Anastomotic leaks after esophagectomy are associated with an increased mortality, prolonged hospital stay, and have been demonstrated to lower overall and progression-free survival [1]. Despite advancement in operative technique and perioperative care, the incidence of anastomotic leaks remains high and ranges from 10 to 40% [2,3,4]. Ischemia to the gastric conduit is thought to play a pivotal role in the etiology of anastomotic leaks, but strategies to overcome this problem remain elusive.

In animal trials published two decades ago, partial devascularization of the stomach improved perfusion of the proximal stomach [5]. Evidence from these trials informed future studies which investigated the effect of this form of gastric ischemic conditioning (GIC) in humans [6]. The general approach involves selective ligation of the blood supply to the stomach, namely the left gastric and short gastric arteries, either through embolization or surgery, with preservation of the right gastroepiploic artery, followed by an interval esophagectomy.

Proponents of GIC have previously reported the positive results on anastomotic leak rates from the retrospective data evidenced in this review. However, there are notable drawbacks that have been raised among the literature. A surgical ligation may incur its own complications and could increase the difficulty of subsequent esophagectomy by generating post-operative adhesions. At the time of resection, posterior dissection and lymphadenectomy may be more challenging which could compromise the overall quality of the oncologic resection. Ischemic conditioning through angiographic means is also not without risk, and there have been reports of splenic infarction and pancreatitis [6, 7]. Currently, there remains no consensus on the type of ischemic conditioning, the duration of delay prior to esophagectomy, or the potential benefit of the intervention [7,8,9,10].

The aim of our present work was to conduct a systematic review to evaluate the current literature on GIC, with our primary outcome serving as incidence of anastomotic leaks and our secondary outcomes being strictures, ischemia, and overall survival.

Materials and methods

Search strategy

This systematic review was conducted with adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A systematic search of MEDLINE, EMBASE, Scopus, Web of Science, and Cochrane Library was performed on February 5th, 2020 by a university librarian (JK) after selection of key terms with the research team. The searches included a combination of free text and controlled vocabulary terms related to ischemic conditioning and esophagectomies. Search terms included “Ischemic Preconditioning OR [variations of the terms]” AND “Esophagectomy OR [variations of the term]” (Supplementary Table 1). In addition to finding literature in bibliographic databases, the research team also reviewed the first 200 results in Google Scholar and the bibliography lists of included studies. A search of bibliographies in seven of the current systematic reviews on the topic was also performed to ensure no studies were overlooked.

Inclusion and exclusion criteria

Inclusion criteria included human participants undergoing esophagectomy with gastric conduit reconstruction, age ≥ 18, greater than five patients, GIC (any method) performed prior to esophagectomy, and studies published in English. Exclusion criteria included conference abstracts with insufficient data, duplicate studies, and case reports. One primary researcher evaluated titles and abstracts (UJ), with potentially eligible studies undergoing a full-text review by two primary researchers (UJ, WS). Disagreement in study inclusion or exclusion criteria between the two primary researchers was resolved by consensus.

Data extraction

For studies meeting inclusion criteria, two primary researchers (UJ, WS) extracted data on study characteristics. These included author, year of publication, study design, n value for ischemic conditioning, n value for control (if applicable), mean baseline patient demographics, method of GIC, time to GIC, and the aforementioned study outcomes. Baseline patient characteristics included age, sex, indication for surgery, neoadjuvant therapy status, and smoking status. Oncologic data extracted included type of cancer, stage of cancer, and location of cancer (proximal, middle, distal). Comparison of pooled data of all studies included in the review are provided in Supplementary Tables 4–6.

Study outcomes

The aim of our study was to determine the impact of GIC on post-operative complications, such as anastomotic leaks, strictures, and ischemia. The primary outcome measure was incidence of anastomotic leaks. The secondary outcome measures were gastric conduit ischemia, anastomotic strictures, and overall survival.

Assessment of methodological quality

Methodological quality of studies was assessed by two independent reviewers (UJ, WS). Methodological quality was formally assessed using the methodological index for non-randomized studies (MINORS) criteria for non-randomized trials. MINORS is an externally validated 12 item index utilized to assess the methodological quality of comparative and non-comparative cohort studies [11].

Statistical analysis

Descriptive categorical data were expressed as percentages and continuous data were expressed as weighted means where appropriate. Where medians were reported, values were converted to means using the method reported by Hozo et al. [12] Meta-analysis was conducted to evaluate differences in anastomotic leaks, anastomotic strictures, conduit ischemia, and overall survival for patients receiving GIC compared to those receiving standard of care. Where direct comparisons and meta-analysis could not be conducted, pooled proportions were used to describe outcomes. Estimated effects were calculated using RevMan 5.4.1 software with a Mantel–Haenszel fixed-effects model. Heterogeneity was quantified by the I2 statistic: (1) low = 25%; (2) moderate = 50%; and (3) high = 75%. Tests for statistical significance were two-tailed with significant p-values defined as < 0.05. Statistical analysis of pooled data was not performed given lack of comparator groups, and results are provided for gross comparison.

