Magnesium

01 What this document is, and four things it is not

This article is a research review of magnesium: the physiology of the Mg2+ ion, the limits of status testing, the chemistry and human evidence for the salts sold as supplements, and the clinical claims that travel with those salts. It is written against a market that treats every ligand as a different medicine and every low-normal serum value as a disease.

Four things follow immediately.

First, this is not a treatment manual. Intravenous magnesium sulfate in pre-eclampsia is a hospital protocol with a randomised evidence base (Altman, Carroli, Duley, et al., 2002). Oral magnesium oxide as an osmotic laxative is a different object (Mori et al., 2019). Neither licences a daily wellness capsule.

Second, it is not a minerals-class article. A sibling title treats minerals as a class. This document is the magnesium deep dive. Speciation, elemental fraction, and outcome trials are argued here at the grain the class title cannot afford.

Third, it is not an endorsement of "best absorbed" marketing. Solubility and urinary recovery differ among salts (Lindberg, Zobitz, Poindexter, and Pak, 1990; Walker, Marakis, Christie, and Byng, 2003; Firoz and Graber, 2001). Those measurements do not, by themselves, rank clinical outcomes.

Fourth, it is not medical advice. This document recommends no use, dose, route, or schedule for any person.

02 The biologically active species is Mg2+

Magnesium is an alkaline-earth cation. PubChem records the element as CID 5462224, atomic weight 24.305 (PubChem, 2026). An adult body contains about 25 g, of which 50 to 60 percent is in bone and most of the remainder is intracellular in soft tissue. Less than 1 percent is in serum, and that fraction is tightly controlled (Institute of Medicine, 1997; de Baaij, Hoenderop, and Bindels, 2015; NIH ODS, 2026). Those compartment numbers are established. They are the reason a serum test is a poor store marker, not a reason to invent a second magnesium.

After an oral salt dissolves, the ligand and the ion part company. What the enterocyte and the kidney handle is Mg2+. The ligand may change solubility, osmolality, and gastrointestinal tolerability. It does not become a second drug unless a human trial of that specific salt, at a stated elemental load, shows an outcome the ion alone does not explain. That rule governs every later form section.

Intestinal uptake uses a saturable transcellular path and a paracellular path. Fine, Santa Ana, Porter, and Fordtran measured absorption from food and from supplements and found that fractional absorption falls as the ingested load rises (Fine, Santa Ana, Porter, and Fordtran, 1991). TRPM6 is the apical channel of the transcellular path in intestine and distal nephron (Voets et al., 2004; de Baaij, Hoenderop, and Bindels, 2015). Those facts are established as physiology. They do not say that a chelate "opens a unique transporter."

03 ATP, enzymes, and the jobs that are not optional

Magnesium is a cofactor for a large set of enzymes. The Institute of Medicine put the number above 300; de Baaij, Hoenderop, and Bindels treated the ion as required for ATP-using reactions, nucleic-acid chemistry, and a wide kinase and cyclase set (Institute of Medicine, 1997; de Baaij, Hoenderop, and Bindels, 2015; NIH ODS, 2026). The chemically important species in phosphoryl transfer is often Mg-ATP, not free ATP. That biochemistry is established.

Romani's review of cellular magnesium homeostasis describes buffered cytosolic free Mg2+, mitochondrial and endoplasmic stores, and regulated transporters rather than a freely mixing pool (Romani, 2011). Jahnen-Dechent and Ketteler's "magnesium basics" paper is a clinical-physiology primer on the same ion (Jahnen-Dechent and Ketteler, 2012). Those reviews are strongly supported as descriptions of mechanism. They are not trials of capsules.

The marketing move is to read the enzyme list as a warrant: because magnesium sits on ATP, a glycinate or a threonate must raise energy, calm nerves, or protect the heart. The enzyme list is real. The warrant is speculative until an outcome trial in the claimed population is produced.

04 Neuromuscular function, contraction, and nerve transmission

Magnesium antagonises calcium at several steps of excitation–contraction coupling and of synaptic transmission. Iseri's older phrase, "nature's physiological calcium channel blocker," is still the teaching shorthand (Institute of Medicine, 1997). de Baaij, Hoenderop, and Bindels review NMDA-receptor and calcium-channel effects as physiology, not as a sleep-product claim (de Baaij, Hoenderop, and Bindels, 2015). Grade for the ion's neuromuscular role: established. Grade for "this salt relaxes muscle better than that salt": speculative, unless a head-to-head human trial says otherwise.

