Prospective comparison of D1 vs modified D2 gastrectomy for carcinoma

To compare the outcomes after D1 gastrectomy with those after modified D2 gastrectomy (preserving pancreas and spleen) performed by specialist surgeons for gastric cancer in a large UK NHS Trust. In all, 118 consecutive patients with gastric adenocarcinoma were referred by postcode, to undergo either a D1 gastrectomy (North Gwent (RJ), n=36, median age 76 years, 21 m) or a modified D2 gastrectomy (South Gwent (WL), n=82, 70 years, 57 m). Operative mortality in the two groups of patients was similar (D1 8.3% vs D2 7.3%, χ2 0.286, DF 1, P=0.593). Overall cumulative survival at 5 years was 32% after D1 gastrectomy compared to 59% after D2 gastrectomy (χ2 4.25, DF 1, P=0.0392). In patients with stage III cancers, survival was 8% after D1, compared with 33% after D2 gastrectomy (χ2 6.43, DF 1, P=0.0112). In a multivariate analysis, T stage (hazard ratio 2.339, 95% CI 1.683–2.995, P=0.01), N stage (hazard ratio 4.026, 95% CI 3.536–4.516, P=0.0001) and the extent of lymphadenectomy (hazard ratio 0.258, 95% CI –0.426–0.942, P=0.0001) were independently associated with durations of survival. In conclusion, modified D2 gastrectomy can improve survival four-fold for patients with stage III gastric cancer, without significantly increasing morbidity and mortality when compared with a D1 gastrectomy.

Opinion over the optimum resection for patients with gastric cancer remains divided, and the literature polarised. The impressive outcomes after D2 gastrectomy published in large retrospective series from Japan (Soga et al, 1979;Maruyama et al, 1987) have not been reproduced in randomised comparative studies from Europe (Bonenkamp et al, 1995(Bonenkamp et al, , 1999Cuschieri et al, 1996Cuschieri et al, , 1999. The two largest randomised studies both report significantly greater operative morbidity and mortality associated with an extended D2 lymphadenectomy when compared with the less aggressive D1 lymphadenectomy, and have failed to demonstrate any survival advantage for a D2 resection. Many of the serious complications associated with D2 resections were associated with resections of the pancreas and spleen (Bonenkamp et al, 1995;Cuschieri et al, 1996), and the best long-term survival was observed in patients undergoing D2 gastrectomy without pancreatico-splenectomy (Cuschieri et al, 1999). Although this latter report concluded than a classical D2 resection offered no survival advantage over a D1 resection, the possibility that a modified D2 resection, preserving pancreas and spleen, might be better than a D1 resection was not dismissed (Cuschieri et al, 1999).
The first reports of outcomes after modified D2 gastrectomy for gastric cancer were originally published in Britain by Sue-Ling et al, 1993(Sue-Ling et al, 1993) and subsequently Griffith et al (1995) from the University Department of Surgery at Leeds. We have subsequently demonstrated that similar results can be achieved in a large British district general hospital Barry et al, 2003). Despite these favourable reports, there remains a widely held assumption, that poor outcomes after surgery for gastric cancer in Britain are due to the greater age, comorbidity, advanced stages of disease and greater body mass indices of Western patients when compared with their Japanese counterparts (Dhar et al, 2000). Furthermore, most oesophagogastric cancer surgery in Britain and much of the West, has by tradition, been performed by general surgeons. Radical lymphadenectomy is a painstaking, technically demanding procedure, which has usually remained within the province of small numbers of specialist upper gastrointestinal surgeons (Roder et al, 1993). Thus, the D1 perigastric lymphadenectomy remains the most commonly performed operation for gastric cancer in the West.
The aim of our study was to investigate whether a modified D2 gastrectomy, allied to a specialist multidisciplinary team approach, would improve outcomes when compared with the traditional D1 gastrectomy. The setting was a large British NHS Trust comprising two District General Hospitals in South Wales serving a total population of 560 000. upper gastrointestinal surgery in Gwent, South Wales. Patients enrolled into the study were to have histologically proven, and potentially curable, gastric carcinoma. Patients were excluded if they had undergone previous gastric surgery, or had serious comorbidity that would preclude a safe D2 gastrectomy. All patients underwent staging laparoscopy to define potentially curable disease. Eligible patients were those who fell within the UICC TNM cancer stages I to III (Kennedy, 1987). Patients with North Gwent postcodes (catchment population 110 000) were treated at Nevill Hall Hospital, Abergavenny (DRBJ), and those with South Gwent postcodes (catchment population 450 000) were treated at the Royal Gwent Hospital, Newport (WGL).
Clinical and pathological information was collected prospectively and indices of multiple deprivation (IMD) were obtained from the Office for National Statistics. In total, 164 patients underwent treatment at Nevill Hall Hospital. Potentially curative resection was possible in 36 out of the 164 patients treated at Nevill Hall Hospital (22%). The median age of these 36 patients was 77 (range 58 -93) years and 22 were male. This comprises our control group (D1). In all, 265 patients underwent treatment at the Royal Gwent Hospital. Potentially curative resection was possible in 82 (31%, w 2 4.108, DF 1, P ¼ 0.043). The median age of this group of patients was 70 (range 27 -86) years (P ¼ 0.0001) and 57 were male. This comprises our comparison group (D2). The definition of a potentially curative resection was that all visible tumour was removed and that both proximal and distal resection margins were free of tumour on histological examination. Ethical approval was granted by the Gwent Research Ethics Committee.

