Stanford neurologist Mitch Miglis lost his NIH long Covid research grant in March 2025. Steven Phillips argues in STAT that the field's own scientific framing made the cut possible.
On March 25, 2025, the National Institutes of Health told Stanford neurologist Mitch Miglis that his long Covid research grant was dead. The termination notice declared the work "incompatible with agency priorities" and said "no modification could bring it into alignment." The grant had been running for two years. Miglis's team had collected its data. The analysis was weeks from completion. The focus was narrow and stubborn: what long Covid does to the autonomic nervous system, the wiring that keeps heart rate and blood pressure stable when a person stands up.
For the patient population the study was built around, the loss is concrete. For the broader field, it is emblematic. In a First Opinion essay published June 11 in STAT News, Steven Phillips argues that the grant's death is a symptom of a problem the field built for itself. Phillips, vice president for science and strategy at the COVID Collaborative and a fellow of the American College of Epidemiology, names the trap plainly: "Attempting to force the condition into the biomedical paradigm was a mistake."
The argument is structural, not merely political. Long Covid presents as a sprawling, multi-system condition with no clean biomarker and no consensus animal model. The flagship federal research effort, NIH's RECOVER program, organized the field around the assumption that a single biomedical cause would eventually be found. That framing was defensible when the disease was being dismissed as psychosomatic. It became fragile once the funding climate turned hostile — and as Phillips's analysis frames it, a field whose legitimacy rests on a discovery that has not arrived is easier to defund than one whose legitimacy rests on the documented experience of patients.
Phillips is naming this as a design flaw in the science, not just a political failure. The implication, which he draws explicitly, is that the field's vulnerability to federal abandonment is partly self-inflicted. A research enterprise organized around a single patron and a single mechanism is, by construction, brittle. When that patron walks away, the field has no structural redundancy to fall back on.
The Miglis grant is the only concrete named termination in the version of the essay available for this analysis. Phillips's broader argument depends on a wider pattern of federal disengagement, not a single case. Whether that pattern holds at the scale he describes is a question for further reporting, not an assertion to be taken from the essay alone.
What the essay does establish is the question. If the field's survival depends on a single federal patron, and that patron's priorities can shift on a six-month notice, then the field needs a different architecture. Phillips does not specify what that architecture looks like in operational detail. He names the requirement: a model that distributes funding, data ownership, and research direction across multiple actors so that no single defunding decision can collapse the whole enterprise.
The Miglis study, had it survived, would have been one piece of a much larger puzzle about autonomic dysfunction in long Covid. Its termination leaves a hole that one terminated grant always leaves: a specific set of questions, asked by a specific set of researchers, on a specific set of patients, that will now go unasked. The bigger question, which Phillips's essay forces into the open, is whether the field will treat that pattern as a temporary setback or as a structural warning.