Predicting recurrence and metastasis of stage III/Dukes' C colorectal cancer with lymph node metastasis.

This study was designed to compare the prediction of recurrence based on detection of occult neoplastic cells (ONCs) in lymph nodes or using specific criteria to identify patients at high risk of recurrence/metastasis among 105 patients with Dukes' C colorectal cancer. Prediction of recurrence based on the detection of ONCs had a sensitivity of 50.0% (22/44), specificity of 80.3% (49/61), and an accuracy of 65.2%. Prediction of recurrence based on positivity for at least 2 of the 3 high-risk criteria had a sensitivity of 54.5% (24/44), specificity of 83.6% (51/61), and an accuracy of 69.1%. Among the 34 patients with ONCs, prediction of recurrence based on positivity for all 3 high-risk criteria had a sensitivity of 27.3% (6/22), specificity of 91.7% (11/12), an accuracy of 59.5%, and a positive predictive value (PPV) of 85.7% (6/7). These results suggest that the predictive value of ONCs and the high-risk criteria was similar, and that recurrence is likely to occur in patients who fulfill < or =2 of the high-risk criteria. Accordingly, combined use of these parameters may be more effective for the early prediction of recurrence/metastasis to assist in the choice of postoperative systemic chemotherapy.