http://www.kidney-international.org
clinical investigation
© 2015 International Society of Nephrology
Magnesium modifies the association between serum phosphate and the risk of progression to end-stage kidney disease in patients with non-diabetic chronic kidney disease Yusuke Sakaguchi1, Hirotsugu Iwatani1, Takayuki Hamano2, Kodo Tomida3, Hiroaki Kawabata1, Yasuo Kusunoki1, Akihiro Shimomura1, Isao Matsui1, Terumasa Hayashi3, Yoshiharu Tsubakihara2, Yoshitaka Isaka1 and Hiromi Rakugi1 1
Department of Geriatric Medicine and Nephrology, Osaka University Graduate School of Medicine, Suita, Osaka, Japan; 2Department of Comprehensive Kidney Disease Research, Osaka University Graduate School of Medicine, Suita, Osaka, Japan and 3Department of Kidney Disease and Hypertension, Osaka General Medical Center, Osaka, Japan
It is known that magnesium antagonizes phosphate-induced apoptosis of vascular smooth muscle cells and prevents vascular calcification. Here we tested whether magnesium can also counteract other pathological conditions where phosphate toxicity is involved, such as progression of chronic kidney disease (CKD). We explored how the link between the risk of CKD progression and hyperphosphatemia is modified by magnesium status. A post hoc analysis was run in 311 nondiabetic CKD patients who were divided into four groups according to the median values of serum magnesium and phosphate. During a median follow-up of 44 months, 135 patients developed end-stage kidney disease (ESKD). After adjustment for relevant clinical factors, patients in the lower magnesium-higher phosphate group were at a 2.07-fold (95% CI: 1.23–3.48) risk for incident ESKD and had a significantly faster decline in estimated glomerular filtration rate compared with those in the higher magnesium-higher phosphate group. There were no significant differences in the risk of these renal outcomes among the higher magnesiumhigher phosphate group and both lower phosphate groups. Incubation of tubular epithelial cells in high phosphate and low magnesium medium in vitro increased apoptosis and the expression levels of profibrotic and proinflammatory cytokine; these changes were significantly suppressed by increasing magnesium concentration. Thus, magnesium may act protectively against phosphate-induced kidney injury. Kidney International advance online publication, 10 June 2015; doi:10.1038/ki.2015.165 KEYWORDS: chronic kidney disease; hyperphosphatemia; magnesium
Correspondence: Yoshitaka Isaka, Department of Geriatric Medicine and Nephrology, Osaka University Graduate School of Medicine, 2-2, Yamadaoka, Suita, Osaka 565-0871, Japan. E-mail: [email protected] Received 11 August 2014; revised 29 March 2015; accepted 9 April 2015 Kidney International
Hyperphosphatemia is one of the major complications of chronic kidney disease (CKD). It is associated not only with an increased risk of cardiovascular mortality1 but also with a faster progression of CKD.2–11 Animal studies showed the benefit of dietary phosphate restriction, or use of non-calcium-containing phosphate binders, for attenuation of deteriorating kidney function.12–15 Moreover, klotho-knockout mice presenting hyperphosphatemia exhibited kidney failure and increased apoptotic cells in the kidney;16 after genetically reducing serum phosphate levels by an additional knockout of the sodium phosphate cotransporter gene, kidney function restored with decreased apoptotic cells. Importantly, the kidney function in these double-knockout mice again worsened when hyperphosphatemia was induced by a high-phosphate diet. These findings suggest a clinical significance of attenuating phosphate toxicity to the kidney to improve renal prognosis of CKD patients. Magnesium (Mg), the fourth most abundant cation in the human body, is an essential mineral for numerous biological processes. Mg deficiency is associated with an increased risk of cardiovascular events and vascular calcification, both in individuals with and without CKD.17–23 Notably, an in vitro study has shown that Mg strongly inhibits phosphate-induced apoptosis of vascular smooth muscle cells,24 a key process of vascular calcification. This finding prompted us to postulate that Mg could also counteract other pathological conditions where phosphate toxicity is involved, such as progression of CKD. Here, we examined how Mg status influences the relationship between hyperphosphatemia and the risk of CKD progression. In addition, in vitro experiments were performed to investigate whether Mg acts protectively against the potential cytotoxicity of phosphate on tubular epithelial cells. 1
clinical investigation
RESULTS Cohort study
Of the total of 367 non-diabetic CKD patients who had participated in the CKD educational program, 56 were excluded from the analysis (7 were followed up for o3 months; 41 had missing data on serum Mg or phosphate; and the medical records of 8 patients had been discarded). Baseline characteristics were not substantially different between the participants and those enrolled in the current study (Supplementary Table S1 online). Baseline characteristics according to Mg-phosphate groups (categorized by the median value of serum Mg and phosphate levels) are summarized in Table 1. Higher phosphate groups tended to have a lower estimated glomerular filtration rate (eGFR) and higher intact parathyroid hormone (PTH) levels, whereas lower Mg groups tended to have a slightly higher urine protein and lower serum albumin levels. During the median follow-up of 44 months, 135 patients progressed to end-stage kidney disease (ESKD) and 13 died. The mean eGFR at dialysis initiation was 6.6 ml/min per 1.73 m2 and was not significantly different among the four Mg-phosphate groups (Supplementary Table S2 online). Only 14 patients had refractory congestive heart failure at dialysis initiation, and these cases were equally distributed across the Mg-phosphate groups (Supplementary Table S2 online). The Kaplan–Meier curves (Figure 1a) revealed that the risk of incident ESKD increased significantly in the higher phosphate groups, with the lower Mg-higher phosphate group having the worst prognosis. It should be noted, however, that baseline eGFR was significantly different among the four groups. Kaplan–Meier curves that were adjusted for baseline eGFR (Figure 1b) showed that the lower Mg-higher phosphate group still had the worst prognosis, but that the risks of the other three groups were almost equivalent. In the Cox models, serum Mg and phosphate were treated as categorical variables because of the following reasons: (1) a nonlinear association between serum Mg and phosphate levels and the natural logarithm of the hazard function, and (2) the violation of the proportional hazard assumption for serum phosphate. The lower Mg-higher phosphate group showed the highest risk of progression to ESKD among the four Mgphosphate groups after adjustment for baseline eGFR (Table 2 and Figure 2a). Similarly, eGFR slope of the lower Mg-higher phosphate group was significantly steeper than those of the other three groups and was approximately two times as fast as that of the higher Mg-higher phosphate group (Table 2). These findings were maintained in multivariate Cox models, which revealed that patients in the lower Mg-higher phosphate group consistently showed the highest risk of progression to ESKD among the four Mg-phosphate groups; patients in this group had a 2.07-fold higher risk compared with those in the higher Mg-higher phosphate group in the fully adjusted model (95% confidence interval: 1.23–3.48; Supplementary Table S3 online). After further adjustment for intact PTH and fibroblast growth factor 23 (FGF23) by using a multiple imputation method, patients in the lower Mg-higher phosphate group still had the highest risk (Supplementary Tables S3 and S4 online). 2
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
No significant risk difference was found between the lower Mglower phosphate and the higher Mg-lower phosphate group. Of note, there were also no significant risk differences among the higher Mg-higher phosphate group and both lower phosphate groups. Similar results were obtained in the Cox models where ESKD and death were treated as a composite outcome (Supplementary Table S3 online). In multivariate linear regression models, the eGFR slope of the lower Mg-higher phosphate group was again steeper compared with those of the other three groups (Table 3 and Figure 2b). The slopes were not significantly different among the higher Mg-higher phosphate group and both the lower phosphate groups. Further adjustment for intact PTH and FGF23 yielded similar results (Table 3 and Supplementary Table S4 online). In vitro experiments Effects of Mg and phosphate on apoptosis in mProx cells.
