Risk of hypermagnesemia in patients with eating disorders taking magnesium oxide preparations: a retrospective study

Abstract

Purpose

This study examined the incidence of hypermagnesemia in patients with eating disorders, its associations with renal function, body mass index, and magnesium oxide use, and the correlation between different methods of estimating renal function and serum magnesium levels.

Methods

This retrospective cohort study was conducted in female patients with eating disorders treated at Nagoya University Hospital between January 2018 and December 2022. Patients diagnosed as eating disorders, based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision, were included. Serum magnesium levels, estimated glomerular filtration rate, body mass index, and magnesium oxide prescriptions were collected. Linear mixed-effects models were used to analyze factors affecting serum magnesium levels. Renal function screening methods were compared in underweight patients.

Results

Among 194 patients, 42 (21.6%) developed hypermagnesemia (≥ 2.5 mg/dL; maximum 5.3 mg/dL). Younger age, lower body mass index, and reduced estimated glomerular filtration rate were linked to higher magnesium levels, whereas magnesium oxide use showed no clear association. Further analysis showed that the alternative estimated glomerular filtration rate method, adjusted for body size, negatively correlated with elevated serum magnesium levels in underweight patients.

Conclusion

The incidence of hypermagnesemia in patients with eating disorders receiving magnesium oxide was comparable to previous studies. Risk factors include low body mass index, impaired renal function, and younger age. Although monitoring is warranted, severe complications were not observed, suggesting magnesium oxide need not be avoided. The use of body size-adjusted estimated glomerular filtration rate may improve the screening for hypermagnesemia in underweight patients.

Level of Evidence: Level III, well-designed cohort or case–control analytic studies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40519-025-01796-3.

Introduction

Eating disorders are serious psychiatric conditions that predominantly affect young women. Among them, anorexia nervosa (AN) is characterized by a distorted body image and an intense fear of weight gain, leading to maladaptive weight control behaviors such as extreme dietary restriction, self-induced vomiting, and excessive physical activity. These behaviors cause physical and biochemical abnormalities due to chronic starvation [1, 2]. Similar to AN, patients with avoidant/restrictive food intake disorder (ARFID) and Other Specified Feeding or Eating Disorders (OSFED) may suffer significant physical damage when experiencing dietary restriction and nutritional deficiency [3, 4].

Renal dysfunction is a common comorbidity among individuals with AN. In a study of 197 patients with AN admitted to psychiatric units, over 80% exhibited an estimated glomerular filtration rate (eGFR) below 90 mL/min, whereas only 18% of hospitalized patients with AN had an eGFR above this threshold [5]. Chronic dehydration resulting from inadequate intake impairs renal perfusion and reduces eGFR [6]. Additionally, purging behaviors (vomiting, laxative or diuretic use) increase the risk of chronic hypokalemia-associated nephropathy, a frequent clinical concern [6].

Furthermore, eating disorders accompanied by malnutrition often lead to reduced muscle mass. As muscle mass influence serum creatinine (Cr)-based eGFR calculations, renal function may be overestimated in these patients. Mild renal impairment may be overlooked, compromising medication safety.

Constipation is also commonly reported among patients with eating disorders, and chronic laxative use is prevalent due to a persistent drive to maintain low body weight [7, 8]. In Japan, saline laxatives, such as magnesium oxide (MgO), are widely prescribed as treatment for constipation [7]. Oral MgO induces laxative effects via intestinal osmosis. Despite low bioavailability, excessive use may cause hypermagnesemia with symptoms like confusion, drowsiness, and headache [9].

Several studies have reported the incidence of oral MgO associated-hypermagnesemia in patients with impaired renal function [10–12]. In response, the Ministry of Health, Labor and Welfare of Japan has advised caution when prescribing oral MgO to older adults with renal dysfunction [13].

