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 versus melatonin for sleep: what does each study actually measure?

By Alejandro del Palacio · · 9 min read

The short answer

No head-to-head trial is reviewed here, so neither source says which to choose. Against placebo, melatonin shortened sleep latency by 7.06 min across 19 RCTs of people with primary sleep disorders (Ferracioli-Oda 2013). In an Oura-ring RCT (n = 76), latency rose 0.89 min with magnesium L-threonate and fell 0.08 min with placebo (Hausenblas 2024).

Key findings

  • Melatonin: a meta-analysis of 19 RCTs (n = 1683, people with primary sleep disorders, not healthy people) found a latency 7.06 min shorter than with placebo. The confidence interval runs from 4.37 to 9.75 (Ferracioli-Oda 2013).

  • Magnesium L-threonate: in 80 adults aged 35-55 with non-clinical insomnia symptoms (n = 76), latency went from 9.07 to 9.96 min with magnesium and from 10.07 to 9.99 min with placebo. The condition-by-time interaction gave p = 0.050 (Hausenblas 2024).

  • Corrigendum: the magnesium paper has a 2025 correction. It adds a retrospective trial registration, dated April 2024, and the full declaration of competing interests. The funder, AIDP Inc., owns patents (Hausenblas 2025).

  • What each one measures: mixed methods in the meta-analysis, among them polysomnography and actigraphy; an Oura ring in the magnesium RCT. Different populations, forms and methods.

  • StatusEstimated

The myth: “Magnesium is the natural version of melatonin and does the same thing”

Short answer: the sources reviewed here do not support it. Against placebo, melatonin shortened latency by 7.06 min across 19 RCTs in people with primary sleep disorders (Ferracioli-Oda 2013). In the magnesium L-threonate RCT with an Oura ring, latency in the magnesium group did not fall: it rose 0.89 min (Hausenblas 2024). They are different studies, in different populations, and they do not say which to choose.

In this article
  1. What does each study say about the time it takes to fall asleep?
  2. Why can't you subtract one number from the other?
  3. How was each one measured?
  4. Who funded the magnesium trial, and what did the 2025 corrigendum add?
  5. What can you check in your own nights?
  6. Where the evidence ends

What does each study say about the time it takes to fall asleep?

Sleep latency is the time between lying down to sleep and falling asleep. Both studies report it. They report it for different groups.

For melatonin, this article reviews one meta-analysis. A meta-analysis pools the results of several trials into one estimate. This one gathered 19 randomized controlled trials (RCTs) with 1683 people in total. All of them had primary sleep disorders, so they were not healthy people. The pooled result was a latency 7.06 min shorter than with placebo. The confidence interval, 4.37 to 9.75, is the range of values the data are compatible with. A wider range means a less precise average (Ferracioli-Oda 2013).

For magnesium, this article reviews one trial. It was a double-blind, placebo-controlled RCT of magnesium L-threonate (Magtein). Its population is 80 adults aged 35-55 with non-clinical insomnia symptoms. The latency figures come from 76 of them. With magnesium, latency went from 9.07 to 9.96 min. With placebo, it went from 10.07 to 9.99 min (Hausenblas 2024).

That is a rise of 0.89 min with magnesium and a fall of 0.08 min with placebo. The authors write that the magnesium group slightly increased latency compared with placebo.

Why can't you subtract one number from the other?

The two numbers look comparable. They are not the same kind of number.

The 7.06 min is a difference between melatonin and placebo, pooled over 19 trials. The 0.89 min is a change inside one group, between two time points in one trial. A change inside a group says nothing about a comparison with placebo until you look at the placebo group too. In the magnesium trial, the placebo group barely moved: it fell by 0.08 min.

That is the reason a trial has a placebo arm. Without it, you cannot tell whether a change comes from the compound or from the passing of days. The magnesium paper compares the two groups with a statistical interaction. Its Table 4 labels the column “Interaction (Condition x Time)”. For sleep latency that interaction is p = 0.050. The table's footnote reads “p < 0.05 indicates significance between groups.” So 0.050 sits right at the edge of that line.

The 80 and the 76 also differ. Table 4 reports n = 38 per group, which adds up to the 76 of the 80 adults described.

How was each one measured?

Each study used a different form of the compound, a different dose, a different population and a different instrument.

Melatonin. The meta-analysis pools trials of immediate-release melatonin taken by mouth, at doses from 0.5-5 mg. Its latency result mixes polysomnography and actigraphy. Polysomnography records sleep in a lab, with sensors on the head and body. Actigraphy estimates sleep from wrist movement. The meta-analysis averages the two together. It does not publish how many days it took for the effect to show.

Magnesium. The RCT gave magnesium L-threonate, 1.0 g a day, for 21 days. Sleep latency came from an Oura ring, not from polysomnography.

The table puts side by side what each study did.