Results

Study selection

A total of 1712 results were retrieved and 1071 unique records remained for the title and abstract screening phase (Fig. 1). Of the abstracts meeting the initial screening criteria, 1008 were excluded either because they did not meet the formal inclusion criteria or met exclusion criteria for our study. This resulted in 63 studies undergoing full-text assessment for eligibility. A total of 23 studies met inclusion criteria, 21 of which were full-text articles and two of which were conference abstracts. Of the included manuscripts, 14 were retrospective studies and 9 were prospective. Two studies from the same institution met inclusion criteria given the significant lapse in time between the studies and lack of significant overlap in the sample sizes [13, 14]. No major methodological flaws were identified, and the results of the search are provided in the PRISMA flow diagram below (Fig. 1).

Fig. 1
figure 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram of included studies

Full size image

Study characteristics

A total of 2203 patients, the majority of which were male (n = 1515, 69.8%) with a mean weighted age of 63 ± 7.1 years, were included (Table 1). GIC was performed in 1178 (53.5%) patients with a mean time from GIC to surgery of 28 days ± 37.3. Of the included studies reporting GIC, 12 were comparator studies and 11 were observational, 17 utilized the laparoscopic method, while 6 utilized embolization (Table 2). Of the studies providing histologic data, there were 842 adenocarcinomas and 371 squamous cell carcinoma patients in the included studies (Supplementary Table 2). The most common stage recorded was stage III, comprising 319 (54%) patients and the most frequent location of tumor reported in studies providing this data were distal, comprising 92 (52%) patients. For those that did include such data, patients who underwent ischemic conditioning generally matched controls. Given the extensive variability and missing data among baseline patient's characteristic data and oncologic data across the included studies, these were not included in the sub-group analysis and are reported in the supplementary tables.

Table 1 Baseline patient characteristics of included studies
Full size table
Table 2 Characteristics of included studies
Full size table

Regarding neoadjuvant therapy, 15 studies contained patients undergoing neoadjuvant therapy prior to ischemic conditioning. Out of these, neoadjuvant chemoradiotherapy was reported in 13 studies. Of the included studies, seven contained data on staging, with stage three being the most common.

Laparoscopic ischemic conditioning was more common than angiographic ischemic conditioning (n = 1835 vs. n = 289, respectively). In the included studies, there was an equal distribution among patients who underwent ischemic conditioning for less than 7 days compared to those who underwent ischemic conditioning for greater than 7 days (n = 1059 vs. n = 1112, respectively). More patients underwent ischemic conditioning with two or more vessels ligated as compared to a single vessel (n = 1333 vs. n = 840, respectively).

Reduced odds of anastomotic leaks in patients with GIC

Anastomotic leaks were reported in 12 comparative studies, of which 9 reported laparoscopic GIC compared to three reporting angiographic GIC. Meta-analysis revealed a significant reduction in anastomotic leaks in patients undergoing GIC (OR 0.67; 95% CI 0.46–0.97; I2 = 5%; p = 0.03) compared to those undergoing esophagectomy alone (Fig. 2).

Fig. 2
figure 2

Odds of post-operative anastomotic leaks among patients undergoing gastric ischemic conditioning versus controls. GIC gastric ischemic conditioning

Full size image

No difference in odds of conduit ischemia between patients with GIC and patients without GIC

Conduit necrosis or ischemia was reported in 11 comparative studies, of which 8 reported laparoscopic GIC compared to three reporting angiographic GIC. No significant reduction in gastric conduit ischemia was noted in patients undergoing GIC (OR 0.40; 95% CI 0.13–1.23; I2 = 0%; p = 0.11) (Fig. 3).

Fig. 3
figure 3

Odds of conduit ischemia among patients undergoing gastric ischemic conditioning versus controls. GIC gastric ischemic conditioning

Full size image

Reduced odds of anastomotic strictures in patients with GIC

Anastomotic strictures were reported in 11 comparative studies, of which eight reported laparoscopic GIC compared to three reporting angiographic GIC. A significant reduction in anastomotic strictures was noted in patients undergoing GIC (OR 0.48; 95% CI 0.29–0.80; I2 = 65%; p = 0.005) (Fig. 4).

Fig. 4
figure 4

Odds of anastomotic strictures among patients undergoing gastric ischemic conditioning versus controls. GIC gastric ischemic conditioning

Full size image

No difference in odds of overall survival in patients with GIC and patients without GIC

Overall survival was reported in nine comparative studies, of which 6 reported laparoscopic GIC compared to three reporting angiographic GIC. No significant improvement in overall survival was noted (OR 0.54; 95% CI 0.29–1.02; I2 = 22%; p = 0.06) (Fig. 5).

