Does magnesium help you sleep? What the trials show

The statistic behind most magnesium marketing is that 48 percent of Americans consume less magnesium than their estimated requirement. That figure is real and comes from NHANES 2013-2016. It is also an intake shortfall calculated from food diaries, not a measurement of anyone's magnesium status.

The National Institutes of Health puts the actual position plainly: no current data on magnesium status in the United States are available, and NHANES has not measured serum magnesium in its participants since 1974. Fifty-two years. The same page notes that symptomatic deficiency from low dietary intake in otherwise healthy people is uncommon, because the kidneys limit how much you excrete.

Testing will not settle it either. An adult body holds about 25 grams of magnesium, 50 to 60 percent of it in bone, and less than 1 percent in blood serum, where it is kept under tight control. NIH states that serum levels have little correlation with total body magnesium or with concentrations in specific tissues. In a magnesium loading test in 88 healthy adults, the authors found no correlation between how much magnesium a person retained and their serum magnesium at all.

So a normal serum result rules out hypomagnesemia and rules nothing else in or out. There is no consumer test worth paying for.

The only meta-analysis on magnesium for insomnia in older adults pooled three trials totalling 151 people, and its headline sleep-onset finding rests on two of them, with 46 and 12 participants.

The best-designed trial to date randomized 155 adults with poor sleep to 250 mg of elemental magnesium as bisglycinate for four weeks. Insomnia severity improved by 1.6 points more than placebo, p=0.049, Cohen's d of 0.2. The authors call it a modest benefit, and that is the strongest result in the field.

The largest trial with a real placebo, 100 adults over 51 on 320 mg for seven weeks, found sleep quality improved substantially regardless of whether people got magnesium or placebo.

Then there is magnesium L-threonate, the premium sleep form. Its flagship trial named six primary outcomes and every one was null: insomnia severity p=0.390, sleep quality p=0.81, getting to sleep p=0.25, and three more. The positive subscales are driven by the placebo group getting worse. That paper carries an erratum disclosing that it was registered retrospectively, nineteen months after the trial closed, and that it was commissioned and funded by a company that owns patents on the ingredient and whose VP of R&D is an author. A second industry-funded threonate trial found no group differences in sleep outcomes either.

The strongest observational test, nearly 3,900 people with actual polysomnography and a genetic magnesium score, found no consistent association between magnesium supplementation and subjective or objective sleep, and no effect on night cramps.

Cochrane reviewed 11 trials and 735 participants. For older adults with night cramps, the proportion of people responding was a relative risk of 1.04 (95% CI 0.84 to 1.29), and Cochrane graded that high certainty, its top rating. Cramps per week showed a difference of -0.18 (95% CI -0.84 to 0.49) at moderate certainty. Their conclusion: it is unlikely that magnesium supplementation provides clinically meaningful cramp prophylaxis to older adults.

The funding pattern is worth seeing. Every independently funded trial is null, including one stopped early for futility. Both trials reporting benefit tested a product made by the company that funded them, and the positive one was registered retrospectively and published no confidence interval for its primary outcome.

The newest trial, prospectively registered with three arms, found magnesium at -0.20 cramps per week (95% CI -1.49 to 1.09, p=0.929) while compression stockings worked, at -1.43 (95% CI -2.36 to -0.50, p=0.001).

For pregnancy cramps the one registered meta-analysis is also null on every outcome.

Search for magnesium and migraine and you will find that the American Academy of Neurology and the American Headache Society rated it Level B, probably effective. That rating comes from a 2012 guideline.

That guideline was retired by the AAN Board of Directors on 16 September 2015, over safety concerns about butterbur, another treatment it recommended. The AAN's own notice states that the recommendations and conclusions in all retired guidelines are considered no longer valid and no longer supported.

The NIH Office of Dietary Supplements page, updated 6 January 2026, still prints the Level B conclusion with no mention of the retirement. So does the National Center for Complementary and Integrative Health. Anyone sourcing from a .gov page inherits a rating that was withdrawn eleven years ago.