CompoundForm and dosePopulationDesignnMethodLatency resultCitationStatus
MelatoninOral, 0.5-5 mg (range across the 19 RCTs)People with primary sleep disorders (not healthy)Meta-analysis of 19 RCTs1683Mixed, among them polysomnography and actigraphy7.06 min shorter than with placebo (interval 4.37 to 9.75)Ferracioli-Oda 2013Estimated
MagnesiumL-threonate (Magtein), 1.0 g a day, 21 days80 adults aged 35-55 with non-clinical insomnia symptomsDouble-blind placebo RCT76Oura ringMagnesium 9.07 to 9.96 min; placebo 10.07 to 9.99; interaction p = 0.050Hausenblas 2024Estimated

When the compound, the form, the dose, the group, the design, the sample size and the instrument all differ, subtracting one number from the other tells you nothing reliable. One result is a summary of 19 trials. The other is a single trial.

Who funded the magnesium trial, and what did the 2025 corrigendum add?

The original 2024 paper states that the research was funded by AIDP Inc. Its competing-interests statement was incomplete.

In 2025 the journal published a corrigendum to that paper. As read for this article, it covers two things. First, the trial was registered after it started: it was retrospectively registered in April 2024 in the ISRCTN registry (ISRCTN14728094). The corrigendum states that no changes were made to the protocol after the trial commenced. Second, it adds the full declaration of interests. AIDP Inc. commissioned and funded the study and owns patents. Three of the authors are paid by AIDP. The other three declare no competing interests.

None of this changes how the latency figures are read here. They come from Table 4 of the original paper. It does mean you should read the result as a single trial, funded by AIDP Inc. and registered after it started.

What can you check in your own nights?

If you track your nights with a ring or a wrist device, you have a series from one body. These two studies do not tell you what to take. They show what a latency number means: a group average, from one instrument, in one population.

A single night says little. A series of nights with the same device, at similar times, shows how your own nights vary. If you change something in your routine, write down the date. That way you know where the series splits.

The group averages above do not tell you how much one person varies from night to night. They also do not say that your own latency would move by 7.06 min or by 0.89 min.

For more on these topics on this site: What the Evidence Says About Magnesium Forms and Sleep, How Many Nights Does Melatonin Need Before You Can Judge It?, and How strong is the evidence for L-theanine and falling asleep?.

Where the evidence ends

This is what the data do not say.

  • It does not compare one compound against the other. There is no direct RCT between melatonin and magnesium in these sources. Each result lives in its own study.
  • No population is “you.” Melatonin was measured in people with primary sleep disorders. Magnesium was measured in adults aged 35-55 with non-clinical insomnia symptoms.
  • The form matters. The trial used magnesium L-threonate. Its result says nothing about other forms of magnesium, such as glycinate.
  • The instruments are not equivalent. The meta-analysis mixes methods, among them polysomnography and actigraphy. The RCT uses a ring. We do not know whether the minutes from one equal the minutes from the other.
  • The timelines differ. The meta-analysis does not report the days until the effect. In the magnesium RCT, the dose was 1.0 g a day for 21 days; that is the length of the trial, not a time to effect.
  • One magnesium trial, with its conflicts. It has p = 0.050, funding from AIDP Inc., and a registration made after the start. On its own, it does not establish that magnesium lengthens latency in general.
  • Only latency is reviewed. The melatonin meta-analysis does not report heart rate variability, temperature or recovery.
  • It does not cover the whole decision. Which problem each one fits, safety, interactions and timing of the dose sit outside this article.

What it means

There is one number for each study. Each comes with its population, its form of the compound, its instrument and, for magnesium, its funder. The question “which one” does not get an answer from here. What you can take from here is a way to read each figure with its context.

Quick questions

What does each study say about the time it takes to fall asleep?

Sleep latency is the time between lying down to sleep and falling asleep. Both studies report it. They report it for different groups. For melatonin, this article reviews one meta-analysis. A meta-analysis pools the results of several trials into one estimate.

Why can't you subtract one number from the other?

The two numbers look comparable. They are not the same kind of number. The 7.06 min is a difference between melatonin and placebo, pooled over 19 trials. The 0.89 min is a change inside one group, between two time points in one trial.

How was each one measured?

Each study used a different form of the compound, a different dose, a different population and a different instrument. Melatonin. The meta-analysis pools trials of immediate-release melatonin taken by mouth, at doses from 0.5-5 mg. Its latency result mixes polysomnography and actigraphy.

Who funded the magnesium trial, and what did the 2025 corrigendum add?

The original 2024 paper states that the research was funded by AIDP Inc. Its competing-interests statement was incomplete. In 2025 the journal published a corrigendum to that paper. As read for this article, it covers two things.

What can you check in your own nights?

If you track your nights with a ring or a wrist device, you have a series from one body. These two studies do not tell you what to take. They show what a latency number means: a group average, from one instrument, in one population. A single night says little.

Sources

Keep reading:

Alejandro del PalacioAuthor · reviewed on 2026-11-17

A measurement tool, not medical advice.

ORIGENHuman Potential Measured

pieza_id p1-magnesium-vs-melatonin-sleep