A new nationwide study from Denmark found no evidence that aluminum salts in childhood vaccines are linked to autism, attention-deficit/hyperactivity disorder (ADHD), asthma, or 47 other chronic conditions. Robert F. Kennedy Jr., a long-time vaccine critic who now heads the United States Department of Health and Human Services, says the work is riddled with bias and should be retracted. A close look at the research methods and the points he raises shows that some of his concerns touch on real limitations, but the bulk of his critique misreads how large registry studies work and overstates what the data actually show.
The Danish team—led by epidemiologist Niklas Worm Andersson with senior researcher Anders Hviid and colleagues at Statens Serum Institut and the University of Copenhagen—set out to test a question that has lingered for decades: do aluminum-adjuvanted shots given in infancy leave children more prone to autoimmune disease, allergies, or developmental disorders later on?
Aluminum salts have been used since the 1920s to boost immune responses in non-live vaccines, yet animal experiments and small case series have kept worries alive that they might carry long-term risks. Denmark proved a natural laboratory because its vaccination program changed several times between 1997 and 2020, increasing or decreasing the aluminum content of standard shots. That policy churn produced natural variation in exposure that was unrelated to parents’ choices, giving researchers a quasi-experimental setting to probe any health effects.
Using nationwide birth, vaccination, hospital, and prescription registers, the investigators assembled a cohort of 1,224,176 children born between 1997 and 2018 who were alive and living in Denmark at their second birthday. Each child’s total aluminum dose from vaccines received before age two ranged from 0 to 4.5 milligrams.
The researchers then analyzed medical records from each child’s second birthday through age five—extending to age eight in additional analyses—to identify the first diagnosis of any of 50 predefined conditions: 36 autoimmune diseases, nine allergic or atopic disorders, and five neurodevelopmental outcomes, including autism spectrum disorders and ADHD.
They used statistical models (Cox models) to estimate how each extra milligram of aluminum from vaccines was associated with the risk of each condition, while adjusting for factors like birth year, season, sex, maternal age, smoking during pregnancy, preterm birth, birthweight, household income, and other variables that might skew the results.
Across all children, each additional milligram of aluminum from vaccines was linked to a hazard ratio of 0.98 for any autoimmune condition, 0.99 for any allergic or atopic disorder, and 0.93 for any neurodevelopmental diagnosis.
In plain terms, there was no sign of increased risk, and for neurodevelopmental outcomes, the rate was slightly lower in children with higher aluminum exposure. Asthma—the most common condition in the study, with more than 28,000 cases—also showed a small decrease in risk, with a hazard ratio of 0.96. The researchers attributed this reduced risk to random variation or residual bias.
Kennedy’s criticism centers on four pillars. First, he argues that the researchers “designed the study not to find harm” by excluding children who died before age two, children with serious early-life illnesses, and about 2.8 percent whose vaccination records indicated unusually high aluminum exposure.
Excluding infants who died young is standard practice in studies like this, since these children often do not complete the vaccination schedule and may not live long enough to receive a diagnosis of a chronic condition. Similarly, children with severe congenital or preexisting conditions—such as immune deficiencies, heart failure, or liver failure—were excluded because these rare disorders can independently affect both vaccine timing and health outcomes, introducing strong confounding.
As for the children with implausibly high aluminum records, the study authors note that these entries were likely data errors—such as duplicate records or coding anomalies—rather than true cases of extreme exposure. Removing these cases reduces the risk of misclassifying exposure levels, which could distort the results more than clarify them.
Second, Kennedy argues that adjusting for the number of general practitioner visits during the first two years “creates collider bias,” suggesting that frequent doctor visits could reflect early symptoms of vaccine injury. But in registry studies, the number of healthcare visits plays a major role in whether a diagnosis gets recorded at all. Children who see doctors more often are more likely to have conditions identified, regardless of cause. If researchers didn’t adjust for this, it could look like certain exposures cause more disease when in fact they just lead to more medical contact—and more opportunities for diagnosis.
