science / health
NIH Cut 383 Clinical Trials; the Patients Inside Them Had No Warning
When a randomized trial halts mid-enrollment, the data inside it cannot be salvaged. The 2025 NIH grant terminations created an evidence gap that no FY2026 appropriation can refill.
A randomized clinical trial is not a spreadsheet you can save and reopen. When enrollment halts mid-study, the control arm contaminates as patients seek treatment elsewhere, the randomization sequence breaks, and the baseline data collected on every participant becomes unusable as evidence of anything. That is the part of the 2025 NIH cuts that no appropriations bill can undo. A peer-reviewed analysis in JAMA Internal Medicine found that 383 federally funded clinical trials, enrolling more than 74,000 patients, were disrupted by grant terminations between February and August 2025. The dollars can be restored. The trials cannot.
Congress has now partially clawed back the cuts, adding $228 million in targeted increases to cancer and Alzheimer’s research after the proposed FY2026 HHS budget called for slashing the NIH from roughly $47 billion to $29 billion. That number is the headline. It is also the wrong story. The patients enrolled in halted Phase 2 and Phase 3 trials did not get their experimental therapies back when the appropriations conference report passed. The randomization in their studies was already broken. The methodologists at the affected sites had already moved on, the IRBs had already closed the protocols, and the multi-year longitudinal data had already been rendered statistically unsalvageable.
This is the part that does not appear in budget tables.
What the JAMA Study Actually Found
The JAMA Internal Medicine analysis, authored by Vishal Patel and colleagues including Anupam Jena at Harvard Medical School, is the first peer-reviewed accounting of what the spring 2025 grant terminations cost in trial-level terms. The authors cross-referenced terminated NIH grants against ClinicalTrials.gov registrations and identified 383 trials affected, representing roughly 1 in 30 of all NIH-supported trials active during the period. The disruption was concentrated, not random.
- 115 of the disrupted trials were oncology studies, according to the JAMA analysis; cancer research was the single most affected category by trial count.
- Mental health research funding fell sharply year-over-year, per the AJMC summary of the dataset, the steepest cut in any disease category.
- New cancer grant awards fell roughly 23 to 24 percent, even as ongoing cancer trials were being terminated, CBS News reported from the same data.
- More than 74,000 enrolled patients were affected across the disrupted studies, Cancer Health reported.
- The proposed FY2026 NIH budget cut totaled roughly $18 billion, with HHS budget documents calling for the consolidation of 27 institutes into 8.
The thesis of the chart is the disproportion. Cancer and behavioral health trials were not affected in proportion to their share of the NIH portfolio; they were affected at a rate roughly twice their baseline share. The grant terminations were carried out by keyword sweep rather than scientific review, and the keywords used (related to gender, equity, and HIV among other categories) intersected most heavily with these therapeutic areas.
Why the Data Cannot Be Recovered
A clinical trial is a designed inferential machine. Its statistical power, the probability that it will detect a real effect if one exists, depends on enrollment numbers, randomization integrity, and the consistency of measurement across the study period. Halt the trial, and each of those degrades.
Patients in the placebo arm who learn the trial is over typically seek active treatment elsewhere, contaminating any future analysis of disease progression in the control group. Patients in the treatment arm who lose access to the experimental therapy mid-protocol drop out at high rates, biasing the surviving cohort toward those who happened to tolerate the intervention. Baseline labs, imaging, biomarker collection (the unglamorous infrastructure of evidence) become orphaned data, scientifically valid for nothing.
This is what Anupam Jena and his JAMA co-authors mean when they write that the trials were not paused but disrupted. The distinction is the entire story. A pause assumes resumption. Disruption produces a multi-year evidence gap, because rebuilding a Phase 3 cancer or HIV trial means starting over: new protocol, new IRB review, new enrollment, new baseline measurements. The FDA’s own description of clinical research places Phase 3 studies on a one- to four-year timeline once enrollment is complete; for the most affected disease areas, the cumulative path back to evidence is measured in years, not budget cycles.
What Congress Restored, and What It Did Not
The FY2026 conference report adds back $128 million for the National Cancer Institute and $100 million for Alzheimer’s research, Science magazine reported on April 28. That $228 million in targeted restorations is real money. It is also approximately 1.3 percent of the originally proposed $18 billion cut, and none of it is structured to reconstitute the specific trials that were terminated. New money funds new grants. The disrupted protocols, with their randomization broken, will need to be redesigned, re-reviewed by institutional review boards, and re-enrolled from a new baseline; at the reported pace of NIH grant terminations through 2025, that restart sequence is itself the evidence gap.
For patients who enrolled in those trials specifically because they had exhausted standard care, the math is harsher. Many will not be alive in 2027 to enroll in the replacement protocol.
The Pattern Below the Headline
The political framing of the 2025 NIH cuts treated research grants as a discretionary line item. The empirical reality, captured in the JAMA dataset, is that the cuts shifted the United States off the late-stage trial evidence track for several therapeutic areas at once. FDA approvals in cancer, HIV prevention, and behavioral health rest on Phase 3 evidence pipelines that take six to ten years to construct and a single budget cycle to dismantle.
The 1993 cancellation of the Superconducting Super Collider cost roughly $2 billion in sunk investment and ceded the energy frontier of physics to CERN for thirty years. The 2025 NIH cuts are in the same category of decision: a budget number that, on paper, looks recoverable, but that in practice scattered the assembled cohorts, contaminated the data, and reset the clock on evidence that was years from maturity. The dollars came back. The trials did not.
The patients did not get a press release about the appropriations conference. They got letters telling them their study had ended.