Introduction

Esophagectomy, lymphadenectomy and restoration of the gastrointestinal continuity via gastric conduit reconstruction represent the gold standard treatment for esophageal malignancies [1]. Anastomotic leak (AL) remains the most feared complication, potentially occurring in up to 40% of patients [2, 3]. Increase in mortality, prolonged hospital stay, delayed oral feeding, risk of reintervention, increased risk of recurrence, and decrease of overall/disease-free survival have been reported to be associated with AL [4, 5]. Anastomotic stricture (AS) may occur in up to 30% of patients and may require endoscopic dilation with a negative effect of postoperative recovery, nutritional status, and quality of life [6].

Gastric ischemic conditioning (GIC) prior to esophagectomy may be an option to improve blood flow in the tubulized stomach and thus decrease the risk of ischemia and subsequent development of AL and AS [7, 8]. Technically, the partial devascularization of the stomach may determine a progressive vascular adaptation with improvement of the submucosal vascular network at the anastomotic site [9, 10]. Previous animal studies described improvement of fundic perfusion, while studies on human suggested a conceivable encouraging effect of preoperative GIC with reduced risk of AL and AS [11,12,13,14,15,16]. Previous meta-analyses have been published with conflicting results, because methodological flaws due to inclusion of single arm studies, search strategies incongruities, inclusion of duplicates with patient overlap, and substantial between-study heterogeneity [17,18,19,20]. Further, there is no consensus on the optimal technique and timing from GIC to esophagectomy.

The aim of our present work was to conduct an updated systematic review to evaluate the effect of GIC on anastomotic complications after esophagectomy.

Materials and methods

We conducted this study according to the Preferred Reporting Items For Systematic Reviews and Meta-Analyses (PRISMA) statement and MOOSE guidelines [21, 22]. All authors of GIC International Collaborative Group agreed on the study protocol.

Institutional review board approval was not required. PubMed, MEDLINE, Scopus, Web of Science, Cochrane Central Library, and ClinicalTrials.gov were used [23]. The last date of search was the March 31st, 2023. A combination of the following MeSH terms (Medical Subject Headings) was used (“esophagectomy” (tiab), OR “esophagectomies” (tiab), OR, “esophagogastric” (tiab), OR “esoph*”(tiab)) AND (“anastomosis” (tiab), OR “suture” (tiab)) AND (“ischemic conditioning” (tiab), OR “ischemic preconditioning” (tiab)) AND (“outcomes” (tiab), OR “complication” (tiab)) AND (“leak” (tiab), OR “leakage” (tiab)) AND (“stricture” (tiab), OR “stenosis” (tiab)). All titles were evaluated and suitable abstracts extracted. The search was completed by consulting the references of each article. The study protocol was registered at the PROSPERO (International prospective register of systematic reviews) (Registration Number: CRD42023423153).

Eligibility criteria

Inclusion criteria: (a) cohort studies and randomized controlled trials (RCTs) comparing outcomes for GIC vs. no GIC among adult patients (> 18 year old) undergoing elective esophagectomy; (b) English written; (c) when two or more papers were published by the same institution, study group, or used the same data-set, articles with the longest follow-up or the largest sample size; (d) in case of duplicate studies with accumulating numbers of patients only the most complete reports were included for quantitative analysis. Exclusion criteria: (a) not English-written; (b) no clear outcome distinction between GIC vs. no GIC; (d) articles with less than five patients per study arm.

Data extraction

The following data were collected: authors, year of publication, country, study design, number of patients, sex, age, body mass index (BMI), American Society of Anesthesiologists (ASA) physical status, comorbidities, surgical indication, tumor characteristics, histological type, tumor location, cancer stage, neoadjuvant chemoradiotherapy, type of GIC (angioembolization vs. surgery), target vessels, timing from GIC to esophagectomy and postoperative outcomes. All data were computed independently by four investigators (AA, DB, GB, LB) and compared at the end of the reviewing process to determined discrepancies.

