Low blood magnesium, and high blood magnesium, are both linked to higher dementia risk

Magnesium does the kind of work that never makes a headline on its own: it is a cofactor for several hundred enzyme systems, involved in nerve signaling, muscle contraction, blood glucose control and blood pressure regulation. A large new records study gives it one anyway, attaching a number to something researchers have suspected since at least 2017: that the body's blood magnesium level, at either extreme, tracks with dementia risk.

The study

Researchers led by Ying-Jen Chang, with coauthors at Chi Mei Medical Center, Chia Nan University of Pharmacy and Science, and E-Da Hospital at I-Shou University in Taiwan, published a retrospective, propensity-score-matched cohort study in Frontiers in Nutrition in 2026. They drew on the TriNetX Global Collaborative Network, a database built from US health system electronic health records, covering 2016 to 2023.

The study identified adults aged 50 and older who had two separate serum magnesium readings below 1.70 mg/dL, taken within a defined window, calling this recurrent low magnesium status. Each of these patients was matched one-to-one, by propensity score, against a patient with two readings in the normal range (1.70 to 2.20 mg/dL). After matching, 162,936 adults were in each group, 325,872 people in total. A one-year landmark period was applied before the researchers started counting new dementia diagnoses, so that dementia already underway at the time of the magnesium reading would not be miscounted as a new outcome, and the two groups were then followed for new diagnoses.

The finding

Adults with recurrent low magnesium had a 33% higher rate of new, all-cause dementia diagnoses than the matched normal-magnesium group: a hazard ratio of 1.33 (95% CI 1.26 to 1.40, p < 0.001). The association was stronger for one dementia subtype specifically, vascular dementia, where the hazard ratio was 1.63 (95% CI 1.42 to 1.88, p < 0.001). The same low-magnesium group also showed higher rates of stroke and of all-cause mortality over the same follow-up.

The researchers ran a separate, exploratory comparison for the opposite extreme: adults with two readings above 2.20 mg/dL, the high end of normal. That group also showed a higher rate of dementia diagnoses than the matched normal group, hazard ratio 1.25, and of vascular dementia specifically, hazard ratio 1.40, though it did not reach statistical significance for Alzheimer's disease as a subtype on its own. Put together, the pattern is U-shaped rather than a straight line: both extremes of blood magnesium, not just the low one, tracked with more dementia diagnoses than the middle of the range.

What this design can and cannot tell you

This is a retrospective analysis of real-world electronic health records, not a trial that assigned people to different magnesium levels. Propensity-score matching balances the two groups on measured characteristics, but it cannot balance them on what a chart does not record, and people who get two low magnesium readings in a US health system differ from people who do not in ways beyond magnesium itself: more medical contact in general, kidney and digestive disease, alcohol use, diuretic use, poor nutrition from other causes. Any of those could independently raise dementia risk and explain part of the association. The authors themselves describe the relationship as requiring prospective confirmation, not as established causation.

It is also worth being specific about how "low" was defined. A reading below 1.70 mg/dL on two separate occasions is a real clinical cutoff, but routine checkups do not typically include a magnesium test at all, so this cohort is not a random slice of the population, it is a slice of people who had their magnesium checked, generally because something prompted it.

This is not the first time this exact U-shaped pattern has shown up. A 2017 Rotterdam Study analysis in the journal Neurology, following 9,569 adults for about eight years, found both the lowest and highest serum magnesium groups ran roughly 30% higher dementia risk than the middle group. That study's lead author was explicit that the finding showed an association, not a cause. The new study is larger and more recent, and lands in the same place.

Why magnesium status might matter for the brain

Magnesium regulates calcium flow into brain cells through NMDA receptors, limiting the kind of chronic low-level overstimulation that has been linked to neuronal damage over time. It also plays a role in vascular tone and in insulin sensitivity, which may explain why the association in this study was strongest for vascular dementia specifically, a type of cognitive decline tied to blood vessel damage in the brain, rather than for Alzheimer's disease. None of this is confirmed as the mechanism behind the new study's numbers; it is the biologically plausible reason researchers keep looking at this nutrient rather than a demonstrated causal chain.