Fig. 5
figure 5

Odds of overall survival among patients undergoing gastric ischemic conditioning versus controls. GIC gastric ischemic conditioning

Full size image

Description of outcomes using pooled proportions

To gain insight into outcomes related to the technique, duration of IC, and number of vessels ligated, pooled data were examined (Supplementary Tables 4–6). Laparoscopic ischemic conditioning had a higher pooled proportion of stricture formations than angiographic (2% vs. 0%). Similar rates of ischemia (1.0% vs. 1.3%) and anastomotic leaks (7.8% vs. 6.8%) were noted among the two techniques. GIC for less than 7 days was associated with higher rates of stricture formation (5.6% vs. 1.2%), while greater than 7 days was associated with higher rates of conduit ischemia (0.4% vs. 1.4%). The observed leak rate was similar among the two different time groups (7.6% vs. 7.3%). Anastomotic leak rates were higher in the single-vessel group (9.5% vs. 5.8%). Incidence of stricture formation was lower in the single-vessel group (1% vs. 4.3%), while conduit ischemia was similar (1.1% vs. 0.9%).

Heterogeneity analysis and risk of bias

Heterogeneity was classified as low for anastomotic leaks (I2 = 5%), low for conduit ischemia (I2 = 0), and low for overall survival (I2 = 22%). Heterogeneity was classified as moderate for anastomotic strictures (I2 = 65%).

Comment

To our knowledge, this is the most recent systematic review and meta-analysis on GIC for patients undergoing esophagectomy. We found that GIC is associated with a significant decrease in anastomotic strictures and anastomotic leaks and was not associated with conduit ischemia. Importantly, we show that patients receiving GIC had 37% lower odds of anastomotic leaks and 52% lower odds of anastomotic stricture than those who did not undergo GIC.

The concept of GIC has drawn increasing interest, with two systematic reviews and meta-analyses published on this topic in recent years. Our findings are consistent with Zhou et al., who have reported a reduction in anastomotic leaks and stricture formation in their recent 2020 meta-analysis [15]. The authors conducted a sub-group analysis on their data, stratifying by method of GIC and noted that both laparoscopic and angiographic GIC reduced anastomotic leaks and stricture formation. Of note, the study did include data from abstract records reported in conference proceedings and did not explore ischemia of the gastric conduit as an outcome. Another review by Kamarajah et al. reported no significant difference in anastomotic leaks or anastomotic strictures in their meta-analysis [16]. Their search strategy resulted in a marked lower number of initial results and there have been further full-text articles published on this topic included in our analysis which may explain the contradictory results reported in our review [17,18,19]. Of note, the authors did not search the Web of Science or Scopus databases which we included in our systematic search and this may explain the decreased number of studies. Our review of the outcomes by pooled proportion reported in the literature for technique of GIC, timing of GIC, and number of vessels selected is an additional novel aspect of our review. While these findings are not statistically significant, the trends among the literature provide valuable and novel insight regarding the overall distribution of outcomes and may serve as an impetus and guide the development of prospective trials.

Our findings are supported by the theories of a microcellular or physiological change induced by ischemic conditioning in the gastric conduit. Urschel et al. hypothesized that ischemic conditioning may provide neovascularization of the remaining vessels, improving blood supply to the gastric conduit and thus reducing the risk of anastomotic breakdown or stricturing [5]. Contrary to this, a recently published randomized control trial by Veeramootoo et al. examined Doppler fluximetry of the gastric conduit comparing patients undergoing ischemic conditioning versus patients who did not and concluded no significant difference in perfusion among the groups [20]. They did not report on clinical outcomes in their study and thus were not included in the meta-analyses of this review. Although severity of the anastomotic leak was not an outcome of interest in this review, Heger et al. demonstrated that anastomotic leaks in ischemic conditioning were more likely to be managed conservatively [21]. While the exact theories through which GIC may impart clinical benefit are not clear, the promising findings of our work highlight the need for further studies which aim to elucidate its underlying mechanisms.

