Magnesium,
does it really help with sleep improvement and eye-twitching relief?

Useful facts when choosing a product

• Within the confirmed scope, under Korean MFDS standards magnesium is a generic nutrient (mineral), not an individually recognized functional ingredient. The recognized functionality wording is twofold: (1) 'needed for energy use' and (2) 'needed to maintain nerve and muscle function.' Thus the recognized expression is only 'needed for maintenance,' not improvement, relief, or treatment of diseases or symptoms.
• Within the confirmed scope, 'sleep quality improvement' functionality is granted not to magnesium but to separate ingredients such as L-glutamic-acid-fermented GABA powder, Ecklonia cava extract, lime peel extract, rice-bran alcohol extract, ashwagandha extract, and milk-protein hydrolysate (Lactium), and magnesium is not on the sleep-health functionality ingredient list. 'Eye-twitching relief' was not confirmed as any generic or individually recognized functionality. MFDS notice/recognition list originals are follow-up items.
• Public data indicate MFDS daily intake recommendation is about 94.5-250mg and the upper limit (UL) about 350mg. The positive deficiency trial Abbasi 2012 used elemental magnesium 500mg/day for 8 weeks, exceeding that upper limit and differing from ordinary market labeled doses (specific numeric comparison is a follow-up item).
• Absorption and dose differ by form: magnesium oxide (MgO) is reported to have low bioavailability, and negative trials such as Gholizadeh 2022 generally used low-dose MgO. Citrate, bisglycinate, and L-threonate are marketed as better absorbed, but in confirmed trials better absorption did not prove sleep or eye-twitching efficacy (absorption and efficacy are separate).
• According to public Severance Hospital guidance, regarding the common belief that under-eye twitching is due to magnesium deficiency, under-eye twitching after accumulated fatigue or excess caffeine is usually temporary, and if it persists or spreads, neurologic disorders such as hemifacial spasm should be considered (original URL comparison is a follow-up item).
• High-dose magnesium, especially oxide or citrate, can cause osmotic diarrhea, and people with impaired kidney function are known to have hypermagnesemia risk. Serious adverse events within the research dose ranges were rarely reported in the confirmed trials.

What the research actually shows

Based on 9 verified papers, 6 adopted as citations and the other 3 used as narrative support. [Sleep] ① Meta-analysis (Mah & Pitre 2021, BMC Complement Med Ther, PMID 33865376): 3 RCTs in older adults with insomnia (N=151), sleep latency was significantly shorter than placebo by -17.36 minutes (95% CI -27.27~-7.44, p=0.0006), but total sleep time +16.06 minutes was nonsignificant (p=0.15). Authors concluded that all trials had medium-to-high risk of bias, evidence quality low to very low (GRADE), and no definitive recommendation was possible. No funding-source subgroup analysis (A2 flag). ② Abbasi 2012 (J Res Med Sci, PMID 23853635): relatively deficient older insomnia patients with dietary Mg RDA<75% + serum Mg<0.95mmol/L received elemental Mg 500mg/day for 8 weeks, improving sleep time, efficiency, ISI, and latency significantly, while early-morning awakening was nonsignificant. Small deficiency-correction context; manufacturer provision of study product unconfirmed. ③ Nielsen 2010 (Magnes Res, PMID 21199787): age 51+ with PSQI>5, magnesium citrate 320mg/day for 7 weeks; PSQI improved in both groups (10.4->6.6) but there was no between-group difference, and the author explicitly noted possible placebo effect. Magnesium's demonstrated effects were limited to low magnesium and inflammatory markers. Funding/study-material suppliers were not visible on checked pages, so independence cannot be determined (A1 caution). ④ Schuster 2025 (Nat Sci Sleep, PMID 40918053): healthy poor sleepers, not deficient, elemental magnesium bisglycinate 250mg/day for 4 weeks; ISI Mg -3.9 vs placebo -2.3 (p=0.049), but Cohen's d=0.2 was 'small,' all markers were self-reported, and author conflicts were disclosed. ⑤ Lopresti 2026 (Front Nutr, PMID 41601871): adults, L-threonate (Magtein) 2g/day, equivalent to about 145mg elemental Mg, for 6 weeks; primary endpoint was cognition, not sleep, and objective Oura sleep markers total sleep, efficiency, and latency all showed no between-group differences, while PROMIS sleep disturbance also did not differ (only PROMIS sleep-related impairment p=0.043). Threotech Inc. funded and supplied product. ⑥ Rawji 2024 systematic review (Cureus, PMID 38817505): among 8 sleep studies, 5 improved, 2 null, 1 mixed, and usefulness may be greater in people with low baseline magnesium. ⑦ Gholizadeh-Moghaddam 2022 (Health Sci Rep, PMID 36620514): 64 PCOS women, MgO 250mg/day for 10 weeks, PSQI between-group p=0.85 null (low-dose MgO alone). ⑧ Zhang 2022 CARDIA (Sleep, DOI 10.1093/sleep/zsab276): observational study, highest dietary Mg quartile had lower short-sleep risk (OR 0.64, 95% CI 0.51~0.81), sleep quality borderline (OR 1.23, p trend 0.051), authors said objective-marker RCTs are needed to establish causality. [Eye twitching] ⑨ Gunes 2024 (Cureus, PMID 39282492): eyelid myokymia patients 103 vs controls 103; blood electrolytes including Mg did not differ between groups (p>0.05), and the actual associated factor was screen time (6.88 vs 4.84 hours/day, p<0.001, r=0.670). No magnesium intervention trial was found (no evidence). Note: ⑥⑦⑧ were used only as narrative support and not among the 6 adopted citations, so separate tracking is limited.

