Indocyanine green fluorescence–guided perfusion vs. standard assessment to prevent clinical anastomotic leak after colorectal resection: a GRADE-assessed systematic review and meta-analysis of randomized controlled trials with site-specific subgroup analysis Authors: Wajahat Mirza, Hania Iqbal, Saeeda Yasmin, Mehak Ejaz Khan, Alishbah Khan, Muhammad Bilal Moeen-Ud-Din, Hadi Mohammad Khan, Hamza Nasir Chatha, Sundus Dadan Published: 2025 ABSTRACT AbstractBackgroundAnastomotic leak remains a devastating complication after colorectal resection, with an incidence rate of 3–19%, and has profound impacts on morbidity, mortality, and quality of life. Indocyanine green (ICG) fluorescence angiography enables real-time visualization of anastomotic perfusion; however, randomized evidence regarding its efficacy in preventing clinical leaks has yielded conflicting results across recent multicenter trials.MethodsWe systematically searched the MEDLINE, Embase, CENTRAL, Web of Science, and Scopus through October 2025 for randomized controlled trials (RCTs) comparing ICG fluorescence-guided perfusion assessment with standard white-light assessment during colorectal anastomosis. The primary outcome was clinical anastomotic leakage (grade B/C). The secondary outcomes included reinterventions, complications, mortality, conversion, operative time, and length of stay. The risk of bias was assessed using the Cochrane RoB 2, and the certainty of evidence was evaluated using the GRADE framework. The systematic review protocol was prospectively registered with PROSPERO (CRD420251162495) before data extraction and analysis.ResultsSeven RCTs enrolling 4577 patients (2287 ICG fluorescence, 2290 standard assessment) met the inclusion criteria. ICG fluorescence-guided perfusion assessment significantly reduced clinical anastomotic leaks compared to the standard assessment (odds ratio [OR] 0.69, 95% confidence interval [CI] 0.56–0.86; p = 0.0009; I²=0%; moderate-certainty evidence), translating to 29 fewer leaks per 1,000 procedures. No significant differences were observed in re-interventions (OR 0.90, 95% CI 0.65–1.25; low certainty), composite complications (OR 0.86, 95% CI 0.74–1.01; low certainty), 90-day mortality (OR 0.86, 95% CI 0.74–1.01; low certainty), conversions to open surgery (OR 1.15, 95% CI 0.83–1.61; low certainty), operative time (mean difference + 2.37 min, 95% CI − 4.22 to + 8.97; low certainty), or hospital length of stay (mean difference + 0.01 days, 95% CI − 0.41 to + 0.42; low certainty).ConclusionModerate-certainty evidence demonstrates that ICG fluorescence-guided perfusion assessment reduces clinical anastomotic leaks by approximately 31% after colorectal resection, supporting its adoption to enhance intraoperative decision making and improve patient outcomes, despite no significant impact on other perioperative endpoints. However, the absence of standardized, objective criteria for ICG interpretation, accounting for patient-specific factors such as cardiovascular disease and obesity, remains a critical barrier to achieving reproducible clinical benefits and represents an essential priority for future validation studies.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12957-025-04163-9.Keywords: Indocyanine green, Fluorescence angiography, Anastomotic leak, Colorectal surgery, Perfusion assessment, Meta-analysis, Randomized controlled trial, GRADE Introduction Anastomotic leak remains one of the most feared complications following colorectal resection, with reported incidence rates ranging from 3% to 19% [1–3]. Clinical consequences include increased mortality, prolonged hospitalization, permanent stoma formation, compromised oncological outcomes, and profoundly diminished long-term quality of life [1, 4]. Inadequate tissue perfusion at the anastomotic site represents a critical pathophysiological determinant of healing failure; however, traditional intraoperative assessment methods—including visual inspection, palpation of vessels, and evaluation of bleeding from resection margins—demonstrate significant limitations in accurately quantifying microvascular perfusion and identifying occult ischemia [5, 6]. Indocyanine green (ICG) fluorescence angiography has emerged as an innovative real-time imaging modality enabling objective visualization of tissue microperfusion during colorectal surgery [7, 8]. Following intravenous administration, ICG binds to plasma proteins and emits near-infrared fluorescence (800–830 nm), delineating vascular perfusion patterns and revealing areas of compromised blood flow invisible to conventional white-light assessment [2, 6, 9]. This technology facilitates dynamic evaluation of anastomotic perfusion, potentially enabling surgeons to modify operative plans by selecting better-perfused bowel segments or converting to a diverting stoma when perfusion is inadequate [5, 10]. Recent large-scale, multicenter randomized controlled trials have yielded conflicting evidence regarding ICG fluorescence-guided perfusion assessment efficacy in preventing clinical anastomotic leaks [9–11]. This variability may reflect differences in ICG interpretation protocols, surgical techniques, anastomotic location, patient selection, and the absence of standardized criteria for defining adequate fluorescence intensity [2, 6]. Previous meta-analyses examining ICG fluorescence angiography have predominantly included observational studies alongside randomized trials, lacked standardized outcome definitions, and did not utilize rigorous certainty of evidence assessment frameworks [1, 4, 5, 12–14]. Furthermore, these syntheses have not adequately addressed potential differential efficacy across anatomical sites (colon versus rectal anastomoses), despite substantial differences in biomechanical stress, blood supply architecture, and baseline leak rates [3, 9–11]. The current systematic review and meta-analysis represent the first comprehensive GRADE-assessed synthesis exclusively focused on randomized controlled trials comparing ICG fluorescence-guided versus standard assessments for preventing clinical anastomotic leak after colorectal resection. This analysis uniquely incorporates prespecified site-specific subgroup analyses distinguishing colon from rectal anastomoses, rigorous sensitivity analyses evaluating robustness across surgical