Survival Outcomes of Hepatectomy in Gastric Cancer Liver Metastasis: A Systematic Review and Meta-Analysis

Gastric cancer liver metastasis (GCLM) is a contraindication for surgical treatment in current guidelines. However, the results of recent studies are questioning this paradigm. We assessed survival outcomes and their predictors following hepatectomy for GCLM in a systematic review of studies published from 2000 to 2022 according to PRISMA guidelines. We identified 42,160 references in four databases. Of these, 55 articles providing data from 1990 patients fulfilled our criteria and were included. We performed a meta-analysis using random-effects models to assess overall survival (OS) and disease-free survival (DFS) at one, three, and five years post-surgery. We studied the impact of potential prognostic factors on survival outcomes via meta-regression. One, three, and five years after surgery, OS was 69.79%, 34.79%, and 24.68%, whereas DFS was 41.39%, 23.23%, and 20.18%, respectively. Metachronous presentation, well-to-moderate differentiation, small hepatic tumoral size, early nodal stage, R0 resection, unilobar compromisation, and solitary lesions were associated with higher overall survival. Metachronous presentation, smaller primary tumoral size, and solitary metastasis were linked to longer DFS. The results of our meta-analysis suggest that hepatectomy leads to favorable survival outcomes in patients with GCLM and provides data that might help select patients who will benefit most from surgical treatment.


Introduction
Gastric cancer (GC) is responsible for 1 in every 12 deaths globally. It represents approximately 8% of cancer-related deaths and is the third leading cause of cancer-related death worldwide [1,2]. In the surgical treatment of GC, complete resection, lymph node dissection, and neoadjuvant and/or adjuvant therapy are the goals for improved survival time [3][4][5][6]. However, despite advances in prevention and treatment strategies, the prognosis has only been modestly improved [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. In Western countries such as the United Kingdom, the majority of gastric cancer cases are detected at an advanced stage. Only 30% of patients are eligible for treatment with curative intent, mainly due to metastatic spread [17][18][19][20][21][22][23][24][25]. The liver, bones, and peritoneum are the most common sites where metastases spread [21][22][23][24][25][26]. Gastric cancer liver metastases (GCLM) are often multifocal, bilobar, or accompanied by simultaneous extrahepatic metastases such as peritoneal lesions or extensive lymph node compromise [27]. The incidence of GCLM ranges from 5% to 9% [28]. They are synchronous (before, during, or within 6 months after gastrectomy) in 3% to 14% of patients and metachronous (6 months or later after gastrectomy) in up to 37% [29]. Current Western guidelines advise against surgical resection of liver metastases, with the majority of patients instead receiving palliative chemotherapy [25][26][27][28][29][30]. According to the National Comprehensive Cancer Network (NCCN), liver metastases are classified as IVb disease and are therefore recommended to be managed with systemic chemotherapy [31]. The median overall survival of GC patients with liver metastasis is around 12 months. The most widely used regimens are doublets or triplets with fluoropyrimidines, platinum derivatives, taxanes, or anthracyclines [32,33]. Given the unsatisfactory prognosis associated with systemic therapy, the clinical management of stage IV disease has been changing over the past few years. Advances in systemic therapy and liver surgery safety have driven the expansion of surgical indications for metastatic disease [34]. However, due to a lack of large, well-designed clinical trials, surgery remains limited to individually selected cases of GCLM. Solely in Japanese treatment guidelines, considering surgery with curative intent is recommended when the number of metastatic nodules is small and no other non-curable factors are present [35]. Here, we aimed to review the literature and provide meta-analytic evidence on survival outcomes following hepatectomy for GCLM. Furthermore, we sought to identify potential predictors of beneficial clinical outcomes.

Materials and Methods
We performed a systematic review and meta-analysis in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The PRISMA checklist is shown in the Supplement S1.

