Hacker News new | ask | show | jobs
by wastra 3160 days ago
I'm a physician. There are a lot of problems with the original article.

The body is excellent at maintaining glucose concentration in the blood with food or with days of fasting. Glucose is the primary short-term energy source for most cells. The brain's energy requirements vary very little with "thinking" despite the assertions to the contrary in the introduction of this article. PET works by noting the slight transient increases in some brain regions with active thought, but moments later (as that 20% of blood flow goes swishing through the brain) more glucose is available. Meanwhile you are digesting sugars and carbohydrates, and your liver is supplying most of your fresh glucose if you are a fasting adult.

Already, before reading the article, I have a low pre-test probability of their hypothesis being true, so they need extraordinarily strong evidence to make their point. Instead, there is a very weak chain of experiments, poorly reported, with borderline statistical significance. Poor charts, no tabulated results, just terrible.

The primary problem is that they assert that glucose fluctuations occur due to the exercise of willpower. They try and fail to show this with their first experiment. However, their "results and discussion" combined section shows that their control group before and after glucose were very close to their post-intervention group glucose. p327 para 1.

This shows how weak their statistics and data are, and from this point, you can throw the whole study in the trash.

6 comments

Before re-reading the comments after approaching this thread again, a few hours after the first time I did, I basically just wanted to comment "this is very very unlikely to be true, and at best we can only say 'reasons that make people self-control make people self-control'"

Suppose I made the argument "car crashes rely upon momentum":

- car crashes is a terse signifier; a high-level recognition of something that no doubt relies upon low-level phenomena

- momentum is a physical, and therefore, low-level, phenomenon

No one would say I said anything useful. They would say "well duh but why were they driving towards where they drove? were they reading a text? were they drunk? were they even to blame when they crashed? was the car fully-functional? what were the weather conditions like? how much experience driving had they had? how old were they? how good was their vision?" etc..

More or less this is what is going on with, at least, the headline implies:

- self control: high level subjective phenomenon (who is to say that every person has self-control via non-dissimilar mechanisms??)

relies upon

- glucose: a correlate. Correlates are mostly useless and used to fool people, even if r > 0, and even if r=~1. Correlation more often indicates that there is a combination of excitement and merely-coincidental yet spurious non-causality in >99.99% of measurements.

I don't blame people for wanting there to be a simple, chemiological explanation for things that are very important to their existence. But I do blame the people who are responsible for preemptively jumping the gun and broadcasting the suggestion that these questions have been answered, especially because they are reporting that it was solved via something as vulgarly facile as in vivo molar concentration of near-universal molecule

Even if it is unlikely that the claim is true in a fundamental and meaningful and not-merely-correlative sense, after so many decades of prior investigators having approached this obvious hypothesis..

> But I do blame the people who are responsible for preemptively jumping the gun and broadcasting the suggestion that these questions have been answered

Usually when such claims are broadcast, it happens in two steps:

1) Some researcher (often a PhD student) finds some correlation (with our without conscious P-hacking) that has p<0.0.5. Their thesis may depend on it. 2) Some news broadcaster makes a story from the article. Their job/income may depend on it.

Now who, exactly, would you blame.

The broadcaster for broadcasting misinformation. The PhD student isn't producing misinformation (unless, they are, which is bad af).

I'm sure that every professional reporter, whether science popularizer or something else, understands and takes seriously the need for obey the Hippocratic Oath in their own profession, too ("first, do no harm").

But.. it's not like the reporter isn't aware that the constraints they are subject to (much higher time pressure + less familiarity with the topic vs the researcher's low time pressure + intimacy with the topic) don't automatically absolve them of their responsibility to own any mistakes they make once the genie's out of the bottle. It mostly hurts their readership if they start employing non-facts once they add them to their repertoire of decision heuristics.

Sure science reporting is definitely a good force in the world for helping people make decisions over important stuff, but.. like.. the researcher's reputation is ruined for causing potential damage to academia, but a reporter doesn't have skin in the game and has little to lose if others trusted them.

