| Scenario | Classification | Key Reasoning |
|---|---|---|
| Participants with social anxiety perform a modified Stroop test using emotional facial expressions. Researchers measure response time differences in color naming. | BESH | Measures a cognitive process (response time), not a health outcome |
| Volunteers are randomly assigned to receive different non-toxic compounds metabolized in the liver. Researchers sequence their Cytochrome P450 genes and measure compound levels in blood and urine. | BESH | Aims to understand metabolic mechanisms, not evaluate a clinical treatment |
| Addiction clinic patients are randomly assigned to one of two behavioral therapies. Researchers compare mental health and addiction measures over time. | Clinical Trial | Evaluates treatment effectiveness against health-related outcomes |
| Researchers analyze retrospective electronic health records from multiple medical centers using AI to identify sepsis predictors. | Observational | No prospective assignment to an intervention |
What Counts as a ‘Clinical Trial’?
How the NIH’s Shifting Definition Could Reshape Transparency in Research
This note is a companion to the Experimental Designs chapter of Global Health Research in Practice, which discusses trial registration on ClinicalTrials.gov and why transparency matters in experimental research.
Late last month, the NIH narrowed the scope of what qualifies as a clinical trial. Effective May 5, 2026, basic experimental studies with humans—known as BESH—will no longer fall under that designation. Many studies involving human participants—brain imaging, stress physiology, cognitive psychology—will no longer need to register on ClinicalTrials.gov or comply with clinical trial reporting requirements.
I think this is the right call. Since the NIH expanded its definition in 2014, researchers running basic experiments have faced requirements designed for drug and device trials. The fit was poor. ClinicalTrials.gov was built around regulatory medicine, not discovery science.
But registration serves a broader purpose than regulatory compliance. Study registration is a scientific public good. It creates a time-stamped record of what a study plans to measure before results are known—the main defense against selective reporting and outcome switching. Removing BESH from the clinical trial category removes the automatic registration requirement, which means some studies may no longer leave a public audit trail.
What Makes Something a “Clinical Trial”?
Since the last update in 2014, the NIH has defined 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.”
“Intervention” has a broad meaning which is not changing—essentially any experimental manipulation of the subject or their environment. This could be giving a new drug or a placebo, but it also includes interventions like manipulating visual stimuli inside an fMRI scanner and randomly assigning participants to view varying types of emotional images.
The change comes in the interpretation of the phrase “health-related biomedical or behavioral outcomes.” Previously, the NIH included any biomedical or behavioral phenomenon or process in this definition. This meant outcomes related to morbidity and mortality, but also outcomes like reaction times, cortisol levels, and neural activation. Under that reading, a basic psychology experiment or a neuroimaging study could be classified as a clinical trial, even though no treatment was being tested and no clinical decision was at stake.
The 2026 change narrows that interpretation. The NIH now reads “health-related biomedical or behavioral outcome” more restrictively, as the notice explains:
“The NIH now considers a health-related biomedical or behavioral outcome as having the potential for direct advancement of health. Although BESH research, which produces fundamental information about biology or behavior, might eventually inform advances in health, it is not conducted with the express intent of changing clinical practice or health but rather aims to understand fundamental aspects of phenomena without immediate clinical applications. Therefore, BESH research is no longer considered to meet the NIH definition of a clinical trial.”
The core definition didn’t change. The interpretation of “health-related” did. The distinction rests on intent: is the study trying to understand how something works, or to evaluate whether a treatment affects health?
Case examples
The NIH provides case studies to illustrate how the classification works:
The clear-cut cases are straightforward. The harder cases sit in between:
- A study measuring cortisol responses to a stress protocol. If framed as “understanding the neuroendocrine stress response,” it’s BESH. If framed as “evaluating a stress-reduction intervention,” it’s a clinical trial. The protocol could be identical.
- An exercise study measuring cognitive performance. If the goal is “understanding how acute physical activity affects working memory,” it’s BESH. If the goal is “testing exercise as an intervention for cognitive decline,” it’s a clinical trial.
- A nutrition study measuring metabolic markers. Basic physiology or clinical intervention? The answer often depends more on how the grant application is written than on what actually happens in the lab.
The NIH advises investigators to consult with their program officers when the distinction is unclear.
