Structural Heart Disease Program, Mid-Atlantic Permanente Medical Group, McLean, VA.
Pharm3r LLC, New York, NY.
Am Heart J. 2018 Apr;198:64-74. doi: 10.1016/j.ahj.2017.10.013. Epub 2018 Feb 3.
Although outcomes data on transcatheter aortic valve replacement (TAVR) and transcatheter mitral valve repair (Mitraclip) are available via the Transcatheter Valve Therapy (TVT) registry, dissemination of TVT data is often delayed. The Food and Drug Administration's Manufacturer and User Facility Device Experience (MAUDE) system collects postmarket outcomes data in public medical device reports. We used natural language processing to compare the event rates for TAVR and Mitraclip in the TVT registry and from MAUDE data.
We identified all medical device reports related to TAVR and Mitraclip from December 2011 through December 2014. Our primary objective was to demonstrate that event rates in TVT and MAUDE were not significantly different. We also compared TVT event rates for TAVRs performed in 2014 to MAUDE-derived event rates for the Sapien XT and CoreValve devices, both Food and Drug Administration-approved in 2014.
Regression analysis demonstrated close correlation between TVT and MAUDE rates for both TAVR and Mitraclip, with R values of 0.86 and 0.77, respectively. The rates for all events except bleeding were not statistically significantly different. We demonstrated similar increased rates of permanent pacemaker implantation in the 2014 TVT and MAUDE data sets consistent with approval of the CoreValve.
We demonstrated that natural language processing technology sorted through raw MAUDE data, allowing identification of the most common events associated with TAVR and Mitraclip procedures, and that MAUDE-derived event rates were not statistically significantly different from TVT event rates. This technology has important public health implications because it improves postmarket surveillance of implantable devices and permits rapid and early dissemination of vital information.
尽管经导管主动脉瓣置换术（TAVR）和经导管二尖瓣修复术（Mitraclip）的治疗结果数据可通过经导管瓣膜治疗（TVT）注册中心获得，但 TVT 数据的传播通常会延迟。食品和药物管理局的制造商和用户设施设备经验（MAUDE）系统在公开的医疗器械报告中收集上市后结果数据。我们使用自然语言处理来比较 TVT 注册中心和 MAUDE 数据中 TAVR 和 Mitraclip 的事件发生率。
我们从 2011 年 12 月至 2014 年 12 月确定了所有与 TAVR 和 Mitraclip 相关的医疗器械报告。我们的主要目标是证明 TVT 和 MAUDE 的事件发生率没有显著差异。我们还比较了 2014 年进行的 TAVR 的 TVT 事件发生率与 2014 年获得 FDA 批准的 Sapien XT 和 CoreValve 设备的 MAUDE 衍生事件发生率。
回归分析表明，TAVR 和 Mitraclip 的 TVT 和 MAUDE 发生率之间存在密切相关性，R 值分别为 0.86 和 0.77。除出血外，所有事件的发生率均无统计学差异。我们证明了在 2014 年 TVT 和 MAUDE 数据集中，与 CoreValve 批准一致，永久性起搏器植入的发生率均有所增加。
我们证明了自然语言处理技术可以对原始 MAUDE 数据进行分类，从而识别与 TAVR 和 Mitraclip 手术最常见的相关事件，并且 MAUDE 衍生的事件发生率与 TVT 事件发生率无统计学差异。这项技术具有重要的公共卫生意义，因为它提高了对植入式设备的上市后监测，并允许快速和早期传播重要信息。
