Innovation in Enrichment: Is Persistence Enough?
Innovation in Enrichment: Is Persistence Enough?
复制标题
丰富创新：坚持就够了吗？
DOI：
10.1097/ccm.0000000000006239
复制
发表时间：
2024
影响因子：
8.8
通讯作者：
Siempos,IliasI
中科院分区：
文献类型：
--
作者：
Schenck,EdwardJ;Siempos,IliasI
针对接受机械通气的患者的临床试验经常失败（1）。这种失败的原因有很多，从无法改变的患者特征导致不良结局到危重护理条件的异质性，时变性，重叠和综合征定义（2）。为了解开异质性，政府领导人（如美国食品和药物管理局[FDA]）已要求临床研究人员找到方法来识别可能在更大人群中被掩盖的治疗效果（3）。富集被认为是一种令人信服的方法，可以隔离更同质的人群，增加从潜在风险治疗中获益的可能性。FDA呼吁浓缩的核心是减少变异性的核心需求（3）。该指令旨在创建更有效的临床试验，例如，不招募可能通过当前护理迅速改善的患者。因此，针对持久性的富集是有价值的（4）。当考虑接受机械通气的低氧性呼吸衰竭（HRF）患者时，针对持久性富集的人群将暴露于治疗足够长的时间以实现目标。此外，持续机械通气通过多种机制、原发疾病过程和由于暴露于镇静、不动和损伤性机械通气的增加而导致的医源性增加了不良结局的风险。为了确定潜在的富集人群的持久性，机械通气临床试验的创新设计可能会有所帮助（5）。这种创新设计可以采用机器学习方法，这些方法擅长将群体内的个体分组，并且可以生成准确的预测模型。在这一期的重症监护医学中，Sathe等人（6）提出了一篇文章，建立了一个利用机器学习方法进行试验富集的框架。利用来自两个大型学术医学中心的两个转化重症监护队列，该团队首先定义了一个新的结果，
Clinical trials targeting patients who receive mechanical ventilation often fail (1). There are numerous causes of this failure, from unmodifiable patient characteristics resulting in poor outcomes to heterogeneous, time-varying, overlapping, and syndromic definitions of critical care conditions (2). To disentangle heterogeneity, governmental leaders (such as the US Food and Drug Administration [FDA]) have charged clinical researchers to find methods to identify treatment effects that may have been obscured in a larger population (3). Enrichment is seen as a compelling method to isolate a more homogeneous population with an increased likelihood of benefiting from a potentially risky therapy. At the core of the FDA’s call for enrichment is the central need to reduce variability (3). The directive is to create more efficient clinical trials by, for example, not enrolling patients who are likely to improve rapidly with current care. It follows that enrichment for persistence is valuable (4).When considering patients with hypoxemic respiratory failure (HRF) receiving mechanical ventilation, a population enriched for persistence would be exposed to a therapy for long enough for it to achieve a goal. Furthermore, persistent mechanical ventilation increases the risk of poor outcomes through several mechanisms, the primary disease process, and iatrogenesis due to an increased exposure to sedation, immobility, and injurious mechanical ventilation. To identify potential enriched populations for persistence, an innovative design of clinical trials in mechanical ventilation (5) might help. Such an innovative design may employ machine learning methods, which excel at grouping individuals within cohorts and can generate accurate predictive models. In this issue of Critical Care Medicine, Sathe et al (6) present an article that establishes a framework for the utilization of machine learning methods for trial enrichment. Leveraging two translational critical care cohorts from two large academic medical centers, the team first defined a novel outcome,