Hydrogen water

[Grumble]

I’ve just come across this on Facebook, home of reliable medical information. Not going to link to the post, but at least they gave a link to a paper on pubmed. I’m going to leave it here, and maybe I will try going through it when I get a little time, but if anyone else wants to have a crack then feel free. It might be perfectly valid, and we should all be drinking hydrogen enriched water, but it feels red-flaggy to me.

https://pmc.ncbi.nlm.nih.gov/articles/PMC9967957/

[IvanV]

This purports to be a meta-study, not an experiment in itself. I observe that all but one of the authors come from the sports departments at their respective universities.

Our findings indicate that drinking HRW can significantly improve lipid status in the clinical populations.

You might think, reading that, that "significantly" means "by a substantial amount". That's what people usually mean by that wording. That is not the sentence you write if you mean that a result is statistically significant, which is what they actually mean. In fact the effect they report is not a "significant[] improvement" at all in the usual understanding of that. The average size effect is small. Moreover there is great diversity in the size effects across the studies, indeed even the sign (plus or minus) is pretty random. What they actually should have written is that the average size effect is small, but its difference from zero is statistically significant. This is a common feature of statistical studies attempting to show that woo is true - you get an effect which is small and statistically different from zero. It is a fair description of a result which is not worth the paper it is written on.

I was entertained by these sentences.

In terms of intake, all participants in the studies ingested HRW.
...
The PEDro scale mean score for the included studies was 9.28, considered as excellent quality.

In other words, none of the studies they looked were RCTs, none had control groups, and none were blinded in any way. I don't know what this PEDro scale for study quality is. But a scale which rates experimental studies of a treatment with these well known design deficiencies as very high quality is evidently not an appropriate metric for this type of study. There should be no ethical problems with proper RCT methodology here.

[science_fox]

I only skimmed it. But didn't see any validation of the H2 content or source. https://encyclopedia.pub/entry/55637#:~ ... %E2%88%923. seems useful. And quite a rebutal of the entire concept. Anything you've done to increase the H2 content in water (and it's barely soluble at best) will very quickly de-gas back out again.

No idea how you'd measure the uptake from H2 into cells before it is lost.

[Grumble]

I only skimmed it. But didn't see any validation of the H2 content or source. https://encyclopedia.pub/entry/55637#:~ ... %E2%88%923. seems useful. And quite a rebutal of the entire concept. Anything you've done to increase the H2 content in water (and it's barely soluble at best) will very quickly de-gas back out again.

No idea how you'd measure the uptake from H2 into cells before it is lost.

In my labs we dissolve hydrogen into water regularly, but we have to give it a good purging and then hold an overpressure. I think we typically aim for about 30 cc/kg (which is a not really SI unit)

[Sciolus]

You might think, reading that, that "significantly" means "by a substantial amount". That's what people usually mean by that wording. That is not the sentence you write if you mean that a result is statistically significant, which is what they actually mean. In fact the effect they report is not a "significant[] improvement" at all in the usual understanding of that. The average size effect is small. Moreover there is great diversity in the size effects across the studies, indeed even the sign (plus or minus) is pretty random. What they actually should have written is that the average size effect is small, but its difference from zero is statistically significant. This is a common feature of statistical studies attempting to show that woo is true - you get an effect which is small and statistically different from zero. It is a fair description of a result which is not worth the paper it is written on.

[rant]
When I rule the world, I will impose the death penalty on anyone who uses the word significant to describe the results of a statistical test. It causes confusion and isn't necessary. Just invent a new word. Here you go: dibulent. Done it for you. Now use it.

There was journal I used to read sometimes, I can't remember what and I've never seen it anywhere else, whose house style was that when a word with a vernacular meaning was being used in a statistical-jargon sense, it was prefixed by s-, thus s-significant, s-expected and so on. This is a tolerable stopgap for people who are too lazy to write statistically significant every time but too scared to use dibulent, even though it's clearly a better term.

It is shameful for science that this unnecessary cause of misunderstanding persists.
[/rant]