In the pediatric intensive care unit (ICU), magnesium (Mg) has the highest prevalence of abnormal values, which are associated with poor outcomes [1], including mortality [2^-4]. As the second most abundant intracellular cation in the body, it plays crucial roles, including acting as a cofactor for many enzymatic reactions, regulating intracellular calcium (Ca) concentrations and smooth muscle tone, and modulating immune responses [5]. The benefits of Mg supplementation include coronary vasodilation, reduced peripheral vascular resistance, and prevention or treatment of arrhythmias. However, the Mg measured clinically is not the active form.

Similar to Ca, circulating extracellular Mg exists in three forms (protein-bound, anion-complexed, and ionized) and is mostly unbound. Ionized Mg (iMg) is the most active and abundant extracellular form, accounting for approximately two-thirds of the total extracellular Mg. There have been conflicting reports on the correlation between total Mg (tMg) and iMg [6, 7]. In a study in an adult surgical ICU, one group reported moderate correlation but poor category agreements (low, normal, high values) between the two, with >80% of “low” tMg associated with normal iMg concentrations [8]. Conversely, although nearly 60% of critically ill children exhibited low iMg in a previous study, more than half of them had normal tMg [9].

The kidneys drive Mg homeostasis [5], which can be affected by kidney function, acid-base status, Ca-vitamin D alterations, and commonly used medications, such as calcineurin inhibitors, antimicrobials, and citrate [10, 11]. Post-kidney transplant (KTx) patients exhibit renal tMg wasting and are administered calcineurin inhibitors, which commonly cause hypomagnesemia [12^-14]. However, little is known about post-KTx iMg concentrations, particularly during the acute period. Given the physiological similarities between Ca and Mg, as similarly sized divalent cations that exist in both free and bound forms and have blood concentrations that depend on similar factors (temperature, albumin, and pH) [15], we sought to investigate whether ionized Ca (iCa) can serve as a surrogate for iMg. In line with previous studies, we hypothesized that in post-KTx patients, the measurement of blood iMg concentrations would result in different category agreements and reveal poor correlations with blood tMg concentrations.

Data collection

The Institutional Review Board (IRB) of Cincinnati Children’s Hospital Medical Center (CCHMC, Cincinnati, OH, USA), approved the retrospective component of this study (approval ID #2021-0581) and waived the requirement for informed consent. We performed chart reviews for all patients receiving a KTx at CCHMC between November 2019 and September 2021, collecting demographic and outcome data; laboratory values; and whether the patients received Mg supplementation, had diarrhea, or required changes in tacrolimus dosing. The decision to provide Mg supplementation was at the discretion of the attending nephrologist, and occasionally, the ICU provider. At our institution, there is no universally agreed-upon Mg value at which one should start enteral replacement, but most agree to begin replacement at a Mg value between 0.41 and 0.53 mmol/L. Thus, practice variation accounted for the varying threshold concentrations at which supplementation was provided. Patients who received a dose of either intravenous MgSO4 or oral MgO before or within 4–48 hrs of a sample draw were included in the supplementation subgroup.

The CCHMC IRB approved the prospective portion of this study. Parents and/or patients provided informed consent before or within 48 hrs of transplantation, and the study design was in agreement with the Declaration of Helsinki. We enrolled pediatric patients who received a KTx at CCHMC between October 2021 and July 2022 for the prospective portion of the project. We collected 0.3–1 mL of whole blood from patients on post-operative days (PODs) 0–28 for iMg and iCa measurements. Each patient provided multiple samples on different days. Patients requiring post-transplant renal replacement therapy or those who underwent multiple organ transplantation were excluded. Demographic, laboratory, and outcome data were extracted from the electronic medical records, with concentrations of tMg (mg/dL), total Ca (tCa, mg/dL), and iCa (mmol/L) paired with values obtained from the sample drawn at the time of iMg blood collection.

