ORIGENHuman Potential MeasuredYour protocol with Oura or WHOOP
A person before a stone threshold and a lake at dusk.

Sourced, not promised.

Research / Mental Work

Magnesium glycinate or L-threonate: what each study actually measured

By Alejandro del Palacio · · 8 min read

The short answer

The two forms were measured with different tools. Glycinate, with an insomnia questionnaire: d = 0.2 at 28 days (2025). L-threonate, with an Oura ring: in minutes, deep sleep did not separate from placebo (p = 0.776; Lopresti & Smith 2026). Among these three trials, none compared the two forms.

Key findings

  • Glycinate: a questionnaire. In healthy adults who report poor sleep, 250 mg a day of elemental magnesium was tested against placebo. The main outcome was the Insomnia Severity Index at week 4, with an effect size of d = 0.2 (2025).

  • L-threonate: a ring. Two placebo-controlled trials used an Oura ring. Minutes of deep sleep did not differ from placebo in either (p = 0.776 and p = 0.070).

  • A score is not minutes. Oura's Deep Sleep Score did separate from placebo (p < 0.001), but mostly because the placebo group's score fell (Hausenblas 2024).

  • StatusEstimated

The myth: “Studies on magnesium and sleep all measure the same thing”

Short answer: no. The glycinate trial measured a questionnaire score, the Insomnia Severity Index (d = 0.2 at week 4; 2025). The L-threonate trials measured sleep with an Oura ring, in minutes and as a score. In minutes of deep sleep, neither of the two L-threonate trials separated from placebo.

In this article
  1. What did the glycinate study measure?
  2. What did the L-threonate studies measure?
  3. Why can’t you line the two forms up?
  4. What does it mean when the placebo group falls?
  5. What can you check with your own data?
  6. Where the evidence ends

What did the glycinate study measure?

The study is titled “Magnesium Bisglycinate Supplementation in Healthy Adults Reporting Poor Sleep” and appeared in Nature and Science of Sleep in 2025. Bisglycinate and glycinate are the same form.

It was a placebo-controlled randomized trial. The participants were healthy adults who said they slept badly. The dose was 250 mg of elemental magnesium a day.

The main outcome was the Insomnia Severity Index, or ISI. The ISI is a questionnaire. People answer it themselves. At week 4, which is 28 days, the effect size was d = 0.2.

Two limits come with that figure. A d of 0.2 is a small effect. And the trial reports no objective measure of how long it took people to fall asleep. The number of participants is not stated here.

What did the L-threonate studies measure?

There are two trials, and both used an Oura ring.

The first is Lopresti & Smith (2026). It was a double-blind, placebo-controlled trial with prospective registration. It enrolled 100 healthy adults aged 18 to 45. The dose was 2 g a day of L-threonate for 6 weeks, or 42 days. The ring was an Oura gen 3. What is reported here is the minutes of deep sleep the ring recorded.

In the active group, deep sleep went from 74.81 to 72.76 minutes (-2.04, p = 0.354). In the placebo group, it went from 76.15 to 74.80 (-1.34, p = 0.535). Between groups, p = 0.776. Both groups went down. In the same trial, sleep latency, efficiency, REM and total sleep were also not significant. The manufacturer of the product, Threotech Inc., funded it.

The second is Hausenblas (2024). It was a double-blind, placebo-controlled trial. It enrolled 80 adults aged 35 to 55 with non-clinical insomnia symptoms (mean ISI 12.46). The n in our source is 76. The dose was 1 g a day for 21 days, about 75 mg of bioavailable elemental magnesium. The Oura ring gave results as scores and also in minutes.

The Deep Sleep Score went from 93.84 to 94.20 in the active group (+0.36 points). In the placebo group it went from 93.58 to 90.70 (-2.88). The interaction was p < 0.001. In minutes, the same table gives p = 0.070, which is not significant.

Sleep latency moved the other way. In the active group it went from 9.07 to 9.96 minutes (+0.89). In the placebo group it went from 10.07 to 9.99 (-0.08). The interaction was p = 0.050, right at the usual significance threshold: a borderline signal, not a firm finding. The paper puts it this way: “MgT slightly increased sleep latency as compared to the placebo.”

Why can’t you line the two forms up?

Because the studies did not ask the same question.

The glycinate trial asked how people rated their own sleep problems on a questionnaire. The L-threonate trials asked what a ring recorded each night. Neither measure substitutes for the other.

The populations differ too. One trial enrolled healthy adults who report poor sleep. Another enrolled healthy adults aged 18 to 45. The third enrolled adults aged 35 to 55 with non-clinical insomnia symptoms. The doses are expressed differently as well: elemental magnesium for glycinate; grams of the compound for L-threonate (in Hausenblas, 1 g equals about 75 mg of elemental magnesium).

So a sentence like “one form beat the other” has nothing to rest on. No trial here compared the two forms head to head.

