Volume outcome relationship in postesophagectomy leak: a systematic review and meta-analysis

Abstract

Background:

Anastomotic leak after esophagectomy carries important short- and long-term sequelae. The authors conducted a systematic review and meta-analysis to determine its association with surgical volume.

Materials and methods:

A systematic literature review was performed to identify all studies reporting on anastomotic leak after esophagectomy. Studies with less than 100 cases were excluded. The primary outcome was postesophagectomy anastomotic leak, while secondary outcomes were operative mortality overall and after anastomotic leak. Pooled event rates (PER) were calculated, and the association with annual esophagectomy volume by center was investigated.

Results:

Of the 3932 retrieved articles, 472 were included (n=177 566 patients). The PER of anastomotic leak was 8.91% [95% CI=8.32; 9.53%]. The PER of early mortality overall and after an anastomotic leak was 2.49% [95% CI=2.27; 2.74] and 11.39% [95% CI=9.66; 13.39], respectively. Centers with less than 37 annual esophagectomies had a higher leak rate compared to those with greater than or equal to 37 annual esophagectomies (9.58% vs. 8.34%; P=0.040). On meta-regression, surgical volume was inversely associated with the PER of esophageal leak and of early mortality.

Conclusions:

The frequency of anastomotic leaks after esophagectomy, perioperative, and leak associated mortality are inversely associated with esophagectomy volume.

Introduction

HIGHLIGHTS

• Anastomotic leak is one of the most serious complications after esophagectomy, which results in significant postoperative morbidity and mortality.

• Centers performing more than 37 esophagectomies per year had a lower leak rate, while centers performing more than 10 esophagectomies per year had a lower overall early mortality.

• Underlying malignancy, and cervical anastomosis were associated with a higher proportion of postoperative leak.

• Thoracic anastomosis, and the annual number of esophagectomies per center were associated with a lower proportion of postoperative leak.

• The data that support the findings of this study are available from the corresponding author upon reasonable request.

Esophagectomy is an extensive procedure that remains a viable treatment option for several pathologies, including both benign and malignant processes.

Anastomotic leak is one of the most serious complications after esophagectomy, which results in significant postoperative morbidity and mortality. Some of the more obvious consequences related to an anastomotic leak are a longer length of hospital stay, readmissions, and the need for additional procedures. Although improved operative techniques and earlier recognition have decreased short-term mortality, it still remains an issue given that it can cause the swift onset of mediastinitis and septicemia^1,2. Furthermore, strictures can also develop and patients who recover may have a lower quality of life secondary to ongoing dysphagia and the need for frequent endoscopic interventions^2,3.

We conducted a systematic review and meta-analysis to determine the association of anastomotic leak after esophagectomy with annual surgical volume.

Methods

This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement⁴ and AMSTAR (Assessing the Methodological Quality of Systematic Reviews)⁵ (Supplementary Material), Supplemental Digital Content 1, http://links.lww.com/JS9/A256, Supplemental Digital Content 2, http://links.lww.com/JS9/A257.

Search strategy

A medical librarian performed a comprehensive literature search to identify all studies with the primary aim of assessing the pooled rate of esophageal leak after esophagectomy. Searches were run on 3 December 2021 in the following databases: Ovid MEDLINE® (In-Process & Other NonIndexed Citations and Ovid MEDLINE® 1946 to Present), Ovid EMBASE (1974 to present), and The Cochrane Library (Wiley). Search terms included all subject headings and associated keywords for ‘esophagectomy’ and ‘leak’. There were no publication date or article type restrictions on the search strategy. The full search strategy for Ovid MEDLINE is shown in Supplementary Table 1, Supplemental Digital Content 3, http://links.lww.com/JS9/A258.

Study selection and inclusion criteria

After search results were de-duplicated, three medically qualified reviewers (M.R., M.B., and J.M.) screened a total of 3932 citations. A fourth independent reviewer (M.G.) confirmed the adequacy of studies based on predefined inclusion and exclusion criteria for titles and abstracts. Articles considered for inclusion were full-text and human studies that evaluated postoperative leak after esophagectomy with online availability of data about leak cases and total esophagectomy cases.