The utility of angiography to perform GIC was first discovered in Japan in the 1990s. Studies published by Akiyama et al. first demonstrated the efficacy of this procedure and reported positive results in terms of reduction of anastomotic leaks post-esophagectomy [5]. Laparoscopic ischemic conditioning was subsequently popularized and has been demonstrated as safe, with post-operative complications occurring rarely. We were unable to conduct meta-analysis comparing these two techniques, as none of the studies obtained provided direct comparison of these two techniques. However, comparison of outcomes by pooled proportion regarding these two techniques depicts higher proportion of leaks and strictures occurring in the laparoscopic group. One reason why this may have occurred is due to the physical manipulation of tissue, which occurs during laparoscopic surgery. This may induce microtrauma to the gastric tissue, increasing its vulnerability to ischemic insult and subsequent leaks and strictures. Despite the less invasive nature of angiographic GIC, however, from our included studies, this technique frequently embolizes the distal splenic artery, whereas this is not performed in laparoscopic GIC. This is associated with reports of pancreatitis and splenic infarction in the literature and thus may explain why angiographic GIC is less frequently performed, despite being the less invasive modality [6, 7]. Laparoscopically, the most commonly selected arteries included the left gastric and short gastric arteries, with some studies including the right gastric artery, as well. In angiographic GIC, it is more common for all three arteries to be selected. The most preferred gastric vessels remain unclear, but the majority of studies have emphasized the selection of the left gastric and short gastric arteries. From the literature to date, it appears that the preferred mechanism of ischemic conditioning has yet to be elucidated, and further research should be directed within this area.

We observed that single-vessel GIC was associated with higher incidence of anastomotic strictures, but lower incidence of stricture formation, compared to multi-vessel GIC. The causal mechanism for why this may occur is unclear. However, presuming the notion that GIC does indeed promote neovascularization, it follows that multi-vessel GIC may elicit a more potent stimuli for neovascularization. This process ultimately increases the blood flow to the anastomosis, promotes healing, and reduces the risk of breakdown and leaks. This may explain the first finding of why single-vessel GIC was associated with higher rates of anastomotic leaks. The rational for why single-vessel GIC was associated with less stricture formation is more nuanced. It has been reported in the literature that stricture formation can be a consequence of relative ischemia to the anastomosis [22]. Perhaps the increased ischemia prior to neovascularization in the multi-vessel GIC group increases the risk of stricture formation. Ischemic conditioning for less than 7 days was associated with higher rates of leaks and strictures. This supports the theory that a greater interval period allows for higher amounts of neovascularization. These observations should be taken with the caveat that the data are obtained from non-standardized studies without direct comparator groups to truly investigate these findings and are subject to confounding variables.

There are a number of limitations to this review. The included studies contain very limited baseline characteristics regarding their study population, not providing smoking status, major cardiopulmonary co-morbidities, and ASA or CCI co-morbidity scores. No randomized control trials meeting the inclusion and exclusion criteria of our search were identified and our findings are largely based on retrospective data. The included studies encompass a broad timeline, from 1996 to 2020, and much has changed in the interim regarding treatment of esophageal cancer. With the publication of the CROSS trial in 2015, neoadjuvant therapy has become standard of care, and it is unclear how the administration of radiation and chemotherapeutic agents impacts ischemic conditioning [23]. Of note, 15 of the 23 studies did report neoadjuvant therapy in patients undergoing GIC, with 9 studies published after the CROSS trial. Furthermore, centers may be reluctant to endorse ischemic conditioning in the post-neoadjuvant, but pre-operative period, given the priority of ensuring an expedited recovery to proceed to esophagectomy in a timely fashion. Introducing an additional procedure in this interim period may ultimately delay esophagectomy or induce a complication precluding curative therapy. Thus, identification of patients who are at high risk for anastomotic complications may allow for the determination of a sub-group who may benefit most from GIC, understanding that this is a calculated risk weighing the benefit of GIC against its potential complications. Regarding our statistical analysis, we used a fixed-effects model, and heterogeneity was demonstrated to be low for anastomotic leaks, conduit ischemia, and overall survival through determination of the I2 statistic. However, there was moderate heterogeneity demonstrated in the meta-analysis of conduit strictures, and thus a random-effects model may have been more appropriate for evaluation of this outcome. We elected to pursue a fixed-effects model to preserve uniformity in the interpretation of our meta-analysis. Given that esophagectomy is most commonly performed for oncologic indications, an additional major limitation of the research is the near complete absence of oncology data among the included studies. Lastly, there still remains many unanswered questions regarding the optimal timing, technique, and number of vessels to ligate during GIC.

Despite the limitations of the current literature, the findings of decreased anastomotic leak and stricture rates obtained in this review have notable implications, particularly in a patient population that is high risk for post-operative complications. Ischemic conditioning of the gastric conduit may reduce rates of anastomotic leaks and stricture formation in patients undergoing esophagectomy. However, the sparsity of prospective data particularly in the form of a randomized control trials evaluating patient outcomes limits the reliability of this data. Furthermore, the potential benefit offered by ischemic conditioning must be balanced against the increased recovery time, operative concerns regarding adhesion genesis, tolerance of chemotherapeutic agents, and cost to both the patient and healthcare system, factors that were not accounted for in the reviewed studies. However, for patients with major risk factors for anastomotic leak, such as heart failure, hypertension, radiation, or renal insufficiency, ischemic conditioning may serve as an important therapeutic tool [3]. Future research on ischemic conditioning should thus be directed at determining the optimal timing, number of vessels ligated, and technique employed, as these remain unanswered questions.