Bill Cassidy’s Footnotes vs. the Executive Branch: Who Actually Runs U.S. Health Policy?

NB: My analysis here does not rest on personal familiarity and long-standing working relationship with Kennedy, but on the evidentiary and methodological standards I have applied to every vaccine-chronic illness study since 2015. This is an important article that deserves your attention and social media shares. -JLW

When Senator Bill Cassidy voted to confirm Robert F. Kennedy Jr. as secretary of Health and Human Services, he did not simply exercise the Senate’s “advice and consent” power. He tried to co‑author the job description.

According to Cassidy’s own floor speech and subsequent reporting, he extracted a series of personal promises from Kennedy and treated those assurances as the price of his support: keep the “vaccines do not cause autism” language on the CDC website, preserve existing vaccine advisory structures, protect access to vaccines, seek Cassidy’s input on key hires, and maintain unusually tight coordination with the Senate health committee chair.

That bargain now lies in pieces. The outcome illustrates a deeper problem than one broken deal: a senator tried to hamstring a cabinet‑level health leader at the level of website copy and committee membership, while the country needs a department capable of leading a hard reset on how sick the United States has become.

Cassidy tried to write bad epidemiology into a confirmation vote. Kennedy treated the bargain as a speed bump. The American public ended up with a singular politically mandated distortion of the science presented as a factoid.

The Bargain That Turned Oversight into Co‑Management

Under the Constitution, the Senate holds a clear power: confirm or reject nominees after public scrutiny. It can legislate guardrails, fund or defund programs, and investigate misconduct. It does not hold a formal power to dictate specific sentences on executive‑branch websites or pre‑approve individual staffing decisions inside an agency.

Yet Cassidy, a physician and chair of the Senate Health, Education, Labor and Pensions Committee, publicly framed his vote for Kennedy as contingent on a list of operational concessions. He cited explicit commitments that:

CDC’s autism page would keep the categorical statement that vaccines do not cause autism. (Cassidy’s claim on the agreement, verbatim: “CDC will not remove statements on their website pointing out that vaccines do not cause autism”)
The HHS secretary would not create parallel vaccine safety structures outside existing systems. (Verbatim: “(Kennedy) would work within the current vaccine approval and safety monitoring systems, and not establish parallel systems.”
The secretary would work “unprecedently” (sic) closely with Cassidy himself and seek his input on senior hires. (Verbatim: “He and I will have an unprecedently (sic) close collaborative working relationship … We will meet or speak multiple times a month.” Kennedy had, in testimony, stated that he had already invited Cassidy to his office on numerous occasions, but that Cassidy had never shown up.

That strategy did not operate as neutral institutional design. Cassidy attempted to insert the legislative branch into executive decision streams at a granular level. He tried to convert the Senate’s gatekeeping authority into a continuing veto over phrases, personnel, and advisory bodies.

That move has two predictable properties.

First, it weakens formal law and strengthens informal deals. Handshake assurances leave no enforceable standard when a secretary reverses course.

Second, it defacto places one Senator in a non-existent supervisory position within the Executive Branch as Sec. Kennedy’s supervisor. Secretary Kennedy already has a supervisor provided by the U.S. Constitution. Cassidy’s wishful employment tempts him to overreach and now fight over symbolic language while children still remain at increased risk of neurodevelopmental disorders and autoimmunity based on policies created without science backing them.

Subscribe now

Kennedy Moves Ahead

Kennedy, as Secretary of Health and Human Services, then used the executive levers that Cassidy had tried to pre‑script, as one does.

In November, under Kennedy’s direction, CDC rewrote its “Autism and Vaccines” page. The site no longer leads with a clear summary of the extensive evidence base showing no causal link between childhood vaccination and autism. Instead, it now asserts that the previous statement “vaccines do not cause autism” lacks an evidence base because studies allegedly have not ruled out a possible contribution from infant vaccines, and it accuses health authorities of ignoring studies that claim a link.

