Minimally Invasive and Open Gastrectomy for Gastric Cancer: A Systematic Review and Network Meta-Analysis of Randomized Clinical Trials

Background and Objectives Optimal surgical management for gastric cancer remains controversial. We aimed to perform a network meta-analysis (NMA) of randomized clinical trials (RCTs) comparing outcomes after open gastrectomy (OG), laparoscopic-assisted gastrectomy (LAG), and robotic gastrectomy (RG) for gastric cancer. Methods A systematic search of electronic databases was undertaken. An NMA was performed as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-NMA guidelines. Statistical analysis was performed using R and Shiny. Results Twenty-two RCTs including 6890 patients were included. Overall, 49.6% of patients underwent LAG (3420/6890), 46.6% underwent OG (3212/6890), and 3.7% underwent RG (258/6890). At NMA, there was a no significant difference in recurrence rates following LAG (odds ratio [OR] 1.09, 95% confidence interval [CI] 0.77–1.49) compared with OG. Similarly, overall survival (OS) outcomes were identical following OG and LAG (OS: OG, 87.0% [1652/1898] vs. LAG: OG, 87.0% [1650/1896]), with no differences in OS in meta-analysis (OR 1.02, 95% CI 0.77–1.52). Importantly, patients undergoing LAG experienced reduced intraoperative blood loss, surgical incisions, distance from proximal margins, postoperative hospital stays, and morbidity post-resection. Conclusions LAG was associated with non-inferior oncological and surgical outcomes compared with OG. Surgical outcomes following LAG and RG superseded OG, with similar outcomes observed for both LAG and RG. Given these findings, minimally invasive approaches should be considered for the resection of local gastric cancer, once surgeon and institutional expertise allows. Supplementary Information The online version contains supplementary material available at 10.1245/s10434-023-13654-6.

Minimally Invasive and Open Gastrectomy … advances, high-quality radical en bloc surgical resection of the tumour remains the cornerstone of management, 7,8 Traditionally, gastrectomy was performed as an open procedure (OG), although more recently, minimally invasive surgical (MIS) approaches have undergone widespread adoption, including laparoscopic-assisted gastrectomy (LAG) and robotic-assisted gastrectomy (RG). 9,10 Advocates of MIS hold this approach to be advantageous for several reasons, particularly reduced morbidity, and enhanced recovery and cosmesis. 11,12 Nevertheless, OG remains the approach of choice for certain cases, as surgical approach is dependent on an array of patient, pathological, and societal parameters, as well as surgeon and institutional expertise. 13 The first LAG was performed for gastric cancer in 1994, 14 and this approach has subsequently been developed and refined. 15 Several studies, including multicentre, prospective, randomized clinical trials (RCTs), have illustrated the non-inferiority of long-term oncological and survival outcomes following LAG and OG. Consequently, LAG is a well-established surgical approach in the management of early gastric carcinoma. [16][17][18][19] Increasingly, LAG is currently emerging as feasible, safe and effective for radical resection of locally advanced distal gastric cancer. 20,21 Nevertheless, there remains debate surrounding the differences in postoperative and oncological outcomes following LAG, as well as the considerable technical aspects and learning curve associated with the laparoscopic approach. [22][23][24][25][26] Robot-assisted surgery has been proposed as a potential platform to overcome some of the limitations of conventional laparoscopy and has recently undergone widespread adoption by many specialists and surgical oncology units for the management of many solid organ cancers. This is principally due to the reported advantages of three-dimensional vision, enhanced skill acquisition (due to a shorter learning curve relative to laparoscopic surgery), increased operator dexterity, improved mobility in narrow areas that have restricted access, and improved ergonomics for the operating surgeon. 27,28 However, while the first RG was performed in Japan 20 years ago by Hashizume et al., 29 the uptake of robotic surgery in upper gastrointestinal surgery has lagged considerably behind other surgical specialties. 30 At present, data available in relation to robotic surgery is primarily obtained from studies of a retrospective design, typically involving outcomes regarding single-centre or single-surgeon experiences of using robotics for gastric cancer resections, with limited long-term oncological outcomes being reported. 7 Therefore, there is a paucity of highquality data evaluating the role of RG for resection of gastric tumours. [31][32][33] Several RCTs and standard pairwise meta-analyses 32,34-37 have attempted to determine the optimal surgical approaches used for the resection of gastric carcinoma, however consensus in relation to the oncological and surgical safety of LAG and RG relative to OG is yet to be determined. Importantly, two recent RCTs have reported short-term postoperative and survival outcomes following RG, 38,39 and anticipation among gastroesophageal surgeons is that such minimally invasive approaches should enhance patient outcomes. 40 Therefore, application of the network meta-analysis (NMA) methodology here is timely to allow simultaneous comparison (direct and indirect) of minimally invasive approaches to gastrectomy with OG using RCT data only. 41,42

METHODS
A systematic review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension statement for reporting of systematic reviews incorporating NMAs of healthcare interventions. 43 This study was registered with the International Prospective Register of Systematic Reviews (PROS-PERO, CRD42022330440).