Search Strategy
Two independent reviewers (D.C., C.R.) performed a systematic search of the literature to identify eligible articles. We searched the databases PubMed, Cochrane, EMBASE, and Google Scholar using the following search terms and Boolean operators: "Gastric cancer AND hepatectomy" OR "gastric AND cancer AND liver metastases" OR "gastric AND cancer AND metastasectomy" OR "stomach AND cancer AND liver metastases", "liver resection", "hepatectomy", "carcinoma", "neoplasm". We also reviewed the reference list of included articles for eligible papers. The complete search strings for all databases are provided in Supplement S2.

Study Selection and Data Collection
Publications were included when they met the following eligibility criteria: (i) studies involving humans and available as full-text articles in English published from January 2000 to June 2022; (ii) studies in patients who had hepatectomy for GCLM as an upfront radical resection with simultaneous management of the primary tumor and the liver metastasis, or as a subsequent procedure after primary gastric cancer treatment; (iii) articles reporting data on 1-,3-, and/or 5-year overall survival and/or disease-free survival of GC patients with only liver metastasis treated surgically; (iv) studies where planning of the surgery was based on the intention to achieve R0 resection. We included prospective or retrospective studies that were of observational or interventional nature. We excluded articles reporting on surgery for GCLM with insufficient reporting of outcome data as well as retrospective observations in less than 10 patients.
Two authors (D.C. and C.R.) independently selected studies by screening titles and abstracts and removed duplicates between articles identified across different databases. Subsequently, they interchanged the lists of articles deriving from their searches to compare the lists in a consensus-based approach. In the case of contradictions between the reviewers, a third reviewer (F.G.) was involved to establish consensus. Afterward, an investigator (D.C.) extracted relevant data from included articles and entered these data into a Microsoft Excel ® version 16.6 database.

Data Synthesis and Quality Assessment
Statistical analysis was performed using the Medcalc ® software package (Version 20.110, MedCalc Software Ltd., Acacialaan 22, 8400 Ostend, Belgium). A p value < 0.05 was considered to be statistically significant.
We calculated one-, three-, and five-year OS and DFS as the proportion of patients being alive and free from the tumor, respectively. Hazard ratio (HR) and estimated standard errors with corresponding 95% confidence intervals (95% CIs) were calculated to assess the association between potential prognostic factors and OS and DFS. HRs and their variance were obtained from the studies or calculated according to the data presentation: annual mortality rates, survival curves, number of deaths, or percentage freedom from death. A random-effects model was used to perform a meta-analysis assuming differences in the treatment effect. Forest plots were generated to illustrate the results of the meta-analysis. Publication bias was assessed with Egger and Begg tests and through funnel plots for graphical inspection. Heterogeneity across the studies was assessed using the Cochran Q test and/or the Higgins test (I 2 statistic to measure the degree of variation not attributable to chance alone). Heterogeneity was graded as low (I 2 < 25%), moderate (I 2 = 25% to 75%), or high (I 2 > 75%). We carried out a meta-regression including variables with a potentially prognostic association with survival outcomes. Models of meta-regression were built with one covariate at the time. We extracted data of sufficient quantity to conduct a meta-regression for age, N stage, time presentation, lobar compromise, number of lesions, size of the metastasis, and kind of hepatectomy for OS. By contrast, we were not able to perform a meta-regression for DFS due to the limited no availability of sufficient data. However, not enough data were obtained to perform a regression with all the factors. Additionally, a regression on the number of patients and year of publication was performed. A p value < 0.05 was considered to be statistically significant. Two raters (D.C. and C.R.) assessed study quality using the Newcastle-Ottawa quality assessment scale independently (36). Afterward, results were compared and a third rater (F.G.) was involved to achieve consensus where necessary.

Results
A total of 42,160 references were obtained by the initial electronic search: 9531 from PubMed, 264 from Cochrane, 22,480 from Google Scholar, and 9885 from Embase. Upon removal of duplicates, 183 papers were eligible for review of their abstracts and full texts. By applying our selection criteria, we included 55 suitable papers into our systematic review and meta-analysis of a total of 1990 patients who underwent hepatic resection for GCLM. All of the included studies were retrospective, and thus none of them had an interventional or prospective observational design. Tables 1 and 2 summarize the main characteristics of the included studies. Figure 1 depicts the flow of information through the phases of our systematic review.
Of the 55 included articles, 42 studies (76%) were conducted in Asian populations and 13 (24%) in Western countries. In all the studies involved, the indications for hepatectomy were good control of the primary tumor, no disseminated disease, and the feasibility of achieving R0 resection.

