Study design: Cohort multiple randomised controlled trial

What is it?

Cohort multiple randomised controlled trials (cmRCT) use a large cohort of patients as a source of participants for multiple randomised controlled trials (RCTs).

A large cohort of patients with the condition of interest is recruited and followed-up over a period of time (as with cohort studies). Patients from this cohort are then selected as participants for an RCT to evaluate an intervention. Outcomes for the whole cohort are measured regularly, providing longitudinal information about the progression of the condition both for patients having usual clinical practice and those having the trial intervention. Multiple RCTs can be carried out, allowing rapid and efficient simultaneous evaluations of different interventions for the same condition within the same cohort.

Consent for the use of data is gained from all cohort members at the offset, whereas consent for participation in a trial is sought only from patients offered the intervention (post-randomisation), which means that blinding is not possible. Electronic medical records can be used to identify large cohorts, with recruitment done through the GP at a consultation (van Staa et al 2014).

The cmRCT study design is a type of pragmatic trial. It is also known as a ‘trial within cohort’ study design (TwiCs).

Figure. Cohort multiple randomised controlled trial


Cluster cmRCTs are a variation on this design using groups (clusters) of patients rather than individuals randomised to different treatments (similar to cluster RCTs). For example, a cluster might be within a GP practice, hospital or community. This can increase the speed of recruitment to the trial and help reduce costs because interventions are administered in fewer places.

Why is it useful?

Related to population:

  • Simple to recruit and carry out: having a large pool of eligible patients means that it is easier to recruit and carry out trials compared with traditional, cluster, and other pragmatic clinical trials.
  • Highly representative sample: use of the cohort of patients may give a more representative sample than those recruited to other types of studies (improving generalisability) but this will depend on the level of any additional eligibility criteria applied for individuals to be included in the trial. For example, RCTs often exclude people from minority ethnic groups or with learning disabilities so the ‘real-world’ population is not represented. The design also enables data to be available on people who refuse treatment.

Related to comparators:

  • Pragmatic comparator: the comparator arm acts exactly as it would in real-life, meaning interventions are compared against current practice as it would be administered outside of trial conditions.
  • Good comparability between trials: as the same cohort is used with the same ‘standard care’ comparator and the same core outcomes measured for multiple RCTs. This may enable more direct and indirect treatment comparisons to be undertaken than with traditional RCT designs.

When is it suitable?

  • Trials for expensive medicines, because simple, cost-efficient recruitment can help offset the costs of the medicines.
  • Trials for treatments viewed as highly desirable by patients.
  • Studies of chronic conditions for which many clinical trials are likely to be conducted in the future (such as obesity, diabetes and chronic pain).
  • Studies with outcomes that can be easily measured in the whole cohort.

What are the limitations?

  • Risk of patients refusing the treatment is high because of consent being obtained for the treatment group; this may result in differences between participants in each group (selection bias) and insufficient overall numbers of patients in the study to be able to adequately demonstrate true treatment differences (although a large initial cohort will offset this drop in statistical power). This is the topic of a case study by GetReal.
  • More difficult to study outcomes that are hard to measure and hard to collect.
  • Issues with attrition may prevent the use of this design in populations that are likely to have high attrition rates (however, this will depend on the selection of the cohort).
  • Not suitable for people with acute conditions or unstable populations where it is difficult to maintain a consistent cohort of patients.
  • Not yet widely used so it may be regarded with scepticism by until more studies are available.

What do stakeholders say?

Performing cmRCT does not have a direct impact on patients’ acceptability or perspectives. Acceptability therefore depends on whether the design option is appropriate to fulfil the research objectives and whether it is considered acceptable by regulators and health technology assessment (HTA) agencies. However, views of regulators and HTA agencies on the use of this study design has not been sought as part of GetReal.

Key contributor

Alexander Pate, University of Manchester