This hypothesis-testing study evaluated malnourished patients with eating disorders across diagnostic categories to examine two hypotheses. (1) In patients with malnourished eating disorders, the under-recognition of renal impairment contributes to an increased risk of MgO-related adverse effects, particularly hypermagnesemia. (2) Reduced muscle mass in this population leads to overestimation of renal function by eGFR, resulting in the underdiagnosis of mild renal impairment.

We retrospectively reviewed charts of patients with eating disorders to assess hypermagnesemia and identify risk factors—particularly renal impairment—examining serum magnesium elevations, MgO exposure, and related symptoms. We also evaluated eGFR–MgO–biochemical relationships and the limits of renal assessment to inform safer prescribing in this high-risk group.

Methods

Study design

This retrospective cohort study was conducted at the Departments of Psychiatry and Child and Adolescent Psychiatry of Nagoya University Hospital between January 1, 2018, and December 31, 2022.

Inclusion criteria were: (1) female patients who received outpatient or inpatient treatment during the study period; and (2) diagnosis of an eating disorder according to DSM-5-TR criteria, including AN restricting type (ANR), AN binge-eating/purging type (ANBP), ARFID, and OSFED.

Exclusion criteria were: (1) patients undergoing dialysis; and (2) patients with incomplete laboratory data.

Pre-2022 diagnoses were re-evaluated by a psychiatrist (MU). Follow-up was censored at relocation, loss to follow-up, or the end of the study period. No deaths occurred during the observation period. Medication use and clinical indicators were confirmed through medical records and patient self-reports.

Follow-up was defined as the interval between the earliest and most recent laboratory data available for each patient and was used as a proxy for nutritional rehabilitation and clinical monitoring. In exploratory analyses, shorter follow-up was interpreted as representing early recovery, when renal function may be temporarily compromised due to dehydration. Impaired renal function was defined as an eGFR of < 60 mL/min/1.73 m², in accordance with the Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Underweight was defined as a body mass index (BMI) of < 15 kg/m², corresponding to the “extreme severity” category of the DSM-5-TR.

As part of an exploratory analysis, the correlations between serum Mg concentration and various body size-adjusted methods of estimating renal function were examined [14–18]. In this study, the participants were drawn from a wide range of socioeconomic backgrounds. Japan’s universal health insurance ensures equitable healthcare access.

Missing data were handled using listwise deletion, whereby only patients with complete data for each variable of interest were included in the final analyses. No data imputation was performed.

During hospitalization, nutritional rehabilitation follows a standardized refeeding protocol. In our hospital, the refeeding protocol for inpatients with severe eating disorders starts with < 1000 kcal/day and increases by approximately 300 kcal per week, aiming for a weight gain of 1–2 kg per week. Some patients required enteral nutrition via a nasogastric tube for medical indications, such as the prevention of intermittent hypoglycemia and the management of gastrointestinal symptoms secondary to superior mesenteric artery syndrome, however, no patients received total parenteral nutrition. Electrolytes and B-vitamins were supplemented as needed. At the beginning of hospitalization, patients were co-managed with the general medicine department. To prevent refeeding syndrome, magnesium, phosphate, and other electrolytes were regularly monitored during the initial refeeding phase. No cases of refeeding syndrome occurred in this cohort.

Survey items

Data on blood biochemical parameters (total protein, blood urea nitrogen [BUN], Cr, albumin [Alb], sodium, potassium, chlorine, Mg, red blood cell, hemoglobin, and hematocrit), eGFR, age, follow-up period from the initial laboratory assessment (in days), eating disorder diagnosis, and BMI were collected. At each data collection time point, information on eating disorder diagnosis, BMI, oral MgO prescriptions, and duration of oral MgO use were recorded. Data on the dosage of MgO were extracted from electronic medical records, including both prescription information and patient-reported medication use on admission.

Additionally, clinical symptoms potentially associated with hypermagnesemia—such as nausea, numbness, dizziness, and muscle weakness—were retrospectively collected from medical records.