Exclusion criteria were non-English articles, case series (including sample size of less than 100 patients), systematic reviews, editorials, pediatric studies, studies in which postesophagectomy leak rate was not mentioned separately, multicenter studies, and studies focused on gastric cancer, corrosive esophagitis, caustic stricture, atresia, or achalasia. In cases of overlapping population, we included the largest sample size. Reference lists of prior meta-analyses were also searched for potential inclusion of eligible studies (i.e. backward snowballing). Full-text articles of initially screened titles and abstracts were then retrieved for a second round of eligibility screening.

The PRISMA flowchart outlining the study selection process is available in Supplementary Figure 1, Supplemental Digital Content 4, http://links.lww.com/JS9/A259. The Cochrane Collaboration’s tool for assessing risk of bias in randomized trials and the Newcastle–Ottawa Quality Assessment Scale (NOS) for critical appraisal of eligible studies were used for randomized controlled trials (RCTs) and observational studies, respectively (Supplementary Table 2, Supplemental Digital Content 5, http://links.lww.com/JS9/A260)^6,7.

Outcomes

The primary outcome was postoperative anastomotic leak after esophagectomy as defined by the authors. Secondary outcomes were early mortality overall and after an anastomotic leak. Early mortality was considered as defined by the authors. The association of all the outcomes with the annual number of esophagectomies performed by the center was investigated.

Data extraction

The following variables were extracted: study name, publication year, study design, number of patients, mean age, BMI, percentage of sex, diabetes mellitus, smoking, radiation treatment, malignancy, thoracic anastomosis, cervical anastomosis, study period (years), annual esophagectomies per center, number of anastomotic leaks, early mortality, and early mortality in patients with esophageal leak.

Statistical analysis

Measurement data were reported as mean±SD. Pooled event rates (PER) with a 95% CI were calculated for all the outcomes.

Subgroup analysis was used to compare the PER of esophageal leak and early mortality based on annual surgical volume by center. The cutoff was determined based on the smallest number of annual cases associated with a significant difference using dummy variables with an ascending cutoff starting with the lowest annual volume (cases/year) and an incremental one-unit increase.

Meta-regression was used to assess the effect of age, BMI, sex, diabetes mellitus, smoking history, radiation treatment, malignancy, thoracic anastomosis, cervical anastomosis, study period (years), and annual esophagectomy volume on the PER of all the outcomes. Meta-regression was reported as regression coefficient (beta)± SD. A positive beta value corresponds to a higher incidence of the outcome with a higher incidence of the assessed variable, while a negative beta value corresponds to a lower incidence of the outcome with a higher incidence of the assessed variable.

The studies’ heterogeneity was assessed using the Cochran Q statistic and the I ² test. Significant heterogeneity (I ²>75%) for the primary outcome was investigated using a leave-one-out sensitivity analysis⁸. Egger’s regression test and a funnel plot were used to assess for potential publication bias⁹. If significant publication bias was found, Duval and Tweedie's trim-and-fill method was used to adjust for the possible bias. A random-effect model (inverse variance (DerSimonian and Laird) method) was used for the whole analysis. Hypothesis testing for equivalence was set at the two-tailed 0.05 level. All statistical analyses were performed using R^10,11 (version 3.3.3 R Project for Statistical Computing) within RStudio (0.99.489, http://www.rstudio.com).

Results

Characteristics of eligible studies and patients included

Among the 3932 de-duplicated searched articles, 472 articles (see references in the Appendix) met the inclusion criteria. A PRISMA flowchart, shown in Supplementary Figure 1, Supplemental Digital Content 4, http://links.lww.com/JS9/A259, provides further details on the exclusion/inclusion process. A summary of the Cochrane Collaboration’s tool for assessing risk of bias in randomized trials and the NOS for critical appraisal of included studies using the NOS is shown in Supplementary Table 2, Supplemental Digital Content 5, http://links.lww.com/JS9/A260.

A total of 177 566 patients were included (3403 from 16 RCTs, 13 431 from 45 prospective studies, and 160 732 from 411 retrospective studies), with baseline characteristics reported in Supplementary Table 3, Supplemental Digital Content 6, http://links.lww.com/JS9/A261. The mean sample size was 376.2 patients (range 100–4132), mean age was 62.55 years (range 39.00–74.00), and the mean BMI was 23.81 kg/m² (range 15.90–28.63). The percentage of females ranged from 0.00 to 79.60% and percentage of smokers ranged from 8.00 to 91.00% as shown in Supplementary Table 3, Supplemental Digital Content 6, http://links.lww.com/JS9/A261.

