The full set of randomized trials is listed in Table 2 [6,7,8,9,10,11,12, 23,24,25,26,27,28,29,30,31,32,33,34,35,36]. There is a Cochrane meta-analysis published in 2015 that includes long-term oncological follow-up data and postoperative mortality data, including 2515 patients from 8 trials [37]. These studies compared D1 versus D2 dissection [8,9,10,11,12, 23]. It’s worth noting that the Taiwanese study comparing D1 vs. D3 dissection (D2 + 13 + 14v without systematic SPC) was included in the D1 vs. D2 comparison and initially introduced heterogeneity between studies, as it was the only study to show a positive impact on overall survival [28]. Of the trials comparing D2 vs. D3 dissection, two compared D2 dissection with D2 extended to group 16 (16a2 + 16b1) [6, 32]. The last trial compared D3 dissection, including groups 12, 13 and 14, to D4 dissection with the addition of group 16 [33]. Two older trials that did not report survival rates were excluded from this meta-analysis [23, 31]. Their results are detailed below.

Table 2 Randomized controlled trials (RCT) comparing different types of lymphadenectomy: morbidity, mortality and long-term survivalAssociated morbidity and mortality of different types of dissectionD1 vs. D2 (or D3) LN dissection

We have the results of 7 randomized trials [8,9,10,11,12, 23,24,25,26].

The South African study by Dent was the first randomized trial to compare D1 dissection with D2 dissection (including excision of the upper mesocolon and pancreatic capsule without mentioning the SPC). No statistical hypothesis was stated, and 43 patients out of 408 who underwent surgery were randomized, with a 3-year follow-up that was purely clinical. The authors concluded that there were more perioperative complications (duration of surgery and transfusion) and postoperative complications (reintervention and length of stay) in the D2 group [23]. The Robertson study compared R3 dissection (D2 dissection with systematic SPC and group 12) (n = 30) with R1 dissection (n = 25) [24]. There was only one death in the R3 group due to intra-abdominal sepsis. There were no major complications in the R1 group, whereas 47% had subphrenic abscesses, 23% required reoperation and 10% had fistulas in the R3 group. These two studies clearly favoured the D1 dissection in terms of post-operative morbidity.

Subsequently, three large multicentre Western trials compared D1 and D2 dissection: a trial from MRC involving 400 patient [9, 26], a Dutch trial involving 721 patients from 80 hospitals [8, 13, 25], and an Italian trial involving 267 patients from 6 centres [10,11,12]. In the first two studies, the authors followed the Japanese authors’ guidelines for dissection closely and received training from them. SPC was routinely performed in total gastrectomy according to the old definition of D2 dissection. The Italian trial was designed in the wake of the other two trials and aimed to assess the benefit of D2 dissection without systematic SPC. In the MRC and Dutch trials [8, 9, 13, 25, 26], D2 dissection increased postoperative morbidity and mortality compared with D1 dissection. A meta-analysis of these two trials confirmed these results, showing that mortality was tripled with D2 dissection, with a relative risk of 2.93 (95% CI 1.45–3.45). The “excess mortality” associated with D2 dissection reported in both trials is attributed by most authors to the learning curve of the surgeons participating in the trials. The number of procedures required to overcome the learning curve was estimated to be 25 [4], which is far higher than the numbers reported in the UK trials (32 surgeons for 400 patients, averaging 12.5 patients per surgeon over 7 years) and the Dutch trials (85 surgeons in 80 hospitals over 4 years, resulting in one resection per surgeon per year). In 2006, a single-centre randomized trial from Taiwan enrolled 221 patients [28, 29] and compared D1 dissection with D3 dissection (D1 + celiac trunk branch + hepatic pedicle + 13 + 14 without group 16). This study was performed with a high-quality methodology. SPC was not routinely performed in cases of total gastrectomy, except when an intraoperative examination was positive for group 1 or 11, which applied to only 12% of patients. Patients requiring SPC for necessity (invasion or large LN masses) were excluded. The results of this study [28, 29] also showed that morbidity was increased with D3 dissection (compared to D1 dissection), including a higher incidence of intra-abdominal infectious complications (8.1% vs. 0%, p = 0.008) and a trend towards more anastomotic fistulas (4.5% vs. 0%, p = 0.060). However, mortality was zero and identical regardless of the extent of dissection, highlighting the expertise of the team involved. The short-term results of the Italian study, conducted in 6 expert centres, comparing D1 dissection (n = 133) with D2 dissection without systematic SPC (n = 134), were published in 2010 [11]. In the intention-to-treat analysis, the morbidity rates after D2 dissection and D1 dissection were 17.9% and 12.0%, respectively (p = 0.178). The in-hospital mortality rate was 3.0% in the D1 group and 2.2% after D2 dissection (p = 0.722). The authors concluded that in specialized centres, the complication rate after D2 dissection without systematic SPC was lower than that reported in previously published Western randomized trials and that it could be considered a safe option in this context.

In a last small study, the authors compared D1 + dissection with standard D2 dissection, including group 12a and systematic splenectomy (n = 36 vs. 37 patients) [27]. Surgical complications were significantly more frequent in the D2 group, including 2 postoperative deaths, favouring D1 + dissection [27].

The results of the meta-analysis on post-operative mortality are clearly against D2 dissection with an odds ratio of 2.02 [1.34;3.04], considering mortality rates of 3.9% and 7.8%, resulting in 38 additional deaths per 1000 patients operated on with D2 dissection instead of D1 [37]. However, analysis of the results of studies that have evaluated the role of SPC as a primary objective or as a post hoc analysis in this excess mortality is clear. These results, together with the morbidity findings of the Italian study, explain the recent changes in the definitions of dissection discussed above.

