Research Design, Biostatistics, and Literature Evaluation

Preliminary research should exist to suggest that the intervention is based on pathophysiology and

repeated outcomes consistent with the existing scientific understanding sufficient to warrant the proposed

testing on patients.

While historically considered the gold standard for research, many limitations exist to the application

and interpretation of randomized controlled trials in the ICU (Front Immunol. 2018;9:1502).

Heterogeneous population: The assumption that all patients would benefit from a treatment is

unlikely.

Gaps in understanding: Prognostic markers and definitions are often limited, raising concern for

inappropriate inclusion and/or exclusion.

Therapeutic optimization often unknown: Preclinical data is often highly different from the

population studied (healthy humans or rodent models) limiting confidence in extrapolation.

opinions on what would be considered the standard of care may differ at local, national, and

international levels in the critically ill. An ethical approach to the comparator arm must therefore be

carefully considered (Circulation 2007;115:1164-9).

Trial enrichment is a mechanism in which a specific subgroup of a population who have higher likelihood

of a response to an intervention are selected over those who may not (Nat Rev Nephrol 2020;16:20-31).

Prognostic enrichment selects for patients with a higher likelihood of disease-related interest (i.e.,

mortality)

Predictive enrichment selects for those patients who are more likely to respond to a treatment based

on underlying biologic mechanisms or physiology (i.e., inflammatory markers in ARDS)

Observation of clinical practice; no intervention is tested

Case-control study

Includes patients who did and did not experience an outcome of interest

Hypothesis generating and retrospective design

Provides cost-effective means to determine the association between the risk factor and the outcome

of interest

experience the outcome to identify the differences and risk factors (i.e., exposures) for developing

the outcome of interest.

Potential for selection bias and confounding because study begins with the “effect” (subjects selected

based on outcome of interest) in order to identify a potential causal association (of note, it is not

possible to definitively determine causation from a single observational study)

Cases and controls are representative of the population with the disease and should be chosen in a

way to minimize selection bias. Controls are randomly selected out of the same population from

which the cases arose.

Cohort study

Includes patients who did and did not experience an exposure of interest. May be prospective or

retrospective in design

Observational study of a given population over a given time to determine the relationship between

exposure and outcome of interest

Describes the natural progression of a disease or syndrome

Case reports/case series

Describes the experience in treating a single patient or the cumulative experience in treating a series

of patients