Statistical analysis

All statistical analyses were performed using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria) and RStudio (R Core Team, 2020). A P value of < 0.05 was considered significant.

Univariate analyses were carried out to determine the frequency of each diagnosis and to calculate the median and range of key variables. A linear mixed-effects model was used to examine the factors associated with elevated serum Mg levels. Independent variables included age, BMI, eGFR, follow-up period, and oral MgO prescription. MgO dosage was categorized into three groups: 0 mg, > 0–750 mg, and > 750 mg. MgO was input into the model as an explanatory variable, with 0 mg serving as the reference group. The lme4 package was used to identify the optimal model based on the Akaike information criterion (AIC) [19, 20]. Three models (Null, random-intercept [RI], random-intercept-and-slope [RIC]) were compared by AIC; the lowest AIC defined the optimal model (Table 1).

Table 1.

Linear mixed effects model

Item	Estimate	95% confidence interval	P value
Age (years)	− 0.0044	(− 0.0067 to − 0.0020)	0.00050*
BMI (kg/m2)	− 0.014	(− 0.021 to − 0.0063)	0.00048*
eGFR (mL/min/1.73m2)	− 0.0038	(− 0.0047 to − 0.0027)	< 0.001*
Follow-up days (days)	− 0.000060	(− 0.00012 to 0.000034)	0.096
MgO prescription (mg/day)
0	Reference
0–750	0.013	(− 0.045 to 0.071)	0.67
> 750	0.046	(− 0.044 to 0.14)	0.31

BMI body mass index, eGFR estimated glomerular filtration rate, MgO magnesium oxide

*Pr < 0.05

Analyses were initially performed on the full cohort and subsequently on a subgroup of patients with a BMI of < 15 kg/m², classified as having extreme severity according to the DSM-5-TR criteria.

Variables significantly associated with serum Mg levels were identified and plotted. Patients were stratified by an arbitrary serum Mg threshold, and mean levels between the groups were compared using Student’s t-test.

The primary outcome was the development of hypermagnesemia, defined as a serum Mg level of ≥ 2.5 mg/dL, according to the hospital’s reference range of 1.8–2.4 mg/dL. Data on physical symptoms observed in patients meeting this criterion were extracted and analyzed descriptively. The serum Mg threshold associated with the onset of symptoms was subsequently estimated.

Subgroup analysis

For the subgroup analysis, data were extracted and analyzed separately for participants with a BMI of < 15 kg/m² who were taking oral MgO at the time of data collection.

Evaluation of renal function in patients with eating disorders

Pearson’s correlation analysis was initially conducted to assess the impact of BMI on standard eGFR. In Japan, renal function in patients with nutritional disorders is assessed according to the KDIGO guidelines [15, 21]. Based on these criteria, an eGFR of ≥ 90 mL/min/1.73 m² was considered normal or elevated (G1), whereas an eGFR of < 60 mL/min/1.73 m² was classified as mildly decreased renal function (G3 or lower). However, in patients with eating disorders, reduced muscle mass may lead to the overestimation of renal function when using standard eGFR. To address this limitation, alternative estimation methods were applied: (1) the five-variable eGFR (Cr-independent); (2) body surface area (BSA)-adjusted eGFR; and (3) CCr estimated using the Cockcroft-Gault (CG) formula [15, 18, 22]. These methods were calculated to provide a more accurate assessment of renal function in this population.

The correlation between serum Mg levels and each method of estimating renal function was examined using scatter plots. Furthermore, data were divided into four quadrants based on a serum Mg threshold of 2.5 mg/dL and the respective renal function cutoff values for each estimation method. Renal function thresholds were defined as follows: 60 mL/min/1.73 m² for standard eGFR, 60 mL/min for five-variable eGFR, 60 mL/min for BSA-adjusted eGFR, and 60 mL/min for CCr estimated using the CG formula.