contexts and trial quality, and systematic GRADE framework assessment to transparently appraise the certainty of evidence for each outcome measure. By exclusively synthesizing high-quality randomized evidence and critically appraising the strength of recommendations through established methodological frameworks, this meta-analysis aims to definitively inform evidence-based clinical decision-making and establish whether ICG fluorescence-guided perfusion assessment should be adopted as standard practice in colorectal surgery. Methods Study design We conducted a GRADE-assessed, randomized controlled trial-only systematic review and meta-analysis comparing indocyanine green (ICG) near-infrared fluorescence perfusion assessment with standard white-light assessment at the time of colorectal anastomosis. The protocol and analysis plan were pre-specified in accordance with PRISMA 2020 and were registered prior to extraction. No a priori limits were placed on language or geography, and conference abstracts were screened for linked full text. Literature search strategy We systematically searched MEDLINE/PubMed, Embase (Ovid), Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science Core Collection, and Scopus, as well as trial registries (ClinicalTrials.gov and WHO ICTRP) from inception to October 2, 2025. Strategies combined controlled vocabulary (e.g., MeSH/Emtree for Indocyanine Green, Fluorescence Imaging, Colorectal Neoplasms, Anastomosis, Surgical) with free-text terms and proximity operators for “indocyanine green/ICG”, “fluorescence/near-infrared/NIR”, “perfusion/angiography/imaging”, “colorectal/colon/rectal/rectum”, “anastomosis/anastomotic leak”, and No language or geography limits were imposed. We also manually searched the reference lists of all included RCTs and key reviews, and performed forward citation tracking. The search strings and initial hit counts for each source are listed in Supplementary Table 1 (pre-deduplication). Records were exported and de-duplicated by a reference manager before title/abstract screening, full-text review, and study selection according to the pre-specified eligibility criteria. Eligibility criteria Eligible studies were parallel-group randomized controlled trials that enrolled adult patients who underwent colorectal resection with primary anastomosis for benign and/or malignant indications. The intervention was ICG fluorescence perfusion assessment (any platform, dose, or timing) and the comparator was a standard white-light assessment without ICG. Studies are required to report clinical anastomotic leak, typically grade B/C, within an author-defined window (approximately 30–90 days) or provide sufficient group-level data to derive it. All operative approaches (open, laparoscopic, and robotic) and anatomical sites (right/left colon and rectum) were considered eligible. Non-randomized or quasi-randomized designs, single-arm series, case–control studies, editorials, letters, and animal/cadaveric/phantom studies were excluded. Studies were ineligible if no primary anastomosis was performed (e.g., planned permanent stoma, Hartmann’s procedure), fluorescence was used in both arms, if non-perfusion indications were addressed only (e.g., ureterography/lymphography), or if clinical leak data were not available and could not be obtained after reasonable contact attempts. When multiple reports overlapped, the most complete/randomised dataset was retained. Data extraction and outcome measures Two authors (W.M. and S.D.) independently screened the titles, abstracts, and full texts, extracted data using a piloted form (study/setting, eligibility, population, and operative details; ICG protocol; whether fluorescence prompted a change of plan, endpoints, windows, and analysis set), and cross-checked all entries. Disagreements were resolved by discussion with (M.B.M.D.). The primary outcome was a clinical anastomotic leak (grade B/C) within a predefined window (approximately 30–90 days). Secondary outcomes included overall leaks (A + B + C), reoperation/reintervention for AL, protective diversion at index surgery, conversions, operative time, blood loss, complications (Clavien–Dindo), readmission, length of stay, and mortality. The ITT was prioritized when both ITT and mITT/As-treated analyses were reported. Quality (risk-of-bias) assessment Two reviewers (W.M. and S.D.) independently assessed each trial using the Cochrane Risk of Bias 2 (RoB 2) across domains (randomization, deviations from intended interventions, missing data, outcome measurement, and selection of reported results) with consensus adjudication by (M.B.M.D.). Domain-level judgment was used to assess the overall risk of bias. Statistical analysis We synthesized dichotomous outcomes as odds ratios (ORs) with 95% CIs and continuous outcomes as mean differences (MDs) (or standardized MDs when scales differed) using random effects models (Hartung–Knapp–Sidik–Jonkman as default). When trials reported medians with IQRs or ranges, we converted them to means and SDs using the Wan et al. method and Hozo et al. approach, selecting the estimator appropriate to the sample size and summary type. For zero-event cells in dichotomous outcomes, we used a 0.5 continuity correction and examined its impact on sensitivity analyses. Heterogeneity was summarized using I² and τ², and when ≥ 5 studies contributed, we also reported a 95% prediction interval. Intention-to-treat data were prioritized; if a trial reported both 30-day and 90-day leak windows, the study-specified primary window informed the main analysis. All tests were two-sided (α = 0.05) and followed a prespecified analysis codebook for reproducibility. Sensitivity analysis, subgroup analysis, and certainty of evidence To probe for robustness, we conducted leave-one-out sensitivity analyses for the meta-analysis exhibiting the highest heterogeneity (by I²) and for any analysis that was disproportionately influenced by a single large trial (study weight > 40%). We prespecified a clinically coherent two-level subgroup contrasting rectal-only RCTs (PILLAR III, EssentiAL, IntAct) versus mixed/colon-inclusive RCTs (FLAG, De Nardi, AVOID, and ICG-COLORAL). The certainty of evidence was appraised per outcome using GRADE, with each domain explicitly rated as not serious, serious (downgraded by one level), or very