A method for individualising the risk of a negative lymph node classification error in cancer of the colon

In cancer of the colon, the number of lymph nodes that should be analysed before a patient is classified as free of lymph node involvement has been widely discussed. A mathematical model is proposed which is based on the Bayes Theorem for calculating the probability of error (PE) similar to that normally used to evaluate a diagnostic test, but adapted to a quantitative variable, the lymph node count. The clinical histories of 480 patients routinely operated on in attempt to cure cancer of the colon were reviewed. Cases with any kind of metastasis were excluded. The proposed formula based on the Bayes Theorem was applied with the aim of calculating the PEs for the complete series and for different patient sub-groups (T2, T3, and T4 tumours). For the probabilities of error of classifying a patient as N negative, which varied between 5% and 1% (near or practically 0), the minimum number of negative lymph nodes required for analysis fluctuated between 7 and 17, respectively, for the complete series. This minimum figure was also variable for the different sub-groups (T2, T3, and T4 tumours) studied. These numbers mainly depended on the case characteristics of a specific study group as regards the prevalence of the N+ cases that they dealt with, and of its historically demonstrated ability to collect and identify positive lymph nodes in those patients that had them. From a mathematical point of view, the minimum number of lymph nodes that have to be analysed in cancer of the colon in order to classify a patient as N negative is not a constant. This depends on the error that is prepared to be assumed for that diagnosis, possibly depending on certain tumour traits, and also may be adapted to the cases of each study group.