Hacker News Reader

Comment by Yen | Hacker News Reader

This is interesting, as before now, I considered most claims of gluten sensitivity (that weren't specifically related to Celiac disease) to be part of a health fad. This study would seem to contradict that.

Are these P numbers actually a large effect, though? This doesn't seem to have freely available full text.

View on HN

This is the most important question in this thread (though it was worded confusingly). There was a statistically discernible difference, but what is the magnitude of that difference?

The reddit thread[0] on this paper has a lot of nonsense, but also has people with actual access to the paper. One informative comment (keeping in mind this study was done only on people who already identified as gluten sensitive):

> 31 of the individuals had an equal response to gluten or placebo, a further 13 had no significant difference. 5 individuals had stronger negative effects to placebo, 9 had stronger effects to Gluten. The three mentioned were over 2 standard deviations above the mean, and hence considered to be sensitive.

> There did appear to be a general increase across the board for symptoms, however the one caveat that was reported in results, but not discussed was the fact that individuals report worse symptoms (p=0.009) in the first week of the trial compared to the second.

> There were also no individual symptoms that showed a significant difference between the placebo and gluten. Only when all symptoms were combined as a sum did they have a difference.

[0] https://www.reddit.com/r/science/comments/2x844y/randomized_...

So only 9 of the 61 individuals that thought they had "gluten-sensitivity" actually reacted to the gluten, and only 3 were strongly affected by the gluten. I don't understand statistics very well... is that really significant enough?

I'm very surprised by those figures.

Is that saying of 58 people studied, if 2 more people had stronger negative effects to placebo than Gluten then there'd be no discernible difference between placebo and Gluten?

How does that show a significant correlation?

The health fad is super unfortunate, mostly because the "gluten free" foods I see are still just as shitty nutritionally as their gluten equivalents.

I'd love for this to be performed on a larger scale, we'll prove that either gluten sensitivity exists or psychics do if the results hold.

Actually, it's really great! Because it allows people who have an observable intolerance to gluten to eat food. That's a nice thing, isn't it?

What's unfortunate is your dismissal of (scare-quoted!) gluten free food as a health fad, but so it goes.

Most of the time I actually eat gluten free by choice, but I don't grab a box of gluten free crackers, I make a salad.

I'm dismissing things like Gluten-free corn dogs... why the hell gluten is in a corn dog is the first place is another problem all by itself.

Shitty nutritionally and taste like crap. We've tried all of the breads and other stuff. Now we just eat that burger with a knife and fork.

As somebody who has to stick to gluten free food at home because of their partner, it's tough because of how bad most things taste, but these guys make bread that's surprisingly decent:

Canyon Bakehouse - http://canyonglutenfree.com/

I'm not sure if it's available outside of Canada, but Celimix is a really decent gluten-free "flour" that I use when my mother-in-law is coming to town (side note: she is all the evidence I need to see that NCGS exists. Night and day differences.)

I've successfully made multiple loaves (and styles) of bread with this stuff, along with tortillas and amazingly (to me) an Angel Food cake from scratch. It's all turned out reasonably well the first time, and excellent the second time after I've had a chance to tweak things a little bit.

I would recommend using it in "scientist mode" though. I use a kitchen scale and measure the mass of all of the ingredients I use, along with taking notes about mixing times and things like that when I'm using it. Mostly because: a) I don't cook gluten-free very often, so I want repeatable results, and b) I really like to be able to figure out what went wrong.

The p-value is the probability that you would observe an effect of this size if the hypothesis were not correct. So you want them to be small. In most fields, you want to get at least p < 0.05 or 95% certainty, which is about two standard deviations from the mean in a normal distribution. That means that your result is far enough away from what you would get by random chance. https://en.wikipedia.org/wiki/Standard_deviation#mediaviewer...

Ok, so their observed effect ("significantly increased overall symptoms") is almost certainly due to gluten. Those are indeed good probabilities.

I'm interested in how large an effect "significantly increased" means. I suspect that this is defined a little more rigorously in the actual text, but it doesn't seem to be in the abstract.

> almost certainly due to gluten

Even if you ignore all the messiness of actually existing scientific practice, the math alone doesn't quite say that. It depends on your prior[0] and the power[1] of the study.

[0]: https://en.wikipedia.org/wiki/Bayes'_theorem [1]: https://en.wikipedia.org/wiki/Statistical_power

Keep in mind that p < 0.05 is actually quite a weak standard, and hard sciences tend to aim for a stronger p < 0.01. The problem with p < 0.05 and p-numbers in general is that current scientific practice allows to get whatever results you want with p < 0.05 by doing various tricks with statistics. Unfortunately this practice of "p-number hacking" is being fought nowhere near as it should be.

It's true that it's easy to abuse or misinterpret p-values. https://xkcd.com/882/ And one journal just banned significance tests! https://news.ycombinator.com/item?id=9109801

Some believe the use of glyphosate in wheat farming materially alters gluten, which alters your body's response to it. No proof one way or the other yet, but certainly an area where more scientific inquiry is appropriate.

https://jonrappoport.wordpress.com/2014/02/24/email-from-aut...

Snopes it. http://www.snopes.com/food/tainted/roundupwheat.asp

Snopes is just a couple with no investigative or journalistic training, who google stuff and express their own biases.

This appears to be the relevant paper: http://sustainablepulse.com/wp-content/uploads/2014/02/Glyph...

Unfortunately I don't read enough scientific journals to vouch for the veracity of it.

Reply on news.ycombinator.com