Blast from the past: Perioperative use of the Maruyama computer program for prediction of lymph node involvement in the surgical treatment of gastric cancer following neoadjuvant chemotherapy

Abstract

BackgroundSurgical quality assurance is a key element of gastric cancer treatment. The Maruyama Computer Program (MCP) allows to predict lymph node involvement in stations no. 1–16. The aim of the current study was to evaluate the accuracy of the MCP predictions in GC patients treated with neoadjuvant chemotherapy (nCTH) followed by gastrectomy with adequate lymphadenectomy. Methods101 patients who underwent preoperative nCTH followed by D2 gastrectomy with curative intent were analysed. The response to nCTH was measured using the tumour regression grade system. ResultsTest sensitivity, specificity, PPV, NPV and accuracy of the MCP were 92%, 33%, 41%, 89%, and 53%, respectively. In patients with response to nCTH, number of false positive (FP) results was significantly higher than in patients who did not respond to nCTH both in the N1 (56.3% vs 28.9%, p < 0.0001) and in the N2 (59% vs 41%, p < 0.0001) trier. The risk for FP results was 6 times higher in N1 (OR = 6.50, 95%CI: 3.91–10.82,; p < 0.0001) and N2 (OR = 5.84, 95%CI: 2.85–11.96; p < 0.0001) triers. In patients with intestinal type GC, the risk for FP results was 4 times higher than in other histologic types of GC in both N1 (OR = 4.23, 95%CI: 2.58–6.95; p < 0.0001) and N2 (OR = 4.23, 95%CI: 2.02–9.62; p = 0.0002) triers. ConclusionsMCP predictions in the GC patients treated with nCTH have low specificity due to significantly high number of FP results. Noticeably low accuracy level of predictions indicate a need for new prediction models, based on Laurén classification, since it may provide some information on expected regression grade.