Mg measurement

Mg was measured in venous whole-blood samples using a Stat Profile Prime Plus analyzer (Nova Biomedical, Waltham, MA, USA). All analyses were performed within 6 hrs of sample collection after confirming analyte stability in our laboratory. iMg concentrations (mmol/L) were unknown to the clinical providers and were not used for clinical decision-making purposes. Based on the literature [16^-18] and manufacturer’s guidelines, we defined as normal an iMg concentration of 0.44–0.65 mmol/L and an iCa concentration of 1.0–1.3 mmol/L.

tMg concentrations were measured in a clinical laboratory using a Siemens Atellica analyzer (Siemens Healthcare Diagnostics, Inc., Tarrytown, NY, USA), which employs a xylidyl blue dye-binding assay but is based on an atomic absorption reference method. tMg concentrations of 0.66–1.07 mmol/L and tCa concentrations of 2.15–2.63 mmol/L were classified as normal based on our center’s clinical laboratory reference ranges. It should be noted that there is considerable variation in standard reference ranges for total Mg owing to analytical method variability as well as within-subject and between-subject biological variations, which typically range from 3% to 6% [19]. For conversion from mmol/L to mg/dL, multiplication factors of 2.43 and 4 are used for Mg and Ca, respectively.

Statistical analysis

All statistical analyses were performed using R (https://www.r-project.org/) or the SPSS software. Normality was assessed using the Shapiro–Wilk test. Normally distributed data are reported as mean±SD and non-normally distributed data as median±in­terquartile range (IQR). Wilcoxon rank-sum testing was used for group-wise comparisons of outcomes. We examined category agreements between iMg and tMg and between iMg and iCa using Fisher’s exact test and the unweighted kappa statistic. We examined the relationships between these concentrations using Pearson’s correlation coefficient and linear regression analyses. We used multiple linear regression to control for serum creatinine (SCr) and albumin concentrations on the sample date. We considered P<0.05 to reflect a statistically significant difference.

Retrospective post-transplant data

Fifty-eight patients received a KTx between November 5, 2019 and September 25, 2021. The median age at the time of KTx was 11.7 (IQR 4.9–16.1) yrs. Among these patients, 27 (47%) were female, 44 (76%) were Caucasian, and 54 (93%) were non-Hispanic. On POD 0, the median tMg was 0.86 (IQR 0.78–0.99) mmol/L. Forty-nine (84%) patients had normal tMg concentrations, and two patients received Mg supplementation on the day of transplantation. In the first 28 days post-KTx, 93% of the patients developed hypomagnesemia (tMg<0.66 mmol/L). Regardless of supplementation, the mean tMg nadir concentration was 0.41 mmol/L, and the median time of the nadir was on POD 7.5 (IQR 4–13, Table 1).

Table 1 . Demographics and outcome data for the retrospective kidney-transplant cohort

Variable	Total (N=58)	Received Mg supplementation (N=28)*	Did not receive Mg supplementation (N=30)*	P†
Age at transplant, yrs	11.7 (4.9–16.1)	9.4 (4.7–14.9)	14.7 (5.0–17.2)	0.126
Female sex, N (%)	27 (47)	14 (50)	13 (43)
Race	44 white, 7 black, 2 Asian/PI, 1 mixed, 4 unknown	22 white, 4 black, 2 unknown	22 white, 3 black, 2 Asian/PI, 1 mixed, 2 unknown
Ethnicity	54 non-Hispanic, 3 Hispanic, 1 unknown	25 non-Hispanic, 2 Hispanic, 1 unknown	29 non-Hispanic, 1 Hispanic
Height, cm	131.5 (97.6–157.9)	125.0 (93.5–152.4)	146.8 (102.6–160.2)	0.131
Weight, kg	34.9 (15.5–52.1)	26.2 (14.1–44.6)	43.5 (16.8–52.4)	1.000
Baseline (POD 0) Mg, mmol/L	0.86 (0.78–0.99)	0.86 (0.74–0.99)	0.90 (0.82–0.99)	0.211
POD of Mg nadir, days	7.5 (4.0–13.0)	7.5 (3.8–16.5)	7.5 (5.0–13.0)	0.882
Mg nadir, mmol/L	0.51±0.077	0.48±0.071	0.55±0.070	0.002
Mg at supplementation, mmol/L		0.55±0.099	NA
ICU LOS, days	3 (2–4)	4 (2–8)	3 (2–3)	0.018
Hospital LOS, days	11.5 (9.0–16.0)	12.0 (9.0–20.0)	11.0 (9.0–14.8)	0.254

Data are expressed as median (interquartile range) or mean±SD.

*Wilcoxon rank-sum tests were used to identify significant differences between the patients who did and did not receive Mg supplementation.

^†P<0.05 was considered to reflect a statistically significant difference.

Abbreviations: Mg, magnesium; PI, Pacific Islander; POD, postoperative day; ICU, intensive care unit; LOS, length of stay.

