In the Netherlands, 80 hospitals participated in a randomized trial to compare morbidity, hospital mortality, survival, and cumulative relapse risk after D1 or D2 lymph node dissection for gastric cancer. Between 1989 and 1993, 996 patients were centrally randomly assigned to different treatment groups; 711 patients (380 assigned to D1 dissection and 331 to D2 dissection) underwent the allocated treatment with curative intent, and 285 patients required palliative treatment. For definition of D1 and D2 dissection, the guidelines of the JRSGC were used. Because these guidelines were not regularly used in the Netherlands, a Japanese surgeon experienced in the treatment of gastric cancer was invited to instruct the participating Dutch surgeons. Continuous quality control was believed to be necessary to maintain the appropriate level of lymph node dissection. This quality control was substantiated by relating the number and location of lymph nodes detected at pathological examination to the guidelines of the protocol (25). If lymph nodes were harvested from stations that were not supposed to be present according to protocol, this was called contamination. If lymph nodes were not harvested from stations that should have been harvested, this was termed noncompliance. These differences from the study protocol could occur in patients undergoing both D1 and D2 dissection. Contamination in the D1 group and noncompliance in the D2 group, in particular, could lead to decreased distinction between the trial arms.

After curative resection, patients in the D2 arm had higher postoperative mortality (10% vs. 4% for the D1 arm; P = 0.004). They also had significantly more complications (43% vs. 25% for the D1 arm; P <0.001), which led to significant prolonged hospital stay for patients with a D2 dissection. Hemorrhage (5% for D2 vs. 2% for D1), anastomotic leakage (9% for D2 vs. 4% for D1), and intraabdominal infection (17% for D2 vs. 8% for D1) were the most frequent complications. After curative resection, D2 patients had higher postoperative mortality (10% vs. 4% for D1; P = 0.004). They also showed significantly more complications (43% vs. 25% for D1; P ≤0.001) that led to significant prolonged hospitalization for patients after D2 dissection. Hemorrhage (5% vs. 2% for D1), anastomotic leakage (9% vs. 4% for D1), and intraabdominal infection (17% vs. 8% for D1) were the most frequent complications. In the most recent evaluation, the median follow-up for all eligible patients is 11 years (range, 6.8–13.1 years). At 11 years, survival rates are 30% for D1 and 35% for D2 (P = 0.53). The risk of relapse is 70% for D1 and 65% for D2 (P = 0.43). If hospital deaths are excluded, survival rates are 32% for D1 (n = 365) and 39% for D2 (n = 299, P = 0.10). The relapse risk of these patients (n = 664) is in favor of the D2 dissection group (P = 0.07).

In a univariate analysis of all 711 patients, no significant impact on survival rates was found for any of the subgroups based on the selected prognostic variables between D1 and D2 dissection. Analysis of interaction between covariate and extent of lymph node dissection shows no significance either. The only subgroup with a trend to benefit is the N2 disease group. If patients with hospital mortality are excluded, there is a significant survival and relapse advantage for patients with N2 disease who had a D2 dissection (P = 0.01). Other stages show no significant difference (N0 P = 0.42; N1 P = 0.31; N3 P = 0.24) in this subset analysis. Furthermore, there is no difference in survival after 11 years whether <15 lymph nodes, between 15 and 25 lymph nodes, or >25 lymph nodes are harvested.

In a prospective randomized trial conducted by the British Medical Research Council, D1 dissection was compared with D2 dissection. Central random assignment of patients to treatment groups followed a staging laparotomy. Of 737 patients with histologically proven gastric adenocarcinoma registered, 337 patients were judged ineligible by staging laparotomy because of advanced disease, and 400 were randomly assigned to treatment (200 to D1 dissection and 200 to D2 dissection). Postoperative mortality was significantly higher in the D2 group (13% vs. 6.5% for D1 dissection; P = 0.04). Postoperative complications were also significantly higher in the D2 group (46% vs. 28% for the D1 group; P <0.001). In this study, anastomotic leakage (26% for D2 vs. 11% for D1), cardiac complications (8% for D2 vs. 2% for D1), and respiratory complications (8% for D2 vs. 5% for D1) were most frequent. The 5-year survival rates were 35% for patients undergoing D1 dissection and 33% for those having D2 dissection (24).

