Performance of the various TNM staging systems, BCLC and Hangzhou criteria systems.

In the univariate model, the highest χ2 by the linear trend test, likelihood ratio (LHR) test, and highest c-index were considered to indicate best prognostic performance in terms of discriminatory ability, homogeneity, and monotonicity of gradients. In the multivariate Cox regression model,the independent contributions of each staging system to the full model were evaluated by removing the system concerned. AIC, Akaike information criterion; TNM, tumor-node- etastasis; BCLC, Barcelona Clinic Liver Cancer; HC, Hangzhou criteria.

在单变量模型中，以线性趋势检验得到的最高卡方（χ²）、似然比（LHR）检验结果，以及最高一致性指数（c-index）作为判定最优预后性能的指标，分别对应判别能力、同质性以及梯度单调性三个维度的表现。在多变量Cox回归模型中，通过逐一移除某一分期系统，即可评估该系统对完整模型的独立贡献程度。其中，AIC为赤池信息准则（Akaike information criterion, AIC）；TNM为肿瘤-淋巴结-转移（tumor-node-metastasis, TNM）分期系统；BCLC为巴塞罗那临床肝癌（Barcelona Clinic Liver Cancer, BCLC）分期系统；HC为杭州标准（Hangzhou criteria, HC）。