Outcomes

Primary outcomes were postoperative AL, AS, and gastric conduit necrosis (GCN). Secondary outcomes were pulmonary complications, overall complications, operative time (OT) (minutes), intensive care unit (ICU) length of stay, hospital length of stay (HLOS) (days) and 30-day mortality. AL was defined as evidence of contrast extravasation at postoperative swallow study and/or CT-scan, or endoscopic visualization of anastomotic dehiscence/fistula, or visible loss of digestive secretions through surgical drains combined with clinical signs. AS was defined based on the need for endoscopic dilatation up to 6 months after the operation. GCN was defined as severe ischemia of the upper gastric conduit necessitating resection and diversion or refashioning of the anastomosis.

Quality assessment

Three authors (AA, AS, LC) independently assessed the methodological quality of included studies. The ROBINS-I tool was used for observational studies [24]. The following domains were considered: confounding bias, selection bias, classification bias, intervention bias, missing data bias, outcomes measurement bias, and reporting bias. Each domain is evaluated with one of the following: “yes”, “probably yes”, “probably no”, or “no”. The categories of judgement for each study are low, moderate, serious, and critical risk of bias. The methodological quality of Randomized Controlled Trials (RCTs) was appraised with the Cochrane risk of bias tool [25]. This tool evaluates the following criteria: (1) method of randomization; (2) allocation concealment; (3) baseline comparability of study groups; and (4) blinding and completeness of follow-up. Trials were graded as follows: A = adequate, B = unclear, and C = inadequate on each criterion. Thus, each RCT was graded as having low, moderate, or high risk of bias. Disagreements were solved by discussion. We used the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) tool to assess the quality of the body of evidence across studies [26].

Statistical analysis

The results of the systematic review were summarized quantitatively into frequentist random effect meta-analysis of pooled Risk Ratio (RR) and standardized mean difference (SMD). An inverse-variance method and DerSimonian–Laird estimator for the variance of the true effect size (τ2) were performed [27]. Heterogeneity among studies was evaluated by the I2 index and Cochran’s Q test [28]. Statistical heterogeneity was considered low, moderate, and high for I2 values of 25, 50, and 75%, respectively, and significant when p < 0.10 [29, 30]. The Wald-type 95% confidence interval (CI) was computed for pooled measurements; otherwise, the 95% CI for the I2 index was calculated according to Higgins and Thompson [31, 32]. The prediction interval for the treatment effect of a new study was calculated according to Borestein [28]. As the sample size was not the same in all studies, we performed a sensitivity analysis by excluding one study each time and rerunning the analysis to verify the robustness of the overall results. The publication bias was also investigated with the Trim and Fill funnel plot and Egger test. A two-sided p value was considered statistically significant when p < 0.05. All analyses and figures were carried out using the R software program, version 3.2.2 [33].

Results

Systematic review

The PRISMA flowchart is reported in Fig. 1. Overall, 1807 publications were identified. After duplicates removal, 1106 titles were screened. Overall, 291 abstracts were reviewed, while 58 articles were found possibly relevant for full-text assessment. After full-text evaluation, 14 studies meet the inclusion/exclusion criteria and were included in the quantitative synthesis. Thirteen studies were retrospective in design, and one study was a RCT. The quality of observational studies and RCTs is reported in Supplementary Table 1 and Supplementary Fig. 1.

Fig. 1
figure 1

The preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram

Full size image

Overall, 1760 patients were included (Table 1). Of those, 732 (41.6%) underwent GIC, while 1028 (58.4%) underwent one-step esophagectomy. The age of the patient population ranged from 24 to 84 years, the BMI ranged from 15.6 to 46 kg/m2 and the majority (82.1%) were males. Tumor location was reported in three studies. Tumor histology was specified in 8 studies; adenocarcinoma and squamous cell carcinoma were diagnosed in 75% and 25% of patients, respectively. Pathological tumor staging according to the 7th and 8th edition of the American Joint Committee on Cancer (AJCC) and Japanese gastric cancer classification (JGCC) was detailed in 5 studies (Stage 0–I: 21.1%, Stage II: 29.7%; Stage III: 49%, and Stage IV: 0.2%). Open, hybrid or totally minimally invasive esophagectomy were performed depending on operating surgeon preference and expertise. A cervical or intrathoracic anastomosis was performed according to tumor location, operating surgeon preferences, and oncological principles. The use of neoadjuvant chemoradiation therapy was heterogeneously reported in 11 studies (i.e. protocols, regimens, etc.). The extent of lymphadenectomy (2-filed and 3-field) was specified in 5 studies and varied depending on surgeon expertise and tumor clinical stage/location.

