A pragmatic trial is usually designed to inform a decision under conditions resembling the settings in which that decision will be made. A more explanatory trial usually emphasizes whether an intervention can produce an effect under selected, controlled conditions. These are purposes and design tendencies, not opposing grades of scientific quality.[1][2]
The distinction matters because applicability is built through choices about who enters, where care occurs, how interventions and comparators are delivered, which outcomes are measured, and how the data are analyzed. Calling a trial “pragmatic,” “real world,” or “multicenter” does not settle whether its result fits a particular population, setting, product, or decision.
Pragmatic and explanatory are not rigid boxes
Trials can be more pragmatic on one dimension and more explanatory on another. A study might use broad eligibility but tightly standardized delivery. Another might enroll a selected population while allowing routine-care flexibility after allocation. The label should not replace a description of those choices.
The 2008 CONSORT pragmatic-trial extension defined a pragmatic randomized trial broadly by its purpose of informing decisions about practice. It expanded reporting recommendations for background, participants, interventions, outcomes, sample size, blinding, participant flow, and generalizability.[1][2] Those items help readers see where a trial sits along multiple dimensions; they do not create a score that proves usefulness.
| Design dimension | More explanatory tendency | More pragmatic tendency | Question for readers |
|---|---|---|---|
| Eligibility | Narrow, selected participants | Broader decision-relevant population | Who was excluded, and would that alter the effect or feasibility? |
| Setting | Specialized or highly resourced sites | Diverse usual-care settings | Does the destination setting have comparable people, skills, and infrastructure? |
| Delivery | Closely standardized and monitored | Flexibility closer to routine practice | What actually varied, and could that variation affect outcomes? |
| Comparator | Placebo or optimized experimental control | Existing care or another real decision option | Is the comparator relevant to the decision now being considered? |
| Follow-up | Intensive contact and retention procedures | Follow-up closer to ordinary systems | Did missingness or weak ascertainment threaten the comparison? |
| Analysis | May emphasize adherence or mechanism | Often emphasizes effects of assignment | Which treatment effect was the trial designed to estimate? |
No row is automatically better. The right choices depend on the question.
Internal validity and applicability answer different questions
Internal validity concerns whether the comparison supports a credible conclusion for the people and conditions studied. Applicability concerns whether that conclusion is informative somewhere else. Broad eligibility and routine-care delivery can improve relevance, but they do not excuse predictable allocation, differential outcome ascertainment, contamination, selective reporting, or avoidable missing data.
Conversely, a carefully controlled trial may have strong internal validity while offering limited evidence about settings that differ materially. A useful interpretation therefore asks two questions in order:
- Is the estimated contrast credible for the trial as conducted?
- If so, how closely does the intended use context match the trial?
Skipping the first question turns “real world” into a shield against bias. Skipping the second turns an internally credible estimate into an unsupported universal claim.
Applicability is claim-specific
A result does not transfer as a single package. Each proposed inference has its own fit requirements.
- Population: age, disease definition, severity, comorbidity, prior exposure, and barriers to participation can modify outcomes or feasibility.
- Setting: specialist centers, community practices, countries, staffing levels, diagnostic access, and follow-up systems can differ.
- Intervention: formulation, identity, delivery process, monitoring, co-interventions, and implementation support must match the studied intervention closely enough for the claim.
- Comparator: usual care changes across place and time. A historical comparator may no longer represent the current decision.
- Outcome: a surrogate, short follow-up, or trial-specific scale may not answer the outcome question important to another audience.
- Care pathway: referral, eligibility screening, adherence support, rescue care, and access after randomization can shape the observed result.
The same trial might be applicable to one policy question but not to a product-specific, long-term, or different-population claim. Applicability should name the destination and the exact inference rather than declare that a trial is “generalizable.”
Setting and recruitment reveal who the trial represents
A site list alone is not enough. Readers need to know the level of care, geographic distribution, investigator experience, recruitment route, and proportion of potentially eligible people who actually enrolled. A trial recruiting volunteers through advertising may represent a different decision population from one enrolling consecutive patients during routine visits.
Eligibility criteria show who could enter; recruitment data show who did enter. Both matter. A broad criterion can still yield a narrow enrolled sample because of consent, travel, digital access, language, clinician selection, or competing demands. Participant flow and baseline data help expose that narrowing.
CONSORT 2025 asks for baseline demographic and clinical characteristics by randomized group and cautions against significance testing of baseline differences; chance differences can occur after proper randomization.[5] Baseline tables support description of the enrolled population, but they do not by themselves show that nonparticipants or other settings would respond similarly.
Delivery flexibility must be described, not assumed
“Usual care” can hide substantial variation. A pragmatic report should identify which components were fixed, which could vary, who made those decisions, what training or reminders were supplied, and what delivery actually occurred. Flexibility may make a comparison relevant to practice, but undocumented flexibility makes the intervention difficult to interpret.
The same principle applies to adherence. An explanatory trial may actively promote protocol adherence to clarify a biological contrast. A pragmatic trial may avoid supports that would not exist in the destination system. Either approach can be coherent if the objective, estimand, support, deviations, and analysis are aligned. Neither permits silently excluding inconvenient participants after randomization.
Comparator delivery also deserves equal detail. If one group receives more contact, monitoring, or rescue options, the estimated effect may reflect the whole care strategy rather than a single named component.
