One weak source, one honest verdict: suggestive but inconclusive, and only in the already-anxious. The single held source is a narrative systematic review (no pooling, no GRADE) whose authors rate the primary-study quality as poor, carrying a magnesium-manufacturer conflict of interest — so magnesium for anxiety is a low-warrant candidate lever, not a finding. Anxiety is a mental-health outcome, admitted here only through its physiological (HPA/stress) intersection with physical outcomes; it stays peripheral, and the Layer-1 ranking keeps it there. (this verdict lead is the wiki’s synthesis of the single source; the evidence and its discounts are tagged in the sections below.)
What the evidence says (and cannot say)
- Direction: «there is suggestive but inconclusive evidence for a beneficial effect of Mg supplementation in mild anxiety» (Boyle et al., 2017) — with the overall caveat that «the quality of the existing evidence is poor» (Boyle et al., 2017).
- Vote-count, not a magnitude: the review does not pool, so there is no effect size. The abstract’s narrative tally is 4/8 anxious-sample studies, 4/7 PMS-sample studies, and 1/2 hypertension studies positive, with no effect on postpartum anxiety (Boyle et al., 2017) — but the results text itself gives a different anxious-sample count: «Three of the eight studies which recruited samples based upon pre-existing levels of mild subjective anxiety reported positive effects» (Boyle et al., 2017). That internal 3-vs-4 discrepancy is itself a tell of loose counting; either way a vote-count over poor-quality, non-independent studies is weak evidence of direction and no evidence of size.
- Stratum is narrow: every study recruited an anxiety-vulnerable sample — «mildly anxious, premenstrual syndrome (PMS), postpartum status, and hypertension» (Boyle et al., 2017). There is no general-population evidence: the review says nothing about whether magnesium reduces anxiety in a reasonably-healthy person.
- The stress claim is essentially unevidenced: «No study administered a validated measure of subjective stress as an outcome.» (Boyle et al., 2017) The title’s “stress” rests on general well-being subscales only.
Why the warrant is low (four independent discounts)
- Co-supplementation confound — decisive. In the anxious stratum, «Those studies which reported positive effects all administered Mg in combination with additional ingredients (e.g., vitamin B6, extract of Hawthorn and California poppy). None of the studies examined the effects of the included ingredients in isolation.» (Boyle et al., 2017) So no positive result isolates magnesium; the effect could be the co-ingredient.
- Manufacturer conflict of interest. Two of the three authors received Sanofi funding for the systematic database search, Sanofi supplied three unpublished in-house Mg intervention studies, and the third author sits on the Sanofi Consumer Healthcare Advisory Board (Boyle et al., 2017); the internal-report codes (MAB6) point to the Mg + vitamin-B6 product line. The source does not state whether those three unpublished studies were positive — but manufacturer funding plus manufacturer-supplied unpublished data is a publication/selection-bias signal, layered on the co-supplementation confound — and two of the positive anxious-sample results were themselves unpublished Mg + vitamin-B6 RCTs of exactly this kind (Boyle et al., 2017).
- Large placebo response + weak controls. All placebo-controlled studies showed significant placebo effects; in one PMS dose-response trial the sorbitol placebo beat every magnesium dose; the anxiolytic-comparison studies had no placebo arm. Outcome = self-report anxiety scales, a surrogate; 6/8 anxious studies used HAM-A as both inclusion and outcome (regression-to-mean risk) (Boyle et al., 2017).
- No dose or form signal. «No clear dose effect of Mg emerges»; positive results at both 75 mg and 360 mg, the one systematic 200/350/500 mg trial null throughout; no consistent effect of magnesium form despite oxide’s low bioavailability (Boyle et al., 2017). So there is no actionable dose or form to recommend.
Mechanism vs evidence — the bottleneck is the evidence
The mechanistic rationale is real and human-corroborated in part: magnesium blocks NMDA-receptor hyperexcitability, supports GABAergic tone, and modulates the HPA axis (reducing central ACTH and peripheral cortisol responses) (Boyle et al., 2017). The authors are explicit that «It is the quality of the available evidence rather than the absence of a potential mechanism which has hindered convincing demonstration of such effects.» (Boyle et al., 2017) This is the physiological-intersection route that admits the outcome at all — but a plausible mechanism is directional weight, never an outcome finding, and whole-organism net effect is exactly what the trials failed to show cleanly. The HPA channel is the same one that connects magnesium to Stress Management and Cardiometabolic Health.
Gaps (G) — what would change the verdict
- ** a magnesium-alone (no co-ingredient), placebo-controlled RCT/SR on subjective anxiety** — the co-supplementation confound cannot be resolved from the held evidence.
- ** a trial recruiting Mg-depleted participants and entering baseline Mg status in the analysis** — «none of the reported studies specifically recruited samples depleted in Mg» (Boyle et al., 2017), so the correct-the-deficiency effect-modification hypothesis (route (b)) is untested, not supported.
- General-population evidence is absent — every stratum here was anxiety-vulnerable; whether magnesium moves anxiety in a reasonably-healthy person is unstudied (insufficient evidence, not a null).
Layer-1 placement
A small, low-certainty, non-substitutable-nothing lever for a peripheral (mental-health) outcome. For a reasonably-healthy person it does not rank: there is no general-population evidence and no isolated-magnesium effect. Distinct from the Dietary Magnesium and Cardiometabolic Outcomes cell (different outcome; do not read a cardiometabolic magnesium benefit as an anxiety benefit) and adjacent to but distinct from Depression and Modifiable Exposures (anxiety is not depression). Correcting a genuine dietary magnesium shortfall has its own justification on the cardiometabolic page; an anxiety-specific magnesium recommendation is not warranted by this evidence.