Background

QTc-prolongation is an independent risk factor for developing life-threatening arrhythmias. Risk management of drug-induced QTc-prolongation is complex and digital support tools could be of assistance. Bindraban et al. and Berger et al. developed two algorithms to identify patients at risk for QTc-prolongation.

Objective

The main aim of this study was to compare the performances of these algorithms for managing QTc-prolonging drug-drug interactions (QT-DDIs).

Materials and Methods

A retrospective data analysis was performed. A dataset was created from QT-DDI alerts generated for in- and outpatients at a general teaching hospital between November 2016 and March 2018. ECGs recorded within 7 days of the QT-DDI alert were collected. Main outcomes were the performance characteristics of both algorithms. QTc-intervals of > 500 ms on the first ECG after the alert were taken as outcome parameter, to which the performances were compared. Secondary outcome was the distribution of risk scores in the study cohort.

Results

In total, 10,870 QT-DDI alerts of 4987 patients were included. ECGs were recorded in 26.2 % of the QT-DDI alerts. Application of the algorithms resulted in area under the ROC-curves of 0.81 (95 % CI 0.79–0.84) for Bindraban et al. and 0.73 (0.70–0.75) for Berger et al. Cut-off values of ≥ 3 and ≥ 6 led to sensitivities of 85.7 % and 89.1 %, and specificities of 60.8 % and 44.3 % respectively.

Conclusions

Both algorithms showed good discriminative abilities to identify patients at risk for QTc-prolongation when using ≥ 2 QTc-prolonging drugs. Implementation of digital algorithms in clinical decision support systems could support the risk management of QT-DDIs.

中文翻译：

---

支持QTc延长药物之间药物相互作用的两种支持药物监测的算法的比较

背景

QTc延长是发展危及生命的心律不齐的独立危险因素。药物引起的QTc延长的风险管理非常复杂，数字支持工具可能会有所帮助。Bindraban等。和Berger等。开发了两种算法来识别有QTc延长风险的患者。

目的

这项研究的主要目的是比较这些算法在管理QTc延长药物-药物相互作用（QT-DDI）中的性能。

材料和方法

进行回顾性数据分析。根据2016年11月至2018年3月期间为普通教学医院的门诊和门诊患者生成的QT-DDI警报创建了数据集。收集了在QT-DDI警报7天内记录的ECG。主要结果是两种算法的性能特征。警报后第一个ECG的QTc间隔> 500 ms作为结果参数，并将其与性能进行比较。次要结果是研究组中风险评分的分布。

结果

总共包括4,987名患者的10,870个QT-DDI警报。在26.2％的QT-DDI警报中记录了ECG。对于Bindraban等人，算法的应用导致ROC曲线下的面积为0.81（95％CI 0.79–0.84）。而Berger等人则为0.73（0.70–0.75）。截止值≥3和≥6导致敏感性分别为85.7％和89.1％，特异度分别为60.8％和44.3％。

结论

当使用≥2种延长QTc的药物时，这两种算法都具有很好的判别能力，可以识别出存在QTc延长风险的患者。在临床决策支持系统中实施数字算法可以支持QT-DDI的风险管理。