Table 1 Comparison of the methodological issues to be considered in the evaluation of single and complex interventions in critical care

   Study question

To determine whether this monoclonal antibody compared to placebo reduces mortality for patients with severe sepsis

To determine whether this resuscitation protocol compared to usual care reduces mortality for patients with severe sepsis

To determine the best way to implement a new protocol for the resuscitation of patients with severe sepsis

   Pre-clinical phase

Linear approach from in vitro studies to animal studies to phase I and phase II clinical trials

Non-linear, iterative approach is needed to examine the effectiveness of each aspect of the protocol, how these aspects interact with current practice and what methods of implementing the protocol as a whole are likely to be most successful

   Pilot studies

Focussed on feasibility of recruitment, compliance with treatment and follow-up

Will help determine feasibility of implementing the protocol as a whole, which components are most commonly implemented or missed

Needed to identify barriers to implementing the protocol, potential means to overcome these barriers, optimal strategies for implementing the protocol

Trial design

   Population

Will be patients with the target condition, for example, two SIRS criteria and evidence of organ dysfunction in patients with suspected or proven infection

May be patients with the target condition, or it may be health service delivery organisations. For example, attempts to determine whether the protocol works may be focussed on patients with severe sepsis or attempts to determine how best to implement the protocol may be focussed on physicians or even hospitals

   Intervention

Clearly defined single drug therapy

Will contain multiple interventions, for example, increased fluids, blood transfusions, vasopressors, additional monitoring devices (arterial lines, lactate measurements, ScvO₂ measurements), as well as specific guidance to clinicians regarding the timing of these interventions

   Comparison group

Placebo

The control group could receive 'usual care' as determined by individual clinicians, a defined protocol of 'usual care', a protocol with different components, or an alternative suite of interventions (for example, computerised reminders) to enhance compliance with the protocol under investigation

   Outcome

Primary outcome: all cause mortality at 90 days

Primary outcome may be mortality or compliance with the protocol may be the primary outcome of interest. As blinding may be less than optimal, well defined and robust outcomes are required

   Context

May relate to the other treatments delivered in conjunction with the monoclonal antibody treatment. Generally reported in a table of co-interventions

Crucial element of trial design. Factors to consider include the existing protocols in place, staffing levels (both numbers and experience), availability of ScvO₂ monitors, resources of the emergency department and current treatment patterns

Trial execution

   Randomization

Individual participants will be randomized

Randomization may be at the individual participant level, particularly for trials designed to determine whether the protocol is effective

Randomization may also need to be at the level of the health care provider or service delivery organization when the aim of the study is to determine how best to implement the protocol

   Blinding

Blinding should be possible

Blinding of the intervention is likely to be difficult or impossible, and may not be desirable if the intention is to determine the best way to implement the protocol. Attempts to blind outcome adjudication, data analysts may be possible and will enhance internal validity

   Analysis

Simple statistical analysis is usually possible

Complex analysis is required for multi-arm trials and cluster-randomized trials

Compliance with the protocol is likely to be of greater interest, and a per-protocol analysis may offer information regarding aspects of the protocol that did or did not add value

Trial reporting

   Reporting

Should follow CONSORT statement

Should follow CONSORT statement or the extension relating to cluster-randomized trials when appropriate