As phosphate is known to induce apoptosis in the kidney cells,16 we examined effects of Mg against the proapoptotic property of phosphate. Incubation of mProx cells, a cell line of mouse proximal tubular cells, with high phosphate (2.0 mmol/l) and low Mg (0.75 mmol/l) increased apoptotic cells compared with physiological phosphate (0.9 mmol/l) and 0.75 mmol/l Mg; however, the high phosphate-induced apoptosis was significantly suppressed by increasing Mg concentration to 1.5 mmol/l (Figure 3a and b). In contrast, Mg did not influence tubular cell apoptosis in the medium with 0.9 mmol/l phosphate. The interaction between the Mg and phosphate concentrations on the proportion of apoptotic cells was significant (Po0.01; two-way analysis of variance (ANOVA)), indicating that the effect of Mg was significantly different between the high and low phosphate conditions. Similarly, the expression of cleaved caspase-3, a major executor of apoptosis, was enhanced in cells treated with 2.0 mmol/l phosphate and 0.75 mmol/l Mg, which was significantly inhibited by increasing Mg level to 1.5 mmol/l (Figure 3c and d). Mg inhibits phosphate-induced depolarization of mitochondrial membrane potential. To examine whether the apoptosis
caused by high phosphate and low Mg was dependent on the mitochondrial pathway, we evaluated mitochondrial membrane potential by tetramethylrhodamine ethyl ester staining. The mitochondrial membrane potential was significantly reduced in cells treated with 2.0 mmol/l phosphate and 0.75 mmol/l Mg compared with those treated with 0.9 mmol/l phosphate and 0.75 mmol/l Mg (Figure 4a and b); the membrane potential was restored by increasing Mg concentration to 1.5 mmol/l. The effect of Mg on the membrane potential was not apparent in the medium with 0.9 mmol/l phosphate. The interaction between the Mg and phosphate concentrations on the membrane potential was significant (Po0.01; two-way ANOVA). Mg suppresses gene expressions of profibrotic and proinflammatory cytokines induced by high phosphate. Tubulointerstitial
fibrosis is the hallmark of CKD progression and is characterized by excessive deposition of extracellular matrix proteins secreted Kidney International
Kidney International
(32) (24) (48) (26)
65 21 27 4 6
(68) (22) (28) (4) (6)
77 (80) 31 (32)
48 21 19 8
(19) (27) (25) (19)
36 14 19 2 12
(52) (20) (28) (3) (17)
55 (80) 25 (36)
29 24 10 6
64.7 (12.2) 46 (67) 23.4 (4.0) 2.11 (0.88) 27.3 (10.0) 9.2 (0.7) 2.3 (0.2) 3.1 (0.4) 73 (41, 119) 3.9 (0.6) 189 (38) 0.4 (0.1, 1.0)
42.1 ⩽ 3.6
⩽ 2.1 ⩽ 3.6
64.4 (10.3) 70 (73) 23.1 (3.3) 1.94 (0.73) 30.0 (10.7) 9.1 (0.6) 1.9 (0.2) 3.1 (0.4) 62 (42, 106) 3.7 (0.5) 189 (47) 0.5 (0.1, 1.3)
Higher Mg-lower P, n = 69
Lower Mg-lower P, n = 96
(24) (21) (15) (26)
39 (57) 20 (29) 18 (26) 5 (7) 3 (4)
63 (91) 16 (23)
37 18 6 8
64.0 (13.2) 42 (61) 24.4 (4.2) 2.70 (1.17) 21.9 (10.4) 9.1 (0.6) 2.0 (0.2) 4.3 (0.7) 154 (53, 210) 3.7 (0.5) 199 (52) 1.0 (0.1, 2.3)
⩽ 2.1 43.6
Lower Mg-higher P, n = 69
(25) (28) (13) (29)
38 (49) 28 (36) 29 (38) 17 (22) 4 (5)
61 (79) 22 (29)
38 25 5 9
66.5 (12.5) 29 (38) 23.4 (4.1) 2.94 (1.30) 18.4 (8.9) 9.0 (0.5) 2.5 (0.4) 4.4 (0.7) 113 (69, 183) 3.8 (0.4) 205 (51) 0.4 (0.1, 1.8)
42.1 43.6
Higher Mg-higher P, n = 77
0.07 0.09 0.4 o0.001 0.01
0.2 0.4
0.2
0.6 o0.001 0.2 o0.001 o0.001 0.4 o0.001 o0.001 o0.001 0.03 0.09 0.02
P-value
Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; Mg, magnesium; P, phosphate; PTH, parathyroid hormone; RAS, renin–angiotensin– aldosterone system. Data are presented as mean (s.d.) or median (interquartile range) or number (%).
(57) (27) (30) (9) (8)
0 0 0 0 0
Prescriptions RAS inhibitor, n (%) Diuretics, n (%) Statins, n (%) Calcium carbonate, n (%) Magnesium oxide, n (%) 178 83 93 28 25
256 (82) 94 (30)
0 0
Comorbidities Hypertension, n (%) Prior history of CVD, n (%)
(100) (100) (100) (100)
64.9 (11.9) 187 (60) 23.5 (3.9) 2.39 (1.10) 24.7 (11.0) 9.1 (0.6) 2.2 (0.3) 3.7 (0.9) 90.5 (51, 156) 3.8 (0.5) 195 (48) 0.5 (0.1, 1.6)
152 88 40 31
0
0 0 3 (1) 0 0 0 0 0 145 (46.6) 0 2 (1) 5 (2)
Total, n = 311
Primary causes of CKD Chronic glomerulonephritis, n (%) Benign nephrosclerosis, n (%) Others, n (%) Unknown, n (%)
Characteristics Age (year) Male, n (%) BMI (kg/m2) Creatinine (mg/dl) eGFR (ml/min per 1.73 m2) Corrected calcium (mg/dl) Magnesium (mg/dl) Phosphate (mg/dl) Intact PTH (pg/ml) Albumin (g/dl) Total cholesterol (mg/dl) Urine protein (g per day)
Serum Mg range (mg/dl) Serum P range (mg/dl)
No. of missing values (%)
Table 1 | Baseline characteristics of 311 non-diabetic chronic kidney disease patients according to Mg-phosphate groups
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
clinical investigation
3
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
1.00 Cumulative event-free survival
Cumulative event-free survival
1.00
0.75
0.50
0.25
Higher Mg-lower P Lower Mg-lower P Higher Mg-higher P Lower Mg-higher P
0.00 0
12
96 69 Higher Mg-lower P 69 Higher Mg-higher P 77 Lower Mg-higher P
36
48
93 58 65 62
81 33 59 53
69 28 49 42
52 18 39 33
0.75
0.50
0.25
0.00 60
72
84
35 10 31 25
0
12
24
36
48
60
72
84
Follow-up time (months)
Follow-up time (months)
Number at risk Lower Mg-lower P
24
Higher Mg-lower P Lower Mg-lower P Higher Mg-higher P Lower Mg-higher P
19 5 21 17
11 4 10 12
Figure 1 | Kaplan–Meier curves for the risk of progression to end-stage kidney disease in 311 non-diabetic chronic kidney disease patients stratified by magnesium-phosphate groups. (a) Unadjusted Kaplan–Meier curves (Po0.001 by log-rank test) and (b) Kaplan–Meier curves adjusted for the estimated glomerular filtration rate (scaled to 15 ml/min per 1.73 m2). The curves were truncated when the percentage of patients remaining at risk was ~ 10%.