The table shows what each trial measured. It has no ranking column on purpose.

StudyForm and dosePopulationInstrumentOutcomeResultStatus
Bisglycinate trial, 2025Glycinate, 250 mg elemental magnesium a dayHealthy adults reporting poor sleepQuestionnaireInsomnia Severity Index at week 4d = 0.2Estimated
Lopresti & Smith 2026L-threonate, 2 g a day, 42 days100 healthy adults, 18 to 45Oura gen 3 ringMinutes of deep sleepBetween groups, p = 0.776Estimated
Hausenblas 2024L-threonate, 1 g a day, 21 days80 adults, 35 to 55 (n = 76)Oura ringDeep Sleep Score; minutes; latencyScore p < 0.001; minutes p = 0.070; latency p = 0.050 (borderline)Estimated

What does it mean when the placebo group falls?

In the Hausenblas trial, the active group’s Deep Sleep Score barely moved (+0.36). The placebo group’s score dropped by 2.88. The p < 0.001 measures the gap between the two changes. It does not say the active group’s score went up.

In Lopresti & Smith, deep sleep fell in both groups: 2.04 minutes in the active group and 1.34 in the placebo group. Neither drop was significant (p = 0.354 and p = 0.535).

That is why the active group’s change is best read next to the placebo group’s. In Hausenblas, the placebo group’s score fell without taking the product.

What can you check with your own data?

With your own data you can see how much your numbers vary from night to night when nothing changes.

Pick the instrument first. A ring gives minutes and scores. A questionnaire gives a rating. A ring cannot check a questionnaire score, and a questionnaire cannot confirm a ring’s minutes.

If you look at deep sleep on a ring, prefer minutes to the score. The Deep Sleep Score is a proprietary Oura score, not a physical quantity, and it does not convert to minutes.

Then ask how much deep sleep has to change before it stands out from ordinary night-to-night variation. One body’s measurement of that is in how much deep sleep has to change before it counts. For another view on the ring and magnesium, there is whether a ring can actually verify magnesium with our thresholds.

If you want an overview of the forms of magnesium and sleep, see what the evidence says about magnesium forms and sleep.

Where the evidence ends

This is what the data does not say.

  • It does not compare glycinate with L-threonate. No trial here gave both forms to the same people.
  • It does not say either form works or fails. It describes what each study measured and what it found.
  • It does not tell you which one to choose.
  • The glycinate figure comes from a questionnaire. It says nothing about minutes of deep sleep.
  • The L-threonate trials used Oura rings. A ring is not a laboratory sleep study.
  • Oura’s Deep Sleep Score is proprietary. It does not convert to minutes.
  • The Lopresti & Smith trial was funded by the manufacturer. This article does not describe the funding of the other trials; check each original paper.
  • The sources give both 80 and 76 for the Hausenblas trial.

What it means

A claim about magnesium and sleep needs three details before it can be compared: the form, the population and the instrument. Glycinate has a questionnaire result at week 4. L-threonate has ring results in minutes and as a score. In minutes of deep sleep they are null. Sleep latency, also in minutes, went up in Hausenblas (+0.89; p = 0.050, borderline).

If you track your own sleep, decide first what you will count: a rating, a score or minutes. That choice shapes what you can see.

Quick questions

What did the glycinate study measure?

The study is titled “Magnesium Bisglycinate Supplementation in Healthy Adults Reporting Poor Sleep” and appeared in Nature and Science of Sleep in 2025. Bisglycinate and glycinate are the same form. It was a placebo-controlled randomized trial. The participants were healthy adults who said they slept badly.

What did the L-threonate studies measure?

There are two trials, and both used an Oura ring. The first is Lopresti & Smith (2026). It was a double-blind, placebo-controlled trial with prospective registration. It enrolled 100 healthy adults aged 18 to 45. The dose was 2 g a day of L-threonate for 6 weeks, or 42 days. The ring was an Oura gen 3.

Why can’t you line the two forms up?

Because the studies did not ask the same question. The glycinate trial asked how people rated their own sleep problems on a questionnaire. The L-threonate trials asked what a ring recorded each night. Neither measure substitutes for the other. The populations differ too.

What does it mean when the placebo group falls?

In the Hausenblas trial, the active group’s Deep Sleep Score barely moved (+0.36). The placebo group’s score dropped by 2.88. The p < 0.001 measures the gap between the two changes. It does not say the active group’s score went up.

What can you check with your own data?

With your own data you can see how much your numbers vary from night to night when nothing changes. Pick the instrument first. A ring gives minutes and scores. A questionnaire gives a rating. A ring cannot check a questionnaire score, and a questionnaire cannot confirm a ring’s minutes.

Sources

Related reading:

Alejandro del PalacioAuthor · reviewed on 2026-11-20

A measurement tool, not medical advice.

ORIGENHuman Potential Measured

pieza_id p1-magnesium-glycinate-vs-threonate