statistical inconsistency to zero (I² = 0%) while leaving the effect estimate non-significant (RR = 1.78, 95% CI 0.65–4.86; p = 0.26) (Supplementary File 3 Fig. S3.5).

Mortality Rate

Four RCTs [28–31] reported postoperative mortality, including a total of 3119 patients (ICG: n = 1561; control: n = 1558). The pooled analysis showed no significant difference in mortality between the ICG and control groups (RR = 1.00, 95% CI 0.47–2.15; p = 0.99). No statistical inconsistency was observed (I² = 0%, p = 0.76) (Fig. 7C). Sensitivity analysis using a risk difference (RD) model demonstrated consistent findings for postoperative mortality, with no significant difference observed between the ICG and conventional groups (RD =  − 0.00, 95% CI − 0.00 to 0.00; p = 0.85; I² = 0%) (Supplementary File 3, Fig. S3.12).

Bibliometric analysis

A bibliometric evaluation of the included trials revealed a progressive expansion of randomized research on ICG fluorescence angiography over the past decade, with publication rates accelerating sharply after 2020 (Fig. 8A). Most RCTs were published in high-impact surgical journals such as The Lancet Gastroenterology & Hepatology, JAMA Surgery, and Surgical Endoscopy (Fig. 8B), indicating that ICG-FA has transitioned from exploratory innovation to an established research domain.

Fig. 8.

Bibliometric panels summarizing the global evidence landscape: A annual scientific production, B the most relevant sources, and C the authorship network. Panels illustrate the post-2020 acceleration in RCTs and the consolidation of international consortia, supporting adequately powered meta-analyses

Authorship and institutional network mapping (Fig. 8C) demonstrated a clear evolution from early single-nation investigations, primarily in Italy and Russia, to multicenter consortia across Europe, Asia, and North America (AVOID, ICG-COLORAL, IntAct, EssentiAL). Shared authorship across trials, for example, Andrea Vignali (Italy) and Roel Hompes (The Netherlands), illustrates intellectual cross-pollination between cohorts, though all trials maintained independent recruitment and datasets. This pattern parallels the methodological refinement observed in Table 2, where later trials standardized ICG dosing and imaging protocols.

Collectively, the bibliometric trends corroborate the quantitative results: the global research community has converged on standardized, multicenter RCTs that now provide sufficient information size and methodological rigor for definitive meta-analytic conclusions.

Comparison with prior meta-analyses

As shown in Table 3, earlier RCT-only meta-analyses by Meyer et al. [9], Elmajdub et al.[14], and Afridi et al. [33] included between three and nine trials, enrolling up to 4047 patients, and collectively suggested a moderate reduction in anastomotic leak with ICG-FA (Table 3). However, these studies lacked trial sequential analysis or formal GRADE appraisal, leaving the conclusiveness of evidence uncertain.

Table 3.