Characteristics of the Patients
The median age of our synthesized population was 64 years (range: 54-74 years), and the coefficient of variation was 7.1%. Men constituted 74.2% (n = 1385) and women 25.8% (n = 481) of the entire population. All patients had undergone surgery for primary GC either in a previous intervention or in the current intervention. All patients were treated with surgery for GCLM, either a synchronous resection in 1215 patients or a metachronous resection in 690 cases. The T classification was divided into two groups: the T1-T2 group consisted of 36.79% of the cases, and the T3-T4 group of 61.26%. The histology was classified according to the differentiation as well-differentiated (48.5%), moderately differentiated (38.1%), and poorly differentiated (33.9%). With respect to nodal compromise, 22.5% of patients were classified as N 0-1 and 38.5% as N 2-3. Of the entire population, 739 (37.13%) patients received adjuvant chemotherapy, and 159 (7.9%) received preoperative chemotherapy. Of all included articles (n = 55), 23 reported data on resection margins. In these studies, overall, 1070 cases had reported R0 resection margins, whereas they were R1 on pathologic analysis in 129 cases. Of the entire synthesized population, unilobar resection of metastasis was performed in 34.5% (n = 687) of patients, whereas multilobar resection was performed in 10.6% (n = 210). Solitary lesions were resected in 45.6% of cases (n = 908), whereas resections of multiple lesions were performed in 26.2% of cases (n = 522). The mean size of the hepatic lesions resected was 3.12 cm (range: 2-5.5 cm). A minor hepatectomy was performed in 54.37% (n = 1082) of the population. By contrast, a major hepatectomy was undertaken in only 18.49% (n = 368) of patients. The mean follow-up in the included studies was 34.6 months (range: 8.9-90.8 months).

Mortality at 30 Days
The mortality rate at 30 days was 1.37% in the synthesized population. The data reported were insufficient to describe the operative complication subtypes and co-morbidities associated with complications.

Overall Survival at 1 Year
Of the 55 included articles, 45 (81.8%) reported OS in the first year after hepatectomy for GCLM. The pooled OS rate was 69.8% (CI 95%: 65.5%, 73.8%) with moderate heterogeneity and random effect size (I 2 = 69.50%, heterogeneity test Q, p < 0.0001). There was no evidence of publication bias, neither on Begg (p = 0.9057) and Egger (p = 0.9057) tests nor on inspection of the funnel plot ( Figure 2). n. Med. 2023, 12, x FOR PEER REVIEW 11

Overall Survival at 3 Years
Of the 55 included articles, 44 (80%) reported OS in the third year after hepatecto for GCLM. The pooled OS rate was 34.8% (CI 95%: 30.7%, 38.9%) with a random e size of and moderate heterogeneity (I 2 = 65.7%, heterogeneity test Q, p < 0.0001). There no evidence of publication bias, neither on Begg (p = 0.85) or Egger (p = 0.90) tests no the inspection of the funnel plot ( Figure 3).

Overall Survival at 5 Years
Of the 55 included articles, 48 (87.2%) reported OS in the fifth year after hepatectomy for GCLM. The pooled OS rate was 24.7% (CI 95%: 21.3%, 28.1%) with random effect size and moderate heterogeneity (I 2 = 62.4%, heterogeneity test Q, p < 0.0001). There was no evidence of publication bias, neither on Begg (p = 0.0997) and Egger (p = 0.2975) tests nor on inspection of the funnel plot ( Figure 4).

Disease-Free Survival at 1 Year
Of the 55 included articles, 16 (29%) reported DFS in the fifth year after hepatectom for GCLM. The pooled DFS rate was 41.4% (CI 95%: 34.4%, 48.5%) with a random eff size and moderate heterogeneity (I 2 = 69.5%, heterogeneity test Q, p < 0.0001). There w no evidence of publication bias, neither on Begg (p = 0.4421) and Egger (p = 0.5197) te nor on inspection of the funnel plot ( Figure 5).