Search Strategy
A formal systematic search of four electronic databases was performed in March 2022 for relevant titles. Details in relation to the search strategy can be found in Appendix 1 in the electronic supplementary material (ESM). Study-specific definitions and the research question determined using the Population, Intervention, Comparison, Outcomes (PICO) framework 44 are outlined in detail in ESM Appendices 2 and 3.

Eligibility Criteria
All published RCTs with full-text manuscripts comparing the outcomes of two or more methods of surgical intervention for gastric cancer (i.e., OG, LAG or RG) were included. The inclusion criteria for studies were (1) compared postoperative surgical outcomes (e.g., complications, estimated blood loss, lymph node yield [LNY], etc.) or long-term oncological and survival outcomes (e.g., disease recurrence, overall survival [OS], etc.); (2) were of a prospective, randomized design; (3) recruited patients aged 18 years or older undergoing surgery for known primary gastric cancer; and (4) studies had to have full-text manuscripts available. Note that when overlapping trial data were reported from two different sources, the source with the longest patient follow-up or largest sample size was included.
The exclusion criteria were (1) studies failing to fulfil the above inclusion criteria; (2) studies that only have results published in abstract form or from conference proceedings; (3) studies not published in the English language; or (4) studies in which the primary indication for the gastrectomy 5546 M. G. Davey et al. was not for gastric cancer (e.g., bariatric sleeve gastrectomy, etc.).

Statistical Analysis
Descriptive statistics were used to outline the characteristics of the included trials. Data pertaining to recurrence, OS, morbidity, complications and readmission were expressed as dichotomous or binary outcomes, reported as odds ratios (ORs) with 95% confidence intervals (CIs). ORs were calculated using crude event RCT data, to compare interventions using per-protocol data, where applicable. Continuous data were calculated using mean values, standard deviations (SDs) and pooled mean variance, with differences expressed as weighted mean differences (WMDs). OG was the principal comparator for all analyses. Bayesian NMAs were conducted using netameta 45 and Shiny packages 46 for R. Point estimates of effect sizes were described with a 95% CI. Results were considered statistically significant at the p < 0.050 level if the 95% CI did not include a value of 1. Estimates of mean and SDs were calculated using standard statistical methods, where applicable. 47,48 Rank probabilities were plotted against the possible ranks for all competing treatments. The confidence in estimates of the outcome was assessed using the Confidence in Network Meta-Analysis (CINeMA) tool. 49

Risk-of-Bias Assessment
Assessment of potential biases within the included RCTs was assessed using the Cochrane Collaboration tool (for RCTs). 50 This assessment tool grades the risk of bias in each study as being high risk (marked in red), low risk (marked in green), or uncertain risk (marked in yellow) of bias across six categories. The critical appraisal was independently completed by two reviewers (HCT and MGD), and in the case of discrepancies in opinion, a third reviewer (NED) was asked to arbitrate.

Overall Survival
Overall, 31.8% of included studies reported outcomes for analysis in relation to OS (7/22 Overall, 50.0% of included studies reported on the number of LNs harvested (11/22). Compared with OG, LAG showed a significantly decreased LN yield (OR − 1.52, 95% CI − 2.75 to − 0.45), whereas there was a non-significant difference in LNs harvested for those who underwent RG (OR − 1.95, 95% CI − 5.77 to 1.25) ( Fig. 3c and ESM Appendix 3c).
In total, 31.8% of the included studies reported on the distance from the proximal margin (7/22). This distance from the proximal margin was significantly lower in those undergoing LAG compared with OG (OR − 0.310, 95% CI − 0.565 to − 0.0992) ( Fig. 3d and ESM Appendix 3d). The rank probability was highest in those undergoing LAG (0.99), indicating a decreased distance from the proximal margin in those undergoing LAG (Table 2).
In total, 27.3% of the included studies reported on the distance from the distal margin (6/22). The distance from distal margin was similar to those who underwent LAG and OG (OR − 0.220, 95% CI − 0.530 to 0.0617) ( Fig. 3e and ESM Appendix 3e). The rank probability was highest in those undergoing LAG (0.95), indicating a decreased distance from the distal margin associated with LAG (Table 2).
In total, 36.4% of the included studies reported on the length of surgical incision (8/22). This was significantly shorter in those undergoing LAG compared with OG (OR − 11.7, 95% CI − 14.5 to − 8.91) ( Fig. 3f and ESM Appendix 3f). The rank probability was highest in those undergoing LAG (0.99), indicating a decreased length of surgical incision associated with LAG (Table 2).