serious (downgraded by two levels) for the risk of bias, inconsistency, indirectness, imprecision, and (where applicable) publication bias. The overall certainty was summarized as high, moderate, low, or very low, with transparent justifications and presentation of absolute risk differences at typical baseline risks. Results We identified 1,012 records across the five databases and removed 312 duplicates before screening, leaving 700 unique records for title/abstract reviews. After excluding 606 citations that clearly failed to meet eligibility criteria, we retrieved and assessed 94 full texts, none of which were unretrievable. Ultimately, seven randomized controlled trials met the inclusion criteria, with most exclusions due to a non-randomized design (n = 54) and ineligible intervention/comparator (n = 10) (Fig. 1). This flow demonstrates a targeted search with disciplined screening focused on RCT evidence of ICG-guided perfusion versus standard assessment. PRISMA 2020 flow diagram of study selection for RCTs comparing indocyanine green fluorescence–guided perfusion with standard assessment in colorectal anastomosis Characteristics of included studies All trials randomized adults undergoing colorectal resection with primary anastomosis to ICG-guided perfusion assessment versus standard care with anastomotic leak as the principal outcome. Two trials were rectal cancer only (EssentiAL, IntAct), while the others included mixed benign/malignant disease and broader colorectal sites, yielding pragmatic generalizability. The primary endpoints were uniformly leak-focused but differed in grading window (30 vs. 90 days) and definition (all grades vs. clinical B/C), which we addressed in pre-specified subgroup/sensitivity analyses anchored to the study-specific primary outcome (Table 1). Table 1Study characteristics of the seven randomized controlled trials comparing indocyanine green (ICG) fluorescence–guided perfusion assessment with standard white-light assessment during colorectal resection with primary anastomosisStudy ID (trial)CountrySettingPopulation—ICG (N)Population—Control (N)Procedure scopeKey inclusion criteriaKey exclusion criteriaPrimary endpoint (window)Main secondary endpoints (examples) Alekseev 2020 [15]Russia (multicenter)Tertiary colorectal units187190Left-sided colon & rectal resections with stapled colorectal anastomosis; lap/robotic/open allowedAdults undergoing elective left-sided/rectal resection with planned primary colorectal anastomosisNo anastomosis created; emergency surgery; contraindication to ICGAnastomotic leak (A + B + C) within ~ 30 daysClinical leak B/C; operative time; blood loss; positive air-leak test; diversion; LOS; complications De Nardi 2020 [7]Italy (multicenter)Tertiary hospitals118122Left colectomy and LAR (laparoscopic/robotic/open)Adults with left-sided colon or rectal resection and planned primary anastomosisNo anastomosis; emergency; ICG contraindication30-day anastomotic leak (A + B + C)Leak grades by Clavien–Dindo; reinterventions; complications; LOS Jafari 2021 [16]USA (multicenter)Academic colorectal centers178169Low/mid rectal cancer resections (LAR) — open/lap/roboticAdults undergoing sphincter-preserving rectal resection ( ≤ ~ 10 cm from anal verge) with planned anastomosisNo anastomosis; abdominoperineal resection; ICG contraindicationAnastomotic leak (A + B + C) (≈ 30–90 days per trial reporting)Reinterventions; stoma complications; ileus; conversions; mortality; LOS Watanabe 2023 [8]Japan (41 centers)Multicenter MIS rectal cancer program422417Minimally invasive rectal cancer resections (lap/robotic)Adults with primary rectal cancer scheduled for MIS resection with primary anastomosisNo anastomosis; T4b tumor; protocol violations; ICG contraindicationAnastomotic leak (A + B + C) (30 days; also, ITT reported)Clinical leak (B + C); change of transection line; time-to-fluorescence; reoperation; complications; LOS; safety of ICG Faber 2024 [11]Netherlands (multicenter)Pragmatic national trial in tertiary hospitals463468Broad colorectal resections (right, left, rectal) — lap/robotic dominantAdults undergoing elective colorectal resection with planned primary anastomosisNo anastomosis; emergency; ICG contraindication90-day clinically relevant leak (B/C) (overall and by side)30-day leak; serious adverse events; reintervention; conversions; operative time; LOS; readmissions; mortality Rinne 2025 [10]Finland (multicenter)Tertiary colorectal units536541Laparoscopic colorectal resections excluding low anterior resectionAdults undergoing elective laparoscopic colorectal resection with planned primary anastomosisLow anterior resections; no anastomosis; emergency; ICG contraindicationOverall anastomotic leak (A + B + C) (≈ 30 days)Right- vs. left-sided leak; change of anastomotic site; conversions; complications; reoperations; LOS Jayne 2025 [9]UK (multicenter)Specialist rectal cancer centers383383Rectal cancer resections (lap/robotic/open)Adults with rectal cancer scheduled for anterior resection and primary anastomosisNo anastomosis; permanent stoma planned; ICG contraindicationClinical leak (B/C) by 90 days (adjusted OR)All leaks (A + B + C); change in planned anastomosis/transection; complications; reinterventions; mortality; LOSAbbreviations: ICG indocyanine green, LAR low anterior resection, MIS minimally invasive surgery, LOS length of stay, A/B/C anastomotic leak grades per Clavien–Dindo-based clinical grading (A = no active intervention; B = active therapy without relaparotomy; C = re-laparotomy), mITT modified intention-to-treat Study characteristics of the seven randomized controlled trials comparing indocyanine green (ICG) fluorescence–guided perfusion assessment with standard white-light assessment during colorectal resection with primary anastomosis Abbreviations: ICG indocyanine green, LAR low anterior resection, MIS minimally invasive surgery, LOS length of stay, A/B/C anastomotic leak grades per Clavien–Dindo-based clinical grading (A = no active intervention; B = active therapy without relaparotomy; C = re-laparotomy), mITT modified intention-to-treat Across the seven RCTs, the baseline characteristics were well-balanced between the ICG and control arms (Table 2). Ages were clustered in the late-50s to early-70s and BMI ~ 26–28 kg/m² (reported as mean ± SD or median with IQR/range), with