Disease-Free Survival at 1 Year
Of the 55 included articles, 16 (29%) reported DFS in the fifth year after hepatectomy for GCLM. The pooled DFS rate was 41.4% (CI 95%: 34.4%, 48.5%) with a random effect size and moderate heterogeneity (I 2 = 69.5%, heterogeneity test Q, p < 0.0001). There was no evidence of publication bias, neither on Begg (p = 0.4421) and Egger (p = 0.5197) tests nor on inspection of the funnel plot ( Figure 5).

Disease-Free Survival at 3 Years
Of the 55 included articles, 14 (25.4%) reported DFS in the third year after hepate tomy for GCLM. The pooled DFS rate was 23.2% (CI 95%: 18.2%, 28.7%) with rando effect size and moderate heterogeneity (I2 = 59%, heterogeneity test Q, p < 0.0001). The was no evidence of publication bias, neither on Begg (p = 0.2259) and Egger (p = 0.089 tests nor on inspection of the funnel plot ( Figure 6).

Disease-Free Survival at 3 Years
Of the 55 included articles, 14 (25.4%) reported DFS in the third year after hepatectomy for GCLM. The pooled DFS rate was 23.2% (CI 95%: 18.2%, 28.7%) with random effect size and moderate heterogeneity (I 2 = 59%, heterogeneity test Q, p < 0.0001). There was no evidence of publication bias, neither on Begg (p = 0.2259) and Egger (p = 0.0893) tests nor on inspection of the funnel plot ( Figure 6).

Disease-Free Survival at 5 Years
Of the 55 included articles, 16 (29%) reported DFS in the fifth year after hepatectomy for GCLM. The pooled DFS rate was 20.2% (CI 95%: 14.3%, 26.7%), with random effect size and moderate heterogeneity (I 2 = 74.6%, heterogeneity test Q, p < 0.0001). There was no evidence of publication bias, neither on Begg (p = 0.8571) and Egger (p = 0.5429) tests nor on inspection of the funnel plot (Figure 7).

Figure 7.
Forest plot of disease-free survival at 5 years after hepatectomy for gastric cance metastasis (GCLM). Description of the proportion and confidence intervals at 95% in the s that reported 5-year disease-free survival.

Prognostic Factors
In the meta-analysis of the risk factors, we found an association between impr OS and the following factors: R0 resection, small diameter of the metastasis, resecti a solitary lesion, unilobar localization, low node compromise, early T stage of the pri tumor, well-to-moderate differentiation grade, and metachronous presentation. In tion to DFS, the factors associated with improved DFS were a metachronous present solitary lesions, and an early T stage (Table 4).

Overall Survival
Disease-Free Survival

Prognostic Factors
In the meta-analysis of the risk factors, we found an association between improved OS and the following factors: R0 resection, small diameter of the metastasis, resection of a solitary lesion, unilobar localization, low node compromise, early T stage of the primary tumor, well-to-moderate differentiation grade, and metachronous presentation. In relation to DFS, the factors associated with improved DFS were a metachronous presentation, solitary lesions, and an early T stage (Table 4). Prognostic factors' hazard ratios on overall survival and disease-free survival, the degree of heterogeneity, and p value.

Meta-Regression
The results of the meta-regression are shown in Table 5. Multivariate models were not constructed due to a large amount of missing data. Among all variables tested, only synchronous presentation displayed a prognostic association with higher OS (Table 5). Meta-regression based on a random effect size model. R 2 identifies the heterogeneity accounted for by the model.