Risk of Bias
All 22 of the included studies were 'low' risk of bias for most categories, using the Cochrane Collaboration risk-of-bias assessment for RCTs. In brief, three of the included RCTs had 'low' risk of bias, 12 RCTs had 'some' risk of bias, and 7 RCTs illustrated evidence of 'high' risk of bias. Comprehensive summaries of the risk-of-bias assessment are outlined in ESM Appendix 10.

DISCUSSION
The management paradigm for gastric carcinoma has evolved such that multimodal therapeutic strategies are now pragmatically tailored to each patient, which has translated into enhanced oncological and survival outcomes. 67 In addition, the inclusion of quality-of-life (QoL) outcome measures is now routinely utilized to establish the impact of both local and systemic therapies on host (or biological) function, to determine whether the oncological benefit of such treatments offset their associated toxicities and morbidity. 68 This NMA was performed to establish the oncological and surgical safety of MIS gastrectomy compared with conventional OG, including data from 6890 patients treated for primary gastric carcinoma in 22 independent RCTs. This analysis illustrated the non-inferiority of LAG compared with OG with respect to survival, while highlighting the improved surgical and recovery outcomes associated with the MIS approaches, supporting their use where possible. Consequently, LAG should be considered for patients with primary resectable gastric cancer, providing that surgeon and institutional expertise allows, echoing the previous comprehensive results of a standard pairwise meta-analysis of RCTs performed by Lou et al. in 2022, 69 albeit limited by the inclusion of studies performed three decades previously.
Traditionally, extensive locoregional resection using OG was the standard of care for resection of gastric carcinoma, which has since been surpassed by the adoption of both laparoscopic and robotic approaches as routine. 70,71 Importantly, this study illustrated the non-inferiority of LAG relative to OG with respect to long-term oncological and survival outcomes, despite reduced nodal yields harvested and closer distances to specimen margins with LAG. OS was identical for both OG and LAG (both 87.0%) at approximately 5 years' follow-up, with similar recurrence observed for both (9.5% vs. 8.7%), which likely represents a significant proportion of patients with early-stage disease, limiting the translatability of these results into the locally advanced setting. However, notably, disease recurrence and OS outcomes are as yet unavailable for patients undergoing RG compared with the other modalities. Therefore, while this study comprehensively establishes the non-inferiority of LAG versus OG for disease recurrence and OS, the absence of RCT data for RG limits the synthesis of any similar level 1 conclusions regarding the oncologic efficacy of RG. In a previous metaanalysis of 19 non-randomized, observational studies including 7275 patients, Ma et al. reported similar OS (hazard   72 Furthermore, a previous propensity-matched analysis performed by Obama et al. demonstrated the noninferiority of RG relative to LAG for disease recurrence (6.7% vs. 5.0%) and OS (8.9% vs. 11.6%). 73 Interestingly, however, there was an increase in locoregional recurrence rates following RG relative to LAG (42.9% vs. 30.8%). In the absence of RCT data, the results from studies such as that by Obama et al. are important to highlight the potential risks and fundamental challenges to introducing and implementing new surgical techniques, albeit being limited due to its single-centre, retrospective design. However, it is important that further rigorous scientific evaluation of RG with prospective, randomized studies are conducted to ensure patient safety and to avoid the unexpected issues that have arose during the early adoption of other new surgical techniques for cancer, 74,75 while remaining cognizant of the fact that premature adoption, inadequate proctoring and suboptimal execution, rather than any issue with the technique itself, may be to blame. 76 While the survival outcomes for LAG and OG are equivocal, it is imperative that the other results in this NMA are considered when selecting the optimal technique for performing gastrectomy for gastric cancer, particularly in terms of the enhanced patient recovery and reduced complication rate associated with the MIS approaches, which coincide with the robust implementation of enhanced recovery after surgery (ERAS) protocols in contemporary surgical oncology. 77 Patients undergoing LAG experienced a reduction in IBL, shorter surgical incisions, reduced distance from proximal margins, shorter postoperative hospital stays, and, most importantly, reduced morbidity post-resection. When interpretating these data, these important findings tip this study in favour of minimally invasive techniques, particularly when these data support the comparability of LAG and OG regarding long-term follow-up. MIS is advantageous as patients tend to be subject to less physiological stress, immunologic burden, faster recovery times, lower complication rates, and less immediate and long-term burden on healthcare resources. 78 Therefore, MIS techniques may prove advantageous in improving cost effectiveness in the long-term, despite the greater direct cost associated with such surgical approaches as described in the current analysis. Accordingly, this study further validates the current paradigm shift towards adopting MIS techniques where possible, as these approaches are associated with longer operative duration relative to OG, as well as RG being significantly more expensive than LAG in the current NMA, coupled with emerging evidence suggesting an increased risk of cardiovascular complications, rendering patient selection imperative if RG is being contemplated during multidisciplinary discussion. Therefore, this study highlights the premise for LAG to be utilized for primary gastric resection where feasible, should institutional expertise allow, with further evaluation of RG approaches required to determine what benefit, if any, this approach may have over LAG.
Despite the absence of survival data, this study does provide preliminary data in support of RG, which was associated with reduced morbidity, major morbidity, and similar cardiorespiratory, pancreatic and other significant postoperative complications in this study compared with OG and LAG. Moreover, when compared with LAG, patients undergoing RG had a significant reduction in overall morbidity (OR 0.43, 95% CI 0.25-0.76), further potentiating RG as a pragmatic minimally invasive approach to gastrectomy in patients with early gastric cancer. These are important findings that strengthen the perceived benefit associated with RG, particularly when the application of robotic technology is an attractive addition to the surgeons' armamentarium, due to the theoretical advantages over conventional laparoscopy, including improved dexterity, enhanced visualization, and superior ergonomics. 79 Shortcomings of robotic surgery include the longer operative time and increased expense associated with this approach, as well as purchasing and maintaining equipment and training operators, a steep learning curve, and poorer cost effectiveness in low-volume centres. [80][81][82][83] While the data suggest reasonable equipoise between RG and LAG surgery in terms of morbidity and recovery, 72 the argument that robotic instrumentation may allow for improved mobility in narrow areas with restricted access, for example, at the diaphragmatic hiatus and when performing anastomoses, seems plausible. 84 Another example is lymphadenectomy, where the removal of the D2 nodes is considered the standard surgical procedure for the majority of patients with resectable gastric carcinoma. 85 Controversary remains in relation to resection beyond D2 for cases of advanced disease, for several reasons, including the reduced operative freedom, the significant difficulty controlling haemorrhage, and the relative ease of trauma to local structures. 86 The results of the present NMA show a significant reduction in LNY following LAG, with similar outcomes observed following OG and RG, indicating that RG may offer an advantage over LAG for technically difficult manoeuvres such as extensive lymphatic resection during 'D2 plus' lymphadenectomy for advanced gastric cancer. 85 The current analysis is subject to limitations. First, as described in detail, none of the two included RCTs reported survival outcomes following RG, limiting the conclusions that may be drawn from the current study. Second, the studies included in this review failed to provide survivorship data that may inform the psychosocial impact of OG, LAG, and RG on patients' QoL following resection. Third, there are several competing factors that may confound the data presented in this study; these include the prescription of (neo) adjuvant chemo-, radio-, or immunotherapeutic agents to improve survival outcomes, as well as the impact of surgical approach in the context of early, locally advanced, or advanced gastric carcinoma. Unfortunately, attempts to perform analyses allowing for correction of such factors have been futile, thus limiting these results. Finally, evaluation of surgeon-specific proficiency and the influence of the centralisation of gastric cancer treatment to high-volume centres was not evaluated to determine their impact on clinical outcomes. Therefore, those responsible for the provision of the next generation of prospective, randomized gastrectomy trials should consider these confounding factors at the time of trial design to ensure the optimisation of data outcomes.

