Discrimination and calibration of Framingham functions (FRS), recalibrated FRS and Health ABC function in men (n = 981).

Abbreviations: FRS: Framingham risk score; CHD: coronary heart disease; coef: coefficient; CI: confidence interval; SD: standard deviation; TC: total cholesterol; HDL-C: high-density lipoprotein cholesterol.
*Some of the Framingham risk factors categories were collapsed to avoid cells with limited numbers of events and /or unpredictive trends. The proportionality assumption was tested using the Therneau and Grambsch statistics, which is based on the Schoenfeld residuals. The assumption was accepted (p = 0.33).
†Based on Wilson et al. [2].
‡Cholesterol categories proposed by the National Cholesterol Education Program [24].
∥Blood pressure categories: Optimal (Systolic<120, Diastolic>80); Normal (Systolic<130, Diastolic>85); High normal (Systolic<140, Diastolic>90); Stage I (Systolic<160, Diastolic<100); Stage II–IV (Systolic ≥160, Diastolic ≥100) [24].
¶Estimated from the Framingham adjusted survival rate (survival rate at the mean value of the risk factors) at 10 years: S0(10) = 90015 [2], as: Ŝ0(7.5) = S0(10)0.75 = 0.9241 (exponential model).
#Kaplan-Meier survival function at t = 7.5 years on HABC data, similar to reference [24].
**Adjusted survival rate at t = 7.5 years obtained on the HABC cohort as the baseline survival functions of the multivariate Cox model, similar to reference [9].
††After bootstrap correction for the optimism (1000 bootstrap samples from the original dataset [20]), c-index = 0.580 (p = 0.90 for comparison with Framingham function).
∥ ∥Adaptation to the Cox model of the Hosmer-Lemeshow test of goodness of fit [33], comparing observed and expected failures within deciles of predicted risk. Larger p values indicate better calibration [21].

缩略语说明：FRS：弗雷明汉风险评分（Framingham risk score）；CHD：冠心病（coronary heart disease）；coef：系数（coefficient）；CI：置信区间（confidence interval）；SD：标准差（standard deviation）；TC：总胆固醇（total cholesterol）；HDL-C：高密度脂蛋白胆固醇（high-density lipoprotein cholesterol）。
*为避免事件数量过少的单元格以及/或无预测价值的趋势，部分弗雷明汉危险因素分类被合并。采用基于舍恩菲尔德残差（Schoenfeld residuals）的Therneau与Grambsch统计量检验比例风险假定，该假定得以满足（p=0.33）。
†基于Wilson等人[2]的研究。
‡美国国家胆固醇教育计划[24]提出的胆固醇分类标准。
∥血压分类标准：理想血压（收缩压<120，舒张压>80）；正常血压（收缩压<130，舒张压>85）；正常高值血压（收缩压<140，舒张压>90）；I级高血压（收缩压<160，舒张压<100）；II~IV级高血压（收缩压≥160，舒张压≥100）[24]。
¶基于弗雷明汉校正后生存率（危险因素均值处的生存率）估算10年生存率：S₀(10)=0.90015[2]，据此可得Ŝ₀(7.5)=S₀(10)^0.75=0.9241（指数模型）。
#基于HABC数据集的t=7.5年时的卡普兰-迈耶生存函数（Kaplan-Meier），与参考文献[24]一致。
**在HABC队列中得到的t=7.5年时的校正生存率，作为多变量Cox模型的基准生存函数，与参考文献[9]一致。
††经Bootstrap法校正偏移后（从原始数据集[20]中抽取1000个Bootstrap样本），一致性指数（c-index）为0.580（与弗雷明汉风险评分函数比较时p=0.90）。
∥∥针对Cox模型的Hosmer-Lemeshow拟合优度检验适配版本[33]，该检验通过比较预测风险十分位数组内的观测失败数与期望失败数进行评估。p值越大，模型校准性能越好[21]。