Twenty-eight (48%) patients received Mg supplementation immediately post-KTx. Intravenous or enteral Mg supplementation was initiated on median POD 4 (IQR 2–9.5), at a mean±SD tMg concentration of 0.55±0.099 mmol/L. Of these 28 patients, 13 (46%) had diarrhea and 20 (71%) had their tacrolimus dose adjusted during supplementation. In addition, 10/13 (77%) patients with diarrhea had their tacrolimus dose adjusted after Mg supplementation. Patients who received Mg supplementation had a longer median ICU length of stay (LOS) than those who did not receive supplementation (4 [IQR 2–8] vs. 3 [IQR 2–3] days, P=0.018) but not a longer hospital LOS (12 [IQR 9–20] vs. 11 [IQR 9–14.8] days, P=0.25).

Prospective post-transplant demographics

Thirteen patients contributed 43 samples after receiving a KTx between November 1, 2021 and July 8, 2022, with 33 (77%) samples drawn on PODs 1–8. Most patients contributed three to four unique samples. The mean age at transplant was 9.5 6.6 yrs. Among these patients, four (31%) were female, nine (69%) were white, and 13 (100%) were non-Hispanic. The median ICU LOS was 2 (IQR 2–4) days, and the mean±SD hospital LOS was 13.7±7.5 days (Table 2).

Table 2 . Demographics and outcome data for the prospective kidney-transplant recipients

Variable	Total (N=13)
Age at transplant, yrs	9.5±6.6
Female sex, N (%)	4 (31)
Race	9 white, 2 black, 1 mixed, 1 unknown
Ethlefticity	13 non-Hispanic
Height, cm	126.3 (89–158.8)
Weight, kg	27 (12.8–59.4)
N of patients who received Mg supplementation, N (%)	4 (31)
N of samples for which Mg supplementation occurred, N (%)	9 (21)
tMg concentration, mmol/L	0.66 (0.53–0.78)
Supplementation	0.53 (0.51–0.78)
No supplementation	0.66 (0.57–0.78)
iMg concentration, mmol/L	0.56 (0.43–0.65)
Supplementation	0.47 (0.42–0.60)
No supplementation	0.59 (0.46–0.66)
ICU LOS, days	2 (2–4)
Hospital LOS, days	13.8±7.5

Data are expressed as median (interquartile range) or mean±SD.

Abbreviations: Mg, magnesium; tMg, total magnesium; iMg, ionized magnesium; ICU, intensive care unit; LOS, length of stay.

The median (IQR) tMg concentration in this cohort was 0.66 (0.53–0.78) mmol/L and the median iMg (IQR) concentration was 0.56 (0.43–0.65) mmol/L. Four of the 13 patients (31%) received Mg supplementation during their postoperative course; none had diarrhea, and three had tacrolimus dose adjustments. Nine samples were collected 4–48 hrs after Mg supplementation; two patients received Mg sulfate and seven received oral MgO. The median tMg (IQR) concentration was 0.53 (0.51–0.78) mmol/L in samples from patients who received concurrent supplementation and 0.66 (0.57–0.78) mmol/L in samples from those who did not. The median (IQR) iMg concentration was 0.47 (0.42–0.60) mmol/L in samples from patients who received concurrent supplementation and 0.59 (0.46–0.66) mmol/L in samples from those who did not.

Prospective post-transplant iMg correlations

The iMg concentrations decreased over time during the immediate post-KTx period. Of the 12 patients who provided more than one sample for analysis, four (33%) had iMg values that decreased from high to low or from normal to low concentrations. Eleven of the 43 iMg concentrations were low (26%), 22 were normal (51%), and 10 were high (23%). In addition, 19 tMg concentrations were low (44%), 24 were normal (56%), and zero were high (0%). We observed agreement between the iMg and tMg categories (Fisher’s exact test, P=0.0003; unweighted kappa=0.27, P=0.013; Table 3). The relationship between the continuous measurements of iMg and tMg was also positive and significant (Pearson’s r=0.71, P<0.001; linear regression B [tMg]=0.315, P<0.001; Fig. 1A). This relationship remained significant after controlling for SCr and albumin concentrations (B [tMg]=0.318, P<0.001).

Table 3 . Category agreement between ionized magnesium and total magnesium or ionized calcium in prospective kidney-transplant recipients^*

Variable		iMg
		Low	Normal	High
tMg	Low	10	8	1
	Normal	1	14	9
	High	0	0	0
iCa	Low	0	0	0
	Normal	6	15	7
	High	5	7	3

*Fisher’s exact test and the unweighted kappa statistic were used to assess category agreement between tMg and iMg (Fisher’s exact P=0.0003, unweighted kappa [95% confidence interval CI]=0.27 [0.056–0.47], P= 0.013) and between iCa and iMg (Fisher’s exact P=0.7, unweighted kappa [95% CI]=0.0074 [–0.19–0.20], P=0.94).