Discussion
The major finding of this systematic review and meta-analysis is that hepatectomy results in beneficial rates of both overall survival and disease-free survival when assessed one, three, and five years post-surgery. The survival outcomes observed at these time points were 69.79%, 34.79%, and 24.68% for OS and 41.39%, 23.23%, and 20.18% for DFS, respectively, with a median survival of 24.5 months. Periprocedural mortality was 1.37%. Our results are in line with previous meta-analyses. The analysis of Petrelli et al. showed a weighted median OS of 22 months at 5 years in 23 studies with 870 patients taken to hepatectomy for GCLM [37]. Another meta-analysis from 2016 assessed 39 studies; the authors described a median survival rate of 68% at 1 year, 31% at 3 years, and 27% at 5 years in GCLM patients [38]. These survival outcomes are consistent with our analysis and differ substantially from those reported with systemic therapy using epirubicin, oxaliplatin, and capecitabine (EOX) in the REAL3 randomized controlled phase III trial at 1 year of 46%, and with a median survival of 11.3 months [39].
Metastatic gastric cancer has long been considered an aggressive disease, and therefore not suitable for surgery [32]. Current guidelines do not support surgery for GCLM [31]. Moreover, there is widespread skepticism about performing surgery in these cases [40]. A survey applied to surgeons in Europe and Japan found that for metachronous GCLM, most of the specialists (50.4%) prefer preoperative chemotherapy followed by liver resection, whereas 30.3% preferred chemotherapy alone and 36% preferred alternative treatments such as ablative radiofrequency ablation (RFA) alone or RFA with chemotherapy [40]. Some of the reasoning for performing surgery on GCLM patients is inspired by the results of research on colorectal LM. In 1439 patients with hepatectomy for colorectal liver metastasis, Adam et al. described survival rates of 33% and 23% at 5 and 10 years, respectively [41]. Recently, a multicenter retrospective review was conducted on 144 patients who underwent hepatectomy in synchronous and metachronous settings. They identified a median OS of 12 months [42]. Historically, OS has been the most commonly used metric for judging the success of treatment [43]. The main disadvantage of this measure is the need for extended follow-up and the potentially diluted death measurement due to other nonmalignant causes. DFS has emerged as a potential candidate for a surrogate of OS in various malignant diseases [43][44][45]. It may complement OS in the measurement of survival outcomes. However, reports of DFS in the literature are scarce. Only 16 out of the 55 articles included in our review reported DFS. A review of studies on a total of 1573 patients who underwent hepatic resection described 1-, 3-, and 5-year DFS of 44%, 24%, and 22%, respectively [46]. Survival is dependent on multiple recognized demographic, tumoral, and metastatic prognostic factors (see Table 3) [43]. Among the primary tumor characteristics, histologic type, serosal invasion, N-stage, and metachronous presentation are prognostic factors (p < 0.05) [32,47]. Tiberio et al. [48] recently demonstrated that T stage, R0 resection, and adjuvant chemotherapy administration are independent factors of OS. T stage ≥ 3, high nodal compromise (N2-3), and poor differentiation were identified as negative prognostic factors for both synchronous and metachronous metastases in their study. They conclude that these patients should be carefully evaluated before hepatic resection is proposed. In this meta-analysis, not achieving an R0 resection margin was associated with the worst prognosis (HR4.04 IC:2.73-7.08, p < 0.001). Even in the face of the worst prognosis of bilobar compromise and the size of the lesion, a propensity score analysis of 119 patients who received multidisciplinary treatments for liver metastasis showed that in the presence of an R0 resection, the distribution and number of liver metastases do not affect the prognosis [49]. Our synthesized analysis highlights the prognostic value of solitary resection over multiple lesions. However, liver resection is not clearly limited to a specific number of metastases. Some reports showed the benefit of resecting 1-3 metastases, and even multiple lesions cannot be considered exclusion criteria for surgery [50].
Neoadjuvant and adjuvant therapies are fundamental in treating advanced GC patients. Any progression during chemotherapy is probably the most relevant contraindication for surgery [46]. Naturally, the use of systemic therapy has improved in recent years, and it seems appropriate to discuss the role of preoperative chemotherapy to increase survival. The FLOT3 trial evaluated the benefit of a regimen of FLOT (5-fluorouracil, oxaliplatin, and docetaxel) followed by surgery and confirmed the potential OS gained [51]. In our review, neoadjuvant chemotherapy administration data were insufficient for inclusion in our analyses. Preoperative chemotherapy was less frequently administered than postoperative chemotherapy in the cohorts we analyzed. In our review, only four studies reported hazard ratios for adjuvant chemotherapy.