similar proportions of men, diabetes, and current smokers by arm. Three studies enrolled patients with rectal cancer only (PILLAR III, EssentiAL, IntAct), one enrolled patient with colon cancer only (ICG-COLORAL), and AVOID/Alekseev/De Nardi included the left-sided colon and/or rectum; per-arm site distributions and rectal levels are shown in Table 2. The surgical approach varied by design (e.g., ICG-COLORAL 100% laparoscopic; EssentiAL MIS-only; AVOID predominantly laparoscopic/robotic), and protective diversion was the highest in rectal trials (Table 2). Table 2Baseline demographics and clinical characteristics reporting across seven randomized trialsStudy IDICG NControl NAge (y)†Male (%) ICG / ControlBMI (kg/m²)†Malignant (%) ICG / ControlBenign (%) ICG / ControlNeoadjuvant RT (%) ICG / ControlNeoadjuvant CT (%) ICG / ControlApproach (% Open / Lap / Robotic) — ICGApproach (% Open / Lap / Robotic) — ControlConversion to open (%) ICG / ControlProtective diversion at index (%) ICG / ControlSite distribution — ICG (% Right / Left / Rectum)Site distribution — Control (% Right / Left / Rectum)Rectal level — ICGRectal level — ControlDiabetes (%) ICG / ControlCurrent smoker (%) ICG / ControlCancer histology (if reported) Alekseev 202018719063 (21–86) / 63 (66–85)49.2 / 48.4NR97.9 / 96.82.1 / 3.2CRT 10.7 / 10.0NR53.5 / 46.5 / 059.5 / 40.5 / 0NR / NR71.1 / 70.50 / 40.6 / 59.40 / 44.7 / 55.3High 9–15 cm: 40.6%; Low < 8 cm: 59.4%High: 44.7%; Low: 55.3%8.0 / 8.9NR / NRColorectal adenocarcinoma; cohort = mixed (benign + malignant). De Nardi 202011812266.1 / 65.150.8 / 54.125.2 / 25.672.9 / 69.727.1 / 30.323.7 / 17.222.9 / 17.2NR / NR / NRNR / NR / NRNR / NRNR / NR0 / 52.5 / 47.50 / 56.6 / 43.4NRNR14.4 / 15.6NR / NRColorectal adenocarcinoma; cohort = mixed (benign + malignant). Jafari 202117816957.2 ± 11.4 / 57.0 ± 11.461.2 / 58.627.8 ± 5.6 / 28.2 ± 5.9100 / 1000 / 0Long-course 63.5 / 65.79.6 / 6.415.6 / 48.0 / 36.413.8 / 44.9 / 41.37.4 / 4.873.7 / 80.40 / 0 / 1000 / 0 / 100Upper: 15.3%; Low/Mid: 81.4%Upper: 14.2%; Low/Mid: 84.6%12.4 / 13.627.5 / 18.3Rectal adenocarcinoma Watanabe 202342241766 (56–73) / 67 (58–74)63 / 65NR / NR100 / 1000 / 0CRT 5.0; SRT 0.2 / CRT 7.2; SRT 0.3NAC 5.7 / 5.80 / 43.6 / 47.90 / 45.6 / 44.10.7 / 0.750.5 / 54.90 / 0 / 1000 / 0 / 100NRNR14.7 / 17.3NR / NRRectal adenocarcinoma Faber 202446346869 [60–75] / 67 [59–76]54 / 5025.8 [23.5–29.9] / 25.9 [23.4–29.0]87 / 9013 / 10~ 5 / ~5~ 1–2 / ~1NR / 78 / 22NR / 78 / 227 / 7NR / NR41 / 38 / 2144 / 34 / 22NRNR13 / 1311 / 13Colorectal adenocarcinoma; cohort = mixed (benign + malignant). Rinne 202553654170 (11) / 70 (11)55.4 / 53.227.6 (5) / 27.6 (5)70.9 / 70.829.1 / 29.2NR / NRNR / NR0 / 100 / 00 / 100 / 05.6 / 4.30.8 / 1.953 / 47 / 054 / 46 / 0NRNRNR / NR10.4 / 11.7Colonic adenocarcinoma; cohort = mixed (benign + malignant). Jayne 202538338362.9 (11.6) / 63.3 (11.5); median 64 [55–72] both66 / 6526.7 (4.7) / 26.5 (4.8)100 / 1000 / 0Long 36; Short 13 / Long 36; Short 139 / 9NR / NR / NRNR / NR / NRNR / NRNR / NR0 / 0 / 1000 / 0 / 100Above PR: 30%; At PR: 28%; Below PR: 41%Above PR: 31%; At PR: 28%; Below PR: 41%NR / NR12 / 13Rectal adenocarcinomaAbbreviations: ICG indocyanine green, RT radiotherapy, CT chemotherapy, CRT chemoradiotherapy, Lap laparoscopic, MIS minimally invasive surgery, PR peritoneal reflection, LAR low anterior resection, NR Not Reported†Age and BMI shown as reported: mean ± SD or median [IQR/range]. Baseline demographics and clinical characteristics reporting across seven randomized trials Abbreviations: ICG indocyanine green, RT radiotherapy, CT chemotherapy, CRT chemoradiotherapy, Lap laparoscopic, MIS minimally invasive surgery, PR peritoneal reflection, LAR low anterior resection, NR Not Reported †Age and BMI shown as reported: mean ± SD or median [IQR/range]. Methodological quality and study integrity The overall methodological quality was acceptable across the five RoB 2 domains (Fig. 2a–b). Bias due to deviations from intended interventions (D2) was uniformly judged “some concerns” because surgeons could not be blinded, and ICG-FA can legitimately alter intra-operative decisions (Fig. 2a). Randomization (D1) was low-risk in the four contemporaneous multicenter RCTs, with “some concerns” in three earlier trials in which allocation concealment was less clearly reported (Fig. 2a). Missing outcome data (D3) was at low risk for most trials, with only one study at “some concern”; outcome measurement (D4) and selective reporting (D5) were at low risk throughout, reflecting objective leak definitions and pre-specified analyses (Fig. 2b). Consequently, the overall risk of bias was “some concerns” across studies, driven primarily by D2 rather than data completeness or outcome assessment (Fig. 2a). This pattern supports, at most, one-level downgrading for risk of bias in GRADE, with sensitivity analyses focusing on D1 differences. Risk of bias (RoB 2.0) for the primary outcome across seven randomized trials. Panel A: traffic-light plot showing study-level judgments across five domains (green = low risk; yellow = some concerns). Panel B: domain summary plot displaying the proportion of studies at each judgment Primary outcomes Clinical anastomotic leak Pooled random-effects OR = 0.69 (95% CI 0.56–0.86; Z = 3.32; p = 0.0009), with no heterogeneity (I²=0%), indicating a statistically significant reduction in clinical leaks with ICG-FA versus standard assessment. The effect was consistent across trials and was driven by larger multicenter RCTs with no opposing outliers. Secondary outcomes Any re-intervention (includes re-operation, IR drainage, endoscopic therapy) The pooled OR was 0.90 (95% CI 0.65–1.25; p = 0.53), with moderate heterogeneity (I²=49%). Although the point estimate trends toward fewer re-interventions with ICG-FA, confidence intervals cross unity, and results are not statistically significant. One trial (EssentiAL) suggested a benefit, whereas the others were neutral; overall, there was no clear difference. Any postoperative complication (composite) Pooled OR = 0.86 (95% CI 0.74–1.01; p = 0.07), with I²=0%. These findings suggest a possible reduction in overall postoperative complications with ICG-FA, but the CI marginally overlaps 1.0, so the effect does not reach statistical significance. The direction favors ICG-FA, but remains imprecise. Mortality (90 days) Pooled OR = 0.86 (95% CI 0.74–1.01), I²=0%. Mortality was not significantly different between the groups, although the point estimate suggests a small potential reduction with ICG-FA; low event rates and narrow CIs straddling unity limit certainty. Conversion to open (from lap/robotic) Pooled OR = 1.15 (95% CI 0.83–1.61; p = 0.40), I²=0%. Conversions were similar between strategies; the point estimate > 1.0, hints at slightly more conversions with ICG-FA, but CIs were wide and cross-unity, with no statistically significant difference. Operative time (minutes) Mean difference (ICG-FA minus standard) + 2.37 min (95% CI − 4.22 to + 8.97; p = 0.48), with moderate heterogeneity (I²=49%). Times were not significantly different overall; study estimates were mixed (some shorter and some longer with ICG-FA). Length of hospital stay (days) Mean difference + 0.01 days (95% CI − 0.41 to + 0.42; p = 0.97), I²=51%. Hospital stay was virtually identical between the groups, with moderate between-study variability, but a pooled effect centered at zero. Across randomized evidence, the only outcome with clear statistical significance was the reduction in clinical anastomotic leakage with ICG-FA (Fig. 3). All other secondary outcomes—re-intervention, composite complications, 90-day mortality, conversion to open surgery, operative time, and hospital stay— showed no statistically significant differences (Figs. 4, 5, 6, 7, 8 and 9), although several trends favored ICG-FA without crossing conventional thresholds. Forest plot—effect of ICG-fluorescence perfusion assessment vs. standard white-light on clinical anastomotic leak (random-effects odds ratio) Forest plot—effect of ICG-fluorescence vs. standard on any re-intervention (re-operation/IR drainage/endoscopic therapy; random-effects odds ratio) Forest plot—effect of ICG-fluorescence vs. standard on any postoperative complication (composite) (random-effects odds ratio) Forest plot—effect of ICG-fluorescence vs. standard on 90-day mortality (random-effects odds ratio) Forest plot—effect of ICG-fluorescence vs. standard on conversion to open from minimally invasive approaches (random-effects odds ratio) Forest plot—effect of ICG-fluorescence vs. standard on operative time (minutes) (random-effects mean difference) Forest plot—effect of ICG-fluorescence vs. standard on length of hospital stays (days) (random-effects mean difference) Sensitivity analysis As shown in Table 3, the primary meta-analysis for any re-intervention was neutral (OR 0.90, p = 0.53; I² = 49%), and the removal of Watanabe 2023 attenuated heterogeneity (I² = 18%) without altering the non-significant effect (OR 0.97, p = 0.83). For operative time, the all-studies result was non-significant (MD + 2.37 min, p = 0.48; I² = 49%), but excluding Watanabe 2023 yielded a small, significant increase favoring standard assessment (MD + 4.67 min, p = 0.02) with eliminated heterogeneity (I² = 0%). A similar pattern was observed for length of stay, which was overall neutral (MD + 0.01 days, p = 0.97; I² = 51%), with a modest but significant increase after exclusion (MD + 0.24 days, p = 0.03; I² = 0%). In summary, the sensitivity checks (Table 3) show that the results for re-intervention are robust, whereas operative time and length of stay are sensitive to the influential EssentiAL trial and, after its exclusion, slightly favor the standard (non-ICG) approach. Table 3Leave-one-out sensitivity analysis for key outcomes (any re-intervention, operative time, and length of hospital stay) comparing ICG-fluorescence–guided perfusion with standard assessment using random-effects modelsOutcomeEffect Estimate (Before)P-value (Before)I²% (Before)Study RemovedEffect Estimate (After)P-value (After)I²% (After)Any re-intervention (OR)0.90 [0.65–1.25]0.5349% Watanabe 2023 (EssentiAL)0.97 [0.77–1.24]0.8318%Operative time (min) (MD)+ 2.37 [− 4.22 to + 8.97]0.4849% Watanabe 2023 (EssentiAL)+ 4.67 [+ 0.66 to + 8.69]0.020%Length of hospital stay (days) (MD)+ 0.01 [− 0.41 to + 0.42]0.9751% Watanabe 2023 (EssentiAL)+ 0.24 [+ 0.03 to + 0.46]0.030% Leave-one-out sensitivity analysis for key outcomes (any re-intervention, operative time, and length of hospital stay) comparing ICG-fluorescence–guided perfusion with standard assessment using random-effects models Subgroup analysis Across the prespecified subgroups, the length of stay (LOS) showed divergent but small patterns. In rectal-only trials (IntAct, EssentiAL), the pooled mean difference was − 0.26 days (95% CI − 0.84 to 0.32), indicating no significant difference and moderate heterogeneity (I² = 42%). In colorectal/colon-inclusive trials (FLAG, AVOID), LOS was slightly longer with ICG-FA (+ 0.29 days, 95% CI + 0.05 to + 0.53; I² = 0%), favoring standard care. The overall effect remained null (+ 0.01 days, 95% CI: −0.41 to + 0.42). A between-subgroup interaction suggested possible effect variation (χ² = 2.93, p = 0.09, I² = 65.8%) but did not reach statistical significance. These findings imply that case mix (rectal-only vs. mixed/colon-inclusive) likely explains much of the observed heterogeneity (Supplementary Fig. 1). GRADE-certainty of evidence Across outcomes, ICG fluorescence probably reduced clinical anastomotic leak with moderate certainty of evidence (OR 0.69, 95% CI 0.56–0.86), translating to approximately 29 fewer leaks per 1,000 procedures compared with standard assessment (Table 4). The effects on any reintervention and composite postoperative complications were uncertain (ORs 0.90 and 0.86, respectively) with low certainty because the CIs spanned benefit and no effect, and the event rates varied (Table 4). Mortality within 90 days was rare and imprecisely estimated (OR 0.86), yielding low certainty and only a small absolute difference of approximately two fewer deaths per 1,000 (Table 4). Conversion to open surgery showed no clear difference (OR 1.15; low certainty), and continuous outcomes (operative time (MD + 2.37 min) and length of stay (MD + 0.01 days)were effectively neutral, both rated as low certainty due to imprecision and/or inconsistency (Table 4). Overall, Table 4 indicates a probable benefit limited to clinical leaks, with other perioperative outcomes showing no convincing advantage of ICG guidance. Table 4GRADE evidence profile for indocyanine-green fluorescence guidance versus standard