mainly from myofibroblasts. Because tubular cells contribute to the development of fibrosis by producing several profibrotic and proinflammatory cytokines in response to various injurious stimuli,25 we investigated whether the expression levels of these cytokines are altered by medium concentrations of phosphate and Mg. The incubation of mProx cells with 2.0 mmol/l phosphate and 0.75 mmol/l Mg increased mRNA expression levels for transforming growth factor-β1 and interleukin-6 compared with incubation with 0.9 mmol/l phosphate and 0.75 mmol/l Mg (Figure 5). Increasing the Mg level to 1.5 mmol/l significantly suppressed the expressions of these genes. Mg did not alter the mRNA expression levels of these genes in the medium with 0.9 mmol/l phosphate. There was significant interaction between Mg and phosphate concentrations on the mRNA expressions of transforming growth factor-β1 and interleukin-6 (Po0.01; two-way ANOVA). DISCUSSION
Previous studies have reported, albeit inconsistently, a significant association between hyperphosphatemia and faster progression of CKD.3–11 We expanded these findings by showing that the risk of ESKD related to high serum phosphate was modified by Mg status; patients with high phosphate had a higher risk of ESKD when they had a low serum Mg level. This finding is in keeping with our hypothesis that Mg acts protectively against phosphateinduced kidney injury. In fact, our in vitro experiments demonstrated that Mg suppressed apoptosis and the production of profibrotic and proinflammatory cytokines of tubular cells induced by high phosphate. Several mechanisms underlying our observation can be assumed. Mitochondrial permeability transition (PT) is one of the key processes in mitochondria-mediated cell death.26,27 PT is caused by opening of the mitochondrial PT pore (mPTP), which makes the mitochondrial inner membrane 4
unselectively permeable to solutes of molecular weight ⩽ 1.5 kDa. As a result, mitochondrial matrix swelling and rupture of the outer membrane occur, leading to activation of mitochondrial apoptotic pathway.28 Of note, Mg and phosphate directly and oppositely act on the mPTP and induce its conformational change; phosphate promotes the mPTP opening, whereas Mg inhibits it.26,27 Therefore, Mg suppresses phosphate-induced apoptosis of tubular cells, possibly by competing the mPTP opening facilitated by phosphate. Although we did not directly evaluate the state of the mPTP opening, we found that Mg restored the mitochondrial membrane potential of tubular cells collapsed owing to high phosphate concentration. There may be another explanation for the protective role of Mg against phosphate toxicity. It is known that nephrocalcinosis is a common pathological change observed in the kidney exposed to high phosphate.12 Serum phosphate is positively associated with calcium content in human kidney biopsy samples.29 In CKD animal models, a low-phosphate diet, or use of non-calcium-containing phosphate binders, was useful in reducing renal calcium content and improving renal histology.12,14,15 As Mg has an anticalcification property, it is conceivable that Mg prevents CKD progression by inhibiting nephrocalcinosis. In fact, rats fed a high phosphate and low Mg diet developed nephrocalcinosis, which was resolved by increasing dietary Mg intake.30 Similarly, high phosphate and low Mg may also contribute to the development of renal artery calcification, which has been suggested to be involved in the progression of CKD.31 Given the relatively short follow-up period, however, the findings in our study may not be fully explained by the progression of nephrocalcinosis or renal artery calcification. Apart from such clinically overt calcification, a calcium-phosphate crystal is also pathologic because it induces oxidative stress, inflammation, and cell injury in various kinds of cells Kidney International
—
P-value Median (IQR)
0.02 0.7
P-value Median (IQR)
− 1.2 (−3.0, 1.1)
61 Median (IQR)
0.5 − 2.1 (−3.9, − 0.8) − 2.0 (−4.0, − 0.5)
P-value
a
91 289 (3) eGFR slope (ml/min per 1.73 m2 per year) No. of patients
HR (95% CI)
0.61 (0.38–0.97) 1.06 (0.65–1.72) 0.01 0.3
P-value HR (95% CI)
0.56 (0.36–0.87) 1.31 (0.81–2.11) Univariate Adjusted for baseline eGFR
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; EKSD, end-stage kidney disease; HR, hazard ratio; IQR, interquartile range; Mg, magnesium; P, phosphate. a Versus the higher Mg-higher P group (post hoc Scheffe test).