Summary of other studies’ data compared with our data

Arms	Meyer et al., 2022 [9]	Elmajdub et al., 2024 [14]	Afridi et al., 2025 [33]	Our study
Number of studies included	3	5	9	8
Number of patients	964	1369	4047	4,712
Type of studies included	RCTs	RCTs	RCTs	RCTs
Overall anastomotic leak	FA was significantly protective against anastomotic leak	The use of ICG significantly reduced the risk of anastomotic leaks compared with the non-ICG group, making the results statistically significant	The use of ICG significantly reduced the risk of anastomotic leaks compared with the non-ICG group, making the results statistically significant	The use of ICG significantly reduced the risk of anastomotic leaks compared with the non-ICG group, making the results statistically significant
Blood loss	NA	The results were not significant	The results were not significant	Statistically significant reduction in blood loss with ICG guidance compared with conventional assessment
Operation time (min)	NA	NA	The ICG group had a significantly longer operating time compared with the non-ICG group	The results were not significant
Length of hospital stay (days)	NA	The results were not significant	The results were not significant	The results were not significant
Reoperation rate	NA	The results were not significant	NA	The results were not significant
Surgical site infection	NA	The results were not significant	The results were not significant	The results were not significant
Conversion to open surgery	NA	NA	NA	The results were not significant
Supgrouping according to the side of the anastomosis	NA	Anastomotic leakage occurrence outcome was divided into three grades based on severity	Anastomotic leakage occurrence outcome was divided into three grades based on severity, and the Clavien-Dindo scale was used to categorize complications into four grades	Anastomotic leakage occurrence outcome was divided into three grades based on severity and into two groups based on side, and the Clavien-Dindo scale was utilized to categorize complications into five grades
Results	One hundred eleven articles were screened, 108 were excluded, and 3 were kept for inclusion. The three included RCTs compared the assessments of bowel perfusion during colorectal anastomosis creation using FA versus standard practice. In meta-analysis, FA was significantly protective against AL (3 RCTs, 964 patients, RR: 0.67, 95% CI: 0.46 to 0.99, I 2: 0%, p = 0.04). The RD of AL was non-significantly decreased by 4 percentage points (95% CI: − 0.08 to 0, I2: 8%, p = 0.06) when using FA. Certainty of evidence was considered moderate	The ICG-FA was associated with a 45% reduction in the risk of overall anastomotic leaks (OR: 0.550, p = 0.012). Subgroup analysis showed a 47% reduction in leaks for low anastomoses (OR: 0.53, p = 0.143) and a 69% reduction in grade A leaks (OR: 0.31, p = 0.008). No significant effects were observed for blood loss, surgery duration, hospital stay, mortality, postoperative ileus, reoperation, or surgical site infections	Overall, anastomotic leak risk was considerably decreased with ICG use (risk ratio [RR] = 0.66; 95% CI: 0.54–0.81; p < 0.0001) and showed no heterogeneity (I2 = 0%). There was no heterogeneity (I2 = 0%) in the Grade A leakage occurrence, which was considerably lower in the ICG group (RR = 0.34; 95% CI: 0.16–0.72; p = 0.005)	ICG-FA reduced overall AL versus control (risk ratio [RR] 0.69, 95% CI 0.58–0.82; I2 = 0%). The absolute risk reduction was 3.6%, yielding a number needed to treat of around 28 to prevent one leak. TSA crossed superiority boundaries for the primary endpoint, suggesting evidence sufficiency at the assumed control risk and effect size. Benefits were more apparent for left-sided anastomoses and for lower-grade leaks (A and A + B), whereas severe (Grade C) leaks and reoperation did not differ significantly. Overall complications showed a small relative reduction, but TSA indicated insufficient information size. Other safety outcomes (surgical site infection, ileus, mortality) were similar between groups
Conclusion	The effect of FA on the prevention of AL in colorectal surgery exists but is potentially of small magnitude. Considering the potential magnitude of effect of FA, we advise that future RCTs have an adequate sample size, include a cost-benefit analysis of the technique, and better define the subpopulation that could benefit from FA	ICG-FA significantly reduces the overall risk of anastomotic leaks, especially grade A leaks, and shows a trend toward fewer leaks in low anastomosis. No significant impact was found on secondary outcomes. Further RCTs are needed to confirm these findings	Although indocyanine green fluorescence imaging may marginally lengthen the recovery period and duration of operation, it dramatically lowers anastomotic leakage and wound infections following colorectal surgery, enhancing results	Across contemporary RCTs, ICG-FA reduces overall anastomotic leaks, particularly in left-sided resections, without clear effects on severe leaks or reoperation. These findings support selective adoption with standardized protocols for dose, timing, and interpretation, alongside further trials powered for patient-important outcomes

AL: anastomotic leak, CI: confidence interval, FA: fluorescence angiography, ICG: indocyanine green, ICG-FA: indocyanine green fluorescence angiography, NA: not applicable/not available, OR: odds ratio, RCTs: randomized controlled trials, RD: risk difference, RR: risk ratio, SD: standard deviation, TSA: trial sequential analysis

By integrating eight contemporary RCTs (n = 4712), including the large multicentric EssentiAL Trial [28], AVOID [29], ICG-COLORAL [31], and IntAct [30] trials, the present analysis achieved both statistical sufficiency and high-certainty evidence for a 31–32% relative risk reduction in overall leaks (I² = 0%).

Discussion

Our systematic review and meta-analysis of eight randomized controlled trials, icluding over 4700 patients from Europe, Asia, and North America, provides high-certainty evidence that indocyanine green fluorescence angiography (ICG-FA) reduces the overall incidence of anastomotic leak in colorectal surgery. The pooled relative risk of 0.69 (95% CI 0.58–0.82; p < 0.0001; I² = 0%) corresponds to an absolute risk reduction of 3.6% and a number needed to treat (NNT) of 24. These results were consistent across diverse healthcare systems and surgical approaches and were corroborated by trial sequential analysis, confirming statistical sufficiency and minimizing the likelihood of type I error (Fig. 6) (Supplementary Fig. S3.3).

