The use of neural networks in identifying risk factors for lymph node metastasis and recommending management of t1b esophageal cancer.

The objective of this study was to establish a prediction model of lymph node status in T1b esophageal carcinoma and define the best squamous and adenocarcinoma predictors. The literature lacks a satisfactory level of evidence of T1b esophageal cancer management. We performed an analysis pooling the effects of outcomes of 2098 patients enrolled into 37 retrospective studies using "neural networks" as data mining techniques. The percentages for lymph node, lymphatic (L+), and vascular (V+) invasion in Sm1 esophageal cancers were 24, 46, and 20 per cent, respectively. The same parameters apply to Sm2 with 34, 63, and 38 per cent as opposed to Sm3 with 51, 69, and 47 per cent. The respective number of patients with well, moderate, and poor histologic differentiation totaled 267, 752, and 582. The rank order of the predictors of lymph node positivity was, respectively: Grade III, (L+), (V+), Sm3 invasion, Sm2 invasion, and Sm1 invasion. Histologic-type squamous and adenocarcinoma (ADC/SCC) was not included in the model. The best predictors for SCC lymph node positivity were sm3 invasion and (V+). As concerns ADC, the most important predictor was (L+). Submucosal esophageal cancer should be managed with surgical resection. However, this is subject to the histologic type and presence of specific predictors that could well alter the perspective of multimodality management.