Background: The appropriate extent of lymph node clearance during gastrectomy for cancer remains controversial.
Methods: Medline, Embase, the Cochrane register and other databases were searched for studies reporting node dissection technique, 5 year survival and mortality after gastrectomy. Comparisons with systematic bias in treatment allocation and patients who received perioperative chemotherapy were excluded. Meta-analysis was performed separately for randomized and non-randomized comparisons.
Results: Two randomized and two non-randomized comparisons of limited (D1) versus extended (D2) node dissection and 11 reports of one dissection type were analysed. For D2 the randomised trials showed no overall survival benefit (Risk ratio (RR) = 0.95, 95 per cent c.i. 0.83-1.09) and an increased postoperative mortality (RR = 2.23, c.i. 1.45-3.45), apparently related to pancreatico-splenectomy and surgical inexperience. A trend towards survival benefit for D2 was observed for T3+ tumours (RR = 0.68, c.i. 0.42-1.10). Non-randomized comparisons found no survival benefit for D2 (RR = 0.92, c.i. 0.83-1.02), but decreased postoperative mortality (RR = 0.65, c.i. 0.45-0.93). Nine observational studies of D2 reported better results than two studies of D1 surgery, but in very different settings.
Conclusions: Evidence for D2 dissection is inconclusive. No overall survival advantage has emerged, but some patients with intermediate stage disease may benefit. Excess operative mortality appears to be associated with pancreatico-splenectomy, low case volume and lack of specialist training.