No current AAN or AHS document gives magnesium any preventive grade. The current AHS consensus statement mentions magnesium exactly twice, both for acute rescue, and its preventive table does not list it. There is no completed Cochrane review, only a protocol filed in November 2025.

The trials themselves are thin: five small adult studies, the largest 81 people, the longest 12 weeks, two of them from 1996. The one trial with a clearly pre-specified primary endpoint was null and stopped for it, with responder rates of 28.6 versus 29.4 percent. When researchers re-applied the AAN's own grading scheme in 2018 they arrived at Grade C, possibly effective, a downgrade. And the dose that produced the positive result, 600 mg, is 1.7 times the safe upper limit for supplements.

Blood pressure. The current synthesis, 38 trials and 2,709 people, finds systolic pressure down 2.81 mmHg (95% CI -4.32 to -1.29). The subgroups are where it lives: 7.68 mmHg in hypertensive people already on medication, 5.97 in people who are actually low in magnesium, and in normotensive groups statistical significance was not reached. There was no dose-response relationship at all. The FDA's own tally of the evidence found that of 38 supplement studies, only 18 showed a statistically significant benefit, and the health claim it permits ends with the words "the evidence is inconsistent and inconclusive". The single best-powered hard-endpoint trial, 164 adults on 450 mg for 24 weeks, found no effect on arterial stiffness or blood pressure from any of three forms, even though citrate demonstrably raised blood magnesium.

Blood sugar. The American Diabetes Association's 2026 Standards of Care are explicit: supplementation with micronutrients such as magnesium is not recommended for glycaemic benefits. The evidence agrees. In the most rigorous synthesis, fasting glucose improved but HbA1c did not (-0.134, 95% CI -0.409 to 0.141, p=0.34). The effect has shrunk steadily as evidence accumulated, from a single 2003 trial to -0.15 percent across 78 trials in 2026, and a network meta-analysis of 178 studies does not list magnesium among the supplements effective for HbA1c at all.

The hinge is baseline status. A trial that recruited only people with confirmed low magnesium found real improvements. A double-blind crossover that recruited only people with normal magnesium found fasting glucose, HbA1c, insulin and HOMA-IR all unchanged, while urinary magnesium rose, so the null was not a compliance failure. For a well-nourished reader, that second trial is the relevant one.

On depression and anxiety, the World Federation of Societies of Biological Psychiatry and CANMAT taskforce grades magnesium (-), its lowest grade, Not Recommended. The most-cited anxiety review says in its own text that three of the four positive studies are unpublished and that all of them gave magnesium combined with other ingredients.

The 2023 joint guideline from the American Gastroenterological Association and the American College of Gastroenterology suggests magnesium oxide for chronic idiopathic constipation. It is a conditional recommendation at very low certainty, resting on two Japanese trials totalling 94 participants, 93 percent of them women.

Two things belong on the label of that recommendation. The dose studied was 1.5 g/day of magnesium oxide, which is about 905 mg of elemental magnesium, roughly 2.6 times the 350 mg upper limit for supplements. This is a pharmacological osmotic laxative effect, not a nutritional one. And the guideline says it directly: only magnesium oxide has been evaluated in randomized trials, and the bioavailability and clinical efficacy of citrate, glycinate, lactate, malate and sulfate for this purpose are unknown.

Magnesium supplements are compounds, and the magnesium is a fraction of the weight. That fraction is arithmetic, and it varies enormously. Our calculations from the molecular formulas: magnesium oxide is 60.3 percent magnesium, citrate is 11 to 16 percent depending on which of three different chemicals the label means, bisglycinate is 12 to 14 percent depending on hydration, and L-threonate is 8.25 percent.

Here is what that does to a shopping decision, from prices read on each brand's own site on 2026-08-22, with the cost arithmetic ours:

• Magnesium oxide, 250 mg elemental per tablet: about $1.86 a month, or $0.74 per 100 mg of magnesium.
• Magnesium citrate, 250 mg elemental: $14.19 a month, $5.68 per 100 mg.
• Magnesium bisglycinate, 200 mg elemental: $26.00 a month, $13.00 per 100 mg.
• Magnesium L-threonate, 144 mg elemental: $31.50 a month, $21.88 per 100 mg.

