Vitamin D3 for Life

Live Enlightened

Research libraryVitamin D3 For Life
Study design and methods

Secondary relevance to vitamin D This paper mainly addresses another topic and also reports a vitamin D analysis, result or scientific synthesis.

This paper mainly addresses another topic and also reports a vitamin D analysis, result or scientific synthesis.

Why this paper is included

Supporting source evidence:

The association between magnesium supplement use and a reduced risk of all-cause death or hospitalization was homogeneous, with no significant interactions, across subgroups of the IPTW-weighted cohort stratified by gender (p = 0.8413), BMI (p = 0.1875), serum vitamin D levels (p = 0.1469), prescribed multivitamin (p = 0.9108), diuretics (p = 0.1114), serum magnesium level (p = 0.7719), and left-ventricular ejection fraction (p = 0.1648).

Source: 3.2.3. Interactions and Stratified Subgroup Analyses. View original paper

Relevance classification does not establish study quality, causation or editorial approval.

Study design Cohort study.

Cohort study

Examines exposures and outcomes over time within a defined group.

Meta-analysis unclear · Blinding unclear / not reported · Human research reported

How this paper was classified

Automatic source-based classification; awaiting editorial verification. This is not a quality score.

This was a retrospective cohort study.

Source: Methods.

Regardless of whether participants were users or non-users, the follow-up began on the eligibility date.

Source: Methods.

View supporting evidence in the original paper

What the study types mean
Systematic review
Uses explicit methods to identify, assess and summarise research on a defined question.
Scoping review
Maps the extent and characteristics of research on a topic.
Narrative review
Discusses existing research; systematic-review methods have not been established by this classification.
Review / evidence synthesis
Reviews existing research; the exact review methods need further assessment.
Randomised controlled trial
Assigns participants or groups to comparison interventions by chance.
Non-randomised intervention study
Examines an intervention without random allocation.
Cohort study
Examines exposures and outcomes over time within a defined group.
Case-control study
Compares people with an outcome against people without it, examining exposures.
Cross-sectional study
Examines a population at a particular point in time.
Observational study
Observes exposures and outcomes without assigning the exposure; the subtype is not established here.
Case report
Describes an individual case.
Case series
Describes a group of individual cases.
Secondary analysis
Analyses existing study data; this paper is not automatically another independent trial.
Laboratory study
Investigates biological processes in a laboratory setting.
Qualitative study
Examines experiences or meanings using qualitative methods.
Genetic instrumental-variable study
Uses genetic variation as an instrument to investigate relationships; this is not a randomised treatment trial.
Study protocol
Describes planned methods and is not evidence that the planned study has reported results.
Multiple reported designs
More than one design is explicitly reported; examine the supporting methods.
Study design awaiting verification
The available source does not establish a sufficiently clear study design.

Meta-analysis combines numerical results from studies. Blinding describes who was unaware of treatment allocation. Neither is a separate quality grade. Study design, vitamin D relevance and document format are independent filters.

Full studies

Association Between Over-the-Counter Magnesium Supplement Use and Health Outcomes in Veterans with Newly Diagnosed Heart Failure

Cohort study · Nutrients · 2025

Original publication title

Association Between Over-the-Counter Magnesium Supplement Use and Health Outcomes in Veterans with Newly Diagnosed Heart Failure

Yan Cheng; Andrew R. Zullo; Ying Yin; Yijun Shao; Senait Tekle; Simin Liu; Qing Zeng-Treitler; Wen-Chih Wu

© 2025 by the authors.

Original publication · Publisher record · CC BY 4.0

Source wording retained. Formatting, section navigation and display-title punctuation adapted for this research record.

Authors’ conclusion

In conclusion, our findings indicate that magnesium supplementation is linked to a lower risk of all-cause mortality or hospitalization in patients with HF. These promising outcomes merit further exploration and validation through RCTs.

An extract from the source publication. Read it in context.

At a glance

Study background

What background does the paper provide?

Magnesium is an essential intracellular mineral involved in numerous key enzymatic processes in the human body [1] and plays a critical role in preventing arrhythmia, improving hemodynamics, and supporting cardiovascular function [2,3]. However, individuals with heart failure (HF) are at increased risk of magnesium deficiency [4]. Findings from two studies suggest that low magnesium dietary intake and low plasma magnesium levels are linked to increased risk of HF-related hospitalizations [5,6].

Additionally, a low plasma magnesium level was also associated with a higher risk of cardiovascular death (hazard ratio [HR]: 1.38; 95% confidence interval [CI]: 1.04 to 1.83; p = 0.024) in 1569 patients with chronic HF [7]. However, the literature does not suggest that dietary intake alone relates to changes in cardiovascular outcomes. Our previous study showed that dietary magnesium intake was not related to mortality in women with HF [8]. Another study showed that although low serum magnesium was associated with a higher risk of arrhythmias, there was no observed association between dietary magnesium intake and arrhythmia risk [9].

Interestingly, magnesium supplementation has been shown to reduce the risk of cardiac arrhythmias in patients with HF in two randomized controlled trials (RCTs) [10]. But it remains unknown if magnesium supplementation has an effect on all-cause mortality or hospitalization in patients with HF.

To assess the effects of magnesium supplementation, rigorously designed RCTs are ideal. However, RCTs are often expensive for a nutritional supplement, as they typically require large participant cohorts and extended follow-up periods to detect relatively small effect sizes [11,12,13]. Target trial emulation is designed to replicate the structure of an RCT on the observational data to evaluate clinical effectiveness while controlling the biases inherent to observational research [14,15,16,17].

This study aimed to assess the association between magnesium supplementation and all-cause hospitalization or death using a target trial emulation.

Source: 1. Introduction Source: 1. Introduction

Material & Methods

2.1. Study Population

Study subjects were patients newly diagnosed with HF between 1 January 2000 and 31 December 2020, identified from the US Veterans Health Administration (VHA) national electronic health records (EHR). HF was defined based on the International Classification of Diseases (ICD) codes (ICD9: 398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.13, 404.91, 428; ICD10: I11.0, I13.0, I13.2, I50). The diagnoses were identified from both inpatient and outpatient encounter data in the VHA system.

2.2. Study Design and Ethics

This was a retrospective cohort study. The study was approved by the VHA Central Institutional Review Board (IRBNET#1710776-1, 23 April 2021), which granted a waiver of informed consent. Due to the sensitive nature of the data involved, access to the dataset is restricted; qualified researchers trained in human subject confidentiality may request access by contacting the Washington DC VA Medical Center via email at helen.sheriff@va.gov.

2.3. Target Trial Emulation

We emulated a target trial by “enrolling” eligible patients into the study cohort. The elements of the target trial and the emulation are summarized in Table 1. A patient would be qualified for inclusion in a trial if, on an outpatient clinical note date, they met the following criteria: (1) had the initial HF diagnosis within 1 year before, (2) had at least two outpatient clinical encounter notes within the previous 12 months (to confirm magnesium supplement use status), (3) had no records of prescribed or supplemental magnesium use prior to this date documented in the electronic medical records (EMR), (4) had no hospitalization within 30 days before or at the eligible date, and (5) had no end-stage-renal-disease (ESRD) any time before or at the eligible date.

The purpose of requiring “no hospitalization within 30 days before or on the eligible date” was to exclude patients who were acutely ill, since these patients would be more likely to die earlier and have less opportunity to initiate magnesium supplements, thereby further reducing selection bias.

2.4. Exposure

The primary exposure of interest was EMR documentation of the patients’ current use of over-the-counter magnesium supplements, as documented in outpatient clinic visit notes. We identified this information using natural language processing (NLP) techniques [18] developed by our team to extract mentions of magnesium supplement use from the clinical notes. The NLP process and performance in identifying magnesium supplement use is described in detail elsewhere [18]. Validation of the NLP method used for dietary supplement and vitamin use showed an AUC of 98.9%, a precision of 96.2%, and a recall of 87.7% [18].

Eligible date: The status of magnesium supplement use on the date of the outpatient clinic visit when a patient became eligible for the target trial is the eligible date (index date, time 0). That is also the date the patient was classified as a magnesium supplement user or non-user (Table 1).

Exposure status during follow-up: For both groups, any patient without clinic visit notes for more than one year was considered lost to follow-up and was censored one year after their last documented clinic visit. For magnesium supplement users, if magnesium supplement use was not mentioned in the notes for over one year, the magnesium supplement use was considered discontinued one year after the last recorded use. If a patient was initially classified as a non-user of magnesium supplement but a clinic note during follow-up indicated magnesium supplement use, this was treated as a change in exposure status, and the patient would be censored at that time point.

To maximize the sample size of magnesium supplement users, we included all patients who were qualified to be a user. Then, we matched a randomly selected group of controls who had a clinic visit that was on the same date as a user in a 1:1 ratio.

The process of the target trial emulation is displayed in Figure 1. The graphical depiction of the study design is displayed in Figure 2. As described earlier, both groups were required to meet the same eligibility criteria. We used the same approach to identify exposure status and define the index date. Regardless of whether participants were users or non-users, the follow-up began on the eligibility date. All these measures were implemented to control for immortal time bias.

2.5. Study Outcomes and Follow-Up

The main study outcome was all-cause hospitalization or death, as identified from the VHA Corporate Data Warehouse (CDW). Death dates were recorded in the patient vital status table and hospitalizations were recorded in the inpatient table. Patients were followed from the index date until the outcome of interest occurred or the censoring date due to magnesium supplement use status change (i.e., stop use in users or start use in non-users), prescribed magnesium replacement was started, or the end of 5th year since the index date, whichever occurred first.

To account for potential differences in follow-up time since magnesium supplement non-users are more likely to maintain the same status than users, a magnesium supplement user and their paired non-user with the same index date were censored at the same date if either side were censored.

2.6. Study Covariates

We included an extensive list of baseline variables as covariates aimed at controlling for potential confounding and to ensure comparability between the two treatment strategy groups. Baseline covariates included demographics, socio-environmental and economic status, duration between the initial HF diagnosis and the index date, comorbid conditions, medication use, laboratory data, vital signs, and healthcare utilization.

Median household income and the socio-environmental percentile ranking were estimated at the patients’ residential address zip code. The socio-environmental percentile Environmental Justice Index (EJI) developed by the Center for Disease Control (CDC) was used to measure the cumulative impact of environmental burden on human health and the environment at the patients’ residence. A higher percentile EJI ranking indicates a greater cumulative socio-environmental burden on the community [19]. Comorbid conditions, identified using ICD codes, were captured any time before or at the index date.

Other covariates were captured within one year prior to and up to the index date, with the measures closest to the index date used as baseline data, which included medication use from prescription/fill records, laboratory tests using LOINC codes, vital signs from vital sign records, and healthcare utilization from outpatient and inpatient encounter records.

2.7. Variables with Missing Values

For vital signs and laboratory values, if the missing value rate was less than 30%, we imputed the values using the mean, conditional on age, gender, race, and ethnicity; otherwise, we converted the variables into categorical ones using missing as one of the possible categories to adjust for potential bias from informative missing. The following variables were imputed, with missing rates (magnesium supplement users vs. non-users) as follows: 4.7% (5.4% vs. 4.1%) for body mass index (BMI), 2.5% (2.9% vs. 2.2%) for systolic blood pressure, 2.6% (3.0% vs. 2.2%) for diastolic blood pressure, 23.4% (26.5% vs. 20.3%) for low-density lipoprotein, 23.9% (27.2% vs. 20.6%) for triglycerides, 22.7% (25.9% vs. 19.5%) for high-density lipoprotein, 22.4% (25.6% vs. 19.2%) for total cholesterol, 13.7% (18.0% vs. 9.3%) for serum sodium, 13.6% (17.9% vs. 9.4%) for serum potassium, 25.6% (28.4% vs. 22.8%) for serum creatinine, 18.4% (22.9% vs. 13.9%) for serum calcium, and 29.7% (32.3% vs. 27.0%) for hemoglobin A1C.

This indicates that magnesium supplement users had a slightly higher rate of missing values.

2.8. Statistical Analysis

We used the per-protocol approach for data analysis as described earlier, and patients were censored when they changed the treatment strategy.

We fit a logistic regression model to estimate the propensity score (i.e., the probability of being a magnesium supplement user) using baseline characteristics including demographics, comorbid conditions, concurrent medications, health care utilization in the past year, vital signs and labs. Then, we applied inverse probability treatment weighting (IPTW) to generate a pseudo-population, in which baseline characteristics of two treatment strategy groups were balanced (ASD < 10%). Stabilized weights (SW) were calculated for each patient. To avoid extreme values, when a weight was above 10, it was trimmed down to 10 [20].

The absolute standardized difference (ASD) for each characteristic was calculated before and after weighting, with an ASD > 10% indicating a significant imbalance in characteristics between the two groups [21].

We used the inverse probability of treatment (IPT)-weighted cohort to estimate the outcome incidence rates of the two groups. Kaplan: Meier curves were plotted, and Cox’s regression modeling was applied to the weighted cohort to assess the association between magnesium supplement use and the risk of all-cause hospitalization or death. The proportional hazards assumption was tested by including an interaction term between magnesium supplement use and time in the Cox’s regression model to examine any time-dependent effects on outcomes.

Additionally, we evaluated statistical interactions between magnesium supplement use and potential effect modifiers, including age, race, serum magnesium level, serum vitamin D level, vitamin D prescription, multivitamin use, and diuretic use, by incorporating a product term in the regression models. When a product term was statistically significant, subgroup analyses were conducted. In these subgroup analyses, we repeated the IPTW process to assess the association between magnesium supplement use and adverse outcomes within each subgroup.

2.9. Sensitivity Analyses

We also conducted two sensitivity analyses. In the first sensitivity analysis, we restricted the cohort to the subgroup who did not die within 6 months after the index date to exclude a very ill population unlikely to have benefited from magnesium supplement use. In the second sensitivity analysis, we restricted the cohort to the subgroup who had at least one additional clinical note mentioning magnesium supplement use and their paired controls to minimize misclassification bias. For each sensitivity analysis, we recalculated the IPTW to make magnesium supplement users and non-users comparable.

All analyses were performed using SAS (Version 9.4). A two-sided p-value of <0.05 was considered statistically significant.

2.10. Target Trial Assumption Evaluation

Consistency: We performed sensitivity analyses mentioned above that assessed the robustness of the results to potential violations of consistency, addressing issues such as exposure-related missing data, time-varying confounding, or measurement error.

Positivity: We evaluated the probability of being exposed or unexposed to magnesium supplement use within each subgroup, based on baseline characteristics, to ensure that observations existed for both exposure status in every subgroup. The positivity assumption was satisfied in this study.

Exchangeability: Although there is always a possibility of unknown confounding, we included an expansive number of variables mentioned above to control for potential confounding bias.

