Prospective External Validation of Three Preoperative Risk Scores for Prediction of New Onset Atrial Fibrillation After Cardiac Surgery
Prospective External Validation of Three Preoperative Risk Scores for Prediction of New Onset Atrial Fibrillation After Cardiac Surgery
复制标题
DOI：
10.1213/ane.0000000000002112
复制
发表时间：
2018-01-01
影响因子：
5.7
通讯作者：
Dupuis, Jean-Yves
中科院分区：
文献类型：
--
作者：
Cameron, Matthew J.;Tran, Diem T. T.;Dupuis, Jean-Yves
背景：术后心房颤动（POAF）与心脏手术患者的早期和晚期发病率和死亡率相关。预防性治疗房颤（AF）已被推荐用于改善心脏手术患者发生POAF的高风险预后。为了实现这一目标，需要可靠的POAF预测模型。本研究试图从外部验证为心脏手术患者术前预测POAF提出的3个风险模型：POAF评分、CHA(2)DS(2)-VASc评分和房颤风险指数。方法：这是一项前瞻性队列研究，纳入了2014年2月至2015年9月在单一心脏外科中心接受非紧急冠状动脉搭桥术和/或瓣膜手术的1416名成年患者。对每位患者分别计算3种预测模型的风险评分。所有患者随访长达2周，或直到出院，观察新发房颤需要治疗的主要结局。用受试者工作特征曲线评估鉴别。使用Pearson(2)拟合优度检验和校准图评估校准。与治疗所有患者或不治疗任何患者的策略相比，根据POAF风险实施房颤预防的评分效用通过净效益分析进行评估。结果：本研究纳入的1416例患者中，478例（33.8%）达到主要结局。在对评分进行验证的人群亚群中，预测POAF的受试者工作特征曲线下的区域如下：POAF评分为0.651(95%可信区间[CI]， 0.621-0.681)， CHA(2)DS(2)-VASc评分为0.593 (95% CI, 0.557-0.629)（与POAF评分相比，P < 0.001，与房颤风险指数相比，P < 0.222），房颤风险指数为0.563 (95% CI, 0.522-0.604)（与POAF评分相比，P < 0.001）。校正分析表明，预测模型在POAF的观测率和预期率之间有较差的拟合。净收益分析显示，与治疗所有患者的预防策略相比，基于这些评分并针对中度或高风险POAF患者的房颤预防策略改善了决策。结论：本研究评估的3种预测评分对心脏外科患者POAF的预测能力有限。尽管如此，与适用于所有患者的预防策略相比，它们可能对针对中度或高风险PAOF患者的预防策略有用。
BACKGROUND: Postoperative atrial fibrillation (POAF) is associated with early and late morbidity and mortality of cardiac surgical patients. Prophylactic treatment of atrial fibrillation (AF) has been recommended to improve outcome in cardiac surgical patients at high risk of developing POAF. Reliable models for prediction of POAF are needed to achieve that goal. This study attempted to externally validate 3 risk models proposed for preoperative prediction of POAF in cardiac surgical patients: the POAF score, the CHA(2)DS(2)-VASc score, and the Atrial Fibrillation Risk Index.METHODS: This was a prospective cohort study of 1416 adult patients who underwent nonemergent coronary artery bypass graft and/or valve surgery in a single cardiac surgical center between February 2014 and September 2015. A risk score for each of the 3 prediction models was calculated in each patient. All patients were followed for up to 2 weeks, or until hospital discharge, to observe the primary outcome of new onset AF requiring treatment. Discrimination was assessed using receiver operating characteristic curves. Calibration was assessed using the Pearson (2) goodness-of-fit test and calibration plots. Utility of the score to implement AF prophylaxis based on the risk of POAF, in comparison to strategies of treating all patients, or not treating any patients, was assessed via a net benefit analysis.RESULTS: Of the 1416 patients included in this study, 478 had the primary outcome (33.8%). The areas under the receiver operating characteristic curve for prediction of POAF in the population subsets for which the scores were validated were as follows: 0.651 (95% confidence interval [CI], 0.621-0.681) for the POAF score, 0.593 (95% CI, 0.557-0.629) for the CHA(2)DS(2)-VASc score (P < .001 versus POAF score, P < .222 versus Atrial Fibrillation Risk Index), and 0.563 (95% CI, 0.522-0.604) for the Atrial Fibrillation Risk Index (P < .001 versus POAF score). The calibration analysis showed that the predictive models had a poor fit between the observed and expected rates of POAF. Net benefit analysis showed that AF preventive strategies based on these scores, and targeting patients with moderate or high risk of POAF, improve decision-making in comparison to preventive strategies of treating all patients.CONCLUSIONS: The 3 prediction scores evaluated in this study have limited ability to predict POAF in cardiac surgical patients. Despite this, they may be useful in preventive strategies targeting patients with moderate or high risk of PAOF in comparison with preventive strategies applied to all patients.