These major randomized studies, the MRC and the DGCT, obviously show the same tendency. Although the timing of randomization was different in the two trials and no quality control was carried out in the British trial, the postoperative mortality and morbidity rates in both trials were significantly higher in the group undergoing D2 dissection. Furthermore, no 5-year survival advantage was found for extended (D2) dissections in either study. The conclusion from these randomized studies was that generally no support exists for the standard use of extended (D2) lymph node dissections in patients with gastric cancer in the West (14,24).

Surgical complications are influenced by the extent of the operation, and a number of studies have addressed this topic. In a Norwegian study, morbidity was significantly lower after subtotal resection (28%) than after total gastrectomy (38%), although in this study proximal gastrectomy had the highest morbidity (52%) (29). In a German study, these differences in morbidity were also found (23% for subtotal resection vs. 48% for total gastrectomy) (30). Gennari et al. found a decreased morbidity for subtotal resections without any significant influence on survival (31). Comparison of their results with those of 15 previous studies led to the conclusion that subtotal gastrectomy should be standard provided that a safe proximal margin is guaranteed. In the DGCT and the MRC, hospital mortality in the groups undergoing D1 dissection and D2 dissection was significantly lower for subtotal gastrectomy (3% and 7%, respectively) than for total gastrectomy (5% and 14%, respectively) (23,32,33). In both trials, the complication rate was also lower after subtotal resections. In the DGCT, this difference was statistically significant. The prognostic value of microscopic resection line involvement in the DGCT was studied by Songun et al. (34). Tumor-positive resection lines were seen in 5.9% of evaluable patients. Resection line involvement was significantly associated with T stage, N stage, tumor location, and tumor differentiation. Presence of resection line involvement was also associated with significantly worse survival. The conclusion from this study was that preoperative frozen-section examination is mandatory in patients undergoing a curative resection for gastric cancer, especially in those with poorly differentiated, signet ring cell, or anaplastic tumors. In this context, arguments can be made for performing a total gastrectomy in all patients with poor tumor differentiation.

Resection of spleen, pancreas, or both plays an important role in surgical complications. Although one study failed to find significant differences (35), most studies find a significant increase of morbidity and hospital mortality if a pancreaticosplenectomy is performed (19,36,37). Two studies in Japan did not show any beneficial effect on survival if pancreaticosplenectomy was combined with total gastrectomy, whereas morbidity was increased in these patients (35,39). In the DGCT, pancreatectomy and type of gastrectomy were the only factors significantly influencing the occurrence of major surgical complications (37). Although the number of dissected lymph nodes increases, septic complications occur more often due to anastomotic leakage, intraabdominal infections, and pancreatic fistula (39).

Preferably, the spleen should also be spared because this might reduce concomitant morbidity (33,40). An increase of anastomotic leakage was seen, particularly in subtotal D2 gastrectomies. The most likely explanation for this finding is that in D2 dissections, the left gastric artery is divided at its origin and the rest of the stomach is dependent on the blood supply of its short gastric arteries. In D1 dissections, in which the left gastric artery is divided more peripherally, the vascularization of the rest of the stomach is probably less compromised. Immunologic factors may play a role in this as well, associated first with resection of the spleen itself (41,42) and second with the immunosuppression induced by blood transfusions, which may be needed for increased hemorrhage (43,44,45). A recent randomized trial on splenectomy indicated no advantages in survival as reported by Yu et al. (46).

Tumor stage is an important prognostic factor for survival in gastric cancer. Although not all studies find size of the tumor to be an independent prognostic factor for survival (47,48), a clear relation is seen between increasing depth of invasion and survival (15). With increasing depth of invasion, a steady increase is seen in the number of cases with positive lymph nodes, from 45.7% when the tumor invades the muscularis propria to 79.6% when adjacent organs are directly invaded. Also, the frequency with which the more distant tiers of nodes (second, third, and fourth) is involved rises steadily with depth of invasion (49).

