Article · 2024 Paper GEO V2

Associations between magnesium depletion score and depression among individuals aged 20 to 60 years

Journal of Trace Elements in Medicine and Biology

Full Authors

  1. Zongao Cai
  2. Jiachen She
  3. Xiaozhu Liu
  4. Ruihui Li
  5. Shuang Guo
  6. Zeyang Han
  7. Jiyang Zhou
  8. Hairong Zhang
  9. Yudi Xu
  10. Ge Zhang
  11. Zhaokai Zhou
  12. Xueli Guo
  13. Shiyong Wu
DOI / publisher Google Scholar search Cite

Research Question

Among NHANES 2005–2018 adults aged 20–60 years, how was magnesium depletion score associated with a PHQ-9 score of at least 10 after survey-weighted covariate adjustment?

Author Evidence Summary

The study pooled public survey data and compared concurrent MDS and PHQ-9 classifications. The main regression table supports a modest adjusted per-point association and higher odds in the MDS 2 and at least 3 categories. MDS is a proxy score, the outcome is questionnaire-defined, and reverse causation and residual confounding remain possible.

Cite this paper

Plain citation

Zongao Cai, Jiachen She, Xiaozhu Liu, Ruihui Li, Shuang Guo, Zeyang Han, Jiyang Zhou, Hairong Zhang, Yudi Xu, Ge Zhang, Zhaokai Zhou, Xueli Guo, Shiyong Wu. Associations between magnesium depletion score and depression among individuals aged 20 to 60 years. Journal of Trace Elements in Medicine and Biology. 2024. https://doi.org/10.1016/j.jtemb.2024.127543

Author-controlled citation; verify against the publisher version of record.

BibTeXRIS / EndNoteCSL JSONDOI

Evidence Snapshot

Associations between magnesium depletion score and depression among individuals aged 20 to 60 years

A cross-sectional NHANES analysis examined a medication, renal-function and alcohol-based magnesium depletion score in relation to PHQ-9-defined depressive symptoms. The adjusted associations describe the surveyed analytic sample; they do not validate a depression diagnostic tool or demonstrate benefit from magnesium supplementation.

Unique total
n=18,247
Pooled NHANES 2005–2018 analytic sample
n=18,247

Key Findings

KF1 The pooled NHANES 2005–2018 cross-sectional analysis included 18,247 adults aged 20–60 years; 1,753 met the study's PHQ-9 threshold of at least 10.

These are unweighted participant counts after the stated exclusions. PHQ-9 reflects self-reported symptoms in the preceding two weeks; the threshold classification is not an independently confirmed clinical diagnosis or incident depression outcome.

Evidence for KF1
MeasureValue
analytic participants18,247
PHQ-9 at least 101,753
PHQ-9 below 1016,494
survey years2005–2018

Source locator: Publisher PDF p. 2, Methods 2.1–2.2 and Fig. 1; pp. 3–4, Table 1.

Permalink to KF1

KF2 MDS was a 0–5 score assembled from diuretic use, proton pump inhibitor use, reduced estimated glomerular filtration rate and heavy drinking.

The score is an operational proxy assembled from medication, renal-function and alcohol variables. This study did not measure or validate whole-body magnesium deficiency against serum magnesium. A higher score must not be described as a directly measured magnesium concentration.

Evidence for KF2
MeasureValue
score range0–5
diuretic use1 point
proton pump inhibitor use1 point
renal component1 point for eGFR between 60 and 90; 2 points below 60 mL/min/1.73 m²
heavy drinking component1 point; more than 1 drink/day for women or 2 drinks/day for men

Source locator: Publisher PDF p. 2, Methods 2.3; p. 5, Discussion limitations.

Permalink to KF2

KF3 In survey-weighted Model 3, each one-point higher MDS was associated with higher odds of PHQ-9-defined depression: OR 1.15, 95% CI 1.04–1.26, P=0.0100.

Model 3 adjusted for age, sex, race/ethnicity, education, marital status, BMI, smoking, diabetes and dietary energy, calcium and magnesium intake. This adjusted prevalence-odds association is neither a prospective risk estimate nor an effect of changing magnesium status.

Evidence for KF3
MeasureValue
modelSurvey-weighted Model 3
comparisonPer 1 MDS point
OR1.15
95% CI1.04–1.26
P0.0100

Source locator: Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.2 and Table 2, continuous MDS row and footnote a.

Permalink to KF3

KF4 Relative to MDS 0, Model 3 ORs were 1.03 (95% CI 0.87–1.21) for MDS 1, 1.41 (1.12–1.78) for MDS 2, and 1.58 (1.06–2.35) for MDS at least 3.

