Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma

META ANALYSES

Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma Muhammed Ashraf Memon, MBBS, MA Clin Ed, DCH, FRCSI, FRCSEd, FRCSEng, FRACS,∗ †‡§ Manjunath S. Subramanya, MBBS, MRCS, Shahjahan Khan, PhD,¶ Md Belal Hossain, MSc,¶ Emma Osland, BHSc, M Phil,∗ ¶ and Breda Memon, RGN, LLB, PGCEd∗ Objectives: To conduct a meta-analysis of randomized controlled trials evaluating the efficacy and drawbacks of limited (D1) versus extended lymphadenectomy (D2) for proven gastric adenocarcinoma. Methods: A search of Cochrane, Medline, PubMed, Embase, Science Citation Index and Current Contents electronic databases identified randomized controlled trials published in the English language between 1980 and 2008 comparing the outcomes of D1 versus D2 gastrectomy for gastric adenocarcinoma. The meta-analysis was prepared in accordance with the Preferred Reporting Items for Systematic reviews and Meta-analyses statement. The 6 outcome variables analyzed included length of hospital stay; overall complication rate; anastomotic leak rate; reoperation rate; 30-day mortality rate and 5-year survival rate. Random effects meta-analyses were performed using odds ratios (OR) and weighted mean differences (WMD). Results: Six trials totaling 1876 patients (D1 = 946, D2 = 930) were analyzed. In 5 of the 6 outcomes the summary point estimates favored D1 over D2 group with a statistically significant reduction of (i) 6.37 days reduction in hospital stay (WMD –6.37, confidence interval [CI] –10.66, –2.08, P = 0.0036); (ii) 58% reduction in relative odds of developing postoperative complications (OR 0.42, CI 0.27, 0.66, P = 0.0002); (iii) 60% reduction in anastomotic breakdown (OR 0.40, CI 0.25, 0.63, P = 0.0001); (iv) 67% reduction in reoperation rate (OR 0.33, CI 0.15, 0.72, P = 0.006); and (v) 41% reduction in 30-day mortality rate (OR 0.59, CI 0.40, 0.85, P = 0.0054). Lastly there was no significant difference in the 5-year survival (OR 0.97, CI 0.78, 1.20, P = 0.7662) between D1 and D2 gastrectomy patients. Conclusions: On the basis of this meta-analysis we conclude that D1 gastrectomy is associated with significant fewer anastomotic leaks, postoperative complication rate, reoperation rate, decreased length of hospital stay and 30day mortality rate. Finally, the 5-year survival in D1 gastrectomy patients was similar to the D2 cohort. (Ann Surg 2011;253:900–911)

From the ∗ Department of Surgery, Ipswich Hospital, Ipswich, Queensland, Australia; †Mayne Medical School, School of Medicine, University of Queensland, Brisbane, Queensland, Australia; ‡Faculty of Health Sciences and Medicine, Bond University, Gold Coast, Queensland, Australia; §Faculty of Health and Social Science, Bolton University, Bolton, Lancashire, UK; and ¶Department of Mathematics and Computing, Australian Centre for Sustainable Catchments, University of Southern Queensland, Toowoomba, Queensland, Australia. Reprints: Professor Muhammed Ashraf Memon, FRCS, FRACS, Ipswich Hospital, Chelmsford Avenue, Ipswich, Queensland, Australia. E-mail: mmemon@ yahoo.com. The authors did not receive any financial support or commercial sponsorship. Conception and design: Muhammed Ashraf Memon. Furthermore, he takes full responsibility for the integrity of the work as a whole, from inception to published article; Acquisition, compilation, and interpretation of the data: Manjunath S. Subramanya, Breda Memon, Muhammed Ashraf Memon; Shahjahan Khan and Md Belal Hossain were responsible for statistical analysis of this article including writing R codes, interpretation of the results and computer output of the data. All authors were involved in drafting the manuscript and revising it critically for important intellectual content and have given final approval of the version to be published. Furthermore, all authors have participated sufficiently in the work to take public responsibility for its content. C 2011 by Lippincott Williams & Wilkins Copyright  ISSN: 0003-4932/11/25305-0900 DOI: 10.1097/SLA.0b013e318212bff6

900 | www.annalsofsurgery.com

G

astric adenocarcinoma is a locoregional disease with a high propensity for nodal metastasis. Therefore, nodal status remains 1 of the most critical independent predictors of patient’s survival after gastrectomy for this disease.1,2 Studies have shown that lymph node involvement occurs in 3% to 5% of cases when gastric cancer is limited to the mucosa; 11% to 25% of cases for those limited to the submucosa; 50% for T2 cancers and 83% for T3 cancers.3,4 D1 gastrectomy entails removing lymph nodes adjacent to the stomach (stations 1 to 6, N1 level), D2 dissection extends this resection to include the nodes around the branches of the celiac axis (stations 7 to 11, N2 level) whereas D3 dissection removes lymph nodes from station 12 to 20 (N3 level) as per second edition of Japanese Classification of Gastric Carcinoma.5 Metastases are expected be found in 8 to 31%, and 40% of N2 lymph nodes in T2 and T3 tumors respectively. Therefore for T2 and T3 tumors, D1 dissection may inadequately stage a substantial portion of patients and leads to noncurative intervention leading to poor patient outcome. In spite of these facts, the relative merits of gastrectomy with limited (D1) versus extended lymphadenectomy (D2) as an oncological treatment of gastric adenocarcinoma remains contentious and unpopular except in Asian countries.6 This is because D2 resection is technically demanding, has a steep learning curve and its benefits remain doubtful based on a number of randomized controlled trials (RCTs) published to date.7–13 The majority of Western surgeons therefore continue to practice D1 resection achieving a 5-year survival rate of 10% to 30%.9 On the other hand surgeons in Asian surgical centers (most notably Japan) routinely perform D2 gastrectomy achieving an impressive 5year survival of 50% to 60% with a low morbidity and mortality.14,15 However, the greatest criticism of these reports from Japan (and the East) demonstrating such an impressive benefit and modest morbidity and mortality from D2 resection has been the retrospective nature of the data. Nonetheless the continuing debate on D1 versus D2 has lead some researchers to address this issue objectively in the form of well-designed RCTs,7–13 however this has further polarized the literature on this subject due to conflicting results obtained. Therefore this meta-analysis which was prepared in accordance with the “Preferred Reporting Items for Systematic reviews and Meta-analyses” statement16 has been undertaken to develop a better understanding of these 2 approaches for the treatment of gastric cancer by pooling data from all of the available RCTs published to date.

