The effectiveness and safety of D2 plus para-aortic lymphadenectomy for resectable gastric cancer: a systematic review and meta-analysis

Background: Gastric cancer (GC) is among the malignant tumors of highest morbidity and mortality in the world, and has a prole of high lymph node metastasis rate. Lymph node clearance is a critical part of gastric cancer surgery, however, the extent of lymph node clearance, for example, whether to perform abdominal aortic lymph node dissection, remains considerably controversial. In this study, we performed a systematic review and meta-analysis to assess the effects of D2 plus para-aortic lymphadenectomy (PALD) on survival and postoperative complications in patients with GC. Methods: An electronic search was conducted through PubMed, Embase and cochrane library. The Q test and I 2 were used to assess heterogeneity. The publication bias was evaluated via funnel plots. All statistical analyses were performed using STATA 14.0 (STATA, College Station, TX). Results: 908 studies were retrieved via literature search and eight studies were nally included. There was no signicant difference between D2 and D2+PALD in the 5-year survival rate after surgery (HR: 1.00, 95% CI: 0.97-1.03, P = 0.897; I 2 = 64.9%). Besides, the 30-day mortality (RR: 1.17, 95% CI: 0.66-2.10, P = 0.590; I 2 = 0.0%) and the overall risk of postoperative complications (RR: 1.15, 95% CI: 0.83-1.59, P = 0.411; I 2 = 35.5%) were comparable between D2 and D2+PALD. Conclusion: Based on current literature body, compared with D2, D2+PALD does not prevail in terms of long-term survival or perioperative outcomes.


Background
Gastric cancer (GC) is among the malignant tumors of highest morbidity and mortality in the world. More than 1 million new cases along with an estimated 783,000 deaths (equivalent to one twelfth of all-cause death worldwide) were reported in 2018, making GC the fth leading cause of cancer incidence and third of cancer-related death 1,2 . The high-risk regions of gastric cancer are mainly in East Asia, namely China, South Korea and Japan [3][4][5] . Due to mild and atypical clinical manifestations and signs in earlier stage, GC patients are usually diagnosed in advanced stage, leading to poor prognosis 6,7 . The therapeutic strategies of GC mainly including surgery, radiotherapy, chemotherapy, etc., among which surgical treatment is especially important. Gastrectomy plus secondstage lymph node dissection (D2) has become a standard curative procedure for advanced GC in East Asian countries, especially in Japan 8, 9 . Five-year survival rates and postoperative complications are well-recognized indicators that are robust to assess the e cacy and safety of surgery. Lymph node metastasis can occur in the early stage of GC, for advanced GC patients, the risk is even higher, so lymph node dissection is particularly critical in GC surgery. However, the extent of lymph node clearance, for example, whether to perform abdominal aortic lymph node dissection, remains considerably controversial. According to the 4 th edition of Japanese Gastric Cancer Treatment Guidelines, for the tumor not invading esophagus, the removal of lymph nodes in the D2 lymphadenectomy for total gastrectomy includes No. 1-7, 8a, 9, 11p, 11d and 12a lymph nodes 9 . The removal of lymph nodes in the D2 plus para-aortic lymphadenectomy (PALD) including D2 lymphadenectomy plus No. 16 lymph node. Studies have shown that more than 20% of patients with advanced gastric cancer have No. 16 lymph node metastasis 10,11 . Some researchers believe that D2+PALD can extract more lymph nodes than traditional D2, achieving better therapeutic effects and improving patient survival 12-14, 3, 15-17 . On the contrary, some researchers believe that D2+PALD can lead to greater trauma, higher blood transfusion rates, longer operative time and hospital stay, etc., without increasing the patient's ve-year survival rate [18][19][20][21][22] . In this study, we evaluated existing literature and conducted a systematic review meta-analysis to compare the clinical effects of the two procedures.