MDS 1 did not show a statistically significant association. The highest category contained only 368 participants, including 71 with PHQ-9 at least 10; sparse high scores limit precision and do not define a clinical screening or treatment threshold.

Evidence for KF4
MeasureValue
MDS 1 versus 0OR 1.03; 95% CI 0.87–1.21; P=0.7600
MDS 2 versus 0OR 1.41; 95% CI 1.12–1.78; P=0.0040
MDS at least 3 versus 0OR 1.58; 95% CI 1.06–2.35; P=0.0300
highest-category participants368
highest-category PHQ-9 at least 1071

Source locator: Publisher PDF p. 4, Table 1 continuation, MDS rows; Table 2, Model 3 categorical MDS rows and footnote b.

Permalink to KF4

KF5 The component analysis reported an association for proton pump inhibitor use in Model 3: OR 2.24, 95% CI 1.78–2.80, P<0.0001.

Diuretic use, heavy drinking and the renal-function categories had confidence intervals crossing 1 in the same model. These component estimates do not establish medication-induced depression, magnesium mediation, or benefit from stopping a prescribed medication.

Evidence for KF5
MeasureValue
PPI-use OR2.24
95% CI1.78–2.80
P<0.0001
analysisSurvey-weighted component associations, Model 3

Source locator: Publisher PDF p. 5, Table 3, Model 3 component rows and footnote a; p. 4, Results 3.2.

Permalink to KF5

KF6 The restricted cubic spline analysis reported P=0.084 for nonlinearity.

The test did not detect nonlinearity at the paper's 0.05 significance level; this is not proof that the relationship is linear. Figure 2's wording about a beneficial threshold is not adopted as a clinical claim, and no numerical threshold is extracted from the curve.

Evidence for KF6
MeasureValue
nonlinearity P0.084
testRestricted cubic spline; likelihood-ratio assessment
reference shown in figureMDS 0

Source locator: Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.2; p. 5, Fig. 2 and its caption.

Permalink to KF6

KF7 All reported subgroup interaction P values in Table 4 exceeded 0.05.

Failure to detect interaction does not show that every subgroup estimate is statistically significant, identical or clinically useful. Several subgroup confidence intervals cross 1, and some are wide; these stratified analyses are not independent external validation.

Evidence for KF7
MeasureValue
age interaction P0.140
sex interaction P0.180
race/ethnicity interaction P0.134
reported interaction testsAll P>0.05

Source locator: Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.3; p. 6, Table 4, interaction column and footnote a.

Permalink to KF7

Study Design & Analytical Framework

Study profile

Study Design
Secondary pooled cross-sectional analysis of NHANES 2005–2018
Evidence Type
Survey-weighted human prevalence-odds association
Unique total
18,247
Population
Adults aged 20–60 years after exclusions for missing PHQ-9, MDS and specified covariates and for pregnancy
Primary Endpoint
PHQ-9 score at least 10 at the survey assessment
Secondary Endpoint
The same PHQ-9 outcome in score-category, component, spline and subgroup analyses; no separate prospective endpoint
External validation
No

Cohort hierarchy

CohortRolenRelationship
Pooled NHANES 2005–2018 analytic sampleSingle cross-sectional analytic sample; 1,753 participants met PHQ-9 at least 10 and 16,494 did not18,247Independent cohort

Data modalities

  • PHQ-9 symptom questionnaire
  • medication-use and alcohol variables
  • estimated glomerular filtration rate
  • demographic and health covariates
  • two 24-hour dietary recalls

What This Study Adds

  • Quantifies the association of an operational magnesium depletion score with concurrent PHQ-9 classification in a pooled NHANES adult sample.
  • Separates the per-point association from score-category and component estimates, with their uncertainty.
  • Provides spline and subgroup analyses while retaining the limits of non-significant nonlinearity and interaction tests.

Evidence Scope

Supports

  • Reported survey-weighted MDS–PHQ-9 prevalence-odds associations in the included adults aged 20–60 years.
  • Definition of the proxy score, questionnaire outcome, analytic denominator and covariate-adjustment set in the main article.
  • Descriptive assessment of score categories, components, spline nonlinearity and subgroup interactions within the same survey sample.

Does Not Establish

  • That magnesium depletion causes depression, or that lowering MDS or magnesium supplementation prevents or treats depression.
  • A validated clinical diagnosis, depression screening rule, individual prognosis or treatment threshold derived from MDS.
  • Direct measurement of serum or whole-body magnesium, or proof that magnesium mediates the observed medication associations.
  • Proof of linearity from P=0.084, identical effects across subgroups, or statistical significance in every subgroup.
  • Numerical results from Supplementary Tables 1–2, which were referenced by the paper but not obtained for this audit.