Median (IQR)
64
a
− 3.2 (−6.1, − 1.3)
P-value
a
− 1.7 (−3.0, − 0.5)
73
— —
P-value HR (95% CI)
Ref. Ref. 0.1 0.006
P-value HR (95% CI) P-value
1.39 (0.90–2.14) 1.84 (1.19–2.84)
69 40 69 29 96 35 311 148
Univariate Adjusted for baseline eGFR
(2) Composite outcome (ESKD+death) No. of patients No. of events
HR (95% CI) P-value
0.01 0.2 0.54 (0.34–0.86) 1.35 (0.82–2.22)
HR (95% CI)
0.56 (0.34–0.92) 1.02 (0.61–1.70)
69 25 96 32 311 135
Kidney International
0.04 0.8
77 44
— —
P-value HR (95% CI)
Ref. Ref. 0.2 0.007
P-value HR (95% CI)
1.38 (0.88–2.16) 1.86 (1.19–2.93)
69 37 P-value
0.02 0.9
77 41
Higher Mg-higher P Lower Mg-higher P Higher Mg-lower P Lower Mg-lower P Total
(1) ESKD No. of patients No. of events
Table 2 | Univariate and baseline eGFR-adjusted models for the hazard ratios of (1) progression to end-stage kidney disease and (2) composite outcome (progression to end-stage kidney disease+death), and (3) the median eGFR slope across the magnesium-phosphate groups
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
clinical investigation
including tubular cells.32,33 As Mg is a potent inhibitor of crystallization of calcium phosphate, Mg might exert its beneficial effects on renal prognosis by alleviating such mineral stress. Future detailed studies are needed to elucidate precise mechanisms by which Mg attenuates the toxic effects of phosphate. Because both high phosphate and low Mg are associated with an increased risk of heart failure,34 it can be speculated that patients in the lower Mg-higher phosphate group had a higher risk for dialysis initiation owing to an increased risk of heart failure. In Japanese CKD patients, however, cardiovascular comorbidities are less prevalent. Only 10.4% of the patients who started dialysis in our cohort had refractory congestive heart failure at dialysis initiation. In addition, the mean eGFR at dialysis initiation was very low and not significantly different among the four groups, suggesting that most patients were followed until they reached the ‘renal’ end point. Furthermore, we also evaluated eGFR slope and found that patients in the lower Mg-higher phosphate group showed a significantly faster decline in eGFR. Taken together, it is unlikely that the increased risk for dialysis initiation in the lower Mg-higher phosphate group was attributed to the possible poor cardiovascular prognosis in this group. In contrast to the current study of non-diabetic CKD, we have reported that there were no obvious effect modification between Mg and phosphate on the risk of ESKD in type 2 diabetic nephropathy.35 The reason for this discrepancy is uncertain but may be attributed to the fact that Mg has a distinct role in the pathophysiology of diabetes. For example, Mg is an essential cofactor for many glycolytic enzymes and tyrosine kinase of insulin receptor;36 thus, Mg deficiency facilitates insulin resistance. In fact, Mg supplementation improves insulin sensitivity.37 More importantly, it has been suggested that Mg deficiency promotes the development of diabetic complications by inhibiting intracellular uptake of inositol through the disruption of its cell membrane transport.38 Mg deficiency is actually associated with the development of various complications of type 2 diabetes.39 Given the specific link between Mg and diabetes, we believe that the effect of Mg on diabetic nephropathy is more independent of phosphate status compared with its effect on non-diabetic CKD. Several limitations should be noted in our study. First, we cannot infer causality of the relationship between serum Mg and phosphate levels and renal prognosis owing to the observational study design. Although we adjusted for several potential confounders, the possibility of residual confounding cannot be excluded. However, our in vitro findings showing the protective effects of Mg against the cytotoxicity of phosphate on tubular cells support the notion that Mg can directly attenuate the risk of CKD progression related to hyperphosphatemia. As this study was a post hoc analysis of the previously reported cohort study, our findings should be confirmed by prospective studies. In particular, interventional trials are required to determine whether Mg supplementation improves renal prognosis of hyperphosphatemic CKD 5
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
4 β-Coefficient
2 1.5 1 0.5
2 0 –2
w
P M gH ig h
M g-
hi gh
P hi gh
lo w M g-
Lo
Lo
H ig h
w
M g-
M g-
lo w
P
P
P hi gh
P hi gh Lo
w
M g-
M gH ig h
Lo
w
M g-
lo w
lo w
P
P
–4
H ig h
Adjusted hazard ratio
3 2.5
Figure 2 | The risk of progression of chronic kidney disease across the magnesium-phosphate groups. (a) Hazard ratios of progression to end-stage kidney disease and (b) β-coefficient for the eGFR slope (ml/min per 1.73 m2 per year). Models were adjusted for (a) baseline estimated glomerular filtration rate and (b) age, sex, body mass index, baseline estimated glomerular filtration rate, comorbidities (hypertension, prior history of cardiovascular disease), medications (use of renin–angiotensin aldosterone blockers, diuretics, statins, calcium carbonate, and magnesium oxide), and laboratory data (log-transformed urine protein, corrected calcium, albumin, and total cholesterol). The error bars represent 95% confidence intervals. eGFR, estimated glomerular filtration rate.
Table 3 | Linear regression analysis between eGFR slope and magnesium-phosphate groups Total, n = 289
Lower Mg-lower P, n = 91 β-Coefficient (95% CI)
Univariate Model 1 Model 2 Model 3 Model 4
− 1.18 (−2.75, − 0.91 (−2.60, − 0.93 (−2.66, − 0.32 (−1.98, − 0.75 (−2.41,
0.40) 0.78) 0.80) 1.34) 0.92)
P-value 0.1 0.3 0.3 0.7 0.4
Higher Mg-lower P, n = 61 β-Coefficient (95% CI) 1.10 1.26 1.58 1.49 1.30
(−0.61, (−0.50, (−0.26, (−0.26, (−0.44,
2.82) 3.02) 3.42) 3.23) 3.05)
P-value 0.2 0.2 0.09 0.1 0.1
Lower Mg-higher P, n = 64 β-Coefficient (95% CI) − 2.79 − 2.12 − 2.24 − 1.85 − 2.07
(−4.53, (−3.89, (−4.03, (−3.56, (−3.85,
− 1.05) − 0.36) − 0.45) − 0.14) −0.29)
Higher Mg-higher P, n = 73
P-value
β-Coefficient (95% CI)
P-value
0.002 0.02 0.01 0.03 0.02
Ref. Ref. Ref. Ref. Ref.
— — — — —
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; Mg, magnesium; P, phosphate. Model 1: Adjusted for age, sex, body mass index, eGFR, and comorbidities (hypertension, prior history of cardiovascular disease). Model 2: Adjusted for covariates in Model 1+medications (use of renin–angiotensin aldosterone blockers, diuretics, statins, calcium carbonate, and magnesium oxide). Model 3: Adjusted for covariates in Model 2+laboratory data (log-transformed urine protein, corrected calcium, albumin, and total cholesterol). Model 4: Adjusted for covariates in Model 3+log-transformed intact parathyroid hormone level (missing values were imputed by multiple imputation method).
