TRAITS

Bayesian methods allow an intuitive approach to guide decision-making in the ongoing TRAITS platform on whether a treatment shows promise in a given trait.  They allow us to make direct calculations of the probability that a treatment will have a clinically important benefit on the outcome of critically ill patients who have that trait.

Repeated interim checks are implemented in two stages of analysis during the trial: 

• In Stage 1, an intermediate outcome measure is selected for each trait that would be expected to show a positive signal if the treatment was truly effective.  Treatments showing promise on the intermediate outcome in Stage 1 analyses (based on a Bayesian calculation of the probability of benefit) then progress to Stage 2. Treatments not showing promise are discontinued from the trial, allowing alternative candidate treatments to be introduced to the platform. 
• In Stage 2, the TRAITS primary outcome – Organ Support Free Days during the first 21 days following randomisation – is analysed to provide a final decision on whether a trait-specific treatment works.   

The Bayesian framework allows multiple Stage 1 and Stage 2 analyses to be undertaken for each treatment until a final conclusion is reached. 

TRAITS data collection

In clinical trials information about trial participants is manually entered into a trial database by research nurses and members of the clinical team at local hospital sites. Once collected, the data are analysed by the research team to produce their results.  The research team may request access to national datasets such as hospital inpatient admissions and combine this with the trial data for further analysis.

For the TRAITS trial, we are automating the flow of information about the participant to the trial database. We are using national datasets, clinical audit systems and live clinical systems to provide data such as co-existing medical conditions, and how sick participants are when they are admitted to ICU. We are also collecting daily information about the type of support participants are needing for their breathing and blood pressure. This data, which would be entered by research staff in the Intensive Care Unit, is pulled automatically (where available) and integrated into the trial database where it is made available to the research team for analysis. 

Pulling the clinical data directly helps avoid any accidental typing errors and streamlines the data collection process. As more clinical data becomes available by this method, the need to enter data at the hospital sites is minimised. This means that more sites can participate in research studies, and research staff can focus their time working with patients and their relatives to help them understand the trial.

Automating data collection is a key step towards precision medicine which aims to accurately predict treatment and prevention strategies for patients.