CDC added a factual footnote: the original “Vaccines do not cause Autism” heading appears with an asterisk that exists only because of Kennedy’s earlier agreement with Cassidy not to remove that sentence

Kennedy also fired the existing members of the Advisory Committee on Immunization Practices and replaced them with a smaller group that includes outspoken vaccine critics, over Cassidy’s objections and calls to delay the panel’s work.

The result:

The executive branch finally uses the CDC masthead to broadcast facts on the insufficient, quality evidence from population‑based studies and meta‑analyses on what is arguably the most important medical question of our time. In doing so, the executive branch finally acknowledges the evidentiary gaps Congress keeps pretending do not exist.
The same senator who tried to control a sentence on a website has no effective recourse besides public complaining and social‑media outrage.

This sequence shows exactly why legislators should not expect that their “consent and advice” applies to granular specifics of complex topics like public health. A secretary determines that he would be negligent in his duties, the senator’s only remaining tools are political rant and, at the outer edge, impeachment proceedings. That would be a step no one seriously contemplates for a website rewrite, but proceedings that would not doubt air the facts on the sorry state of science on the question, including that not all vaccines in the childhood schedule have been tested for association with autism; that association does not test causality; and that, in the end, for the public good, science must trump politics.

What the Evidence on Vaccines and Autism Actually Shows

While mainstream reports parrot that large epidemiologic and meta‑analytic studies covering more than a million children have “tested” the vaccine–autism hypothesis from multiple angles, these reports fail to bring forward evidence of analytic malfeasance and potential fraud due to the re-analysis of data in ways designed to make positive findings supportive of causality disappear. But that’s been the pattern, over and over.

Large Vaccine–Autism Studies Under a Microscope

Mainstream reports regularly present a small canon of large epidemiologic and meta‑analytic studies—together covering well over a million children—as if they have definitively “tested” and closed the vaccine–autism question. Coverage tends to cite the bottom‑line conclusion (“no association”) while skating past the modeling choices, outcome definitions, censoring rules, and conflicts of interest that shape those conclusions. Critics have raised allegations of p‑hacking, selective reporting, and biased re‑analysis in this literature, but those allegations have not been confirmed by formal findings of fabrication or fraud. What can be documented, from the papers and from published critiques, are structural limitations and design features that systematically favor a null result.

The 2002 Danish MMR cohort (Madsen et al., 2002, NEJM; ~537,000 children)

Madsen et al. (2002) used Danish registries to follow 537,303 children born between 1991 and 1998 and reported “strong evidence” against an association between MMR and autism.

Key limitations and pressure points:

Shallow clinical validation of cases. Only 40 of 316 autism cases underwent detailed clinical record review. The rest relied on registry diagnoses, which letters in NEJM flagged as inadequate for a nuanced autism phenotype, especially for regressive subtypes.

Use of person ‑ years and differential follow ‑ up. The original analysis used person ‑ years rather than individuals as the unit of analysis. Critics noted that vaccinated children contributed less follow ‑ up time than unvaccinated children (about 3.7 vs 5.0 years), which can dilute case detection in the vaccinated group if not handled correctly. Madsen’s reply argued that a time ‑ dependent vaccination covariate fixed this, but the concern illustrates how analytic choices can push estimates toward the null.

Power claims that ignore small susceptible subgroups. The authors claimed 80% power to detect a relative risk of 1.5, but as Spitzer noted in NEJM correspondence, if only a modest subgroup (e.g., 10% of autistic children) had vaccine ‑ triggered autism with a true odds ratio near 4, the aggregate risk ratio would still cluster near 1.0 and appear null.

Diagnostic drift and calendar effects. The study spanned a period during which autism diagnoses increased sharply. Even with adjustments for age and calendar period, registry ‑ based work in a fast ‑ changing diagnostic environment risks misclassifying both timing and prevalence in ways that can hide time ‑ locked associations.

So while the Madsen study is large and methodologically conventional, it does not meaningfully address narrow risk strata, subtle phenotype differences, or misclassification that would preferentially bury a small but real effect.

The 2014 Vaccine meta ‑ analysis (Taylor et al., 2014, Vaccine; 5 cohort + 5 case ‑ control)

Taylor et al. (2014) pooled 5 cohort studies (1,256,407 children) and 5 case ‑ control studies (9,920 children), concluding that vaccination, MMR, thimerosal, and mercury were not associated with autism or ASD.