assessment in colorectal resectionOutcomeRisk of BiasInconsistencyIndirectnessImprecisionPublication BiasEffect Estimate (pooled)Absolute Effect (per 1,000) *GRADE CertaintyClinical anastomotic leak (primary)SeriousNot serious (I²≈0%)Not seriousNot serious (CI excludes no effect)UndetectedOR 0.69 (95% CI 0.56–0.86)101/1,000 in control → 72/1,000 with ICG (29 fewer, 95% CI 13–42 fewer)Moderate ⊕⊕⊕◯Any re-intervention (re-op/IR/endo)SeriousNot serious (I²≈49%)Not seriousSerious (CI spans benefit/harm)UndetectedOR 0.90 (0.65–1.25)109/1,000 in control → 99/1,000 with ICG (10 fewer, 95% CI 35 fewer to 24 more)Low ⊕⊕◯◯Any postoperative complication (composite)SeriousNot serious (I²≈0%)Not seriousSerious (borderline CI includes little benefit to no effect)UndetectedOR 0.86 (0.74–1.01)295/1,000 in control → 265/1,000 with ICG (30 fewer, 95% CI 58 fewer to 3 more)Low ⊕⊕◯◯Mortality (≤ 90 days)SeriousNot serious (I²≈0%)Not seriousSerious (rare events; wide CI around no effect)UndetectedOR 0.86 (0.74–1.01)≈ 12/1,000 in control → ≈10/1,000 with ICG (~ 2 fewer, CI includes no effect)Low ⊕⊕◯◯Conversion to openSeriousNot serious (I²≈0%)Not seriousSerious (CI spans benefit/harm)UndetectedOR 1.15 (0.83–1.61)43/1,000 in control → 49/1,000 with ICG (6 more, 95% CI 7 fewer to 24 more)Low ⊕⊕◯◯Operative time (min)SeriousNot serious (I²≈49%)Not seriousSerious (wide CI; clinically trivial to moderate)UndetectedMD + 2.37 min (− 4.22 to + 8.97)Not applicableLow ⊕⊕◯◯Length of hospital stay (days)SeriousSerious (I²≈51%)Not seriousNot serious (narrow CI around null)UndetectedMD + 0.01 days (− 0.41 to + 0.42)Not applicableLow ⊕⊕◯◯CER denotes the control event rate; pooled effects are expressed as odds ratios (ORs) with 95% confidence intervals (CIs) for dichotomous outcomes and mean differences (MDs) for continuous outcomesThe GRADE domains—risk of bias, inconsistency, indirectness, imprecision—are judged as Not serious, Serious, or Very serious based on ROB2 appraisals, magnitude of heterogeneity, applicability of populations/interventions/outcomes, and precision of the pooled estimate; publication bias is judged Undetected (insufficient studies for formal testing). Overall certainty is summarized with symbols: ⊕⊕⊕◯ = Moderate, ⊕⊕◯◯ = Low (higher or lower grades would be denoted by adding or removing ⊕ accordingly) GRADE evidence profile for indocyanine-green fluorescence guidance versus standard assessment in colorectal resection CER denotes the control event rate; pooled effects are expressed as odds ratios (ORs) with 95% confidence intervals (CIs) for dichotomous outcomes and mean differences (MDs) for continuous outcomes The GRADE domains—risk of bias, inconsistency, indirectness, imprecision—are judged as Not serious, Serious, or Very serious based on ROB2 appraisals, magnitude of heterogeneity, applicability of populations/interventions/outcomes, and precision of the pooled estimate; publication bias is judged Undetected (insufficient studies for formal testing). Overall certainty is summarized with symbols: ⊕⊕⊕◯ = Moderate, ⊕⊕◯◯ = Low (higher or lower grades would be denoted by adding or removing ⊕ accordingly) Discussion This systematic review and meta-analysis of seven randomized controlled trials enrolling 4,577 patients demonstrated that ICG fluorescence-guided perfusion assessment significantly reduced clinical anastomotic leaks compared to standard white-light assessment (OR 0.69, 95% CI 0.56–0.86; p = 0.0009; I²=0%; moderate-certainty evidence), translating to 29 fewer leaks per 1,000 procedures at a typical baseline risk of 10.1%. This represents a clinically meaningful 31% relative risk reduction in most morbid complications of colorectal surgery, despite neutral effects on reinterventions, composite complications, mortality, conversions, operative time, and hospital stay. As the first RCT-only, GRADE-assessed synthesis incorporating pre-specified site-specific subgroup analyses, this work establishes moderate-certainty evidence supporting ICG adoption for high-risk anastomoses while highlighting persistent uncertainties regarding standardization, cost-effectiveness, and generalizability across anatomical sites and patient phenotypes. The mechanism underlying leak reduction centers on ICG-enabled objective perfusion mapping triggers concrete intraoperative decision modifications—more proximal transection to the better-perfused bowel, anastomotic relocation, or protective diversion when fluorescence reveals marginal flow [5]. Across the seven included trials [7–11, 15, 16], change-of-plan rates ranged from 5% to 15% in ICG arms versus < 1% with standard assessment, with altered transection sites occurring in 8–12% of ICG-guided cases and increased protective diversion in borderline perfusion scenarios. ICG dosing varied from 0.05 to 0.2 mg/kg administered 1–5 min pre-assessment, with surgeons applying qualitative fluorescence criteria—homogeneous mucosal/serosal enhancement, symmetric marginal arterial filling, absence of dark zones—to guide adequacy judgments [2, 6]. This real-time visualization overcomes the subjective limitations of visual inspection, palpation, and bleeding assessment, which fail to reliably identify occult ischemia and predict anastomotic viability [3, 6]. The target population for maximal ICG benefit comprises high-risk anastomoses: low- and mid-rectal resections with baseline leak rates of 10–20%, narrow pelvic workspace limiting vascular assessment, post-chemoradiotherapy fibrotic tissue, tenuous inferior mesenteric artery ligation margins, and patient-level risk factors, including cardiovascular disease, obesity, diabetes, and smoking [3, 9–11]. Three rectal-only trials [9, 10, 16] enrolled such cohorts, with 36–65% receiving neoadjuvant chemoradiotherapy and 50–80% requiring protective diversion. Conversely, well-vascularized right colectomies with robust ileocolic pedicles and baseline leak rates of < 5% demonstrate marginal incremental value, raising cost-effectiveness concerns for routine universal application [10, 11]. This risk stratification aligns with our prespecified subgroup framework distinguishing rectal-only from colon-inclusive trials, where baseline leak gradients and biomechanical stresses differ substantially, suggesting tailored adoption strategies rather than blanket implementation across all colorectal