Safe for whom? At what rate? Compared to what? For a substance injected into virtually every infant, the default is no until safety is demonstrated. Zero placebo-controlled trials in children. The two in adults both found harm. Three animal studies found brain aluminium where the models said it would not be. The most comprehensive review does not cite any of them.
When injected aluminium adjuvant forms granulomas, the aluminium stays put. When it doesn't, macrophages carry it to the brain. A controlled experiment in mice found 50 times more brain aluminium at the lowest dose than at higher doses. The question is not whether more is worse than less, but how much escapes the injection site.
No safety signal in surveillance, so no new studies needed. No study exists, so no signal found. Each step is rational. Together they form a loop that cannot generate the evidence it would need to close the question.
The strongest defence of thin foundational evidence is post-market surveillance. But the systems catch acute, distinctive, temporally clustered signals. They were not built to detect what the foundational studies could not rule out.
The global regulatory case for aluminium adjuvant safety traces to a pharmacokinetic model built on one study of 2 rabbits per adjuvant type, observed for 28 days, with lost bone samples and one lost brain. The three brains they did measure all contained aluminium. That study has never been replicated.
A 2022 systematic review found 102 trials studying aluminium adjuvants. Only 2 compared aluminium alone against an inert placebo. Together they enrolled 84 subjects.
The phrase "no evidence of harm" can mean four very different things. Only one of them is reassuring.
If you need to vaccinate 64,000 people to prevent one death, but your trial only has 341 people, "no evidence of harm" means almost nothing.
Take nobody's word for it. This blog is about reading vaccine safety evidence yourself.