Interventional Studies: Definition, Regulatory Classification, and Design Types

An interventional study is one in which the investigator assigns participants to a specific intervention according to a protocol, in order to evaluate that intervention’s effect — as opposed to an observational study, where the investigator measures exposures and outcomes as they occur naturally without assigning who gets what. That single design decision, made before a single participant is enrolled, determines which regulatory framework applies, what registration and ethics-review obligations attach, and what kind of causal claim the resulting data can support. Because the consequences are so large and the definitions used by different regulators genuinely differ from one another, “is this study interventional?” is one of the first questions a research office, IRB, or sponsor must answer — and answering it wrong has real downstream costs.

The defining feature: assignment, not observation

Every clinical or health-related study design sits on one side of a foundational split, determined by a single question: does the investigator assign the exposure or intervention under study, or merely observe it as it occurs in routine practice?

• Interventional (experimental) study — participants are prospectively assigned, according to a protocol, to receive one or more interventions (which may include a placebo or an active comparator as a control). The investigator is an author of the exposure: they decide, by a pre-specified method such as randomisation, who receives what.
• Observational study — the investigator measures exposures, characteristics, and outcomes as they occur naturally, without assigning who receives what. The investigator is a bystander to the exposure, not an author of it. Cohort, case-control, and cross-sectional designs are the three main analytical observational designs.

This is a design-type distinction that CASRAI’s clinical study design orientation guide introduces at a landscape level; this guide goes deeper specifically into what “interventional” means, how the major regulators define it (and where those definitions diverge), and what obligations follow from the classification.

Interventional vs. observational: the practical comparison

Dimension	Interventional study	Observational study
Who decides the exposure/intervention	The investigator, per protocol	Not the investigator — occurs in routine care or naturally
Assignment method	Often randomised; may also be non-randomised (e.g. single-arm)	No assignment; groups are defined by pre-existing exposure or characteristic
Typical registries	ClinicalTrials.gov, EU CTIS, ISRCTN, other WHO ICTRP primary registries	Some observational studies register (e.g. on ClinicalTrials.gov) but registration is not universally mandated the way it is for trials
Ethics/IRB review depth	Typically full board review; informed consent for the assigned intervention itself	Can qualify for expedited review or, for some secondary-data-only designs, exemption
Regulatory trigger	May require an IND (drugs/biologics) or IDE (devices) in the US, or CTIS authorisation in the EU	Generally does not trigger IND/IDE or CTIS trial authorisation, though it may still need ethics and data-protection approval
Causal claims supported	Can support causal inference about the intervention’s effect, strength depending on design rigor (randomisation, blinding, control)	Generally supports association, not causation, due to potential confounding from non-random exposure

See CASRAI’s RCT vs. observational study comparison for a deeper look at the specific case of randomised interventional trials against cohort/case-control designs, and the dictionary entries for cohort study and case-control study for the two most common observational designs.

Why the classification matters

Whether a study is interventional is not a stylistic label — it is the trigger for a cascade of obligations:

• Registration. ICMJE member journals require prospective registration in a WHO-recognised primary registry (ClinicalTrials.gov, EU CTIS, ISRCTN, and others) as a condition of considering the study for publication, and the ICMJE’s own definition of “clinical trial” is deliberately broad and phase-agnostic.
• Regulatory authorisation. An interventional study testing an unapproved drug, biologic, or device, or an approved product outside its labelled use, can trigger an FDA Investigational New Drug (IND) or Investigational Device Exemption (IDE) application in the US, or a clinical trial application through the EU’s Clinical Trials Information System (CTIS).
• IRB/ethics review depth. Assigning an intervention to a human subject generally carries different, often greater, risk exposure than observing existing practice, which shapes how an IRB scopes its review and what informed consent must disclose.
• What conclusions the data can support. Because interventional designs can control for confounding through randomisation and blinding, they support stronger causal claims than most observational designs — which is exactly why regulators require them for the pivotal claims behind product approval, and why misclassifying a study doesn’t just create a paperwork problem, it can undermine the evidentiary weight the study was meant to produce.

Because these consequences differ by regulator and by product type, and because a study can be “interventional” under one definition and fall into a lighter-touch category under another, the classification has to be checked against each applicable regulatory definition individually — not inferred from a single generic rule of thumb.

The regulatory definitions — and where they genuinely differ

This is the part that trips research offices up most often: the major bodies that define “interventional” or “clinical trial” use similar-sounding but not identical language, and the differences are consequential, not academic.

ICMJE / WHO definition

The International Committee of Medical Journal Editors (ICMJE) defines a clinical trial as “any research project that prospectively assigns people or a group of people to an intervention, with or without concurrent comparison or control groups, to study the relationship between a health-related intervention and a health outcome.” This definition is deliberately broad and phase-agnostic — it explicitly includes early-phase (Phase 1) trials, not just later confirmatory trials. It is the definition that drives ICMJE’s prospective-registration requirement as a condition of publication in member journals, and it is closely aligned with the World Health Organization’s own clinical trial registration criteria used across the WHO International Clinical Trials Registry Platform (ICTRP) network of primary registries.