Table 1

Elements of the target trial and emulation for magnesium supplement treatment strategies.

Target TrialEmulation in Observational Data
AimEvaluate the association between magnesium supplement use and all-cause hospitalization or death in the U.S. military veterans with HF
Eligibility(1) Newly diagnosed with HF no more than 1 year
(2) Never used magnesium supplement or replacement treatment before
(3) No hospitalization in the past 30 days
(4) No end-stage renal disease (ESRD)
(1) Had the initial HF diagnosis within 1 year before
(2) Had at least two outpatient clinical notes within the previous 12 months (to confirm magnesium supplement use status)
(3) Had no record of prescribed or supplemental magnesium use prior to this date
(4) Had no hospitalization within 30 days before or at the eligible date
(5) Had no ESRD any time before or at the eligible date
Treatment Strategies(1) Magnesium supplement intervention: Initiation of magnesium supplement
(2) Magnesium supplement non-intervention: No initiation of magnesium supplement
(1) Magnesium supplement users: new documentation of the use of magnesium supplements in an outpatient clinic encounter
(2) Magnesium supplement non-users: No documentation of magnesium supplement use in an outpatient clinic encounter
AssignmentRandomly assigned 1:1 to each treatment strategy in a parallel designVeterans were assigned to treatment arms based on EMR documentation. All magnesium supplement users were included. For the non-user group, randomization was emulated by randomly selecting a group of non-user patients with the same eligible date as a user. Inverse probability weighting was used to balance observed confounders between treatment groups.
Follow-upFrom date of randomization up until end of 5 years, outcome, or censoring due to events such as violation of the assigned treatment strategyFollowed from the time when a patient became eligible for the trial (index date) until the outcome of interest occurred, or the censoring date due to magnesium supplement use status change (i.e., stop use in users or start use in non-users), prescribed magnesium replacement was started, or the end of 5th year from the index date, whichever occurred first.
OutcomesAll-cause mortality or all-cause hospitalization (time to events)Same
Causal ContrastThe difference in risk of all-cause mortality or all-cause hospitalization had everyone followed strategy (1) vs. strategy (2)Same
Statistical AnalysisCox’s regressionPer-protocol analysis using Cox’s regression weighted by inverse probability of treatment
Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods Source: 2. Material & Methods

Results: Baseline Characteristics

Between 1 January 2000, and 31 December 2020, a total of 1,194,095 veterans received a first-time diagnosis of HF within the VHA system. For the target trial emulation, we enrolled all 9900 individuals who met eligibility criteria and used magnesium supplements and matched them to 9900 randomly selected non-users who had a clinic visit on the same index date. Patient characteristics before and after weighting are presented in Table 2 and Table 3 and Supplementary Figure S1.

Prior to weighting, many baseline characteristics were unbalanced (ASD > 10%) between treatment groups. Magnesium supplement users were older (73.3 vs. 72 years) and less likely to be African American (8.8% vs. 15.7%) than non-users. They also had a more favorable socio-environmental percentile ranking (46.3 vs. 49.6) and a shorter average interval between HF diagnosis and the index date (105.6 days vs. 153.7 days). Magnesium users were less likely than non-users to have a history of alcohol abuse (12.0% vs. 16.5%) or smoking (25.1% vs. 36.4%).

They had lower rates of hypertension (84.2% vs. 89.7%), hyperlipidemia (79.3% vs. 83.1%), myocardial infarction (15.5% vs. 20.5%), angina (13.6% vs. 20.7%), and other comorbid conditions, including history of chronic obstructive pulmonary disease (COPD), anemia, arthritis, cancer, urinary tract infection (UTI), pneumonia, respiratory failure, sepsis, depression, and fracture. However, they were more likely to have atrial fibrillation (36.8% vs. 30.9%) and low serum magnesium levels (7.6% vs. 2.5%) at baseline. In addition, magnesium supplement users were less likely to have a known left-ventricular ejection fraction (43.6% vs. 32.6% unknown) and were less likely to be treated with medications including insulin, antihypertensives, cholesterol medications, aspirin, glucocorticoids, and platelet inhibitors.

The prevalence of left-ventricular ejection fraction ≤ 40% was similar between the groups (26.9% vs. 25.4%).

The propensity score distributions of magnesium supplement users and non-users are shown in Figure 3, indicating adequate overlap and no violation of the positivity assumption. The mean and standard deviation of IPTWs were 1.01 ± 0.58 overall (0.99 ± 0.52 for non-magnesium users and 1.02 ± 0.62 for magnesium supplement users). The ASD values for all measured baseline characteristics in the weighted cohort were below 10%, indicating balance after weighting (Table 3). The IPT-weighted pseudo-cohort had a mean age of 72.6 ± 11.1 years and comprised 12.6% African Americans, 3.3% Hispanics, and 3.4% women.

Among these patients in the pseudo-cohort, 87.2% had hypertension, 33.9% had atrial fibrillation, 18.4% had myocardial infarction, and 12.7% had ischemic stroke. Furthermore, 14.5% and 31.5% of patients had a history of alcohol abuse and smoking, respectively. Overall, 39.8% of patients were treated with anti-diabetes medications and 77.1% with anti-hypertensive medications.

Table 2

Baseline characteristics of study participants before IPT weighting.

Baseline CharacteristicsOverall
(N = 19,800)
Magnesium
Users
(N = 9900)
Magnesium
Non-Users
(N = 9900)
ASD Before IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.173.310.772.011.412
Male19,13396.6%954496.4%958996.9%3
Race
White15,22076.9%787279.5%734874.2%13
Black242612.3%8698.8%155715.7%21
Others215410.9%115911.7%99510.1%5
Ethnicity
Non-Hispanics17,96590.7%899190.8%897490.6%1
Hispanics6403.2%2442.5%3964.0%8
Unknown11956.0%6656.7%5305.4%5
Married11,58358.5%626263.3%532153.7%20
Median Income (lowest to highest) of Zip Code
1st Quartile412820.8%191419.3%221422.4%8
2nd Quartile481124.3%247825.0%233323.6%3
3rd Quartile487124.6%247225.0%239924.2%2
4th Quartile553027.9%286528.9%266526.9%4
Unknown4602.3%1711.7%2892.9%8
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code48.022.646.322.249.622.915
Duration between HF Diagnosis and Index Date, Days129.6117.7105.6118.1153.7112.242
Comorbid Conditions (31 Variables)
Alcohol Disorder281814.2%118812.0%163016.5%13
Smoking608330.7%248325.1%360036.4%25
Hypertension17,21887.0%834084.2%887889.7%16
Hyperlipidemia16,07981.2%784979.3%823083.1%10
Myocardial Infarction356618.0%153615.5%203020.5%13
Atherosclerosis477024.1%220822.3%256225.9%8
Angina339417.1%134613.6%204820.7%19
Atrial Fibrillation670533.9%364636.8%305930.9%12
Cardiac Valve Disease343017.3%157415.9%185618.7%7
Ischemic Stroke246012.4%108310.9%137713.9%9
Hemorrhage Stroke1720.9%620.6%1101.1%5
Transient Ischemic Attack10615.4%4294.3%6326.4%9
Chronic Obstructive Pulmonary Disease712536.0%332433.6%380138.4%10
Asthma18209.2%8568.6%9649.7%4
Chronic Kidney Disease561228.3%267227.0%294029.7%6
Anemia654133.0%300830.4%353335.7%11
Arthritis934747.2%439844.4%494950.0%11
Cancer675334.1%312931.6%362436.6%11
Neurological Disorders755638.2%357236.1%398440.2%8
Parkinson4172.1%1811.8%2362.4%4
Hypothyroidism288714.6%151115.3%137613.9%4
Osteoporosis8664.4%4204.2%4464.5%1
Osteomyelitis5002.5%1831.8%3173.2%9
Urinary Tract Infection12386.3%4874.9%7517.6%11
Liver Disease18759.5%8288.4%104710.6%8
Pneumonia248112.5%9179.3%156415.8%20
Respiratory Failure10875.5%3733.8%7147.2%15
Sepsis6623.3%2102.1%4524.6%14
Opioid Use Disorder4202.1%1501.5%2702.7%8
Depression688334.8%299830.3%388539.2%19
Fracture222311.2%9159.2%130813.2%13
Concurrent Medications (22 Variables)
Insulin504125.5%221422.4%282728.6%14
Metformin394919.9%185518.7%209421.2%6
Glucagon-Like Peptide-1780.4%340.3%440.4%2
Sodium: Glucose Cotrasporter-2 Inhibitors1951.0%1001.0%951.0%0
Other Diabetes Medication322116.3%147014.8%175117.7%8
Vitamin D Prescription361818.3%143314.5%218522.1%20
Multivitamin (Rx)9724.9%3903.9%5825.9%9
Angiotensin-Converting Enzyme Inhibitor802140.5%322932.6%479248.4%33
Angiotensin Receptor Blocker313715.8%144614.6%169117.1%7
Calcium Channel Blocker544227.5%221122.3%323132.6%23
Loop Diuretics876544.3%375037.9%501550.7%26
Thiazide Diuretics326216.5%137213.9%189019.1%14
Selective Beta Blocker799640.4%328733.2%470947.6%30
Non-Selective Beta Blocker442522.3%191319.3%251225.4%15
Other Antihypertensive593130.0%259826.2%333333.7%16
Statins11,12556.2%467947.3%644665.1%36
Other Lipid-Lowering Medication12586.4%5735.8%6856.9%5
Proton-Pump Inhibitors666233.6%294029.7%372237.6%17
Aspirin589429.8%206720.9%382738.7%40
Digoxin3341.7%1541.6%1801.8%2
Glucocorticoids244312.3%9729.8%147114.9%16
Platelet Inhibitor305815.4%128012.9%177818.0%14
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,86075.1%836484.5%649665.6%45
1303515.3%9769.9%205920.8%31
2+19059.6%5605.7%134513.6%27
Number of visits in past year31.529.526.828.036.230.232
Addiction Medicine4832.4%1861.9%2973.0%7
Cardiology855143.2%317032.0%538154.4%46
Emergency Care580029.3%202520.5%377538.1%39
Endocrinology9905.0%4254.3%5655.7%6
Hospice Medicine1380.7%540.5%840.8%4
Intensive Care Unit12556.3%3793.8%8768.8%21
Internal Medicine618431.2%262826.5%355635.9%20
Oncology9804.9%4044.1%5765.8%8
Palliative Care3381.7%1151.2%2232.3%8
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.930.26.929.86.96
Systolic Blood Pressure, mmHg129.519.9127.919.8131.019.916
Diastolic Blood Pressure, mmHg72.111.771.311.572.911.914
Serum Magnesium, mg/dL
<1.7 mg/dL10035.1%7577.6%2462.5%23
1.7 to 2.5 mg/dL645432.6%260826.3%384638.8%27
>2.5 mg/dL1891.0%900.9%991.0%1
Unknown12,15461.4%644565.1%570957.7%15
Serum 25-hydroxy vitamin D < 20 ng/dL9274.7%3823.9%5455.5%8
Low-Density Lipoprotein, mg/dL85.934.285.134.586.533.94
Triglycerides, mg/dL146.3111.5147.1110.3145.6112.61
High-Density Lipoprotein, mg/dL43.214.243.314.443.014.02
Cholesterol, mg/dL155.342.6154.542.9156.042.34
Sodium, mmol/L138.93.2138.73.3139.03.19
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.59.20.50
Hemoglobin A1C, %6.81.56.91.56.81.57
Ejection Fraction
≤40%517126.1%260026.9%251125.4%3
>40%709135.8%292829.6%416342.1%26
Unknown753838.1%431243.6%322632.6%23

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Table 3

Baseline characteristics of study participants after IPT weighting.