Methodological issues:

Meta ‑ analysis on top of observational designs. The paper aggregates studies that share the same core limitations: non ‑ random assignment to vaccination, reliance on administrative data, and imperfect control of confounders such as healthcare ‑ seeking behaviour and early developmental concerns. Pooling them does not eliminate those biases; it standardizes them.

Publication bias and study selection. The authors used only published studies and reported that no unpublished data were available. Critics have noted that the lack of completely unvaccinated comparison groups and the possibility of unpublished positive findings were not addressed in detail. While the authors list “Publication Bias” as a MeSH term, they do not fully interrogate the implications of selective publication on their summary conclusions.

Heterogeneity in exposures and outcomes. The meta ‑ analysis combines studies of MMR, thimerosal ‑ containing vaccines, total mercury dose, and “any vaccination,” across different countries and time periods, with varied diagnostic criteria for autism/ASD. That heterogeneity increases statistical noise and can weaken signals in specific exposure ‑ phenotype combinations.

Lack of individual ‑ level data. The authors work from published summary statistics, not raw data. That prevents independent exploration of alternative models, subgroup definitions, or exposure windows and makes it impossible to fully assess p ‑ hacking or unreported analytic paths in the underlying studies.

In short, Taylor et al. summarize the published negative literature efficiently, but the paper cannot rule out risks that sit in narrow subgroups, specific exposure combinations, or unpublished analyses.

The 2015 JAMA sibling cohort (Jain et al., 2015, JAMA; ~96,000 children, ~1,900 high ‑ risk)

Jain et al. (2015) assembled a cohort of 95,727 privately insured U.S. children with older siblings, including 1,929 children whose older sibling had ASD. They reported no harmful association between MMR receipt and ASD, even in the higher ‑ risk subgroup.

Important limitations:

Claims data rather than clinical records. Autism diagnoses and vaccination status both come from insurance claims. The authors acknowledge under ‑ reporting and misclassification: MMR coverage in their data lagged National Immunization Survey estimates by 4 ‑ 14 percentage points, and their ASD algorithm had an estimated positive predictive value of about 87%. Misclassifying vaccinated children as unvaccinated, in particular, biases estimates toward “no association.”

Restricted population and generalizability. The cohort consists of continuously enrolled, privately insured children, with race/ethnicity and income distributions that differ from the broader U.S. population. The authors explicitly state that the findings “may not be completely generalizable” to more socioeconomically and ethnically diverse populations with different care patterns.

Limited power in the highest ‑ risk subgroup. Only 1,929 younger siblings had an older sibling with ASD, and just 134 of them were diagnosed with ASD. The authors highlight that the upper bound of confidence intervals seldom exceeds 1.5, arguing that large effects are unlikely. That still leaves room for meaningful risk in specific sub ‑ phenotypes or exposure timings that the model does not isolate.

Industry ‑ linked data infrastructure. The data come from Optum, a UnitedHealth subsidiary; several authors are employees of Optum or The Lewin Group (an Optum company), although the primary stated funder is NIMH. The paper discloses these ties; disclosure is not proof of bias, but it does matter in a field where conflicts of interest can shape analytic decisions.

Post ‑ publication correction. JAMA issued a correction in 2016 after a variable was re ‑ defined. That sort of correction is not fraud, but it underscores how sensitive results can be to coding decisions that are usually invisible to the public.

Overall, Jain et al. reduce one set of concerns (recall bias, tiny samples) while inheriting others (healthy user bias, claims ‑ based misclassification, selection into private insurance, modest power in the subgroup that parents worry about most).

The 2019 Danish nationwide cohort (Hviid et al., 2019, Annals of Internal Medicine; ~657,000 children)

Hviid et al. (2019) tracked 657,461 children born from 1999–2010 and found no increased autism risk after MMR vaccination, including in predefined “susceptible” subgroups.