Beyond confirming efficacy, this analysis clarifies how fluorescence imaging influences intraoperative judgment. Three RCTs reported a higher frequency of intraoperative change in surgical plan (CISP), typically proximal resection revision or diversion prompted by perfusion concerns, under ICG guidance. This supports the concept that ICG enhances real-time decision-making rather than prolonging operative time or hospital stay. Because CISP definitions varied across trials, it was treated as a composite process metric, and its consistency in sensitivity analyses reinforces its value as a mechanistic correlation rather than a definitive mediator. In practice, CISP can be regarded as a “safety valve,” reflecting adaptive surgical modification based on perfusion findings rather than procedural inefficiency.

Interpretation within mechanistic context

The protective effect of ICG-FA is anatomically and pathophysiologically selective rather than universal. Its benefit was concentrated in left-sided anastomoses (RR = 0.56, 95% CI 0.37–0.85; p = 0.006), where perfusion is inherently limited because of the terminal nature of the inferior mesenteric arterial supply and variable collateralization of the marginal artery. In contrast, no significant reduction was observed for right-sided anastomoses (RR = 0.84, p = 0.52). These findings, supported by TSA sufficiency for the left-sided subgroup, align with the concept that ICG primarily mitigates perfusion-related risk rather than mechanical or technical failure. Comparable location-stratified analyses in major RCTs, including those by Faber et al. [29], Rinne et al. [31], and Alekseev et al. [26], reinforce this interpretation.

The core mechanism through which ICG-FA achieves this protective effect is by facilitating a Change in Surgical Plan (CIP), primarily through the identification of inadequate perfusion at the initially planned resection margin. Evidence from multiple RCTs confirms this necessary intervention. The CIP rate ranged across trials, reaching almost 20% of patients in the FLAG trial (36/187) [26], leading to an extra resection of the proximal bowel up to 5 cm. In the IntAct trial, a change in planned anastomosis occurred in 12% (47/377) of the ICG group, significantly higher than in the standard care group (7%, adjusted OR 2.152; p = 0.0057) [30]. The AVOID trial similarly reported that the surgical plan was altered in 9% (41/463) of patients based on the fluorescence signal [29], with these adjustments primarily directed at the proximal colon transection level and resulting in a median proximal shift of 2.0 cm. The frequency of these adjustments validates the technology's utility in identifying perfusion deficits missed by subjective assessment.

Prior non-randomized studies and early meta-analyses of observational data consistently suggested this protective effect, indicating that ICG-FA was associated with reduced leak rates. For instance, published data from an observational study on low anterior resection for rectal cancer reported that the AL rates of Clavien-Dindo (CD) grades ≥ II and ≥ III were 10.4% (22/211) and 9.5% (20/211) in the non-ICG-FI group and 4.7% (10/211) and 2.8% (6/211) in the ICG-FI group [34–36]. Earlier systematic reviews that included non-randomized evidence also concluded that the surgical plan was changed in 10.8% of cases after ICG-FA application, and the AL rate was significantly reduced (e.g., from 7.4% to 3.4%) [12].

Furthermore, ICG-FA’s efficacy is concentrated in preventing low-grade morbidity, showing a significant reduction in Grade A leaks (RR = 0.55, p = 0.007) and a combined Grade A and B reduction (RR = 0.54, p < 0.0003). This finding is biologically plausible: ICG-FA identifies subtle, potentially correctable micro-perfusion deficits, which are the most likely contributors to minor, low-grade anastomotic complications.

At the opposite end of the spectrum, ICG-FA did not influence severe or non-perfusion-driven outcomes. Rates of Grade C leaks (RR = 1.04, p = 0.89), reoperation (RR = 0.97, p = 0.85), and mortality (RR = 1.00, p = 0.99) were unchanged. These null findings underscore that ICG optimizes tissue viability but cannot offset failures arising from mechanical tension, stapler misfiring, or systemic factors. The PILLAR III trial—focused on high-risk low anterior resections with heavy neoadjuvant exposure—illustrated this ceiling effect: despite rigorous fluorescence assessment, overall leak rates were similar (9.0% vs 9.6%) [27]. The AVOID trial further contextualized this by implicating local biological factors, such as collagen degradation by Enterococcus faecalis, as contributors to non-perfusion leaks [29]. Collectively, these results delineate the mechanistic boundary of ICG’s benefit: it enhances perfusion-related decision-making and prevents minor ischemic failures but does not eliminate multifactorial catastrophic leaks.