Baseline CharacteristicsIPT-Weighted
Overall
(N = 19,925)
IPT-Weighted Magnesium
Users
(N = 10,096)
IPT-Weighted Magnesium
Non-Users
(N = 9829)
ASD After IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.172.510.872.611.41
Male19,24096.6%974896.6%949296.6%0
Race
White15,26076.6%771176.4%755076.8%1
Black251812.6%130012.9%121912.4%2
Others214710.8%108610.8%106110.8%0
Ethnicity
Non-Hispanics18,10590.9%917190.8%893590.9%0
Hispanics6543.3%3363.3%3183.2%1
Unknown11665.9%5895.8%5775.9%0
Married11,56558.0%582857.7%573758.4%1
Median Income (lowest to highest) of Zip Code
1st Quartile419721.1%213921.2%205820.9%1
2nd Quartile479724.1%242424.0%237324.1%0
3rd Quartile482524.2%243824.1%238724.3%0
4th Quartile562428.2%284228.1%278228.3%0
Unknown4822.4%2532.5%2292.3%1
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code47.922.748.022.947.822.61
Duration between Diabetes Diagnosis and Index Date, Days135.0118.4136.0125.8134.0110.52
Comorbid Conditions (31 Variables)
Alcohol Disorder288814.5%147114.6%141814.4%1
Smoking626731.5%321031.8%305831.1%2
Hypertension17,38287.2%880487.2%857887.3%0
Myocardial Infarction367018.4%186618.5%180418.4%0
Atherosclerosis488424.5%247824.5%240524.5%0
Angina354817.8%181718.0%173117.6%1
Atrial Fibrillation675633.9%342433.9%333133.9%0
Cardiac Valve Disease356517.9%182418.1%174117.7%1
Ischemic Stroke253012.7%128412.7%124612.7%0
Hemorrhage Stroke1850.9%971.0%880.9%1
Transient Ischemic Attack10925.5%5525.5%5395.5%0
Chronic Obstructive Pulmonary Disease726336.5%370336.7%356036.2%1
Asthma19229.6%9899.8%9349.5%1
Chronic Kidney Disease576929.0%291928.9%285029.0%0
Anemia676734.0%345234.2%331433.7%1
Arthritis953247.8%482047.7%471247.9%0
Cancer690534.7%348834.6%341734.8%0
Neurological Disorders778539.1%395339.1%383239.0%0
Parkinson4212.1%2182.2%2042.1%1
Hypothyroidism294414.8%149814.8%144614.7%0
Osteoporosis8884.5%4594.6%4284.4%1
Osteomyelitis5212.6%2682.7%2542.6%1
Hyperlipidemia16,30681.8%824481.7%806282.0%1
Urinary Tract Infection13446.7%7117.0%6336.4%2
Liver Disease19719.9%100610.0%9659.8%1
Pneumonia259213.0%134113.3%125112.7%2
Respiratory Failure11755.9%6226.2%5535.6%3
Sepsis6993.5%3623.6%3383.4%1
Opioid Use Disorder4372.2%2292.3%2082.1%1
Depression705035.4%359535.6%345435.1%1
Fracture237811.9%123512.2%114411.6%2
Concurrent Medications (22 Variables)
Insulin521926.2%265626.3%256426.1%0
Metformin409920.6%204920.3%205120.9%1
Glucagon-Like Peptide-1840.4%420.4%420.4%0
Sodium: Glucose Cotrasporter-2 Inhibitors2081.0%1061.0%1031.0%0
Other Diabetes Medication331516.6%166316.5%165216.8%1
Vitamin D Prescription378919.0%194119.2%184818.8%1
Multivitamin (Rx)10445.2%5365.3%5075.2%0
Angiotensin-Converting Enzyme Inhibitor828041.6%422741.9%405341.2%1
Angiotensin Receptor Blocker325316.3%163316.2%162016.5%1
Calcium Channel Blocker566328.4%289228.6%277128.2%1
Loop Diuretics917646.1%466646.2%451045.9%1
Thiazide Diuretics333916.8%168316.7%165716.9%1
Selective Beta Blocker827841.5%420241.6%407741.5%0
Non-Selective Beta Blocker463623.3%236223.4%227423.1%1
Other Antihypertensive619531.1%314331.1%305231.1%0
Statins11,39157.2%578857.3%560357.0%1
Other Lipid-Lowering Medication12766.4%6306.2%6466.6%2
Proton-Pump Inhibitors690334.6%348734.5%341534.7%0
Aspirin618031.0%318131.5%299930.5%2
Digoxin3321.7%1641.6%1681.7%1
Glucocorticoids258413.0%132013.1%126412.9%1
Platelet Inhibitor317115.9%160515.9%156715.9%0
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,67273.6%737273.0%730074.3%3
1321716.1%165716.4%156115.9%1
2+203610.2%106810.6%9689.9%2
Number of visits in past year32.830.333.131.632.629.02
Addiction Medicine4912.5%2512.5%2402.4%1
Cardiology887444.5%455045.1%432444.0%2
Emergency Care613430.8%314631.2%298830.4%2
Endocrinology10585.3%5295.2%5295.4%1
Hospice Medicine1640.8%920.9%710.7%2
Intensive Care Unit13226.6%6846.8%6386.5%1
Internal Medicine641632.2%329332.6%312331.8%2
Oncology10525.3%5485.4%5045.1%1
Palliative Care3892.0%2132.1%1751.8%2
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.929.97.030.06.81
Systolic Blood Pressure, mmHg129.720.2129.820.9129.619.61
Diastolic Blood Pressure, mmHg72.312.072.412.272.211.72
Serum Magnesium, mg/dL
<1.7 mg/dL10705.4%5115.1%5595.7%3
1.7 to 2.5 mg/dL671133.7%344834.2%326333.2%2
>2.5 mg/dL1901.0%950.9%951.0%1
Unknown11,95460.0%604259.8%591260.2%1
Serum 25-hydroxy vitamin D < 20 ng/dL9564.8%4924.9%4644.7%1
Low-Density Lipoprotein, mg/dL85.834.585.836.285.832.90
Triglycerides, mg/dL145.5111.2145.5109.7145.6112.50
High-Density Lipoprotein, mg/dL43.214.243.314.543.213.91
Cholesterol, mg/dL155.143.2155.145.4155.141.00
Sodium, mmol/L138.93.3138.93.4138.93.10
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.69.20.50
Hemoglobin A1C, %6.81.56.81.66.81.50
Ejection Fraction
≤40%530726.6%268326.6%262426.7%0
>40%726936.5%370836.7%356136.2%1
Unknown734936.9%370436.7%364437.1%1

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Source: 3.1. Baseline Characteristics Source: 3.1. Baseline Characteristics Source: 3.1. Baseline Characteristics Source: 3.1. Baseline Characteristics Source: 3.1. Baseline Characteristics

Magnesium Supplement Use and Adverse Outcomes

Results: Magnesium Supplement Use and Adverse Outcomes

3.2.1. Descriptive Analysis

The IPT-weighted cohort was followed for up to 5 years (median follow-up: 0.7 years for magnesium supplement non-users and 0.6 years for magnesium supplement users), with a mean (±standard deviation [SD]) follow-up duration of 0.85 ± 0.96 years for magnesium supplement users and 0.82 ± 0.95 years for non-users.

Among the 9900 magnesium supplement users, 3532 (35.7%) had no follow-up clinical note referencing magnesium supplement use after the eligible date; for these patients, magnesium supplementation was considered discontinued one year after the last recorded clinical note. For the remaining 6368 (64.3%) magnesium users, the mean duration of magnesium supplement use was 1.9 ± 2.0 years (median: 1.1 years; interquartile range [IQR]: 0.6 to 2.6 years) without accounting for the timing of outcomes or censoring events and 0.8 ± 1.0 years (median 0.4 years; IQR: 0.1 to 1.0 years) when accounting for the occurrence of an outcome or a censoring event.

The incidence of all-cause hospitalization or death was 24.3% (297.8 events per 1000 person-years) among magnesium supplement users and 30.7% (360.2 events per 1000 person-years) among non-users in the weighted cohort (Table 4).

3.2.2. Main and Sensitivity Analyses

The weighted cohort showed that magnesium users had a significantly higher death-or-hospitalization-free survival compared to non-users (p < 0.0001 for the log-rank test), as shown in Figure 4a. Cox’s regression modeling indicated that magnesium supplement use was associated with a significantly reduced risk of incident all-cause hospitalization or death (HR: 0.81; 95% CI: 0.77 to 0.86; p < 0.0001).

Similar associations were observed in the sensitivity analyses of the subgroup who survived at least 6 months (n = 17,959; HR: 0.91; 95% CI: 0.85 to 0.97; p = 0.0024, Figure 4b) from the index date and in the restricted cohort who had at least one more clinic visit note during follow-up confirming sustained magnesium supplement use along with their paired controls (n = 12,736; HR: 0.77; 95% CI: 0.72 to 0.82; p < 0.0001, Figure 4c).

3.2.3. Interactions and Stratified Subgroup Analyses

The interaction between magnesium supplement use and follow-up time was not statistically significant (p = 0.7689), indicating that the proportional hazard assumption was not violated. The association between magnesium supplement use and a reduced risk of all-cause death or hospitalization was homogeneous, with no significant interactions, across subgroups of the IPTW-weighted cohort stratified by gender (p = 0.8413), BMI (p = 0.1875), serum vitamin D levels (p = 0.1469), prescribed multivitamin (p = 0.9108), diuretics (p = 0.1114), serum magnesium level (p = 0.7719), and left-ventricular ejection fraction (p = 0.1648).

The association between magnesium supplement use and reduced outcome risk appeared stronger in older individuals than younger ones (p = 0.0131 for interaction with age) and in the European American population than other racial subgroups (p = 0.0292 for interaction with race) (Figure 5a). Vitamin D prescription also significantly modified the association between magnesium supplement use and adverse outcomes, with an additional 8% risk reduction in adverse outcome in patients who were not prescribed with vitamin D compared to those who were prescribed (p = 0.0268 for interaction with vitamin D prescription).

Although diuretics did not have a differential effect on the association between magnesium supplementation and the outcomes, a slightly stronger effect size for magnesium supplement use was observed in diuretic users than in non-users.

Furthermore, no heterogeneity was observed in the association between magnesium supplement use and the outcomes when stratified by knowledge of serum magnesium levels (Figure 5b).

Table 4

All-cause death or hospitalization over 5 years of follow-up in the IPT-weighted cohort.

# Events# Patients at RiskEvent/Patients at Risk (%)Mean (STD) Follow-Up YearsMedian (IQR) Follow-Up YearsTotal Person YearsIncidence Rate (per 1000 Person-Years)
Magnesium non-users3013982930.7%0.85 (0.96)0.70
(0.10 to 1.00)
8365.5360.2
Magnesium users24521009624.3%0.82 (0.95)0.60
(0.10 to 1.00)
8234.1297.8

Abbreviations: #, number of; IPT, inverse probability of treatment; STD, standard deviation; IQR: interquartile range.

Source: 3.2. Magnesium Supplement Use and Adverse Outcomes Source: 3.2. Magnesium Supplement Use and Adverse Outcomes Source: 3.2. Magnesium Supplement Use and Adverse Outcomes Source: 3.2. Magnesium Supplement Use and Adverse Outcomes

Discussion

Our study found that patients with documented use of magnesium supplements within one year after diagnosed HF were associated with a 19% reduction in the risk of all-cause mortality or hospitalization compared to those without any recorded magnesium supplement use. The effect size was even greater among individuals with sustained use, showing a 23% relative risk reduction. However, when the analysis was limited to patients who survived at least six months from the index date, the effect size was less pronounced, with the KM curves beginning to diverge at around two years after initiation as opposed to within the first year.

Magnesium is essential for maintaining normal heart rhythm [22] and plays a critical role in the prevention of atrial fibrillation [23] and other types of cardiac arrhythmia [24]. Although many studies have shown that magnesium replacement is linked to a reduced risk of adverse outcomes related to HF [25,26,27], there is lack of research on the effects of magnesium supplementation. Only a few studies related to dietary magnesium have been reported.

A reverse association between dietary magnesium and the 10-year risk of atherosclerotic cardiovascular disease events was observed in a cross-sectional study conducted among 2980 participants aged 40 to 70 years, based on data from the National Health and Nutrition Examination Survey (1999 to 2018) [28]. Another observational study found that magnesium intake was significantly associated with a reduction in 28-day all-cause mortality (HR: 0.68; 95% CI: 0.54 to 0.86) among 1970 patients with HF with preserved ejection fraction (HFpEF) who had at least one admission to the intensive care unit (ICU) [29].

Our findings expand the prior studies to examine magnesium supplement use in a large cohort of veteran patients with various types of HF, with a follow-up period of up to 5 years.

These finding have important implications, given that approximately 6 million Americans are currently living with HF [30]. Dietary supplements, such as magnesium, are relatively low-cost and generally considered safe. The results suggest that magnesium supplementation in patients with HF who do not have end-stage renal disease may offer potential health benefits. The observed benefits were more pronounced in older individuals, particularly those aged 80 years or older. Since age-related declines in magnesium absorption and the use of some medications in the elderly may reduce magnesium levels [31,32], our findings add evidence to support dietary magnesium supplementation in patients at risk of magnesium deficit.

Regarding the heterogeneity of magnesium supplements by racial groups, it is likely related to variations in average dietary intake of magnesium and prevalence of certain chronic health conditions rather than a difference in how magnesium works in the body. Given the low sample size of non-white races in the study cohort, further investigation is needed.

Vitamin D and magnesium interact closely: vitamin D cannot function effectively when magnesium levels are low [33]. It is likely that individuals prescribed with vitamin D have low vitamin D levels, which may alter magnesium distribution and utilization. This may explain why magnesium supplementation had a stronger effect size in patients without a vitamin D prescription, presumably due to likely normal vitamin D levels.

Since diuretics may cause hypomagnesemia [34,35], we conducted a subgroup analysis stratified by diuretic use and observed slight heterogeneity, with more beneficial effects in diuretic users than in non-users. This finding suggests that magnesium supplementation should be recommended for diuretic users, as supported by the literature [36].

Our study has limitations. First, magnesium supplement use was identified through clinician-documented history rather than prescription data. This raises the possibility that some patients may have begun using magnesium supplements before their first reported use, which would reflect prevalent rather than incident use. In addition, data on adherence to magnesium supplements would also have the same misclassification bias. To minimize this bias, we used continuously documented use to document adherence. There is a likelihood of misclassifying the treatment strategies (i.e., magnesium users as non-users and vice versa due to self-report or documentation deficits), which may bias our estimates towards the null.

We also performed sensitivity analyses to assess the robustness of our results, which remained consistent. Also, the lack of a unified formulation or dose of magnesium supplements raises the question of what the ideal amount of elemental magnesium supplementation is and whether a dose: response relationship exists. According to the literature, the most commonly used formulations for magnesium supplements in the US (magnesium glycinate, citrate, and oxide) contain a range of elemental magnesium from 27 to 242 mg [37], most of which attempt to match the Recommended Dietary Allowance (RDA) dose for adult men and women (400 to 420 mg and 310 to 360 mg, respectively) [38].

Although we adjusted for a wide range of potential confounders, including social and environmental factors, to generate IPTWs, residual confounding may still be present, such as dietary and exercise factors. The social and environmental variables, including residential area median income and the EJI socio-environmental percentile ranking, were included to account for differences in health outcomes that may be influenced by healthier diets, lifestyles, and the activeness of health-seeking behaviors. Conversely, the study population was considered a high-risk and hard-to-recruit population in the real life, as reflected by the high number of adverse events occurring within the first six months.

However, the sensitivity analysis restricted to patients who survived at least six months showed consistent results and suggests that the results can also be generalizable to lower-risk patients with HF. Lastly, the generalizability of our findings to women with HF or community-based HF patients may be limited due to the predominantly elderly male veteran population included in the study. Despite this limitation, this study was still able to include 667 women. We chose veterans as the study population because the VHA databases contain comprehensive variables related to health history and records: such as comorbidities, laboratory results, pharmacy data, health care utilization, and vital signs: which may help address potential confounding and indication biases associated with both magnesium supplement use and the outcome.

Future studies need to be conducted in other populations.

Source: 4. Discussion Source: 4. Discussion Source: 4. Discussion Source: 4. Discussion Source: 4. Discussion Source: 4. Discussion Source: 4. Discussion

Results

3.1. Baseline Characteristics

Between 1 January 2000, and 31 December 2020, a total of 1,194,095 veterans received a first-time diagnosis of HF within the VHA system. For the target trial emulation, we enrolled all 9900 individuals who met eligibility criteria and used magnesium supplements and matched them to 9900 randomly selected non-users who had a clinic visit on the same index date. Patient characteristics before and after weighting are presented in Table 2 and Table 3 and Supplementary Figure S1.

Prior to weighting, many baseline characteristics were unbalanced (ASD > 10%) between treatment groups. Magnesium supplement users were older (73.3 vs. 72 years) and less likely to be African American (8.8% vs. 15.7%) than non-users. They also had a more favorable socio-environmental percentile ranking (46.3 vs. 49.6) and a shorter average interval between HF diagnosis and the index date (105.6 days vs. 153.7 days). Magnesium users were less likely than non-users to have a history of alcohol abuse (12.0% vs. 16.5%) or smoking (25.1% vs. 36.4%).

