Beauty Tools & Accessories
Dr. Vinay Prasad Embraces Policy-Based Evidence Making
Aug
The phrase “policy-based evidence making” sounds like evidence-based policymaking after someone put the words in a blender. The difference, however, is not cosmetic. Evidence-based policy begins with a problem, examines the best available information, and allows the findings to influence the decision. Policy-based evidence making starts with the desired decision and goes shopping for research that will make it look scientifically inevitable.
That accusation has been directed at Dr. Vinay Prasad, a hematologist-oncologist, researcher, commentator, and former senior Food and Drug Administration official. The criticism initially centered on his enthusiastic endorsement of Dr. Jay Bhattacharya as director of the National Institutes of Health and Prasad’s proposed research agenda for revisiting pandemic policies.
The title of this article therefore describes a published criticism, not an uncontested fact. Prasad and his supporters would likely argue that he was calling for neglected questions to receive serious study. Critics counter that his language appeared to announce both the research program and its preferred verdict before investigators had collected the evidence. That tension offers a useful case study in how science, politics, funding, and institutional power can become tangled together like charging cables in a desk drawer.
What Is Policy-Based Evidence Making?
Evidence-informed policymaking does not require public officials to behave like emotionless laboratory robots. Governments must consider ethics, costs, feasibility, public preferences, legal authority, and uncertainty. Scientific evidence is crucial, but it rarely makes every policy choice automatically.
Policy-based evidence making crosses a different line. It occurs when an institution commissions, selects, interprets, or promotes research mainly to justify a conclusion that leaders have already embraced. The central problem is not that policymakers have values or priorities. The problem is that the research process becomes a stage production in which the ending has been written before the first act.
Warning signs include narrowly framed questions, selective outcome measures, one-sided grant announcements, suppression of inconvenient findings, exaggerated confidence, and different evidentiary standards for favored and disfavored policies. A rigorous study may still emerge from a politically motivated funding program, but the risk of bias grows when researchers know which answer will make the sponsor happiest.
Why Prasad’s Remarks Attracted Criticism
The Bhattacharya Endorsement
In November 2024, Prasad published an article praising President Donald Trump’s selection of Jay Bhattacharya to lead the NIH. Bhattacharya, a Stanford physician and health economist, became nationally prominent through his opposition to broad COVID-19 lockdowns and school closures. The Senate later confirmed him, and he took office as NIH director in April 2025.
Prasad predicted that Bhattacharya would establish funding to examine the harms of lockdowns, school closures, vaccine mandates, and other pandemic interventions. Studying those issues is entirely legitimate. School disruption, delayed medical care, economic damage, mental health effects, learning loss, and unequal social burdens are serious subjects deserving careful investigation.
The controversy arose from how Prasad described the goal. He suggested that the resulting scholarship would deter future public health “tyrants.” Critics argued that this wording framed the policies as both harmful and tyrannical before the proposed research had begun. In their view, Prasad was not merely identifying unanswered questions. He was announcing the prosecution’s theory and asking science to assemble the exhibits.
A Research Question or a Predetermined Destination?
A neutral research agenda might ask: What were the benefits, harms, distributional effects, and opportunity costs of different pandemic interventions under varying conditions? It would examine whether timing, duration, disease prevalence, community behavior, age, socioeconomic status, and implementation quality changed the results.
A predetermined agenda instead asks researchers to document the damage caused by policies already labeled ineffective or oppressive. That framing can influence which proposals receive funding, which outcomes investigators measure, and which findings receive institutional publicity.
This does not prove that every study in such a portfolio would be biased. Independent researchers can produce uncomfortable results even when a sponsor expects something else. Still, leadership language matters. Scientists notice whether officials reward genuine uncertainty or treat research as ammunition in an old political battle.
The Strongest Defense of Prasad’s Position
A balanced analysis must acknowledge that research funding is never completely neutral. Congress, federal agencies, foundations, universities, and private companies routinely select priorities. They decide that cancer, Alzheimer’s disease, climate-related illness, maternal mortality, or pandemic preparedness deserves more attention. Choosing a topic is not the same as dictating a conclusion.
Prasad’s defenders could reasonably argue that the adverse consequences of pandemic restrictions received less attention than infection control and vaccine development during the emergency. Governments needed answers quickly, and research naturally focused on reducing hospitalization and death. Questions about educational setbacks, missed screenings, small-business closures, social isolation, and the long-term effects of mandates often developed more slowly.
Creating grants to investigate those outcomes could correct an imbalance. It might also improve future emergency planning by showing when restrictions were useful, when they were excessive, and which alternatives produced better results.