Third, he faults the authors for not conducting a simple vaccinated versus unvaccinated comparison. The study’s primary analysis, however, used a more robust dose-response model, treating aluminum exposure as a continuous variable. This design is less susceptible to the profound confounding that exists between vaccinated and completely unvaccinated families, who often differ in many health-seeking behaviors. While the unvaccinated children were used as a reference group in some secondary analyses, focusing the main analysis on variation among vaccinated children minimizes these biases.
Fourth, Kennedy points to a figure buried in the supplement: among children born between 2007 and 2018, the hazard ratio for Asperger syndrome per milligram of aluminum was 1.67. On the surface, that sounds alarming. But the estimate is based on just 51 cases in that subgroup, and the confidence interval—ranging from 1.01 to 2.77—is wide and barely clears the threshold for statistical significance. A small shift in the number of diagnoses would erase the signal entirely.
The pattern didn’t appear in the older birth group, wasn’t present in the full cohort, and disappeared when the researchers extended follow-up to age eight. In large studies that test many outcomes and subgroups, a few results will cross the significance threshold by chance alone. Kennedy’s selective focus on one small subgroup is a classic example of the “multiple comparisons problem,” where running dozens of analyses is expected to produce some chance findings.
Finally, he points to institutional conflicts: three authors work at Statens Serum Institut, which produces some vaccines used in Denmark. The institute is state owned but sells serum products worldwide, so the concern is not trivial. The journal required disclosure, and the affiliations are printed on the paper. Whether those ties invalidate the analysis is a matter of judgment. Independent replication in another setting would help address any lingering doubt.
Taken together, the study has clear limits. It is observational, not randomized. It covers Danish exposure levels, which can be lower than those in the United States, primarily because Denmark only recommends the aluminum-containing Hepatitis B vaccine to at-risk infants, whereas the U.S. recommends it universally starting on the first day of life.
It relies on registry diagnoses. While this may miss the mildest cases managed exclusively by a general practitioner without a specialist referral or prescription, the use of the nationwide prescription registry for conditions like asthma and ADHD captures a vast number of cases that never result in a hospital visit, making the data far more comprehensive than hospital records alone.
The study also cannot probe rare genetic susceptibilities or interactions between aluminum and other environmental toxins. Still, the data set is enormous, the confounder control is thorough, and the natural experiment created by policy shifts offers a strong test of population-level risk. Most of Kennedy’s objections either misinterpret standard epidemiological practice or single out a single noisy subgroup result without weighing the broader picture.
What did the researchers really discover? At the population level in Denmark, where most children received between two and four milligrams of aluminum from vaccines in the first two years of life, there was no detectable rise in autoimmune disease, atopic illness, autism spectrum disorder, or ADHD. The narrow confidence bands around many estimates rule out even modest risk increases. If aluminum adjuvants carried a hazard large enough to explain the dramatic rise in autism diagnoses seen in many countries, it would almost certainly have appeared in a cohort of more than a million children.
Kennedy is right that science needs transparency. Danish privacy rules restrict public release of individual health records, and independent teams will need special approvals to reanalyze the raw data. Yet calls for retraction over alleged “fatal flaws” lack support in the methods or results. The study adds one of the largest pieces of evidence to date that standard infant vaccines, at the aluminum doses commonly given in Denmark, do not raise the risk of the chronic disorders parents most fear.
The Danish study is not the final word on vaccine safety, but it offers one of the most comprehensive tests yet of a widely circulated concern. While no study is perfect, the evidence here strongly suggests that aluminum in early childhood vaccines, at the levels given in Denmark, is not driving a hidden epidemic of chronic illness.
The study, “Aluminum-Adsorbed Vaccines and Chronic Diseases in Childhood: A Nationwide Cohort Study,” was authored by Niklas Worm Andersson, Ingrid Bech Svalgaard, Stine Skovbo Hoffmann, and Anders Hviid.