D2 (or D3) vs. D2 dissection combined with para-aortic LN dissection

Five randomized trials [6, 30,31,32,33] have been published. There is considerable variation in the definition of D2 + dissection. It has been referred to as D2+ [30, 31, 33] or D3 [32]. In the most recent study, it was even referred to as D4 and compared to D3 dissection (D2 + 12, 13, 14) vs. D4 = D3 + 16) [33]. It is also worth noting that, with the exception of one study in which splenectomy was performed systematically [32], splenectomy and pancreatectomy were only performed in cases of LN metastases or local invasion of the spleen or pancreas. Of these 5 trials, only one showed a statistically significant increase in morbidity after D2 dissection combined with para-aortic LN dissection (D3) [32]. In the other four trials, morbidity was similar. Finally, postoperative mortality was not increased after lombo-aortic dissection in addition to standard D2 dissection in all 5 trials. These results were consistent with the meta-analysis, which included only 3 of the trials, but concluded that there was no excess postoperative mortality, with no heterogeneity between trials [37].

Role of (spleno) pancreatectomy

A total of 6 randomized trials [8, 9, 27, 30, 31, 33] have shown a strong independent association between postoperative morbidity and mortality and resection of the spleen and tail of the pancreas. For example, in the study by Cuschieri et al. [9], the authors demonstrated that morbidity and mortality were significantly higher in cases of splenectomy (59% vs. 22%, p < 0.001 and 17% vs. 6%, p < 0.001, respectively). These results were confirmed by the Dutch study [8]. Wu et al. [28, 29] also observed an increase in morbidity with splenectomy, but not in mortality. Finally, in other studies [32, 36], pancreatectomy was the most significant predictive factor for postoperative complications and was associated with an increase in morbidity, whether or not LN dissection included para-aortic LN.

D2 LN dissection with or without splenectomy

Three studies [7, 35, 36] have been published with conflicting results. These trials compared postoperative morbidity and mortality and survival according to whether or not splenectomy was performed as standard with or without D2 LN dissection during total gastrectomy, excluding patients with splenic continuity invasion and LN metastases requiring mandatory splenectomy. The third study also excluded patients with tumour invasion of the greater curvature (including gastric linitis) [7]. In two studies [7, 35], morbidity was significantly increased with splenectomy (with D2 LN dissection). In the study by Yu et al. [36], morbidity was not significantly affected by splenectomy. Mortality was similar in all three studies [7, 35, 36].

Oncological outcomes (5-year survival) of different types of lymph node dissectionD1 vs. D2 (or D3) LN dissection

In the Dent study, D2 lymphadenectomy did not show a significant benefit in 3-year survival [23]. In the Galizia study comparing D1 + vs. D2 with splenectomy, the primary site of tumour recurrence and 5-year disease-free survival were not different between the two groups. The incidence of involved LN in the additional resection groups was 5%. These results favoured D1 + lymphadenectomy [27]. In the Robertson trial, survival was even better in the R1 group (1511 days vs. 922, p < 0.05) [24]. The British and Dutch studies [25, 26] showed no survival benefit for D2 lymphadenectomy compared to D1. However, it was mainly postoperative mortality that negatively affected survival in both trials. In 2010, the 15-year results of the Dutch study were published [13]. D2 lymphadenectomy was ultimately associated with a reduction in locoregional recurrence and cancer-related mortality compared with patients who underwent D1 lymphadenectomy (p = 0.01). Overall survival (OS) was significantly lower in patients who underwent splenectomy and pancreatectomy in both the D1 and D2 arms. Subgroup analysis of patients who did not undergo pancreatectomy or splenectomy showed a significantly higher 15-year OS in the D2 group (35% vs. 22%), leading the authors to recommend D2 without systematic SPC lymphadenectomy for resectable gastric cancer. The 5-year results of the Italian trial showed that OS and disease-specific survival (DSS) were 66.5% and 71% after D1 lymphadenectomy and 64.2% and 72.6% after D2 without systematic SPC lymphadenectomy, with no significant difference between the two groups (OS p = 0.695, DSS p = 0.916) [10]. However, there was significant contamination in the D1 group. Furthermore, subgroup analysis showed a trend towards a benefit of D2 dissection in patients with locally advanced gastric cancer > T1 (DSS 55% for D1 vs. 69% for D2 with p = 0.143) and in N + patients (OS rate of 35% for D1 vs. 51% for D2 and DSS rate of 38% for D1 vs. 59% for D2) and in patients with T2-T4 and N+. The long-term results of this trial (15 years), published in 2021, confirmed the absence of a significant difference in OS and DSS between the two groups in the overall population [12]. Subgroup analysis showed a significantly higher DSS in the D2 without systematic SPC lymphadenectomy group in patients with locally advanced gastric cancer > pT1N+ (29.4% vs. 51.4%, p = 0.035), confirming the benefit of D2 without systematic SPC lymphadenectomy in these patients. Conversely, DSS was significantly better after D1 lymphadenectomy in early-stage patients and those over 70 years of age (p = 0.001) [12].

Only the Taiwanese study, in which postoperative mortality was zero regardless of the extent of lymphadenectomy, showed for the first time that extensive lymphadenectomy (D3) resulted in significantly improved survival [28,