A search of PubMed, Embase, Web of Science, The Cochrane Library, Chinese National Knowledge Infrastructure, VIP Database, China Biology Medicine Disc, and Wan-fang database was conducted on 10 April 2021 by two independent researchers. The retrieval terms included ICG and esophagectomy. Both database subject heading fields (Emtree in EMBASE, MeSH in MEDLINE) and text word fields were searched. References of retrieved articles and reviews were also manually screened to obtain additional relevant studies.

The inclusion criteria included: (1) esophagectomy with gastric conduit reconstruction; (2) comparative study design: control group using color, temperature, and pulsation of vessels; experimental group using ICG to assess perfusion; (3) age ≥18 years; (4) sufficient clinical outcome data for further analysis; and (5) sample size ≥10. The exclusion criteria were as follows: (1) abstract or review; (2) data insufficient; (3) repeated publications; and (4) included in a multicenter study.

We used the Methodological Index for Non-Randomized Studies tool (MINORS) (score from 0 to 24). MINORS is a 12-point validated tool designed specifically to evaluate the methodological quality of non-randomized surgical trials. For the comparative studies, the ideal score was 24. A MINORS score below 12 indicates poor quality. A MINORS score above 18 indicates good quality (16, 17). Disagreements in the quality assessment were resolved through discussion.

Data were extracted by two independent researchers (Z-NH and LH) and entered into an EXCEL file including author, published time, country, study design, BMI, sample size, gender, history of smoking, American Society of Anesthesiologists (ASA) status, histological type, preoperative albumin, preoperative comorbidity (including hypertension, diabetes mellitus, obstructive lung disease), neoadjuvant therapy, tumor location, pathological tumor category, TNM stage, anastomotic method (stapler/hand sewn), route of gastric conduit, operation type, operative time, intraoperative blood loss, AL number, other complications, postoperative hospital stay, and hospital cost.

We used the inconsistency statistic (I²) to evaluate the extent of heterogeneity. Relative risk was used to evaluate the binary variable and the weighted mean difference for the continuous variable. An I² value greater than 50% was considered to indicate substantial heterogeneity. A fixed model was used when I² <50%, and a random model was used when I² ≧50%. A 2-sided test at the 5% level was defined as indicating statistical significance. We only calculated the existing data and did not fill in the missing data. Subgroup and sensitivity analyses were conducted to find the heterogeneity source. Publication bias would be assessed by funnel plots, Begg’s test, and Egger’s tests. Statistical analysis was conducted with Stata version 15 (Stata Corp, College Station, TX, USA), and Revman 5.4.

Electronic database search results are available in the PRISMA flow diagram ( Figure 1 ). From an initial total of 501 studies, 42 underwent full-text review, with 9 studies being included in our analysis (18–26). Many related publications focus on the flow dynamics of ICG perfusion.

A total of 1,162 patients from 4 countries were included in our analysis, of whom 419 patients underwent ICG. Although most of the publications were retrospective, there was only 1 prospective study. The sample size ranged from 40 to 285. There was some heterogeneity regarding the ICG dose, near-infrared system, operation type, and ICG intervention time. The ICG and near-infrared system were varied, which makes it difficult for further analysis. Detailed information about the included studies can be found in Table 1 .

Five studies were assessed to be of good quality based on the MINORS, with scores of 19 or more. The other four studies were assessed to be of moderate quality, with scores of 13 or more ( Table 2 ).

The age, body mean index, history of smoking, pathological type, diabetes mellitus, cardiovascular disease, obstructive lung disease, tumor location, pathological tumor category (T1–2/T3–4), TNM stage (I–II/III–IV), anastomotic procedure using a stapler, and gastric conduit through the posterior mediastinal were comparable in the ICG and control groups. However, the ICG group was lower in ASA I–II status (RR of 0.63, 95% CI 0.4–0.99), neoadjuvant therapy (RR of 0.59, 95% CI 0.41–0.85), and preoperative albumin (WMD −0.14 g/L, 95% CI −0.28–0.002). The ICG group underwent more thoracoscopy (RR of 2.73, 95% CI 1.40–5.37) and laparoscopy (RR of 2.17, 95% CI 1.18–3.96) ( Table 1 ).