MATERIAL AND METHODS Literature Search and Study Selection RCTs of any size that compared outcomes of D1 with D2 gastrectomy for the treatment of gastric adenocarcinoma between January 1980 and December 2008 in full peer-reviewed journals were considered for inclusion (Fig. 1). Only those studies which have reported on at least 1 clinically relevant outcome were included. Unpublished RCTs, nonrandomized prospective and retrospective comparative trials were excluded. Furthermore abstracts of RCTs presented at national and international meetings were also excluded to prevent the duplication of data. Studies not published in English were also Annals of Surgery r Volume 253, Number 5, May 2011

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma

FIGURE 1. Flow chart providing information through the different phases of systematic review. excluded. Finally, studies which reported on gastric cancers other than adenocarcinoma such as lymphomas were excluded because of different biological behavior and treatment options for these tumors.

Type of Participants Adult patients requiring surgical intervention for histologically proven gastric adenocarcinoma were the target population for this meta-analysis.

Types of Intervention Surgical interventions assessed the differences in outcomes between D1 gastrectomy (limited lymphadenectomy) and D2 gastrectomy (extended lymphadenectomy) as curative management for gastric adenocarcinoma.

Types of Outcome Measures Analyzed The 6 outcome variables analyzed included (a) length of hospital stay; (b) overall complication rate; (c) anastomotic leak rate; (d) reoperation rate; (e) 30-day mortality rate; and (f) 5-year survival rate. These outcomes were thought to be important because they exert influence over practical aspects of surgical practice and policy decisions within institutions.

Data Collection Trials were identified by conducting a comprehensive search of Medline, Embase, Science Citation Index, Current Contents and PubMed databases, using medical subject headings “D1 gastrectomy,” “D2 gastrectomy,” “gastric cancer,” “comparative study,” “prospective  C 2011 Lippincott Williams & Wilkins

studies,” “randomized controlled trials,” “random allocation,” and “clinical trial”. Manual search of the bibliographies of relevant articles was also carried out to identify trials for possible inclusion (Fig. 1). Data extraction and critical appraisal were carried out by 3 authors (MSS, BM, and MAM) for compliance with inclusion criteria and methodological quality. Standardized data extraction forms17 were used by authors to independently and blindly summarize all the data available in the RCTs meeting the inclusion criteria. The authors were not blinded to the source of the document or authorship for the purpose of data extraction. The data were compared and discrepancies were addressed with discussion until consensus was achieved.

Methodological Quality Evaluation of methodological quality of identified studies was conducted using the Jadad scoring system18 in which each study was assigned a score of between zero (lowest quality) and 5 (highest quality) based on reporting of randomization, blinding, and withdrawals reported within the study period.

STATISTICAL ANALYSIS Meta-analyses were performed using odds ratios (ORs) for binary outcomes and weighted mean differences (WMDs) for continuous outcome measures. A slightly amended estimator of OR was used to avoid the computation of reciprocal of zeros among observed counts/values in the calculation of the original OR.19 Random effects models using the inverse variance weighted method approach were used to combine the data.20 Heterogeneity among studies was assessed using the Q statistic proposed by Cochran20–22 and I2 index www.annalsofsurgery.com | 901

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Memon et al

introduced by Higgins and Thompson.23,24 For the computations of the confidence intervals (CI) estimates of mean and standard deviation are required. However, some of the published clinical trials did not report the mean and standard deviation, but rather reported the size of the trial, the median and range. From these available statistics, estimates of the mean and standard deviation were obtained using formulas proposed by Hozo et al.25 Funnel plots were obtained to determine the presence of publication bias in the meta-analysis. Both total sample size and precision (reciprocal of the standard error) were plotted against the treatment effects (log OR for the dichotomous variables and WMD for the continuous variables).20,26,27 All estimates were obtained using computer programs written in R.28 All plots were obtained using the “rmeta” package.29 In the case of tests of hypotheses, the article reports the value of the test statistics and associated P-values for different study variables. Effect measures have been considered to be statistically significant if the P-value is less than or equal to 5%. Ninety-five percent CI are presented for each effect measure.

RESULTS

There was almost a perfect agreement (κ = 0.99) between the 3 authors (MSS, BM, MAM) regarding the inclusion and exclusion of various RCTs. On the basis of this agreement, a total of 6 RCTs (Europe = 3, Asia = 2, Africa = 1)7–12 that included 1876 gastrectomies (D1 = 946 and D2 = 930) were considered suitable for meta-analysis (Table 1).