Results

Patient characteristics

A total of 209 participants met inclusion criteria (5817 observations). After excluding one dialysis patient and incomplete records, 194 patients (3828 observations) remained (Fig. 1). Of these, 70 (36.1%) received MgO (Table 2). Hypermagnesemia occurred in 42/194 (21.6%); 20/42 (47.6%) were MgO users. Maximum magnesium was 5.3 mg/dL.

Fig. 1.

Flowchart illustrating the participant enrollment process. Mg, Serum Mg level, eGFR estimated glomerular filtration rate

Table 2.

Demographics of patients

Item	All (N = 194)			Age under 20 years (N = 85)			Age 20 years or older (N = 109)
	Average	Median	[Range]	Average	Median	[Range]	Average	Median	[Range]
Age (years)	25.4	21	[11–65]	15.3	16	[11–19]	33.3	30	[20–65]
Body Weight (kg)	34.5	33.9	[18.9–58.0]	33.7	33.5	[18.9–57.8]	35	34.1	[22.5–58.0]
BMI (kg/m2)	14.3	14	[9.5–24.0]	14.3	14	[10.6–21.6]	14.4	14.1	[9.5–24.0]
Creatinine (mg/dL)	0.7	0.67	[0.31–2.92]	0.7	0.69	[0.39–1.16]	0.69	0.65	[0.31–2.92]
eGFR (mL/min/1.73m2)	94.6	92.1	[17.4–201.8]	102.3	98.5	[58.4–201.8]	88.6	87.9	[17.4–198.4]
Magnesium (mg/dL)	2.1	2.1	[1.5–5.3]	2.2	2.2	[1.8–2.9]	2.1	2.1	[1.5–5.3]
Follow-up days (days)	446.6	245	[1–1817]	297.6	81	[1–1767]	499.2	316	[1–1817]
MgO prescription (mg/day)	109.9	0	[0–2250]	36.2	0	[0–1000]	167.3	0	[0–2250]
subtype	N (%)
ANR	107 (55.2)			59 (69.4)			48 (44.0)
ANBP	49 (25.3)			13 (15.3)			36 (33.0)
ARFID	25 (12.9)			9 (10.6)			16 (14.7)
OSFED	13 (6.7)			4 (4.7)			9 (8.3)

BMI body mass index, eGFR estimated glomerular filtration rate, MgO magnesium oxide, ANR anorexia nervosa restricting type, ANBP anorexia nervosa binge-eating/purging type, ARFID avoidant/restrictive food intake disorder, OSFED Other Specified Feeding or Eating Disorders

Age, body weight, BMI, serum creatinine, eGFR, and serum magnesium levels were obtained from the baseline observation data

Of the 194 patients included, 170 (87.6%) were hospitalized at least once during the study period, whereas 24 (12.4%) were treated only as outpatients. Among hospitalized patients, the median follow-up was 139 days (mean 423), while among outpatients it was 0 days (mean 247), as most visited only once. Patients with CCr < 60 had significantly more hospitalizations than those with CCr ≥ 60 (20.7 vs. 8.2, Student’s t-test, p < 0.01). The cutoff of 60 mL/min was chosen in accordance with the KDIGO GFR categories.

Factors associated with elevated serum Mg levels

In the linear mixed-effects model analysis, the RIC model had the lowest AIC (− 176.70), compared with the null (677.86) and RI (349.62). Among the variables included—age, BMI, eGFR, follow-up period, and MgO prescription (included as an explanatory variable)—younger age, lower BMI, and impaired renal function were significantly associated with elevated serum Mg levels (Table 1).

In the subgroup analysis of patients with extreme underweight (BMI < 15 kg/m²), oral MgO prescription was not significantly associated with an increased risk of adverse effects (Table 3).

Table 3.

