The Pragmatic Trial Has Been Hijacked. Here Is What It Was Actually Supposed to Do.

Pull up the protocol for almost any Phase IV study filed in the last five years and you will find the word “pragmatic” somewhere in the design rationale. It appears in the eligibility criteria section to justify broad enrollment. It appears in the comparator justification to explain why usual care is acceptable. It appears in the statistical plan to defend a wider-than-standard confidence interval. The word has become a rhetorical Swiss Army knife, deployed wherever a sponsor needs methodological flexibility and a reviewer might push back. And that is precisely the problem that a recent JAMA piece on pragmatic clinical trials exposes with unusual directness: the label “pragmatic” has been stretched so far from its original meaning that it now obscures more than it reveals.

The stakes are not semantic. When a regulatory reviewer reads “pragmatic design,” they need to understand what kind of evidence they are actually evaluating. A trial built to detect a biological mechanism and a trial built to inform a treatment choice in community practice require completely different evidentiary frameworks. Conflating them does not produce a hybrid that satisfies both purposes. It produces evidence that satisfies neither, and the consequences land on patients, payers, and the sponsors who believed their design would survive scrutiny.

What Schwartz and Lellouch Actually Said

In 1967, D. Schwartz and J. Lellouch introduced the terms “explanatory” and “pragmatic” to describe two fundamentally different attitudes toward clinical research, not two points on a design checklist. An explanatory trial tests a causal hypothesis: does this intervention produce a measurable biological effect under conditions controlled tightly enough to isolate that effect? A pragmatic trial asks a different question entirely: given that this intervention exists, does choosing it over the available alternatives produce better outcomes for patients in the settings where those patients actually receive care?

Schwartz and Lellouch were not proposing a looser version of the RCT. They were proposing a different instrument for a different instrument problem. A cardiologist managing heart failure in a community hospital does not need to know whether a drug suppresses a specific neurohormonal pathway under laboratory conditions. That question was answered by the explanatory trial. The cardiologist needs to know whether prescribing this drug, to patients who look like the ones sitting across from her today, in a clinic that operates the way her clinic operates, produces better outcomes than what she is currently doing. Schwartz and Lellouch saw that the clinical trial enterprise of their era was generating enormous quantities of explanatory evidence and very little of the second kind.

That insight was correct in 1967. It remains correct now. The problem is that somewhere in the intervening six decades, the word “pragmatic” migrated from describing the question a trial was designed to answer to describing a set of methodological choices sponsors make when they want broader populations, simpler operations, or cheaper execution. Those are not the same thing.

The ONE PRINCIPLE at the center of this confusion is what might be called the Question-Before-Design sequence. Every trial must begin with a clear statement of its decision problem: who needs to choose, between which options, in which population, under which conditions? That decision problem determines whether the trial is explanatory or pragmatic. The design features follow from the question. When sponsors reverse this sequence, selecting pragmatic design features first and retrofitting a decision question afterward, they produce trials that look flexible but answer nothing clearly.

The PRECIS-2 framework, published in the BMJ in 2015, represents the most disciplined attempt to operationalize this distinction. It scores a trial across nine domains, including eligibility criteria, recruitment, setting, flexibility of intervention delivery, and the nature of the primary outcome, to characterize where the trial sits on the explanatory-to-pragmatic continuum. A trial cannot simply declare itself pragmatic. Under PRECIS-2, it must demonstrate that its design choices across all nine domains are coherent with the decision problem it claims to address. An eligibility criterion that excludes patients with common comorbidities, for instance, pulls a trial toward the explanatory end regardless of what the protocol cover page says. PRECIS-2 makes that tension visible and quantifiable.

What sponsors and sponsors’ regulatory counsel have increasingly done with PRECIS-2 is select the favorable domains, score them high on pragmatic alignment, and treat the framework as a partial defense rather than a comprehensive diagnostic. A trial can score pragmatic on setting and comparator while scoring explanatory on eligibility and follow-up intensity, producing a PRECIS-2 radar chart that looks roughly centered while the underlying design remains incoherent. The framework was designed to prevent exactly this kind of selective application, but the incentive structure pushes against it.

Where the FDA Ran the Experiment

The regulatory consequences of this definitional drift are not theoretical. In 2022, the FDA Oncology Center of Excellence launched Project Pragmatica specifically to stress-test pragmatic design elements in trials for approved oncology products. The initiative’s explicit purpose was to determine when pragmatic features, including real-world data integration and broader eligibility criteria, could support regulatory decision-making without sacrificing the inferential validity that approval decisions require. The very fact that the FDA needed a dedicated initiative to work this out confirms the central observation in the methodological literature: the field had been using “pragmatic” without a shared operational definition, and the regulatory machinery was straining under the ambiguity.