Description of Various RCTs Including Inclusion and Exclusion Criterias The study size in half of these RCTs was less than 100 patients in each arm except for the United Kingdom10 (200 each arm), Dutch9 (>300 each arm), and Taiwanese12 (>100) trials (Table 1). All the patients had histologically proven gastric adenocarcinoma by endoscopic biopsies before being enrolled in the trial. They also had a preoperative cancer staging predominantly via computerized tomography, and less frequently with ultrasound scan. The patients were also excluded from the study if distant metastases involving liver or lungs were found during preoperative staging; if they had inoperable gastric cancer at laparotomy; were >80 or 95%.47–49 Additionally Japanese patients tend to be younger and have fewer co-morbidities as compared with their Western counterparts and therefore radical gastrectomy in Asian practice is the norm rather than the exception. Because of the above reasons and the fact that Japanese surgeons have been practicing D2 resection for a longer period of time has lead to an extensive surgical experience in the more radical forms of gastrectomy. This implies that the superior surgical results reported in the Asian literature for D2 gastrectomy are achieved as a result of a greater degree of technical experience through increased exposure over a prolonged period of time.  C 2011 Lippincott Williams & Wilkins

A number of RCTs7–12 have been undertaken to investigate the issues of efficacy and drawbacks of limited versus extended lymphadenectomy. Two of these largest RCTs have been published from the European centers9,10 where the experience of participating surgeons in radical gastric surgery has been minimal. The landmark trial from the Netherlands conducted an evaluation of their first 389 patients to identify factors leading to protocol deviations, ie, noncompliance (ie, performance of less dissection than specified) and contamination (ie, performance of more extensive dissection than specified).50 These violations have an impact on the feasibility of the trial and the reliability of TNM staging in general. The proportion of patients operated on with curative intent was 66% (ie, 255 patients). Noncompliance occurred in 84% of D1 and D2 cases. In D1 gastrectomy, noncompliance was only minor (1 to 2 stations) in 75% of the cases (vs. 51% in D2), whereas in D2 gastrectomy, major www.annalsofsurgery.com | 905

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Memon et al

FIGURE 5. Forest plot for anastomotic leak favoring D1.

FIGURE 6. Forest plot for reoperation favoring D1.

FIGURE 7. Forest plot for mortality favoring D1. noncompliance (≥3 stations) occurred in 33% of the cases (vs. 9% in D1). Contamination occurred in 48% of the D1 and 52% of D2 cases and the magnitude of contamination was modest and distributed over D1 and D2 gastrectomy. In D1 and D2 cases, contamination was restricted to 1 or 2 stations in 45% of the cases. Contamination occurred at 3 stations in only 3% of the D1 cases. The results showed substantial 906 | www.annalsofsurgery.com

protocol violations by the surgical-pathologic teams due to extending or limiting lymphadenectomy, leading either to undertreatment or overtreatment in both the D1 and D2 groups impacting both short and long-term results (ie, locoregional recurrences and overall survival). Finally, the protocol deviations also have an impact on accurate TNM staging (stage migration).51 Therefore, one can summarize that  C 2011 Lippincott Williams & Wilkins

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma

FIGURE 8. Forest plot for 5-year survival indicating no advantage for either group. TABLE 3. Summary Statistics of Pooled Data Comparing D1 versus D2 Gastrectomy Test for Overall Effect Outcome Variables Length of hospital stay Complication rate Anastomotic leak rate Reoperation rate 30-day mortality rate 5-year survival rate

Test for Heterogeneity

Pooled OR or WMD (95% CI)

Z

P value

Q

P value

I2 Index (CI)

−6.37 (−10.66, −2.08) 0.42 (0.27, 0.66) 0.40 (0.25, 0.63) 0.33 (0.15, 0.72) 0.59 (0.40, 0.85) 0.97 (0.78; 1.20)

−2.9097 −3.7635 −3.9633 −2.7485 −2.7795 −0.2973

0.0036 0.0002 0.0001 0.006 0.0054 0.7662

36.34 11.85 1.72 3.52 1.86 1.67

0.0001 0.0369 0.8868 0.3179 0.8684 0.797

86.2% (72.2%; 93.2%) 57.8% (0%; 82.9%) 0% (0%; 26.1%) 14.8% (0%; 87%) 0% (0%; 31.7%) 0% (0%; 50.1%)

surgical inexperience and the learning curve occurring throughout the trial may have impacted the overall results in both arms of these studies. Other confounding factors include patient population and their selection, operative techniques and outcome descriptors have fuelled the ongoing debate on the risks and benefits of D1 versus D2 gastrectomy despite reasonable attempts being made by the authors to provide trials of high quality. The authors of this article have undertaken a meta-analytical review based on the available RCTs data in an attempt to provide some clarification. To date D2 gastrectomy has shown better results than D1 gastrectomy in mainly retrospective studies.39,40,52 However, this has not been seen in 2 of the largest RCTs published from Europe.9,10 In the Dutch trial,53 where the median duration of follow-up was 15 years, survival rates were 21% for D1 and 29% for D2 groups (which were not statistically significant). The risk of local and regional recurrence were 22% and 19% for D1 and 12% and 13% for D2 which were not significantly different as well. These results when viewed in the context of noncompliance or contamination in the extent of lymphadenectomy performed in Dutch RCT31,50,54 needs to be interpreted with extreme caution. All 6 studies7–12 have reported on the length of hospital stay. Of these, 5 had a significantly longer hospital stay for D2 group compared with D1 group (Fig. 3). The reasons for the longer stay in D2 cohort could be multiple and include (a) prolonged and more complex surgical procedures; (b) more perioperative complications; (c) the higher rate for reoperation; and (d) greater amount of blood transfusions.7–9,12 Four of 6 RCTs7–9,12 have consistently shown a statistically significant higher requirement for transfusion in D2 patients. Perioperative blood transfusion has a negative impact on the length of hospital stay and increases morbidity and mortality. This  C 2011 Lippincott Williams & Wilkins