Demographic data of the study patients (N = 153) (BMI < 15 kg/m²)

Explanatory Variable	Median	[Range]	Average	Regression Coefficient (Estimate)	95% confidence interval	P value
Age (years)	28	[11–68]	30.3	− 0.0045*	(− 0.070 to − 0.020)	0.002*
BMI (kg/m2)	13.3	[8.5–14.9]	13.01	− 0.016*	(− 0.028 to − 0.0043)	0.011*
eGFR (mL/min/1.73m2)	90.1	[9.1–309.9]	94.3	− 0.0038*	(− 0.050 to − 0.027)	< 0.001*
Follow-up days (days)	227.5	[1–1817]	447.6	− 0.000081*	(− 0.00020 to 0.000035)	0.19
MgO intake (mg/day)	0	[0–2250]	184.3
0				Reference
0–750				0.011*	(− 0.058 to 0.080)	0.76
> 750				0.063	(− 0.012 to 0.14)	0.14

BMI body mass index, eGFR estimated glomerular filtration rate, MgO magnesium oxide

In a second analysis, the AIC values were 809.14 (null model), 586.14 (RI model), and − 34.00 (RIC model). The RIC model was again selected as optimal based on the lowest AIC value. In this model, age, BMI, eGFR, and follow-up period—but not MgO prescription—were associated with elevated serum Mg levels (Table 3).

No significant effects by subtypes(ANR, ANBP and ARFID) were observed in primary and secondary analyses, all p-values were > 0.05. OSFED was not analyzed because the number of cases was not enough.

Narrative complaints with elevation of serum Mg levels

Given their strong associations, we focused on eGFR and BMI. Reduced eGFR (< 60 mL/min/1.73 m²) and low BMI (< 15 kg/m²) were linked to higher magnesium (both p < 0.001; Figs. 2, 3). These thresholds correspond to the KDIGO classification for renal function and DSM-5-TR severity criteria, respectively.

Fig. 2.

a Scatter plot showing the relationship between eGFR and serum Mg levels. b Scatter plot showing the relationship between BMI and serum Mg levels. BMI body mass index, eGFR estimated glomerular filtration rate

Fig. 3.

Boxplot showing the serum Mg levels across the eGFR and BMI groups. eGFR estimated glomerular filtration rate, BMI body mass index

Among 42 patients with hypermagnesemia, 10 (23.8%) had a single symptom and 5 (11.9%) had multiple symptoms; lightheadedness, weakness, vomiting, and numbness were most frequent, with symptom counts rising at higher magnesium levels (Supplementary Table 1 and Fig. 1).

Alternative methods to assess renal function in patients with eating disorders

In our analysis, a weak negative correlation was observed between eGFR and BMI (r = − 0.095, p < 0.001 [p = 3.3 × 10⁻⁹]; 95% confidence interval: − 0.127 to − 0.064) (Fig. 4). Based on the standard eGFR calculation, 50.1% (1916/3828) of patients had an eGFR of ≥ 90 mL/min/1.73 m², whereas 17.5% (670/3828) had an eGFR of < 60 mL/min/1.73 m² (Supplementary Fig. 2). Using the five-variable eGFR, 49.5% (1608/3246) of patients had an eGFR of ≥ 90 mL/min/1.73 m², whereas 17.7% (576/3246) had an eGFR of < 60 mL/min/1.73 m² [15]. When adjusted for BSA, 15.9% (610/3828) of patients had an eGFR of ≥ 90 mL/min, whereas 42.9% (1641/3828) had an eGFR of < 60 mL/min [22]. Using CCr estimated with the CG formula, 18.6% (712/3828) of patients had an eGFR ≥ 90 mL/min, whereas 38.2% (1461/3828) had an eGFR of < 60 mL/min [15, 18].

Fig. 4.