Abbreviations: iCa, ionized calcium; iMg, ionized magnesium; tMg, total magnesium.

Figure 1. Correlations between iMg and tMg or iCa in prospective kidney transplant recipients. (A) Pearson’s correlation coefficients and simple linear regression analysis were used to assess the relationship between iMg and tMg (Pearson’s r=0.707). The dashed line represents the linear regression equation (P<0.001). (B) Correlation between iMg and iCa, with the dashed line representing the linear regression equation (Pearson’s r=–0.252, P=0.103). For conversion from mmol/L to mg/dL, multiplication factors of 2.43 and 4 are used for Mg and Ca, respectively.
Abbreviations: iMg, ionized magnesium; tMg, total magnesium; iCa, ionized calcium.

Given the feasibility of iCa measurements and their similar electrochemical properties, the relationship between iMg and iCa was assessed. We observed zero (0%) low, 28 (65%) normal, and 15 (35%) high iCa concentrations. Among the 11 low iMg concentrations, iCa was normal in six samples and high in five samples. We observed poor category agreement between iMg and iCa (Fisher’s exact test, P=0.7; unweighted kappa=0.0074, P=0.94; Table 3). Further analyses indicated no relationship between continuous measurements of these concentrations (Pearson’s r=–0.25, P=0.1034; linear regression B [iCa]=–0.333, P= 0.103; Fig. 1B). The relationship remained insignificant after controlling for SCr and albumin concentrations (B [iCa]=–0.227, P= 0.321).

In evaluating category agreements among patients who received Mg supplementation, eight of nine (89%) tMg concentrations were low (normal value: 0.70 mmol/L), and three (33%), four (44%), and two (22%) iMg concentrations were low, normal, and high, respectively. Category agreement was not assessed statistically given the low number of samples; however, one-third (33%) showed category agreement (low/low), whereas four had low tMg/normal iMg, one had normal tMg/high iMg, and one had low tMg/high iMg. In this very small subset of samples, the relationship between the continuous measurements of iMg and tMg was very similar to that of the whole group; thus, they likely did not skew the associations (Pearson’s r=0.78, P=0.013; linear regression B [tMg]=0.67, P=0.013).

Given the discrepancies in the literature and the lack of available data on pediatric patients in the acute post-KTx period, we assessed the frequency of hypomagnesemia and the adverse effects of Mg supplementation post-KTx, and correlated the concentrations of Mg with those of tMg and iCa. We found that iMg and tMg showed category agreement and a strong correlation, whereas iMg and iCa did not. Therefore, tMg may reliably represent active Mg. However, iCa should not be used as a surrogate for active Mg in post-KTx patients.

Although it is well known that renal Mg handling is altered post-KTx, we assessed and compared iMg concentrations with more readily available tests (tMg and iCa measurements) in the immediate post-KTx period. Several studies have examined post-KTx iMg concentrations, but these involved adults with a long-term transplant status or >6 months post-KTx, with a focus on the effects of calcineurin inhibitors on iMg [11, 14, 20, 21]. Koch, et al. [7] were the first to investigate the correlation between iMg and tMg concentrations in adult patients (N=34) in an ICU setting. The excellent correlation they found (r=0.903) in their small cohort has not yet been matched, but the results of another large ICU study involving adults showed very similar correlations to ours (r=0.71, P<0.0001) [7, 16]. In a single-center pediatric study examining all patients admitted to a pediatric or a cardiac ICU, 60% of patients had ionized hypomagnesemia, and 60% of those patients had normal tMg. It is likely that the substantially lower rate of ionized hypomagnesemia in our study was related to the unique aberrations that occur in the peri-transplant period in pediatric KTx recipients. Despite the lack of category agreement between iMg and tMg in heterogeneous populations (including patients with cardiac disease), this group also showed a poor correlation between iMg and iCa [9].