is because it exposes the patients to acute or delayed adverse effects such as acute lung injury, volume overload, hypothermia, graft versus host disease, and immunomodulatory effects. The greater length of stay has a negative impact on cost and efficiency in the surgical units, exposing patients to further hospital acquired complications and imparts a higher overall cost to the procedure. On the other hand, early-discharge is associated with lower direct medical, direct nonmedical, and indirect costs leading to significant cost savings. Furthermore, early discharge reduces the pressure on hospital bed occupancy. This in turn increases the efficiency and turnover of elective surgical procedures with positive impact on health care finances. These issues (ie, of cost and efficiency) have not been addressed in any of the RCTs. All 6 trials7–12 have reported significantly higher complication rates for D2 group (Fig. 4). The pooled data is also in line with these RCTs (Table 3). Complications ranged from simple wound infection, intraabdominal abscess formation to anastomotic leak. Different studies have enumerated complications on different bases, which made summation of the results difficult. Most of the studies have reported complication under 2 major headings, surgical and nonsurgical (Table 4). Among the surgical complications, intraabdominal or subphrenic abscess formation was not only commonly seen but also required reoperation in the majority of cases. The high incidence of subphrenic abscess in the D2 cohort may be related to pancreatic injury (duct leakage, direct pancreatic injury or ischaemia).7–12 The intraabdominal sepsis was also responsible for other complications such as secondary haemorrhages and death.11 These complications were seen exclusively in D2 group. Similarly in various nonrandomized comparative studies, D2 patients seem to have a higher incidence www.annalsofsurgery.com | 907

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Memon et al

TABLE 4. Surgical and Nonsurgical Complications of RCTs Authors

Surgical Complications

Dent et al7

Robertson et al8

Bonenkamp et al9

Cuschieri et al10

Degiuli et al11

Wu et al12

Nonsurgical Complications

Type Wound Infection Retrocolic intestinal hernia Subphrenic abscess Gastrojejunal leak Colonic infarction Subphrenic abscess Haemorrhage Oesophagojejunal leak Haemorrhage Wound infection/dehiscence Anastomotic leaks Intraabdominal infection Pancreatic leak Haemorrhage Anastomotic leak

D1 0 0 0 0 0 0 0 0 8 15 16 30 3 6 11

D2 0 1 1 1 1 14 3 3 15 30 30 55 10 4 26

Duodenal stump leak Pancreatic leak Colonic perforation Acute pancreatitis Haemorrhage Abdominal abscess Abdominal abscess Anastomotic leaks Duodenal stump leak Pnacreatic fistula Chylous leak Bleeding Cholecystitis Wound infection Splenic injury

0 1 1 0 2 0 0 0 0 0 0 0 1 2 6

1 1 0 2 1 2 9 5 1 1 1 1 0 5 19

of intraabdominal sepsis and other complications.55,56 Among the nonsurgical complications, pulmonary and cardiac were the 2 most common ones seen in both groups. All 6 trials7–12 have reported on anastomotic leak rate (Fig. 5) which was higher in the D2 group. The overall pooled data also confirmed this finding. The United Kingdom10 and the Dutch9 trials, however have not provided any information on the site(s) of these leaks or following which procedures (such as distal or total gastrectomy) these leaks occurred. In the Hong Kong study8 3 anastomotic leaks were recorded, all in D2 group at oesophagojejunal junction (ie, after a total gastrectomy) which were treated conservatively with total parenteral nutrition with favorable outcome. In the Taiwanese trial,12 5 leaks were recorded all in D2 group. Three were at gastrojejunal site, whereas 2 at oesophagojejunal site. All of these were minor according to the authors and were managed conservatively with nutritional support. Disappointingly not all the RCTs (especially the 2 larger ones9,10 ) have documented their operating time for the D1 and D2 groups. Dent et al,7 Robertson et al,8 and Wu et al12 have shown statistically significant longer operating time for D2 gastrectomy. A protracted operating time exposes the patient to a longer duration of anesthesia, and a greater risk of thermic, thromboembolic, cardiac, 908 | www.annalsofsurgery.com

Type Pulmonary

D1 3

D2 3

DVT PE

0 0

0 0

Pulmonary Cardiac UTI Thromboembolic

23 14 6 2

49 17 5 7

Pulmonary Cardiac PE Pulmonary Cardiac

5 2 1 2 2

8 8 1 7 0

Cardiac Post-op psychosis Renal failure Thermal injury

0 1 1 1

1 0 0 0

and respiratory complications. In addition, it negatively impacts on the efficiency of the operating staff and the operating complex as a whole, and imparts a higher overall cost to the procedure. Four of the 6 trials7–9,11 have documented reoperation rate (Fig. 6). The pooled data revealed a statistically significant higher reoperation rate for D2 group. The most common cause reported for reoperation was a subphrenic abscess. Other causes include intraperitoneal haemorrhage, anastomotic leaks, duodenal stump leak, colonic perforation, colonic infarction, retro-colic hernia, intraabdominal abscesses, and acute pancreatitis. The reasons for reoperation in the Dutch study9 are not given and the UK study10 failed to provide any information on this variable all together. The commonest cause for reoperation in the Hong Kong study8 was subphrenic abscess, all in D2 group. Two of 4 studies7,8 had no incidence of reoperation in the D1 group. All 6 trials7–12 have reported on 30-day mortality rates (Fig. 7). Pooled data showed a statistically significant higher mortality rate for D2 cohort compared with their D1 counterpart. The inadequacy of pretrial training in D2 gastrectomy may have had a significant impact on the mortality and morbidity as explained by minimal surgical complications from Italian and Tiawanese trials11,12 when compared with their UK, Dutch, Hong Kong, and South African counterparts7–10  C 2011 Lippincott Williams & Wilkins