Scatter plot showing the relationship between BMI and eGFR. BMI body mass index, eGFR estimated glomerular filtration rate

In the analysis using the five-variable eGFR, age, BMI, and eGFR were significantly associated with serum Mg levels. When using the BSA-adjusted eGFR and CCr estimated following the CG formula, age, BMI, eGFR, and follow-up period were identified as significant factors (Supplementary Table 2). Similar analyses were performed for each renal function measure, employing the same approach used for the standard eGFR. In the linear mixed-effects models, the RI model consistently yielded the lowest AIC values. For the five-variable eGFR, the AIC values were 593.49 for the null model, 331.41 for the RI model, and − 99.69 for the RIC model, with the RIC model demonstrating the lowest AIC value. For the BSA-adjusted eGFR, the AIC values were 677.86 for the null model, 334.00 for the RI model, and 744.35 for the RIC model, with the RI model providing the lowest AIC value. For the CG formula-estimated CCr, the AIC values were 677.86 for the null model, 310.46 for the RI model, and 657.29 for the RIC model, with the RI model demonstrating the lowest AIC value.

The distributions of serum Mg levels and renal function indices (standard eGFR, five-variable eGFR, BSA-adjusted eGFR, and CG formula-estimated CCr) demonstrated negative correlations between serum Mg levels and each renal function measure (Supplementary Fig. 3a–d). In the scatter plots, the bottom-right quadrant represents patients with both hypermagnesemia and impaired renal function. A comparative analysis of data point frequency within this quadrant indicated a higher prevalence of hypermagnesemia when renal function was assessed using the BSA-adjusted eGFR and CG formula-estimated CCr compared with the standard eGFR (Supplementary Table 3).

Discussion

This retrospective study identified two key findings in Japanese patients with eating disorders: (1) hypermagnesemia incidence over 20% and (2) links to low BMI, renal impairment, and younger age.

Degree of elevated serum Mg levels and clinical manifestations

Prior Japanese studies reported approximately 23% hypermagnesemia among Japanese patients prescribed oral MgO, which is consistent with our findings and suggests that patients with eating disorders are not at disproportionately higher risk. [11, 23].

Previous studies have demonstrated that the symptoms of hypermagnesemia are generally nonspecific—such as nausea, dizziness, and weakness—and typically begin to manifest at serum magnesium levels of approximately 4–7 mg/dL. More severe symptoms have been reported at levels exceeding 8–9 mg/dL, with life-threatening events reported at levels above 15 mg/dL [9, 24]. In the present study, the highest recorded Mg level was 5.3 mg/dL, and no severe symptoms were documented, although the possibility of underreporting cannot be excluded. Symptoms appeared as low as 3.1 mg/dL (Supplementary Table 1) although patients with AN often report vague complaints, making it difficult to establish a definitive causal relationship between elevated Mg levels and symptoms.

These limited observations suggest that while careful monitoring remains necessary when prescribing oral MgO to malnourished patients with eating disorders, severe complications were not observed in our cohort. Therefore, MgO need not be avoided in this population, consistent with previous studies [11, 23].

Administration factors of MgO preparations associated with hypermagnesemia

With regard to the effect of MgO prescription on serum Mg levels, a significant association was found between the patient groups; however, no consistent increase in serum Mg levels was noted. Although the influence of MgO prescription varied across models within the elevated group, low eGFR, low BMI, and younger age consistently emerged as significant risk factors. Among these, low eGFR (Fig. 2a) and low BMI (Fig. 2b) showed the strongest associations with elevated serum Mg levels.

MgO prescription was not consistently associated with higher magnesium, whereas low eGFR, low BMI, and younger age were. Renal handling dominates magnesium homeostasis, and malnutrition may worsen renal function via dietary deficits [10, 25].

Younger age and shorter follow-up, reflecting early recovery, were linked to elevated serum Mg, likely from delayed development and immature renal function. This may indicate incomplete renal recovery during nutritional rehabilitation. Further studies are needed.

Previous reports also highlight hypermagnesemia in eating disorders under specific conditions: elevated magnesium in AN with advanced renal impairment [26]; hypermagnesemia in AN with chronic laxative abuse; and a fatal case in AN with renal failure and magnesium-containing laxatives [27, 28]. These findings suggest that renal dysfunction and laxative or Mg preparations act synergistically to increase risk.