They had lower rates of hypertension (84.2% vs. 89.7%), hyperlipidemia (79.3% vs. 83.1%), myocardial infarction (15.5% vs. 20.5%), angina (13.6% vs. 20.7%), and other comorbid conditions, including history of chronic obstructive pulmonary disease (COPD), anemia, arthritis, cancer, urinary tract infection (UTI), pneumonia, respiratory failure, sepsis, depression, and fracture. However, they were more likely to have atrial fibrillation (36.8% vs. 30.9%) and low serum magnesium levels (7.6% vs. 2.5%) at baseline. In addition, magnesium supplement users were less likely to have a known left-ventricular ejection fraction (43.6% vs. 32.6% unknown) and were less likely to be treated with medications including insulin, antihypertensives, cholesterol medications, aspirin, glucocorticoids, and platelet inhibitors.

The prevalence of left-ventricular ejection fraction ≤ 40% was similar between the groups (26.9% vs. 25.4%).

The propensity score distributions of magnesium supplement users and non-users are shown in Figure 3, indicating adequate overlap and no violation of the positivity assumption. The mean and standard deviation of IPTWs were 1.01 ± 0.58 overall (0.99 ± 0.52 for non-magnesium users and 1.02 ± 0.62 for magnesium supplement users). The ASD values for all measured baseline characteristics in the weighted cohort were below 10%, indicating balance after weighting (Table 3). The IPT-weighted pseudo-cohort had a mean age of 72.6 ± 11.1 years and comprised 12.6% African Americans, 3.3% Hispanics, and 3.4% women.

Among these patients in the pseudo-cohort, 87.2% had hypertension, 33.9% had atrial fibrillation, 18.4% had myocardial infarction, and 12.7% had ischemic stroke. Furthermore, 14.5% and 31.5% of patients had a history of alcohol abuse and smoking, respectively. Overall, 39.8% of patients were treated with anti-diabetes medications and 77.1% with anti-hypertensive medications.

3.2. Magnesium Supplement Use and Adverse Outcomes

3.2.1. Descriptive Analysis

The IPT-weighted cohort was followed for up to 5 years (median follow-up: 0.7 years for magnesium supplement non-users and 0.6 years for magnesium supplement users), with a mean (±standard deviation [SD]) follow-up duration of 0.85 ± 0.96 years for magnesium supplement users and 0.82 ± 0.95 years for non-users.

Among the 9900 magnesium supplement users, 3532 (35.7%) had no follow-up clinical note referencing magnesium supplement use after the eligible date; for these patients, magnesium supplementation was considered discontinued one year after the last recorded clinical note. For the remaining 6368 (64.3%) magnesium users, the mean duration of magnesium supplement use was 1.9 ± 2.0 years (median: 1.1 years; interquartile range [IQR]: 0.6 to 2.6 years) without accounting for the timing of outcomes or censoring events and 0.8 ± 1.0 years (median 0.4 years; IQR: 0.1 to 1.0 years) when accounting for the occurrence of an outcome or a censoring event.

The incidence of all-cause hospitalization or death was 24.3% (297.8 events per 1000 person-years) among magnesium supplement users and 30.7% (360.2 events per 1000 person-years) among non-users in the weighted cohort (Table 4).

3.2.2. Main and Sensitivity Analyses

The weighted cohort showed that magnesium users had a significantly higher death-or-hospitalization-free survival compared to non-users (p < 0.0001 for the log-rank test), as shown in Figure 4a. Cox’s regression modeling indicated that magnesium supplement use was associated with a significantly reduced risk of incident all-cause hospitalization or death (HR: 0.81; 95% CI: 0.77 to 0.86; p < 0.0001).

Similar associations were observed in the sensitivity analyses of the subgroup who survived at least 6 months (n = 17,959; HR: 0.91; 95% CI: 0.85 to 0.97; p = 0.0024, Figure 4b) from the index date and in the restricted cohort who had at least one more clinic visit note during follow-up confirming sustained magnesium supplement use along with their paired controls (n = 12,736; HR: 0.77; 95% CI: 0.72 to 0.82; p < 0.0001, Figure 4c).

3.2.3. Interactions and Stratified Subgroup Analyses

The interaction between magnesium supplement use and follow-up time was not statistically significant (p = 0.7689), indicating that the proportional hazard assumption was not violated. The association between magnesium supplement use and a reduced risk of all-cause death or hospitalization was homogeneous, with no significant interactions, across subgroups of the IPTW-weighted cohort stratified by gender (p = 0.8413), BMI (p = 0.1875), serum vitamin D levels (p = 0.1469), prescribed multivitamin (p = 0.9108), diuretics (p = 0.1114), serum magnesium level (p = 0.7719), and left-ventricular ejection fraction (p = 0.1648).

The association between magnesium supplement use and reduced outcome risk appeared stronger in older individuals than younger ones (p = 0.0131 for interaction with age) and in the European American population than other racial subgroups (p = 0.0292 for interaction with race) (Figure 5a). Vitamin D prescription also significantly modified the association between magnesium supplement use and adverse outcomes, with an additional 8% risk reduction in adverse outcome in patients who were not prescribed with vitamin D compared to those who were prescribed (p = 0.0268 for interaction with vitamin D prescription).

Although diuretics did not have a differential effect on the association between magnesium supplementation and the outcomes, a slightly stronger effect size for magnesium supplement use was observed in diuretic users than in non-users.

Furthermore, no heterogeneity was observed in the association between magnesium supplement use and the outcomes when stratified by knowledge of serum magnesium levels (Figure 5b).

Table 2

Baseline characteristics of study participants before IPT weighting.

Baseline CharacteristicsOverall
(N = 19,800)
Magnesium
Users
(N = 9900)
Magnesium
Non-Users
(N = 9900)
ASD Before IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.173.310.772.011.412
Male19,13396.6%954496.4%958996.9%3
Race
White15,22076.9%787279.5%734874.2%13
Black242612.3%8698.8%155715.7%21
Others215410.9%115911.7%99510.1%5
Ethnicity
Non-Hispanics17,96590.7%899190.8%897490.6%1
Hispanics6403.2%2442.5%3964.0%8
Unknown11956.0%6656.7%5305.4%5
Married11,58358.5%626263.3%532153.7%20
Median Income (lowest to highest) of Zip Code
1st Quartile412820.8%191419.3%221422.4%8
2nd Quartile481124.3%247825.0%233323.6%3
3rd Quartile487124.6%247225.0%239924.2%2
4th Quartile553027.9%286528.9%266526.9%4
Unknown4602.3%1711.7%2892.9%8
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code48.022.646.322.249.622.915
Duration between HF Diagnosis and Index Date, Days129.6117.7105.6118.1153.7112.242
Comorbid Conditions (31 Variables)
Alcohol Disorder281814.2%118812.0%163016.5%13
Smoking608330.7%248325.1%360036.4%25
Hypertension17,21887.0%834084.2%887889.7%16
Hyperlipidemia16,07981.2%784979.3%823083.1%10
Myocardial Infarction356618.0%153615.5%203020.5%13
Atherosclerosis477024.1%220822.3%256225.9%8
Angina339417.1%134613.6%204820.7%19
Atrial Fibrillation670533.9%364636.8%305930.9%12
Cardiac Valve Disease343017.3%157415.9%185618.7%7
Ischemic Stroke246012.4%108310.9%137713.9%9
Hemorrhage Stroke1720.9%620.6%1101.1%5
Transient Ischemic Attack10615.4%4294.3%6326.4%9
Chronic Obstructive Pulmonary Disease712536.0%332433.6%380138.4%10
Asthma18209.2%8568.6%9649.7%4
Chronic Kidney Disease561228.3%267227.0%294029.7%6
Anemia654133.0%300830.4%353335.7%11
Arthritis934747.2%439844.4%494950.0%11
Cancer675334.1%312931.6%362436.6%11
Neurological Disorders755638.2%357236.1%398440.2%8
Parkinson4172.1%1811.8%2362.4%4
Hypothyroidism288714.6%151115.3%137613.9%4
Osteoporosis8664.4%4204.2%4464.5%1
Osteomyelitis5002.5%1831.8%3173.2%9
Urinary Tract Infection12386.3%4874.9%7517.6%11
Liver Disease18759.5%8288.4%104710.6%8
Pneumonia248112.5%9179.3%156415.8%20
Respiratory Failure10875.5%3733.8%7147.2%15
Sepsis6623.3%2102.1%4524.6%14
Opioid Use Disorder4202.1%1501.5%2702.7%8
Depression688334.8%299830.3%388539.2%19
Fracture222311.2%9159.2%130813.2%13
Concurrent Medications (22 Variables)
Insulin504125.5%221422.4%282728.6%14
Metformin394919.9%185518.7%209421.2%6
Glucagon-Like Peptide-1780.4%340.3%440.4%2
Sodium: Glucose Cotrasporter-2 Inhibitors1951.0%1001.0%951.0%0
Other Diabetes Medication322116.3%147014.8%175117.7%8
Vitamin D Prescription361818.3%143314.5%218522.1%20
Multivitamin (Rx)9724.9%3903.9%5825.9%9
Angiotensin-Converting Enzyme Inhibitor802140.5%322932.6%479248.4%33
Angiotensin Receptor Blocker313715.8%144614.6%169117.1%7
Calcium Channel Blocker544227.5%221122.3%323132.6%23
Loop Diuretics876544.3%375037.9%501550.7%26
Thiazide Diuretics326216.5%137213.9%189019.1%14
Selective Beta Blocker799640.4%328733.2%470947.6%30
Non-Selective Beta Blocker442522.3%191319.3%251225.4%15
Other Antihypertensive593130.0%259826.2%333333.7%16
Statins11,12556.2%467947.3%644665.1%36
Other Lipid-Lowering Medication12586.4%5735.8%6856.9%5
Proton-Pump Inhibitors666233.6%294029.7%372237.6%17
Aspirin589429.8%206720.9%382738.7%40
Digoxin3341.7%1541.6%1801.8%2
Glucocorticoids244312.3%9729.8%147114.9%16
Platelet Inhibitor305815.4%128012.9%177818.0%14
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,86075.1%836484.5%649665.6%45
1303515.3%9769.9%205920.8%31
2+19059.6%5605.7%134513.6%27
Number of visits in past year31.529.526.828.036.230.232
Addiction Medicine4832.4%1861.9%2973.0%7
Cardiology855143.2%317032.0%538154.4%46
Emergency Care580029.3%202520.5%377538.1%39
Endocrinology9905.0%4254.3%5655.7%6
Hospice Medicine1380.7%540.5%840.8%4
Intensive Care Unit12556.3%3793.8%8768.8%21
Internal Medicine618431.2%262826.5%355635.9%20
Oncology9804.9%4044.1%5765.8%8
Palliative Care3381.7%1151.2%2232.3%8
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.930.26.929.86.96
Systolic Blood Pressure, mmHg129.519.9127.919.8131.019.916
Diastolic Blood Pressure, mmHg72.111.771.311.572.911.914
Serum Magnesium, mg/dL
<1.7 mg/dL10035.1%7577.6%2462.5%23
1.7 to 2.5 mg/dL645432.6%260826.3%384638.8%27
>2.5 mg/dL1891.0%900.9%991.0%1
Unknown12,15461.4%644565.1%570957.7%15
Serum 25-hydroxy vitamin D < 20 ng/dL9274.7%3823.9%5455.5%8
Low-Density Lipoprotein, mg/dL85.934.285.134.586.533.94
Triglycerides, mg/dL146.3111.5147.1110.3145.6112.61
High-Density Lipoprotein, mg/dL43.214.243.314.443.014.02
Cholesterol, mg/dL155.342.6154.542.9156.042.34
Sodium, mmol/L138.93.2138.73.3139.03.19
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.59.20.50
Hemoglobin A1C, %6.81.56.91.56.81.57
Ejection Fraction
≤40%517126.1%260026.9%251125.4%3
>40%709135.8%292829.6%416342.1%26
Unknown753838.1%431243.6%322632.6%23

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Table 3

Baseline characteristics of study participants after IPT weighting.