The defense becomes weaker, however, when the sponsor describes the expected conclusion in moral or partisan terms. “Study an underexamined possibility” is science. “Build scholarship that will punish the villains we have already identified” begins to resemble advocacy wearing a laboratory coat.
Randomized Trials, Observational Evidence, and a Consistency Problem
Prasad built much of his professional reputation by demanding stronger clinical evidence. He has criticized drug approvals based on surrogate endpoints, small uncontrolled studies, weak comparisons, and outcomes that do not clearly improve patients’ survival or quality of life. Those critiques often raise important questions, particularly in oncology and rare-disease medicine, where treatments may be extremely expensive and supported by limited data.
He has also emphasized randomized controlled trials. Randomization can reduce confounding and provide highly credible estimates of whether an intervention caused an outcome. Yet randomized trials are not a magic wand, and the absence of one does not make every other form of knowledge disappear in a puff of statistical smoke.
Some public health measures cannot be randomized easily, quickly, ethically, or at sufficient scale. Researchers may need natural experiments, interrupted time-series analyses, cohort studies, case-control designs, mathematical models, qualitative research, and comparisons across jurisdictions. Each method has weaknesses, but careful triangulation can be more informative than pretending that only one design counts.
The central criticism of Prasad is therefore not simply that he likes randomized trials. It is that his evidentiary demands may appear asymmetrical. Critics say he has sometimes demanded exceptionally strong proof for policies he opposes while accepting suggestive or nonrandomized evidence when it supports his preferred conclusions.
Consistency is essential. A researcher should not require a double-blind trial to accept one claim and then treat a convenient ecological correlation as decisive proof of another. The scientific rulebook should not change uniforms at halftime.
From Public Commentary to FDA Authority
The debate became more consequential when Prasad entered federal regulatory leadership. In 2025, he was appointed to lead the FDA’s Center for Biologics Evaluation and Research, which oversees vaccines, blood products, and many cell and gene therapies. He also served as the agency’s chief medical and scientific officer.
Prasad and FDA Commissioner Marty Makary subsequently outlined a revised approach to COVID-19 vaccine regulation. Their framework favored continued access for older adults and people at elevated risk while seeking stronger clinical-outcome evidence for broad vaccination of healthy, lower-risk populations. An FDA decisional memorandum later emphasized uncertainty surrounding surrogate immune markers, limitations in observational effectiveness studies, lower contemporary rates of severe disease, and the need to reconsider the benefit-risk calculation for healthier people.
Supporters viewed this as a overdue return to stronger regulatory standards. They argued that pharmaceutical manufacturers should demonstrate meaningful clinical benefits rather than relying indefinitely on antibody responses and extrapolation from earlier products.
Critics responded that the new approach discounted useful real-world evidence, imposed impractical trial requirements, and changed established vaccine policy without sufficient transparency or external consultation. The disagreement intensified after Prasad circulated an internal message asserting that an FDA review had linked COVID-19 vaccination to the deaths of at least 10 children. Detailed evidence supporting that number was not publicly presented at the time, and vaccine specialists questioned whether reports submitted to passive surveillance systems could establish causation.
Twelve former FDA commissioners later warned that the proposed changes threatened evidence-based vaccine regulation and public health security. Prasad’s allies maintained that entrenched institutions were resisting overdue scrutiny. His critics saw the episode as precisely the danger identified earlier: a strong policy conclusion appearing before a transparent, reproducible evidentiary record.
After a turbulent tenure involving vaccines and several disputed rare-disease decisions, Prasad left the FDA in April 2026. His departure did not settle the broader argument. It merely moved the debate from one office to the larger question of how regulators should handle uncertainty without allowing political preferences to select the evidence.
How NIH Could Study Pandemic Policies Without Rigging the Answer
Publish Neutral Questions in Advance
A credible pandemic research program should define broad questions before selecting investigators. It should examine benefits and harms, including disease outcomes, mental health, education, employment, inequality, delayed care, public trust, and unintended behavioral effects. Questions should permit findings that support, oppose, or complicate prior policies.
Fund Competing Hypotheses
Grant portfolios should include researchers with different methodological approaches and prior views. Intellectual diversity is useful when it produces testable disagreement rather than partisan theater. A panel composed entirely of people who already share the director’s conclusion is not a scientific advisory group; it is a group chat with a federal budget.
Protect Independent Peer Review
Scientific merit should be evaluated by qualified reviewers who are insulated from pressure to favor politically convenient projects. Agency leaders may set broad priorities, but they should not quietly convert ideological agreement into a grant criterion.
Preregister Methods and Outcomes
Investigators should specify hypotheses, populations, analytic plans, and primary outcomes before analyzing the data whenever feasible. Preregistration cannot eliminate bias, but it makes it harder to search through dozens of measures and present only the result that supports the preferred narrative.