patients. Second, some laboratory data were collected retrospectively and had a large proportion of missing values (i.e., intact PTH and FGF23) even though these data were imputed by the multiple imputation method. Third, the sample size was relatively small. Although serum Mg and phosphate levels were dichotomized and patients were categorized into four Mg-phosphate groups, further detailed categorization makes the groups too small to obtain robust results. Larger-scale studies are needed to determine a dose–response relationship between Mg and outcomes. Finally, our patients were participants in a CKD education program and this may have created a selection bias toward individuals who were more compliant with therapy. In particular, serum phosphate level was relatively low because these patients were educated on dietary protein restriction. In such a patient group, the impact of phosphate on CKD progression may become obscure. Our findings should be confirmed in patients with higher phosphate levels. In conclusion, our study showed that low serum Mg significantly enhanced the risk of incident ESKD associated 6
with high serum phosphate in non-diabetic CKD patients. In vitro experiments found that Mg suppressed apoptosis and expression of profibrotic and proinflammatory cytokines of tubular cells induced by high phosphate. These findings suggest that Mg is a protective mineral against phosphate toxicity in the kidney. Interventional studies are warranted to elucidate effects of Mg supplementation on renal prognosis in CKD patients with hyperphosphatemia. MATERIALS AND METHODS Cohort study Study design. This was a post hoc analysis of a previously published cohort study.35 The detailed methods were described elsewhere.35 Briefly, 521 non-dialysis CKD stage 3–5 outpatients of the nephrology unit at Osaka General Medical Center, who were participants of a CKD educational program, were enrolled in the study between April 2001 and December 2007. This program included CKD patients in stable general condition and mainly aimed to provide nutritional counseling by a dietitian. Patients who had uremic symptoms were not targeted for this program. Patients were followed from the day of participation in the program to either the Kidney International
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
0.9
0.9
2.0
Mg (mmol/l) 0.75
1.5
0.75
Propidium iodide
P (mmol/l)
2.0 1.5
105
105
105
105
104
104
104
104
103
103
103
103
102
102 102
103
104
102
102
105
102
103
104
105
102
103
104
105
102
103
104
105
Annexin V
Apoptotic cells (%)
20
*
*
15
β-Actin NS
10
Cleaved caspase-3
5
0 P (mmol/l)
1 0.9
2 0.9
2.0 3
4 2.0
Mg (mmol/l) 0.75
1.5
0.75
1.5
Interaction P-value
clinical investigation
© 2015 International Society of Nephrology
Magnesium modifies the association between serum phosphate and the risk of progression to end-stage kidney disease in patients with non-diabetic chronic kidney disease Yusuke Sakaguchi1, Hirotsugu Iwatani1, Takayuki Hamano2, Kodo Tomida3, Hiroaki Kawabata1, Yasuo Kusunoki1, Akihiro Shimomura1, Isao Matsui1, Terumasa Hayashi3, Yoshiharu Tsubakihara2, Yoshitaka Isaka1 and Hiromi Rakugi1 1
Department of Geriatric Medicine and Nephrology, Osaka University Graduate School of Medicine, Suita, Osaka, Japan; 2Department of Comprehensive Kidney Disease Research, Osaka University Graduate School of Medicine, Suita, Osaka, Japan and 3Department of Kidney Disease and Hypertension, Osaka General Medical Center, Osaka, Japan
It is known that magnesium antagonizes phosphate-induced apoptosis of vascular smooth muscle cells and prevents vascular calcification. Here we tested whether magnesium can also counteract other pathological conditions where phosphate toxicity is involved, such as progression of chronic kidney disease (CKD). We explored how the link between the risk of CKD progression and hyperphosphatemia is modified by magnesium status. A post hoc analysis was run in 311 nondiabetic CKD patients who were divided into four groups according to the median values of serum magnesium and phosphate. During a median follow-up of 44 months, 135 patients developed end-stage kidney disease (ESKD). After adjustment for relevant clinical factors, patients in the lower magnesium-higher phosphate group were at a 2.07-fold (95% CI: 1.23–3.48) risk for incident ESKD and had a significantly faster decline in estimated glomerular filtration rate compared with those in the higher magnesium-higher phosphate group. There were no significant differences in the risk of these renal outcomes among the higher magnesiumhigher phosphate group and both lower phosphate groups. Incubation of tubular epithelial cells in high phosphate and low magnesium medium in vitro increased apoptosis and the expression levels of profibrotic and proinflammatory cytokine; these changes were significantly suppressed by increasing magnesium concentration. Thus, magnesium may act protectively against phosphate-induced kidney injury. Kidney International advance online publication, 10 June 2015; doi:10.1038/ki.2015.165 KEYWORDS: chronic kidney disease; hyperphosphatemia; magnesium
Correspondence: Yoshitaka Isaka, Department of Geriatric Medicine and Nephrology, Osaka University Graduate School of Medicine, 2-2, Yamadaoka, Suita, Osaka 565-0871, Japan. E-mail: [email protected] Received 11 August 2014; revised 29 March 2015; accepted 9 April 2015 Kidney International
Hyperphosphatemia is one of the major complications of chronic kidney disease (CKD). It is associated not only with an increased risk of cardiovascular mortality1 but also with a faster progression of CKD.2–11 Animal studies showed the benefit of dietary phosphate restriction, or use of non-calcium-containing phosphate binders, for attenuation of deteriorating kidney function.12–15 Moreover, klotho-knockout mice presenting hyperphosphatemia exhibited kidney failure and increased apoptotic cells in the kidney;16 after genetically reducing serum phosphate levels by an additional knockout of the sodium phosphate cotransporter gene, kidney function restored with decreased apoptotic cells. Importantly, the kidney function in these double-knockout mice again worsened when hyperphosphatemia was induced by a high-phosphate diet. These findings suggest a clinical significance of attenuating phosphate toxicity to the kidney to improve renal prognosis of CKD patients. Magnesium (Mg), the fourth most abundant cation in the human body, is an essential mineral for numerous biological processes. Mg deficiency is associated with an increased risk of cardiovascular events and vascular calcification, both in individuals with and without CKD.17–23 Notably, an in vitro study has shown that Mg strongly inhibits phosphate-induced apoptosis of vascular smooth muscle cells,24 a key process of vascular calcification. This finding prompted us to postulate that Mg could also counteract other pathological conditions where phosphate toxicity is involved, such as progression of CKD. Here, we examined how Mg status influences the relationship between hyperphosphatemia and the risk of CKD progression. In addition, in vitro experiments were performed to investigate whether Mg acts protectively against the potential cytotoxicity of phosphate on tubular epithelial cells. 1
clinical investigation
RESULTS Cohort study
Of the total of 367 non-diabetic CKD patients who had participated in the CKD educational program, 56 were excluded from the analysis (7 were followed up for o3 months; 41 had missing data on serum Mg or phosphate; and the medical records of 8 patients had been discarded). Baseline characteristics were not substantially different between the participants and those enrolled in the current study (Supplementary Table S1 online). Baseline characteristics according to Mg-phosphate groups (categorized by the median value of serum Mg and phosphate levels) are summarized in Table 1. Higher phosphate groups tended to have a lower estimated glomerular filtration rate (eGFR) and higher intact parathyroid hormone (PTH) levels, whereas lower Mg groups tended to have a slightly higher urine protein and lower serum albumin levels. During the median follow-up of 44 months, 135 patients progressed to end-stage kidney disease (ESKD) and 13 died. The mean eGFR at dialysis initiation was 6.6 ml/min per 1.73 m2 and was not significantly different among the four Mg-phosphate groups (Supplementary Table S2 online). Only 14 patients had refractory congestive heart failure at dialysis initiation, and these cases were equally distributed across the Mg-phosphate groups (Supplementary Table S2 online). The Kaplan–Meier curves (Figure 1a) revealed that the risk of incident ESKD increased significantly in the higher phosphate groups, with the lower Mg-higher phosphate group having the worst prognosis. It should be noted, however, that baseline eGFR was significantly different among the four groups. Kaplan–Meier curves that were adjusted for baseline eGFR (Figure 1b) showed that the lower Mg-higher phosphate group still had the worst prognosis, but that the risks of the other three groups were almost equivalent. In the Cox models, serum Mg and phosphate were treated as categorical variables because of the following reasons: (1) a nonlinear association between serum Mg and phosphate levels and the natural logarithm of the hazard function, and (2) the violation of the proportional hazard assumption for serum phosphate. The lower Mg-higher phosphate group showed the highest risk of progression to ESKD among the four Mgphosphate groups after adjustment for baseline eGFR (Table 2 and Figure 2a). Similarly, eGFR slope of the lower Mg-higher phosphate group was significantly steeper than those of the other three groups and was approximately two times as fast as that of the higher Mg-higher phosphate group (Table 2). These findings were maintained in multivariate Cox models, which revealed that patients in the lower Mg-higher phosphate group consistently showed the highest risk of progression to ESKD among the four Mg-phosphate groups; patients in this group had a 2.07-fold higher risk compared with those in the higher Mg-higher phosphate group in the fully adjusted model (95% confidence interval: 1.23–3.48; Supplementary Table S3 online). After further adjustment for intact PTH and fibroblast growth factor 23 (FGF23) by using a multiple imputation method, patients in the lower Mg-higher phosphate group still had the highest risk (Supplementary Tables S3 and S4 online). 2
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
No significant risk difference was found between the lower Mglower phosphate and the higher Mg-lower phosphate group. Of note, there were also no significant risk differences among the higher Mg-higher phosphate group and both lower phosphate groups. Similar results were obtained in the Cox models where ESKD and death were treated as a composite outcome (Supplementary Table S3 online). In multivariate linear regression models, the eGFR slope of the lower Mg-higher phosphate group was again steeper compared with those of the other three groups (Table 3 and Figure 2b). The slopes were not significantly different among the higher Mg-higher phosphate group and both the lower phosphate groups. Further adjustment for intact PTH and FGF23 yielded similar results (Table 3 and Supplementary Table S4 online). In vitro experiments Effects of Mg and phosphate on apoptosis in mProx cells.