Nerve transmission and cardiac conduction sit on the same calcium–potassium–magnesium relationship. Sodium–potassium ATPase requires magnesium. Cellular potassium wasting in magnesium depletion is a recognised clinical pairing (Institute of Medicine, 1997; de Baaij, Hoenderop, and Bindels, 2015). That pairing explains why frank hypomagnesemia is treated as a medical finding. It does not explain a capsule for a replete adult with a mid-reference serum value.

05 Bone, cardiovascular physiology, and glucose metabolism

Bone holds most of the body's magnesium. A fraction is exchangeable and can support the extracellular concentration (Institute of Medicine, 1997). Structural bone magnesium is not the same object as a serum number, and a supplement trial that does not measure bone is not a bone paper.

Cardiovascular physiology is where the enzyme list, the calcium-antagonist teaching, and the epidemiology most often get stacked. Dietary magnesium intake associates, in prospective cohorts, with lower rates of cardiovascular disease, type 2 diabetes, and all-cause mortality (Fang et al., 2016). That paper is strongly supported as a dose-response meta-analysis of cohorts. It is food-and-pattern evidence. People who eat magnesium eat plants, nuts, and less refined grain. Fang and colleagues did not randomise a salt.

Randomised evidence for pills is narrower. Zhang, Li, Del Gobbo, and colleagues, in a meta-analysis of double-blind placebo-controlled trials, found a modest blood-pressure reduction from magnesium supplements (Zhang et al., 2016). Kass, Weekes, and Carpenter reached a similar modest conclusion in an earlier meta-analysis (Kass, Weekes, and Carpenter, 2012). Modest is the correct adjective. Neither analysis is a licence to treat magnesium oxide as a cardiology drug.

Glucose metabolism has a parallel split. Veronese and colleagues meta-analysed double-blind trials of supplementation on glucose parameters in people with or at risk of diabetes (Veronese et al., 2016; Veronese et al., 2021). Guerrero-Romero and Rodriguez-Moran reported improved insulin sensitivity or reduced insulin resistance in non-diabetic subjects with low serum magnesium after oral supplementation (Guerrero-Romero and Rodriguez-Moran, 2004; Guerrero-Romero and Rodriguez-Moran, 2011). Those papers are emerging to strongly supported for the populations they enrolled. They are not a proof that a replete adult needs a premium chelate.

06 The kidney is the regulator

The kidney filters and reabsorbs magnesium. There is no meaningful tubular secretion. About 65 percent of filtered magnesium is reabsorbed in the loop of Henle and 20 to 30 percent in the proximal tubule (Institute of Medicine, 1997; de Baaij, Hoenderop, and Bindels, 2015). TRPM6 in the distal convoluted tubule is the fine-tuning step (Voets et al., 2004). Urinary magnesium falls to very low levels within a few days of experimental depletion (Institute of Medicine, 1997). That renal conservation is established.

The practical consequences are two, and they point in opposite directions.

When the gut delivers more magnesium than the intestine absorbs, the unabsorbed salt is an osmotic laxative. That is a gastrointestinal event, not a proof of superior "detox."

When the kidney cannot excrete a load — chronic kidney disease, and some acute settings — supplemental and pharmacologic magnesium can raise serum magnesium into a dangerous range. Jahnen-Dechent and Ketteler treat hypermagnesemia as a renal-excretion failure more than a dietary-food failure (Jahnen-Dechent and Ketteler, 2012). Grade for food magnesium as a hypermagnesemia cause in people with intact kidneys: low. Grade for supplement and laxative magnesium in reduced glomerular filtration: established caution.

Loop and thiazide diuretics increase urinary magnesium. Long-term proton-pump inhibitor use is associated with hypomagnesemia, including severe cases (Cundy and Dissanayake, 2008; Markovits et al., 2014). Those are drug-safety findings. They are not a wellness argument for the same molecule in people who are not on those drugs.

07 Dietary sources and reference intakes

Chlorophyll is a magnesium–porphyrin chelate, so green leaves are a food source. Unrefined grains, legumes, nuts, and seeds are the other dense sources. Refining that removes bran and germ removes magnesium (Institute of Medicine, 1997; NIH ODS, 2026). Tap and bottled water vary from about 1 mg/L to more than 120 mg/L (Azoulay, Garzon, and Eisenberg, 2001). Typical fractional absorption from mixed diets is about 30 to 40 percent in the ODS summary, and Fine's perfusion work showed the fraction falling as the load rose (NIH ODS, 2026; Fine, Santa Ana, Porter, and Fordtran, 1991).

The Food and Nutrition Board's Recommended Dietary Allowances for adults run from 310 mg to 420 mg per day by age and sex (Institute of Medicine, 1997; NIH ODS, 2026). Those values are established as reference intakes. They are not diagnoses.