Baseline CharacteristicsIPT-Weighted
Overall
(N = 19,925)
IPT-Weighted Magnesium
Users
(N = 10,096)
IPT-Weighted Magnesium
Non-Users
(N = 9829)
ASD After IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.172.510.872.611.41
Male19,24096.6%974896.6%949296.6%0
Race
White15,26076.6%771176.4%755076.8%1
Black251812.6%130012.9%121912.4%2
Others214710.8%108610.8%106110.8%0
Ethnicity
Non-Hispanics18,10590.9%917190.8%893590.9%0
Hispanics6543.3%3363.3%3183.2%1
Unknown11665.9%5895.8%5775.9%0
Married11,56558.0%582857.7%573758.4%1
Median Income (lowest to highest) of Zip Code
1st Quartile419721.1%213921.2%205820.9%1
2nd Quartile479724.1%242424.0%237324.1%0
3rd Quartile482524.2%243824.1%238724.3%0
4th Quartile562428.2%284228.1%278228.3%0
Unknown4822.4%2532.5%2292.3%1
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code47.922.748.022.947.822.61
Duration between Diabetes Diagnosis and Index Date, Days135.0118.4136.0125.8134.0110.52
Comorbid Conditions (31 Variables)
Alcohol Disorder288814.5%147114.6%141814.4%1
Smoking626731.5%321031.8%305831.1%2
Hypertension17,38287.2%880487.2%857887.3%0
Myocardial Infarction367018.4%186618.5%180418.4%0
Atherosclerosis488424.5%247824.5%240524.5%0
Angina354817.8%181718.0%173117.6%1
Atrial Fibrillation675633.9%342433.9%333133.9%0
Cardiac Valve Disease356517.9%182418.1%174117.7%1
Ischemic Stroke253012.7%128412.7%124612.7%0
Hemorrhage Stroke1850.9%971.0%880.9%1
Transient Ischemic Attack10925.5%5525.5%5395.5%0
Chronic Obstructive Pulmonary Disease726336.5%370336.7%356036.2%1
Asthma19229.6%9899.8%9349.5%1
Chronic Kidney Disease576929.0%291928.9%285029.0%0
Anemia676734.0%345234.2%331433.7%1
Arthritis953247.8%482047.7%471247.9%0
Cancer690534.7%348834.6%341734.8%0
Neurological Disorders778539.1%395339.1%383239.0%0
Parkinson4212.1%2182.2%2042.1%1
Hypothyroidism294414.8%149814.8%144614.7%0
Osteoporosis8884.5%4594.6%4284.4%1
Osteomyelitis5212.6%2682.7%2542.6%1
Hyperlipidemia16,30681.8%824481.7%806282.0%1
Urinary Tract Infection13446.7%7117.0%6336.4%2
Liver Disease19719.9%100610.0%9659.8%1
Pneumonia259213.0%134113.3%125112.7%2
Respiratory Failure11755.9%6226.2%5535.6%3
Sepsis6993.5%3623.6%3383.4%1
Opioid Use Disorder4372.2%2292.3%2082.1%1
Depression705035.4%359535.6%345435.1%1
Fracture237811.9%123512.2%114411.6%2
Concurrent Medications (22 Variables)
Insulin521926.2%265626.3%256426.1%0
Metformin409920.6%204920.3%205120.9%1
Glucagon-Like Peptide-1840.4%420.4%420.4%0
Sodium: Glucose Cotrasporter-2 Inhibitors2081.0%1061.0%1031.0%0
Other Diabetes Medication331516.6%166316.5%165216.8%1
Vitamin D Prescription378919.0%194119.2%184818.8%1
Multivitamin (Rx)10445.2%5365.3%5075.2%0
Angiotensin-Converting Enzyme Inhibitor828041.6%422741.9%405341.2%1
Angiotensin Receptor Blocker325316.3%163316.2%162016.5%1
Calcium Channel Blocker566328.4%289228.6%277128.2%1
Loop Diuretics917646.1%466646.2%451045.9%1
Thiazide Diuretics333916.8%168316.7%165716.9%1
Selective Beta Blocker827841.5%420241.6%407741.5%0
Non-Selective Beta Blocker463623.3%236223.4%227423.1%1
Other Antihypertensive619531.1%314331.1%305231.1%0
Statins11,39157.2%578857.3%560357.0%1
Other Lipid-Lowering Medication12766.4%6306.2%6466.6%2
Proton-Pump Inhibitors690334.6%348734.5%341534.7%0
Aspirin618031.0%318131.5%299930.5%2
Digoxin3321.7%1641.6%1681.7%1
Glucocorticoids258413.0%132013.1%126412.9%1
Platelet Inhibitor317115.9%160515.9%156715.9%0
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,67273.6%737273.0%730074.3%3
1321716.1%165716.4%156115.9%1
2+203610.2%106810.6%9689.9%2
Number of visits in past year32.830.333.131.632.629.02
Addiction Medicine4912.5%2512.5%2402.4%1
Cardiology887444.5%455045.1%432444.0%2
Emergency Care613430.8%314631.2%298830.4%2
Endocrinology10585.3%5295.2%5295.4%1
Hospice Medicine1640.8%920.9%710.7%2
Intensive Care Unit13226.6%6846.8%6386.5%1
Internal Medicine641632.2%329332.6%312331.8%2
Oncology10525.3%5485.4%5045.1%1
Palliative Care3892.0%2132.1%1751.8%2
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.929.97.030.06.81
Systolic Blood Pressure, mmHg129.720.2129.820.9129.619.61
Diastolic Blood Pressure, mmHg72.312.072.412.272.211.72
Serum Magnesium, mg/dL
<1.7 mg/dL10705.4%5115.1%5595.7%3
1.7 to 2.5 mg/dL671133.7%344834.2%326333.2%2
>2.5 mg/dL1901.0%950.9%951.0%1
Unknown11,95460.0%604259.8%591260.2%1
Serum 25-hydroxy vitamin D < 20 ng/dL9564.8%4924.9%4644.7%1
Low-Density Lipoprotein, mg/dL85.834.585.836.285.832.90
Triglycerides, mg/dL145.5111.2145.5109.7145.6112.50
High-Density Lipoprotein, mg/dL43.214.243.314.543.213.91
Cholesterol, mg/dL155.143.2155.145.4155.141.00
Sodium, mmol/L138.93.3138.93.4138.93.10
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.69.20.50
Hemoglobin A1C, %6.81.56.81.66.81.50
Ejection Fraction
≤40%530726.6%268326.6%262426.7%0
>40%726936.5%370836.7%356136.2%1
Unknown734936.9%370436.7%364437.1%1

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Table 4

All-cause death or hospitalization over 5 years of follow-up in the IPT-weighted cohort.

# Events# Patients at RiskEvent/Patients at Risk (%)Mean (STD) Follow-Up YearsMedian (IQR) Follow-Up YearsTotal Person YearsIncidence Rate (per 1000 Person-Years)
Magnesium non-users3013982930.7%0.85 (0.96)0.70
(0.10 to 1.00)
8365.5360.2
Magnesium users24521009624.3%0.82 (0.95)0.60
(0.10 to 1.00)
8234.1297.8

Abbreviations: #, number of; IPT, inverse probability of treatment; STD, standard deviation; IQR: interquartile range.

Source: 3. Results Source: 3. Results Source: 3. Results Source: 3. Results Source: 3. Results

Conclusions

In conclusion, our findings indicate that magnesium supplementation is linked to a lower risk of all-cause mortality or hospitalization in patients with HF. These promising outcomes merit further exploration and validation through RCTs.

Source: 5. Conclusions
Authors’ abstract

Background: Individuals with heart failure (HF) are at increased risk of magnesium deficiency. Magnesium supplements are widely available and being used without clear evidence of efficacy in HF. Methods: We emulated a target trial to assess the association between magnesium supplements and adverse outcomes in U.S. veterans with newly diagnosed HF. Eligible patients were outpatients who received ambulatory care in the Veterans Health Administration between 1 January 2000 and 31 December 2020. Veterans with a hospitalization within 30 days prior to the eligible date, previous magnesium supplement or replacement use, or end-stage renal disease were ineligible for the trial.

Initial self-reported magnesium supplement use (measured at eligible date) was identified in medical records using natural language processing and then checked repeatedly to confirm continuous use. The outcome was all-cause hospitalization or death. Patients were followed for up to five years from the eligible date and were censored if they changed from the assigned treatment strategy or initiated prescribed magnesium replacement. We applied inverse probability treatment weighting and Cox’s regression to estimate hazard ratios (HRs), with sensitivity analyses in patients surviving ≥ 6 months and those with continuous documentation of magnesium supplement use.

Results: We enrolled 9900 magnesium supplement users and 9900 matched non-users. In the weighted cohort (mean age 72.6 years; 12.6% African American; 3.4% women; median follow-up 0.7 years), users had significantly better survival in both primary and sensitivity analyses (HR in primary analysis: 0.81 [0.77 to 0.86], p < 0.0001; HRs in sensitivity analyses: 0.91 [0.85 to 0.97], p = 0.0025 and 0.77 [0.72 to 0.82], p < 0.0001, respectively). Conclusions: magnesium supplement use was associated with a reduced risk of all-cause mortality or hospitalization among veterans with HF.

Read the full paper

1. Introduction

Magnesium is an essential intracellular mineral involved in numerous key enzymatic processes in the human body [1] and plays a critical role in preventing arrhythmia, improving hemodynamics, and supporting cardiovascular function [2,3]. However, individuals with heart failure (HF) are at increased risk of magnesium deficiency [4]. Findings from two studies suggest that low magnesium dietary intake and low plasma magnesium levels are linked to increased risk of HF-related hospitalizations [5,6].

Additionally, a low plasma magnesium level was also associated with a higher risk of cardiovascular death (hazard ratio [HR]: 1.38; 95% confidence interval [CI]: 1.04 to 1.83; p = 0.024) in 1569 patients with chronic HF [7]. However, the literature does not suggest that dietary intake alone relates to changes in cardiovascular outcomes. Our previous study showed that dietary magnesium intake was not related to mortality in women with HF [8]. Another study showed that although low serum magnesium was associated with a higher risk of arrhythmias, there was no observed association between dietary magnesium intake and arrhythmia risk [9].

Interestingly, magnesium supplementation has been shown to reduce the risk of cardiac arrhythmias in patients with HF in two randomized controlled trials (RCTs) [10]. But it remains unknown if magnesium supplementation has an effect on all-cause mortality or hospitalization in patients with HF.

To assess the effects of magnesium supplementation, rigorously designed RCTs are ideal. However, RCTs are often expensive for a nutritional supplement, as they typically require large participant cohorts and extended follow-up periods to detect relatively small effect sizes [11,12,13]. Target trial emulation is designed to replicate the structure of an RCT on the observational data to evaluate clinical effectiveness while controlling the biases inherent to observational research [14,15,16,17].

This study aimed to assess the association between magnesium supplementation and all-cause hospitalization or death using a target trial emulation.

2. Material & Methods

2.1. Study Population

Study subjects were patients newly diagnosed with HF between 1 January 2000 and 31 December 2020, identified from the US Veterans Health Administration (VHA) national electronic health records (EHR). HF was defined based on the International Classification of Diseases (ICD) codes (ICD9: 398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.13, 404.91, 428; ICD10: I11.0, I13.0, I13.2, I50). The diagnoses were identified from both inpatient and outpatient encounter data in the VHA system.

2.2. Study Design and Ethics

This was a retrospective cohort study. The study was approved by the VHA Central Institutional Review Board (IRBNET#1710776-1, 23 April 2021), which granted a waiver of informed consent. Due to the sensitive nature of the data involved, access to the dataset is restricted; qualified researchers trained in human subject confidentiality may request access by contacting the Washington DC VA Medical Center via email at helen.sheriff@va.gov.

2.3. Target Trial Emulation

We emulated a target trial by “enrolling” eligible patients into the study cohort. The elements of the target trial and the emulation are summarized in Table 1. A patient would be qualified for inclusion in a trial if, on an outpatient clinical note date, they met the following criteria: (1) had the initial HF diagnosis within 1 year before, (2) had at least two outpatient clinical encounter notes within the previous 12 months (to confirm magnesium supplement use status), (3) had no records of prescribed or supplemental magnesium use prior to this date documented in the electronic medical records (EMR), (4) had no hospitalization within 30 days before or at the eligible date, and (5) had no end-stage-renal-disease (ESRD) any time before or at the eligible date.

The purpose of requiring “no hospitalization within 30 days before or on the eligible date” was to exclude patients who were acutely ill, since these patients would be more likely to die earlier and have less opportunity to initiate magnesium supplements, thereby further reducing selection bias.

2.4. Exposure

The primary exposure of interest was EMR documentation of the patients’ current use of over-the-counter magnesium supplements, as documented in outpatient clinic visit notes. We identified this information using natural language processing (NLP) techniques [18] developed by our team to extract mentions of magnesium supplement use from the clinical notes. The NLP process and performance in identifying magnesium supplement use is described in detail elsewhere [18]. Validation of the NLP method used for dietary supplement and vitamin use showed an AUC of 98.9%, a precision of 96.2%, and a recall of 87.7% [18].

Eligible date: The status of magnesium supplement use on the date of the outpatient clinic visit when a patient became eligible for the target trial is the eligible date (index date, time 0). That is also the date the patient was classified as a magnesium supplement user or non-user (Table 1).

Exposure status during follow-up: For both groups, any patient without clinic visit notes for more than one year was considered lost to follow-up and was censored one year after their last documented clinic visit. For magnesium supplement users, if magnesium supplement use was not mentioned in the notes for over one year, the magnesium supplement use was considered discontinued one year after the last recorded use. If a patient was initially classified as a non-user of magnesium supplement but a clinic note during follow-up indicated magnesium supplement use, this was treated as a change in exposure status, and the patient would be censored at that time point.

To maximize the sample size of magnesium supplement users, we included all patients who were qualified to be a user. Then, we matched a randomly selected group of controls who had a clinic visit that was on the same date as a user in a 1:1 ratio.

The process of the target trial emulation is displayed in Figure 1. The graphical depiction of the study design is displayed in Figure 2. As described earlier, both groups were required to meet the same eligibility criteria. We used the same approach to identify exposure status and define the index date. Regardless of whether participants were users or non-users, the follow-up began on the eligibility date. All these measures were implemented to control for immortal time bias.

2.5. Study Outcomes and Follow-Up

The main study outcome was all-cause hospitalization or death, as identified from the VHA Corporate Data Warehouse (CDW). Death dates were recorded in the patient vital status table and hospitalizations were recorded in the inpatient table. Patients were followed from the index date until the outcome of interest occurred or the censoring date due to magnesium supplement use status change (i.e., stop use in users or start use in non-users), prescribed magnesium replacement was started, or the end of 5th year since the index date, whichever occurred first.

To account for potential differences in follow-up time since magnesium supplement non-users are more likely to maintain the same status than users, a magnesium supplement user and their paired non-user with the same index date were censored at the same date if either side were censored.

2.6. Study Covariates

We included an extensive list of baseline variables as covariates aimed at controlling for potential confounding and to ensure comparability between the two treatment strategy groups. Baseline covariates included demographics, socio-environmental and economic status, duration between the initial HF diagnosis and the index date, comorbid conditions, medication use, laboratory data, vital signs, and healthcare utilization.

Median household income and the socio-environmental percentile ranking were estimated at the patients’ residential address zip code. The socio-environmental percentile Environmental Justice Index (EJI) developed by the Center for Disease Control (CDC) was used to measure the cumulative impact of environmental burden on human health and the environment at the patients’ residence. A higher percentile EJI ranking indicates a greater cumulative socio-environmental burden on the community [19]. Comorbid conditions, identified using ICD codes, were captured any time before or at the index date.

Other covariates were captured within one year prior to and up to the index date, with the measures closest to the index date used as baseline data, which included medication use from prescription/fill records, laboratory tests using LOINC codes, vital signs from vital sign records, and healthcare utilization from outpatient and inpatient encounter records.

2.7. Variables with Missing Values

For vital signs and laboratory values, if the missing value rate was less than 30%, we imputed the values using the mean, conditional on age, gender, race, and ethnicity; otherwise, we converted the variables into categorical ones using missing as one of the possible categories to adjust for potential bias from informative missing. The following variables were imputed, with missing rates (magnesium supplement users vs. non-users) as follows: 4.7% (5.4% vs. 4.1%) for body mass index (BMI), 2.5% (2.9% vs. 2.2%) for systolic blood pressure, 2.6% (3.0% vs. 2.2%) for diastolic blood pressure, 23.4% (26.5% vs. 20.3%) for low-density lipoprotein, 23.9% (27.2% vs. 20.6%) for triglycerides, 22.7% (25.9% vs. 19.5%) for high-density lipoprotein, 22.4% (25.6% vs. 19.2%) for total cholesterol, 13.7% (18.0% vs. 9.3%) for serum sodium, 13.6% (17.9% vs. 9.4%) for serum potassium, 25.6% (28.4% vs. 22.8%) for serum creatinine, 18.4% (22.9% vs. 13.9%) for serum calcium, and 29.7% (32.3% vs. 27.0%) for hemoglobin A1C.

This indicates that magnesium supplement users had a slightly higher rate of missing values.

2.8. Statistical Analysis

We used the per-protocol approach for data analysis as described earlier, and patients were censored when they changed the treatment strategy.

We fit a logistic regression model to estimate the propensity score (i.e., the probability of being a magnesium supplement user) using baseline characteristics including demographics, comorbid conditions, concurrent medications, health care utilization in the past year, vital signs and labs. Then, we applied inverse probability treatment weighting (IPTW) to generate a pseudo-population, in which baseline characteristics of two treatment strategy groups were balanced (ASD < 10%). Stabilized weights (SW) were calculated for each patient. To avoid extreme values, when a weight was above 10, it was trimmed down to 10 [20].

