Development of new designs for intervention studies: Cluster randomised crossover trials with application to intensive care research

Principal Investigator: 
Prof. Andrew Forbes
Chief Investigators: 
Prof. John Carlin

Of patients admitted to Intensive Care Units (ICUs), approximately 12% die prior to hospital discharge. Although the development of better approaches to their care is an important public health priority, ascertainment of the impact of universal, everyday ICU interventions is difficult because the survival benefit of optimising any given treatment is generally small.

Cluster randomised trials are the only feasible design for universal interventions in the ICU setting, however the loss of precision due to within-ICU clustering of patient outcomes in these designs precludes their use for detecting small effects. This research theme focusses on more statistically efficient designs in this setting, and in particular on cluster-randomised crossover (CRXO) trials. In brief, CRXO trials involve ICUs randomised to deliver one intervention for a certain period of time and then ‘cross over’ to deliver the other intervention, with different groups of patients observed in each treatment period. The main benefit of this design is that the similarity of prognosis of patients within the same ICU unit but across treatment periods can be exploited to counterbalance the precision loss due to similarity of patient outcomes within the same treatment period. Statistical properties of these designs has been determined to a limited extent for continuous outcomes, but is almost non-existent for binary outcomes.

The proposed program of research will initially develop an array of possible analytical approaches for the two-period CRXO design for binary data by combining methods for crossover trials of individuals with those for cluster randomised trials. These will include methods based on cluster summaries, and various individual-patient models, both marginal and conditional. The CRXO design also offers enormous potential through the extension to multi-period designs and trials of two or more interventions, including factorial trials, and this potential will be explored.

Collaborators:

  • Professors Rinaldo Bellomo and Jamie Cooper, Australian and New Zealand Intensive Care Research Centre (ANZIC-RC), Monash University