A pragmatic clinical trial is a clinical trial designed primarily to determine how an intervention performs under conditions resembling routine practice. It asks whether a treatment, diagnostic strategy, or care-delivery approach produces useful outcomes in the settings where it is intended to be used. Pragmatic trials commonly use random allocation, but pragmatic describes the purpose and design of a trial rather than its allocation method. Their defining feature is alignment with a practical decision, not simply a large sample, broad eligibility, or use of routinely collected data.
Pragmatic and explanatory purposes
The distinction between pragmatic and explanatory trials was articulated by Daniel Schwartz and Joseph Lellouch in 1967. An explanatory trial primarily asks whether an intervention can produce an effect under selected, controlled conditions. A pragmatic trial primarily asks what happens when that intervention is offered or delivered in ordinary care.
This distinction is often expressed as efficacy versus effectiveness:
- Efficacy concerns an intervention’s effects under specified, often tightly controlled conditions.
- Effectiveness concerns its effects in the circumstances in which it is actually used.
These are complementary questions, not competing standards of scientific quality. A tightly controlled trial can establish whether an intervention has a particular effect, while a pragmatic trial can establish whether deploying it produces meaningful benefits in practice. Both can use a randomized controlled trial design.
Pragmatism is a continuum rather than a binary classification. A trial may recruit a broad patient population but require unusually frequent follow-up, or use routine clinicians while tightly standardizing treatment delivery. Its pragmatic character therefore needs to be assessed across several design features.
Main design characteristics
Pragmatic trials are organized around a clearly specified decision: which available treatment to offer, whether to introduce a service, or how to organize care within a health system.
Participants and recruitment
Eligibility criteria generally aim to include people who would receive the intervention in practice. This can include older adults, people with multiple long-term conditions, and patients taking other medications, when their inclusion is appropriate to the research question and safety requirements.
Recruitment through ordinary clinical encounters can reduce the differences between trial participants and the population intended to benefit. Nevertheless, broad eligibility alone does not ensure representativeness. Consent requirements, language barriers, clinician participation, and unequal access to care can still shape who enters a trial.
Settings and delivery
Interventions are often delivered by ordinary clinical staff in hospitals, community services, or primary care, using resources reasonably available in those settings. Protocols may allow clinicians to adjust doses, select among permitted treatments, or manage additional conditions as they normally would.
Flexibility does not mean that the intervention is undefined. The protocol must distinguish its essential components from elements that may vary locally. Otherwise, it may become unclear what strategy the trial actually tested.
Comparators
A pragmatic comparator is commonly an active treatment or usual care rather than a placebo. The choice depends on the decision being studied; placebo controls are not categorically incompatible with pragmatic aims.
“Usual care” requires careful description because it can differ between sites and change over time. A comparison against poorly characterized usual care may be difficult to interpret or apply elsewhere.
Outcomes and follow-up
Pragmatic trials often prioritize clinical endpoints that directly matter to patients or decision-makers, such as hospitalization, symptoms, functional status, or survival. Patient-reported outcomes and health-related quality of life can capture effects not adequately represented by laboratory measurements.
Follow-up may rely partly on electronic health records, administrative claims, or registries. These sources can reduce additional visits and support large-scale follow-up, but information collected for care or payment is not automatically complete or suitable for research.
Assessing pragmatism: PRECIS-2
The Pragmatic–Explanatory Continuum Indicator Summary 2, usually called PRECIS-2, was published in 2015 as a tool for aligning trial design with its intended purpose. It assesses nine domains:
| Domain | Central question |
|---|---|
| Eligibility | How similar are participants to those who would receive the intervention in usual care? |
| Recruitment | How much additional effort is used to recruit participants beyond ordinary care processes? |
| Setting | How similar are trial settings to intended practice settings? |
| Organization | How similar are the resources and expertise required to those available in usual care? |
| Flexibility of delivery | How much flexibility is allowed in delivering the intervention? |
| Flexibility of adherence | How much additional effort is used to encourage adherence? |
| Follow-up | How much more intensive is follow-up than routine care? |
| Primary outcome | How directly relevant is the primary outcome to participants? |
| Primary analysis | To what extent does the analysis include participants regardless of adherence? |
Each domain is rated from 1, very explanatory, to 5, very pragmatic. Ratings are interpreted relative to the intended setting and research question. PRECIS-2 is a design and discussion tool, not a general risk-of-bias instrument; a high score does not establish that a trial is scientifically sound.
