Glycine and deep sleep: what 3 g actually do

I take 3 g of glycine about 45 minutes before going to bed. Nothing happens at the time. No drowsiness, no heaviness, nothing remotely like a sleeping pill. The difference shows up the next morning.

My watch gives me a lot of REM sleep, around 32% across nine hour nights, and little deep sleep, often thirty to forty minutes. On glycine, the deep sleep it displays climbs to an hour ten or an hour twenty, and I wake up markedly less tired.

This is not a one-off impression. I have alternated with and without several times, on regular nights, with the same device: when I stop, the displayed deep sleep drops back. That is not a clinical trial, and it is not a «I feel better» either.

It is interesting because the literature on glycine does not say quite the same thing as my watch. Both deserve laying out.

Why nothing happens at the time

Glycine is not a sedative, and its mechanism explains why taking it is a non-event. In rats, oral glycine causes dilation of blood vessels in the skin, measured at the sole of the foot and proportional to the dose. The body sheds heat, and core temperature falls.

That temperature drop is the physiological signal for sleep onset, not a consequence of it. In the same study, glycine induced non-REM sleep and shortened the time needed to enter it. The effect runs through NMDA receptors in the suprachiasmatic nucleus, the brain's internal clock: destroying that nucleus abolishes the effect entirely, while a blocker of glycine's own receptors changes nothing.

Two practical consequences. First, there is nothing to feel when swallowing the capsule: a thermal shift does not announce itself. Second, it needs time to happen, which makes a dose forty-five minutes before bed more coherent than one taken while closing your eyes.

The only trial that used electrodes

In 2007 a Japanese team gave 3 g of glycine before bed to volunteers complaining of unsatisfactory sleep, with polysomnography recording, the reference measurement that tracks brain activity across a whole night.

The results point the right way. Subjective sleep quality improves. So does sleep efficacy, meaning the share of time actually slept out of the time spent in bed. Sleep onset latency shortens, and so does the latency to deep sleep. Daytime sleepiness falls, and performance on memory recognition tasks improves.

Then comes the precision almost nobody repeats, and it sits in the same sentence: all of that without changes in the sleep architecture.

Read it again. Glycine shortens the time it takes to reach deep sleep. It does not change how much of it you get. Almost every page on the subject flattens that distinction, and it changes what can legitimately be expected from the product.

So why do I wake up less tired?

Because that is precisely what was measured. The 2007 trial reports less daytime sleepiness. A second trial went further: in healthy volunteers whose sleep had been cut by 25% for three consecutive nights, 3 g of glycine before bed produced a significant drop in reported fatigue, a trend towards less sleepiness, and above all a significant improvement on the psychomotor vigilance test, which measures reaction time and cannot be talked into a result.

The same trial checked the hormonal route and did not find it: neither plasma melatonin nor the internal clock genes moved. So it is not a chronobiotic in disguise.

In other words, the result I observe on waking is exactly the one the trials measured. What stays open is the route it takes, and that is where my watch and the trial diverge.

My watch and the trial are not measuring the same thing

That leaves the gap. The trial finds no extra deep sleep, my watch displays two to three times more. Before picking a side, it helps to know what each one measures, and a validation study gives half the answer: seven consumer devices compared with polysomnography in 34 adults, over three nights in a lab including one deliberately disrupted.

For detecting sleep, the devices are good, with sensitivity of at least 0.93. For assigning stages, the assessments are inconsistent. Reading an absolute deep sleep figure is therefore fragile, and I know my own watch tends to overestimate REM.

But that study answers a precise question: does the device agree with polysomnography on a given night? It does not ask whether the device tracks a change correctly in the same person, on the same model, under stable conditions. A systematic bias distorts the value, it does not stop a gap from showing when the gap repeats.

What stands is therefore precise: a reproducible difference, in the same direction, between nights with and nights without. What falls is certainty about the exact number of minutes. That is not the same as «this number is worthless».

Three ways to reconcile the two

The first is the gap between an average and a person. The trial covers a small group of dissatisfied sleepers, and its result is an average. It says architecture does not move on average, not that it moves in nobody. That is exactly the tension the health score article deals with: a statistical norm and an individual trajectory do not answer the same question.

The second is an unusual starting profile. Plenty of REM sleep and little deep sleep is not the average starting point of a trial participant. An effect that shifts the beginning of the night does not produce the same result depending on what there was to shift.

The third follows directly from the trial result, and it is the one I find most solid. If deep sleep arrives earlier, it occupies a place REM would have taken. In someone whose REM overflows, bringing deep sleep forward can mechanically raise its share across the whole night. A shortened latency and a rise in deep sleep would then not be two competing claims, but the same one seen in two places.

None of the three is demonstrated. They are hypotheses, and I give them as such. But they explain the gap without having to throw out either measurement, which already beats picking a side.

The dose, and why one gram is not enough

Both human trials cited here use 3 g. Not 1 g, not 500 mg: 3 g, taken before bed. It is the only amount the sleep literature rests on.

Many commercial capsules nevertheless recommend 1 g per day. That dose was evaluated in none of this work. It is not proof that it does nothing, it is the observation that it falls outside what has been tested, and personally I noticed nothing below 3 g.

And digestion?

That is the effect I was not expecting, and it needs a word of context. I am autistic, and I have always had digestive problems. That is not an individual quirk: a meta-analysis of 15 controlled studies finds markedly more gastrointestinal symptoms in autistic children, with an odds ratio of 4.42, meaning odds of occurrence more than four times higher. The gap shows up separately for diarrhoea, constipation and abdominal pain.

That work covers children and adolescents, and I flag it because it is a real limit: nobody has shown the association stops at eighteen, and nobody has measured it after either.

On that front, glycine resolved nothing. It improved things, with better formed stools, clearly enough that I noticed without looking for it. I also take probiotics continuously, but glycine I first took in stretches, and it was across those alternations that the difference showed.

There is a clean mechanism besides. Glycine is a significant component of the bile acids secreted into the small intestine, the ones needed to digest dietary fat and absorb long-chain fatty acids.

Two things meet, then. An effect I was not looking for and did not expect, and a documented role for glycine at the exact place where I notice a change. That convergence demonstrates nothing, and it beats a coincidence: I could not have expected an effect whose mechanism I did not know about.

What is missing is simple to state: no trial has tested supplemental glycine on digestion. No negative trial, no positive trial, no trial. That is an absence of evidence, which is not evidence of absence, and it remains the only honest thing to say about it.

What this article does not say

• That glycine increases deep sleep in everyone. The only human trial with electrodes did not see architecture move, on average, in a small group.
• That my watch minutes are a laboratory measurement. The exact number is fragile. The repeated difference between nights with and without is far less so, and that is what I report.
• That it has no effect for all that. Two trials find less fatigue and better vigilance the next day. Small groups, but measured.
• That it treats insomnia. Established insomnia gets managed, and the first validated option is not a supplement.

« A trial measures an average. A watch measures a gap. Neither one measures the other. »

What Helix does

This case sums up the problem Helix tries to make readable. One product, three claims that circulate together, and three different levels of evidence: a human trial with electrodes for the latency, a placebo-controlled trial for next-day fatigue, an animal model for the mechanism, and nothing at all for digestion.

The engine shows that distinction molecule by molecule, instead of filing everything under «benefits». And when an effect rests only on what your watch reports, it says that too.