Table 5

Potential recommendations to assist in the design of future clinical trials

Recommendation

Potential benefits

Potential issues

More extensive inclusion and exclusion criteria that are more descriptive of the population to be enrolled

Less opportunity for patient variability and sites having to 'learn as they go'

Lower likelihood of extrapolating efficacy observed in the clinical trial to effectiveness in clinical practice

Standardize the major facets of severe sepsis concomitant care

Reduced variability as caring for patients with severe sepsis may be a more complex 'procedure' than many commonly performed surgical procedures

May be questions related to the applicability of the study results to a more general severe sepsis population, in which concomitant care has not been standardized

Given the heterogeneity of severe sepsis patients, different populations of patients may require unique sets of inclusion and exclusion criteria (for example, medical patients and surgical patients).

Optimizes inclusion and exclusion criteria without the extra time and resources that would be needed to run two separate studies

May be issues with interpretation of data and powering if the treatment effect differs significantly between the two populations, in which two separate studies may be preferable

Use a clinical coordinating center to assist study sites in enrollment of eligible patients.

Helps to optimize protocol compliance

May be questions related to the applicability of the study results to a more general severe sepsis population

Site selection should be based on having good clinical trial and critical care experience.

Helps to minimize variability and (potentially) protocol violations

May be questions related to the applicability of the study results to a more general severe sepsis population

The use of severity scoring systems in clinical trials may require training and validation of the training to ensure proper collection of severity of illness information.

Helps to ensure the collection of accurate data

Additional time and resources required

Given the potential influence of site experience on outcomes, randomization stratified by site should be considered a requirement for studies in complex disease states.

Helps to minimize effect of differences between sites and enrollment sequence effect

May limit ability to stratify by additional parameters

Planned enrollment per site should be based on the block size used for randomization. Expected enrollment per site should be at least two full blocks of patients.

Helps to minimize enrollment sequence effect

May only be able to have larger sites in the study, raising questions related to generalizability of the results

Futility stopping rules should incorporate the potential for learning curves to obscure a beneficial treatment effect in the early stages of a clinical trial.

Helps to avoid type II error

May be a delay in identifying futility signals if no enrollment sequence effect is present

Statistical analysis plans should explore the potential for learning curves within the clinical trial dataset.

Prospectively defined analyses have greater weight and may help to explain study findings

Additional workload

Clinical studies should have a prospectively defined monitoring plan. Source data verification and documentation of protocol violations should be performed on the first few patients enrolled at a site until the site demonstrates adequate understanding of the protocol.

Helps to minimize protocol violations

Additional time and resources required

Laterre et al. Critical Care 2008 12:R117   doi:10.1186/cc7011