On Tuesday, Aaron Siri, personal lawyer and close adviser to Robert F. Kennedy Jr., presented his “smoking gun” at a Senate Permanent Subcommittee on Investigations hearing on vaccine science. Siri, who has represented Kennedy in multiple lawsuits against federal health agencies and reportedly helped interview candidates for Department of Health and Human Services positionsunveiled a study riddled with the exact flaws that peer review is designed to catch: fundamental study design errors, statistical impossibilities inconsistent with known prevalence, and results that collapse under routine epidemiologic scrutiny. Notably, even this study’s own data showed no association between vaccines and autism, the condition most frequently cited by vaccine critics.
The study, known as the Henry Ford Health system analysis, was completed years ago and remains unpublished. It purports to show that vaccinated children have dramatically higher rates of chronic conditions than unvaccinated children. That it remains unpublished isn’t suppression, as Siri alleged, but rather quality control. The bitter irony of this hearing’s title — “How the Corruption of Science Has Impacted Public Perception” — is that, as I testified to the subcommittee, the real corruption on display isn’t in journals rejecting flawed studies. It’s in bypassing peer review entirely, shopping for any analysis that supports predetermined conclusions, then presenting it as evidence on a Senate stage. The analysis contains fundamental mistakes that any credible journal would flag.
The most glaring problem is detection bias, which occurs when one group gets examined more frequently than another, leading to more diagnoses regardless of actual disease rates. In the Henry Ford data, vaccinated children had substantially more health care visits than unvaccinated children. Conditions requiring clinical evaluation to diagnose — ADHD, learning disorders, speech delays, ear infections — will inevitably be recorded more often in the frequently seen group. Yet the authors never correct for this gap. Their only check was to drop children who never had a single encounter with a health care provider, which still leaves one group averaging seven visits a year and the other averaging two. That doesn’t level the playing field; it simply bakes the bias into the results. What they’re measuring is exposure to medical observation, not the effects of vaccines.
Take the claim of a six- to eight-fold increase in ear infections among vaccinated children. This is medically implausible but perfectly explained by detection bias. A child who rarely sees a clinician won’t have “otitis media” coded in their record, even if they’ve had ear pain. While untreated ear infections can resolve on their own, they subject children to unnecessary pain, potential hearing damage, and risk of complications like mastoiditis or meningitis. The study repeatedly conflates absence of diagnosis with absence of disease.
The statistical red flags accumulate. The authors report near-zero cases of common conditions like ADHD and learning disabilities among thousands of unvaccinated children. Data show these conditions affect roughly 11% and 9% of children, respectively. Finding essentially none suggests these conditions went undiagnosed and unrecorded in children who rarely saw doctors. In several categories, hazard ratios can’t even be calculated because all cases occur in the vaccinated group — exactly what happens when diagnoses are missed in the comparison group.
Real vaccine adverse events do exist, and surveillance systems successfully detect them: anaphylaxis at roughly 1.3 per million doses and febrile seizures after MMR vaccine at approximately 333 per million doses. A system that can detect signals at these magnitudes while filtering out spurious ones is working as intended.
Properly conducted vaccinated-versus-unvaccinated comparisons look very different from the Henry Ford analysis. A 2014 meta-analysis of more than 1.25 million children published in Vaccine found no link between vaccines and autism. Denmark’s nationwide registry studies found no association with Type 1 diabetes or autism. A study published in the Journal of Pediatrics confirmed these findings. Germany’s nationally representative KiGGS study of 13,453 children, including 94 completely unvaccinated, found no meaningful differences in chronic conditions when appropriate methods were used.
If one accepts the Henry Ford paper’s methodology, you’d also have to accept its finding of no link to autism. You can’t declare a study definitive when it seems to support your views and suddenly invalid where it doesn’t. That would be selective interpretation, not science.
Elevating an unpublished, methodologically flawed analysis on a Senate stage has real consequences. The irony is striking: witnesses claiming that vaccine science has been corrupted present as evidence a study with detection bias so fundamental that any epidemiology student would catch it. If the scientific establishment truly were suppressing data, they’d presumably choose something more sophisticated than failing to publish work that confuses health care visits with actual illness. This undermines evidence-based policymaking, confuses parents trying to make informed decisions, and distracts from work that actually improves vaccine safety: investing in surveillance systems, funding research to identify those at higher risk of adverse events, and communicating transparently about trade-offs. When Congress elevates unpublished analyses while dismissing peer-reviewed evidence, it sets a dangerous precedent for how science is weighed in policymaking.
Parents deserve better than political theater. They deserve a system that acknowledges real risks, quantifies them honestly, and updates guidance as data accumulate. They deserve policymakers who can distinguish between methodologically sound analyses and those with fatal flaws. And they deserve to know that the overwhelming body of high-quality research, using proper controls and accounting for health care utilization, finds no increased risk of chronic conditions from routine childhood vaccination, even as we continue studying safety and improving our systems.
When public health policy is shaped by studies that can’t distinguish health care utilization from health outcomes, we all lose. The science on vaccine safety remains imperfect and incomplete, but substituting unpublished analyses with fundamental errors for peer-reviewed evidence won’t make children safer. It will only make parents less certain about whom to trust.
Jake Scott, M.D., is an infectious diseases physician and clinical associate professor at Stanford University School of Medicine and curator of the Controlled Vaccine Trial Database. The views expressed are his own and do not necessarily reflect those of Stanford University or Stanford Medicine.