As phosphate is known to induce apoptosis in the kidney cells,16 we examined effects of Mg against the proapoptotic property of phosphate. Incubation of mProx cells, a cell line of mouse proximal tubular cells, with high phosphate (2.0 mmol/l) and low Mg (0.75 mmol/l) increased apoptotic cells compared with physiological phosphate (0.9 mmol/l) and 0.75 mmol/l Mg; however, the high phosphate-induced apoptosis was significantly suppressed by increasing Mg concentration to 1.5 mmol/l (Figure 3a and b). In contrast, Mg did not influence tubular cell apoptosis in the medium with 0.9 mmol/l phosphate. The interaction between the Mg and phosphate concentrations on the proportion of apoptotic cells was significant (Po0.01; two-way analysis of variance (ANOVA)), indicating that the effect of Mg was significantly different between the high and low phosphate conditions. Similarly, the expression of cleaved caspase-3, a major executor of apoptosis, was enhanced in cells treated with 2.0 mmol/l phosphate and 0.75 mmol/l Mg, which was significantly inhibited by increasing Mg level to 1.5 mmol/l (Figure 3c and d). Mg inhibits phosphate-induced depolarization of mitochondrial membrane potential. To examine whether the apoptosis
caused by high phosphate and low Mg was dependent on the mitochondrial pathway, we evaluated mitochondrial membrane potential by tetramethylrhodamine ethyl ester staining. The mitochondrial membrane potential was significantly reduced in cells treated with 2.0 mmol/l phosphate and 0.75 mmol/l Mg compared with those treated with 0.9 mmol/l phosphate and 0.75 mmol/l Mg (Figure 4a and b); the membrane potential was restored by increasing Mg concentration to 1.5 mmol/l. The effect of Mg on the membrane potential was not apparent in the medium with 0.9 mmol/l phosphate. The interaction between the Mg and phosphate concentrations on the membrane potential was significant (Po0.01; two-way ANOVA). Mg suppresses gene expressions of profibrotic and proinflammatory cytokines induced by high phosphate. Tubulointerstitial
fibrosis is the hallmark of CKD progression and is characterized by excessive deposition of extracellular matrix proteins secreted Kidney International
Kidney International
(32) (24) (48) (26)
65 21 27 4 6
(68) (22) (28) (4) (6)
77 (80) 31 (32)
48 21 19 8
(19) (27) (25) (19)
36 14 19 2 12
(52) (20) (28) (3) (17)
55 (80) 25 (36)
29 24 10 6
64.7 (12.2) 46 (67) 23.4 (4.0) 2.11 (0.88) 27.3 (10.0) 9.2 (0.7) 2.3 (0.2) 3.1 (0.4) 73 (41, 119) 3.9 (0.6) 189 (38) 0.4 (0.1, 1.0)
42.1 ⩽ 3.6
⩽ 2.1 ⩽ 3.6
64.4 (10.3) 70 (73) 23.1 (3.3) 1.94 (0.73) 30.0 (10.7) 9.1 (0.6) 1.9 (0.2) 3.1 (0.4) 62 (42, 106) 3.7 (0.5) 189 (47) 0.5 (0.1, 1.3)
Higher Mg-lower P, n = 69
Lower Mg-lower P, n = 96
(24) (21) (15) (26)
39 (57) 20 (29) 18 (26) 5 (7) 3 (4)
63 (91) 16 (23)
37 18 6 8
64.0 (13.2) 42 (61) 24.4 (4.2) 2.70 (1.17) 21.9 (10.4) 9.1 (0.6) 2.0 (0.2) 4.3 (0.7) 154 (53, 210) 3.7 (0.5) 199 (52) 1.0 (0.1, 2.3)
⩽ 2.1 43.6
Lower Mg-higher P, n = 69
(25) (28) (13) (29)
38 (49) 28 (36) 29 (38) 17 (22) 4 (5)
61 (79) 22 (29)
38 25 5 9
66.5 (12.5) 29 (38) 23.4 (4.1) 2.94 (1.30) 18.4 (8.9) 9.0 (0.5) 2.5 (0.4) 4.4 (0.7) 113 (69, 183) 3.8 (0.4) 205 (51) 0.4 (0.1, 1.8)
42.1 43.6
Higher Mg-higher P, n = 77
0.07 0.09 0.4 o0.001 0.01
0.2 0.4
0.2
0.6 o0.001 0.2 o0.001 o0.001 0.4 o0.001 o0.001 o0.001 0.03 0.09 0.02
P-value
Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; Mg, magnesium; P, phosphate; PTH, parathyroid hormone; RAS, renin–angiotensin– aldosterone system. Data are presented as mean (s.d.) or median (interquartile range) or number (%).
(57) (27) (30) (9) (8)
0 0 0 0 0
Prescriptions RAS inhibitor, n (%) Diuretics, n (%) Statins, n (%) Calcium carbonate, n (%) Magnesium oxide, n (%) 178 83 93 28 25
256 (82) 94 (30)
0 0
Comorbidities Hypertension, n (%) Prior history of CVD, n (%)
(100) (100) (100) (100)
64.9 (11.9) 187 (60) 23.5 (3.9) 2.39 (1.10) 24.7 (11.0) 9.1 (0.6) 2.2 (0.3) 3.7 (0.9) 90.5 (51, 156) 3.8 (0.5) 195 (48) 0.5 (0.1, 1.6)
152 88 40 31
0
0 0 3 (1) 0 0 0 0 0 145 (46.6) 0 2 (1) 5 (2)
Total, n = 311
Primary causes of CKD Chronic glomerulonephritis, n (%) Benign nephrosclerosis, n (%) Others, n (%) Unknown, n (%)
Characteristics Age (year) Male, n (%) BMI (kg/m2) Creatinine (mg/dl) eGFR (ml/min per 1.73 m2) Corrected calcium (mg/dl) Magnesium (mg/dl) Phosphate (mg/dl) Intact PTH (pg/ml) Albumin (g/dl) Total cholesterol (mg/dl) Urine protein (g per day)
Serum Mg range (mg/dl) Serum P range (mg/dl)
No. of missing values (%)
Table 1 | Baseline characteristics of 311 non-diabetic chronic kidney disease patients according to Mg-phosphate groups
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
clinical investigation
3
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
1.00 Cumulative event-free survival
Cumulative event-free survival
1.00
0.75
0.50
0.25
Higher Mg-lower P Lower Mg-lower P Higher Mg-higher P Lower Mg-higher P
0.00 0
12
96 69 Higher Mg-lower P 69 Higher Mg-higher P 77 Lower Mg-higher P
36
48
93 58 65 62
81 33 59 53
69 28 49 42
52 18 39 33
0.75
0.50
0.25
0.00 60
72
84
35 10 31 25
0
12
24
36
48
60
72
84
Follow-up time (months)
Follow-up time (months)
Number at risk Lower Mg-lower P
24
Higher Mg-lower P Lower Mg-lower P Higher Mg-higher P Lower Mg-higher P
19 5 21 17
11 4 10 12
Figure 1 | Kaplan–Meier curves for the risk of progression to end-stage kidney disease in 311 non-diabetic chronic kidney disease patients stratified by magnesium-phosphate groups. (a) Unadjusted Kaplan–Meier curves (Po0.001 by log-rank test) and (b) Kaplan–Meier curves adjusted for the estimated glomerular filtration rate (scaled to 15 ml/min per 1.73 m2). The curves were truncated when the percentage of patients remaining at risk was ~ 10%.