A total of 115 studies (24.36%) were from China, 91 studies (19.28%) from Japan, 78 studies (16.53%) from the United States, 38 studies (8.05%) from the United Kingdom, 26 studies (5.51%) from Germany, 12 studies from Hungary, 9 studies from France, and 3 studies from Sweden.

Meta-analysis

Overall, 16 092/177 566 (9.06%) patients experienced a postoperative leak (PER 8.91% [95% CI=8.32; 9.53]). Centers performing greater than or equal to 37 annual esophagectomies had lower esophageal leak rates compared to centers performing less than 37 annual esophagectomies [PER 8.34% (95% CI=7.54; 9.22) vs. 9.58% (95% CI=8.76; 10.46)], respectively, P-value for interaction = 0.043. (Fig. 1, Fig. 2). Cervical anastomosis was associated with higher leak rates [10.06% (8.99; 11.24)] compared to thoracic anastomosis [5.55% (4.49; 6.85), P-value for interaction <0.001] but lower early mortality from leak (6.22% vs. 19.78%, P for interaction <0.001) (Fig. 1). The incidence of anastomotic esophageal leak in different country subgroups is shown in Supplementary table 4, Supplemental Digital Content 7, http://links.lww.com/JS9/A262 (P for interaction <0.0001); Pakistan had the lowest leak incidence of 1.58% versus the highest in Thailand (35.91%). Studies from the US and Canada had leak incidence of 8.56% and 14.43%, respectively (Supplementary Table 4, Supplemental Digital Content 7, http://links.lww.com/JS9/A262).

Figure 1.

Outcomes summary. ¶ Cutoff was determined based on the smallest annual number of cases associated with a significant difference (Supplementary figure 3 for more details). ¶¶ P-value for subgroup difference (Early mortality from anastomotic leak vs. no-leak) <0.0001. † Obtained from studies exclusively reported 100% cervical or thoracic anastomosis.

Figure 2.

Determination of interaction P-value at different annual volume/center for postesophagectomy leak cutoff of annual esophagectomy cases per center.

Early mortality was 11.39% (95% CI=9.66; 13.39) in patients with a leak (Fig. 1). Pooled 30-day and 90-day mortality rates were 1.89% (95% CI=1.55; 2.30) and 2.43% (95% CI=1.89; 3.13), respectively. Threshold analysis revealed that centers that performed greater than or equal to 10 annual esophagectomies had lower early mortality rates when compared to centers that performed less than 10 annual esophagectomies [PER 2.45% (95% CI=2.23; 2.71) vs. 3.21% (95% CI=2.09; 4.90), respectively], P-value for the interaction=0.0097 (Fig. 3).

Figure 3.

Determination of interaction P-value at different annual volume/center for postesophagectomy early mortality cutoff of annual esophagectomy cases per center.

Sensitivity analysis using leave-one-out analysis for the primary outcome revealed the robustness of the results. However, a funnel plot of observed and imputed studies (trim-and-fill method) revealed the presence of publication bias (Egger’s intercept =−1.41±0.08, P<0.001) (Supplementary Figure 2, Supplemental Digital Content 10, http://links.lww.com/JS9/A265). The bias adjusted PER for postoperative leak was 14.07% (95% CI=13.14; 15.04). The publication bias adjusted results of the other outcomes are summarized in Supplementary Table 5, Supplemental Digital Content 8, http://links.lww.com/JS9/A263.

An analysis on the use of stents or endoscopy for anastomotic leaks did not show differences between low- and high-volume centers (P=0.776 for stent, P=0.5674 for endoscopy), Supplementary Table 6, Supplemental Digital Content 9, http://links.lww.com/JS9/A264.

Meta-regression

Underlying malignancy (Beta=0.012, P=0.034), and cervical anastomosis (Beta=0.006, P<0.0001) were associated with higher PER of postoperative leak, while thoracic anastomosis (Beta=−0.006, P<0.0001), and annual number of esophagectomies per center (Beta=−0.0008, P=0.018) were associated with lower PER of postoperative leak (Supplementary Figure 3 (A–B), Supplemental Digital Content 11, http://links.lww.com/JS9/A266 and Table 1). Difference in volume/outcome among different continents was shown in Figure 4 where it was evident among studies from China/Japan only (Beta=-0.001, P=0.014).

Table 1.