Table 1 Demographic, clinical, and operative data for patients undergoing gastric ischemic conditioning (GIC) before esophagectomy and one stage esophagectomy (C, control)
Full size table

Laparoscopic GIC was performed in 593 (81%) and angioembolization in 139 (19%) patients. Multi-vessel GIC with ligation of the left gastric artery, short gastric vessels, and right gastric artery or single vessel GIC with selective ligation of the left gastric artery were performed in 442 (60.3%) and 290 (39.7%) patients, respectively. The time from GIC to esophagectomy varied from 3 to 196 days.

Meta-analysis—primary outcomes

AL was reported in 14 studies (1760 patients). The cumulative incidence of AL was reduced for GIC vs. no GIC (8.8% vs. 14.4%). Compared to no GIC, GIC was associated with a significantly reduced AL risk (RR = 0.63; 95% CI 0.47–0.86; p < 0.01) (Fig. 2A). The prediction lower and upper limits were 0.45 and 0.89, respectively. The heterogeneity was zero (I2 = 0.0%, 95% CI 0.0–51%; p < 0.01) and τ2 = 0.01. The Funnel plot (Fig. 2B) and the Egger test (p = 0.49) did not show evidence of publication bias.

Fig. 2
figure 2

Anastomotic leak. Forrest (A) and Funnel (B) plot. RR risk ratio, 95% CI confidence interval

Full size image

AS was reported in 13 studies (1722 patients). The cumulative incidence of AS was reduced for GIC vs. no GIC (4.8% vs. 18.8%). Compared to no GIC, GIC was associated with a significantly reduced AS risk (RR = 0.51; 95% CI 0.29–0.91; p = 0.02) (Fig. 3A). The prediction lower and upper limits were 0.16 and 1.63, respectively. The heterogeneity was low (I2 = 20%, 95% CI 0.0–58%; p = 0.24) and τ2 = 0.18. The Funnel plot (Fig. 3B) and the Egger test (p = 0.04) showed that publication bias could not be excluded.

Fig. 3
figure 3

Anastomotic stricture. Forrest (A) and Funnel (B) plot. RR risk ratio, 95% CI confidence interval

Full size image

GCN was reported in 11 studies (1271 patients). The cumulative incidence of GCN was reduced for GIC vs. no GIC (0.2% vs. 1.9%). No differences were found for GIC vs. no GIC (RR = 0.56; 95% CI 0.19–1.61; p = 0.28) (Fig. 4A). The prediction lower and upper limits were 0.16 and 1.90, respectively. The heterogeneity was zero (I2 = 0.0%, 95% CI 0.0–0.0%; p = 1.00) and τ2 = 0.0. The Funnel plot (Fig. 4B) and the Egger test (p = 0.04) showed that publication bias could not be excluded. For all primary outcomes, the sensitivity analysis showed the robustness of findings in terms of point estimation, relative confidence intervals, and heterogeneity.

Fig. 4
figure 4

Gastric conduit necrosis. Forrest (A) and Funnel (B) plot. RR risk ratio, 95% CI confidence interval

Full size image

Meta-analysis—secondary outcomes

Pulmonary complications (RR = 1.09; 95% CI 0.83–1.44; p = 0.99; I2 = 0.0%), overall complications (RR = 0.87; 95% CI 0.73–1.04; p = 0.19; I2 = 28%), OT (SMD − 0.58; 95% CI − 1.12, 0.03; p = 0.07; I2 = 78%), ICU length of stay (SMD − 0.76; 95% CI − 2.15, 0.63; p = 0.29; I2 = 93%), HLOS (SMD 0.66; 95% CI − 0.10, 1.43; p = 0.09; I2 = 90%), and 30-day mortality (RR = 0.69; 95% CI 0.38–1.25; p = 0.22; I2 = 0.0%) were similar for GIC vs. no GIC. Using the GRADE tool, we rated the quality of evidence supporting each outcome as low-moderate mainly because of limitations in study design (Supplementary Table 2).

