Yesterday, U.S. Secretary of Health and Human Services Robert F. Kennedy, Jr. published a sharp and comprehensive critique of the recent study by Andersson et al., which was published in the Annals of Internal Medicine. The study made headlines for claiming that aluminum-adjuvanted vaccines administered in early childhood are not associated with increased risks of autoimmune, allergic, or neurodevelopmental disorders.

Kennedy did not mince words. He described the study as “so deeply flawed it functions not as science but as a deceitful propaganda stunt by the pharmaceutical industry.” Among the many questionable features he identified, one stood out to me in particular. Kennedy wrote:

“These sleights of hand magnify the potential for allowing the authors to reach their absurd suggestion that higher aluminum exposure is somehow protective against asthma, allergies, and neurodevelopmental disorders, including autism.”

This sentence stopped me in my tracks because I had noticed the exact same thing. While Kennedy voiced this concern from the standpoint of public health advocacy, I approached the same issue from an academic and data-driven perspective. What I found not only aligns with his observation but adds further empirical grounding to it. In fact, this very point was at the heart of a formal comment I submitted to the Annals of Internal Medicine. The authors of the study responded — but, in my view, did not adequately address the core contradiction. In this short article, I lay out the full story, supported by the data, to show why this implausible pattern of protective effects cannot be ignored.

A Pattern Too Good to Be True

In their central figure (see the screenshot below), Andersson et al. report hazard ratios for 34 different health outcomes, comparing children with different levels of aluminum exposure through vaccination. At first glance, the figure seems balanced and comprehensive. But a closer look reveals a startling trend: 25 of the 34 estimates (73.5%) leaned in the same direction—suggesting that more aluminum exposure was associated with lower risk. And not merely by random chance: over half of these “protective” associations were statistically significant (with 95% confidence intervals excluding 1.0). Strikingly, children who received higher doses of aluminum appeared to have lower risks of developing conditions such as food allergies, autism, and ADHD. …

More importantly, the authors do not address the central anomaly: the consistent and statistically significant protective associations observed across a wide range of outcomes (see the screenshot above). This is not a case of random noise or a few spurious findings; it is a systematic pattern pointing in an implausible direction.

A Pattern Too Perfect to Trust

So, where does that leave us? If healthy vaccinee bias isn’t responsible for these strange results, then we are left with a far more troubling possibility: that the dataset itself has been compromised (even if unintentionally), whether through unrecognized flaws, distortions, or structural artifacts.

In this light, healthy vaccinee bias becomes the least concerning explanation. It offers a familiar, unintentional source of error. But rejecting it leaves us confronting the specter of contaminated science—exactly what Secretary Kennedy warned about. His critique, like mine, is not about being “pro-” or “anti-” vaccine. It is about holding science to its own standards. And when findings seem too good to be true, we owe it to the public to ask whether they are not just unlikely, but actually misleading.