The absolute standardized difference (ASD) for each characteristic was calculated before and after weighting, with an ASD > 10% indicating a significant imbalance in characteristics between the two groups [21].

We used the inverse probability of treatment (IPT)-weighted cohort to estimate the outcome incidence rates of the two groups. Kaplan: Meier curves were plotted, and Cox’s regression modeling was applied to the weighted cohort to assess the association between magnesium supplement use and the risk of all-cause hospitalization or death. The proportional hazards assumption was tested by including an interaction term between magnesium supplement use and time in the Cox’s regression model to examine any time-dependent effects on outcomes.

Additionally, we evaluated statistical interactions between magnesium supplement use and potential effect modifiers, including age, race, serum magnesium level, serum vitamin D level, vitamin D prescription, multivitamin use, and diuretic use, by incorporating a product term in the regression models. When a product term was statistically significant, subgroup analyses were conducted. In these subgroup analyses, we repeated the IPTW process to assess the association between magnesium supplement use and adverse outcomes within each subgroup.

2.9. Sensitivity Analyses

We also conducted two sensitivity analyses. In the first sensitivity analysis, we restricted the cohort to the subgroup who did not die within 6 months after the index date to exclude a very ill population unlikely to have benefited from magnesium supplement use. In the second sensitivity analysis, we restricted the cohort to the subgroup who had at least one additional clinical note mentioning magnesium supplement use and their paired controls to minimize misclassification bias. For each sensitivity analysis, we recalculated the IPTW to make magnesium supplement users and non-users comparable.

All analyses were performed using SAS (Version 9.4). A two-sided p-value of <0.05 was considered statistically significant.

2.10. Target Trial Assumption Evaluation

Consistency: We performed sensitivity analyses mentioned above that assessed the robustness of the results to potential violations of consistency, addressing issues such as exposure-related missing data, time-varying confounding, or measurement error.

Positivity: We evaluated the probability of being exposed or unexposed to magnesium supplement use within each subgroup, based on baseline characteristics, to ensure that observations existed for both exposure status in every subgroup. The positivity assumption was satisfied in this study.

Exchangeability: Although there is always a possibility of unknown confounding, we included an expansive number of variables mentioned above to control for potential confounding bias.

3. Results

3.1. Baseline Characteristics

Between 1 January 2000, and 31 December 2020, a total of 1,194,095 veterans received a first-time diagnosis of HF within the VHA system. For the target trial emulation, we enrolled all 9900 individuals who met eligibility criteria and used magnesium supplements and matched them to 9900 randomly selected non-users who had a clinic visit on the same index date. Patient characteristics before and after weighting are presented in Table 2 and Table 3 and Supplementary Figure S1.

Prior to weighting, many baseline characteristics were unbalanced (ASD > 10%) between treatment groups. Magnesium supplement users were older (73.3 vs. 72 years) and less likely to be African American (8.8% vs. 15.7%) than non-users. They also had a more favorable socio-environmental percentile ranking (46.3 vs. 49.6) and a shorter average interval between HF diagnosis and the index date (105.6 days vs. 153.7 days). Magnesium users were less likely than non-users to have a history of alcohol abuse (12.0% vs. 16.5%) or smoking (25.1% vs. 36.4%).

They had lower rates of hypertension (84.2% vs. 89.7%), hyperlipidemia (79.3% vs. 83.1%), myocardial infarction (15.5% vs. 20.5%), angina (13.6% vs. 20.7%), and other comorbid conditions, including history of chronic obstructive pulmonary disease (COPD), anemia, arthritis, cancer, urinary tract infection (UTI), pneumonia, respiratory failure, sepsis, depression, and fracture. However, they were more likely to have atrial fibrillation (36.8% vs. 30.9%) and low serum magnesium levels (7.6% vs. 2.5%) at baseline. In addition, magnesium supplement users were less likely to have a known left-ventricular ejection fraction (43.6% vs. 32.6% unknown) and were less likely to be treated with medications including insulin, antihypertensives, cholesterol medications, aspirin, glucocorticoids, and platelet inhibitors.

The prevalence of left-ventricular ejection fraction ≤ 40% was similar between the groups (26.9% vs. 25.4%).

The propensity score distributions of magnesium supplement users and non-users are shown in Figure 3, indicating adequate overlap and no violation of the positivity assumption. The mean and standard deviation of IPTWs were 1.01 ± 0.58 overall (0.99 ± 0.52 for non-magnesium users and 1.02 ± 0.62 for magnesium supplement users). The ASD values for all measured baseline characteristics in the weighted cohort were below 10%, indicating balance after weighting (Table 3). The IPT-weighted pseudo-cohort had a mean age of 72.6 ± 11.1 years and comprised 12.6% African Americans, 3.3% Hispanics, and 3.4% women.

Among these patients in the pseudo-cohort, 87.2% had hypertension, 33.9% had atrial fibrillation, 18.4% had myocardial infarction, and 12.7% had ischemic stroke. Furthermore, 14.5% and 31.5% of patients had a history of alcohol abuse and smoking, respectively. Overall, 39.8% of patients were treated with anti-diabetes medications and 77.1% with anti-hypertensive medications.

3.2. Magnesium Supplement Use and Adverse Outcomes

3.2.1. Descriptive Analysis

The IPT-weighted cohort was followed for up to 5 years (median follow-up: 0.7 years for magnesium supplement non-users and 0.6 years for magnesium supplement users), with a mean (±standard deviation [SD]) follow-up duration of 0.85 ± 0.96 years for magnesium supplement users and 0.82 ± 0.95 years for non-users.

Among the 9900 magnesium supplement users, 3532 (35.7%) had no follow-up clinical note referencing magnesium supplement use after the eligible date; for these patients, magnesium supplementation was considered discontinued one year after the last recorded clinical note. For the remaining 6368 (64.3%) magnesium users, the mean duration of magnesium supplement use was 1.9 ± 2.0 years (median: 1.1 years; interquartile range [IQR]: 0.6 to 2.6 years) without accounting for the timing of outcomes or censoring events and 0.8 ± 1.0 years (median 0.4 years; IQR: 0.1 to 1.0 years) when accounting for the occurrence of an outcome or a censoring event.

The incidence of all-cause hospitalization or death was 24.3% (297.8 events per 1000 person-years) among magnesium supplement users and 30.7% (360.2 events per 1000 person-years) among non-users in the weighted cohort (Table 4).

3.2.2. Main and Sensitivity Analyses

The weighted cohort showed that magnesium users had a significantly higher death-or-hospitalization-free survival compared to non-users (p < 0.0001 for the log-rank test), as shown in Figure 4a. Cox’s regression modeling indicated that magnesium supplement use was associated with a significantly reduced risk of incident all-cause hospitalization or death (HR: 0.81; 95% CI: 0.77 to 0.86; p < 0.0001).

Similar associations were observed in the sensitivity analyses of the subgroup who survived at least 6 months (n = 17,959; HR: 0.91; 95% CI: 0.85 to 0.97; p = 0.0024, Figure 4b) from the index date and in the restricted cohort who had at least one more clinic visit note during follow-up confirming sustained magnesium supplement use along with their paired controls (n = 12,736; HR: 0.77; 95% CI: 0.72 to 0.82; p < 0.0001, Figure 4c).

3.2.3. Interactions and Stratified Subgroup Analyses

The interaction between magnesium supplement use and follow-up time was not statistically significant (p = 0.7689), indicating that the proportional hazard assumption was not violated. The association between magnesium supplement use and a reduced risk of all-cause death or hospitalization was homogeneous, with no significant interactions, across subgroups of the IPTW-weighted cohort stratified by gender (p = 0.8413), BMI (p = 0.1875), serum vitamin D levels (p = 0.1469), prescribed multivitamin (p = 0.9108), diuretics (p = 0.1114), serum magnesium level (p = 0.7719), and left-ventricular ejection fraction (p = 0.1648).

The association between magnesium supplement use and reduced outcome risk appeared stronger in older individuals than younger ones (p = 0.0131 for interaction with age) and in the European American population than other racial subgroups (p = 0.0292 for interaction with race) (Figure 5a). Vitamin D prescription also significantly modified the association between magnesium supplement use and adverse outcomes, with an additional 8% risk reduction in adverse outcome in patients who were not prescribed with vitamin D compared to those who were prescribed (p = 0.0268 for interaction with vitamin D prescription).

Although diuretics did not have a differential effect on the association between magnesium supplementation and the outcomes, a slightly stronger effect size for magnesium supplement use was observed in diuretic users than in non-users.

Furthermore, no heterogeneity was observed in the association between magnesium supplement use and the outcomes when stratified by knowledge of serum magnesium levels (Figure 5b).

4. Discussion

Our study found that patients with documented use of magnesium supplements within one year after diagnosed HF were associated with a 19% reduction in the risk of all-cause mortality or hospitalization compared to those without any recorded magnesium supplement use. The effect size was even greater among individuals with sustained use, showing a 23% relative risk reduction. However, when the analysis was limited to patients who survived at least six months from the index date, the effect size was less pronounced, with the KM curves beginning to diverge at around two years after initiation as opposed to within the first year.

Magnesium is essential for maintaining normal heart rhythm [22] and plays a critical role in the prevention of atrial fibrillation [23] and other types of cardiac arrhythmia [24]. Although many studies have shown that magnesium replacement is linked to a reduced risk of adverse outcomes related to HF [25,26,27], there is lack of research on the effects of magnesium supplementation. Only a few studies related to dietary magnesium have been reported.

A reverse association between dietary magnesium and the 10-year risk of atherosclerotic cardiovascular disease events was observed in a cross-sectional study conducted among 2980 participants aged 40 to 70 years, based on data from the National Health and Nutrition Examination Survey (1999 to 2018) [28]. Another observational study found that magnesium intake was significantly associated with a reduction in 28-day all-cause mortality (HR: 0.68; 95% CI: 0.54 to 0.86) among 1970 patients with HF with preserved ejection fraction (HFpEF) who had at least one admission to the intensive care unit (ICU) [29].

Our findings expand the prior studies to examine magnesium supplement use in a large cohort of veteran patients with various types of HF, with a follow-up period of up to 5 years.

These finding have important implications, given that approximately 6 million Americans are currently living with HF [30]. Dietary supplements, such as magnesium, are relatively low-cost and generally considered safe. The results suggest that magnesium supplementation in patients with HF who do not have end-stage renal disease may offer potential health benefits. The observed benefits were more pronounced in older individuals, particularly those aged 80 years or older. Since age-related declines in magnesium absorption and the use of some medications in the elderly may reduce magnesium levels [31,32], our findings add evidence to support dietary magnesium supplementation in patients at risk of magnesium deficit.

Regarding the heterogeneity of magnesium supplements by racial groups, it is likely related to variations in average dietary intake of magnesium and prevalence of certain chronic health conditions rather than a difference in how magnesium works in the body. Given the low sample size of non-white races in the study cohort, further investigation is needed.

Vitamin D and magnesium interact closely: vitamin D cannot function effectively when magnesium levels are low [33]. It is likely that individuals prescribed with vitamin D have low vitamin D levels, which may alter magnesium distribution and utilization. This may explain why magnesium supplementation had a stronger effect size in patients without a vitamin D prescription, presumably due to likely normal vitamin D levels.

Since diuretics may cause hypomagnesemia [34,35], we conducted a subgroup analysis stratified by diuretic use and observed slight heterogeneity, with more beneficial effects in diuretic users than in non-users. This finding suggests that magnesium supplementation should be recommended for diuretic users, as supported by the literature [36].

Our study has limitations. First, magnesium supplement use was identified through clinician-documented history rather than prescription data. This raises the possibility that some patients may have begun using magnesium supplements before their first reported use, which would reflect prevalent rather than incident use. In addition, data on adherence to magnesium supplements would also have the same misclassification bias. To minimize this bias, we used continuously documented use to document adherence. There is a likelihood of misclassifying the treatment strategies (i.e., magnesium users as non-users and vice versa due to self-report or documentation deficits), which may bias our estimates towards the null.

We also performed sensitivity analyses to assess the robustness of our results, which remained consistent. Also, the lack of a unified formulation or dose of magnesium supplements raises the question of what the ideal amount of elemental magnesium supplementation is and whether a dose: response relationship exists. According to the literature, the most commonly used formulations for magnesium supplements in the US (magnesium glycinate, citrate, and oxide) contain a range of elemental magnesium from 27 to 242 mg [37], most of which attempt to match the Recommended Dietary Allowance (RDA) dose for adult men and women (400 to 420 mg and 310 to 360 mg, respectively) [38].

Although we adjusted for a wide range of potential confounders, including social and environmental factors, to generate IPTWs, residual confounding may still be present, such as dietary and exercise factors. The social and environmental variables, including residential area median income and the EJI socio-environmental percentile ranking, were included to account for differences in health outcomes that may be influenced by healthier diets, lifestyles, and the activeness of health-seeking behaviors. Conversely, the study population was considered a high-risk and hard-to-recruit population in the real life, as reflected by the high number of adverse events occurring within the first six months.

However, the sensitivity analysis restricted to patients who survived at least six months showed consistent results and suggests that the results can also be generalizable to lower-risk patients with HF. Lastly, the generalizability of our findings to women with HF or community-based HF patients may be limited due to the predominantly elderly male veteran population included in the study. Despite this limitation, this study was still able to include 667 women. We chose veterans as the study population because the VHA databases contain comprehensive variables related to health history and records: such as comorbidities, laboratory results, pharmacy data, health care utilization, and vital signs: which may help address potential confounding and indication biases associated with both magnesium supplement use and the outcome.

Future studies need to be conducted in other populations.

5. Conclusions

In conclusion, our findings indicate that magnesium supplementation is linked to a lower risk of all-cause mortality or hospitalization in patients with HF. These promising outcomes merit further exploration and validation through RCTs.

Acknowledgments

The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health, the Department of Veterans Affairs, or the U.S. Government.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/nu17233687/s1, Figure S1: Plots of ASDs before and after IPT-weighting.

Supplement file at the original source: open the original paper. This file is not included in this text import.

Author Contributions

Y.C. and Y.S. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: A.R.Z., S.L., Q.Z.-T. and W.-C.W.; acquisition, analysis, or interpretation of data: Y.C., A.R.Z., Y.Y., Y.S., Q.Z.-T. and W.-C.W.; drafting of the manuscript: Y.C., S.T., Q.Z.-T. and W.-C.W.; critical review of the manuscript for important intellectual content: A.R.Z., Q.Z.-T. and W.-C.W.; statistical analysis: Y.C.; administrative, technical, or material support: S.T.; supervision: Q.Z.-T. and W.-C.W.

All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Washington DC VA Medical Center (IRBNET#1710776-1, 23 April 2021).