Randomization and analysis
Randomization remains valuable because it supports causal comparisons by balancing measured and unmeasured prognostic factors between groups, subject to chance variation. Pragmatic trials may randomize individuals or groups.
In a cluster randomized trial, clinics, wards, practices, or communities receive an assigned strategy. This can suit interventions directed at organizational practice or reduce contamination between treatment groups. However, outcomes within a cluster are correlated, requiring appropriate sample-size calculations and analysis. Recruitment after cluster allocation can also introduce selection bias if knowledge of the assigned strategy influences enrollment.
Pragmatic trials commonly use intention-to-treat analysis, comparing participants according to their assigned group. This can estimate the effect of assigning or offering a strategy under the trial’s delivery conditions. It does not necessarily estimate the effect of consistently receiving treatment.
Nonadherence and treatment crossover can reduce the difference between groups. Analyses restricted to adherent participants may address additional questions, but simple exclusions can undermine the protection provided by randomization.
Allocation concealment remains important even when blinding after assignment is impractical. Blinded outcome assessment or standardized outcome definitions may still reduce measurement bias.
Applications and related approaches
Pragmatic trials support evidence-based medicine by studying choices encountered in routine practice. Applications include comparisons of established medications, diagnostic pathways, rehabilitation services, and methods of organizing care.
An embedded pragmatic trial integrates research procedures into existing care workflows and information systems. Embedding can make studies more feasible, but dependence on local infrastructure may limit applicability elsewhere.
Pragmatic trials overlap with comparative-effectiveness research, implementation research, and studies using real-world data, but these concepts are not interchangeable:
- Comparative-effectiveness research compares available strategies and can use randomized or observational methods.
- Implementation research examines how interventions are adopted, delivered, and sustained.
- Real-world data describes data sources and their context of collection, not whether treatment assignment was randomized.
An observational study does not become a pragmatic trial merely because it uses routine records.
Limitations and interpretation
Pragmatic design can improve relevance to practice, but it does not guarantee generalizability. Results remain dependent on the participating population, baseline risks, clinical resources, comparator, and patterns of implementation.
Important limitations include incomplete routine data, inconsistent coding, unblinded outcome assessment, contamination between groups, and changes in care during follow-up. Some outcomes—especially symptoms, function, and treatment burden—may be poorly captured unless collected specifically for the study.
Resource requirements also affect interpretation. An intervention supported by extra trial staff or unusually advanced information systems may not perform similarly elsewhere. Conversely, a small observed effect may reflect limited uptake rather than an absence of benefit when treatment is received.
Claims of noninferiority require particular care: poor adherence and treatment crossover can make groups appear similar without demonstrating that the interventions have comparable effects.
Reporting guidance, including the pragmatic-trial extension to the CONSORT statement, emphasizes enough information about participants, settings, interventions, and outcomes to judge the applicability of results.
Ethics and oversight
Pragmatic trials remain human-subjects research. Their integration into ordinary care does not remove requirements for ethical review, proportionate risk assessment, and appropriate protection of participants.
An institutional review board or equivalent body evaluates consent arrangements under applicable rules. Modified consent procedures or waivers may be permissible in some circumstances, but routine-care interventions and cluster allocation do not automatically justify them. Permission from a hospital or practice leader is not, by itself, equivalent to individual informed consent.
Use of linked clinical records also raises data privacy and confidentiality concerns. Safety monitoring, transparent communication, and clear responsibilities remain necessary even when a trial adds few procedures beyond usual care.