Publish Negative and Inconclusive Results
Null findings are not failed research. If a study finds little evidence that a policy caused a predicted harm, the result should remain available. Data, code, assumptions, and sensitivity analyses should be released when privacy and legal requirements allow independent verification.
Separate Scientific Findings From Value Judgments
Evidence may estimate how many infections, school days, jobs, or medical visits were affected by a policy. It cannot independently determine how society should weigh every outcome. Officials should clearly distinguish empirical findings from moral and political choices instead of presenting their values as if they arrived directly from a spreadsheet.
Why the Debate Matters Beyond One Physician
The controversy surrounding Vinay Prasad is larger than Prasad. Every administration enters office with priorities, beliefs, allies, and grievances. The temptation to use research agencies to validate those commitments is bipartisan and enduring.
Public trust depends less on officials claiming to be perfectly objective than on institutions making bias difficult. Transparent methods, competing hypotheses, independent review, accessible data, and honest uncertainty offer better protection than any leader’s promise to follow “the real science.” That phrase has been borrowed by so many factions that it now needs its own lost-and-found department.
Studying the failures of pandemic policy is necessary. So is studying its successes. Researchers should investigate whether specific interventions helped, harmed, or produced mixed results under particular circumstances. They should not be expected to create a permanent scholarly monument to the political preferences of the person controlling the grant budget.
Conclusion
Dr. Vinay Prasad’s call for research into the harms of lockdowns, school closures, and vaccine mandates addressed legitimate and important questions. The objection was not that these policies should remain beyond examination. It was that his stated purpose appeared to contain the verdict before the investigation.
That distinction separates a defensible research priority from policy-based evidence making. Science can begin with concern, suspicion, or a provocative hypothesis. It must still preserve the possibility that the evidence will surprise, disappoint, or contradict the people who funded the work.
The best response to contested pandemic history is not to replace one official narrative with another. It is to create research systems capable of testing both, exposing uncertainty, and publishing the results even when nobody gets the satisfying victory speech they ordered.
Experience-Based Lessons From the Policy and Evidence Debate
Several practical experiences common to research institutions help explain why this controversy resonates. Consider a grant-review panel asked to fund studies on the “damage caused by school closures.” Even without an explicit order, applicants quickly understand the desired framing. A proposal examining whether temporary closures reduced transmission may appear less responsive than one promising to document learning loss. The wording of the funding announcement has already tilted the playing field before reviewers score a single application.
A better announcement would request studies of the short- and long-term effects of different school policies, including educational, medical, social, and economic outcomes. This small change does not forbid research into harm. It simply leaves the answer open. Experienced researchers know that neutral wording is not bureaucratic decoration; it is part of the study’s intellectual architecture.
Hospital committees provide another useful example. Suppose leaders strongly believe that a new screening program will improve care. They may choose an outcome that is easy to increase, such as the number of tests completed, while overlooking false positives, unnecessary procedures, costs, anxiety, and whether patients actually live longer or feel better. The committee can then declare success because it measured the outcome most likely to applaud the policy.
This is why good evaluation plans establish meaningful benefits, potential harms, and stopping rules in advance. They also invite skeptics into the design process. A thoughtful critic can identify blind spots that an enthusiastic project team misses. Skepticism is most useful before implementation, not after the press release has been drafted.
Newsrooms face a related problem. A reporter may begin with a strong theory about a controversial public figure and collect quotations that support it. Every quotation may be accurate, yet the final story can still mislead if conflicting evidence, relevant context, or reasonable alternative interpretations are omitted. Accuracy at the sentence level does not guarantee fairness at the article level.
The same principle applies to scientific reviews. Researchers can cite only genuine studies and still produce a distorted conclusion through selective inclusion. Transparent search strategies, prespecified criteria, risk-of-bias assessments, and disclosure of excluded evidence help readers determine whether the review explored a question or merely assembled a case.
Finally, experienced policymakers learn that uncertainty must be communicated without paralysis. Officials sometimes must act before ideal evidence exists. The responsible approach is to explain what is known, what remains uncertain, why a temporary decision was made, and what evidence would trigger revision. Policies should include mechanisms for reassessment rather than becoming identity badges that leaders feel compelled to defend forever.
These experiences suggest a simple test for distinguishing evidence-informed leadership from policy-based evidence making: Would the institution accept, publish, and act upon a well-conducted study that reached the opposite conclusion from its leaders? If the honest answer is no, the research program is not searching for knowledge. It is searching for supporting paperwork.
Editorial note: The headline reflects a documented criticism of Dr. Vinay Prasad’s approach. Disputed claims and interpretations are attributed to their proponents or critics rather than presented as settled conclusions.