Informed Consent Statement

Patient consent was waived due to because the study involved secondary analysis of existing database.

Data Availability Statement

The datasets presented in this article are not readily available because of the sensitive nature of the data involved and the access to the dataset is restricted. Requests to access the datasets should be directed to the Washington DC VA Medical Center. Only qualified researchers trained in human subject confidentiality may request access by contacting via email at helen.sheriff@va.gov.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. Paolisso G. Scheen A. D’Onofrio F. Lefebvre P. Magnesium and glucose homeostasis Diabetologia 1990 33 511 514 10.1007/BF00404136 2253826
2. Douban S. Brodsky M.A. Whang D.D. Whang R. Significance of magnesium in congestive heart failure Am. Heart J. 1996 132 664 671 10.1016/S0002-8703(96)90253-7 8800040
3. DiNicolantonio J.J. Liu J. O’Keefe J.H. Magnesium for the prevention and treatment of cardiovascular disease Open Heart 2018 5 e000775 10.1136/openhrt-2018-000775 30018772 PMC6045762
4. Gottlieb S.S. Importance of magnesium in congestive heart failure Am. J. Cardiol. 1989 63 G39 G42 10.1016/0002-9149(89)90218-X 2650514
5. Nishihara T. Yamamoto E. Sueta D. Fujisue K. Usuku H. Oike F. Takae M. Arima Y. Araki S. Takashio S. Clinical significance of serum magnesium levels in patients with heart failure with preserved ejection fraction Medicine 2019 98 e17069 10.1097/MD.0000000000017069 31567942 PMC6756625
6. Taveira T.H. Ouellette D. Gulum A. Choudhary G. Eaton C.B. Liu S. Wu W.C. Relation of Magnesium Intake with Cardiac Function and Heart Failure Hospitalizations in Black Adults: The Jackson Heart Study Circ. Heart Fail. 2016 9 e002698 10.1161/CIRCHEARTFAILURE.115.002698 27056880 PMC4826717
7. Adamopoulos C. Pitt B. Sui X. Love T.E. Zannad F. Ahmed A. Low serum magnesium and cardiovascular mortality in chronic heart failure: A propensity-matched study Int. J. Cardiol. 2009 136 270 277 10.1016/j.ijcard.2008.05.006 18672302 PMC2721016
8. Levitan E.B. Shikany J.M. Ahmed A. Snetselaar L.G. Martin L.W. Curb J.D. Lewis C.E. Calcium, magnesium and potassium intake and mortality in women with heart failure: The Women’s Health Initiative Br. J. Nutr. 2013 110 179 185 10.1017/S0007114512004667 23199414 PMC5373650
9. Misialek J.R. Lopez F.L. Lutsey P.L. Huxley R.R. Peacock J.M. Chen L.Y. Soliman E.Z. Agarwal S.K. Alonso A. Serum and dietary magnesium and incidence of atrial fibrillation in whites and in African Americans: Atherosclerosis Risk in Communities (ARIC) study Circ. J. 2013 77 323 329 10.1253/circj.CJ-12-0886 23047297 PMC4228988
10. Sueta C.A. Patterson J.H. Adams K.F. Jr. Antiarrhythmic action of pharmacological administration of magnesium in heart failure: A critical review of new data Magnes. Res. 1995 8 389 401 8861138
11. The ASCEND Study Collaborative Group Bowman L. Mafham M. Wallendszus K. Stevens W. Buck G. Barton J. Murphy K. Aung T. Haynes R. Effects of n-3 Fatty Acid Supplements in Diabetes Mellitus N. Engl. J. Med. 2018 379 1540 1550 10.1056/NEJMoa1804989 30146932
12. Laville M. Segrestin B. Alligier M. Ruano-Rodriguez C. Serra-Majem L. Hiesmayr M. Schols A. La Vecchia C. Boirie Y. Rath A. Evidence-based practice within nutrition: What are the barriers for improving the evidence and how can they be dealt with? Trials 2017 18 425 10.1186/s13063-017-2160-8 28893297 PMC5594518
13. Manson J.E. Bassuk S.S. Lee I.M. Cook N.R. Albert M.A. Gordon D. Zaharris E. Macfadyen J.G. Danielson E. Lin J. The VITamin D and OmegA-3 TriaL (VITAL): Rationale and design of a large randomized controlled trial of vitamin D and marine omega-3 fatty acid supplements for the primary prevention of cancer and cardiovascular disease Contemp. Clin. Trials 2012 33 159 171 10.1016/j.cct.2011.09.009 21986389 PMC3253961
14. Danaei G. Rodriguez L.A. Cantero O.F. Logan R. Hernan M.A. Observational data for comparative effectiveness research: An emulation of randomised trials of statins and primary prevention of coronary heart disease Stat. Methods Med. Res. 2013 22 70 96 10.1177/0962280211403603 22016461 PMC3613145
15. Matthews A.A. Danaei G. Islam N. Kurth T. Target trial emulation: Applying principles of randomised trials to observational studies BMJ 2022 378 e071108 10.1136/bmj-2022-071108 36041749
16. Hernan M.A. Robins J.M. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available Am. J. Epidemiol. 2016 183 758 764 10.1093/aje/kwv254 26994063 PMC4832051
17. Hernan M.A. Wang W. Leaf D.E. Target Trial Emulation: A Framework for Causal Inference From Observational Data JAMA 2022 328 2446 2447 10.1001/jama.2022.21383 36508210
18. Cheng Y. Zullo A.R. Yin Y. Shao Y. Liu S. Zeng-Treitler Q. Wu W.C. Nonprescription Magnesium Supplement Use and Risk of Heart Failure in Patients with Diabetes: A Target Trial Emulation J. Am. Heart Assoc. 2025 14 e038870 10.1161/JAHA.124.038870 40135571 PMC12132832
19. ATSDR EJI Frequently Asked Questions Available online: https://www.atsdr.cdc.gov/place-health/php/eji/eji-frequently-asked-questions.html(accessed on 3 December 2024)
20. Elze M.C. Gregson J. Baber U. Williamson E. Sartori S. Mehran R. Nichols M. Stone G.W. Pocock S.J. Comparison of Propensity Score Methods and Covariate Adjustment: Evaluation in 4 Cardiovascular Studies J. Am. Coll. Cardiol. 2017 69 345 357 10.1016/j.jacc.2016.10.060 28104076
21. Ali M.S. Groenwold R.H. Klungel O.H. Best (but oft-forgotten) practices: Propensity score methods in clinical nutrition research Am. J. Clin. Nutr. 2016 104 247 258 10.3945/ajcn.115.125914 27413128
22. Negru A.G. Pastorcici A. Crisan S. Cismaru G. Popescu F.G. Luca C.T. The Role of Hypomagnesemia in Cardiac Arrhythmias: A Clinical Perspective Biomedicines 2022 10 2356 10.3390/biomedicines10102356 36289616 PMC9598104
23. Curran J. Ross-White A. Sibley S. Magnesium prophylaxis of new-onset atrial fibrillation: A systematic review and meta-analysis PLoS ONE 2023 18 e0292974 10.1371/journal.pone.0292974 37883337 PMC10602269
24. Shiga T. Wajima Z. Inoue T. Ogawa R. Magnesium prophylaxis for arrhythmias after cardiac surgery: A meta-analysis of randomized controlled trials Am. J. Med. 2004 117 325 333 10.1016/j.amjmed.2004.03.030 15336582
25. Na L. Chang J. Li X. Che X. Sun Y. Cui W. Zhang W. Xue X. Association Between Intravenous Magnesium Sulfate and All-Cause Mortality in Patients with Acute Heart Failure: A Propensity Score-Matched Cohort Study in MIMIC-IV Rev. Cardiovasc. Med. 2025 26 39206 10.31083/RCM39206 41089788 PMC12516736
26. Stepura O.B. Martynow A.I. Magnesium orotate in severe congestive heart failure (MACH) Int. J. Cardiol. 2009 134 145 147 10.1016/j.ijcard.2009.01.047 19367681
27. Guasch-Ferre M. Bullo M. Estruch R. Corella D. Martinez-Gonzalez M.A. Ros E. Covas M. Aros F. Gomez-Gracia E. Fiol M. Dietary magnesium intake is inversely associated with mortality in adults at high cardiovascular disease risk J. Nutr. 2014 144 55 60 10.3945/jn.113.183012 24259558
28. Yang Z. Zhang Y. Gao J. Yang Q. Qu H. Shi J. Association between dietary magnesium and 10-year risk of a first hard atherosclerotic cardiovascular disease event Am. J. Med. Sci. 2024 368 355 360 10.1016/j.amjms.2024.05.014 38801946
29. Song L. Ying J. Li M. Ying L. Zhao C. Propensity score matched cohort study on magnesium supplementation and mortality in critically ill patients with HFpEF Sci. Rep. 2025 15 1944 10.1038/s41598-025-85931-1 39809943 PMC11733161
30. Savarese G. Lund L.H. Global Public Health Burden of Heart Failure Card. Fail. Rev. 2017 3 7 11 10.15420/cfr.2016:25:2 28785469 PMC5494150
31. Hollifield J.W. Magnesium depletion, diuretics, and arrhythmias Am. J. Med. 1987 82 30 37 10.1016/0002-9343(87)90130-6 2436474
32. Barbagallo M. Veronese N. Dominguez L.J. Magnesium in Aging, Health and Diseases Nutrients 2021 13 463 10.3390/nu13020463 33573164 PMC7912123
33. Uwitonze A.M. Razzaque M.S. Role of Magnesium in Vitamin D Activation and Function J. Am. Osteopath. Assoc. 2018 118 181 189 10.7556/jaoa.2018.037 29480918
34. Martin B.J. Milligan K. Diuretic-associated hypomagnesemia in the elderly Arch. Intern. Med. 1987 147 1768 1771 10.1001/archinte.1987.00370100082014 3662705
35. Kuller L. Farrier N. Caggiula A. Borhani N. Dunkle S. Relationship of diuretic therapy and serum magnesium levels among participants in the Multiple Risk Factor Intervention Trial Am. J. Epidemiol. 1985 122 1045 1059 10.1093/oxfordjournals.aje.a114186 4061439
36. Guerrera M.P. Volpe S.L. Mao J.J. Therapeutic uses of magnesium Am. Fam. Physician 2009 80 157 162 19621856
37. African American Policy Forum Table 2 Common Magnesium Formulations and Dosages Available online: https://www.aafp.org/pubs/afp/issues/2009/0715/p157/jcr:content/root/aafp-article-primary-content-container/aafp_article_main_par/aafp_tables_content1.print.html(accessed on 13 November 2025)
38. National Institutes of Health Magnesium: Fact Sheet for Health Professional Available online: https://ods.od.nih.gov/factsheets/Magnesium-HealthProfessional/(accessed on 17 November 2025)
Figure 1

Cohort Assembly.

Figure image at the original source: open the original paper. This file is not included in this text import.
Figure 2

Graphical depiction of study design.

Figure image at the original source: open the original paper. This file is not included in this text import.
Figure 3

Propensity score distribution.

Figure image at the original source: open the original paper. This file is not included in this text import.
Figure 4

(a) KM Curves of all-cause death or hospitalization comparing inverse probability of treatment-weighted magnesium supplement users vs. non-users. (b) KM curves of all-cause death or hospitalization comparing inverse probability of treatment-weighted magnesium supplement users vs. non-users (sensitivity analysis in the subgroup who did not die within 6 months after the index date). (c) KM curves of all-cause death or hospitalization comparing inverse probability of treatment-weighted magnesium supplement users vs. non-users (sensitivity analysis in the subgroup who had at least one additional clinical note mentioning magnesium supplement use and their paired controls).

Figure image at the original source: open the original paper. This file is not included in this text import.
Figure 5

(a) Subgroup analysis by variables with significant interactions. (b) Subgroup analysis by serum magnesium.

Figure image at the original source: open the original paper. This file is not included in this text import.
Table 1

Elements of the target trial and emulation for magnesium supplement treatment strategies.

Target TrialEmulation in Observational Data
AimEvaluate the association between magnesium supplement use and all-cause hospitalization or death in the U.S. military veterans with HF
Eligibility(1) Newly diagnosed with HF no more than 1 year
(2) Never used magnesium supplement or replacement treatment before
(3) No hospitalization in the past 30 days
(4) No end-stage renal disease (ESRD)
(1) Had the initial HF diagnosis within 1 year before
(2) Had at least two outpatient clinical notes within the previous 12 months (to confirm magnesium supplement use status)
(3) Had no record of prescribed or supplemental magnesium use prior to this date
(4) Had no hospitalization within 30 days before or at the eligible date
(5) Had no ESRD any time before or at the eligible date
Treatment Strategies(1) Magnesium supplement intervention: Initiation of magnesium supplement
(2) Magnesium supplement non-intervention: No initiation of magnesium supplement
(1) Magnesium supplement users: new documentation of the use of magnesium supplements in an outpatient clinic encounter
(2) Magnesium supplement non-users: No documentation of magnesium supplement use in an outpatient clinic encounter
AssignmentRandomly assigned 1:1 to each treatment strategy in a parallel designVeterans were assigned to treatment arms based on EMR documentation. All magnesium supplement users were included. For the non-user group, randomization was emulated by randomly selecting a group of non-user patients with the same eligible date as a user. Inverse probability weighting was used to balance observed confounders between treatment groups.
Follow-upFrom date of randomization up until end of 5 years, outcome, or censoring due to events such as violation of the assigned treatment strategyFollowed from the time when a patient became eligible for the trial (index date) until the outcome of interest occurred, or the censoring date due to magnesium supplement use status change (i.e., stop use in users or start use in non-users), prescribed magnesium replacement was started, or the end of 5th year from the index date, whichever occurred first.
OutcomesAll-cause mortality or all-cause hospitalization (time to events)Same
Causal ContrastThe difference in risk of all-cause mortality or all-cause hospitalization had everyone followed strategy (1) vs. strategy (2)Same
Statistical AnalysisCox’s regressionPer-protocol analysis using Cox’s regression weighted by inverse probability of treatment
Table 2

Baseline characteristics of study participants before IPT weighting.