The short-term clinical outcomes are summarized in Table 3 . The operation time and intraoperative blood loss were comparable in the ICG and control groups. There was no significant difference regarding overall postoperative mortality, reoperation rate, arrhythmia rate, vocal cord paralysis rate, pneumonia rate, or surgical wound infection. However, the ICG group had a shorter postoperative stay, with a 2.66-day reduction (WMD, 95% CI −3.77–1.55, p = 0.000).

The overall AL rate was 17.6% (n = 131) in the control group and 4.5% (n = 19) in the ICG group. The overall RR for AL was 0.29 (95% CI 0.18–0.47, p = 0.000), suggesting that ICG was associated with a statistically significant decrease in rates of AL ( Figure 2 ). The sensitivity analysis results did not indicate any publications with obvious heterogeneity ( Supplementary 1 ). Subgroup analysis was conducted by anastomosis site, ICG group sample size, and intervention time. The application of ICG in cervical anastomosis significantly reduced the incidence of AL (RR of 0.31, 95% CI 0.18–0.52), but the application of ICG fluorescence in intrathoracic anastomosis did not significantly reduce the incidence of AL rate (RR of 0.35, 95% CI 0.09–1.43) ( Figure 3 ). Regarding the sample size in the ICG group, publications with a sample size of >50 had a lower RR of 0.24 (95% CI 0.12–0.48). However, publications with a sample size of <50 still had a statistical RR of 0.35 (95% CI 0.18–0.68) ( Figure 4 ). Regarding the intervention time of ICG, the application of ICG both before and after gastric construction showed a better RR of 0.25 (95% CI 0.07–0.89) ( Figure 5 ).

A funnel plot analysis based on AL was performed. The funnel plot was asymmetrical ( Figure 6A ), which suggested that smaller studies favoring control had been omitted. Thus, we further conducted the Egger’s test (p = 0.65) ( Figure 6B ) and Begg’s test (p = 0.47) ( Figure 6C ), which both indicated no potential publication bias. The trim and fill test was stable without any trim or fill ( Figure 6D ).

Despite the evidence-based findings, there still exist some limitations in our study, Methodological limitations include: (1) Only one study was prospective, and no randomized control trial (RCT) studies were included in meta-analysis. Based on the GRADE scales, the evidence is low. Thus, a well-designed RCT is necessary for further confirm the effect of ICG fluorescence. We would like to update this meta-analysis when there are more reliable studies, especially RCTs. (2) There were some minor differences in baseline data between the ICG group and control group, which may cause potential influence on the incidence of AL. (3) Most studies have not reported AL grading based on the Clavien-Dindo classification. Whether ICG fluorescence could prevent patients from severe AL is unclear. (4) The number of included publications on Ivor-Lewis (conducting intrathoracic anastomosis) is limited, more high-quality publications are necessary to confirm the usefulness of ICG fluorescence in intrathoracic anastomosis. (5) Few studies used ICG quantification to measure the blood flow speed, and most studies ignored the importance of venous congestion. How to reduce the AL rate in patients with poor perfusion is really a question (34). Recently, Takeda FR et.al reported the supercharged cervical anastomosis for esophagectomy procedure may reduce the occurrence of anastomotic leakage and improve perfusion in the anastomotic area via vein and arterial micro-anastomoses based on the ICG quantification technique (35).

The application of ICG fluorescence could effectively reduce the incidence of anastomotic leak and shorten postoperative hospital stay for patients undergoing cervical anastomosis, but not for patients undergoing intrathoracic anastomosis. The application of ICG fluorescence before and after gastric management can better prevent patients from AL. This prevention works both in large (>50 cases) and small (<50 cases) medical centers.