mainly from myofibroblasts. Because tubular cells contribute to the development of fibrosis by producing several profibrotic and proinflammatory cytokines in response to various injurious stimuli,25 we investigated whether the expression levels of these cytokines are altered by medium concentrations of phosphate and Mg. The incubation of mProx cells with 2.0 mmol/l phosphate and 0.75 mmol/l Mg increased mRNA expression levels for transforming growth factor-β1 and interleukin-6 compared with incubation with 0.9 mmol/l phosphate and 0.75 mmol/l Mg (Figure 5). Increasing the Mg level to 1.5 mmol/l significantly suppressed the expressions of these genes. Mg did not alter the mRNA expression levels of these genes in the medium with 0.9 mmol/l phosphate. There was significant interaction between Mg and phosphate concentrations on the mRNA expressions of transforming growth factor-β1 and interleukin-6 (Po0.01; two-way ANOVA). DISCUSSION
Previous studies have reported, albeit inconsistently, a significant association between hyperphosphatemia and faster progression of CKD.3–11 We expanded these findings by showing that the risk of ESKD related to high serum phosphate was modified by Mg status; patients with high phosphate had a higher risk of ESKD when they had a low serum Mg level. This finding is in keeping with our hypothesis that Mg acts protectively against phosphateinduced kidney injury. In fact, our in vitro experiments demonstrated that Mg suppressed apoptosis and the production of profibrotic and proinflammatory cytokines of tubular cells induced by high phosphate. Several mechanisms underlying our observation can be assumed. Mitochondrial permeability transition (PT) is one of the key processes in mitochondria-mediated cell death.26,27 PT is caused by opening of the mitochondrial PT pore (mPTP), which makes the mitochondrial inner membrane 4
unselectively permeable to solutes of molecular weight ⩽ 1.5 kDa. As a result, mitochondrial matrix swelling and rupture of the outer membrane occur, leading to activation of mitochondrial apoptotic pathway.28 Of note, Mg and phosphate directly and oppositely act on the mPTP and induce its conformational change; phosphate promotes the mPTP opening, whereas Mg inhibits it.26,27 Therefore, Mg suppresses phosphate-induced apoptosis of tubular cells, possibly by competing the mPTP opening facilitated by phosphate. Although we did not directly evaluate the state of the mPTP opening, we found that Mg restored the mitochondrial membrane potential of tubular cells collapsed owing to high phosphate concentration. There may be another explanation for the protective role of Mg against phosphate toxicity. It is known that nephrocalcinosis is a common pathological change observed in the kidney exposed to high phosphate.12 Serum phosphate is positively associated with calcium content in human kidney biopsy samples.29 In CKD animal models, a low-phosphate diet, or use of non-calcium-containing phosphate binders, was useful in reducing renal calcium content and improving renal histology.12,14,15 As Mg has an anticalcification property, it is conceivable that Mg prevents CKD progression by inhibiting nephrocalcinosis. In fact, rats fed a high phosphate and low Mg diet developed nephrocalcinosis, which was resolved by increasing dietary Mg intake.30 Similarly, high phosphate and low Mg may also contribute to the development of renal artery calcification, which has been suggested to be involved in the progression of CKD.31 Given the relatively short follow-up period, however, the findings in our study may not be fully explained by the progression of nephrocalcinosis or renal artery calcification. Apart from such clinically overt calcification, a calcium-phosphate crystal is also pathologic because it induces oxidative stress, inflammation, and cell injury in various kinds of cells Kidney International
—
P-value Median (IQR)
0.02 0.7
P-value Median (IQR)
− 1.2 (−3.0, 1.1)
61 Median (IQR)
0.5 − 2.1 (−3.9, − 0.8) − 2.0 (−4.0, − 0.5)
P-value
a
91 289 (3) eGFR slope (ml/min per 1.73 m2 per year) No. of patients
HR (95% CI)
0.61 (0.38–0.97) 1.06 (0.65–1.72) 0.01 0.3
P-value HR (95% CI)
0.56 (0.36–0.87) 1.31 (0.81–2.11) Univariate Adjusted for baseline eGFR
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; EKSD, end-stage kidney disease; HR, hazard ratio; IQR, interquartile range; Mg, magnesium; P, phosphate. a Versus the higher Mg-higher P group (post hoc Scheffe test).
