The prognostic role of perineural invasion (PNI) in gastric cancer remains unclear. We hypothesized that the diameter of the tumor-involved nerves might be a useful indicator for prognosis. By labeling nerves and cancer cells in 204 cases of gastric cancer with single or double immunochemistry, we found that 146 cases were PNI positive and that 58 were PNI negative. For each case with PNI, the maximum diameter of the involved nerve was measured microscopically. Then, we correlated this parameter with the patients' 5-year overall survival, and receiver operating curves were used to determine the cutoff value. We found that the optimal cutoff value for predicting 5-year survival was 65 µm (sensitivity 76.9%, specificity 70.0%). Next, all 204 patients were classified into two groups as follows: Group A, PNI-positive cases in which the largest involved nerves were ≥65 µm in diameter (110 cases); Group B, PNI-positive cases in which the largest involved nerves were P˂65 µm and all PNI-negative cases (94 cases). Compared with Group A, Group B had a better 5-year survival (74.5% vs 27.3%) and a better 5-year disease-free survival (63.8% vs 23.6%). Multivariate analysis suggested that a ≥65 µm maximum diameter of the involved nerves was an independent risk factor for both recurrence (P˂0.001) and gastric cancer-related death (P˂0.001) within 5 years. However, if all patients were classified simply based on whether PNI existed (regardless of the nerve size), this did not provide more information than traditional clinicopathological variables. In conclusion, the presence of cancer-involved nerves with a diameter ≥65 µm was a valuable prognostic factor for gastric cancer.