Appropriate screening for serum Mg elevation considering body size

As Cr levels tend to underestimate true renal function in underweight individuals with reduced muscle mass—including patients with eating disorders—the standard eGFR formula commonly used in Japan likely overestimates renal function in this population [29]. In our analysis, a weak negative correlation was observed between standard eGFR and BMI (Fig. 4). However, this did not align with previous studies, which suggest lower body weight is associated with reduced renal function [6, 14, 29]. This discrepancy highlights that standard eGFR may be unreliable in undernourished patients. Impaired renal function was a significant risk factor for elevated Mg; clinicians should be cautious of hypermagnesemia, even when eGFR appears normal.

Both BSA-adjusted eGFR and CG formula-estimated CCr generally yielded lower renal function values compared with the standard eGFR. These indicators may help assess renal function when serum Mg is elevated. However, comparisons should be interpreted cautiously due to lack of a gold standard and differing units (mL/min and mL/min/1.73m²). The present findings suggest that the BSA-adjusted eGFR may reduce the risk of renal function misclassification when evaluating serum Mg abnormalities.

However, the five-variable eGFR may have limited utility as a screening method, as only a few patients exhibited an eGFR of < 60 mL/min/1.73 m² compared with the standard eGFR. Although the inclusion of BUN and Alb was intended to mitigate the effect of Cr, the limited sensitivity of this formula may be attributed to the tendency for Alb and BUN levels to be reduced in malnourished patients.

Clinicians should remain vigilant for hypermagnesemia in underweight patients, even when the prescribed MgO dosage appears safe and standard eGFR values fall within the normal range. Body size-adjusted renal function indices may enhance screening accuracy in this population.

Strengths and limits

This is the largest Japanese study on hypermagnesemia in eating disorder patients and the first to assess its relationship with renal function indices, including five-variable eGFR, BSA-adjusted eGFR, and CG-estimated CCr. These findings provide new insights into risk stratification and renal function screening in malnourished patients, underscoring the need for tailored clinical approaches in this vulnerable population.

First, MgO intake was assessed using prescription records and medications brought at admission, so privately procured products may have been missed; however, no clinically significant symptoms were observed, suggesting oral MgO safety was not compromised.

Second, only female patients were included, as few men with eating disorders were admitted and different eGFR equations are used for each sex.

Third, the retrospective design may have limited the detection of hypermagnesemia-related symptoms, as undocumented complaints may have gone unreported by patients.

Fourth, accurate GFR measurement using exogenous markers was not performed, and vitamin D or diuretic use—potential confounders—was excluded due to their low frequency [11, 22, 30–32].

Fifth, variability in sampling timing and limited information on hydration, nutrition, and concomitant medications may have affected the results. In addition to oral MgO, several types of enteral nutritional formulas containing magnesium were used during treatment. Although both enteral nutrition and regular meals contributed to magnesium intake, their exact amounts could not be quantified and were therefore not measured in this study.

Sixth, analysis by daily MgO dose and cumulative duration would have been valuable but was precluded by record limitations; future prospective studies should address this issue.

Seventh, diuretic use was confirmed in 1% of patients. As both in- and outpatients were included, laxative/diuretic abuse could not be reliably assessed; it is restricted during hospitalization but uncertain in outpatients. Evaluating associations with renal dysfunction would require focusing only on inpatients under controlled conditions, yet past abuse would remain difficult to assess.

What is already known on this subject?

Patients with eating disorders, especially AN, often develop physical comorbidities such as renal dysfunction and chronic constipation. In this group, eGFR may overestimate renal function due to low muscle mass. While oral MgO, widely used in Japan, is generally safe due to low bioavailability, it can cause hypermagnesemia in those with renal impairment—a risk noted in prior studies and health authority warnings. However, susceptibility to hypermagnesemia in this population remains insufficiently studied.