The Tolerable Upper Intake Level of 350 mg per day applies to supplements and medications, not to food magnesium (Institute of Medicine, 1997; NIH ODS, 2026). The usual adverse effect that set the UL is osmotic diarrhoea from unabsorbed salt. Food magnesium has no UL of that kind because the limiting event is different. Collapsing the UL with the RDA — "do not exceed 350 mg" as if it were total intake — is the elemental-dose confusion this article is written against.

Surveys have long shown that many adults, especially women, sit below the EAR or RDA for magnesium from food (Institute of Medicine, 1997; NIH ODS, 2026). That is intake inadequacy relative to a reference. It is not a prevalence of clinical hypomagnesemia, and it is not a serum-test result.

08 What deficiency actually is

Frank hypomagnesemia is a laboratory and clinical finding: serum magnesium below about 0.75 mmol/L, often with neuromuscular irritability, arrhythmia risk, refractory hypokalemia, or hypocalcemia (Elin, 2010; de Baaij, Hoenderop, and Bindels, 2015; NIH ODS, 2026). Causes that are established include gastrointestinal losses, alcohol use with poor intake, loop and thiazide diuretics, and long-term proton-pump inhibitors (Cundy and Dissanayake, 2008; Markovits et al., 2014; Institute of Medicine, 1997).

"Subclinical magnesium deficiency" is a different claim. It usually means: intake below a reference, or a serum value inside the reference interval that some authors would like to raise, or a symptom cluster that magnesium might help. Workinger, Doyle, and Bortz reviewed how hard that claim is to lock with present tests (Workinger, Doyle, and Bortz, 2018). Costello, Smith, and colleagues argued that the lower bound of the serum reference interval itself may be set too low (Costello et al., 2016). Those papers are emerging as a debate about cut-points. They are not a licence to tell a person with a serum of 0.82 mmol/L that they are deficient.

09 Why serum magnesium is a limited test

Elin's assessment paper is the clean statement: serum total magnesium is what laboratories report; it does not track total-body or tissue magnesium (Elin, 2010). ODS repeats the same limit (NIH ODS, 2026). Witkowski, Hubert, and Mazur, in a systematic review of assessment methods, found no single satisfactory test (Witkowski, Hubert, and Mazur, 2011). Grade for serum as a detector of frank hypomagnesemia: established. Grade for serum as a store or intake marker in an otherwise well adult: poor, and treated here as a category error when used that way.

Ionized magnesium, erythrocyte magnesium, buccal-cell or EXAtest-style assays, and 24-hour urine each have a literature. None has replaced serum as a population or clinic standard, and several are analytically uneven (Witkowski, Hubert, and Mazur, 2011; Workinger, Doyle, and Bortz, 2018). A magnesium-loading or tolerance test — parenteral magnesium followed by urinary recovery — is considered by some reviewers the least-bad research tool for depletion and is not a consumer test (NIH ODS, 2026; Witkowski, Hubert, and Mazur, 2011).

10 Intracellular and tissue assessment

If serum is weak, the temptation is to buy a "cellular magnesium" test. Witkowski, Hubert, and Mazur's systematic review is the brake: erythrocyte, ionized, saliva, and urine methods were surveyed; none was satisfactory as a standalone status tool (Witkowski, Hubert, and Mazur, 2011). Workinger, Doyle, and Bortz reached the same practical conclusion from a methods-and-diagnosis angle (Workinger, Doyle, and Bortz, 2018). Muscle biopsy and 31P-NMR are research tools, not shelf products.

A 24-hour urine can show renal wasting versus conservation. It does not, by itself, diagnose a tissue deficit in a person eating a variable diet. The loading test remains a research procedure (NIH ODS, 2026). This document does not treat any commercial intracellular kit as established.

11 How to read a magnesium salt

Four numbers matter, in this order.

Elemental magnesium. Labels that say "400 mg magnesium citrate" without saying how many milligrams are magnesium are not comparable to a label that says "400 mg elemental magnesium as citrate." The first may be mostly citrate. Fine's absorption work and every bioavailability paper below are elemental-load experiments whether they say so or not (Fine, Santa Ana, Porter, and Fordtran, 1991; Ranade and Somberg, 2001).

Solubility. ODS states the governing qualitative rule: forms that dissolve well tend to be absorbed better than poorly soluble oxides and sulfates; aspartate, citrate, lactate, and chloride tend to outrank oxide and sulfate on bioavailability (NIH ODS, 2026). Lindberg, Walker, and Firoz are the human papers that make that sentence concrete for oxide versus citrate (Lindberg, Zobitz, Poindexter, and Pak, 1990; Walker, Marakis, Christie, and Byng, 2003; Firoz and Graber, 2001).