Registry ‑ only outcome ascertainment with no chart review. The authors list “no individual medical charts were reviewed” as the primary limitation. Registry diagnoses are valuable for scale but can miss subtler phenotypes, timing nuances, or misdiagnoses that matter when testing a specific mechanistic hypothesis.

Age and follow ‑ up window. Follow ‑ up extended through August 2013, yielding about 5 million person ‑ years. Children born in 2010 were only followed to age 3.6 on average, well before the peak diagnosis age (~7.2 years). While the authors used time-to-event analysis and censoring cutoffs, the possibility of under ‑ ascertainment remains for late-diagnosed cases.

Vaccine schedule and formulation issues. A 2025 critical review by Hammond and colleagues questions Hviid et al.’s exposure classification and timing assumptions, arguing that their handling of formulation changes and dose timing could obscure dose ‑ response patterns. These critiques are published but not yet answered in mainstream journals and should be considered provisional.

Under ‑ ascertainment of autism relative to other Danish data. National rates at the time and Schendel & Thorsteinsson (2018) estimate autism prevalence in similar cohorts at ~2%, whereas Hviid’s cohort shows ~0.99% diagnosed by mean age 8.6. Critics argue that the short follow-up for younger children accounts for this difference and may suppress later diagnoses.

Vaccination misclassification. Danish data audits (e.g., Holt et al.) show that national vaccination registries can misclassify vaccinated children as unvaccinated when doses are given outside billing systems. Such misclassification tends to pull relative risk estimates toward the null.

Institutional conflicts of interest. The study’s authors were based at Statens Serum Institut, which both conducts vaccine research and supplies vaccines to the Danish national program, with primary funding from the Novo Nordisk Foundation and the Danish Ministry of Health—entities with clear financial and reputational stakes in stable vaccine policy. The paper discloses these ties; critics argue that such embedded institutional roles warrant extra scrutiny of analytic choices.

Taken together, the Hviid study pushes statistical power higher than earlier cohorts and explicitly attempts to test “susceptible subgroup” hypotheses. At the same time, it rests on registry abstractions, incomplete follow ‑ up of younger cohorts, and data systems with known misclassification issues, and it has provoked detailed methodological challenges that remain unresolved in peer ‑ reviewed form.

What Survives After You Inspect the Methods

If you walk through these four studies with a rationalist’s red pen, two things remain true at once:

They do not support a large, population‑wide autism risk from MMR vaccination. Across hundreds of thousands of children in Denmark and tens of thousands in U.S. cohorts, crude and adjusted estimates repeatedly cluster near 1.0, not 3 or 4. That constrains any model in which routine MMR use is a dominant driver of the overall autism prevalence.
They leave real epistemic gaps. They are underpowered or structurally unsuited to rule out:
- Risk concentrated in narrowly defined biological or exposure phenotypes.
- Interactions between MMR and other vaccines, adjuvants, or environmental toxicants across sensitive developmental windows.
- Effects that manifest through diagnostic timing, regression patterns, or co‑morbid immune and GI phenotypes that registries code poorly.

Claims that these four papers “prove vaccines do not cause autism” overshoot what their methods can honestly deliver. Claims that they have been conclusively exposed as fraudulent overshoot in the opposite direction. The only scientifically defensible position is more demanding and less comfortable: recognize that these studies argue against large, average‑effect MMR‑autism models, while acknowledging that the design, execution, and incentive structures behind them still warrant serious, adversarial audit if HHS, CDC, or anyone else wants to make categorical statements in either direction.

When the CDC now claims that the statement “vaccines do not cause autism” lacks an evidence base, it does describe the actual state of the science. It does not describe the worldview of its current political supervisor.

That matters for executive‑legislative design. The job of HHS and CDC is not to mirror whatever narrative is desired to support a given policy. The job is to integrate the strongest available evidence and update guidance as that evidence changes. If the evidence falls short, HHS should say so.

The self-assigned burden now falls to the Secretary to design studies with sufficient power, appropriate controls, transparent methods, and credible peer review.