NIH definition: the four-question test

The US National Institutes of Health defines a clinical trial as “a research study in which one or more human subjects are prospectively assigned to one or more interventions (which may include placebo or other control) to evaluate the effects of those interventions on health-related biomedical or behavioral outcomes.” NIH operationalises this into a four-question test; a study is an NIH clinical trial only if the answer to all four is yes:

1. Does the study involve human participants?
2. Are the participants prospectively assigned to an intervention?
3. Is the study designed to evaluate the effect of the intervention on the participants?
4. Is the effect being evaluated a health-related biomedical or behavioral outcome?

Like the ICMJE definition, NIH’s test is phase-agnostic: an early-phase pharmacokinetics, safety, or maximum-tolerated-dose study still counts if all four questions are answered yes. NIH’s classification framework has also evolved recently — a study that meets both NIH’s basic-research definition and all four clinical-trial questions was previously classified as a “Basic Experimental Study Involving Humans” (BESH) and treated as a clinical-trial subtype; NIH notice NOT-OD-26-032 (issued January 29, 2026) announced that BESH will no longer be classified as a clinical trial going forward, a change research offices tracking NIH-funded interventional work should note when applying the four-question test to older versus newer guidance.

FDA: “clinical investigation”

FDA’s regulations use a related but separately defined term. Under 21 CFR 312.3 (governing investigational new drugs), a “clinical investigation” means any experiment in which a drug is administered or dispensed to, or used involving, one or more human subjects; for this purpose, an experiment is any use of a drug except the use of a marketed drug in the course of medical practice. The device equivalent sits in 21 CFR Part 812. This is narrower in framing than ICMJE’s or NIH’s definitions — it is anchored to product administration/use rather than to “assignment to an intervention” in the abstract — but functionally captures the same core fact pattern: a human subject receiving a drug, biologic, or device under an experimental protocol rather than as ordinary marketed-product use. Whether a specific study meets FDA’s definition determines whether an IND or IDE is required before enrollment can begin.

EU Clinical Trials Regulation: three categories, not two

The EU’s Clinical Trials Regulation (Regulation (EU) No 536/2014), the framework behind the EU Clinical Trials Regulation and CTIS, is structured around three categories rather than a simple interventional/observational binary:

• Clinical trial — the standard, full-obligation category: assignment of a subject to a pre-defined treatment strategy involving an investigational medicinal product, going beyond normal clinical practice.
• Low-intervention clinical trial — a distinct, lighter-touch subcategory (Article 2(2)(3)) for trials that pose only minimal additional risk to subject safety compared with normal clinical practice, typically because the investigational medicinal product already holds a marketing authorisation and is used within, or on evidence-based grounds outside, that authorisation, and any additional diagnostic or monitoring procedures involve no more than minimal additional risk or burden compared with normal clinical practice.
• Non-interventional study — falls outside the CTR/CTIS authorisation regime entirely; see CASRAI’s dedicated non-interventional study entry for the Article 2(2)(4) definition and how it differs from both trial categories above.

The practical significance of the “low-intervention” middle category is procedural: it qualifies for a shortened regulatory assessment timeline and reduced monitoring/insurance requirements under the CTR, while still requiring full CTIS authorisation as a clinical trial — it is not a lighter version of “non-interventional,” it is a lighter version of “clinical trial.”

The common conflation to avoid: a study can be squarely “interventional” under ICMJE and NIH’s broad, phase-agnostic definitions while qualifying for the EU’s low-intervention category, or while falling outside the FDA’s IND requirement entirely if it exclusively uses marketed drugs within labeling in a way that meets an IND-exemption condition. None of these four definitions is a subset of another; each has to be checked on its own terms against the specific study, sponsor, and jurisdiction.

Types of intervention

“Intervention” is not limited to investigational drugs. NIH, FDA, and WHO trial-registration frameworks all recognise a broad range of intervention types, including:

• Drug — a pharmacologic agent, investigational or an approved product used in a new way
• Biologic — vaccines, blood products, gene and cell therapies, and other biologically derived products
• Device — a medical device studied under an investigational protocol
• Procedure — a surgical, diagnostic, or therapeutic procedure assigned as part of the protocol
• Behavioural — a counseling, education, or behavior-change intervention
• Dietary — a nutritional supplement, dietary pattern, or feeding protocol
• Diagnostic — a diagnostic test or screening strategy being evaluated for its effect on outcomes or management decisions
• Radiation — a radiation-therapy or imaging protocol assigned as the studied exposure
• Health-services / systems-level — changes to how care is organised or delivered (e.g. a staffing model, a care pathway, a reminder system), often studied with cluster-randomised or stepped-wedge designs

Design families for interventional studies

Once a study is classified as interventional, the next design decision is which structural family fits the research question, the intervention, and the practical constraints of the setting:

