The paper says: "The adjusted population attributable fraction for all-cause mortality was 18% (95% CI 10·9–26·1), equivalent to 412 000 (95% CI 250 000–598 000) deaths each year."
My opinion is still that translates better to "kills 412k per year" than "might kill up to 412k per year."
I could see "might kill up to 598k per year" or "kills at least 250k per year" from their confidence interval.
Why do you think it should say "might" and "up to"?
The site guidelines ask you not to editorialize in titles: https://news.ycombinator.com/newsguidelines.html. Cherry-picking one detail and putting it in the title is not just editorializing, it's the leading form of it.
If you want to say what you think is important about an article, please do so in a comment. Then your view is on a level playing field with everyone else's.
These are associations, not causal implications. "Lead kills" is a causal statement, and even "lead associated with X deaths" is a bit leading, and I think that the authors were careful to not say that directly.
Given the experimental evidence mentioned in the introduction of the paper, it's likely that some or many of these deaths could be due to lead, but they could also be due to shared causes. There could be other things that lead to both greater lead exposure and greater cardiac death, for example.
One has to be very very careful with drawing conclusions from surveys such as this one. The standard in epidemiology is usually, association + one potential mechanism, but even with that it doesn't mean that the one potential mechanism is an explanation for all the observed association.
Keeping that type of phrasing, saying something like "Lead might kill up to 412k/year" would be in the realm of reasonable.
This type of headline manipulation is how the media turns good science into bad science.