Materials And Methods
Literature and search strategy Databases such as PubMed, Embase, and Cochrane library database were independently searched by two authors to identify original articles published until the end of June, 2019, using the following terms: "gastric", "stomach", "cancer", "tumor", "malignancy", "neoplasm", "para-aortic", "D3", "D4", and "lymphadenectomy".. Manual screening of references was also conducted. No language restriction was performed.

Study selection
This meta-analysis is performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) 23 . The inclusion and exclusion criteria for this meta-analysis were as follows: (1) All patients underwent radical gastrectomy with pathological diagnosis of primary GC. (2) Patients with distant metastasis, residual GC, severe cardiovascular, respiratory, liver or renal function diseases were excluded. (3) There should be su cient data, including survival rate, postoperative complications, etc., and studies with inadequate data were excluded. (4) Data and related results can be extracted directly or indirectly from the original study. Conference abstracts and reports were excluded due to incomplete information. (5) Only comparable original articles were included, excluding reviews, systematic reviews and meta-analyses, letter to editor, case reports, and single-arm studies. The study selection process was conducted by two independent reviewers in accordance with the following procedure: the retrieved literature were rstly reviewed via title and abstract screening, the remaining were then scrutinized by full-text.
Data extraction and quality assessment The following data were independently extracted by two authors: rst author, year of publication, country, study period, sample size, age, TNM stage, abdominal aortic lymph node condition, type of surgery, time of surgery, intraoperative blood loss, blood transfusion, postoperative complications, postoperative mortality and survival data. The Newcastle-Ottawa quality assessment scale (NOS) was used to evaluate the methodological quality of the included studies 24 . According to the evaluation criteria of the NOS scale, studies with more than 5 points were judged as moderate quality studies, while studies with 7 or more points were considered high quality studies.

Statistical analysis
Continuous data were processed using weighted mean difference (WMD) and 95% Con dence interval (CI), and postoperative complications and survival relative risk (RR) or risk ratio (HR) and 95% CI. Quantitative syntheses were conducted using random effects model to provide more conservative results, considering the inevitable heterogeneity. Cochran Q test and I 2 statistic were used to quantify and evaluate study heterogeneity. I 2 > 50% and/or P <0.1 indicate signi cant statistical heterogeneity 25,26 . Heterogeneity was detected by sensitivity analysis. Publication bias was evaluated via funnel plots. Since the number of studies included in each meta-analysis is less than 10, funnel plots, instead of Egger's test, were used to assess publication bias, considering the lack of statistical power 27 . All statistical analyses were performed using Stata 14.0 software (Stata Corporation, College Station, TX USA). P <0.05 was considered to be statistically signi cant.

Results
Search results and study characteristics The ow chart of the literature search is shown in Figure 1. Based on a prede ned search strategy, 908 studies were initially identi ed through database and manual search. After excluding duplicates, screening title and abstracts, 879 unrelated studies were excluded and the remaining 29 studies were further evaluated by full-text view. Among these 29 studies, 21 articles were excluded due to incomplete information or other reasons. In the end, the meta-analysis included eight eligible studies, all of which were retrospective observational cohort studies 28,18,29,3,30,20,19,31 .
All included studies were published between 2003 and 2014, with sample sizes ranging from 117 to 1792. Among these studies, three were from Japan, four from China, and one from Poland. According to the results of NOS assessment, among the eight studies, two with eight and three with seven were considered of high-quality, the remaining three scored six points and were deemed moderate-quality. The detailed baseline characteristics of all included studies are summarized in Table 1.

Impact of D2+PALD on long-term survival of GC patients
We rst evaluated the effect of D2+PALD on 5-year survival. A total of seven studies were pooled. The results showed no signi cant difference between the two groups (HR: 1.00, 95% CI: 0.97-1.03, P = 0.897; I 2 = 64.9%) (Fig. 2). No signi cant asymmetry was observed in funnel plot.
No signi cant asymmetry was observed in funnel plots of each quantitative syntheses.

Publication bias
According to the results of the funnel plot observation, there is no signi cant publication bias in this study.