What this study adds?

This study found a high incidence of hypermagnesemia (> 20%), with serum magnesium levels up to 5.3 mg/dL. Unlike previous reports, MgO use was not consistently linked to elevated Mg, suggesting relative safety as a laxative. Instead, younger age, lower BMI, and reduced renal function were key risk factors. Notably, standard eGFR may overestimate renal function in underweight patients. Our findings show body size-adjusted indices (BSA-adjusted eGFR, CG-estimated CCr) may more accurately detect renal impairment, improving hypermagnesemia screening.

Conclusion

This study found a hypermagnesemia incidence of ~ 20% in eating disorder patients, with low BMI and impaired renal function as key factors. Monitoring serum Mg levels is important regardless of MgO use. Standard eGFR may overestimate renal function in severely underweight patients, indicating the need for size-adjusted assessment. Future studies using inulin clearance could better identify those at risk.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Mariko Uematsu, Satoshi Tanaka, Miho Imaeda, and Masashi Ikeda Methodology: Takahiro Imaizumi, Satoshi Tanaka, Miho Imaeda, and Masashi Ikeda Formal analysis and investigation: Mariko Uematsu, Takahiro Imaizumi, Shintaro Oyama, Hirotake Hida, Hiroki Okumura, and Masashi Ikeda Data curation: Mariko Uematsu, Takahiro Imaizumi, Shintaro Oyama, Hirotake Hida, Hiroki Okumura, and Akemi Morohashi Writing–original draft: Mariko Uematsu and Shiori Ogawa Writing–review and editing: Takahiro Imaizumi, Satoshi Tanaka, Miho Imaeda, and Shiori Ogawa Funding acquisition: Takahiro Imaizumi, Satoshi Tanaka, Miho Imaeda, Shintaro Oyama, Norio Ozaki, and Masashi Ikeda Supervision: Tomoko Oya-Ito, Yoshinari Yasuda, Norio Ozaki, and Masashi Ikeda Project administration: Satoshi Tanaka and Masashi Ikeda Visualization: Mariko Uematsu and Satoshi Tanaka.

Funding

This study was supported by the Grant-in-Aid from the Japan Society for the Promotion of Science KAKENHI (JP20K07942 and JP19K17087) and the Industry-University Collaborative Project for Human Resource Development to Accelerate AI R&D in the Health and Medical Fields.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Review Committee of Nagoya University Graduate School of Medicine (2022-0349) and conducted in accordance with the Declaration of Helsinki. An opt-out consent process, including consent for publication of anonymized data, was implemented through a public notice on the hospital website, and no patients declined participation. Although the study was not preregistered, it was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for observational studies. The requirement for written informed consent was waived because this study was conducted retrospectively using anonymized medical records.

Competing interests

MIk received speakers’ honoraria from Sumitomo Pharma, Eisai, Otsuka, Tanabe Mitsubishi, Mochida, Takeda, Meiji Seika Pharma, EA Pharma, Viatris, MSD, Janssen, Lundbeck, and Yoshitomi. NO received research support, speakers’ honoraria, and/or royalties from or has served as a joint researcher or consultant for Sumitomo Pharma, Otsuka, KAITEKI, Takeda, Ricoh, Meiji Seika Pharma, Taisho Pharma, Mochida, Shionogi, Mitsubishi Tanabe, Tsumura, EA Pharma, Eli Lilly, Daiichi Sankyo, MSD, Lundbeck Japan, Viatris, Eisai, Mochida, Kyowa Pharmaceutical Industry, Nihon Medi-Physics, Kyowa Kirin, Janssen Pharmaceuticals, Yoshitomi Pharmaceutical, Nippon Chemiphar, Medical Review, Nippon Boehringer Ingelheim, Ono Pharmaceutical, Woolsey Pharmaceuticals, and SUSMED, outside the submitted work.