Meta-regression of different variables on different outcomes

Incidence of esophageal leak	Meta-regression results (Beta*±SD, P-value)
Age (mean)	0.0079±0.0101, P=0.4376
Female percent	−0.0017±0.0036, P= 0.6315
DM percent	0.0047±0.0075, P=0.5295
BMI (mean)	−0.0219±0.0218, P=0.3162
Smoking percent	0.0012±0.0033, P=0.7217
Radiation percent	0.0009±0.0019, P=0.6400
Malignancy percent	0.0125±0.0059, P =0.0343
Thoracic anastomosis percent	−0.0064±0.0010, P <0.0001
Cervical anastomosis percent	0.0065±0.0010, P <0.0001
Study period (years)	−0.0097±0.0064, P=0.1282
Annual esophagectomies per center	−0.0008±0.0003, P =0.0183
Early mortality among esophageal leak
Age (mean)	0.0279±0.0244, P=0.2542
Female percent	−0.0055±0.0089, P=0.5365
DM percent	−0.0170±0.0156, P=0.2749
BMI (mean)	0.0246±0.0799, P=0.7586
Smoking percent	−0.0003±0.0102, P=0.9799
Radiation percent	0.0074±0.0052, P=0.1546
Malignancy percent	−0.0042±0.0109, P=0.7010
Thoracic anastomosis percent	0.0160±0.0023, P <0.0001
Cervical anastomosis percent	−0.0160±0.0023, P <0.0001
Study period (years)	0.0275±0.0151, P=0.0681
Annual esophagectomies per center	−0.0012±0.0008, P=0.1277
Overall early mortality
Female percent	−0.0008±0.0047, P=0.8631
DM percent	−0.0081±0.0114, P=0.4777
BMI (mean)	0.0692±0.0362, P=0.0558
Smoking percent	−0.0130±0.0051, P =0.0107
Radiation percent	0.0066±0.0023, P =0.0044
Malignancy percent	0.0013±0.0073, P=0.8554
Annual esophagectomies per center	−0.0031±0.0005, P <0.0001
Total leak (%)	0.0184±0.0053, P =0.0005

Bold values are statistically significant variables.

^*Positive beta (regression coefficient) value corresponds to higher incidence of the outcome with higher incidence of the assessed variable while negative beta corresponds to lower incidence of the outcome with higher incidence of the assessed variable.

DM, diabetes mellitus.

Figure 4.

Meta-regression of annual esophagectomy cases per center on incidence of anastomotic leak (Volume outcome was evident among studies from China/Japan with lower leak among centers with high annual esophagectomy cases). Beta: is the regression coefficient. Positive beta reflects higher incidence of the outcome (leak) with lower incidence of the covariate (annual esophagectomy cases).

Annual number of esophagectomies (Beta=-0.003, P<0.0001) was inversely associated with early mortality, while percentage of leak was directly associated with it (Beta=0.018, P=0.0005) (Table 3, Supplemental Digital Content 13, http://links.lww.com/JS9/A268 and Supplementary Figure 4, Supplemental Digital Content 12, http://links.lww.com/JS9/A267).

Cervical anastomosis (Beta =−0.016, P<0.0001) was inversely associated with early mortality in patients with esophageal leak (Table 3, Supplemental Digital Content 13, http://links.lww.com/JS9/A268).

Discussion

Esophagectomy is an extensive procedure associated with high morbidity. It is important to better understand complications associated with higher early mortality and morbidity after this procedure to improve postoperative outcomes. In this meta-analysis, we looked at single center studies on anastomotic leak after esophagectomy, one of the most dreaded complications, and how this relates to short-term mortality while also taking center volume into consideration.

Anastomoses after esophagectomy follow the same general principles of healing as any other bowel anastomosis and are influenced by both local and systemic factors¹. Ideal conditions for anastomotic healing are absence of tension, good vascular supply for adequate oxygenation, and thorough anastomotic technique. Additionally, there is also a known difference in consequences of leak related to anastomotic location. While cervical leaks tend to be contained due to their location and known delayed presentation around the seventh day or later postoperatively, thoracic leaks are more prone to fulminant contamination due to negative intrapleural pressure, which makes their management more difficult and the chances of sepsis higher³. Situations that can lead to leak include the need to further mobilize the gastric conduit, increased anastomotic tension and higher chances of mucosal retraction leading to improper mucosa-to-mucosa apposition in more proximal locations³. These are the keys to prevent postesophagectomy anastomotic leak and are important to keep in mind when trying to understand patterns in its occurrence.

