Postoperative short-term outcomes of minimally invasive versus open esophagectomy for patients with esophageal cancer: An updated systematic review and meta-analysis.

Background We performed a systematic review and meta‐analysis to synthesize the available evidence regarding short‐term outcomes between minimally invasive esophagectomy (MIE) and open esophagectomy (OE). Methods Studies were identified by searching databases including PubMed, EMBASE, Web of Science and Cochrane Library up to March 2019 without language restrictions. Results of these searches were filtered according to a set of eligibility criteria and analyzed in line with PRISMA guidelines. Results There were 33 studies included with a total of 13 269 patients in our review, out of which 4948 cases were of MIE and 8321 cases were of OE. The pooled results suggested that MIE had a better outcome regarding all‐cause respiratory complications (RCs) (OR = 0.56, 95% CI = 0.41–0.78, P = <0.001), in‐hospital duration (SMD = −0.51; 95% CI = −0.78−0.24; P = <0.001), and blood loss (SMD = −1.44; 95% CI = −1.95−0.93; P = <0.001). OE was associated with shorter duration of operation time, while no statistically significant differences were observed regarding other outcomes. Additionally, subgroup analyses were performed for a number of different postoperative events. Conclusions Our study indicated that MIE had more favorable outcomes than OE from the perspective of short‐term outcomes. Further large‐scale, multicenter randomized control trials are needed to explore the long‐term survival outcomes after MIE versus OE.


Introduction
Esophageal cancer is the seventh most common cause of cancer-related death globally. 1 The overall five-year survival is below 20%. 2,3 The main course of treatment is surgical resection, which is usually combined with chemotherapy or chemoradiotherapy for locally advanced tumors. 4 Conventional surgical treatment involves open esophagectomy (OE) using transthoracic or transhiatal approaches which are associated with high morbidity and mortality. Respiratory complications (RCs) are common with OE and can increase the risk of death up to 20%. [5][6][7] In recent decades, minimally invasive esophagectomy (MIE) has become an alternative to OE. MIE encompasses a number of techniques including total MIE (tMIE), hybrid minimally invasive esophagectomy (hMIE) and robotic surgery. 8 Given the technical complexity of MIE, a number of concerns exist regarding the benefits of MIE compared with OE in terms of postoperative complications and short-term mortality. On one hand, even though a number of previously performed studies have established MIE as a relatively safe procedure in terms of post-operative outcomes, [9][10][11][12][13] on the other, studies performed by Seesing et al. and Mariette et al. state the opposite. 14,15 With a number of emerging studies regarding MIE and OE in recent years, there has been a lack of a systematic study to investigate the short-term outcomes after MIE versus OE. Furthermore, a detailed and updated metaanalysis concerning the two approaches might help surgeons with their surgical decisions. Ergo, the purpose of this systematic review and meta-analysis was not only to use the latest and largest population-based data to extensively compare and summarize the postoperative complications after MIE versus OE for esophageal cancer, but also to clarify whether MIE could improve the post-operative outcomes and overall survival of patients with esophageal cancer.

Literature search strategy
This study was performed according to Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. Literature was identified by searching databases including PubMed, EMBASE, Web of Science and Cochrane Library up to June 2019 without language restrictions. The search terms used for literature identifications include "esophageal carcinoma, esophageal cancer, esophagectomy, minimally invasive esophagectomy, open esophagectomy and thoracoscopic laproscopic esophagectomy".

Eligibility criteria for literature selection
Literature included in the study had to meet the following criteria: (i) studies comparing MIE with OE; (ii) studies published in English only; (iii) studies including at least 20 or more patients; (iv) studies with assigned NOS (Newcastle-Ottawa quality assessment scale) score of seven or higher; (iv) prospective, randomized controlled trials or retrospective studies only; and (v) studies where full text was available.

Data extraction and quality assessment
Literature included in the study was independently assessed for methodological quality purposes (N.A and D.D). First, the titles and abstracts were screened to assess the eligibility of included literature, and then the full text was reviewed. Any discrepancies were resolved in discussion with a third author (C.D). The information recorded for each study is given in Table 1.

Definition of study endpoints
In total, we discussed 11 endpoints in our study: one primary and 10 secondary endpoints. All-cause respiratory complications (RCs) were chosen to be discussed as the primary endpoint. These RCs included atelectasis, pneumonia, acute respiratory distress syndrome (ARDS), pleural effusion, pneumothorax and respiratory insufficiency. The details of 10 secondary endpoints are given below. All-cause cardiac complications (CCs) which included cardiac arrest, myocardial infarction, atrial & ventricular dysrhythmia, congestive heart failure and pericarditis; allcause anastomotic leakage (AL) defined as full thickness GI defect involving esophagus, anastomosis, staple line, or conduit irrespective of presentation or method of identification; total length of in-hospital stay; total operation time; total blood loss; R0 resection; 30-day mortality; 90-day mortality; all-cause in-hospital mortality; and reoperation rate.

Statistical analysis
SPSS software was used for general data analysis. Data was extracted and entered into review manager. Continuous variables were expressed as median and interquartile ratio or range, and the mean and SD were estimated from the available data. The Mantel-Haenszel method for dichotomous data was used. Fixed or random-effects models were used in this study. Forest plots were provided to illustrate pooled odds ratios (ORs), and corresponding 95% confidence intervals (CIs). Cochran's Q test and Higgins I 2 were used to test the heterogeneity of different studies. A P-value of less than 0.1 was considered significant. Heterogeneity was interpreted according to the thresholds outlined in the Cochrane Handbook. With significant heterogeneity, a pooled effect was calculated with a random-effects model; otherwise, a fixed-effects model was applied. The reasons for interstudy heterogeneity were explored by using subgroup analysis. We also conducted sensitivity analysis by omission of each single study to evaluate stability of the results. Publication bias was assessed by using funnel plots.