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma

(Table 4). It is entirely possible that surgical experience may have been an independent risk factor for predicting mortality in various trials. Other factors which may have had an impact on the mortality and morbidity was pancreatic or splenic resection, which was inconsistent throughout the trials. Pancreatosplenectomy recommended as part of the D2 gastrectomy in the second edition of Japanese classification5,57 was not routinely performed in all the RCTs.7,10,12–14 Although surgeons in Hong Kong and Taiwanese trials8,12 performed pancreatic and splenic resection routinely, the rest of the surgeons performed these resections only when the upper two-thirds of the stomach was involved7,9,10 or if the cancer had extended into these organs.11 Cuschieri et al10 felt that higher morbidity rate in D2 cohort were secondary to pancreatosplenic resection which was similarly seen in the Tawianese trial.12 However Bonenkamp et al9 found no such association. Immunological abnormalities especially after splenectomy such as higher rate of septic complications, higher rate of blood transfusion or pancreatic leakage may all contribute to higher morbidity seen after pancreatosplenectomy. A number of authors have suggested that spleen and pancreatic preservation during gastric cancer resection produces not only better short-term outcome but have no impact on 5-year survival. Csendes et al58 have produced the only RCT in the Western countries addressing specifically the risk and benefits of splenectomy in D2 patients. In several Japanese studies59–63 it has been demonstrated that micrometastasis in station 10 lymph nodes varies between 12% and 20% and therefore avoiding routine splenectomy may compromise tumor free and overall survival in patients with gastric cancer. Csendes et al58 detected micrometastasis in 9% of their patients. However, the 5-year survival was not significantly different in their patients with or without splenectomy. On the other hand, they found postoperative septic complications to be significantly higher in patients with the splenectomy. In yet another RCT from Korea64 comparing 104 with and 103 patients without splenectomy, the authors have shown similar results to the Csendes et al58 study. A number of other studies61,63,64–68 evaluating the role of routine splenectomy during gastrectomy have shown similar results in the past, although none were of randomized nature. Two large retrospective studies67,68 involving almost 500 patients with or without splenectomy after total gastrectomy found that splenectomy failed to improve the long-term survival but instead increased the postoperative complications. In a retrospective Finnish study56 comparing D1 versus D2–3 gastrectomy, of the 22 patients who underwent pancreas resection, 40.9% had at least 1 surgical complication postoperatively compared with 22.8% without pancreas resection. Two of the RCTs9,10 although not designed to specifically answer the impact of pancreatic or splenic resection in D2 gastrectomy, have shown unfavorable results with caudal pancreatectomy and concomitant splenectomy in a subset analysis. The overall consensus is that routine splenectomy and distal pancreatectomy during D2 dissection has no long-term survival benefit and may even be counter productive. However it may be performed for selected patients with T3 tumors or direct invasion or metastasis at the splenopancreatic hilum. The most important outcomes after any cancer treatment is the actual 5-year survival rate. Interestingly this meta-analysis does not show any difference in the 5-year survival rate between the 2 groups (Fig. 8) which questions the very rationale for performing aggressive D2 gastrectomy for gastric cancer with its higher incidence of morbidity and mortality. There are some limitations associated with this meta-analysis. First, publication bias was detected on funnel plot analysis for the outcomes of length of stay and postoperative complications. This is of concern as the selective publication of studies may result in pooled effect sizes obtained from studies exclusively located from the published scientific literature demonstrating a more significant  C 2011 Lippincott Williams & Wilkins

result in terms of the magnitude of harm or benefit of an intervention than in actuality.69,70 However, it should be noted there are number of limitations to the assessment of publication bias in this meta-analysis. First, the numbers of studies included are inadequate to sensitively detect such a bias.22,24 Second, a random effects model of metaanalysis was utilized for this analysis, and this model is known to exaggerate the presence of publication bias due to attributing heavier weighting to smaller studies than occurs in the fixed effects model of meta-analysis. Finally, funnel plots are not an ideal method for assessing71,72 publication bias due to their inability to differentiate between forms of bias present, however they remain the accepted method for assessment of publication bias.73 A second possible limitation within this meta-analysis is the presence of heterogeneity detected within several outcomes, specifically length of stay and postoperative complications. Although some degree of heterogeneity is inevitable in a medical meta-analysis due to the realities of clinical practice,69,74 the degree of between-study heterogeneity present may undermine the quality and legitimacy of the results obtained.75 A visual assessment of the forest plots for the 2 outcomes with significant degrees of heterogeneity clearly demonstrate that the Robertson et al study8 reports results considerably stronger in favor of the D1 procedure than any other included study. A review of the included articles failed to demonstrate significant differences in the practices reported that may indicate the Robertson et al article8 was inappropriate for inclusion with this meta-analysis: true heterogeneity may therefore account for the differences observed. In view of this it is also interesting to consider that it was in these outcomes that publication bias was detected, as funnel plots are also sensitive to the presence of true heterogeneity.27,72 The exclusion of studies published in languages other than English is another potential limitation to this work, particularly in view of the fact that the main advocates for the D2 procedure are from Japan and largely published in Japanese. The latter can have little bearing on the outcomes of this meta-analysis as the Japanese studies supporting the use of D2 procedures would not have met the RCT inclusion criteria for this meta-analysis. The impact of the inclusion of English only studies was also challenged by rerunning the search terms without this limit applied–no eligible studies in languages other than English were located. Finally, the small number of studies included in this metaanalysis remains a largely unavoidable limitation of this and many other meta-analyses conducted in surgical fields. Producing a large number of good quality clinical trials is expensive, logistically difficult and time consuming,76 and as such only a limited number with comparable interventions and methods are generally conducted on any given topic. This makes the production of a reliable metaanalysis difficult as limited study numbers have implications for statistical power for all statistical tests conducted, including those for determining point estimates, confidence interval, heterogeneity test, and publication bias.24,69,72,77