The incidence of metastasis and 5-year survival rate shows a strong correlation. Moreover, with increasing distance between involved node and the primary tumor, the proportion of 5-year survivors decreases. Involvement of node station 13 is associated with a zero 5-year survival rate (49). In the DGCT, surgery with an involved N4 node was regarded as a noncurative operation. Benefit from extended dissections (stations 7–12 and 16) in Japanese studies is estimated to be between 0% and 10.5% (49), although as discussed previously, this benefit was not found in randomized studies in the West (14,32). In Japan, dissections even beyond the D2 level are now being studied in two randomized studies, results are not expected before 2008 (50).

Recommendations for stage grouping from the JRSGC and from the fourth edition of the AJCC/UICC manual are more or less comparable (8,9,10). Most studies until now have used these classifications. The fifth edition of the UICC manual, however, classifies N stage by the number of affected regional lymph node metastases and not by the location of the lymph node metastases (51,52). Details of the sixth edition AJCC staging system can be found in Chapter 22. This new staging method has three advantages:

Hermanek et al. compared this staging system with the previously used systems and found an improvement in estimation of outcome with the new staging system (53). Other studies comparing these staging systems are now underway. Also, the data of the DGCT are being used for evaluation of this new system.

With more accurate staging, a stage migration phenomenon may arise (54). As a result of extended lymphadenectomy, a proportion of patients will be assigned to a more advanced stage than would otherwise be the case, although their prognosis is the same. If this occurs, the overall results for each stage improve, and the proportion of patients in more advanced stages increases. This stage migration is often held responsible for the differences in survival between Japanese and Western patients (52).

In the DGCT, 5-year survival for tumor, node, metastasis (TNM) stage II was 38% after D1 dissection and 43% after D2 dissection. For TNM stage IIIA, 5-year survival was 10% after D1 dissection and 29% after D2 dissection (14). Stage migration occurred in 30% of the patients undergoing D2 dissection. Given the observed 5-year survival rates, we calculated
that stage migration leads to a drop in TNM stage-specific survival of 3% for UICC stage I, 8% for stage II, 6% for stage IIIA, and 12% for stage IIIB (55). In a large German study, the disease stages of patients that benefited most in 5-year survival from D2 dissections were stages II and IIIA (stage II: standard dissection 27% and radical dissection 55%; stage IIIA: standard dissection 25.3% and radical dissection 38.4%) (17). The significant or marginally significant differences in survival for patients undergoing D1 and D2 dissections for disease of TNM stages II and IIIA have now shown to be largely attributable to stage migration.

With the aging of Western society, gastric cancer will be diagnosed in more elderly patients. Population-based data from the Netherlands from 1982 to 1992 show that 27% of newly diagnosed patients were older than 80 (56). A study of gastric cancer in the elderly by Kranenbarg et al. (57) found no difference in resectability and curability rates between different age groups (Table 23.3), but hospital mortality increases with age, especially in those older than 75. Performance of extended (D2) dissections in elderly patients was associated with significantly higher hospital mortality. Multivariate analysis of these data showed that the older than 65 age group had a relative risk of 4.35 for hospital death, compared to patients younger than 65 (33). Also, 5-year survival was found to be significantly better for patients younger than 65. Some investigators do not consider age to be an important prognostic factor (58); however, the present authors believe that, although gastrectomies should not be withheld from elderly patients, extended dissections should be avoided in Western patients older than 70.

Maruyama et al. compiled a computer-based database containing pathological data from 3,040 patients (59). With the knowledge of tumor size, position, and depth of invasion (judged preoperatively by endoscopy and double-contrast barium meal or by endosonography), the likelihood of lymph node metastasis in each of the 16 lymph node stations can be predicted accurately. This enables the correct level of lymph node dissection to be determined. Applicability of this program to Western patients is shown by Peeters et al. (60). A blinded, retrospective analysis of Dutch (D1 vs. D2) trial data suggests that low Maruyama Index (MI) surgery is associated with significantly increased survival. A dose–response effect with respect to the MI and survival is also apparent. We advocate using the Maruyama program, a computerized tool based on patient experience, to identify nodal stations at risk, either preoperatively or intraoperatively, to customize surgical lymphadenectomy and routinely generate a low-MI operation.