Discussion

This study shows that GIC prior to esophagectomy seems to be associated with a reduced RR of AL and AS with a trend toward a reduced risk for GCN. Pulmonary complications, overall complications OT, ICU length of stay, HLOS, and 30-day mortality seem similar between treatments.

The esophagogastric anastomosis is the most challenging part of esophagectomy and severe related complications such as AL and AS have been described. Factors that potentially contribute to anastomotic failure are tension, technique, malnutrition, patient comorbidities, and inadequate blood supply of the gastric conduit [49, 50]. Patients’ vascular status assessed by amount of calcifications in the thoracic aorta or stenosis of the celiac axis may represent an objective risk factor for post-esophagectomy AL. Therefore, optimization of gastric conduit perfusion and maintenance of an efficient submucosal vascular network at the anastomotic site is desirable. Physiologic gastric perfusion is supported by large capillaries, distributed perpendicular to both gastric curvatures, and by small capillary branches that run parallel to both curvatures thus forming a complex vascular network [51]. Ligation of the left gastric artery, left gastroepiploic artery, and short gastric vessels determine a rearrangement of this pattern. Therefore, perfusion dynamics in the gastric conduit occur through the right gastroepiploic artery at the greater curvature, while the tip of the conduit relies on small longitudinally oriented collateral connections. Occasionally, this vascular arcade may be prematurely interrupted. The knowledge of these anatomical key points is crucial for the construction of the gastric conduit to prevent ischemia at the anastomotic site [52]. With this aim, GIC prior to esophagectomy has been proposed to allow progressive vascular adaptation and consequent improvement of the submucosal vascular network of the gastric substitute [38,39,40,41,42,43,44,45,46,47,48].

We found that GIC is associated with a significantly reduced RR for AL compared to no GIC. The related heterogeneity was 0.0%. Jogiat et al. [20] also reported reduced odds for AL after ischemic conditioning (OR = 0.67; p = 0.03), whereas Kamarajah et al. [18] did not show any difference between GIC and no GIC (OR = 0.79; p = 0.3). Furthermore, GIC has been suggested to be associated with a tendency toward minimized leak severity and less need for redo surgery in case of leak occurrence. Specifically, Heger et al. [17] found that AL after GIC tended to be less severe and more likely managed conservatively or with endoscopic stenting, thus reducing the risk for reintervention (25% vs. 69%). Also, Schroder et al. [38] concluded that GIC led to more endoscopic stenting in case of AL rather than reintervention. Similarly, GIC was associated with a significantly reduced risk for AS. The related heterogeneity was low (I2 = 20%). Our result is different from what reported by Kamarajah et al. (OR = 0.74; p = 0.5), and Heger and colleagues [17] (OR = 1.1; p = 0.76), who concluded that there were no differences in terms of AS but is similar to what reported by Jogiat et al. [20] who supported reduced odds for AS (OR = 0.48; p = 0.005) in favor of GIC. The likely explanation for a reduced AL is a better vascular/oxygen supply to the anastomotic site induced by preoperative GIC. This is indirectly supported by previous studies using Doppler flowmetry, fluorescence microscopy, hyperspectral imaging, and angiography [7, 12, 13, 53]. Histological changes with neoangiogenesis, increased microvessels count, and vessel hypertrophy have also been reported [44]. Similarly, a possible explanation for reduced AS may be an increased preservation of the muscularis propria layer and decreased collagen deposition, with reduced fibrosis at the anastomotic site [19]. Despite the low heterogeneity, our results should be carefully interpreted, because the potential effect of confounders associated with an increased risk for AL and AS such as anastomosis location (thoracic vs. cervical), anastomotic technique (linear vs. circular stapled, omental wrapping), timing of GIC (< 14 days vs. > 14 days), technique of GIC (single vs. multi-vessel GIC, and laparoscopy vs. embolization), blood flow assessment with indocyanine green, patient age and comorbidities, steroid use, patient body morphometrics, past medical history, intraoperative interventions (i.e. blood transfusion, intraoperative hypotension, etc.), different anaesthesia protocols, neoadjuvant treatment, and tumor characteristics (i.e. grading, size, location, R0 status) [49]. Finally, despite the lack of statistical significance, a trend toward clinically reduced risk for GCN was observed for preoperative GIC. This is similar to what reported by Kamarajah and colleagues [18] who found a significantly reduced risk for conduit necrosis after ischemic conditioning (OR = 0.29; p = 0.013). Looking at secondary outcomes, no differences were noted for pulmonary complications, overall complications, OT, ICU length of stay, HLOS, and 30-day mortality.