CONCLUSIONS On the basis of this meta-analysis and its limitations, we conclude that D1 is associated with significantly fewer anastomotic leaks, postoperative complications, reoperations, decreased length of hospital stay and 30-day mortality rate compared with D2 gastrectomy. Furthermore, the 5-year survival rate in D1 gastrectomy patients was similar to the D2 cohort. Therefore, at present best clinical evidence comparing D1 and D2 surgery does not favor the D2 resection, We therefore need to consider whether further well-designed RCTs with more rigorous quality control in performing D2 gastrectomy are necessary or the addition of other modalities such as chemotherapy may provide a better patient outcome as opposed to more extensive surgical operation. Other important issues that need urgent attention www.annalsofsurgery.com | 909

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Memon et al

include (a) surgical experience in performing D2 resections to prevent breaches and contaminations; (b) number of gastrectomies needed to be performed yearly to participate in such trials; (c) standardization of D2 dissection (eg, with or without pancreatosplenectomy); (d) training requirement and standards; (e) higher versus lower volume centers (ie, volume–quality or outcome relationship) participating in the trials; (f) accurate pathological staging; and (g) cost-benefit analysis. Finally, with the developing appreciation of the biomolecular basis for these cancers obtained with the genetic decoding technology, the future of this debate and surgical procedures may be rendered void as biomolecular agents become the first line of treatment for these cancers.

REFERENCES 1. Seto Y, Nagawa H, Muto T. Impact of lymph node metastasis on survival with early gastric cancer. World J Surg. 1997;21:186–189. 2. Siewert JR, B¨ottcher K, Stein HJ, et al. Relevant prognostic factors in gastric cancer: ten-year results of the German Gastric Cancer Study. Ann Surg. 1998;228:449–461. 3. O˜nate-Oca˜na LF, Aiello-Crocifoglio V, Mondrag´on-S´anchez R, et al. Survival benefit of D2 lympadenectomy in patients with gastric adenocarcinoma. Ann Surg Oncol. 2000;7:210–217. 4. De Gara CJ, Hanson J, Hamilton S. A population-based study of tumor-node relationship, resection margins, and surgeon volume on gastric cancer survival. Am J Surg. 2003;186:23–27. 5. Japanese Gastric Cancer Association. Japanese Classification of Gastric Carcinoma, 2nd English Edition. Gastric Cancer. 1998;1:10–24. 6. Mansfield PF. Lymphadenectomy for gastric cancer. J Clin Oncol. 2004;22:2759–2761. 7. Dent DM, Madden MV, Price SK. Randomized comparison of R1 and R2 gastrectomy for gastric carcinoma. Br J Surg. 1988;75:110–112. 8. Robertson CS, Chung SC, Woods SD, et al. A prospective randomized trial comparing R1 subtotal gastrectomy with R3 total gastrectomy for antral cancer. Ann Surg. 1994;220:176–182. 9. Bonenkamp JJ, Songun I, Hermans J, et al. Randomised comparison of morbidity after D1 and D2 dissection for gastric cancer in 996 Dutch patients. Lancet. 1995;345:745–748. 10. Cuschieri A, Weeden S, Fielding J, et al. Patient survival after D1 and D2 resections for gastric cancer: long-term results of the MRC randomized surgical trial. Surgical Co-operative Group. Br J Cancer. 1999;79:1522–1530. 11. Degiuli M, Sasako M, Calgaro M, et al. Italian Gastric Cancer Study Group. Morbidity and mortality after D1 and D2 gastrectomy for cancer: interim analysis of the Italian Gastric Cancer Study Group (IGCSG) randomised surgical trial. Eur J Surg Oncol. 2004;30:303–308. 12. Wu CW, Hsiung CA, Lo SS, et al. Randomized clinical trial of morbidity after D1 and D3 sugery for gastric cancer. Br J Surg. 2004;91:283–287. 13. Wu CW, Hsiung CA, Lo SS, et al. Nodal dissection for patients with gastric cancer: a randomised controlled trial. Lancet Oncol. 2006;7:309–315. 14. Soga J, Kobayashi K, Saito J, et al. The role of lymphadenectomy in curative surgery for gastric cancer. World J Surg. 1979;3:701–708. 15. Maruyama K, Okabayashi K, Kinoshita T. Progress in gastric cancer surgery in Japan and its limits of radicality. World J Surg. 1987;11:418–425. 16. Moher D, Liberati A, Tetzlaff J, et al. The PRISMA Group. Preferred reporting items for systematic reviews and meta-analysis: the PRISMA statement. PLoS Med. 2009;6:e1000097. 17. Moher D, Cook DJ, Eastwood S, et al. Improving the quality of reports of metaanalyses of randomised controlled trials: the QUOROM statement. Quality of Reporting of Meta-analyses. Lancet. 1999;354:1896–1900. 18. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary? Control Clin Trials. 1996;17: 1–12. 19. Liu IM, Agresti A. Mantel-Haenszel-type inference for cumulative odds ratios with a stratified ordinal response. Biometrics. 1996;52:1223–1234. 20. Sutton AJ, Abrams KR, Jones DR, et al. Methods for Meta-Analysis in Medical Research. John Wiley: Chichester; 2000. 21. Wermuth N, Cochran WG. Detecting systematic errors in multi-clinic observational data. Biometrics. 1979;35:683–686. 22. Hedges LV, Olkin I. Statistical Methods for Meta Analysis. Orlando: Academic Press; 1985.