A big problem is that an observed decline in the level of glucose is being ipso facto attributed to the consumption of that glucose. No consideration seems to be given to the more likely possibility that the decline simply indicates the uptake of glucose into some tissue (reshuffling of glucose among bodily compartments, so to speak).
I agree. And they don't even show there was a decline in their initial experiment on which their entire chain of reasoning is based:

control group glucose before: 102+/-21, after 103+/-18 intervention glucose before: 107+/-21, after 101+/-18 (+/-SD)

Wide confidence intervals, and critically important to note, the pre-test intervention baseline of 107 is much higher than all the other numbers. They do not show that this is not by chance, i.e. null hypothesis of no difference is not rejected. They don't state a P-value (even though this is now out of vogue, the paper is ten years old), likely because it is greater than the mythical 0.05.

These numbers are laughable, so much so that I do not even think they are fabricated. They also quote two decimal places even though their SD is ~20! Every detail of this paper is amateurish.

I also have a problem with the ethics of this study. They recruited students and paid them in course credits. This strikes me as being coersive.

I'm sure I could find more gaping holes in their reasoning, statistics, but I hope I've already shot it down. I'm not a neuroscientist by any means, but I do digest a lot of scientific papers. My guess for a real cause of this effect would be neurotransmitter depletion in key neurons involved in self-discipline.

>My guess for a real cause of this effect would be neurotransmitter depletion in key neurons involved in self-discipline.

Thanks for your detailed refutation of this article. I pinboard'd the article, then read your comment, and un-pinboard'd it.

Your guess for the real mechanism is a depletable neurotransmitter. It brings up the matter of discipline - is it a consumable resource, or a buildable resource?

I think traditionally most people think of discipline/self-control being a limited resource. In the course of my own life, trying to manage habits( example, strength training, eating healthily) I find the less I have to rely on at-the-moment willpower, and more on habit, the more adherence I demonstrate. That leads me to think that discipline is a (precious) limited resource.

On the other hand, was listening to a Jocko Willink podcast episode this weekend, and he claims that exercising discipline leads to further capacity for more discipline.

I don't totally reject his idea, but it implies there isn't a limited neurotransmitter, but more some kind of positive feedback loop.

Truth is probably some kind of murky combination I suppose.

From what I understand, behaviorally there are two systems. There's active management, i.e. exerting willpower, and there's habitual behavior, i.e. doing what you usually do.

In the same way that "You" are not a monolithic entity, "You" don't pay attention to everything - take the old "think about your breathing" troll, which causes a redirection of attention to an otherwise autonomous function.

In this world, the role of "willpower" is in habit formation - doing something routinely enough that you train 'yourself' to do it automatically. So exercising this ability can help you train yourself to do it more easily - you're meta-training yourself to train yourself.

But the active exercise of will is by no means energetically free, and that's the origin of the original paper's hypothesis. I'd guess that this exercise of will is not especially amenable to alteration beyond gross physical trauma, but I don't think it has much to do with blood glucose.

Would you care to show your work? I'm sure there are many here that would be interested, not only in your results, but how you arrived at them, particularly if they're presented in a way that would allow others to apply them to other papers they come across.
User Wastra didn't claim to have done any work beyond reading the paper critically, or have any results. He/she gave a concise review of the paper: that's the output of the reading. What are you asking for?
You're right with respect to reading the paper critically. What I'm looking for in particular is how 'wastra arrived at "there is a very weak chain of experiments, poorly reported, with borderline statistical significance. Poor charts, no tabulated results, just terrible."
I thought I'd just mention that it appears the issue here is there is a need to rephrase your request for substantiation. The burden of proof is rightly on the claim of new discovery, and user wastra should not be burdened with proving or providing work that the above claim is false, even when the user is being critical.

