Platform of Randomized Adaptive Clinical Trials in Critical Illness

研究简介（原文）

PRACTICAL is a randomized multifactorial adaptive platform trial for acute hypoxemic respiratory failure (AHRF). This platform trial will evaluate novel interventions for patients with AHRF across a range of severity states (i.e., not intubated, intubated with lower or higher respiratory system elastance, requiring extracorporeal life support) and across a range of investigational phases (i.e., preliminary mechanistic trials, full-scale clinical trials). AHRF is a common and life-threatening clinical syndrome affecting millions globally every year. Patients with AHRF are at high risk of death and long-term morbidity. Patients who require invasive mechanical ventilation are at risk of ventilator-induced lung injury and ventilator-induced diaphragm dysfunction. New treatments and treatment strategies are needed to improve outcomes for these very ill patients. Utilizing advances in Bayesian adaptive trial design, the platform will facilitate efficient yet rigorous testing of new treatments for AHRF, with a particular focus on mechanical ventilation strategies and extracorporeal life support techniques as well as pharmacological agents and new medical devices. The platform is designed to enable evaluation of novel interventions at a variety of stages of investigation, including pilot and feasibility trials, trials focused on mechanistic surrogate endpoints for preliminary clinical evaluation, and full-scale clinical trials assessing the impact of interventions on patient-centered outcomes. A domain is defined as a set of interventions that are intended to act on specific mechanisms of injury using different variations of a common therapeutic strategy. A domain may also be a non-interventional study that addresses observational research questions by collecting specific data or outcomes that are not collected as part of other domains. Domains are intended to function independently of each other, allowing independent evaluation of multiple therapies and mechanistic pathways within the same patient. Once feasibility is established, Bayesian adaptive statistical modelling will be used to evaluate treatment efficacy at regular interim adaptive analyses of the pre-specified outcomes for each intervention in each domain. These adaptive analyses will compute the posterior probabilities of superiority, futility, inferiority, or equivalence for pre-specified comparisons within domains. Each of these potential conclusions will be pre-defined prior to commencing the intervention trial. Decisions about trial results (e.g., concluding superiority or equivalence) will be based on pre-specified threshold values for posterior probability. The primary outcome of interest, the definitions for superiority, futility, etc. (i.e., the magnitude of treatment effect) and the threshold values of posterior probability required to reach conclusions for superiority, futility etc., will vary from intervention to intervention depending on the phase of investigation and the nature of the intervention being evaluated. All of these parameters will be pre-specified as part of the statistical design for each intervention trial. In general, domains will be designed to evaluate treatment effect within four discrete clinical states: non-intubated patients, intubated patients with low respiratory system elastance (\<2.5 cm H2O/(mL/kg)), intubated patients with high respiratory system elastance (≥2.5 cm H2O/(mL/kg)), and patients requiring extracorporeal life support. Where appropriate, the model will specify dynamic borrowing between states to maximize statistical information available for trial conclusions. In this perpetual trial design, different interventions may be added or dropped over time. Where possible, the platform will be embedded within existing data collection repositories to enable greater efficiency in outcome ascertainment. Standardized systems for acquiring both physiological and biological measurements are embedded in the platform, to be acquired at sites with appropriate training, expertise, and facilities to collect those measurements.

适应症