Discussion
Surgical treatment is the only way to achieve radical cure for resectable GC 32 ,and has been signi cantly improved through decades of development. The Japan Gastric Cancer Association was the rst to de ne the three stations of GC in 1998, and de ned the terminology of the operations as D1, D2, and D3 according to the range of lymph nodes in the operation 33 , and updated the lymph node dissection range of GC according to the extent of gastric resection in 2011 34,35 . Wu et al published a study comparing D1 and D2 in 2006, and the results showed that the 5-year survival rates of patients with advanced GC after D1 and D2 were 53.6% and 59.5%, respectively (P=0.041). Compared with D1 surgery, patients with advanced GC undergoing D2 surgery achieved better survival outcome 36 . Plenty of retrospective studies have shown that D2 resection can improve survival in patients with higher T &N-staged tumors, and in this regard, D2 surgery has been widely used to treat advanced GC in Asia [37][38][39][40][41][42][43] .
Controversy lies in whether to perform D2 or D2+ PALD surgery on patients with advanced GC 44 . According to relevant studies, once gastric tumors invade the subserosal (T3 phase), serosal (T4a phase) or adjacent structures (T4b phase), the metastatic risk of para-aortic lymph nodes (PANs) increases to 10% to 30% [45][46][47] . PALD may help to remove potential metastases, and in the meantime, surely help to collect more lymph nodes, which is critical to staging and prognosis prediction [48][49][50] 52 . However, Hu et al showed that D2 plus PALD was not signi cantly superior, survival-wise, to D2 in patients with T3-4, N2 staging 18 . Besides, Sasako et al showed that D2 lymphadenectomy plus prophylactic PALD did not improve the survival rate of curable GC compared with D2 lymphadenectomy, even though D2+PALD did not increase the risk of anastomotic leakage, pancreatic stula, and abdominal infection 28 . In line with the conclusion of the work by Sasako et al, the results of our comprehensive analysis on prophylactic and therapeutic D2+PALD showed that, compared with D2, D2+PALD did not improve the patients' long-term survival, but nonetheless, did not increase the risk of postoperative complications.
It is well known that D2+PALD is more di cult to perform than D2, with longer operative time, more bleeding, and longer hospital stays, requiring experienced surgeons 18,3 . However, with the development of modern technology, the extensive use of advanced medical equipment has reduced the operation time, operative mortality and surgery-related morbidity. D2+PALD can also perform as safely as D2 in professional medical centers with well-trained surgeons 45 .
Several limitations reside in our work. First, all the included studies are retrospective, which affects the level of evidence of this study to some extent. Second, we only focused on overall survival as e cacy outcome in this study, other outcomes like recurrence free survival and rate of recurrence were not covered due to lack of data. Third, we were unable to access the personalized data of included studies to perform any further analyses to adjust for confounding factors.
Fourth, the surgical results are closely related to the surgeon's experience. The expertise and approach of surgeons differs and matters, in which resides heterogeneity 54 . Fifth, the postoperative intervention, such as postoperative radiotherapy and chemotherapy, on the patients also has a great impact on the outcome, however, due to the lack of relevant data, we did not account for this confounding factor. Sixth, the number of studies included is relatively small, especially for the studies concerning therapeutic D2+PALD, which might affect the credibility of the results of the analyses concerning long-term survival and perioperative complications. Besides, the lack of comparable studies on therapeutic D2+PALD makes it impossible to draw a conclusion, in this regard, more elaborately designed multi-center studies with larger sample size are needed in the future to illuminate the advantages and disadvantages of D2+PALD more comprehensively.

Conclusion
In summary, D2 + PALD can be as safe as standard D2, but requires experienced surgeons to operate. Prophylactic D2 + PALD did not increase postoperative complications compared to standard D2, but no long-term survival bene t was achieved. The role of therapeutic D2 + PALD requires more researches to further con rm. More well-designed multi-center studies of larger scale are needed in the future to explore D2 + PALD. Yes. Figure 1 The ow diagram of study selection.

Figure 2
Forest plots evaluating the impact of D2+PALD on long-term survival.  Funnel plot for the analysis of long-term survival.