We found that centers performing more than 37 esophagectomies per year had a lower leak rate, while centers performing more than 10 esophagectomies per year had a lower overall early mortality. An inverse correlation between center case volume and improved patient outcome has been shown in prior series as well. The Leapfrog national expert panel specified the minimum hospital volume standard to be eleven high-risk procedures for which there was a strong volume outcome relationship. For esophageal resection for cancer cases, this value was set at 20¹². Similarly, in 2017, Fuchs et al.,¹³ performed a retrospective review of the Nationwide Inpatient Sample to determine the effects on mortality after esophagectomy based on center volume. They also demonstrated improved mortality at high volume centers compared to low volume centers, which were defined as those performing greater than or equal to 20 esophagectomies per year and those performing less than six esophagectomies per year, respectively¹³. These previous findings support our results and demonstrate the importance of center experience in dealing with postoperative complications.

Limitations of the study were those inherent to meta-analysis, with the inclusion of reporting bias from each of the studies analyzed to obtain pooled event rates. There were a few prospective and RCTs included which avoid this bias. Additionally, the use of the NOS to guarantee that studies included were of high quality does improve the reliability of these results. We also compensated for this with an initial broad, all-inclusive search along with subsequent rigorous review based on preset inclusion and exclusion criteria to ensure that no relevant outcome data was spared. Although it would be interesting to look at surgeon case volume and its impact on leak rate, it was out of our aim during settlement of the inclusion criteria and the majority of the included papers did not specify the leak rate per surgeon. While Esophagectomy Complications Consensus Group categorized anastomotic leak into three types, we included all authors’ reported leak cases due to the significant variation in the leak definition globally. Finally, presence of publication bias reflects that incidence of leak is high in highly active centers, but this incidence will most likely be even higher in average activity centers. Future prospective studies, entailing standardized clinical classification and dedicated imaging, are needed to further assess incidence of, mechanisms for, and different preventions for anastomotic leak postesophagectomy.

Conclusions

Anastomotic leak after esophagectomy is seen to be higher in association with centers performing less than 37 annual esophagectomies. Early mortality related to this potentially devastating complication can be decreased with vigilance to help with early detection and aggressive management.

Ethical approval

This is a meta-analysis of published data, thus no ethical approval was required.

Sources of funding

None.

Author contribution

M.R., N.A., M.G.: idea; M.R., J.M., M.B., S.S., K.M., M.H., M.D.: literature search and data collection; M.R., M.B.: data analysis; M.R., J.M., M.B., F.M. K., N.K.A., and M.G.: writing. Critical revision is done by all authors.

Conflicts of interest disclosure

The authors declare that they have no financial conflict of interest with regard to the content of this report.

Research registration unique identifying number (UIN)

1. Name of the registry: PROSPERO

2. Unique Identifying number or registration ID: CRD42019123947

3. Hyperlink to your specific registration (must be publicly accessible and will be checked): https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42019123947

Guarantor

Mario Gaudino, MD, PhD. Department of Cardiothoracic Surgery, Weill Cornell Medicine of Cornell University, 525 East 68th Street, NY 10065, USA. E-mail: mfg9004@med.cornell.edu. Mohamed Rahouma, MD, MS. Department of Cardiothoracic Surgery, Weill Cornell Medicine of Cornell University, 525 East 68th Street, NY 10065, USA. E-mail: mmr2011@med.cornell.edu; mhmdrahouma@gmail.com

Provenance and peer review

Not commissioned, externally peer-reviewed.

Supplementary Material

js9-110-2349-s001.pdf ^{(57.8KB, pdf)}

js9-110-2349-s002.jpg ^{(326.6KB, jpg)}

js9-110-2349-s003.jpg ^{(83.1KB, jpg)}

js9-110-2349-s004.jpg ^{(78.7KB, jpg)}

js9-110-2349-s005.pdf ^{(250.8KB, pdf)}

js9-110-2349-s006.docx ^{(15.4KB, docx)}

js9-110-2349-s008.docx ^{(66.8KB, docx)}

js9-110-2349-s009.docx ^{(17.3KB, docx)}

js9-110-2349-s010.docx ^{(19.7KB, docx)}

js9-110-2349-s011.docx ^{(18.7KB, docx)}

js9-110-2349-s012.docx ^{(15KB, docx)}

js9-110-2349-s013.docx ^{(18.7KB, docx)}