Currently, no prospective data exist about the comparison between non-resectional management and surgical management in patients with GCLM. Most of the studies in this review have no systemic therapy arm [52]. Shinohara et al. retrospectively compared OS in 22 patients receiving surgery in GCLM to 25 patients who did not receive surgery and found a significant difference between the groups (median survival time: 22 vs. 7 months, respectively, p = 0.001) [53]. Despite the benefits of gastrectomy plus hepatectomy over non-resectional management in patients with GCLM, it must be pointed out that all data came from retrospective studies and systematic reviews.
The use of targeted therapies has increased given the advances in the understanding of the molecular mechanisms of GC. Currently, the most studied therapy is related to HER2 expression. While early studies disagreed on the prognostic relevance of HER status, recent evidence highlights its value [54,55]. In 2010, the Toga Study evaluated the use of the HER2-targeted monoclonal antibody trastuzumab with standard chemotherapy in 584 patients. The addition of trastuzumab increased the median survival in HER2 positive patients to 13.8 months compared to 11.1 months with chemotherapy alone (HR 0.74 (95% CI 0.60-0.91, p = 0.0046)) [56]. More recent trials support the use of targeted therapy in HER2 patients (Keynote 811 and Destiny Gastric 01) [57]. However, the therapy has some drawbacks: HER2 expression in gastric cancer is only around 9-38%, the antibody shows high heterogeneity, the benefit is more evident in patients with high levels of HER2 expression, and there are different testing methods for HER2 measurement (2). Moreover, in GCLM, the evidence is limited. From the studies evaluated, only 1 study discriminated HER2 patients and received trastuzumab associated with chemotherapy [58]. A study performed in 94 patients with GCLM found no relevance of HER2 positivity as an independent prognostic factor (HR 0.918, IC: 0.185-4.5), but the analysis seems to be affected by the underpowered sample [59]. Future studies should include this as a potential prognostic factor.
The relevant difference between synchronous and metachronous diseases may be related to the insufficient data in the literature on metachronous cases. This may be explained by the difficulty in finding patients with potential surgical indications, tumoral aggressiveness, or, in most cases, the simultaneous spread of metastases. In a 2017 retrospective cohort study, the authors compared the outcomes of 653 patients with metachronous disease; 34 were treated surgically, while 619 were treated non-surgically. In this study, surgically treated patients displayed higher survival outcomes than non-surgically treated patients (1YOS: 73.5% vs. 19.7%, 3YOS: 36.9% vs. 6.6%, 5YOS: 24.53% vs. 4.4%, p < 0.001) [60]. Cui et al. found the metachronous hepatectomy to be a favorable factor for OS, but the number of liver metastases was not [61]. Overall, a metachronous presentation was associated with improved OS in this study. On the other hand, in this study's metaregression, synchronous presentation was associated with better OS in the first year. In addition to survival differences between metachronous and synchronous presentations, Tatsubayashi et al. described postoperative complications as more common in patients with synchronous GCLM compared to metachronous disease. The length of hospital stay in these patients was also prolonged (p = 0.003) [62].
This work contributes to the growing evidence supporting hepatectomy for GCLM. It has strengths. Our systematic review and meta-analysis provide synthesized analyses of a large population of patients with GCLM, complementing and partially exceeding the scope of previous reviews. It supports the value of DFS as a complementary marker of survival in GC patients. However, our analysis is limited by the retrospective nature of the included studies. Thus, selection bias and institutional bias cannot be ruled out.

Conclusions
In conclusion, the results of our systematic review and meta-analysis in a large population of patients with GCLM indicate that hepatectomy leads to favorable survival outcomes in these patients. While our analysis provides data that might help select patients who will benefit most from surgical treatment, large, well-designed prospective studies are needed to confirm these observations. Based on our analysis and the current literature, we advocate using DFS as a complementary survival outcome parameter in research on GC.

Data Availability Statement:
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.