Laxative effect. Unabsorbed magnesium salt is an osmotic cathartic. Hydroxide, oxide, citrate at bowel-prep loads, and sulfate (Epsom salt) are used as laxatives on purpose. A salt that "doesn't bother the stomach" may simply be delivering less unabsorbed ion — or a smaller elemental load. Laxation is not a toxicity in the UL sense so much as the UL's defining adverse effect (Institute of Medicine, 1997).

Human pharmacokinetics versus outcome trials. Urinary or serum increments after a single salt are emerging to strongly supported for a handful of head-to-heads. Outcome trials that compare two salts on migraine, blood pressure, sleep, or anxiety at a matched elemental dose are scarce. Ranade and Somberg's review already warned that bioavailability data are uneven across salts (Ranade and Somberg, 2001). Schuchardt and Hahn updated the absorption file and did not convert it into an outcome league table (Schuchardt and Hahn, 2017).

12 Elemental content, calculated and checked

Elemental percentages below use IUPAC magnesium 24.305 and the conventional anhydrous (or stated hydrate) formula. They were calculated for this compilation and cross-checked against PubChem molecular weights where the CID resolved to the intended salt. Several marketed names resolve on PubChem to a hydrate, a fragment, or a different species; those mismatches are named. Hydration water lowers the percentage. A "citrate" label that does not name the hydrate is not a precise chemical claim.

PubChem CID 14792 confirms magnesium oxide, molecular weight 40.305 (PubChem, 2026). CID 84645 confirms magnesium glycinate, molecular weight 172.42, matching 14.10 percent. CID 6099958 for "magnesium citrate" returns 488.8, a hydrate or alternate stoichiometry, not the anhydrous trimagnesium dicitrate used in the table. CID 45051722 for a threonate query returned L-threonic acid (102.37), not the magnesium salt. CID 91826668 for acetyltaurinate returned 290.16, not the 360.66 implied by C8H16MgN2O8S2. Those CID mismatches are why the table leads with formulas, not with database nicknames.

A 100 mg elemental dose therefore requires about 166 mg of oxide and about 1.2 g of L-threonate salt. Price-per-bottle comparisons that ignore that ratio are not price-per-magnesium comparisons.

13 Oxide, hydroxide, carbonate, chloride, sulfate

Magnesium oxide is the high-percentage, low-solubility, inexpensive salt. Lindberg, Zobitz, Poindexter, and Pak compared citrate and oxide and found higher bioavailability from citrate (Lindberg, Zobitz, Poindexter, and Pak, 1990). Walker, Marakis, Christie, and Byng, in a randomised double-blind study, found citrate more bioavailable than oxide and than an amino-acid chelate (Walker, Marakis, Christie, and Byng, 2003). Firoz and Graber, sampling US commercial preparations, found oxide poorly bioavailable relative to other products (Firoz and Graber, 2001). Those papers are strongly supported as human recovery comparisons. They are not migraine or sleep trials.

Oxide is also the salt in Mori, Tomita, Fujimura, and colleagues' randomised, double-blind, placebo-controlled constipation trial, and in a later senna-versus-oxide constipation RCT (Mori et al., 2019; Morishita et al., 2021). In that job, poor absorption is the mechanism: unabsorbed salt holds water. Calling oxide "low quality" and then citing it as a laxative is two different uses of the same chemistry.

Magnesium hydroxide is the Milk of Magnesia salt. Elemental fraction 41.68 percent. It is an established osmotic laxative and antacid. It is not a bioavailability champion, and it does not need to be.

Magnesium carbonate is an antacid salt, 28.83 percent elemental, poorly soluble. Antacid use is labelled chemistry. Supplement-aisle "gentle magnesium" claims for carbonate are speculative.

Magnesium chloride is among the salts ODS groups as better absorbed than oxide (NIH ODS, 2026). Anhydrous material is 25.53 percent elemental; the hexahydrate, which is what many liquids and Slow-Mag-class products actually are, is 11.96 percent. Human outcome RCTs that isolate chloride from other soluble salts are thin. Liquid chloride products are often marketed as "high absorption" because they are already dissolved. Dissolved is step 1 in Figure 3.

Magnesium sulfate is two drugs under one name. Oral Epsom salt is a cathartic (heptahydrate 9.86 percent elemental). Intravenous magnesium sulfate is the hospital drug in the Magpie pre-eclampsia trial and in several arrhythmia and asthma protocols (Altman, Carroli, Duley, et al., 2002). Topical "transdermal" Epsom-salt claims are not supported by an absorption file that can carry a clinical outcome. Treating Magpie as evidence for a bath salt or a capsule is a route error.

