Moving the Goalposts, Quietly: The Case for Endpoint Stewardship in Industry-Sponsored Trials
Photo: clinical research protocol documents with checklist and pen, via clinessentials.com
Somewhere between ClinicalTrials.gov and the pages of a major journal, trials change. The primary endpoint that was registered at study initiation becomes a secondary outcome. A new co-primary endpoint appears. The timepoint of assessment shifts from twelve months to six. None of this is announced; it is simply discovered by researchers who take the trouble to compare records — and most readers never do.
The empirical literature on this is sobering. Audits of registered trials versus their publications have repeatedly found that a substantial share of industry-sponsored studies report a primary outcome that differs from the registered one, and that these switches correlate with favorable results. A trial that fails its original endpoint but succeeds on a substitute one still gets published, still gets marketed, and still informs prescribing. The registration record exists precisely to prevent this, yet enforcement in the United States remains largely aspirational.
Why Endpoint Switching Is Not a Technicality
Defenders of the practice note that endpoints sometimes evolve for legitimate scientific reasons: new data, regulatory feedback, patient-relevant measures. That is true. But the direction of the changes matters. Endpoint modifications overwhelmingly trend toward significance — soft outcomes replace hard ones, shorter timepoints replace longer ones, responder definitions loosen. When changes flow in one direction, 'scientific evolution' starts to look like outcome optimization.
The deeper problem is epistemological. Randomization protects a trial against bias in treatment assignment; prespecification protects it against bias in analysis. A switched endpoint dissolves the second protection. The multiplicity of possible endpoints, timepoints, and analysis populations gives sponsors enormous degrees of freedom, and readers of the final paper cannot reconstruct what was tried and discarded.
The Transatlantic Gap
The FDA requires prospective registration and has taken enforcement action on registration lapses, but the agency's oversight of endpoint consistency between registry and publication remains thin. Contrast this with developments elsewhere: European funders and several national regulators increasingly tie publication and reimbursement decisions to registry-publication concordance, and the ICMJE's policies — while stronger on paper in the US than in practice — have counterparts in Europe enforced with greater institutional teeth.
The result is a regulatory arbitrage. Sponsors can design globally, register flexibly, and publish selectively, with the weakest jurisdiction's standards setting the effective floor.
What Endpoint Stewardship Would Look Like
An enforceable stewardship framework would have three components. First, locked registries: any change to a primary or key secondary endpoint after enrollment begins would require a documented amendment visible on the registry, with the original entry preserved and timestamped. Second, mandatory concordance statements: journals would require authors to affirm — and provide a link demonstrating — that reported endpoints match registered ones, with discrepancies explained in the paper itself, not in a supplement. Third, consequences: registration noncompliance and unexplained endpoint switching would carry the same professional weight as other forms of research misconduct, affecting funding eligibility and institutional review.
None of this requires new science. It requires treating endpoint definitions with the seriousness they deserve — as the precommitted contract between a trial and its future readers.
The Stakes
Clinicians make decisions based on endpoint hierarchies. A drug approved on a surrogate or a softened composite may be prescribed for years before harder outcomes clarify its value — and by then, the endpoint switch that shaped the evidence base is long forgotten. Endpoint integrity is not a bookkeeping matter. It is the difference between evidence that was designed and evidence that was engineered. Making endpoint stewardship enforceable would be the most consequential trial-accountability reform since mandatory registration itself.