Project Pragmatica also surfaced a friction point that a gap the methodological literature identifies but does not fully name. Pragmatic trials generate effectiveness evidence: what happens to patients in routine care. The FDA’s evidentiary standard for most approval decisions requires efficacy evidence: what happens under conditions controlled enough to establish causality. These are not the same evidentiary bar. A pragmatic trial can be methodologically rigorous and still produce evidence that does not meet the standard for a new indication, a label expansion, or a comparative effectiveness claim that payers will accept for formulary positioning. Sponsors who designed pragmatic trials expecting regulatory flexibility were frequently surprised to find that the FDA’s flexibility had limits defined by the decision the agency needed to make, not by the design the sponsor had chosen.

The CASRAI definition of pragmatic trials captures this boundary clearly: a pragmatic RCT is designed to assess whether an intervention works under real-world conditions encountered by clinicians and healthcare systems. That definition contains a quiet but important word: “works.” Not “works through a specified mechanism.” Not “works better than placebo under controlled conditions.” Works, as in produces the outcome that matters, in the setting where care is delivered, for the patients who actually receive it. When a trial designed to answer that question gets submitted in support of a regulatory claim that requires the explanatory version of “works,” the mismatch surfaces in the review cycle, and the cost lands on the sponsor’s timeline.

The Design Choice That Cannot Be Retrofitted

The deeper systemic problem the JAMA editorial reveals is that the pragmatic-versus-explanatory choice cannot be made at the analysis stage. It must be made at the protocol stage, before a single patient is enrolled, because it determines the comparator, the eligibility window, the follow-up duration, the primary endpoint definition, and the statistical model. Retrofitting a pragmatic label onto an explanatory design after enrollment closes does not transform the evidence. It produces a document that describes the trial as pragmatic while the data structure prevents the trial from answering a pragmatic question.

Consider what this means for a sponsor running an oncology trial under Project Pragmatica’s framework. If the protocol enrolled patients from academic cancer centers exclusively, applied stringent performance status criteria, and mandated protocol-specified dose modifications, the resulting dataset reflects care delivery patterns at a small fraction of the settings where the drug will actually be used. A PRECIS-2 score can flag that misalignment, but only if the sponsor applies the tool honestly across all nine domains before locking the protocol. By the time the data is in, the choices are irreversible.

The methodological literature has documented this failure mode since at least the PMC analysis that formalized the effectiveness-efficacy distinction in trial design, which observed that pragmatic trials must be coherent in their purpose from inception: the entire design architecture, from who is eligible to what counts as an outcome, must reflect the decision context the trial is meant to inform. A single explanatory design choice embedded in an otherwise pragmatic protocol does not merely shift the trial slightly toward the explanatory end of the PRECIS-2 spectrum. In some cases, it invalidates the pragmatic claim entirely, because the population or the comparator no longer represents the real-world decision the trial was supposed to illuminate.

That is the finding the methodological literature on pragmatic trials crystallizes, and it has a direct operational translation for every sponsor currently designing a post-approval study, a label expansion trial, or a real-world evidence study intended to support a payer submission. Write down the decision problem first. Name the decision-maker, the population, the alternatives they are actually choosing between, and the setting in which they make the choice. Then evaluate every design element against that decision problem using PRECIS-2 as a diagnostic, not a marketing tool. If any design element drifts toward the explanatory end without a specific justification tied to the decision problem, remove it or document why the tradeoff is acceptable. The protocol you file after that process will be coherent. The one you file without it will be called pragmatic in the cover letter and questioned in the review.

The first sponsor to submit a genuinely pragmatic oncology label-expansion trial under Project Pragmatica’s framework, with a PRECIS-2 profile that is coherent across all nine domains and a decision problem written clearly enough that a formulary committee and an FDA reviewer reach the same conclusion about what the evidence means, will set an evidentiary standard the rest of the field will spend years trying to replicate.

References

1. JAMA, “The Pragmatic Clinical Trial”
2. Queen Mary University of London, PCTU, “What Is a Pragmatic Clinical Trial?” (Schwartz and Lellouch, 1967)
3. Rethinking Clinical Trials, “Pragmatic Elements: An Introduction to PRECIS-2” (BMJ, 2015)
4. FDA Oncology Center of Excellence, “Project Pragmatica” (2022)
5. CASRAI, “Pragmatic Trial | Real-World Trial Design” (definition)
6. PMC, “Pragmatic Trials: Effectiveness vs. Efficacy in Trial Design”