Magnesium Intake and Depression in Adults

Emily K. Tarleton, MS, RD; Benjamin Littenberg, MD


J Am Board Fam Med. 2015;28(2):249-256. 

In This Article

Subjects and Methods

Data Source and Subjects

To investigate the question of whether there is an association between depression and magnesium intake, we conducted a cross-sectional study using the National Health and Nutritional Examination Survey (NHANES). NHANES participants undergo extensive interviews and laboratory assessments, including measures of dietary intake, dietary supplements, socioeconomic factors, clinical characteristics, and personal habits.[21] By applying a weighting scheme supplied by the Centers for Disease Control and Prevention, NHANES can be used to represent the sex-, age-, race-, and ethnicity-adjusted noninstitutionalized population of the United States. To increase the power of the analyses, we combined data from 2 separate waves of the survey (2007 to 2008 and 2009 to 2010).[22] We included all subjects at least 20 years old with complete data for the outcome, predictor, and all candidate confounders.


The main predictor variable was total magnesium intake in milligrams per day calculated from 24-hour dietary and supplement recall data. Intake was used because of the unreliability of serum magnesium concentrations[16] and because it is directly modifiable and could serve as an intervention. Low magnesium intake was defined as intake in the lowest quintile (<184 mg/day). Magnesium deficiency was defined with age- and sex-varying thresholds taken from the EAR as intake <350 mg/day for men >30 years old, <330 mg/day for younger men, <265 for women >30 years old, and <255 mg/day for younger women.[23]

The outcome variable was the score on the 9-item Patient Health Questionnaire (PHQ-9), a validated survey tool for measuring the presence and severity of depression in adults.[24] The PHQ-9 score is the sum of the responses to 9 items representing symptoms of depression. Each is graded by the patient according to how often they have experienced the symptoms over the previous 2 weeks, from 0 (not at all) to 3 (nearly every day). PHQ-9 scores range from 0 to 27 and were dichotomized into depressed (PHQ-9 score of 5–27) or not (PHQ-9 score of 0–4).

Based on a review of the literature and our clinical experience, we considered age, sex, race, ethnicity, education, marital status, household income, food security, tobacco use, alcohol intake, diabetes, kidney disease, and folate intake as potential confounders of the relationship between depression and magnesium intake. Race and ethnicity were combined into a single dichotomous variable of non-Hispanic white versus all others. Education was dichotomized as having a high school diploma (or equivalent) versus not. Marital status was characterized as married or living as married versus single, divorced, widowed, or separated. Household income was dichotomized as low if it was reported to be ≤$35,000 per year. Food insecurity was present if the subject endorsed any of the following 3 statements: "(I/we) worried whether (my/our) food would run out before (I/we) got money to buy more" or "The food that (I/we) bought just did not last, and (I/we) did not have money to get more" or "(I/we) could not afford to eat balanced meals." Tobacco use was considered present if the patient endorsed current smoking versus absent for former smokers and those who never smoked. Alcohol use was coded as the average number of units consumed per day over the past year. A unit of alcohol is 1 can of beer, 1 glass of wine, or 1 ounce of liquor. Nondrinkers were coded as zero. Diabetes and kidney disease were considered present if the patient endorsed that a doctor or other health professional had told them they had such a diagnosis. Folate intake (micrograms/day) included dietary folate equivalents of food plus supplements and was dichotomized at <230 μg/day (the lowest quintile of daily folate intake).

Statistical Analysis

The primary hypothesis was that depression is associated with magnesium intake while adjusting for possible confounders. We used unadjusted nonparametric Wilcoxon tests of trend to assess the relationships between quintiles of magnesium intake and other subject characteristics.[25] We used logistic regression to model the relationship between the presence of depression (PHQ-9 score ≥5) and low magnesium intake (<184 mg/day, the lowest quintile) and tested the hypothesis by examining the odds ratio (OR) and relative risk[26] (RR) of magnesium intake and their 95% confidence intervals (CIs). Each potential confounder was tested in a separate univariate logistic regression for association with the outcome (depression) and the main predictor (low magnesium intake). If the variable was associated with both the outcome and predictor (each P < .1), it was considered a potential confounder and included in the multivariate model. We also explored the use of magnesium as a function of energy intake (milligrams of magnesium/1000 calories) as the predictor by following the same procedure. Because both magnesium intake[27] and depression[1] vary with age and sex, we constructed additional models including interaction terms to explore the possibility of interactions of magnesium with sex and magnesium with age. All analyses used the stratification and weighting scheme recommended for NHANES by the National Center for Health Statistics[22] using Stata software version13.1 (StataCorp, College Station, TX). P values <.05 were considered statistically significant.