Perioperative outcomes

ICG integration did not prolong operative time (MD = 0.80 min; p = 0.82), confirming that fluorescence imaging can be incorporated into the routine workflow without procedural delay. This time-neutral finding, consistent across sensitivity analyses, reflects improved familiarity and standardization in recent trials. A small but consistent reduction in intraoperative blood loss (MD = –4.77 ml; p = 0.03; I² = 0%) likely represents the avoidance of unnecessary dissection once perfusion is confirmed rather than a clinically meaningful hemostatic effect.

Length of stay (LOS) (MD = 0.12 days; p = 0.54; I² = 48%) was unaffected. Variability across trials probably reflects differences in perioperative pathways—particularly the inclusion of cohorts such as the Japanese EssentiAL trial, where Enhanced Recovery After Surgery (ERAS) protocols are less intensive than in most European centers [37–39]. These exploratory perioperative outcomes should be interpreted cautiously, as they are not direct mechanistic consequences of fluorescence-guided perfusion assessment.

Comparison with the existing literature

These findings refine and extend prior meta-analyses that were constrained by smaller datasets and limited statistical power. The early SRMA by Meyer et al., which included only three RCTs (n = 964), showed a modest leak reduction (RR 0.67) with moderate certainty [9]. Our pooled estimate (RR 0.68–0.69) confirms this effect with high certainty and trial sequential sufficiency, establishing the evidence base as statistically complete. Similarly, Elmajdub et al. analyzed five RCTs and reported a 45% reduction in overall leaks and a 69% reduction in Grade A leaks [14]. Our expanded dataset corroborates this pattern, showing that ICG’s protective effect is concentrated in minor, perfusion-related leaks (RR 0.55 for Grade A).

Unlike Arfidi et al., who reported an 8.26-min prolongation of operative time [33], our larger, standardized dataset demonstrates time neutrality (MD 0.80 min), indicating that with refined protocols and increasing familiarity, ICG integration no longer incurs a workflow penalty.

Collectively, these updated results align with international consensus. The European Association for Endoscopic Surgery (EAES) recommends ICG-FA to lower the leak risk [40], while the World Society of Emergency Surgery (WSES) endorses its selective use when perfusion assessment is critical [41]. Together, these endorsements and high-certainty pooled data position ICG-FA as a validated, evidence-based adjunct in contemporary colorectal surgery.

Bibliometric evaluation

The evolution of ICG-FA research reflects a rapid maturation of evidence quality after 2020, marked by an acceleration of randomized trials and the establishment of multinational consortia. Early single-center efforts, such as the FLAG trial in Russia [26] and the Italian study by De Nardi et al. [25], paved the way for large multicenter RCTs including IntAct [30], AVOID [29], ICG-COLORAL [31], and EssentiAL [28]. This shift standardized fluorescence protocols, transitioning from variable weight-based dosing to fixed 5-mg (0.1-mg/kg) regimens and dual-injection techniques (before and after anastomosis) and enabled cross-regional reproducibility. The resulting methodological convergence and high internal consistency (I² = 0% for overall AL) allowed the present analysis to achieve trial-sequential sufficiency and deliver a definitive, high-certainty conclusion regarding ICG-FA’s efficacy.

Strengths and limitations

This meta-analysis synthesizes the largest and most geographically diverse randomized evidence base to date, encompassing 8 high-quality RCTs and more than 4700 patients from Europe, Asia, and North America. Inclusion of recent multicenter trials, EssentiAL, IntAct, AVOID, and ICG-COLORAL [28–31], enhanced both statistical power and external validity. Exclusive reliance on randomized evidence minimized the confounding inherent in observational studies, while bias, certainty, and information sufficiency were systematically evaluated using Cochrane RoB 2, GRADE, and trial sequential analysis, respectively. The achievement of TSA superiority boundaries for the primary outcome establishes the evidence as statistically conclusive.

Subgroup analyses clarified the clinical domains of benefit—left-sided resections and low-grade leaks—while demonstrating negligible statistical inconsistency, reinforcing the robustness and interpretability of these findings. Methodological evolution across later trials, particularly the adoption of standardized fixed dosing and dual-injection protocols, further supports internal consistency.

Nonetheless, several limitations must temper interpretation. Trial-level statistical inconsistency persisted in inclusion criteria, leak definitions, and detection methods—ranging from mandatory imaging to clinical suspicion alone—introducing potential information bias. Lack of surgeon blinding represents an unavoidable performance bias in fluorescence-guided trials, compounded by the subjective interpretation of perfusion signals and a recognized learning curve [25, 27, 30, 31]. Some studies were underpowered or prematurely terminated, such as PILLAR III and De Nardi et al. [25, 27], while others included large numbers of right-sided resections, which may have diluted the observed effect [29, 30]. Neoadjuvant radiotherapy or chemoradiotherapy exposure was also inconsistently reported across trials, precluding reliable subgroup analyses according to radiation status and leaving potential residual confounding related to impaired anastomotic healing. Some continuous perioperative outcomes demonstrated substantial dispersion and possible right-skewness, which may limit the interpretability of mean-based pooled estimates.