910 | www.annalsofsurgery.com

23. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002,15;21(11):1539–1558 24. Huedo-Medina TB, S´anchez-Meca J, Mar´ın-Mart´ınez F, et al. Assessing heterogeneity in meta-analysis: qtatistic or I2 index? Psychol Methods. 2006;11:193–206. 25. Hozo SP, Djulbegovic B, Hozo I. Estimating the mean and variance from the median, range, and the size of a sample. BMC Med Res Methodol. 2005;5:13. 26. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315:629–634. 27. Tang JL, Liu JL. Misleading funnel plot for detection of bias in meta-analysis. J Clin Epidemiol. 2000;53:477–484. 28. R: A language and environment for statistical computing [computer program]. Version 2.8.0. Vienna: Foundation for Statistical computing; 2008. 29. Lumley T. The rmeta Package Version 2.14. http://cran.rproject.org/web/ packages/rmeta/index.html. Accessed April 9, 2010. 30. Maruyama K. Surgical Treatment and End Results of Gastric Cancer. Toyko: National Cancer Center Press 1985: 7–24. 31. Bonenkamp JJ, Hermans J, Sasako M, et al. Quality control of lymph node dissection in the Dutch randomized trial of D1 and D2 lymph node dissection for gastric cancer. Gastric Cancer. 1998;1:152–159. 32. Solomon MJ, Laxamana A, Devore L, et al. Randomized controlled trials in surgery. Surgery. 1994;115:707–712. 33. McLeod RS, Wright JG, Solomon MJ, et al. Randomized controlled trials in surgery: issues and problems. Surgery. 1996;119:483–486. 34. Horton R. Surgical research or comic opera: questions, but few answers. Lancet. 1996;347:984–985. 35. Edwards P, Blackshaw GR, Lewis WG, et al. Prospective comparison of D1 vs modified D2 gastrectomy for carcinoma. Br J Cancer. 2004;90:1888–1892. 36. Lewis WG, Edwards P, Barry JD, et al. D2 or not D2? The gastrectomy question. Gastric Cancer. 2002;5:29–34. 37. D´ıaz de Lia˜no A, Yarnoz C, Aguilar R, et al. Rationale for gastrectomy with D2 lymphadenectomy in the treatment of gastric cancer. Gastric Cancer. 2008;11:96–102. 38. Sue-Ling HM, Johnston D, Martin IG, et al. Gastric cancer: a curable disease in Britain. BMJ. 1993;307:591–596. 39. Pacelli F, Doglietto GB, Bellantone R, et al. Extensive versus limited lymph node dissection for gastric cancer: a comparative study of 320 patients. Br J Surg. 1993;80:1153–1156. 40. Siewert JR, B¨ottcher K, Roder JD, et al. Prognostic relevance of systematic lymph node dissection in gastric carcinoma. German Gastric Carcinoma Study Group. Br J Surg. 1993;80:1015–1018. 41. Sue-Ling HM, Johnston D. D1 versus D2 dissection for gastric cancer. Lancet. 1995;345:1515–1516. 42. Wu CW, Hsieh MC, Lo SS, et al. Results of curative gastrectomy for carcinoma of the distal third of the stomach. J Am Coll Surg. 1996;183:201–207. 43. Imamura Y, Yoshimi I. Comparison of cancer mortality (stomach cancer) in five countries; France, Italy, Japan, UK and USA from the WHO mortality database (1960–2000). Jpn J Clin Oncol. 2005;35:48–51. 44. http://seer.cancer.gov/statfacts/html/stomach.html. Accessed April 9, 2010. 45. http://info.cancerresearchuk.org/cancerstats/types/stomach/index.htm?script =true. Accessed April 9, 2010. 46. http://www.cancer.org/downloads/STT/Global_Facts_and_Figures_2007_ rev2.pdf. Accessed April 9, 2010. 47. Nitti D, Marchet A, Mammano E, et al. Extended lymphadenectomy (D2) in patients with early gastric cancer. Eurp J Surg Oncol. 2005;31:875–881. 48. Itoch H, Oohata Y, Nakamura K, et al. Complete ten-year postgastrectomy follow-up of early gastric cancer. Am J Surg. 1989;158:14–16. 49. Yoshikawa T, Tsuburaya A, Kobayashi O, et al. Is D2 lymph node dissection necessary for early gastric cancer? Ann Surg Oncol. 2002; 9: 401–405. 50. Bunt AMG, Hermans J, Boon MC, et al. Evaluation of the extent of lymphadenectomy in a randomized trial of Western-versus Japanese-Type surgery in gastric cancer. J Clin Oncol. 1994;12:417–422. 51. Feinstein AR, Sosin DM, Wells CK. The Will Rogers phenomenon. Stage migration and new diagnostic techniques are a source of misleading statistics for survival in cancer. N Engl J Med. 1985;312:1604–1608. 52. Volpe CM, Koo J, Miloro SM, et al. The effect of extended lymphadenectomy on survival in patients with gastric adenocarcinoma. J Am Coll Surg. 1995;181:56–64. 53. Songun I, Putter H, Kranenbarg EM, et al. Surgical treatment of gastric cancer: 15-year follow-up results of the randomised nationwide Dutch D1D2 trial. Lancet Oncol. 2010;11:439–449.

