The current method for staging in gastric cancer is not sufficient as even after a complete primary tumor resection, patients with node-negative gastric cancer suffer from disease recurrence. In this study, the relation between disease recurrence and the presence of occult tumor cells (OTC) in lymph nodes from gastric cancer patients was evaluated. In a case-control design, lymph nodes from 40 cases (disease recurrence) and 41 controls (no disease recurrence and followed for at least five years) with gastric cancer were examined for the presence of OTC, that comprised micrometastases (MM; >0.2 mm and < or =2.0 mm) and isolated tumor cells (ITC; < or =0.2 mm). The original hematoxylin and eosin-stained sections of all lymph nodes from cases and controls were previously considered as tumor-negative by the local pathologist. Fresh hematoxylin and eosin-stained sections were screened by conventional microscopy. Histologic sections stained by immunohistochemistry with anticytokeratin antibodies CAM5.2 were screened by conventional and automated microscopy. Tumor cells were detected in lymph nodes from 40 of 81 (49%) patients. There was no significant difference in the presence of OTC, MM, or ITC between the case and control groups (P = 0.658, P = 0.691, P = 0.887, respectively). However, significantly more cases presented with 20% or more OTC-positive lymph nodes (P = 0.015). A multivariate logistic regression analysis showed that examination of less than five lymph nodes (odds ratio, 13.8; 95% confidence interval, 1.6-120.6, P = 0.018) was the only significant independent risk factor for disease recurrence, especially for locoregional disease recurrence (odds ratio, 20.4; 95% confidence interval, 2.2-190.8, P = 0.008). A similar analysis for distant disease recurrence showed a percentage of 20% or more OTC-positive lymph nodes to be the only significant independent risk factor (odds ratio, 15.6, 95% confidence interval, 1.6-151.4, P = 0.018). The sensitivity of immunohistochemistry evaluated by microscopy to identify cases with 20% or more OTC-positive lymph nodes increased from 8% for conventional microscopy to 22% for automated microscopy (McNemar's test, P = 0.063). The mere presence of OTC-positive lymph nodes in gastric cancer patients did not predict disease recurrence. However, the number of examined lymph nodes and the percentage of OTC-positive lymph nodes were independent risk factors for locoregional disease recurrence and distant disease recurrence, respectively. Automated microscopy was essential in identifying patients with 20% or more OTC-positive lymph nodes. Therefore, a maximum number of lymph nodes should be removed and meticulously examined for OTC to identify high-risk patients. These patients should be considered for additional treatment.