Magnesium benefits: what the human trials actually show

Blood pressure: a small effect, in specific groups

Magnesium is an essential mineral, and supplements are marketed for nearly everything. This review takes the claims one at a time and asks what human trials found. Blood pressure has the most consistent evidence of any of them.

A 2025 meta-analysis pooled 38 randomized trials in 2,709 people, each lasting at least four weeks. The median dose was 365 mg a day of elemental magnesium for a median of 12 weeks. Across everyone, systolic pressure fell by 2.81 mmHg and diastolic by 2.05 mmHg versus placebo (PMID 41000008).

The averages hide the useful detail. The reduction was larger in people already taking blood pressure medication (7.68 mmHg systolic) and in people with low blood magnesium (5.97 mmHg systolic). In people with normal blood pressure, the drop was not statistically significant. The authors found no dose-response relationship and flagged high variation between trials.

Evidence strength: moderate for a small average effect, weaker for any individual person. It is not a substitute for treating diagnosed hypertension.

Sleep: inconsistent, and rated low certainty

The sleep trials are small, use different doses and salts, and disagree. A 2026 systematic review of 12 randomized trials, described in our source material, found results that varied between questionnaires, EEG, actigraphy, and wearables. It rated certainty low to very low and did not support magnesium as a routine insomnia treatment.

The most encouraging single trial gave 500 mg a day to 46 older adults with diagnosed insomnia for eight weeks. Sleep-onset latency and sleep efficiency improved more than with placebo (PMID 23853635). A widely cited meta-analysis of three older trials, 151 people aged 59 to 75, found sleep onset about 17 minutes faster. Those trials used oxide or citrate, not glycinate, and total sleep time did not change significantly.

A 2025 trial of 250 mg a day of magnesium bisglycinate in adults with poor sleep found a marginal improvement on one score and nothing on the secondary measures. Two industry-funded L-threonate trials gave mixed results. In one, 1 g a day for three weeks improved sleep and daytime function versus placebo (PMID 39252819). In the other, 2 g a day for six weeks, every objective sleep-tracker measure showed no difference (PMID 41601871).

For scale, a meta-analysis of 19 melatonin trials found sleep onset about 7 minutes faster and about 8 minutes more total sleep (PMID 23691095). Nobody has compared the two head to head. Evidence strength: weak and inconsistent.

Migraine: a real signal, at high doses, in people with migraine

An older 12-week placebo-controlled trial gave 81 people with diagnosed migraine 600 mg a day of magnesium as trimagnesium dicitrate. Attack frequency fell significantly more than with placebo (41.6% versus 15.8% in the final month, per the source material). Attack duration and intensity did not change significantly, and 18.6% of participants had diarrhea (PMID 8792038).

Pooled analyses disagree on the details. A meta-analysis of three trials in 226 people found no significant reduction in severity and called the evidence low strength (PMID 30764890). A 2025 meta-analysis found fewer attacks, lower severity, and fewer monthly migraine days, without naming a single dose or form (PMID 39404918).

Another meta-analysis found that intravenous magnesium helped acute attacks and oral magnesium helped prevention. Neither result concerns everyday headaches, and neither isolates glycinate (PMID 26752497). Evidence strength: low to moderate, and only for people with diagnosed migraine.

Depression, anxiety, and stress: the placebo problem

For depression, the results swing with trial design. An open-label trial in 112 adults with mild-to-moderate depression used 248 mg a day of elemental magnesium as chloride for six weeks. It found a net 6.0-point improvement on the PHQ-9, but compared magnesium with no treatment at all (PMID 28654669).

A 2018 systematic review found a very large effect in trials without a placebo group (Hedges' g of -1.60) and a much smaller one in placebo-controlled trials (g of -0.21). It concluded the overall evidence is weak (PMID 29897029). A 2023 meta-analysis of seven placebo-controlled trials, 325 people, then found a large reduction in depression scores, though its authors asked for larger trials (PMID 38213402).

