Thrombotic events are a major complication of hypertrophic cardiomyopathy (HCM). There are only two available risk prediction models for thrombotic events, HCM Risk-CVA score and French HCM score developed by foreign scholars, yet the former one has been found to have limited predictive value in Chinese HCM patients.

To develop a risk prediction model for thrombotic events in Chinese patients with HCM.

A retrospective cohort study design was used. Five hundred and thirty-seven HCM patients who admitted to West China Hospital of Sichuan University from 2010 to 2018 were recruited. Post-discharge health status was collected by use of telephone follow-up or checking the treatment status recorded in the electronic medical record system once every 6 to 12 months until a composite thrombotic event (defined as the endpoint event) or death occurred or the determined thrombotic risk assessment day of this study (2019-12-31). Univariate and multvariate Cox regression analyses were applied to build a thrombotic risk prediction model, and its internal validation was tested in a resample using the bootstrapping technique.

Due to data missing, 24 cases were excluded, and the other 513 cases were finally included. During a median follow-up of 4.2 years (IQR: 1.3-6.2 years), thrombotic events occurred in 42 cases (8.18%), with an annual morbidity rate of 2.10%〔95%CI (1.47%, 2.73%) 〕. By multivariate Cox regression analysis, age, prior thrombotic event and left ventricular ejection fraction (LVEF) were identified (P<0.05) and used for constructing the formula of SAAE score (S=prior stroke and other thrombotic events, A=atrial fibrillation, A=age, E=LVEF) for predicting thrombotic events after being weighted based on the regression coefficient. Internal validation suggested that SAAE score could discriminate thrombotic events in the whole population {Harrell's C-index=0.773〔95%CI (0.688, 0.858) 〕}, with a calibration slope of 1.006, and could well discriminate 1-year, 3-year and 5-year thrombotic events (Harrell's C-index=0.790, 0.799, 0.735), with a good calibration ability. In addition, the SAAE score also performed well in distinguishing thrombotic events in patients with or without atrial fibrillation {Harrell's C-index=0.669〔95%CI (0.548, 0.791) 〕, 0.647〔95%CI (0.498, 0.795) 〕}, with good calibration ability. Besides that, SAAE score could partially discriminate 1-year, 3-year and 5-year thrombotic events in these two groups, with certain calibration ability. For three groups (whole study population, patients with/without atrial fibrillation), SAAE score could discriminate the risk of thrombotic events (either low, moderate or high) excellently. For the whole study population, SAAE score was better than HCM Risk-CVA score in distinguishing thrombotic events (P=0.013). Decision curve analysis showed the net benefit of SAAE score was better than HCM Risk-CVA score at different prediction time points (1, 3 and 5 years) .

This thrombotic events risk prediction model developed by us for Chinese HCM patients, namely SAAE score, could well stratify the risk of thrombotic events.

血栓栓塞（TE）事件是肥厚型心肌病（HCM）的重要并发症。目前针对HCM患者TE事件的风险预测，仅国外学者构建了两个模型：HCM Risk-CVA及French HCM score，然而，现有研究发现HCM Risk-CVA模型对于中国HCM患者的临床价值较为有限。

本研究拟构建适合中国HCM患者的TE事件风险预测模型。

本研究系回顾性队列研究，收集2010—2018年在四川大学华西医院就诊的537例HCM患者的病例资料。本研究通过电话随访或电子病历系统查询患者就诊记录，每6~12个月随访1次，直至出现终点事件或死亡或研究拟定的评估日期（2019-12-31），终点事件定义为复合性TE事件。采用单因素和多因素Cox回归分析构建风险预测模型，并使用自助重抽样的方法进行内部验证。

537例患者中，24例患者有不同程度的数据缺失，最终纳入513例患者。中位随访时间为4.2（1.3，6.2）年，随访过程中42例（8.18%）发生TE事件，年发病率为2.10%〔95%CI（1.47%，2.73%）〕。根据多因素Cox回归模型构建TE事件风险预测模型，最终纳入年龄、既往TE事件、心房颤动及左心室射血分数（LVEF）（P<0.05）。根据回归系数进行权重后，分别赋值构建SAAE score评分模型，即：S=既往脑卒中（stroke）等TE事件，A=心房颤动（atrial fibrillation），A=年龄（age），E=左心室射血分数（left ventricular ejection fraction）。内部验证提示SAAE score在整体人群中区分TE事件的Harrell's C-指数为0.773〔95%CI（0.688，0.858）〕，校准斜率为1.006；同时，SAAE score在整体人群中对1、3、5年区分TE事件的Harrell's C-指数分别为0.790、0.799及0.735，校准能力较好。此外，SAAE score在合并/不合并心房颤动的人群中区分TE事件的Harrell's C-指数分别为0.669〔95%CI（0.548，0.791）〕及0.647〔95%CI（0.498，0.795）〕，校准能力也较好；同时，在此两类人群中，该模型对1、3、5年TE事件也具有一定价值的区分及校准能力。对于整体人群、合并/不合并心房颤动人群，根据SAAE score进行高危、中危及低危分层后，均可较好区分TE事件的发生。对于整体人群，SAAE score对于TE的区分优于HCM Risk-CVA模型（P=0.013）。临床决策曲线结果提示在不同预测时间点（1、3、5年），SAAE score的净获益均优于HCM Risk-CVA。

本研究在中国HCM患者中构建了针对TE事件的风险预测模型，即SAAE score，该模型可较好地对HCM患者进行TE事件风险分层。