tl;dr: When browsing online, don’t judge based on if you agree or if it sounds “true.” Ask how much the author (influencer) would benefit from making the headline more dramatic than the evidence warrants.
Nearly everyone is motivated to get clicks at any cost. That is not a path to truth.
Here is one example from science “journalism” (but it applies to everything, including financial “advice” and especially politics):
Long-term melatonin use linked to 90% higher heart failure risk
Q’s analysis:
The first thing that
jumped out at me is that ScienceDaily is presenting this on August 29, 2026
as though it were fresh. It isn’t. This is a repackaging of an American
Heart Association conference abstract from November 2025. As of now, I can
find the conference abstract and commentary about it, but not a full
peer-reviewed paper reporting this analysis. The AHA itself explicitly says
conference abstracts are not peer reviewed and should be considered
preliminary.
The headline number is
scary: “90% higher heart failure risk.” But the actual result was 4.6%
versus 2.7% over five years, an absolute difference of 1.9 percentage
points. That's worth investigating, certainly. But before deciding melatonin
caused even part of that difference, there are some enormous problems.
The biggest problem is
exposure classification. “Melatonin user” meant melatonin appeared in the medical
record. “Nonuser” meant it didn't. But in the US, people buy melatonin at CVS,
Costco, Amazon, etc., without telling their doctor. The authors didn't even
know which countries individual patients came from. So some unknown fraction of
their supposedly melatonin-free controls were almost certainly taking
melatonin. The authors themselves acknowledge this.
More importantly, I
think there is a massive confounding-by-indication problem here. Who
gets melatonin formally entered in their medical chart and keeps taking it for
a year? Probably not a random sample of people with mild occasional trouble
sleeping. They're plausibly people with worse or more persistent insomnia,
more medical contact, more psychiatric illness, more complicated health
problems, or combinations thereof. Insomnia itself is associated with
cardiovascular risk. The investigators did propensity-match on 40 measured
variables, which is better than a crude observational comparison, but they did
not know insomnia severity or psychiatric diagnoses sufficiently well. Even
the study authors explicitly identify worse insomnia, depression, and anxiety
as possible explanations.
And here's where my
eyebrow goes way up: heart-failure hospitalization was supposedly 19.0%
versus 6.6%, even though new heart-failure diagnosis was only 4.6%
versus 2.7%. That's internally weird enough that the authors had to explain
that the hospitalization endpoint captured a bunch of related billing codes
rather than clean incident-heart-failure events. I would put very little weight
on that spectacular “3.5× hospitalization” number.
Likewise, all-cause
mortality was 7.8% versus 4.3%. If melatonin were genuinely causing
something close to a doubling of five-year mortality in middle-aged people,
that would be an absolutely gigantic pharmacologic effect. Given decades of
widespread melatonin exposure, clinical trials, pharmacovigilance, and
observational research, I'd expect some corroborating signal elsewhere.
Instead, a 2026 Journal of Pineal Research commentary specifically
criticizing this work points out that its result sits awkwardly against the
broader melatonin safety literature.
There is also a
particularly relevant piece of prior evidence: a 2024 prospective analysis
of three US cohorts examined melatonin supplement use and cardiovascular
disease/diabetes. That's a very different design and doesn't settle the
question, but it is part of why I wouldn't update dramatically from one TriNetX
database analysis.
So my rough Bayesian
translation would be:
Before this study: I wouldn't have assumed
chronic melatonin is perfectly harmless, because we really don't have superb
decades-long randomized safety evidence.
After this study: I'd move slightly
toward “worth studying; perhaps don't take it indefinitely for no reason.”
I would not move
anywhere remotely close to “melatonin raises heart-failure risk by 90%.”
That causal interpretation is unsupported.
And specifically for
you, I wouldn't change anything on the basis of this article. Your diary
mentions melatonin as an occasional part of various sleep experiments, rather
than anything resembling the study's “documented chronic melatonin treatment
for ≥365 days” exposure. Your recent cardiac stress test was also reassuring. [After 1+ year wearing the Google Watch, Google Health claims my VO2 Max is 55.9, at age 58.*] So this isn't setting off a personal alarm bell for me.
My bullshit-meter score: perhaps 3/10 for the
underlying research question, 7/10 for the impression created by the headline.
The study found a real correlation in a huge database. The leap from that to
“long-term melatonin use linked to 90% higher heart failure risk” is
mathematically true but medically much more dramatic than the evidence
warrants.
And the funniest/most irritating part: ScienceDaily has resurrected a ten-month-old, non-peer-reviewed conference abstract as “news.” Your New Scientist klaxon may remain armed. 😜
*
 |
| Not bad for someone w/o two full lungs. |
No comments:
Post a Comment