 C 2011 Lippincott Williams & Wilkins

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Annals of Surgery r Volume 253, Number 5, May 2011

Meta-Analysis of D1 Versus D2 Gastrectomy for Gastric Adenocarcinoma

54. Bunt TMG, Bonenkamp HJ, Hermans J, et al. Factors influencing noncompliance and contamination in a randomized trial of “Western” (r1) versus “Japanese” (r2) type surgery in gastric cancer. Cancer. 1994;73:1544–1551. 55. Yildirim E, Celen O, Berberoglu U. The Turkish experience with curative gastrectomies for gastric carcinoma: is D2 dissection worthwhile? J Am Coll Surg. 2001;192:25–37. 56. Danielson H, Kokkola A, Kiviluoto T, et al. Clinical outcome after D1 vs D2–3 gastrectomy for treatment of gastric cancer. Scand J Surg. 2007;96:35–40. 57. Aiko T, Sasako M. The new Japanese Classification of Gastric Carcinoma: Points to be revised. Gastric Cancer. 1998;1:25–30. 58. Csendes A, Burdiles P, Rojas J, et al. A prospective randomized study comparing D2 total gastrectomy versus D2 total gastrectomy plus splenectomy in 187 patients with gastric carcinoma. Surgery. 2002;131:401–407. 59. Maruyama K, Sasako M, Knishita T, et al. Pancreas-Preserving total gastrectomy for proximal gastric cancer. World J Surg. 1995;19:532–536. 60. Machara Y, Moriguchi S, Yoshida M, et al. Splenectomy does not correlate with length of survival in patients undergoing curative total gastrectomy for gastric carcinoma. Cancer. 1991;67:3006–3009. 61. Kodera Y, Yamamura Y, Shimizu Y, et al. Lack of benefit of combined pancreato-splenectomy in D2 resection for proximal third gastric cacner. World J Surg. 1997;21:622–628. 62. Kinoshita T, Maruyama K, Sasako M, et al. Lymph node dissection around the splenic artery: a comparative study of pancreatectomy and pancreas preserving operation. Nippon Geka Gakkai. 1992;93:128–132. 63. Okayima K, Isozaki H. Splenectomy for treatment of gastric cancer; Japanese experience. World J Surg. 1995;19:537–540. 64. Yu W, Choi GS, Chung HY. Randomized clinical trial of splenectomy versus splenic preservation in patients with proximal gastric cancer. Br J Surg. 2006;93:559–563. 65. Adachi Y, Kamakura T, Mori M, et al. Role of lymph node dissection and splenectomy in node positive gastric carcinoma. Surgery. 1994;116:837–841.

 C 2011 Lippincott Williams & Wilkins

66. Griffith JP, Sue-Ling HM, Martin I, et al. Preservation of the spleen improves survival after radical surgery for gastric cancer. Gut. 1995;36:684– 690. 67. Kwon SJ. Prognostic impact of splenectomy on gastric cancer: results of the Korea gastric cancer study group. World J Surg. 1997;21:837–844. 68. Lee KY, Noh SH, Hyung WJ, et al. Impact of splenectomy for lymph node dissection on long-term surgical outcome in gastric cancer. Ann Surg Oncol. 2001;8:402–406. 69. Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56:455– 463. 70. Sauerland S, Seiler CM. Role of systematic reviews and meta-analysis in evidence-based medicine. World J Surg. 2005;29:582–587. 71. Greenland S. Invited commentary: a critical look at some popular meta-analytic methods. Am J Epidemiol. 1994;140:290–296, 72. Stern JA, Egger M. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemol. 2001;54:1046–1055. 73. Higgns JP, Green S. Analysing and Presenting Results. Cochran’s Handbook for Systematic Reviews of Interventions 4.2.6 (updated September 2006). Section 8. In: The Cochrane Library, Issue 4. Chichester, UK; John Wiley & Sons Ltd. 2006. 74. Ioannidis JP, Lau J. Can quality of clinical trials and meta-analyses be quantified? Lancet. 1998;352:590–591. 75. Ng TT, McGory ML, Ko CY, et al. Meta-analysis in surgery: methods and limitations. Arch Surg. 2006;141:1125–1130; discussion 1131. 76. Smith GD, Egger M. Incommunicable knowledge? Interpreting and applying the results of clinical trials and meta-analyses. J Clin Epidemiol. 1998;51:289– 295. 77. Ioannidis JP, Patsopoulos NA, Evangelou E. Uncertainty in heterogeneity estimates in meta-analyses. BMJ. 2007;335:914–916.

www.annalsofsurgery.com | 911

Copyright © 2011 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.