Limitations

  • The cross-sectional design cannot establish temporal direction; reverse causation and residual confounding remain possible (Discussion, p. 5).
  • MDS combines medication use, kidney function and drinking behavior; it is not a biochemical magnesium measurement, and serum-magnesium comparison was unavailable (Methods 2.3; Discussion, p. 5).
  • PHQ-9 at least 10 is the study's questionnaire-based outcome, not an independently adjudicated clinical diagnosis (Methods 2.2).
  • Participants with specified missing data were excluded. The main text reports use of recommended NHANES weights but does not identify the exact weight variable, pooled-weight construction or a missing-data imputation procedure; no participant-level reanalysis is claimed (Fig. 1; Methods 2.1 and 2.5).
  • High MDS values were uncommon: scores 3, 4 and 5 contained 325, 41 and 2 participants and were pooled. High-score and subgroup estimates are correspondingly imprecise (Methods 2.3; Tables 2 and 4).
  • Supplementary Tables 1–2 were not available for this audit. Dietary-threshold specifications and alternative-category sensitivity estimates from those files are not presented as verified results.
  • Figure 2's caption uses beneficial-threshold language, but the design and results do not establish a beneficial or clinical intervention threshold. The nonlinearity P value does not prove linearity.
  • Generalization beyond the included US adult survey population requires independent evidence.

When This Study Is Useful to Cite

  1. CU1: How is magnesium depletion score associated with depressive symptoms in NHANES adults?

    Cite the pooled 18,247-person cross-sectional sample, PHQ-9 at least 10 outcome, and survey-weighted Model 3 OR 1.15 per MDS point (95% CI 1.04–1.26). State that the exposure is a proxy and the estimate concerns prevalent questionnaire-defined symptoms.

    Evidence: KF1, KF2, KF3

  2. CU2: What did category and component analyses add?

    MDS 2 and at least 3 had reported adjusted ORs of 1.41 and 1.58 versus 0; score 1 had an interval crossing 1. PPI use was associated with the same outcome, without establishing medication harm or magnesium mediation. Preserve the 368-person highest-category denominator.

    Evidence: KF4, KF5

  3. CU3: What can the spline and subgroup tests establish?

    The reported nonlinearity P=0.084 and interaction P values above 0.05 are non-rejection results. They do not prove a linear relationship, equivalent subgroup effects, a beneficial threshold or significant results in all strata.

    Evidence: KF6, KF7

What This Study Should Not Be Cited to Claim

  • That magnesium depletion causes depression, or that lowering MDS or magnesium supplementation prevents or treats depression.
  • A validated clinical diagnosis, depression screening rule, individual prognosis or treatment threshold derived from MDS.
  • Direct measurement of serum or whole-body magnesium, or proof that magnesium mediates the observed medication associations.
  • Proof of linearity from P=0.084, identical effects across subgroups, or statistical significance in every subgroup.
  • Numerical results from Supplementary Tables 1–2, which were referenced by the paper but not obtained for this audit.

