Table 1 Summary of trial designs

Randomized controlled trial

To determine proof of causality

Heterogeneous patient populations reduce ability of the trial to ascertain differences in treatment effect.

Logistic burden and cost

Participating centers should have a phase-in period that allows demonstration of the capacity of the center to perform the study safely and effectively.

Use of a smaller, well-defined, and homogeneous population may increase the probability of determining a true treatment effect

Observational study

Hypothesis generation

Do not provide proof of causality

Collaborations among centers with large databases.

Standardization of reporting metrics

Individual patient meta-analysis

To avoid the limitations and complications inherent in comparing disparate studies.

Allows interpretation of patient-specific events that would not normally be identified in aggregate trial data

Hugely time and resource intensive

The analysis will be confounded by the same limitations as present in the original trials

Cluster randomization

To reduce variations in process of care for complex interventions as a confounding factor in randomized controlled trials

Requires more complex statistical methodology to account for the effects of clustering

May be particularly suitable for complex interventions, such as glucose control in the critically ill