CONCLUSION
This analysis demonstrates the non-inferiority of oncological and surgical outcomes for OG and LAG in patients being treated predominantly for early gastric cancer following 5 years of follow-up. Moreover, surgical outcomes following minimally invasive gastrectomy superseded those following OG, with enhanced outcomes observed for both LAG and RG. This highlights the oncological and surgical safety of LAG relative to OG, while also illustrating that the short-term surgical, morbidity and recovery outcomes following RG are comparable with LAG. Given these findings, LAG may be considered for patients with primary resectable gastric cancer, providing that surgeon and institutional expertise allows, however further RCTs are warranted before definitive conclusions may be drawn in the setting of advanced disease. However, given the challenges in adopting high-quality LAG for all cases, evaluation of new techniques that might make the procedure easier and thereby reduce the incidence of poor oncological and functional outcome are warranted. The concept of utilizing RG in these circumstances offers a number of theoretical but as yet unproven advantages, apart from LNY as was demonstrated in this analysis. In time, as more institutions employ an RG approach, it is likely that surgical and institutional expertise will facilitate further evaluation of the MIS techniques to determine the relative advantages and disadvantages of both techniques, which have not been extensive explored in this study. In the interim, technique selection should be based on individual tumour characteristics and patient expectations, as well as surgeon and institutional expertise.
ACKNOWLEDGEMENT The authors would like to thank Dr. Amirhossein Jalali, School of Medicine, University of Limerick, Limerick, Ireland, for his expertise and assistance with the study.
FUNDING Open Access funding provided by the IReL Consortium. No funding was received for this study.
DATA AVAILABILITY Data can be made available upon reasonable request from the corresponding author. ETHICAL APPROVAL Not applicable as this is a review of previously published studies.

DISCLOSURE
OPEN ACCESS This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.