Burke H B, Goodman P H, Rosen D B, Henson D E, Weinstein J N, Harrell F E, Marks J R, Winchester D P, Bostwick D G
Department of Medicine, New York Medical College, Valhalla 10595, USA.
Cancer. 1997 Feb 15;79(4):857-62. doi: 10.1002/(sici)1097-0142(19970215)79:4<857::aid-cncr24>3.0.co;2-y.
The TNM staging system originated as a response to the need for an accurate, consistent, universal cancer outcome prediction system. Since the TNM staging system was introduced in the 1950s, new prognostic factors have been identified and new methods for integrating prognostic factors have been developed. This study compares the prediction accuracy of the TNM staging system with that of artificial neural network statistical models.
For 5-year survival of patients with breast or colorectal carcinoma, the authors compared the TNM staging system's predictive accuracy with that of artificial neural networks (ANN). The area under the receiver operating characteristic curve, as applied to an independent validation data set, was the measure of accuracy.
For the American College of Surgeons' Patient Care Evaluation (PCE) data set, using only the TNM variables (tumor size, number of positive regional lymph nodes, and distant metastasis), the artificial neural network's predictions of the 5-year survival of patients with breast carcinoma were significantly more accurate than those of the TNM staging system (TNM, 0.720; ANN, 0.770; P < 0.001). For the National Cancer Institute's Surveillance, Epidemiology, and End Results breast carcinoma data set, using only the TNM variables, the artificial neural network's predictions of 10-year survival were significantly more accurate than those of the TNM staging system (TNM, 0.692; ANN, 0.730; P < 0.01). For the PCE colorectal data set, using only the TNM variables, the artificial neural network's predictions of the 5-year survival of patients with colorectal carcinoma were significantly more accurate than those of the TNM staging system (TNM, 0.737; ANN, 0.815; P < 0.001). Adding commonly collected demographic and anatomic variables to the TNM variables further increased the accuracy of the artificial neural network's predictions of breast carcinoma survival (0.784) and colorectal carcinoma survival (0.869).
Artificial neural networks are significantly more accurate than the TNM staging system when both use the TNM prognostic factors alone. New prognostic factors can be added to artificial neural networks to increase prognostic accuracy further. These results are robust across different data sets and cancer sites.
TNM分期系统的出现是为了满足对准确、一致、通用的癌症预后预测系统的需求。自20世纪50年代引入TNM分期系统以来，已识别出了新的预后因素，并开发了整合预后因素的新方法。本研究比较了TNM分期系统与人工神经网络统计模型的预测准确性。
对于乳腺癌或结直肠癌患者的5年生存率，作者将TNM分期系统的预测准确性与人工神经网络（ANN）的预测准确性进行了比较。应用于独立验证数据集的受试者工作特征曲线下面积是准确性的衡量指标。
对于美国外科医师学会患者护理评估（PCE）数据集，仅使用TNM变量（肿瘤大小、区域阳性淋巴结数量和远处转移），人工神经网络对乳腺癌患者5年生存率的预测显著比TNM分期系统更准确（TNM为0.720；ANN为0.770；P<0.001）。对于美国国立癌症研究所的监测、流行病学和最终结果乳腺癌数据集，仅使用TNM变量，人工神经网络对10年生存率的预测显著比TNM分期系统更准确（TNM为0.692；ANN为0.730；P<0.01）。对于PCE结直肠癌数据集，仅使用TNM变量，人工神经网络对结直肠癌患者5年生存率的预测显著比TNM分期系统更准确（TNM为0.737；ANN为0.815；P<0.001）。在TNM变量中加入常见收集的人口统计学和解剖学变量进一步提高了人工神经网络对乳腺癌生存率（0.784）和结直肠癌生存率（0.869）预测的准确性。
当单独使用TNM预后因素时，人工神经网络比TNM分期系统显著更准确。可以将新的预后因素添加到人工神经网络中以进一步提高预后准确性。这些结果在不同数据集和癌症部位均具有稳健性。
