[Role of a computer program in gastric cancer surgery - beyond the evidence]

Forty percent of patients with gastric cancer have an unnecessarily extended lymph node dissection with a higher rate of morbidity and mortality. While the Maruyama computer program (MCP) can estimate the lymph node involvement before the surgery, the Maruyama Index (MI) could be a good predictor of overall and disease free survival.

To measure the probability calculations by MCP, we had to define different "cut-off" levels, with using the calculation of the receiver-operating characteristics analysis. The long term oncological results, as the overall survival (OS) and disease free survival (DFS) were calculated in correlation with the extension of lymphadenectomy (D1 versus D2) and Maruyama Index (MI < 5 versus MI ≥ 5).

74 patients were investigated by the Maruyama computer program preoperatively for the short-term results, and the data of 101 patients were eligible for evaluation of the long-term oncological outcomes. The MCP had a 90.2% of sensitivity, 63.3% of specificity and 78.4% of accuracy. The positive predictive value was 75.5% and the negative predictive value was 84%. In D1 group the DFS was 93.6 months and 68.7 months in D2 group (p = 0.41; HR = 1.34), and the OS was 74.6 and 72.2 months respectively (p = 0.66; HR = 0.87). In patients with MI < 5 the DFS was 92 months and 62.5 months in patients with MI ≥ 5 (p = 0.31; HR = 1.4), while the OS was 86 months and 60.4 months (p = 0.17; HR = 1.52).

Our results proved, that the computerized prediction of LN metastases is efficient and the long term results suggest, that the MI < 5 has a better impact on survival, than the D-level guided surgery.