Cassidy’s Real Error: Micromanaging Instead of Legislating Standards

Cassidy clearly did not want this outcome. He now denounces the new CDC language, repeats that childhood vaccines for diseases such as measles and polio do not cause autism, and warns that the revised site will scare parents away from he claims are effective tools- while headlines on mumps outbreaks in fully vaccinated schools and studies on asymptomatic transmission of measles say different.

He could have pursued a different strategy from the start.

Instead of tying his support for Kennedy to line‑item pledges, Cassidy could have used the confirmation window to advance statutory and structural constraints that would bind any secretary:

Evidence thresholds written into law for when federal health agencies may reverse major safety claims. (Kennedy would have gladly seen this applied).
Transparency requirements for any official assertion, including a formal docket that lists the studies, effect sizes, and methodological critiques.
Clear insulation of CDC’s scientific content from direct political editing by Pharma influenced legislators, similar in spirit to norms that protect the Federal Reserve’s technical assessments from day‑to‑day White House messaging.

Congress holds the power to legislate such standards. It holds the purse strings. It can demand that HHS or CDC meet specified evidence thresholds or lose authority over specific domains. Regardless, now that Kennedy has pledged to provide real evidence on the question, future HHS leadership will have a high bar to pass.

Cassidy did not do that. He chose soft leverage—an agreement about phrases, hiring input, and informal consultation—because soft leverage trades better in short‑cycle politics. That choice guaranteed maximum exposure when Kennedy decided to govern according to his long‑standing commitment to bring objective, strong-inference science to questions of vaccines rather than Cassidy’s pre-determined conclusions.

Kennedy Brings What the Public Actually Wants from HHS: Health, Not Semantic Trench Warfare

While the branches fight over a sentence on a website, the underlying health profile of the United States remains brutal.

CDC estimates that by 2023 about 76% of U.S. adults had at least one of a set of major chronic conditions, and about 51% had multiple chronic conditions.
Obesity prevalence in adults reached roughly 40% in the 2021–2023 window, with at least one in five adults living with obesity in every state and at least one in three adults living with obesity in twenty‑plus states.
Chronic and mental health conditions drive about 90% of the nation’s nearly $5 trillion annual health spending.
Life expectancy has just clawed back to 78.4 years after a steep pandemic‑era drop, yet still trails peer nations by roughly four years, with high burdens from cardiovascular disease, metabolic illness, overdose, and violence.

Kennedy sees this. The American public can see this. Families live inside this reality every day in ways that dwarf any website headline. They do not ask Sen. Cassidy or HHS for another round of semantic warfare over a single sentence. They ask—directly or indirectly—for leadership that affects their probability of living a long, functional life:

Food systems that reduce ultra‑processed intake and improve metabolic resilience.
Environments that allow physical activity rather than fight it.
Air, water, and chemical regulation that reduces toxic exposures rather than externalizes them.
Behavioral health infrastructure that answers overdose, depression, and anxiety with something stronger than slogans.
Safer, cleaner medicines.

Nothing in Cassidy’s issuances on matter of vaccines facilitated progress these domains. His leverage focused entirely on vaccine access, vaccine messaging, and vaccine advisory personnel. His attempts are to keep narrativ enforcers in place.

That narrow frame tracks the long‑running D.C. habit of treating “public health” as a synonym for vaccination campaigns, with chronic disease burden handled as an actuarial problem for Medicare and private insurers to absorb. Kennedy sees the full picture.

Executive vs. Legislative Power

Kennedy now uses executive power aggressively. He directs CDC to align with his the exact epistemic state of knowledge about vaccines and autism. He reshapes advisory structures to ensure that critics of vaccination set the bar of evidence. He cancels certain mRNA development funds and rewrites COVID‑19 vaccination recommendations with a sharp contraction in eligible age groups per the new, and appropriate, risk-stratification model.

Cassidy and his colleagues hold legislative tools they could use to respond, if they dare think it appropriate:

They can hold evidentiary hearings and put methodological scrutiny under oath.
They can amend enabling statutes that govern CDC’s scientific independence.
They can create an explicit firewall between political appointees and specific analytic outputs.
They can tie appropriations to adherence with evidence hierarchies rather than to compliance with one senator’s preferred phrase.