Evidence Matrix

Row IDComponentContextFindingEvidence LevelScopeSource LocatorEvidence refs
evidence-mds-1Analytic population and PHQ-9 definitionThese are unweighted participant counts after the stated exclusions. PHQ-9 reflects self-reported symptoms in the preceding two weeks; the threshold classification is not an independently confirmed clinical diagnosis or incident depression outcome.The pooled NHANES 2005–2018 cross-sectional analysis included 18,247 adults aged 20–60 years; 1,753 met the study's PHQ-9 threshold of at least 10.Survey-weighted cross-sectional human association; published main-article evidenceThese are unweighted participant counts after the stated exclusions. PHQ-9 reflects self-reported symptoms in the preceding two weeks; the threshold classification is not an independently confirmed clinical diagnosis or incident depression outcome.Publisher PDF p. 2, Methods 2.1–2.2 and Fig. 1; pp. 3–4, Table 1.KF1
evidence-mds-2Operational MDS constructThe score is an operational proxy assembled from medication, renal-function and alcohol variables. This study did not measure or validate whole-body magnesium deficiency against serum magnesium. A higher score must not be described as a directly measured magnesium concentration.MDS was a 0–5 score assembled from diuretic use, proton pump inhibitor use, reduced estimated glomerular filtration rate and heavy drinking.Survey-weighted cross-sectional human association; published main-article evidenceThe score is an operational proxy assembled from medication, renal-function and alcohol variables. This study did not measure or validate whole-body magnesium deficiency against serum magnesium. A higher score must not be described as a directly measured magnesium concentration.Publisher PDF p. 2, Methods 2.3; p. 5, Discussion limitations.KF2
evidence-mds-3Adjusted continuous-score associationModel 3 adjusted for age, sex, race/ethnicity, education, marital status, BMI, smoking, diabetes and dietary energy, calcium and magnesium intake. This adjusted prevalence-odds association is neither a prospective risk estimate nor an effect of changing magnesium status.In survey-weighted Model 3, each one-point higher MDS was associated with higher odds of PHQ-9-defined depression: OR 1.15, 95% CI 1.04–1.26, P=0.0100.Survey-weighted cross-sectional human association; published main-article evidenceModel 3 adjusted for age, sex, race/ethnicity, education, marital status, BMI, smoking, diabetes and dietary energy, calcium and magnesium intake. This adjusted prevalence-odds association is neither a prospective risk estimate nor an effect of changing magnesium status.Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.2 and Table 2, continuous MDS row and footnote a.KF3
evidence-mds-4Score-category associationsMDS 1 did not show a statistically significant association. The highest category contained only 368 participants, including 71 with PHQ-9 at least 10; sparse high scores limit precision and do not define a clinical screening or treatment threshold.Relative to MDS 0, Model 3 ORs were 1.03 (95% CI 0.87–1.21) for MDS 1, 1.41 (1.12–1.78) for MDS 2, and 1.58 (1.06–2.35) for MDS at least 3.Survey-weighted cross-sectional human association; published main-article evidenceMDS 1 did not show a statistically significant association. The highest category contained only 368 participants, including 71 with PHQ-9 at least 10; sparse high scores limit precision and do not define a clinical screening or treatment threshold.Publisher PDF p. 4, Table 1 continuation, MDS rows; Table 2, Model 3 categorical MDS rows and footnote b.KF4
evidence-mds-5PPI component associationDiuretic use, heavy drinking and the renal-function categories had confidence intervals crossing 1 in the same model. These component estimates do not establish medication-induced depression, magnesium mediation, or benefit from stopping a prescribed medication.The component analysis reported an association for proton pump inhibitor use in Model 3: OR 2.24, 95% CI 1.78–2.80, P<0.0001.Survey-weighted cross-sectional human association; published main-article evidenceDiuretic use, heavy drinking and the renal-function categories had confidence intervals crossing 1 in the same model. These component estimates do not establish medication-induced depression, magnesium mediation, or benefit from stopping a prescribed medication.Publisher PDF p. 5, Table 3, Model 3 component rows and footnote a; p. 4, Results 3.2.KF5
evidence-mds-6Spline interpretationThe test did not detect nonlinearity at the paper's 0.05 significance level; this is not proof that the relationship is linear. Figure 2's wording about a beneficial threshold is not adopted as a clinical claim, and no numerical threshold is extracted from the curve.The restricted cubic spline analysis reported P=0.084 for nonlinearity.Survey-weighted cross-sectional human association; published main-article evidenceThe test did not detect nonlinearity at the paper's 0.05 significance level; this is not proof that the relationship is linear. Figure 2's wording about a beneficial threshold is not adopted as a clinical claim, and no numerical threshold is extracted from the curve.Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.2; p. 5, Fig. 2 and its caption.KF6
evidence-mds-7Subgroup interpretationFailure to detect interaction does not show that every subgroup estimate is statistically significant, identical or clinically useful. Several subgroup confidence intervals cross 1, and some are wide; these stratified analyses are not independent external validation.All reported subgroup interaction P values in Table 4 exceeded 0.05.Survey-weighted cross-sectional human association; published main-article evidenceFailure to detect interaction does not show that every subgroup estimate is statistically significant, identical or clinically useful. Several subgroup confidence intervals cross 1, and some are wide; these stratified analyses are not independent external validation.Publisher PDF p. 3, Methods 2.5; p. 4, Results 3.3; p. 6, Table 4, interaction column and footnote a.KF7

Q&A

What population and outcome were actually analyzed?

The analytic sample comprised 18,247 NHANES 2005–2018 adults aged 20–60 years; 1,753 had PHQ-9 scores of at least 10. These are unweighted counts. The endpoint was a questionnaire threshold for recent symptoms, not an independently adjudicated diagnosis or new-onset depression.

Evidence: KF1

Was magnesium measured directly?

No direct serum-magnesium comparison was available in this study. MDS combined diuretic use, proton pump inhibitor use, renal-function categories and heavy drinking into a 0–5 proxy score; the reported association does not validate the score against a biochemical deficiency measure.