Outcomes and follow-up should match the intended decision
A pragmatic purpose does not guarantee a decision-relevant outcome. Readers should ask whether the primary outcome is meaningful to the target audience, measured consistently across settings, and observed for long enough to inform the proposed conclusion.
Routine data can reduce research burden and widen participation, but their original purpose may differ from the trial question. Outcome definitions, coding changes, incomplete capture outside a network, and differences in surveillance can affect ascertainment. A large record system does not automatically produce a valid or complete outcome.
Follow-up procedures can also change behavior and retention. Intensive visits may improve measurement while moving the experience away from usual care. Minimal contact may preserve routine conditions while increasing missingness. The report should make that tradeoff visible rather than treating either approach as inherently pragmatic.
“Real world,” “multicenter,” and “pragmatic” are not synonyms
“Real-world data” usually describes a data source or context, not a complete trial design. “Multicenter” means that more than one site participated; the sites may still be highly selected and use identical specialized procedures. “Pragmatic” describes an intention supported by multiple design choices.
None of these words proves representative enrollment, unbiased comparison, complete follow-up, or transfer to a new product. A multicenter explanatory trial can be highly controlled. A single-region pragmatic trial can be directly useful for a local decision yet poorly transferable elsewhere. A randomized trial using routine data can have a pragmatic purpose while retaining strict eligibility or centralized outcome adjudication.
Reporting guidance has a specific, current status
The 2008 pragmatic extension remains listed on the SPIRIT–CONSORT extensions overview under extensions to CONSORT 2010.[3] It is a reporting extension: it helps authors disclose design choices so readers can judge applicability. It is not a regulatory requirement, a risk-of-bias certificate, or proof that a trial achieved its intended degree of pragmatism.
A paper published online in June 2026 is a protocol for developing an updated CONSORT-PRECIS extension covering trials with declared pragmatic or explanatory intentions.[4] It describes planned literature review, consensus work, guidance development, and pilot testing. It is not the completed replacement guideline. Until an updated statement is completed and published, the protocol should not be cited as if its future recommendations already govern reporting.
When a pragmatic trial randomizes clinics
A care strategy may be assigned to entire clinics or hospitals when individual assignment would allow the intervention to spill into the comparison group. That changes the unit of randomization: participants within a clinic can have correlated outcomes. Readers should check the number of clusters, their sizes, the intracluster correlation used in planning, and how the analysis accounted for clustering. A large patient count alone does not establish precision. These reporting questions are described in the CONSORT extension for cluster randomized trials.
Also establish when individuals entered the study relative to cluster allocation. If recruiters already know a clinic’s assignment, selection into the study can differ between groups. The report should explain whether all eligible individuals were included, whether sampling was used, and whether recruitment was masked to allocation. Examine participant flow at both the clinic and individual levels. Cluster randomization can fit routine workflows, but it does not remove the need for comparable recruitment and a design-appropriate analysis.
An applicability reading checklist
- State the decision the trial was intended to inform.
- Identify pragmatic and explanatory tendencies across separate design dimensions.
- Assess randomization, concealment, outcome ascertainment, missing data, and selective reporting first.
- Compare eligibility with the destination population.
- Compare recruitment, sites, resources, and care pathways with the destination setting.
- Describe intervention and comparator flexibility, support, and actual delivery.
- Check whether the outcome and follow-up match the proposed decision.
- Name material differences in formulation, identity, monitoring, or co-interventions.
- Preserve uncertainty when important destination groups were uncommon or absent.
- Treat applicability as a bounded judgment for a named claim.
Important limitations
- Pragmatic and explanatory are not synonyms for high and low quality.
- A trial can be pragmatic on some dimensions and explanatory on others.
- Broad eligibility does not repair bias in allocation, measurement, follow-up, or reporting.
- A usual-care comparator can vary across sites and become outdated.
- Diverse sites do not guarantee representative participants.
- Routine-data outcomes still require valid definitions and adequate capture.
- Reporting-guideline adherence does not prove applicability.
- A result for a named intervention cannot authenticate or establish comparability with a separately sourced material.
Frequently asked questions
Is a pragmatic trial always conducted in routine care?
No. A pragmatic purpose is expressed across several design dimensions. Some procedures may resemble routine care while others remain centralized or tightly controlled. The report should describe each dimension rather than rely on one label.
Is an explanatory trial less useful than a pragmatic trial?
No. An explanatory design may be well suited to a mechanistic or proof-of-concept question. Usefulness depends on whether the design matches the question and whether conclusions stay within the studied conditions.
Does broad eligibility make a result generalizable?
Not by itself. Recruitment, consent, site selection, actual baseline characteristics, delivery, outcome capture, and follow-up can still produce a study sample or experience unlike the destination context.
Does “multicenter” mean the result applies everywhere?
No. Sites can share specialized resources, selection practices, or care pathways. Readers need the distribution and characteristics of sites and participants, not only the number of centers.
Can a pragmatic trial have weak internal validity?
Yes. Pragmatism does not excuse flaws in randomization, allocation concealment, outcome measurement, missing-data handling, or selective reporting. Applicability is worth judging only after the comparison is credible.
Is the 2026 CONSORT-PRECIS paper the new completed guideline?
No. It is a development protocol published online in June 2026. It describes how an update will be produced; it is not a completed replacement for the listed 2008 pragmatic extension.[3][4]
Can a pragmatic trial validate a different commercial material?
No. Applicability requires a defensible match to the studied intervention, including identity and other claim-relevant attributes. A design label cannot establish that an untested material is the same.