Anxiety is thinner. A systematic review of 18 studies found mixed results and rated the evidence poor. Every study recruited people already vulnerable to anxiety, such as those with mild anxiety, premenstrual syndrome, or hypertension (PMID 28445426). For stress, a trial of 300 mg a day in stressed, magnesium-low adults had no placebo arm. Adding 30 mg of vitamin B6 helped only the most severely stressed subgroup (PMID 30562392).

Evidence strength: weak for depression, poor for anxiety and stress.

Blood sugar, PCOS, and nerve damage: mixed and population-specific

A 2021 meta-analysis of 25 double-blind trials, summarized in our source material, found lower fasting glucose in people with diabetes. It also found better glucose tolerance and insulin-sensitivity markers in people at high risk. HbA1c, the longer-term marker, did not change significantly. Every participant had diabetes or elevated risk, so it says little about healthy people.

Dose may matter. In people with type 2 diabetes, about 1,000 mg a day of elemental magnesium raised magnesium levels and improved a blood sugar marker. About 500 mg a day did not differ from placebo (PMID 9589224). In women with PCOS, a 2025 meta-analysis of six trials found no significant effect on any cardiometabolic or hormonal marker (PMID 40005397).

For neuropathy, a five-year trial in 110 magnesium-depleted people with type 1 diabetes and existing nerve damage gave 300 mg a day of elemental magnesium or nothing. Nerve staging improved in 39% versus 8% and worsened in 12% versus 61%, but there was no placebo (PMID 15319143). A 2024 review of 40 studies on supplements for chronic nerve pain found only one tested magnesium, combined with calcium, and recommended no supplement (PMID 37654090).

Evidence strength: modest for glucose markers in diabetes, null in PCOS, and unreplicated for nerve damage.

Exercise performance, cramps, and restless legs: mostly null

A review of 12 human trials, using 116 to 500 mg a day, found that most evidence favored no effect on strength, aerobic, or anaerobic performance (PMID 10959930). A later review of 130 experiments across minerals rated magnesium's evidence base among the stronger ones. It still concluded there is little evidence that any mineral improves athletic performance (PMID 30909645).

On cramps, a 2020 Cochrane review (per the source material) covered 11 trials and 735 people. It judged magnesium unlikely to provide clinically meaningful prevention for idiopathic cramps in older adults. It found no randomized trial at all on exercise-related cramps. For restless legs syndrome, a systematic review found a single small randomized trial with no significant effect, and called the evidence insufficient either way (PMID 31678660).

Evidence strength: weak to negative.

Cholesterol, inflammation, weight, and bones: null findings

• Cholesterol: a 2017 meta-analysis of randomized trials found no significant effect on total cholesterol, LDL, HDL, or triglycerides (PMID 28180945). A 2025 dose-response analysis found the same for total cholesterol, LDL, and triglycerides, plus a small HDL rise (about 1.2 mg/dL) with extremely high heterogeneity (PMID 39905454).
• Inflammation: the largest meta-analysis found no significant effect on CRP, IL-6, or TNF-α at any dose or duration (PMID 34143369). In metabolic syndrome, a smaller analysis found a modest CRP drop sitting right at the edge of significance (p = 0.048) (PMID 41245414).
• Weight: 28 trials in 2,013 people showed no change in body weight, BMI, or body fat. Waist circumference fell about 2.09 cm in an obesity subgroup only (PMID 32718360).
• Bones: the evidence is observational. Higher intake tracked with slightly higher hip bone density in older adults, but with no lower fracture risk, and no supplementation trials were found (PMID 34666201; PMID 26556742).

Evidence strength: low for all four. These are fair claims to be skeptical about.

Other claims, and who should be careful

The oral asthma result should not be confused with hospital care. Intravenous and nebulized magnesium for severe attacks have meta-analytic support, while long-term oral supplementation does not improve control in adults (PMID 17218820).

On safety, severe magnesium excess is rare and occurs almost only when kidney function is already substantially reduced and intake from supplements, laxatives, or antacids is high (PMID 30220246). Reviews also warn that supplementing in advanced kidney disease needs careful monitoring (PMID 39256329). Magnesium can reduce absorption of some antibiotics and bisphosphonates (PMID 21942976) and of levothyroxine (PMID 41221788), so spacing doses matters.