Why this is classified as D

Because this is a compound claim, endpoints were scored separately and then combined (methodology v0.4 rule ③ for compound claims). [Sleep] Human RCTs and meta-analyses exist, so by evidence type alone it is not D. However, under methodology 2-1① and ②: ① primary endpoint priority: in the trial measuring objective sleep markers (Lopresti 2026), total sleep, efficiency, and latency all showed no between-group difference, total sleep time in meta-analysis was also nonsignificant, and significant findings were mainly self-reported surrogate markers such as sleep latency, ISI, and PSQI with small effect sizes (sleep alone has at most C-level evidence). ② independence priority: in the conflict-of-interest-zero/high-quality layer, effects in general adults are not confirmed. The only general-adult trial measuring objective markers (Lopresti) was null and manufacturer-funded; positive trials (Schuster, Lopresti for some self-report items) carry strong conflicts; Nielsen 2010 had no between-group difference and funding/material sources are not confirmed, so it cannot be used as independent positive evidence. Strong positive signals concentrate in deficiency-correction contexts (Abbasi, low-magnesium subgroup, manufacturer provision unknown) and cannot be generalized to general adults. Methodology ② prioritizes the conflict-of-interest-zero layer rather than the whole pool, so absence of confirmation there leads to D. [Eye twitching] There are 0 intervention trials giving magnesium for eye twitching, and the only observational study does not support a magnesium-eye twitching association. Under ③ this no-evidence endpoint further lowers the compound claim. [Overall] Sleep is unconfirmed in the conflict-of-interest-zero layer and eye twitching has no intervention evidence, so the combined claim is D. The eye-twitching part is separately noted as practically 'not judgeable due to lack of literature.' MFDS-recognized functionality ('needed for maintenance') is neutral metadata, not an upward factor, and sleep and eye twitching are both outside magnesium's recognized functionality.

Counterpoint. [People arguing upward to C or higher] 'The meta-analysis found statistically significant sleep-latency shortening (p=0.0006), several RCTs exist, and a clear effect in deficient people has practical meaning. Since evidence conflicts, it is C, not D.' -> That signal is not denied. However, methodology 2-1② says that when evidence conflicts, the conflict-of-interest-zero/high-quality layer should be read first, and in that layer (objective-marker trials and independent trials), effects in general adults were not confirmed. Significant signals concentrate in self-reported surrogate markers, conflict-bearing trials, and deficient subgroups, so under the general-adult/independent layer it is D. [People arguing downward to F] 'Since general-adult/objective-marker trials were repeatedly null, it is effectively no effect (F).' -> Positive RCTs and significant meta-analysis signals in deficiency-correction contexts exist, so evidence is insufficient to declare 'disproved (F).' A conditional effect in specific subgroups such as deficient people may remain, so lack of evidence for a general-adult claim fits D. [Eye-twitching objection] 'Clinically, magnesium is recommended for eye twitching and some improve.' -> That clinical experience was not confirmed by verifiable intervention studies, and the confirmed observational study did not support a magnesium-eye twitching association. The grade can change if evidence appears.

Cross-check — Codex and Claude

Evidence Table

Receipt — 6 References

Every cited source was opened and checked against the live page on 2026-07-06.

What this document does and does not do

Chamgap is an information source. It reports what research has and has not confirmed; it does not tell readers what to take or buy. That decision belongs to readers and, when needed, medical or legal professionals. This verdict reflects literature available up to the search date and may change as new research appears. Nothing here is medical advice.