Finally, the change-in-surgical-plan (CISP) outcome was explored post hoc and varied in definition across trials; it should therefore be regarded as supportive but hypothesis-generating evidence of mechanism. Long-term outcomes, such as recurrence, function, or quality of life, remain insufficiently reported.

Future directions

Future research should focus on objectifying and standardizing fluorescence interpretation. Current reliance on subjective visual assessment limits reproducibility. Quantitative fluorescence metrics, such as time-to-peak intensity, slope, and Fmax, are being validated to create measurable perfusion thresholds [42]. Minimal dose strategies also matter for repetitive angiography (resection side, then anastomosis): prospective evidence supports 0.02–0.05 mg/kg to avoid signal saturation while permitting re-dosing [43]. Emerging studies combining ICG with AI-enhanced imaging or machine learning models (e.g., to create a patient-calibrated transection recommender) show promise for establishing objective, consistent intraoperative decision-making thresholds and overcoming subjective interpretation [44, 45].

Targeted evaluation in high-risk populations is also warranted. Subgroup analyses from IntAct [30], EssentiAL [28], and ICG-COLORAL [31] suggest the protective effect of ICG is strongest in left-sided anastomoses and low-grade leaks but remains uncertain in technically complex, biologically compromised, or ultra-low anastomoses. The lack of benefit in right-sided resections (RR = 0.84) suggests routine ICG use is not warranted in these lower-risk populations. Conversely, definitive trials are still required in ultra-low rectal resections (e.g., < 5 cm from the anal verge) and in patients receiving heavy neoadjuvant chemoradiotherapy, where the risk of AL is driven by fibrosis and inflammation, likely overriding perfusion benefit. Dedicated trials in these specific high-risk groups are needed to refine indications and identify populations with the highest absolute benefit. Based on current evidence, integrating ICG NIR fluorescence imaging into surgical guidelines for patients undergoing rectosigmoid resections for cancer is already advisable [29].

Finally, forthcoming studies should extend beyond short-term safety to encompass cost-effectiveness and patient-centered outcomes. Early data from the IntAct trial suggest that ICG may achieve cost neutrality or modest savings (~ £73 per patient) while maintaining equivalent quality-adjusted life years [30]. Multinational evaluations should test this finding across healthcare systems and integrate long-term metrics, recurrence, function, and quality of life to fully define the value of fluorescence guidance in modern colorectal surgery.

Conclusion

This comprehensive RCT-level meta-analysis provides high-certainty evidence that indocyanine green fluorescence angiography (ICG-FA) significantly reduces overall anastomotic leak rates in colorectal surgery, particularly in left-sided resections. The evidence base now meets the trial-sequential sufficiency criterion, supporting both the statistical reliability and clinical relevance of the findings.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AL

Anastomotic leak

BMI

Body mass index

CI

Confidence interval

EAES

European Association for Endoscopic Surgery

ERAS

Enhanced recovery after surgery

GRADE

Grading of recommendations, assessment, development, and evaluations

ICG

Indocyanine green

ICG-FA

Indocyanine green fluorescence angiography

ISREC

International Study Group of Rectal Cancer

LOS

Length of hospital stay

MD

Mean difference

NNT

Number needed to treat

OT

Operative Time

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PROSPERO

International Prospective Register of Systematic Reviews

RCT

Randomized controlled trial

RoB 2

Risk of Bias 2 (tool)

RR

Risk ratio

SD

Standard deviation

SSI

Surgical site infection

TSA

Trial sequential analysis

Author contributions

Conceptualization: M.A. Screening: M.M., N.L., M.A. Data curation: M.M., N.L., M.A. Data analysis: H.A., N.H. Writing – original draft: M.A., M.H.Z., M.A.S., A.A. Wrote final draft: M.H.Z., M.A.S. Revision: H.A, M.H.Z., M.A.S., M.A.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary materials. Additional data are available from the corresponding author on reasonable request.

Declarations

Conflicts of interest

The authors declare no competing interests.

Ethical approval

This article is based on previously published studies and does not involve new studies with human participants or animals performed by any of the authors.

Informed consent

Not applicable, as no new human data were collected or analyzed.