Surgical treatment
The entire surgical management was conducted by two consultant surgeons working independently in the two major district general hospitals within Gwent Healthcare NHS Trust (DRBJ, Nevill Hall Hospital, North Gwent, and WGL, Royal Gwent Hospital, South Gwent). The operative details of the patients were defined in terms of the extent of the gastric resection, the macroscopic tumour-free margins and the level of the lymphadenectomy. Preoperative staging was done with the aid of both spiral computerised tomography and laparoscopy. All tumours were staged in accordance with the 1987 Unified Classification of gastric cancer (Kennedy, 1987), until 1997, when we adopted the recently published TNM Classification of Malignant Tumours (Sobin and Wittekind, 1997). The details of the patients undergoing potentially curative resection together with their ASA grades and stages of disease are shown in Tables 1 and 2. The details of their surgery are shown in Table 3.
D1 resections entailed removal of the lymph nodes within 3.0 cm of the tumour en bloc with the greater omentum and stomach. D2 resections necessitated the additional removal of the omental bursa, the hepatoduodenal nodes (antral tumours) and the splenic artery nodes. In patients with tumours of the posterior proximal stomach invading the tail of pancreas or splenic hilum, a splenectomy and distal pancreatectomy were also performed. In both groups of patients, a distal gastrectomy up to and including the duodenal bulb with a minimum of 5 cm proximal tumour-free margin was performed for antral tumours, whereas total gastrectomy was performed for middle and proximal tumours.

Follow-up
Patients undergoing both D1 and D2 resections were reviewed every 3 months for the first year and every 6 months thereafter. Only one of the 118 patients was lost to follow-up and 80 patients were followed up for a minimum of 5 years or until death (24 D1, 56 D2). The median duration of follow-up was 36 months. Endoscopy and computed tomography were arranged if recurrent disease was suspected. Causes of death were sought from case notes, general practitioners' records or via the Office for National Statistics.

Statistical analysis
Sample size calculations were based on a prestudy literature survey of six studies, which indicated that the baseline 5-year survival rate

RESULTS
Stages of disease at presentation (Table 1) There was no significant difference in the stages of disease in the two groups of patients, although the proportion of patients with stage I and II cancers was greater in patients undergoing D1 gastrectomy (58%) in North Gwent when compared with the patients undergoing D2 gastrectomy (48%) in South Gwent (w 2 1.162, DF 1, P ¼ 0.281).

Operative morbidity and mortality
There was no significant difference between the ASA grades of the patients undergoing potentially curative surgery in the D1 group (ASA I, II, III, IV: 4, 11, 21, 0) compared to the patients undergoing potentially curative surgery in the D2 group (ASA I, II, III, IV: 6, 39, 37, 0). Operative morbidity was 25% after D1 gastrectomy compared with 23% after D2 gastrectomy (w 2 ¼ 0.046, DF 1, P ¼ 0.830). Details of the major operative morbidity are shown in Table 4. Operative mortality 30 days after operation was 8.3% (three out of 36 patients) after D1 gastrectomy compared with 7.3% overall (six out of 82 patients) after a modified D2 gastrectomy (w 2 ¼ 0.037, DF 1, P ¼ 0.848). None of the patients who died after a D2 gastrectomy had undergone a splenectomy or pancreatectomy.

Survival
Corrected cumulative survival by treatment, calculated by life-table analysis, is shown in Figure 1. Cumulative survival for the 36 patients undergoing a D1 gastrectomy was 32% at 5 years. In contrast, survival for the 82 patients undergoing a D2 gastrectomy was 59% at 5 years (w 2 4.25, DF 1, P ¼ 0.0392). For patients with stage I disease the cumulative survival for the patients undergoing a D1 gastrectomy was 70% at 5 years compared with 80% for the patients undergoing a D2 gastrectomy (w 2 1.78, DF 1, P ¼ 0.1817). For patients with stage III disease the cumulative survival for the patients undergoing a D1 gastrectomy was 8% at 5 years compared with 33% for the patients undergoing a D2 gastrectomy (w 2 6.43, DF 1, P ¼ 0.0112) Figure 2. The factors found to be significantly associated with the duration of survival on univariate analysis are shown in Table 5.

Multivariate analysis
The prognostic variables entered into the model were age, sex, ASA grade, stage of the tumour, location of the tumour, type of gastrectomy (subtotal or total), level of lymphadenectomy, and level of the resection of the spleen and pancreas. Forward and backward stepwise regression was used.