Almost 80% of included patients in our study underwent laparoscopic GIC. Possible disadvantages of this approach are increased difficulty in dissection and lymphadenectomy at the time of esophagectomy because of adhesions. Also, use of staplers for en bloc ligation of the left gastric artery pedicle may be associated with significant scarring around the staple line. Furthermore, laparoscopic GIC represents a supplementary surgical intervention with increased costs, risk for GIC-related complications (i.e., bleeding requiring transfusions), and need for general anaesthesia. There is also risk of arterial injury and conversion to open, but this has not yet been described. In contrast, GIC via embolization may be regarded as a less invasive technique, although splenic infarction/necrosis, necrotizing cholecystitis, distal pancreatitis, and bleeding at the femoral access have been labeled [10, 35, 54, 55]. To date, a precise incidence of complications following both surgical and radiological GIC cannot be outlined, because the majority of enrolled articles are focused on the clinical advantage that GIC may have on the esophagogastric anastomosis during esophagectomy. Besides, further studies looking at the cost-analysis of GIC are needed.

There are sparse data describing the optimal timing between GIC and esophagectomy. It has been theorized that waiting > 2 weeks may be associated with better outcomes compared to a shorter period (< 2 weeks) [19]. Hypothetically, waiting longer intervals may be associated with improved neoangiogenesis. Furthermore, it remains unclear which vessels must be ligated to allow an adequate fundic revascularization [20]. Some authors advocate dividing only the left gastric artery while other suggested a multi-vessel approach. Unfortunately, a formal quantitative analysis to assess the ideal timing between GIC and esophagectomy and the potential benefits of a multi-vessel approach was not feasible. Opponents of GIC may argue that longer overall operative times, increased risk for GIC-related morbidity, and increased health-care costs combined with limited advantages do not justify the utilization of this strategy. Therefore, it should be considered that the reduced AL and AS risk may mitigate the initial expenses with ultimate global cost effectiveness. Furthermore, not all the patients probably will benefit from GIC prior to esophagectomy, but it is likely that some high-risk individuals could be good candidates [56,57,– 58]. More specifically, patients with poor preoperative vascular status, major calcification of the thoracic aorta, and stenosis of the celiac axis could hypothetically benefit from GIC before esophagectomy [58].

Notably, both AL and AS may be also influenced by surgeon proficiency, learning curves, structured training/mentorship programs and hospital volumes [59,60,61]. It has been shown that case-load centralization in high-volume centres significantly reduce mortality and may improve outcomes [62]. Based on these considerations, it should be considered that AL and AS may not entirely reflect the effect of preoperative GIC but are also influenced by surgeon experience. Finally, the introduction of new technologies such as fluorescence imaging with indocyanine green to assess the gastric conduit perfusion/anastomosis may hypothetically further improve outcomes.

Our study suffers from the typical limitations of a meta-analysis including observational studies, i.e., lack of inclusion criteria defined a priori, lack of homogenous surgical approach, and lack of globally defined postoperative management protocols. Second, AL and AS were not uniformly defined and classified among included studies. Third, the different techniques for GIC (laparoscopy vs. embolization), number of sealed vessels, and timing prior to esophagectomy were diverse and may potentially affect outcomes. Fourth, surgeon experience and volume were not measured with a potential bias and influence on outcomes. Lastly, the effect of GIC in the setting of neoadjuvant treatments (CROSS vs. FLOT) in locally advanced disease is still unclear and need to be defined. Specifically, GIC may be accomplished during staging laparoscopy prior neoadjuvant treatment therefore the interval time for resection may be delayed up to 5–7 weeks after the end of treatment while delivery of chemotherapeutic agents to junctional tumors after ischemic conditioning is unknown.

Conclusions

GIC prior to esophagectomy seems associated with a reduced risk for AL and AS. Further studies are necessary to identify the subset of patients who can benefit from this procedure, the optimal technique, and the timing of GIC prior to esophagectomy.