The misclassification of iMg has clinical implications, especially in the post-KTx population. Ionized hypo- and hypermagnesemia have been associated with poor outcomes, including mortality [3, 4]. However, Mg supplementation is not benign. Analysis of our retrospective cohort showed that >90% of the patients had low tMg concentrations in the first 28 days post-KTx, which led to supplementation in nearly half of the patients. The values at which Mg was supplemented was provider-dependent; the fact that tMg nadired at a concentration of 0.41 mmol/L in our cohort suggests reluctancy to and late initiation of supplementation. A more liberal supplementation threshold would result in far more Mg supplementation. Supplementation can lead to diarrhea, which may contribute to altered calcineurin absorption and frequent dose adjustments. The results of our short-term retrospective study suggest that dose adjustments may prolong the ICU LOS. If Mg supplementation leads to abnormal calcineurin concentrations in the subacute/long term, it may lead to calcineurin inhibitor toxicity or an increased risk of graft rejection.

The impact of Mg supplementation on iMg and tMg remains variable and possibly dependent on the type of Mg administered; studies have shown that iMg can increase without an increase in tMg [22, 23]. Although less than a quarter of our samples were drawn right before or after Mg supplementation, we investigated whether this skewed the correlation. Category agreement was observed for one-third of the samples, and the correlation was stronger in the supplemented than in the non-supplemented group.

The lack of a standard reference range is an important reason why routine iMg testing does not occur and why its use in research settings is limited. Although different reference ranges have been proposed based on age and race, normal ranges in healthy adults have spanned from 0.39 to 0.71 mmol/L [16, 17, 24, 25]. The largest pediatric cohort studied to date showed a normal range of 0.52–0.63 mmol/L [18]. We chose a wider normal range of 0.44–0.65 mmol/L to decrease the chance of false-positive abnormal category assignments to favor the null hypothesis (i.e., fewer patients will have low or high iMg). As mentioned previously, tMg testing is not immune to this problem. Within- and between-person variations of 3%–6% exist, and multiple methods of analysis result in different reference standards among laboratories, making the generalizability of studies very difficult [19]. We attempted to overcome these shortcomings using linear regression models to examine the correlations, and the results were not affected by using standardized reference ranges.

This study had several limitations. First, we did not correct iMg for pH or control for medications such as tacrolimus, although we did control for albumin and SCr, and each patient’s tacrolimus doses were modified such that they fell within a narrow therapeutic range. Studies have shown that iMg was approximately one-third as sensitive to pH changes as iCa, and in a large adult study, there was no significant impact of pH (or albumin) on the iMg concentration in correlation analyses [26, 27]; therefore, we believe the lack of controlling for pH did not impact our results. We did not include a control group; instead, we relied on reference ranges reported in the literature. We felt that a control group would not answer our clinical question about whether normally tested values (tMg and iCa) could help predict the active form of Mg and, thus, relied on previously published control cohorts. Trace element concentrations, including iMg, can be influenced by a patient’s condition or group; therefore, these correlations should not be extrapolated beyond the post-transplant population. This single-center study was conducted at a large children’s hospital in a midwestern region of the USA. For these reasons, along with the small patient size, the generalizability of our results is limited. Lastly, and most importantly, we were unable to capture the untoward clinical consequences of the reported ionized hypomagnesemia, namely arrhythmias (which are rare in the pediatric population) and hypotension. It remains unknown which tMg and iMg concentrations are associated with clinical morbidity in this population and how Mg supplementation affects iMg concentrations. These questions warrant further investigation.

In conclusion, iMg and tMg correlated strongly shortly post-KTx with good category agreement. It is likely that tMg represents active Mg in these patients, which should be considered when providers make decisions regarding supplementation; however, iCa should not be used as a surrogate for iMg concentrations. Mg supplementation is not without side effects; therefore, future studies should focus on how supplementation affects iMg and determine the concentration at which it is clinically necessary to initiate supplementation.

Nova Biomedical provided the machine and supplies to perform this study. However, they played no role in any part of study design and data analysis, nor were they involved in any part of the drafting of the manuscript.

Hasson DC and Mohan S performed the study conception, implementation, data analysis, and patient recruitment, and Hasson DC performed interpretation of results and drafting of the manuscript. Rose JE and Benoit SW oversaw the laboratory and assisted with use of the iMg machine. Merrill KA performed the statistical analysis. Goldstein SL and Varnell CD provided study oversight and aided in the editing/revising of the manuscript.

Financial interests: Stuart L Goldstein reports receiving personal fees from Baxter Healthcare, BioPorto Inc., CHF Solutions, Fresenius, MediBeacon, and Medtronic. Non-financial interests: DCH and SLG have patents for the use of a novel urinary biomarker; this biomarker is completely unrelated to this study. The remaining authors declare no conflicts of interest.

None declared.