基于列线图的肿瘤风险预测及其效果评价

Prediction of Tumor Risk Based on Nomograms and Evaluation of Its Effect

目的:基于R软件介绍肿瘤风险预测模型的可视化-列线图的绘制和模型评价。方法:以SPSS软件自带案例乳腺癌生存数据为例,基于R软件包survival,rms拟合Cox回归模型并绘制生存概率风险列线图,用一致性指数(C-index)和校准曲线(calibration plots)评价模型区分度和一致性。结果:模型预测因素包括年龄、孕酮受体状态、淋巴结转移、病理学肿瘤大小。基于模型绘制的列线图评估个体化发病预后风险简单直观,模型一致性指数:0.70(0.62-0.80),校准曲线3年、5年、10年的预测生存概率与实际生存概率基本一致,反映该模型有较高的区分度(Discrimination)和一致性(Accuracy)。结论:基于Cox模型的列线图可通过多种预测因素量化肿瘤发病风险,应用简单直观、使用方便.可以为临床决策提供支持。

Objective: To introduce how to draw and evaluate nomogram, a medical visualization of cancer prediction model in R software. Methods: Taking the survival database for breast cancer from SPSS as an example, we drew nomogram based on cox model using survival and rms packages with R software. C-index and the calibration plot were used to evaluate model discrimination and accuracy. Results: Age, progesterone receptor status, lymphatic metastasis, tumor size were independent prognostic factors. The value of C-index was 0. 70 (0. 62 ~ 0. 80), The predicted survival probability by the calibration plot (3/5/10 year) was basically consistent with the actual survival probability, which indicates that the model has high discrimination and accuracy. Conclusion: Nomograms based on cox model can be used to predict the incidence of cancer through various factors. This easy and direct tool may help clinicians in decision making.