14 Citrate, glycinate, lactate, malate, gluconate, aspartate

Magnesium citrate is the human bioavailability comparator that oxide keeps losing to (Lindberg, Zobitz, Poindexter, and Pak, 1990; Walker, Marakis, Christie, and Byng, 2003). Anhydrous trimagnesium dicitrate is 16.16 percent elemental; commercial hydrates are lower. Citrate is also a bowel-prep laxative at a different dose class. A 150 mg elemental citrate capsule and a citrate prep solution are not the same experiment. Grade for "citrate is better absorbed than oxide": strongly supported. Grade for "citrate is the best magnesium for blood pressure / sleep / anxiety": speculative as a salt-specific claim.

Magnesium glycinate / bisglycinate is 14.10 percent elemental on the bisglycinate formula (PubChem CID 84645). It is marketed for sleep, anxiety, and "no laxation." The glycine ligand has its own sleep literature; that literature is about glycine, not about a magnesium salt inheriting glycine's trials. A title search for human glycinate bioavailability returned no locked head-to-head. Walker included an amino-acid chelate that lost to citrate on recovery (Walker, Marakis, Christie, and Byng, 2003). Boyle, Lawton, and Dye's anxiety review mixed salts and did not isolate glycinate (Boyle, Lawton, and Dye, 2017). Grade for "glycinate is the sleep and anxiety magnesium": speculative. Grade for "glycinate is often better tolerated than oxide at a similar elemental load": plausible, as a laxation-and-load observation, not as a proven anxiolytic mechanism.

Magnesium lactate sits in the ODS "tends to absorb better than oxide" group (NIH ODS, 2026). Anhydrous 12.01 percent; dihydrate 10.19 percent. Salt-specific outcome trials are thin.

Magnesium malate is sold for "energy" and for fibromyalgia-adjacent fatigue. The hydrogen-malate formula used here is 15.54 percent elemental. PubChem CID 9833818 resolved to a 1533.8 molecular weight species and was discarded as the wrong object. Human PK and outcome evidence that isolate malate from other organic salts were not locked. Grade for malate-specific energy claims: speculative.

Magnesium gluconate is 5.86 percent elemental anhydrous — a bulky salt. Older US products used it. Low elemental density is a pill-burden fact, not a purity fact.

Magnesium aspartate is grouped by ODS with the better-absorbed organics (NIH ODS, 2026). Elemental fraction 8.66 percent on the formula used here. PubChem CID 165679 did not resolve. The aspartate ligand does not make this an "ATP salt" in any trial sense.

15 Taurate, orotate, L-threonate, acetyl-taurate

Magnesium taurate is 8.92 percent elemental on Mg(C2H6NO3S)2. The marketing sentence is that taurine is "for the heart," therefore the salt is a heart magnesium. Taurine has its own literature. This article does not import it. Locked human outcome trials of magnesium taurate as a named salt were not found at a standard that could carry a cardiovascular claim. Animal taurate papers exist and stay in the animal file.

Magnesium orotate is 7.27 percent elemental anhydrous. Stepura and Martynow reported the MACH study in severe congestive heart failure (Stepura and Martynow, 2009). That paper is a small heart-failure file, emerging at most, and is not a contemporary standard of care. "Orotate as a mineral transporter" remains a marketing theory.

Magnesium L-threonate is the salt built to raise cerebrospinal-fluid magnesium in animals. Slutsky, Abumaria, Wu, and colleagues reported enhanced learning and memory after elevating brain magnesium with this salt in rats (Slutsky et al., 2010). That paper is strongly supported as rodent neuroscience. It is not a human trial.

Liu, Weinger, Lu, Xue, and Sadeghpour reported MMFS-01, a magnesium L-threonate product, in a randomised, double-blind, placebo-controlled trial in older adults with cognitive complaints (Liu et al., 2016). That trial is the human cognition object. It is small, proprietary, and not a general-population brain-health licence. Grade: emerging for the trial's own composite in its enrolled sample. Grade for "this is the magnesium that crosses the blood–brain barrier in shoppers": speculative as consumer copy.

Wu, Jin, Ma, and colleagues later randomised magnesium L-threonate as an opioid-sparing adjunct in advanced cancer (Wu et al., 2023). That is a hospital analgesia question. It is not a cognition replication, and it is not a wellness use.

Magnesium acetyl-taurate (ATA-Mg, magnesium N-acetyltaurinate) is a low-percentage, high-price salt with a rodent synaptic-plasticity paper (Fassin, Danhier, and Ris, 2020) and a 2026 preclinical comparison claiming neurophysiological advantages over L-threonate (Kumar et al., 2026). Both are preclinical. No independent human outcome series was locked. Using either paper to sell a human "brain magnesium" is the threonate error with a new ligand.