How other systems define “clinical trial”
The NIH is not the only body that defines what counts as a clinical trial. The World Health Organization, International Committee of Medical Journal Editors (ICMJE), and European Union each have their own definitions, and they don’t all draw the line in the same place. The definitions range from narrow to broad: the EU regulation is the most restrictive, limited to medicinal products; the WHO and ICMJE occupy the middle, requiring a health-related intervention evaluated against a health outcome; and the NIH pre-2026 definition was the most expansive—any intervention, any behavioral or biomedical outcome, no intent requirement.
The 2026 change adds an intent requirement—direct health advancement—that brings the NIH from the broadest definition into closer alignment with the WHO and ICMJE.
| Criterion | NIH (pre-2026) | NIH (post-2026) | WHO / ICTRP | ICMJE | EU CTR 536/2014 |
|---|---|---|---|---|---|
| Human subjects | ✅ | ✅ | ✅ | ✅ | ✅ |
| Prospective assignment | ✅ | ✅ | ✅ | ✅ | ✅ |
| Intervention | ✅ | ✅ | ✅ | ✅ | ✅ |
| Intervention must be health-related | ❌ | ❌ | ✅ | ✅ | ✅ |
| Health-related outcomes | ✅ | ✅ | ✅ | ✅ | ✅ |
| Intent: direct health advancement | ❌ | ✅ | — | — | — |
| Limited to medicinal products | ❌ | ❌ | ❌ | ❌ | ✅ |
Registration Benefits More Than Just Clinical Trials
The 2026 change resolves a definitional inconsistency. But because registration requirements are tied to what counts as a “clinical trial,” narrowing the definition narrows the scope of mandatory registration. That matters because the problems registration addresses are not unique to drug or device studies.
Pre-registration creates a public, time-stamped record of a study’s research questions, primary outcomes, and analysis plan before results are known. Without that record, there is no way to distinguish a pre-specified hypothesis from one constructed after examining the data. Researchers inevitably face many defensible analytic choices—how to define outcomes, which covariates to include, what time windows to use, which subgroups to explore. Individually legitimate, these choices collectively create enough flexibility to tilt literatures toward clean, positive stories.
Comparisons of registry entries and publications routinely reveal added, dropped, or redefined outcomes, with statistically significant findings disproportionately represented in print. Registered Reports—a publishing format where peer review happens before data collection—yield far fewer “positive” results than traditional publications, which suggests the standard pipeline filters out unfavorable findings. None of these dynamics are specific to pharmaceuticals. Basic experimental studies are subject to the same incentives.
Paragraph 35 of the 2013 Declaration of Helsinki calls for prospective registration of all research involving human subjects—not just studies testing treatments. It is an ethical benchmark rather than a regulatory mandate, but its scope is broader than the operational definitions used by funders and journals. By narrowing the clinical trial label, the NIH resolved a regulatory mismatch—but widened the gap between what is required and what Helsinki says should be standard practice.
Where to Register
If you are planning a study that is a clinical trial by NIH/ICMJE/WHO standards, register in a WHO primary registry or another registry that meets ICMJE standards. ClinicalTrials.gov is one such registry, as are the EU Clinical Trials Register, ISRCTN, and others. This is what journals expect.
For BESH and other studies outside that frame, ClinicalTrials.gov is usually the wrong tool. It is tied to legal reporting regimes, and its data model is built around trial phase, FDA status, and sponsor class. For a behavioral lab randomizing participants to memory tasks, most of that infrastructure is irrelevant. Better alternatives exist for researchers who want to preregister studies that are not clinical trials:
- OSF Registries accepts preregistrations for any study type across any discipline. Flexible templates fit studies with no clinical component.
- AsPredicted provides a streamlined, nine-question form designed to be completed in minutes.
- EGAP serves researchers in political science, public policy, and governance.
- AEA RCT Registry covers randomized controlled trials in economics and development.
The NIH made the right call. BESH studies were never a natural fit for clinical trial infrastructure, and the 2026 change corrects that mismatch. But the pressures that prospective registration is meant to counter—outcome switching, selective reporting, and publication bias—operate regardless of how a study is classified.
If you prospectively assign participants, measure outcomes, and draw inferences, the case for a public, time-stamped record still applies. The label has narrowed. The rationale for transparency has not.