Baseline CharacteristicsOverall
(N = 19,800)
Magnesium
Users
(N = 9900)
Magnesium
Non-Users
(N = 9900)
ASD Before IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.173.310.772.011.412
Male19,13396.6%954496.4%958996.9%3
Race
White15,22076.9%787279.5%734874.2%13
Black242612.3%8698.8%155715.7%21
Others215410.9%115911.7%99510.1%5
Ethnicity
Non-Hispanics17,96590.7%899190.8%897490.6%1
Hispanics6403.2%2442.5%3964.0%8
Unknown11956.0%6656.7%5305.4%5
Married11,58358.5%626263.3%532153.7%20
Median Income (lowest to highest) of Zip Code
1st Quartile412820.8%191419.3%221422.4%8
2nd Quartile481124.3%247825.0%233323.6%3
3rd Quartile487124.6%247225.0%239924.2%2
4th Quartile553027.9%286528.9%266526.9%4
Unknown4602.3%1711.7%2892.9%8
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code48.022.646.322.249.622.915
Duration between HF Diagnosis and Index Date, Days129.6117.7105.6118.1153.7112.242
Comorbid Conditions (31 Variables)
Alcohol Disorder281814.2%118812.0%163016.5%13
Smoking608330.7%248325.1%360036.4%25
Hypertension17,21887.0%834084.2%887889.7%16
Hyperlipidemia16,07981.2%784979.3%823083.1%10
Myocardial Infarction356618.0%153615.5%203020.5%13
Atherosclerosis477024.1%220822.3%256225.9%8
Angina339417.1%134613.6%204820.7%19
Atrial Fibrillation670533.9%364636.8%305930.9%12
Cardiac Valve Disease343017.3%157415.9%185618.7%7
Ischemic Stroke246012.4%108310.9%137713.9%9
Hemorrhage Stroke1720.9%620.6%1101.1%5
Transient Ischemic Attack10615.4%4294.3%6326.4%9
Chronic Obstructive Pulmonary Disease712536.0%332433.6%380138.4%10
Asthma18209.2%8568.6%9649.7%4
Chronic Kidney Disease561228.3%267227.0%294029.7%6
Anemia654133.0%300830.4%353335.7%11
Arthritis934747.2%439844.4%494950.0%11
Cancer675334.1%312931.6%362436.6%11
Neurological Disorders755638.2%357236.1%398440.2%8
Parkinson4172.1%1811.8%2362.4%4
Hypothyroidism288714.6%151115.3%137613.9%4
Osteoporosis8664.4%4204.2%4464.5%1
Osteomyelitis5002.5%1831.8%3173.2%9
Urinary Tract Infection12386.3%4874.9%7517.6%11
Liver Disease18759.5%8288.4%104710.6%8
Pneumonia248112.5%9179.3%156415.8%20
Respiratory Failure10875.5%3733.8%7147.2%15
Sepsis6623.3%2102.1%4524.6%14
Opioid Use Disorder4202.1%1501.5%2702.7%8
Depression688334.8%299830.3%388539.2%19
Fracture222311.2%9159.2%130813.2%13
Concurrent Medications (22 Variables)
Insulin504125.5%221422.4%282728.6%14
Metformin394919.9%185518.7%209421.2%6
Glucagon-Like Peptide-1780.4%340.3%440.4%2
Sodium: Glucose Cotrasporter-2 Inhibitors1951.0%1001.0%951.0%0
Other Diabetes Medication322116.3%147014.8%175117.7%8
Vitamin D Prescription361818.3%143314.5%218522.1%20
Multivitamin (Rx)9724.9%3903.9%5825.9%9
Angiotensin-Converting Enzyme Inhibitor802140.5%322932.6%479248.4%33
Angiotensin Receptor Blocker313715.8%144614.6%169117.1%7
Calcium Channel Blocker544227.5%221122.3%323132.6%23
Loop Diuretics876544.3%375037.9%501550.7%26
Thiazide Diuretics326216.5%137213.9%189019.1%14
Selective Beta Blocker799640.4%328733.2%470947.6%30
Non-Selective Beta Blocker442522.3%191319.3%251225.4%15
Other Antihypertensive593130.0%259826.2%333333.7%16
Statins11,12556.2%467947.3%644665.1%36
Other Lipid-Lowering Medication12586.4%5735.8%6856.9%5
Proton-Pump Inhibitors666233.6%294029.7%372237.6%17
Aspirin589429.8%206720.9%382738.7%40
Digoxin3341.7%1541.6%1801.8%2
Glucocorticoids244312.3%9729.8%147114.9%16
Platelet Inhibitor305815.4%128012.9%177818.0%14
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,86075.1%836484.5%649665.6%45
1303515.3%9769.9%205920.8%31
2+19059.6%5605.7%134513.6%27
Number of visits in past year31.529.526.828.036.230.232
Addiction Medicine4832.4%1861.9%2973.0%7
Cardiology855143.2%317032.0%538154.4%46
Emergency Care580029.3%202520.5%377538.1%39
Endocrinology9905.0%4254.3%5655.7%6
Hospice Medicine1380.7%540.5%840.8%4
Intensive Care Unit12556.3%3793.8%8768.8%21
Internal Medicine618431.2%262826.5%355635.9%20
Oncology9804.9%4044.1%5765.8%8
Palliative Care3381.7%1151.2%2232.3%8
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.930.26.929.86.96
Systolic Blood Pressure, mmHg129.519.9127.919.8131.019.916
Diastolic Blood Pressure, mmHg72.111.771.311.572.911.914
Serum Magnesium, mg/dL
<1.7 mg/dL10035.1%7577.6%2462.5%23
1.7 to 2.5 mg/dL645432.6%260826.3%384638.8%27
>2.5 mg/dL1891.0%900.9%991.0%1
Unknown12,15461.4%644565.1%570957.7%15
Serum 25-hydroxy vitamin D < 20 ng/dL9274.7%3823.9%5455.5%8
Low-Density Lipoprotein, mg/dL85.934.285.134.586.533.94
Triglycerides, mg/dL146.3111.5147.1110.3145.6112.61
High-Density Lipoprotein, mg/dL43.214.243.314.443.014.02
Cholesterol, mg/dL155.342.6154.542.9156.042.34
Sodium, mmol/L138.93.2138.73.3139.03.19
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.59.20.50
Hemoglobin A1C, %6.81.56.91.56.81.57
Ejection Fraction
≤40%517126.1%260026.9%251125.4%3
>40%709135.8%292829.6%416342.1%26
Unknown753838.1%431243.6%322632.6%23

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Table 3

Baseline characteristics of study participants after IPT weighting.

Baseline CharacteristicsIPT-Weighted
Overall
(N = 19,925)
IPT-Weighted Magnesium
Users
(N = 10,096)
IPT-Weighted Magnesium
Non-Users
(N = 9829)
ASD After IPTW (%)
Mean/NSD/%Mean/NSD/%Mean/NSD/%
Demographics (8 Variables)
Age, Years72.611.172.510.872.611.41
Male19,24096.6%974896.6%949296.6%0
Race
White15,26076.6%771176.4%755076.8%1
Black251812.6%130012.9%121912.4%2
Others214710.8%108610.8%106110.8%0
Ethnicity
Non-Hispanics18,10590.9%917190.8%893590.9%0
Hispanics6543.3%3363.3%3183.2%1
Unknown11665.9%5895.8%5775.9%0
Married11,56558.0%582857.7%573758.4%1
Median Income (lowest to highest) of Zip Code
1st Quartile419721.1%213921.2%205820.9%1
2nd Quartile479724.1%242424.0%237324.1%0
3rd Quartile482524.2%243824.1%238724.3%0
4th Quartile562428.2%284228.1%278228.3%0
Unknown4822.4%2532.5%2292.3%1
Environmental Justice Index Socio-Environmental Percentile Ranking of Zip code47.922.748.022.947.822.61
Duration between Diabetes Diagnosis and Index Date, Days135.0118.4136.0125.8134.0110.52
Comorbid Conditions (31 Variables)
Alcohol Disorder288814.5%147114.6%141814.4%1
Smoking626731.5%321031.8%305831.1%2
Hypertension17,38287.2%880487.2%857887.3%0
Myocardial Infarction367018.4%186618.5%180418.4%0
Atherosclerosis488424.5%247824.5%240524.5%0
Angina354817.8%181718.0%173117.6%1
Atrial Fibrillation675633.9%342433.9%333133.9%0
Cardiac Valve Disease356517.9%182418.1%174117.7%1
Ischemic Stroke253012.7%128412.7%124612.7%0
Hemorrhage Stroke1850.9%971.0%880.9%1
Transient Ischemic Attack10925.5%5525.5%5395.5%0
Chronic Obstructive Pulmonary Disease726336.5%370336.7%356036.2%1
Asthma19229.6%9899.8%9349.5%1
Chronic Kidney Disease576929.0%291928.9%285029.0%0
Anemia676734.0%345234.2%331433.7%1
Arthritis953247.8%482047.7%471247.9%0
Cancer690534.7%348834.6%341734.8%0
Neurological Disorders778539.1%395339.1%383239.0%0
Parkinson4212.1%2182.2%2042.1%1
Hypothyroidism294414.8%149814.8%144614.7%0
Osteoporosis8884.5%4594.6%4284.4%1
Osteomyelitis5212.6%2682.7%2542.6%1
Hyperlipidemia16,30681.8%824481.7%806282.0%1
Urinary Tract Infection13446.7%7117.0%6336.4%2
Liver Disease19719.9%100610.0%9659.8%1
Pneumonia259213.0%134113.3%125112.7%2
Respiratory Failure11755.9%6226.2%5535.6%3
Sepsis6993.5%3623.6%3383.4%1
Opioid Use Disorder4372.2%2292.3%2082.1%1
Depression705035.4%359535.6%345435.1%1
Fracture237811.9%123512.2%114411.6%2
Concurrent Medications (22 Variables)
Insulin521926.2%265626.3%256426.1%0
Metformin409920.6%204920.3%205120.9%1
Glucagon-Like Peptide-1840.4%420.4%420.4%0
Sodium: Glucose Cotrasporter-2 Inhibitors2081.0%1061.0%1031.0%0
Other Diabetes Medication331516.6%166316.5%165216.8%1
Vitamin D Prescription378919.0%194119.2%184818.8%1
Multivitamin (Rx)10445.2%5365.3%5075.2%0
Angiotensin-Converting Enzyme Inhibitor828041.6%422741.9%405341.2%1
Angiotensin Receptor Blocker325316.3%163316.2%162016.5%1
Calcium Channel Blocker566328.4%289228.6%277128.2%1
Loop Diuretics917646.1%466646.2%451045.9%1
Thiazide Diuretics333916.8%168316.7%165716.9%1
Selective Beta Blocker827841.5%420241.6%407741.5%0
Non-Selective Beta Blocker463623.3%236223.4%227423.1%1
Other Antihypertensive619531.1%314331.1%305231.1%0
Statins11,39157.2%578857.3%560357.0%1
Other Lipid-Lowering Medication12766.4%6306.2%6466.6%2
Proton-Pump Inhibitors690334.6%348734.5%341534.7%0
Aspirin618031.0%318131.5%299930.5%2
Digoxin3321.7%1641.6%1681.7%1
Glucocorticoids258413.0%132013.1%126412.9%1
Platelet Inhibitor317115.9%160515.9%156715.9%0
Health Care Utilization (11 Variables)
Hospitalization(s) in past year
014,67273.6%737273.0%730074.3%3
1321716.1%165716.4%156115.9%1
2+203610.2%106810.6%9689.9%2
Number of visits in past year32.830.333.131.632.629.02
Addiction Medicine4912.5%2512.5%2402.4%1
Cardiology887444.5%455045.1%432444.0%2
Emergency Care613430.8%314631.2%298830.4%2
Endocrinology10585.3%5295.2%5295.4%1
Hospice Medicine1640.8%920.9%710.7%2
Intensive Care Unit13226.6%6846.8%6386.5%1
Internal Medicine641632.2%329332.6%312331.8%2
Oncology10525.3%5485.4%5045.1%1
Palliative Care3892.0%2132.1%1751.8%2
Vital Signs and Labs (14 Variables)
Body Mass Index, kg/m230.06.929.97.030.06.81
Systolic Blood Pressure, mmHg129.720.2129.820.9129.619.61
Diastolic Blood Pressure, mmHg72.312.072.412.272.211.72
Serum Magnesium, mg/dL
<1.7 mg/dL10705.4%5115.1%5595.7%3
1.7 to 2.5 mg/dL671133.7%344834.2%326333.2%2
>2.5 mg/dL1901.0%950.9%951.0%1
Unknown11,95460.0%604259.8%591260.2%1
Serum 25-hydroxy vitamin D < 20 ng/dL9564.8%4924.9%4644.7%1
Low-Density Lipoprotein, mg/dL85.834.585.836.285.832.90
Triglycerides, mg/dL145.5111.2145.5109.7145.6112.50
High-Density Lipoprotein, mg/dL43.214.243.314.543.213.91
Cholesterol, mg/dL155.143.2155.145.4155.141.00
Sodium, mmol/L138.93.3138.93.4138.93.10
Potassium, mmol/L4.30.54.30.54.30.50
Creatinine, mg/dL1.30.61.30.61.30.60
Calcium, mg/dL9.20.59.20.69.20.50
Hemoglobin A1C, %6.81.56.81.66.81.50
Ejection Fraction
≤40%530726.6%268326.6%262426.7%0
>40%726936.5%370836.7%356136.2%1
Unknown734936.9%370436.7%364437.1%1

Abbreviations: SD: standard deviation; ASD: absolute standardized difference; IPTW: inverse probability of treatment weighting.

Table 4

All-cause death or hospitalization over 5 years of follow-up in the IPT-weighted cohort.

# Events# Patients at RiskEvent/Patients at Risk (%)Mean (STD) Follow-Up YearsMedian (IQR) Follow-Up YearsTotal Person YearsIncidence Rate (per 1000 Person-Years)
Magnesium non-users3013982930.7%0.85 (0.96)0.70
(0.10 to 1.00)
8365.5360.2
Magnesium users24521009624.3%0.82 (0.95)0.60
(0.10 to 1.00)
8234.1297.8

Abbreviations: #, number of; IPT, inverse probability of treatment; STD, standard deviation; IQR: interquartile range.

Publication and source details

Original title: Association Between Over-the-Counter Magnesium Supplement Use and Health Outcomes in Veterans with Newly Diagnosed Heart Failure

Yan Cheng; Andrew R. Zullo; Ying Yin; Yijun Shao; Senait Tekle; Simin Liu; Qing Zeng-Treitler; Wen-Chih Wu

Nutrients · 2025

* Correspondence: wen-chih.wu@va.gov; Tel.: +401-273-7100© 2025 by the authors.2025https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).National Health Lung and Blood Institute of the NIH Office of the Director5R01HL156518Office of Research and Development, Health Services Research and DevelopmentWashington DC VA Medical CenterThe research reported in this manuscript was supported by the National Health Lung and Blood Institute of the NIH Office of the Director under award number 5R01HL156518 and with resources from the Office of Research and Development, Health Services Research and Development, and the use of facilities at the Washington DC VA Medical Center.

Repository JATS snapshot; version as supplied

Source retrieved: 2026-11-11T01:33:59.757066+00:00

DOI: 10.3390/nu17233687 · PMC12694171