Magnesium deficiency and long-term incident dementia among COVID-19 survivors: a propensity score-matched cohort study.

BACKGROUND

COVID-19 survivors face an elevated risk of long-term cognitive decline, yet potentially modifiable factors that shape post-COVID dementia trajectories remain insufficiently characterized. Low magnesium status has been linked to dementia risk in general populations, but whether low magnesium status is associated with long-term dementia risk among COVID-19 survivors remains unclear.

METHODS

This propensity score-matched retrospective cohort study used the TriNetX Global Collaborative Network to examine the association between low serum magnesium ( < 1.7 mg/dL on two measurements) and incident dementia among hospitalized COVID-19 survivors aged ≥ 55 years (2020-2023). A 1-year landmark period was imposed to reinforce temporal separation between exposure and outcome. The primary outcome was overall incident dementia over 5 years. Secondary outcomes included Alzheimer's disease, vascular dementia, other dementia types, and mild cognitive impairment.

RESULTS

After matching, 58,991 patients with low magnesium and 58,991 controls were included. Magnesium deficiency was associated with a higher risk of overall dementia [2.12% vs. 1.77%; hazard ratio (HR) 1.39, 95% confidence interval (CI) 1.28-1.51; P < 0.001; E-value 2.13). Among secondary outcomes, low magnesium was associated with higher risks of Alzheimer's disease (HR 1.34; P = 0.005), vascular dementia (HR 1.50; P = 0.001), other dementia types (HR 1.38; P < 0.001), and mild cognitive impairment (HR 1.27; P < 0.001). The association between magnesium deficiency and overall incident dementia persisted across sensitivity analyses and sex-stratified subgroups.

CONCLUSION

In this observational study, low serum magnesium was associated with increased long-term dementia risk among COVID-19 survivors. Because eligibility required repeated magnesium testing and follow-up healthcare visits were marginally more frequent in the low-magnesium group, selection bias and differential dementia detection cannot be excluded. These hypothesis-generating findings warrant prospective validation.