Instead, Cassidy treated confirmation like a private contract and now treats Kennedy’s progress somehow as a communications scandal.

The net effect:

The legislative branch acts as partisan narrator of disputed science on singular issue. The executive branch is enforcing the standards of evidence that should have been in play all along. The legislative branch performs outrage theater over a text change that it tried to pre‑write, instead of agreeing that HHS develop and employ durable standards that would protect scientific integrity regardless of which party holds the White House.

Popular Rationalism: What a Serious Health Platform Would Actually Require

A rationalist health platform—legislative or executive—has to begin by confronting evidence honestly rather than laundering conclusions through institutional habit. That means rejecting categorical claims that outrun the data and rejecting political edits that substitute narrative for analysis.

First, acknowledge what the major vaccine–autism studies do and do not show. Large registries, insurance-based cohorts, and meta-analyses consistently report null average effects, but those results depend on design choices that—study by study—cluster around misclassification, incomplete follow-up, population restrictions, registry-only diagnoses, and institutional conflicts of interest. They argue against large, population-wide effects; they do not close the question for susceptible subgroups, specific exposure windows, synergistic toxicants, or regressive phenotypes that registries poorly capture. A rational platform insists on methodological integrity and refuses both absolutist slogans and absolutist dismissals.

Second, treat the vaccine–autism literature as only one slice of a much larger health-failure landscape. The United States is a country where roughly 40% of adults live with obesity, where over three-quarters carry at least one chronic condition, and where metabolic, environmental, nutritional, psychiatric, and toxicological burdens interact in ways that dwarf any single intervention. The public health catastrophe is systemic—not a marketing problem, not a messaging problem, and not solvable by asserting or denying any single causal pathway.

Third, build institutions in which evidence—not personalities, not bargains, not senatorial hostage notes—drives guidance and policy. That requires:

CDC to publish analytic positions grounded in cumulative evidence with transparent methods, explicit assumptions, and audit-ready models.
HHS leadership to set strategic direction, fund proper research, and establish quality thresholds—not rewrite scientific claims by swapping website sentences.
Congress to legislate process, disclosure, and evidentiary thresholds instead of micromanaging individual phrases or relying on handshake concessions during confirmation votes.

In that architecture, a health secretary who wants to revise established guidance must do so by producing or funding credible, replicable research, not by retrofitting a federal website to their priors. And a senator concerned with public welfare must build enforceable statutory standards, not fragile private deals that collapse on first contact with executive authority.

Kennedy has decided to do all of this with his position not because of an Act of Congress – but because it is the right thing to do.

Is Cassidy’s Platform, in Practice, “Keep America Sick”?

Cassidy now presents himself as the defender of vaccine policy against a rogue health secretary who dared represent the truth about the science. That story holds a narrow reality; Kennedy’s positions on childhood vaccines line up the evidence base, and Cassidy’s wished-for policy does not.

Yet the deeper record looks harsher.

Cassidy accepted Kennedy’s nomination after securing promises that any ethical Secretary should break. He concentrated on preserving a website sentence and an advisory committee configuration rather than demanding a health agenda that attacked the chronic‑disease tsunami that shortens and worsens American lives. He used legislative power to micromanage a future partner rather than to build the scaffolding of a healthier country.

Kennedy then used executive power to report on the actual state of the science, and then promote better science, leading a broad health renewal.

Cassidy’s failure hides in what he did not even try to do. He did not treat this appointment as a chance to force a serious reorientation of HHS toward metabolic health, environmental exposures, and structural drivers of disease. He treated it as a chance to defend a narrow piece of status quo orthodoxy and then watched that orthodoxy evaporate at the first contact with a determined secretary.

When a senator uses the Senate’s constitutional leverage to fight over wording instead of health outcomes, when chronic disease continues to expand under a status quo that he largely accepts, one question becomes unavoidable:

If this is the shape of his health strategy in practice—symbolic fights over copy, no structural attack on the drivers of sickness—does Bill Cassidy’s platform now function, in effect, as “Keep America Sick”?

Thanks for reading Popular Rationalism! This post is public so feel free to share it.