Asking wastra to expand or show why he/she claims what they claim is fine. However, what I think people are taking issue with is the need for the aforementioned distinction in burden of evidence--it appears your comment on 'work and results' may be unintentionally misinterpreted in this way.

Thanks for the feedback. Agreed on all points.
Going out on a limb:

HN user wastra read a paper which is obviously rife with reports from weak chains of experiments, poorly reported and with borderline statistical significance.

Then, HN user wastra commented accordingly.

Too late to edit: it looks like people are reading my comments as obliquely questioning 'waster's credentials. I can see how one could read my comments that way and given how conversations on the internet can go, how one may believe that to be the case; but that's really not my intent. If I took issue with their findings, I'd come out and say so. I'm just interested in finding out more where their coming from.
You're welcome to just go "I don't trust wastra's self-identified credentials" and keep scrolling.
Why is it that glucose levels correlate significantly with self reported “energy” and mood levels?
This is one of the earliest tricks of the paper. In the initial discussion they cite numerous authors on why low glucose, i.e. hypoglycemia, affects cognition. I have seen many hypoglycemic diabetics in bad trouble from this. However, the author then silently moves on to asserting that lower levels within the normal glucose range also have this effect, but without any supporting evidence.

In addition, the commonly stated myth that people "feel hypoglycemic" could be easily disproved in 99% of people if it was worth the time and effort to do so. The awareness of hunger, awareness of time since last meal, and the nocebo suggestion that such a thing as hypoglycemia in (most) normal people exists. It is a real thing with some medical problems, and a very few otherwise healthy people, but far overstated.

If we're into anecdotal levels of evidence, I eat once a day, and _feel_ the least energetic after eating, for 30-60 minutes.

We’ll sure, food coma is a thing. But at least my own anecdote is that sipping a sugary drink helps my focus and attention last longer. I haven’t done a formal double blind study but it is something I noticed after the fact—coffee doesn’t really work for me, but a sugary, milky coffe does. I found this by doing regression on a spreadsheet of my intake for a few months and self reported energy and focus levels.

It was after he fact that I found out this a commonly reported correlation. I wouldn’t be so quick to discount it as placebo.

I know from diets that low calorie intake does affect the mind, usually in a positive manner. That's not necessarily something I would call "hypoglycemia" though. I never checked, so far.

Both glucose and insulin levels vary throughout the day, even in healthy people, so I don't see why blood sugar can't be "low", at least with respect to the individual's daily range.

Thank you for your comment. I have more faith on peer review system.
Exactly. I fast for 6 days, no problem (after 3 days of torture). Right now, I have enough excess weight to survive for 40 days without food. And I know what hypoglycemia is like: lots of swimming all day as a kid; it feels like you're gonna die unless you eat something. Hypoglycemia should be rare unless someone either over-exerts themselves and depletes all glycogen reserves (IIRC fat burning isn't a fast process) or have some metabolic issues like diabetes.

To risk being that "hey, will you fix my computer/heath/car" jerk: I received this odd recommendation from my PCP: metformin for non-T2 diabetics is supposedly "dangerous" because "it lowers blood sugar." I was tolerating 1500 mg, no problem, with few side-effects other than losing weight faster (which was great). This doesn't make sense because there's nothing in the literature or warnings about it lowering blood glucose dangerously. So then how in tarnation would it (a now cheap generic, no less) in an FDA phase 4 clinical trial as the first and, so far, only "anti-aging" medication because it does something to promote healthier metabolic and other pathways?

https://clinicaltrials.gov/ct2/show/NCT02432287

@burntrelish Would you fast for 40 days? I’ve been looking into extended fasts for health benefits. I’ve been following intermittent fasting and enjoying it.
There's a decent chance you'll die if you fast for 40 days. You can google stories of this happening. Even if you don't die, you could really hurt yourself. 40 days of fat stores won't prevent your body from breaking down organs and muscles.

Fasting for health benefits is typically done for one to several days.

I'm not a doctor or the OP, but felt someone should respond to your question because of the risks.