Oxide delivers 29 times more elemental magnesium per dollar than L-threonate. And the front of that L-threonate bottle advertises 2,000 mg while the supplement facts panel states 144 mg of magnesium. A six-cent oxide tablet contains more magnesium than a $1.05 three-capsule threonate serving.

Does the expensive form absorb better? Nobody has properly tested it. The entire premium-form category rests on five small studies from 1990 to 2003 that contradict each other, one of which is a two-page letter with no abstract, and one of which was null overall with the glycinate advantage coming from a four-person subgroup. Two later and larger studies found magnesium oxide performed best. In the newest four-way comparison, bisglycinate produced no significant rise in magnesium at any timepoint.

More to the point, the comparisons buyers care about have never been run. Glycinate versus citrate on any endpoint: zero randomized trials. Magnesium malate versus anything in humans: zero. Threonate versus any other form: zero. The words glycinate, threonate, malate, taurate and orotate do not appear once in the NIH's professional magnesium fact sheet.

The upper limit is 350 mg a day, and it applies only to supplements and medications, not to magnesium in food. It was set using diarrhea as the endpoint, which is why exceeding it is usually uncomfortable rather than dangerous. Risk rises with impaired kidney function, because the kidneys are how magnesium leaves the body.

Japan is the country to look at, because it treats magnesium oxide as a prescription laxative used by roughly 10 million people. In 2015 its regulator mandated new label text after 29 reported cases of hypermagnesemia, four of them fatal. The mandated wording states that serious outcomes occurred even when renal function was normal and the dose was at or below what is recommended, and requires dose reduction in older patients and periodic measurement of serum magnesium during prolonged use. Twenty-one of the 29 cases were in people over 65, and 22 were taking it for constipation. The regulator noted that hypermagnesemia often went undetected until serious outcomes such as loss of consciousness occurred.

By contrast, the current US guideline's entire magnesium safety content is to avoid it in renal insufficiency, with no eGFR threshold, no age threshold and no monitoring advice.

On drug interactions, take magnesium apart from these, per the FDA-approved labels: alendronate says wait at least half an hour before other oral medicines; doxycycline and tetracycline absorption is impaired by magnesium-containing antacids; ciprofloxacin says take it at least 2 hours before or 6 hours after. Separately, proton pump inhibitors can cause low magnesium: the esomeprazole label warns of symptomatic and asymptomatic hypomagnesemia after three months or more, sometimes with tetany, arrhythmias and seizures.

Eat the food. Magnesium is concentrated in leafy greens, legumes, nuts, seeds and whole grains, and one careful study found magnesium from almonds was just as bioavailable as from a soluble magnesium salt. The observational evidence linking higher magnesium intake to lower diabetes risk is genuinely strong, but it is dietary magnesium, which travels with those foods, and it has never been tested as a pill: no randomized trial has ever examined magnesium supplementation for preventing diabetes.

If you want to supplement anyway, it is cheap and mostly harmless below the upper limit, and the honest reason to take it is that your diet is short on those foods, not that it will fix your sleep. Buy on elemental milligrams per dollar, which means oxide or citrate, and ignore the form marketing until somebody runs the trial. If oxide upsets your stomach, that is the known trade-off, and citrate at roughly $5.68 per 100 mg is the reasonable next step.

And if you are taking magnesium for a specific problem, the honest ranking is: constipation yes, at laxative doses and ideally with a clinician if you have any kidney impairment; blood pressure maybe, if you are hypertensive and on medication; everything else no, on current evidence.

Before supplementing if you have reduced kidney function, and before using magnesium at laxative doses at any age. If you take a proton pump inhibitor long term, a diuretic, or digoxin, ask whether your magnesium should be checked, because those are situations where a genuine deficiency can develop. And if you are taking magnesium for insomnia, migraine or low mood, those all have treatments with far better evidence behind them.