Other marketed names — "microencapsulated," "sustained-release," "oil-suspended," blended "calming" complexes — inherit the ion. They do not inherit a trial unless they have one.

16 Price, pill burden, and value

Oxide is cheap per elemental milligram and poorly absorbed, and it is the salt with the cleanest constipation RCT locked here (Mori et al., 2019). Citrate costs more, absorbs better in recovery studies, and is a laxative at high load. Glycinate and threonate cost the most per elemental milligram because the ligand is heavy and the brand story is heavy. Value is not "best absorbed." Value is whether the extra money buys a demonstrated outcome in the person in front of the label. For most wellness claims, it does not.

Coudray, Rambeau, Feillet-Coudray, and colleagues compared ten organic and inorganic salts in magnesium-depleted rats using stable isotopes (Coudray et al., 2005). The paper is useful as animal chemistry. It is not a human ranking and is not used here as one.

17 Deficiency correction

Repleting frank hypomagnesemia is medical care. The oral salts used in that setting are chosen for elemental load and tolerability, not for a brand story. This document does not specify a regimen. The evidence class is established as the indication and not transferable to a replete adult (Elin, 2010; de Baaij, Hoenderop, and Bindels, 2015).

PPI-associated hypomagnesemia can be severe and can recur if the drug continues (Cundy and Dissanayake, 2008; Markovits et al., 2014). That is a drug-safety file.

18 Constipation

This is the indication where "poorly absorbed" is the point. Mori and colleagues randomised magnesium oxide against placebo in chronic constipation and found a benefit on that endpoint (Mori et al., 2019). A later randomised comparison of senna versus magnesium oxide treated both as active laxatives (Morishita et al., 2021). Hydroxide and high-dose citrate belong to the same osmotic class. Grade: established as osmotic laxation. Grade as a "detox" or "gut healing" claim: speculative.

19 Migraine

Peikert, Wilimzig, and Köhne-Volland reported a prospective, multicentre, placebo-controlled, double-blind randomised study of oral magnesium for migraine prophylaxis (Peikert, Wilimzig, and Köhne-Volland, 1996). That trial is the historical oral object. Teigen and Boes, reviewing oral magnesium in preventive treatment, treated the file as mixed and methodologically uneven (Teigen and Boes, 2015). Chiu, Yeh, Huang, and Chen meta-analysed intravenous and oral magnesium for migraine and kept the routes separate (Chiu, Yeh, Huang, and Chen, 2016). Grade for oral magnesium as migraine prophylaxis: emerging to strongly supported in selected trial cohorts, not a proof for every salt. Grade for "glycinate is the migraine magnesium": speculative. Intravenous emergency treatment is a different route and a different evidence class.

20 Blood pressure

Zhang and colleagues' 2016 meta-analysis of double-blind placebo-controlled trials is the pill paper (Zhang et al., 2016). Kass, Weekes, and Carpenter is the earlier meta-analysis (Kass, Weekes, and Carpenter, 2012). Both find a small average systolic and diastolic reduction. Fang and colleagues' cohort meta-analysis is the food paper (Fang et al., 2016). The gap between those designs is the argument. Grade for a modest average blood-pressure effect of supplementation, especially where intake or status is low: strongly supported as a meta-analytic mean. Grade for a clinically decisive antihypertensive in replete adults: not shown. Grade for salt-specific superiority: not shown.

21 Glycemic control

Veronese and colleagues' 2016 and 2021 meta-analyses of double-blind trials are the pill papers for glucose parameters (Veronese et al., 2016; Veronese et al., 2021). Guerrero-Romero and Rodriguez-Moran's randomised studies in non-diabetic subjects with low serum magnesium are the individual-trial objects (Guerrero-Romero and Rodriguez-Moran, 2004; Guerrero-Romero and Rodriguez-Moran, 2011). Fang's cohorts associate dietary magnesium with lower type 2 diabetes incidence (Fang et al., 2016). Grade for a modest effect on selected glucose parameters in people with or at risk of diabetes, particularly if status is low: emerging to strongly supported. Grade for a diabetes drug substitute, or for a replete adult on a premium chelate: speculative.

22 Arrhythmia contexts

Intravenous magnesium is used in specific hospital arrhythmia and perioperative settings. That file is not reviewed here as a supplement file. Oral supplementation as a general anti-arrhythmic in outpatients was not locked to a trial that could carry the claim. de Baaij, Hoenderop, and Bindels review cardiac physiology; they do not convert it into a capsule indication (de Baaij, Hoenderop, and Bindels, 2015). Grade for oral magnesium as a consumer arrhythmia treatment: speculative. Frank hypomagnesemia with arrhythmia is a medical finding, not a supplement-aisle story.