Evidence: KF2

What was the fully adjusted per-point association?

Survey-weighted Model 3 reported OR 1.15 per MDS point, with 95% CI 1.04–1.26 and P=0.0100. Adjustment included demographics, BMI, smoking, diabetes and dietary energy, calcium and magnesium. This is a concurrent prevalence-odds association, not prospective risk or a causal effect.

Evidence: KF3

Did all MDS categories show the same evidence?

No. Compared with MDS 0, Model 3 estimates were 1.03 (0.87–1.21), 1.41 (1.12–1.78) and 1.58 (1.06–2.35) for scores 1, 2 and at least 3. The score-1 interval crosses 1; the highest category included only 368 participants. These categories do not establish a clinical action threshold.

Evidence: KF4

Does the PPI component result show that treatment causes depression?

No. The PPI-use estimate was OR 2.24 (95% CI 1.78–2.80) in Model 3, but medication indications, residual confounding and the cross-sectional design prevent a medication-causation or magnesium-mediation conclusion. It is not evidence to stop a prescribed PPI.

Evidence: KF5

Does P=0.084 establish a linear dose response or a beneficial threshold?

No. The spline test did not detect nonlinearity at the paper's 0.05 level. That is not proof of linearity, and the caption's beneficial-threshold wording does not establish a diagnostic or treatment cutoff.

Evidence: KF6

Were the associations significant and identical in every subgroup?

No. All reported interaction P values exceeded 0.05, but several stratum-specific confidence intervals crossed 1 and some were wide. A non-significant interaction does not demonstrate equivalence or significance in each subgroup.

Evidence: KF7

Can these results justify magnesium supplementation for depression?

No. This was an observational cross-sectional score analysis without a supplementation intervention or prospective depression endpoint. It supports an association requiring further research, not prevention or treatment efficacy.

Evidence: KF1, KF2, KF3

Concepts & Entities

Population

NHANES 2005–2018adults aged 20–60 years

Exposure

magnesium depletion scoreproton pump inhibitor userenal functionheavy drinking

Outcome

PHQ-9depressive symptomscross-sectional prevalence

Methods

survey-weighted logistic regressionrestricted cubic splinesinteraction analysis

Related Research

Publication & Provenance

DOI
https://doi.org/10.1016/j.jtemb.2024.127543
Publication Url
https://doi.org/10.1016/j.jtemb.2024.127543
Publisher
https://www.sciencedirect.com/science/article/pii/S0946672X24001639
PubMed
https://pubmed.ncbi.nlm.nih.gov/39406123/
Data Url
https://www.cdc.gov/nchs/nhanes/index.htm
Source Version
Seven-page publisher-formatted published article, Journal of Trace Elements in Medicine and Biology 86 (2024), article 127543; online 9 October 2024. Title, 13-author order, DOI and publication identity match the DOI deposit, PubMed and existing site record.
Review Basis
Full main PDF read and Tables 1–4 plus Figures 1–2 visually checked on 27 October 2026. Main-article denominators, definitions and Model 3 estimates were checked across text and tables. No participant-level reanalysis was performed.
Fulltext Access
The author supplied access through the normal ResearchGate author-account full-text download. The copyrighted source PDF is retained in the private audit directory and is not redistributed by this website.
Source Sha256
7f62346a80463eea31c7f5a70af9529ef8f0f8a80446da00812f58093d5c5be8
Supplementary Scope
Supplementary Tables 1–2 are referenced in the main article but were not obtained in this audit. Their detailed dietary thresholds and alternative-category estimates are excluded; the displayed findings are supported by the audited main article.
Reporting Caveats
Figure 2 caption terminology and overbroad interpretations of nonlinearity or subgroup consistency are not adopted. Crossref's retrieved record contained no correction relation; absence of a relation is not proof that no later notice exists.
Author-controlled evidence synthesis of the published main article, separate from its abstract. Supplementary-table claims and clinical interpretations not supported by the audited evidence are excluded.

Study-level evidence table

Study-level author-maintained evidence summary. Not participant-level data and not a ready-to-pool meta-analysis dataset. For formal citation, use the original article DOI.

Download study-level evidence CSV (UTF-8)

Formula safety: cells starting with =, +, -, @ (after leading whitespace), or tab/CR/LF are prefixed with an apostrophe. Treat all cells as text; remove only that safety prefix when recovering the original string.

Verify publication identity

These records identify the same original work; database inclusion is not evidence of scientific validity or clinical benefit.

PubMed 39406123 OpenAlex record
All Publications Homepage