resections [12–14]. Site-specific subgroup analysis revealed consistent leak reduction across rectal and colon-inclusive trials without significant between-subgroup heterogeneity, although rectal-only studies [8, 9, 16] tended toward larger absolute benefits given higher baseline risks compared to mixed colon-rectal cohorts [7, 10, 11, 15]. For length of stay, rectal-only trials showed neutral effects (MD − 0.26 days, 95% CI − 0.84 to + 0.32; I²=42%), whereas colon-inclusive trials demonstrated marginally longer stays with ICG (MD + 0.29 days, 95% CI + 0.05 to + 0.53; I²=0%), suggesting possible effect modification by case-mix (χ²=2.93, p = 0.09, I²=65.8%). Leave-one-out sensitivity analyses confirmed robustness of the primary leak outcome, but exclusion of one influential trial [8] eliminated heterogeneity in operative time and length of stay, revealing slight increases favoring standard assessment (MD + 4.67 min, p = 0.02; MD + 0.24 days, p = 0.03). These patterns underscore differential perioperative workflows and diversion practices across rectal versus colonic surgery, with ICG-prompted protective stoma creation potentially extending hospital duration without altering complication profiles, thereby explaining neutral downstream metrics despite the reduced anastomotic failure rates. Our trial-purified findings reconcile prior inconsistencies in mixed-design meta-analyses that included observational studies with inherent selection bias, variable leak definitions (combining grade A biochemical leaks with clinical B/C leaks), and heterogeneous follow-up windows [1, 4, 5, 12–18]. Earlier syntheses [1, 4, 5, 12–18] reported ORs ranging from 0.38 to 0.72 for leak reduction, but pooled RCTs and cohort studies without site stratification or rigorous certainty assessment. Narrative reviews [2, 6, 19, 20] and comprehensive meta-analyses [21–23] highlighted technical promise but noted the absence of standardized fluorescence criteria and interobserver variability challenges [6, 24–26]. By exclusively synthesizing seven multicenter RCTs with clinically relevant grade B/C leak endpoints, employing GRADE certainty appraisal, and conducting pre-specified rectal-versus-colon subgroup analyses, this study provides definitive moderate-certainty evidence supporting ICG efficacy while exposing persistent protocol heterogeneity and the critical need for objective, quantitative perfusion thresholds to maximize reproducibility and clinical translation across diverse surgical contexts. The neutral secondary endpoints—reinterventions, composite complications, 90-day mortality, conversions, operative time, and hospital stay—reflect multifactorial influences: low event rates underpowering mortality comparisons (OR 0.86, 95% CI 0.74–1.01), rescue interventions such as percutaneous drainage blunting reintervention differences (OR 0.90, 95% CI 0.65–1.25), and standardized enhanced recovery pathways narrowing length-of-stay variability (MD + 0.01 days, 95% CI − 0.41 to + 0.42). ICG-prompted protective diversion may prevent severe leaks but extend hospitalization without reducing overall complication rates, while minimal operative time increases (+ 2.37 min) reflect efficient fluorescence integration into established workflows [8, 10, 11]. Critically, preventing clinical grade B/C anastomotic leaks, requiring active intervention, re-laparotomy, or prolonged morbidity, represents a major patient-important outcome with profound quality-of-life implications, permanent stoma risk, and oncological consequences in cancer patients [1, 3, 4, 9], justifying ICG adoption even without secondary endpoint differences. Propensity-matched cohort studies [27, 28] and real-world retrospective analyses [29–32] have similarly demonstrated leak reduction under ICG guidance, supporting the external validity of these randomized findings across diverse practice settings. This analysis has important limitations: unavoidable performance bias from unblinded surgeons (downgrading GRADE certainty by one level), protocol heterogeneity in ICG dosing (0.05–0.2 mg/kg), timing (1–5 min pre-assessment), and number of injections (single versus multiple), lack of validated quantitative thresholds for adequate fluorescence intensity or time-to-peak ingress, documented interobserver variability in fluorescence interpretation, and site-mix indirectness when generalizing rectal-specific benefits to low-risk colonic resections [6, 24–26]. Insufficient trial numbers precluded formal publication bias assessment using funnel plots or Egger’s test, although comprehensive database searches and trial registry screening minimized retrieval bias. Moderate certainty for the primary leak outcome reflects these methodological constraints alongside the need for objective, reproducible perfusion metrics, including quantitative fluorescence indices, standardized timing protocols, and validated interpretation criteria [24, 25, 33]—to translate enhanced intraoperative visualization into consistent clinical benefit across varied surgical settings, patient phenotypes, and operator experience levels. Clinicians should consider routine ICG fluorescence guidance for high-risk scenarios such as low rectal anastomoses, post-chemoradiotherapy pelvic surgery, compromised vascular pedicles, patients with cardiovascular disease or obesity limiting marginal perfusion, and any anastomosis with borderline intraoperative viability concerns [3, 8–11, 16]. Selective use for low-risk right colectomies balances incremental costs with marginal benefit [10, 11, 23, 34]. Future priorities include developing standardized, quantitative perfusion metrics (time-to-peak fluorescence, relative intensity indices, washout kinetics) [24, 25, 33], structured training programs to reduce interobserver variability [26]; site-stratified cost-effectiveness analyses incorporating quality-of-life gains and stoma avoidance [19, 21–23]; patient-specific algorithms integrating comorbidities and tissue quality [29–32, 35]; and prospective registries or adaptive platform trials powered for long-term oncological outcomes, permanent stoma rates, and patient-reported quality-of-life endpoints [2, 9, 20, 34]. Ultimately, realizing ICG’s full potential of ICG requires transitioning from subjective qualitative assessment to objective, reproducible perfusion thresholds that