Selection of eligible studies
The PRISMA flowchart diagram is shown in Figure 1. In summary, our literature search strategy initially identified 150 articles. Finally, 33 articles qualified to be included in our meta-analysis study.

Characteristics of included literature
A total of 13 269 patients were included in this metaanalysis study, out of which 4948 cases were of MIE and 8321 cases were of OE. Table 1 provides detailed characteristics of the articles included. In summary, six studies had a RCT study design, 12 had a prospective study design and the remaining 15 had a retrospective study design. Test of heterogeneity showed considerable heterogeneity (I 2 = 77% and P = <0.001). Subgroup analyses were conducted to explore potential sources of that heterogeneity ( Table 2). The pooled ORs of most subgroups were not markedly changed by the study characteristics. However, the subgroup analysis by intervention type showed considerable significance for tMIE/OE (P = <0.001; I 2 = 91%) as compare to hMIE/OE (P = 0.07; I 2 = 35%) which was less significant. We also noted the changes in statistical heterogeneity in the subgroup analysis of different institutes and facilities (single center, I 2 = 64%; multicenter, I 2 = 83%), initial inclusion period (<2008, I 2 = 68%; ≥2008 I 2 = 88%), study design (RCT, I 2 = 74%; prospective, I 2 = 79%; retrospective, I 2 = 54%), and NOS score (7, I 2 = 56%; 8, I 2 = 74%; 9, I 2 = 87%). Sensitivity analysis was conducted by omission of each single study to evaluate the stability of results indicating an unaffected pooled OR. The funnel plots displaying the publication bias of all cause RCs is shown in Figure S2b.

Discussion
This study compared the outcomes of OE with both tMIE and hMIE. Due to the complexity of esophagectomy, different types of surgical approaches might lead to different kinds of surgical complications, but the main morbidities remain the same which include RCs, CCs, AL and the aforementioned.
Most of the meta-analysis studies comparing the outcomes of MIE and OE previously performed were either based on retrospective studies only, or had a small sample size. 11,[47][48][49] Although, Lv et al. had a relatively larger sample size of 6025 patients from 20 studies, their study only included literature up to 2016. 12 Since then, a considerable number of updated studies have been published, showing new findings and discrepancies in their results. 9,[13][14][15][16]18,20,23,26,27,33,42 In contrast, we included 33 studies in total involving 13 269 patients in our meta-analysis to provide the latest and more robust outcomes comparing MIE and OE.
Postoperative RCs are of great importance and could impact the prognosis of patients, which are also the most frequent morbidity events after esophagectomy. Some previous studies have shown contradictory results regarding the advantages of MIE over OE with respect to postoperative RCs. Two retrospective studies showed no significant differences regarding RCs between two groups. 13,50 On the other hand, two RCTs showed a significantly lower incidence of respiratory complications after MIE than OE. 36,51 Pooled data from our study also showed that patients who underwent MIE experienced fewer postoperative RCs compared to those who underwent OE (Fig. 2a). The association of MIE with fewer postoperative RCs could be explained by the elegance of the MIE operation procedure which decreases surgical trauma to the chest wall and does less harm to pulmonary tissues.
The results from our study showed that MIE was associated with a longer operative time as compared to OE. These results were consistent with other recently published studies and could be attributed to the technical difficulty in MIE and a limited operating space for surgeons to perform the delicate procedure. 16,18,42 Data analyses also demonstrated that patients who underwent MIE experienced shorter postoperative in-hospital stay and had less in-operative blood loss, as compared to those who underwent OE. Both these results were in accordance with previous studies and can be associated with the less intrusive nature of MIE. 23,26,36 Notably, pooled results and subgroup analyses from our study showed no significant correlation between neoadjuvant therapy and improvement of postoperative outcomes, either after MIE or OE.

Principle findings and limitations
Our meta-analysis provides strong evidence for the association of MIE with overall better short-term outcomes (Table 3). When stratified by publication year, initial inclusion period, number of cases, types of surgical intervention, and NOS quality score, the results remained mostly constant. Meanwhile, the heterogeneity in subgroup analyses was shown to be not considerable in general. In addition, with the application of some advanced statistical methods, the results have demonstrated that the outcomes tend to be much more stable with the increasing number of studies over time.
There are several limitations to our study that should also be acknowledged. First, as shown in Table 1, the pathological TNM staging and ASA classification is missing from several included studies, which resulted in undeniable differences in their quality and strength. Second, patients of different ethnical groups were placed together into MIE or OE groups, which would also have effects on the results of this study. Third, different MIE methods (tMIE or hMIE) were used in different included studies, which makes it difficult to more specifically point out if there was any particular MIE technique that was the most beneficial for better outcomes. Fourth, there is also a possibility that patients with beneficial prognostic variables, for example, younger age and less comorbidity, were more readily selected for MIE rather than OE. Finally, even though our study included several RCTs, the lack of larger number of multi-institutional RCTs might reduce the effectiveness of the research. Consequently, the work definitely needs to be improved when there are more RCTs. Although advanced statistical methods were applied, publication bias was inevitable as shown in Fig. S2.
In conclusion, while OE was associated with shorter operation time and a slightly better surgical clearance of the tumor (R0 resection rates) compared with MIE, MIE  was associated with fewer RCs, lesser blood loss, shorter postoperative in-hospital stay and better overall postoperative outcomes. Further large-scale, multicenter RCTs are needed to continue to explore further long-term survival outcomes of patients with MIE and OE.

Supporting Information
Additional Supporting Informationmay be found in the online version of this article at the publisher's website: Figure S1.