• Parallel-group — the default design: each participant is assigned to one arm (intervention or control) for the duration of the study. Simplest to analyse, no carryover concerns.
• Crossover — each participant receives more than one intervention in sequence, separated by a washout period, and serves as their own control. Efficient for chronic, stable conditions where carryover and period effects can be managed; unsuitable for curative or irreversible interventions.
• Factorial — two or more interventions are evaluated simultaneously in combination (e.g. a 2×2 design testing drug A, drug B, both, or neither), allowing efficient evaluation of multiple questions and interaction effects from one trial.
• Cluster-randomised — whole groups (clinics, wards, schools, communities) rather than individuals are randomised to arms, typically because the intervention operates at the group level or individual randomisation would cause contamination between arms.
• Stepped-wedge — a cluster-based design in which every cluster eventually receives the intervention, but the order and timing of rollout is randomised, generating within-cluster before/after data as well as between-cluster comparisons. Common where withholding an intervention indefinitely from a control cluster is impractical or unacceptable.
• Adaptive — pre-specified rules allow planned modifications to the trial (e.g. sample size, arm allocation, or stopping) based on accumulating data, without undermining the trial’s statistical validity.
• Single-arm — all enrolled participants receive the same intervention with no concurrent control arm, and outcomes are compared against an external benchmark (historical control, natural history, or performance goal). Still interventional — the investigator still assigns the intervention — but weaker for causal inference than a controlled design, generally reserved for settings like rare diseases or oncology where a concurrent control is infeasible or unethical.

CASRAI’s guide on designing a clinical trial covers how endpoint selection, sample size, and randomisation scheme choices interact with these design families in practice.

Registration obligations

Classifying a study as interventional typically activates one or more registration obligations:

• ClinicalTrials.gov — the US registry; registration is a statutory requirement under FDAAA 801 for “applicable clinical trials” meeting specific product and phase criteria, and separately an NIH policy requirement for NIH-defined clinical trials, in addition to being the registry most US institutions default to for ICMJE compliance.
• EU CTIS — mandatory for clinical trials and low-intervention clinical trials authorised under the EU Clinical Trials Regulation; see the EU CTR/CTIS entry for the authorisation mechanics.
• ISRCTN and other WHO ICTRP primary registries (ANZCTR, CTRI, ChiCTR, DRKS, JPRN, and others) — used depending on sponsor location and study population; all feed into WHO’s ICTRP global search.
• ICMJE prospective registration — ICMJE member and many other journals require registration in a WHO-recognised primary registry before enrollment of the first participant, as a condition of considering the resulting manuscript for publication; see prospective clinical trial registration and clinical trial registration for the mechanics and consequences of registering late.

Observational studies are registered far less consistently — some journals and funders encourage or require it, but there is no equivalent universal, cross-registry mandate the way there is for interventional trials, which is itself a downstream consequence of the interventional/observational classification.

Why this is a first question for a research office

Because the interventional/observational determination cascades into registration, IND/IDE applicability, IRB review pathway, and what causal claims the eventual publication can defend, research offices generally need to settle it before a protocol is finalised — not after enrollment has started. Answering it wrong has concrete costs: a study that is interventional but was registered or reviewed as if it were observational can face late-registration penalties, an ICMJE journal declining to consider the resulting manuscript, an unauthorised trial finding at an FDA or EU inspection, or an IRB determination that consent already obtained was inadequate for the actual risk profile of an assigned intervention. Because the ICMJE, NIH, FDA, and EU definitions are not identical, the safest practice is to check the specific study against each applicable definition individually, document the determination and its rationale, and revisit it if the protocol is amended in a way that changes what is being assigned to participants.

Frequently asked questions

Is a study interventional if it only involves an approved drug used within its label?

Not necessarily by that fact alone — what matters is whether participants are prospectively assigned to a treatment strategy per protocol to evaluate an effect. A study that assigns participants to an on-label drug versus a comparator to test a health outcome is still interventional under the ICMJE and NIH definitions, even though it may not trigger an FDA IND if it meets one of the IND-exemption conditions in 21 CFR 312.2(b) (e.g. not intended to support a new labeling claim, no significant safety risk introduced). Regulatory-authorisation triggers and clinical-trial classification are related but separate questions.

Can a single-arm study be interventional?

Yes. The defining feature of “interventional” is that the investigator assigns the intervention according to a protocol, not that there is a concurrent control arm. A single-arm study still assigns every participant to receive the studied intervention; it is interventional, just without a randomised comparator, which is why it generally supports weaker causal claims than a controlled design.

Does an interventional study always require IRB full-board review?

Not automatically — some minimal-risk interventional studies can qualify for expedited review under applicable human-subjects regulations, depending on the specific research activity and risk level, not solely on the interventional/observational label. The interventional classification is one input into the IRB’s review-pathway determination, not the only one.

Is “non-interventional” the same as “observational”?

They overlap heavily but are defined by different bodies for different purposes. “Observational” is the general research-methods term for a study design that does not assign exposure. “Non-interventional study” is a specific EU Clinical Trials Regulation category (Article 2(2)(4)) with its own regulatory definition and consequences for CTIS authorisation. See the dedicated non-interventional study entry for the distinction.