23 Exercise, cramping, and muscle function

Garrison, Korownyk, Kolber, and colleagues' Cochrane review of magnesium for skeletal muscle cramps found little evidence of benefit in the populations studied (Garrison et al., 2020). That is strongly supported as a negative-leaning evidence synthesis. Exercise-associated cramp marketing that cites the ion's role in contraction, without a trial, is the enzyme-list error. IOM already recorded that magnesium supplements did not enhance marathon performance in a cited runner study (Institute of Medicine, 1997). Muscle function in frank deficiency is a different file.

24 Sleep and anxiety

Abbasi, Kimiagar, Sadeghniiat, and colleagues randomised magnesium in elderly subjects with primary insomnia (Abbasi et al., 2012). The trial is small. Mah and Pitre systematically reviewed oral magnesium for insomnia in older adults and found the evidence limited (Mah and Pitre, 2021). Boyle, Lawton, and Dye systematically reviewed subjective anxiety and stress and found a possible signal in vulnerable samples, with mixed salts and mixed quality (Boyle, Lawton, and Dye, 2017). Grade for a general sleep or anxiolytic indication in healthy adults: emerging at best, speculative as glycinate-specific marketing. Glycine is not glycinate.

25 Bone

Bone is the storehouse. Dietary magnesium and bone outcomes travel mostly in observational files and in the IOM bone-related DRI chapter (Institute of Medicine, 1997). A dedicated, adequately powered fracture RCT of a named oral salt was not locked. Grade for food-pattern associations: plausible to emerging. Grade for a supplement as an osteoporosis drug: not shown.

26 Pregnancy contexts

The Magpie Trial randomised magnesium sulfate against placebo in pre-eclampsia and found a reduction in eclampsia (Altman, Carroli, Duley, et al., 2002). That paper is established as hospital intravenous evidence. It is not an oral-prenatal-vitamin paper. Oral magnesium for pregnancy cramps sits inside the same Cochrane cramp file that is unimpressive (Garrison et al., 2020). Using Magpie to sell a capsule to a well pregnant woman is a route-and-indication error.

27 Safety, at the same volume as efficacy

Osmotic diarrhoea is the leading oral adverse effect and the basis of the supplemental UL (Institute of Medicine, 1997; NIH ODS, 2026). Grade: established.

Hypermagnesemia in reduced glomerular filtration, and after large cathartic loads, is established (Jahnen-Dechent and Ketteler, 2012). This is the kidney-disease safety file the market under-talks.

Drug-associated wasting (PPI, diuretics) is established as a cause of low magnesium, not as a reason for unsupervised high-dose supplementation (Cundy and Dissanayake, 2008; Markovits et al., 2014).

Replete adults. Meta-analytic blood-pressure and glucose effects are small and tend to sit in lower-status samples (Zhang et al., 2016; Veronese et al., 2021). Treating those means as a warrant for high-dose chelates in people who already eat nuts and greens is the repletion-versus-surplus error.

This document specifies no dose. Every milligram figure above is a reported study or reference-intake parameter attached to a named population.

28 Exhaustive form comparison

29 Head-to-head evidence matrix

30 Critical questions

Is there a "best absorbed magnesium"? There is a better-documented pair: citrate recovers more than oxide in human studies (Lindberg, Zobitz, Poindexter, and Pak, 1990; Walker, Marakis, Christie, and Byng, 2003; Firoz and Graber, 2001). ODS's qualitative ranking puts soluble organics and chloride above oxide and sulfate (NIH ODS, 2026). That is not a crown for glycinate, threonate, or acetyl-taurate, and it is not an outcome ranking. The slogan is rejected as a general claim.

Does magnesium L-threonate raise human brain magnesium and improve cognition as advertised? Slutsky and colleagues showed a rodent learning effect after raising brain magnesium (Slutsky et al., 2010). Liu and colleagues showed a small proprietary human cognition signal (Liu et al., 2016). Wu and colleagues showed an opioid-sparing signal in advanced cancer (Wu et al., 2023). Those are three different objects. Consumer brain-health copy treats them as one. The collapse is rejected.

Does glycinate treat sleep and anxiety because it is glycinate? Boyle, Lawton, and Dye did not isolate the salt (Boyle, Lawton, and Dye, 2017). Abbasi and Mah did not make a glycinate claim (Abbasi et al., 2012; Mah and Pitre, 2021). Walker’s chelate lost a recovery contest to citrate (Walker, Marakis, Christie, and Byng, 2003). The marketing sentence is speculative.

Are most people magnesium deficient? Many adults eat less than the EAR or RDA (Institute of Medicine, 1997; NIH ODS, 2026). That is not the prevalence of hypomagnesemia. Serum surveys and intake surveys are different instruments. The collapsed prevalence claim is rejected.