Where magnesium actually comes from

Amounts below are per realistic serving, ordered by density, from the NIH Office of Dietary Supplements:

• Pumpkin seeds, roasted, 1 oz (about 28 g): 156 mg
• Chia seeds, 1 oz (2 tbsp): 111 mg
• Almonds, dry roasted, 1 oz (about 23 almonds): 80 mg
• Spinach, boiled, ½ cup: 78 mg
• Cashews, dry roasted, 1 oz: 74 mg
• Peanuts, oil roasted, ¼ cup: 63 mg
• Soymilk, plain or vanilla, 1 cup: 61 mg
• Black beans, cooked, ½ cup: 60 mg
• Edamame, shelled, cooked, ½ cup: 50 mg
• Peanut butter, smooth, 2 tbsp: 49 mg
• Potato, baked with skin, 3.5 oz: 43 mg
• Rice, brown, cooked, ½ cup: 42 mg
• Yogurt, plain, low-fat, 8 oz: 42 mg
• Breakfast cereal, fortified, 1 serving: about 40 mg
• Oatmeal, instant, 1 packet: 36 mg
• Kidney beans, canned, ½ cup: 35 mg
• Banana, 1 medium: 32 mg
• Salmon, Atlantic, farmed, cooked, 3 oz: 26 mg
• Milk, 1 cup: 24 to 27 mg
• Avocado, ½ cup cubed: 22 mg
• Chicken breast, roasted, 3 oz: 22 mg
• Bread, whole wheat, 1 slice: 23 mg

How much you need

The US RDA for magnesium, from the National Academies:

• Men 19 to 30: 400 mg; men 31 and up: 420 mg
• Women 19 to 30: 310 mg; women 31 and up: 320 mg
• Pregnancy: 350 to 360 mg, depending on age
• Breastfeeding: 310 to 320 mg, depending on age

The tolerable upper intake level, 350 mg a day, applies only to magnesium from supplements and medications, not to magnesium from food. Dietary magnesium in excess of what the body needs is excreted by healthy kidneys, which is why there is no food-based upper limit, but supplemental doses above that figure commonly cause diarrhea and cramping before anything more serious.

Too little, and too much

Too little builds gradually. Early signs are loss of appetite, nausea, vomiting, fatigue and weakness. Left uncorrected, it can progress to numbness, tingling, muscle contractions and cramps, seizures, personality changes and abnormal heart rhythms. Risk is elevated with gastrointestinal diseases that cause chronic diarrhea or malabsorption, such as Crohn's disease and celiac disease, with poorly controlled type 2 diabetes, with long-term heavy alcohol use, and with some diuretics and proton pump inhibitors.

Too much from food alone is not a real risk in someone with functioning kidneys, since the kidneys clear the excess. Hypermagnesemia, generally defined as a blood level above about 2.6 mg/dL, is almost always caused by high-dose supplements or magnesium-containing laxatives and antacids combined with reduced kidney function, not by diet. Early symptoms are diarrhea, nausea and cramping; at much higher levels, the StatPearls clinical reference describes loss of deep tendon reflexes, heart block, respiratory paralysis and cardiac arrest, which is a medication-overdose picture, not something ordinary eating produces.

Tracking it in AIM

AIM does not track magnesium as standard, so add it as a custom metric: give it a name, a unit (mg) and a daily target of 420 mg for men or 320 mg for women (350 to 360 mg during pregnancy, 310 to 320 mg while breastfeeding), and it is estimated on every meal from then on, pinned to your cards or left behind "more". Any custom nutrient works the same way, with no target required if you would rather just watch the number.

Estimates from a photograph carry real uncertainty for micronutrients, because content varies with soil, cultivar, storage and cooking in ways no image contains. That is true of magnesium specifically: it varies with soil mineral content more than most nutrients on this list, so a log is a reasonable trend line, not a lab value. If you have a reason to suspect your level is actually abnormal, high or low, that is a blood test and a conversation with a doctor, not something a food log can substitute for.