Background:
Randomised clinical trials (RCTs) are the gold standard for evaluating intervention effects, however, conventional RCTs are bureaucratic, costly, inflexible, and often inconclusive. Adaptive platform trials are increasingly used as they can reduce barriers and are more flexible, and thus come with a higher probability of obtaining conclusive results faster at lower costs.
Objectives:
The Intensive Care Platform Trial (INCEPT) will be used to assess the effects of interventions used in adults acutely admitted to the intensive care unit (ICU).
Design:
INCEPT is an investigator-initiated, pragmatic, randomised, embedded, multifactorial, international, adaptive platform trial. INCEPT uses adaptive stopping and arm-dropping rules, as well as fixed and response-adaptive randomisation. Specific domains may be either open label or blinded.
Domains and interventions:
Comparable groups of interventions will be nested in domains, which have conceptual similarities with stand-alone randomised trials. Domains will continuously be added to INCEPT and conducted following domain-specific appendices to the core protocol.
Inclusion and exclusion criteria:
Adults acutely admitted to the ICU will be screened if they are eligible for at least one active domain. The only platform-level exclusion criteria are 1) informed consent after inclusion expected to be unobtainable and 2) patients admitted under coercive measures. Additional inclusion and exclusion criteria will be domain-specific.
Stakeholder involvement:
Stakeholder involvement is central in INCEPT and ensured through a central advisory board comprising various key stakeholders, and consultations with national and international research panels consisting of ICU survivors, family members, clinicians, and researchers. Stakeholders will be involved in the development of the overall platform trial and specific domains with pre-specified minimum requirements for involvement.
Outcomes:
Each domain will use one of the core outcomes (defined elsewhere in the registration) as the primary outcome and the guiding outcome driving all adaptations.
Statistical methods Primary analyses will generally be conducted in the intention-to-treat population of each domain. INCEPT primarily uses Bayesian statistical methods with neutral priors conveying either minimal information or some scepticism, although specific domains may use conventional, frequentist statistical methods. Outcomes will generally be analysed using logistic and linear regression models adjusted for pre-specified anticipated prognostic baseline characteristics, followed by calculation of sample-average estimates and intervention effects using G-computation. Results will be presented for each intervention and comparisons presented on both the absolute (risk differences and mean differences) and relative (risk ratios and ratios of means) scales with 95% credible intervals and probabilities of superiority. INCEPT will generally use constant, symmetric stopping rules for superiority/inferiority based on the guiding outcome; domains may use stopping rules for practical equivalence or futility based on the posterior distribution of the guiding outcome on the absolute scale. All stopping rules will be binding. Response-adaptive randomisation, either with or without restrictions, may be used based on the posterior distribution for the guiding outcome. Missing data will be multiply imputed. Additional secondary analyses (e.g., per-protocol analyses), sensitivity analyses, and analyses of heterogeneity in intervention effects according to pre-defined baseline characteristics may be specified for each domain and undertaken once a domain has stopped. Domains will be designed and evaluated using statistical simulation.