enable precision-guided anastomotic decision-making and maximize clinical benefits across diverse colorectal surgical populations. Strengths and limitations This systematic review and meta-analysis has several important strengths. First, we employed a rigorous methodology by exclusively including parallel-group randomized controlled trials, the highest level of evidence for therapeutic interventions, and systematically excluding non-randomized designs that are prone to selection bias and confounding. Second, we applied the GRADE framework to transparently assess the certainty of evidence for each outcome, providing explicit justification for downgrading across domains (risk of bias, inconsistency, indirectness, imprecision, and publication bias) and calculating absolute risk differences at typical baseline rates. Third, our pre-specified analysis plan, prospective registration in PROSPERO, and adherence to the PRISMA 2020 reporting standards enhanced reproducibility and minimized selective outcome reporting. Fourth, we conducted comprehensive sensitivity analyses, including leave-one-out analyses for outcomes with high heterogeneity or disproportionate single-study influence, and clinically coherent subgroup analyses stratifying rectal-only versus colo-rectal/colon-inclusive trials, to evaluate robustness and explore sources of heterogeneity. Fifth, our systematic search strategy spanned five major databases and trial registries with no language or geographic restrictions and included forward/backward citation tracking to maximize the retrieval of eligible studies. Despite these strengths, several limitations warrant consideration. First, all included trials were unblinded, as surgeons must visualize fluorescence data to modify their operative plan, introducing potential performance and detection bias; although leak definitions were objective and typically assessed by clinicians unaware of intraoperative fluorescence findings, differential perioperative management cannot be entirely excluded. Second, substantial heterogeneity existed in ICG protocols across trials, including variations in dosing, timing of injection, number of injections, and criteria for defining adequate perfusion, limiting our ability to establish optimal standardized protocols and potentially contributing to variability in effect estimates. Third, while our primary outcome (clinical anastomotic leak) achieved moderate certainty, all secondary outcomes were rated with low certainty due to imprecision (wide confidence intervals spanning benefits and no effects) and, in some cases, moderate heterogeneity, which constrained confidence in the null findings for re-interventions, complications, and mortality. Conclusions This GRADE-assessed systematic review and meta-analysis of randomized controlled trials demonstrated that indocyanine green fluorescence-guided perfusion assessment probably reduced clinical anastomotic leaks after colorectal resection by approximately 31% compared to standard white-light assessment, translating to 29 fewer leaks per 1,000 procedures with moderate certainty of evidence. While ICG fluorescence did not significantly impact secondary outcomes such as reinterventions, composite complications, mortality, operative time, or length of stay (all low certainty), the consistent and clinically meaningful reduction in anastomotic leak, one of the most devastating complications in colorectal surgery, supports its consideration as an adjunct to enhance surgical decision making and improve patient outcomes. However, a critical limitation is the absence of standardized, objective criteria for interpreting ICG fluorescence intensity that account for patient-specific physiological variables, such as cardiovascular comorbidities and obesity, which significantly alter ICG pharmacokinetics. Future research must prioritize the development and validation of uniform, quantitative perfusion thresholds, risk-stratified interpretation algorithms, identification of patient subgroups most likely to benefit, and formal cost-effectiveness analyses to achieve the validity and reliability necessary for definitive clinical adoption and maximize the potential of real-time perfusion assessment in reducing the burden of anastomotic leaks in colorectal surgery. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements We express our sincere gratitude to Dr. Hadi Mohammad Khan (H.M.K.), the senior-most author on this study, for his invaluable clinical guidance, strategic oversight, and mentorship throughout the project. Authors’ contributions W.M. conceived the study, developed the protocol, led the methodology, and had full access to all data; W.M. coordinated the project, supervised screening and extraction, conducted the primary statistical analyses, and drafted the manuscript’s first and revised versions. W.M. and S.D. conducted the database searches, performed independent screening, and extracted data; M.B.M.D. and H.I. adjudicated discrepancies, and M.E.K. verified extracted data against the full texts and prepared the study characteristics and outcomes tables. Risk-of-bias assessment (RoB 2) was performed independently by W.M. and S.D., with methodological oversight from S.Y. and M.B.M.D. GRADE certainty appraisals and sensitivity/subgroup analyses were undertaken by W.M., H.N.C., and A.K. Statistical analysis and figure generation were led by W.M., with critical input from H.N.C. and M.B.M.D. The initial draft was written by W.M. and H.I.; all authors (S.Y., M.E.K., A.K., H.M.K., H.N.C., S.D., M.B.M.D.) critically revised the manuscript for important intellectual content, approved the final version, and agreed to be accountable for the work. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. This research did not receive any external grant support. However, internal research-support expenses—including access to academic software, statistical and language-refinement tools, Ithenticate services, and paid retrieval of articles behind paywalls—were covered by Dr. Saeeda Yasmin. The authors gratefully acknowledge this contribution. Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate Not applicable (systematic review of published aggregate data). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement No datasets were generated or analysed during the current study.