Gastric resections are performed less frequently, especially because H₂ antagonists have reduced the need for them and because the incidence of malignant gastric disease throughout the world has declined. Nevertheless, resection remains the only possible cure for gastric cancer. With the improvement of surgical techniques, postoperative mortality and morbidity have decreased in the past decades, and anesthesia, metabolic care, and nutrition still play important roles. Because extended resections are recommended to increase survival, the question arises as to whether the influence of surgical skill on outcomes has grown (61).

McCulloch et al. retrospectively studied the results of 206 gastrectomies performed by 17 consultant surgeons and showed a considerable variation among surgeons with respect to judgments of resectability, adequacy of resection, anastomotic leakage, and patient mortality (62). Although in this study the number of resections performed by a surgeon had no significant influence, the conclusion was that having fewer surgeons perform all resections might improve outcomes. In the German Gastric Cancer Study, the experience of surgeons was also studied (30). A significantly higher rate of anastomotic leakage occurred when patients were operated on by surgical trainees under supervision than when they were operated on by experienced surgeons (19% vs. 6%, respectively). In the DGCT, all D2 dissections were supervised by referent surgeons, whereas nearly all D1 dissections were attended by the study coordinator. All referent surgeons had at least some experience with D2 dissections before the trial. To standardize the procedure, all referent surgeons were trained by a single Japanese surgeon experienced in this surgery (23). Referent surgeons performed, on average, 41 D2 dissections (range, 23–61) during the trial (Table 23.4) (63). No serious heterogeneity in morbidity and postoperative mortality rates was seen among referent operators, nor between referent surgeons and the Japanese
instructing surgeon in operations on Dutch patients (64). Neither univariate nor multivariate analysis showed a learning curve for the referent surgeons throughout the study (33). In his own experience, McCulloch found that he had to perform at least 30 dissections independently before reaching a plateau (65). In a 3-year prospective study of his own learning curve for D2 gastrectomy, a significant decrease in morbidity and postoperative mortality was shown to occur in the third year, which suggests a learning curve lasting 18 to 24 months, or 15 to 25 procedures, before a plateau is reached. Explanations given for this improvement in outcome, besides operative skill, include better patient selection and performance of less extensive dissections (66). As McCulloch gradually moved away from pancreaticosplenectomy, the idea that only the change of tactics, and not the operative skill, improved outcome during these years is not unthinkable. A plateau may have been reached before or may not yet have been reached. Furthermore, the learning period differs for every surgeon—some are quick learners and some are slow learners. Thus, the number of operations or time needed to reach a plateau for a certain procedure differs for each surgeon. Finally, some surgeons are just better surgeons. In addition, the means of learning may be of influence. Supervision and training very likely lead to quicker and better learning of a certain procedure than “self-tuition.”

The experience of the hospital might also influence outcomes for gastric resection. Comparison of results for resections at Japanese (n = 845), German (n = 564), and Dutch (n = 50) hospitals showed a significantly better outcome for Japanese patients. The reasons for these differences between Japanese and Western results have often been discussed (15,18,59). No difference in outcomes was noted between the German hospital and the smaller Dutch hospital. The impact of hospital volume on postoperative mortality was also studied by Begg et al. (67). In this retrospective cohort study of 5,013 patients older than 65, a low hospital volume was shown to be strongly associated with excess mortality. In the German Gastric Cancer Study, 19 hospitals participated and together performed 1,654 resections (range, 12–243). In a multivariate analysis in this study, the experience of the hospitals had a significant influence on morbidity and postoperative mortality (30). These findings are in line with those of McCulloch (62,66).

Both doxorubicin and epirubicin have been used in the adjuvant setting. Several of the combinations also included mitomycin C.