Does a normal serum magnesium mean stores are full? No (Elin, 2010; Witkowski, Hubert, and Mazur, 2011; Workinger, Doyle, and Bortz, 2018). Does a mid-reference serum mean a person should supplement? That does not follow. Costello's cut-point argument changes how one reads the lower bound; it does not diagnose an individual (Costello et al., 2016).

Do supplements work in replete individuals? Average effects in blood-pressure and glucose metas are small and sit closer to low-status samples (Zhang et al., 2016; Veronese et al., 2021). Surplus supplementation in people who already meet the RDA from food is not the experiment those papers were built to win.

Is kidney disease a safety issue? Yes. Reduced excretion plus supplemental or laxative magnesium is the hypermagnesemia setting (Jahnen-Dechent and Ketteler, 2012; Institute of Medicine, 1997). Food magnesium is not the usual cause.

What is the elemental-dose confusion? Three mistakes travel together: reading milligrams of salt as milligrams of magnesium; reading the 350 mg supplemental UL as a total-intake cap; and reading a high oxide percentage as proof of a high absorbed dose. Fine's load–fraction curve and the oxide-versus-citrate recovery papers are the corrective (Fine, Santa Ana, Porter, and Fordtran, 1991; Lindberg, Zobitz, Poindexter, and Pak, 1990).

31 What is known, and what is sold

What is known is sharp. Magnesium is an essential intracellular cation with established roles in Mg-ATP chemistry, neuromuscular excitability, and renal conservation (de Baaij, Hoenderop, and Bindels, 2015; Institute of Medicine, 1997; Romani, 2011). Serum is a poor store test (Elin, 2010). Citrate recovers more than oxide in human bioavailability studies (Lindberg, Zobitz, Poindexter, and Pak, 1990; Walker, Marakis, Christie, and Byng, 2003). Oxide is an established osmotic laxative (Mori et al., 2019). Intravenous sulfate is an established pre-eclampsia drug (Altman, Carroli, Duley, et al., 2002). Pills have a modest average blood-pressure effect and a mixed, status-sensitive glucose file (Zhang et al., 2016; Veronese et al., 2021). Cramps, sleep, anxiety, and "brain magnesium" are thinner than their shelf talk (Garrison et al., 2020; Mah and Pitre, 2021; Liu et al., 2016).

What is sold is a collapse. One ion, fifteen salts, a "best absorbed" crown, a sleep glycinate, a brain threonate, and a prevalence statistic that confuses the EAR with the ICU. The ligand is real chemistry. The crown is not.

A reader who wants a single sentence can have it. Magnesium is an established essential ion; citrate is strongly supported as better recovered than oxide; oxide is established as a laxative salt; hospital magnesium sulfate is established in pre-eclampsia and is not an oral-capsule trial; blood-pressure and glucose pill effects are modest and status-sensitive; glycinate sleep claims and L-threonate brain claims remain emerging at best and speculative as general consumer copy; and a serum number in the middle of the reference interval is not a store assay (de Baaij, Hoenderop, and Bindels, 2015; Lindberg, Zobitz, Poindexter, and Pak, 1990; Mori et al., 2019; Altman, Carroli, Duley, et al., 2002; Zhang et al., 2016; Elin, 2010; Liu et al., 2016; NIH ODS, 2026).

The extra length is the point. Separate the fifteen salts from the single ion, the sleep and brain claims from their preferred forms, and the population deficiency statistic from the intensive-care one, and magnesium is an essential mineral with real repletion value under a marketing layer its trials do not underwrite.

32 References

33 Evidence handling

Findings are labelled in the reporting sentence by design: randomised trial, meta-analysis of trials, prospective cohort meta-analysis, pharmacokinetic comparison, systematic review of methods, regulator or DRI instrument, or database record. Animal and in-vitro results are named as such and are not used to imply a human outcome.

Conflicting evidence is kept in the same section. Fang's food cohorts and Zhang's pill meta-analysis are not averaged. Slutsky's rats and Liu's MMFS-01 trial are not stacked into one brain claim. Peikert and Teigen are not forced into a single grade.

Elemental percentages were calculated from conventional formulas and IUPAC magnesium 24.305, then checked against PubChem where the CID resolved to the intended salt. CID mismatches were discarded as authorities for the percentage.

Quantitative claims were locked to verified NCBI records or to the IOM DRI chapter and the NIH ODS magnesium fact sheet fetched for this build. Candidate PubMed identifiers that resolved to a different paper were discarded and not cited.

References are generated from those verified records plus the non-PubMed database and DRI sources listed with them. Internal production logs are not part of this apparatus.