OBJECTIVES: In patients with gastric cancer, identifying those at early stage is important to decide between endoscopic or surgical resection. We aimed to review the value of magnetic resonance (MRI), computed tomography (CT) and endoscopic ultrasound (EUS) to assess disease extension in terms of tumour size (T) and node involvement (N).

METHODS: We searched Medline, Cochrane Library and EMBASE (February, 2018) to identify studies that: a) reported accuracy of EUS, CT and MRI to detect T (T vs Tor Tvs T) and N (N0 vs N+) disease in patients with proven gastric adenocarcinoma, or b) reported a cost-effectiveness analysis. For accuracy studies, we included only those which assessed patient prior to surgery, directly compared pairs of tests and used post-surgery pathology as reference standard. We performed meta-analysis using bivariate mixed-effects model to estimate sensitivity and specificity. We used the GRADE system to assess the certainty of evidence.

RESULTS: Sixteen studies reported direct comparisons, a) MRI vs CT (six studies, 285 patients): to assess N+ disease, sensitivity was greater for MRI (80% vs 58%), while specificity was greater for CT (67% vs 91%); b) CT vs EUS (ten studies, 1543 patients): for N+ disease, sensitivity was greater for CT (73% vs 68%) and specificity was greater for EUS (77% vs 86%); c) MRI vs EUS (two studies, 90 patients): to detect Tinvolvement specificity was greater for EUS (70% vs 90%), while for assess N+ disease sensitivity was greater for EUS (77% vs 84%). Overall, tests showed lower accuracy when used smaller T disease threshold. Certainty of evidence was low due to imprecision. We did not identify cost-effectiveness studies.

CONCLUSIONS: Staging test mainly differ in the assessment of N disease. However, studies with larger sample size or evaluating clinical outcomes are required to assess the value of these tests.