This study compared the prediction of recurrence based on detection of occult neoplastic cells (ONCs) in lymph nodes or using 3 criteria to identify high-risk patients among 72 patients who had Dukes' A colorectal cancer with or without proper muscle invasion. Predicting recurrence based on the detection of ONCs had a sensitivity of 40.0% (2/5) and a false-negative rate of 60.0% (3/5), while there was a specificity of 97.0% (65/67) and false-positive rate of 3.0% (2/67), resulting in an accuracy of 68.5%, PPV of 50.0% (2/4), and NPV of 95.6% (65/68). Predicting recurrence based on the presence of at least 2 of the 3 high-risk criteria showed a sensitivity of 60.0% (3/5) and a false-negative rate of 40.0% (2/5), while it had a specificity of 74.6% (50/67) and a false-positive rate of 25.4% (17/67), resulting in an accuracy of 67.3%, PPV of 15.0% (3/20), and NPV of 96.2% (50/52). These results suggest that a prediction based on ONCs was similar to use of the high-risk criteria, with both methods having a high specificity for recurrence/metastasis of Dukes' A colorectal cancer.