DISCUSSION
Gastric cancer continues to be described as one of the 'captains of the men of death' in a major western text (Bailey and Love, 2000). The findings of this study demonstrate that gastrectomy with extended D2 lymphadenectomy, originally described by Japanese surgeons (Soga et al, 1979;Maruyama et al, 1987), can be introduced safely into a British Cancer Unit and may significantly improve the outcome for operable gastric cancer. In particular, our modified D2 gastrectomy (with preservation of the spleen) was associated with a three-fold improvement in survival for patients with stage III gastric cancer compared with a contemporaneous group of stage III patients undergoing a traditional D1 gastrectomy. Moreover, the proportion of patients for whom potentially curative resection was possible was significantly greater in patients undergoing a D2 gastrectomy compared to patients undergoing a D1 gastrectomy. There are several potential criticisms of this study. Clearly, this is not a randomised trial, because patients were referred to the relevant surgeon according to their postal code. This may introduce two confounding variables; in studies of breast cancer and colorectal cancer postcode addresses of patients have been associated with survival differences, perhaps because social deprivation is commoner in certain postal districts (Sainsbury et al, 1995, Brewster andBlack, 1998). Secondly, the postal address of each patient determined which one of two surgeons performed their operation. It is well recognised that the identity of the surgeon performing the operation is an independent determinant of stage-related survival (MacDonald et al, 2001). Yet the improved outcomes after D2 gastrectomy cannot be explained by unexpectedly poor results in our control group because outcome after D1 gastrectomy compared favourably with that reported in a large UK population-based study (Allum et al, 1989). In this latter study, less than 1% of patients had stage I disease (compared with 7% from our study and 18% from the MRC ST01 trial), and curative resection was possible in only 25% of patients (compared to 21% in the present study). Operative mortality was 16% after potentially curative D1 resection (compared to 8.3% in our study and 6.5% from the MRC ST01 trial) and 5-year survival was 20% (compared to 38% in our study and 35% from the MRC ST01 trial). The surgical management of patients with gastric cancer at Nevill Hall Hospital was therefore at least as good as that of most other centres performing D1 gastrectomies in the United Kingdom.
The reluctance of Western surgeons to adopt extended lymphadenectomy is understandable in the light of the results of the MRC ST01 trial of D1 vs D2 resection for gastric cancer (Cuschieri et al, 1996(Cuschieri et al, , 1999 and the Dutch study (Bonenkamp et al, 1995(Bonenkamp et al, , 1999. Each reported similar mortality following D2 gastrectomy (13 and 10%, respectively), and both include evidence to suggest that the excess mortality is accounted for by splenectomy (with or without the resection of the pancreas) rather than the extended lymphadenectomy. Moreover, in MRC ST01 the best survival was obtained in patients who underwent D2 resections without pancreatico-splenectomy. The incidence of lymph node metastases along the splenic artery and splenic hilum is reported to vary from 15 to 27% (Fass and Schumpelick, 1989;Mendes de Almedida et al, 1995;Tsubaraya et al, 1995). Many of these nodes can be cleared with preservation of the pancreas and spleen. D2 gastrectomy with preservation of the spleen has already been shown to confer potential survival benefits with low morbidity and mortality (Griffith et al, 1995).
Both MRC ST01 and the Dutch trial have received criticism over the relative inexperience of many different surgeons performing D2 lymphadenectomy. Furthermore, the existence of a relationship between caseload and operative mortality remains controversial. The specialist surgical unit in Leeds has described a long learning curve during the adoption of D2 gastrectomy, with 10 years elapsing before operative mortality fell to 5% (Sue-Ling et al, 1993). Our results parallel this finding, with surgical subspecialisation reducing operative mortality after D2 gastrectomy from 12.5 to 2.4% over a period of 5 years. Operative mortality fell during this period even though there were no differences in stage of  cancers at presentation or postoperative complications. This serves to emphasise that low operative mortality is achieved not only by specialised operative technique, but also by specialised anaesthetic and postoperative care from an experienced multidisciplinary team.
In conclusion, this nonrandomised study points to potential survival benefits arising from modified D2 gastrectomy in patients with gastric cancer of pathological stage T3 NX M0. The results strongly support the establishment of a multicentre randomised controlled trial of modified D2 vs D1 gastrectomy, each performed by expert gastric surgeons and targeted towards patients with tumours of intermediate stage. The practical problem that we face in Britain is that most patients present with advanced stage III and IV cancers, and preoperative staging modalities remain of limited sensitivity (Daly et al, 1999;Barry et al, 2002). Moreover, additional confounding variables, such as surgical subspecialisation and multi disciplinary team learning curves may result in treatment contamination and noncompliance, to the extent that implementation would require robust quality control (McCulloch et al, 2003). Stage-directed management tailored to individual patients is clearly the way forward if the benefits of greater subspecialisation and a multidisciplinary team approach are to be realised (Furukawa et al, 2002). The exact role of D2 gastrectomy following neoadjuvant chemotherapy (Allum et al, 2003) or prior to chemoradiotherapy (MacDonald et al, 2001) remains unclear at present, but our own experience is encouraging and suggests that in otherwise fit patients with gastric tumours that are perceived to be of stage III, without distant metastatic disease, a modified D2 gastrectomy preserving pancreas and spleen where possible, and performed by specialist surgeons, may have advantages over the traditional standard D1 gastrectomy.