Do magnesium supplements improve sleep, anxiety, and blood pressure? Decoding the claim

Magnesium is marketed for an unusually wide range of complaints at once: poor sleep, everyday anxiety, and high blood pressure are the three most common reasons adults reach for it without a doctor telling them to. The pitch draws on something real, magnesium is a cofactor in more than 300 enzyme systems governing muscle, nerve, and cardiovascular function, according to the NIH Office of Dietary Supplements. But a role in hundreds of enzyme reactions is not the same as a proven effect on any one of these three symptoms in a person whose magnesium levels are already normal, and the trial evidence behind each of the three claims stands on very different ground.

The blood pressure claim: real, but small and concentrated in specific people

The largest, most direct test is Zhang and colleagues’ 2016 meta-analysis in Hypertension, which pooled 34 randomized, double-blind, placebo-controlled trials in 2,028 participants. At a median dose of 368 mg per day for a median of three months, magnesium supplementation produced a real but modest reduction in blood pressure: systolic down 2.00 mmHg, diastolic down 1.78 mmHg, compared with placebo. A newer and larger 2025 meta-analysis in the same journal (Argeros and colleagues), pooling 38 trials in 2,709 participants, found a similar overall effect, systolic down 2.81 mmHg and diastolic down 2.05 mmHg, but with an important qualifier: the benefit was concentrated almost entirely in people already being treated for hypertension or with documented low magnesium levels, where systolic reductions reached 6 to 8 mmHg. In normotensive people with ordinary magnesium status, the reduction did not reach statistical significance. The NIH Office of Dietary Supplements summarizes the same pattern plainly, noting that magnesium supplementation “only marginally lowers blood pressure” and citing a Cochrane review that found just a 2.2 mmHg reduction in diastolic pressure. This is a real, replicated signal, not a null result, but it is a small effect that appears to depend heavily on starting from a hypertensive or magnesium-deficient baseline, not a blanket blood-pressure benefit for anyone who takes a supplement.

The sleep claim: a real but very thin evidence base

The best available synthesis on magnesium and sleep is a 2021 systematic review and meta-analysis by Mah and Pitre, which found only three randomized controlled trials meeting their criteria, totaling 151 older adults with insomnia. Pooling those three trials, magnesium supplementation reduced sleep onset latency by 17.36 minutes compared with placebo, a large-sounding number from a very small evidence base. The authors themselves rated the underlying trials as carrying moderate-to-high risk of bias, and graded the quality of evidence as low to very low using GRADE, the standard framework for weighing trial quality. Total sleep time improved by roughly 16 minutes in the magnesium group, but that result did not reach statistical significance. Three small, methodologically flawed trials pointing the same direction is a real signal worth further study, not proof of a reliable sleep benefit, and this evidence base has not been meaningfully expanded by any large, well-controlled trial since.

The anxiety claim: the thinnest of the three

Boyle, Lawton, and Dye’s 2017 systematic review in Nutrients reviewed 18 studies on magnesium and subjective anxiety or stress and reached a blunt conclusion: “the quality of studies was generally poor,” with issues including inappropriate sample selection, missing placebo controls, and weak statistical analysis. Where a benefit did appear, it clustered in samples specifically selected for anxiety vulnerability, such as premenstrual syndrome or pre-existing mild anxiety, rather than the general population, and the reviewers flagged large placebo responses that made it difficult to isolate any effect specific to magnesium itself. Their own stated conclusion was that “well-designed randomised controlled trials are required” before any recommendation could be made. Of the three claims examined here, this is the one with the least to stand on.

Deficiency correction is a different question than supplementing a replete person

The trials above almost universally tested magnesium in people who were not screened for, or confirmed to have, low magnesium status, which is precisely why the 2025 blood-pressure meta-analysis found such a striking difference between hypomagnesemic and normal-status participants. The NIH Office of Dietary Supplements sets the recommended dietary allowance for adults at 400 to 420 mg per day for men and 310 to 320 mg per day for women, and identifies specific groups at elevated risk of a true deficiency, including people with gastrointestinal disease, type 2 diabetes, alcohol dependence, and older adults with reduced dietary variety. Whether a given person’s intake, health conditions, or symptoms indicate an actual magnesium shortfall, as opposed to a normal level being topped up in hopes of a symptom benefit that the trial evidence does not clearly support, is a question for a clinician working from an actual dietary history and, where indicated, lab values, not something a supplement label can determine.

The honest bottom line

Magnesium’s three headline claims do not all rest on the same footing. The blood pressure effect is the most solidly replicated across two independently run meta-analyses eight years apart, but it is small in the general population and concentrated in people who are hypertensive or already magnesium-deficient. The sleep effect rests on only three small trials of low-to-very-low quality evidence, real enough to note but far from settled. The anxiety effect is the weakest of the three, described by its own reviewers as poor-quality evidence concentrated in vulnerable subgroups with substantial placebo response. That uneven pattern, one modest and reasonably confirmed effect, one thin but plausible signal, and one largely unproven claim, is why magnesium lands as a mixed verdict rather than a clean yes across the board.