Several trials have used the fluorouracil-doxorubicin-mitomycin (FAM) regimen or a variant. Coombes et al. reported the results of a trial in which patients undergoing curative resection received either FAM or observation only (80). A total of 281 patients were evaluable. Entrance into the trial was allowed as late as 6 weeks postoperation. After a median follow-up of 68 months, 61% of patients in the control arm experienced recurrence versus 56% in the treated arm. No significant difference was seen in either diseasefree or overall survival (overall survival was 45.7% in the FAM arm vs. 35.4% in the observation arm). A trend toward improved outcome was noted for patients with T3 or T4 tumors who had positive lymph nodes (P = 0.07 in favor of the FAM group). In a similar study, investigators from the Southwest Oncology Group examined the use of FAM therapy versus expectant observation. Of 193 eligible patients, 100 were randomly assigned to observation and 93 to receive FAM chemotherapy (81). After an overall median follow-up of 9.5 years, no significant differences were found in diseasefree or overall survival (overall survival at 5 years was 37% for the FAM group vs. 32% for the observation group; P = 0.59). As was the case in the study by Coombes et al., an unplanned subgroup analysis indicated a trend toward benefit for patients with stage III disease. In a third FAM study, Lise et al. randomly assigned 159 patients to receive surgery only and 155 to receive a modification of the FAM regimen (82). After a median follow-up of 80 months, no significant differences in survival were seen, with approximately 43% of patients in each arm surviving for 5 years. An improvement was found in diseasefree survival (P = 0.02). Hallissey et al. reported the results of a three-arm study in which patients undergoing resection were followed with observation only, received radiation therapy to 45 Gy, or received the FAM regimen. One hundred and forty-five patients were observed; 153 received radiation, and 138 received chemotherapy (83). The FAM regimen was modified from that in the original treatment report. No significant differences in outcome were seen; 5-year survival rates were 20% for the control group, 12% for the group receiving radiation, and 19% for the group receiving FAM chemotherapy.

Tsavaris et al. (84) examined the use of FAM chemotherapy versus observation in a group of 84 patients. Sixty-four percent of patients receiving chemotherapy experienced recurrence or death versus 81% of patients in the control group. For this small group of patients, however, this difference was not statistically significant.

Krook et al. studied 120 evaluable patients who were randomly assigned either to observation or to three cycles of fluorouracil and doxorubicin (85). The median survival was 31 months for patients undergoing observation versus 36 months for patients receiving therapy, and 5-year survival was almost identical for the two groups (33% vs. 32%, respectively). These differences were not statistically significant. Two treatment-related deaths occurred.

Dutch investigators performed a small random assignment trial using the fluorouracil-Adriamycin-methotrexate (FAMTX) regimen (86). Fifty-six patients were entered into this trial, which was closed early because of poor accrual. Four cycles of chemotherapy were given prior to surgery with the control arm undergoing operation alone. The study was initially powered for a total of 225 patients in each arm, with an interim analysis after 100 patients had been studied. The objectives
included an improvement in curable resectability rate and diseasefree survival. Fifty-six percent of patients in the FAMTX arm had R0 resections versus 62% in the surgery-only arm. There was no significant difference in “downstaging.” Toxicity in the FAMTX arm included neutropenia. There was no significant difference in median survival.

Epirubicin has also been used in the adjuvant setting. In a small trial, Neri et al. compared an epirubicin-fluorouracil-leucovorin regimen to expectant observation in a group of 137 patients (87). Chemotherapy was delivered over a 7-month period. The median survival for patients receiving this therapy was 31 months and was superior to survival of those undergoing observation only (18 months) (P = 0.01). At 5 years, 30% of patients receiving adjuvant therapy remained alive versus 13% of patients randomly assigned to observation. The results are different from those of the study by Krook et al. (85), who treated a similar number of patients with a similar anthracycline regimen. A confirmatory trial is not yet available.

Recently, the final results of a phase III trial performed by Italian investigators have been reported in abstract form. This study used systemic chemotherapy, including epirubicin, leucovorin, fluorouracil, and etoposide, after operation versus operation alone in patients having at least a D1 lymph node dissection. Two hundred twenty-eight patients were randomized to surgery only or surgery followed by chemotherapy. With a median follow-up of 60 months, 5-year overall survival was 48% for those receiving postoperative chemotherapy and 45% for those undergoing surgery only (P = 0.6). There also was no difference in diseasefree survival. They concluded that adjuvant therapy using this regimen had a 4% to 5% difference in overall diseasefree survival, which was not statistically different between the two arms. Of note, this is the same magnitude of difference reported from meta-analyses (88).