Median (IQR)
64
a
− 3.2 (−6.1, − 1.3)
P-value
a
− 1.7 (−3.0, − 0.5)
73
— —
P-value HR (95% CI)
Ref. Ref. 0.1 0.006
P-value HR (95% CI) P-value
1.39 (0.90–2.14) 1.84 (1.19–2.84)
69 40 69 29 96 35 311 148
Univariate Adjusted for baseline eGFR
(2) Composite outcome (ESKD+death) No. of patients No. of events
HR (95% CI) P-value
0.01 0.2 0.54 (0.34–0.86) 1.35 (0.82–2.22)
HR (95% CI)
0.56 (0.34–0.92) 1.02 (0.61–1.70)
69 25 96 32 311 135
Kidney International
0.04 0.8
77 44
— —
P-value HR (95% CI)
Ref. Ref. 0.2 0.007
P-value HR (95% CI)
1.38 (0.88–2.16) 1.86 (1.19–2.93)
69 37 P-value
0.02 0.9
77 41
Higher Mg-higher P Lower Mg-higher P Higher Mg-lower P Lower Mg-lower P Total
(1) ESKD No. of patients No. of events
Table 2 | Univariate and baseline eGFR-adjusted models for the hazard ratios of (1) progression to end-stage kidney disease and (2) composite outcome (progression to end-stage kidney disease+death), and (3) the median eGFR slope across the magnesium-phosphate groups
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
clinical investigation
including tubular cells.32,33 As Mg is a potent inhibitor of crystallization of calcium phosphate, Mg might exert its beneficial effects on renal prognosis by alleviating such mineral stress. Future detailed studies are needed to elucidate precise mechanisms by which Mg attenuates the toxic effects of phosphate. Because both high phosphate and low Mg are associated with an increased risk of heart failure,34 it can be speculated that patients in the lower Mg-higher phosphate group had a higher risk for dialysis initiation owing to an increased risk of heart failure. In Japanese CKD patients, however, cardiovascular comorbidities are less prevalent. Only 10.4% of the patients who started dialysis in our cohort had refractory congestive heart failure at dialysis initiation. In addition, the mean eGFR at dialysis initiation was very low and not significantly different among the four groups, suggesting that most patients were followed until they reached the ‘renal’ end point. Furthermore, we also evaluated eGFR slope and found that patients in the lower Mg-higher phosphate group showed a significantly faster decline in eGFR. Taken together, it is unlikely that the increased risk for dialysis initiation in the lower Mg-higher phosphate group was attributed to the possible poor cardiovascular prognosis in this group. In contrast to the current study of non-diabetic CKD, we have reported that there were no obvious effect modification between Mg and phosphate on the risk of ESKD in type 2 diabetic nephropathy.35 The reason for this discrepancy is uncertain but may be attributed to the fact that Mg has a distinct role in the pathophysiology of diabetes. For example, Mg is an essential cofactor for many glycolytic enzymes and tyrosine kinase of insulin receptor;36 thus, Mg deficiency facilitates insulin resistance. In fact, Mg supplementation improves insulin sensitivity.37 More importantly, it has been suggested that Mg deficiency promotes the development of diabetic complications by inhibiting intracellular uptake of inositol through the disruption of its cell membrane transport.38 Mg deficiency is actually associated with the development of various complications of type 2 diabetes.39 Given the specific link between Mg and diabetes, we believe that the effect of Mg on diabetic nephropathy is more independent of phosphate status compared with its effect on non-diabetic CKD. Several limitations should be noted in our study. First, we cannot infer causality of the relationship between serum Mg and phosphate levels and renal prognosis owing to the observational study design. Although we adjusted for several potential confounders, the possibility of residual confounding cannot be excluded. However, our in vitro findings showing the protective effects of Mg against the cytotoxicity of phosphate on tubular cells support the notion that Mg can directly attenuate the risk of CKD progression related to hyperphosphatemia. As this study was a post hoc analysis of the previously reported cohort study, our findings should be confirmed by prospective studies. In particular, interventional trials are required to determine whether Mg supplementation improves renal prognosis of hyperphosphatemic CKD 5
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
4 β-Coefficient
2 1.5 1 0.5
2 0 –2
w
P M gH ig h
M g-
hi gh
P hi gh
lo w M g-
Lo
Lo
H ig h
w
M g-
M g-
lo w
P
P
P hi gh
P hi gh Lo
w
M g-
M gH ig h
Lo
w
M g-
lo w
lo w
P
P
–4
H ig h
Adjusted hazard ratio
3 2.5
Figure 2 | The risk of progression of chronic kidney disease across the magnesium-phosphate groups. (a) Hazard ratios of progression to end-stage kidney disease and (b) β-coefficient for the eGFR slope (ml/min per 1.73 m2 per year). Models were adjusted for (a) baseline estimated glomerular filtration rate and (b) age, sex, body mass index, baseline estimated glomerular filtration rate, comorbidities (hypertension, prior history of cardiovascular disease), medications (use of renin–angiotensin aldosterone blockers, diuretics, statins, calcium carbonate, and magnesium oxide), and laboratory data (log-transformed urine protein, corrected calcium, albumin, and total cholesterol). The error bars represent 95% confidence intervals. eGFR, estimated glomerular filtration rate.
Table 3 | Linear regression analysis between eGFR slope and magnesium-phosphate groups Total, n = 289
Lower Mg-lower P, n = 91 β-Coefficient (95% CI)
Univariate Model 1 Model 2 Model 3 Model 4
− 1.18 (−2.75, − 0.91 (−2.60, − 0.93 (−2.66, − 0.32 (−1.98, − 0.75 (−2.41,
0.40) 0.78) 0.80) 1.34) 0.92)
P-value 0.1 0.3 0.3 0.7 0.4
Higher Mg-lower P, n = 61 β-Coefficient (95% CI) 1.10 1.26 1.58 1.49 1.30
(−0.61, (−0.50, (−0.26, (−0.26, (−0.44,
2.82) 3.02) 3.42) 3.23) 3.05)
P-value 0.2 0.2 0.09 0.1 0.1
Lower Mg-higher P, n = 64 β-Coefficient (95% CI) − 2.79 − 2.12 − 2.24 − 1.85 − 2.07
(−4.53, (−3.89, (−4.03, (−3.56, (−3.85,
− 1.05) − 0.36) − 0.45) − 0.14) −0.29)
Higher Mg-higher P, n = 73
P-value
β-Coefficient (95% CI)
P-value
0.002 0.02 0.01 0.03 0.02
Ref. Ref. Ref. Ref. Ref.
— — — — —
Abbreviations: CI, confidence interval; eGFR, estimated glomerular filtration rate; Mg, magnesium; P, phosphate. Model 1: Adjusted for age, sex, body mass index, eGFR, and comorbidities (hypertension, prior history of cardiovascular disease). Model 2: Adjusted for covariates in Model 1+medications (use of renin–angiotensin aldosterone blockers, diuretics, statins, calcium carbonate, and magnesium oxide). Model 3: Adjusted for covariates in Model 2+laboratory data (log-transformed urine protein, corrected calcium, albumin, and total cholesterol). Model 4: Adjusted for covariates in Model 3+log-transformed intact parathyroid hormone level (missing values were imputed by multiple imputation method).
patients. Second, some laboratory data were collected retrospectively and had a large proportion of missing values (i.e., intact PTH and FGF23) even though these data were imputed by the multiple imputation method. Third, the sample size was relatively small. Although serum Mg and phosphate levels were dichotomized and patients were categorized into four Mg-phosphate groups, further detailed categorization makes the groups too small to obtain robust results. Larger-scale studies are needed to determine a dose–response relationship between Mg and outcomes. Finally, our patients were participants in a CKD education program and this may have created a selection bias toward individuals who were more compliant with therapy. In particular, serum phosphate level was relatively low because these patients were educated on dietary protein restriction. In such a patient group, the impact of phosphate on CKD progression may become obscure. Our findings should be confirmed in patients with higher phosphate levels. In conclusion, our study showed that low serum Mg significantly enhanced the risk of incident ESKD associated 6
with high serum phosphate in non-diabetic CKD patients. In vitro experiments found that Mg suppressed apoptosis and expression of profibrotic and proinflammatory cytokines of tubular cells induced by high phosphate. These findings suggest that Mg is a protective mineral against phosphate toxicity in the kidney. Interventional studies are warranted to elucidate effects of Mg supplementation on renal prognosis in CKD patients with hyperphosphatemia. MATERIALS AND METHODS Cohort study Study design. This was a post hoc analysis of a previously published cohort study.35 The detailed methods were described elsewhere.35 Briefly, 521 non-dialysis CKD stage 3–5 outpatients of the nephrology unit at Osaka General Medical Center, who were participants of a CKD educational program, were enrolled in the study between April 2001 and December 2007. This program included CKD patients in stable general condition and mainly aimed to provide nutritional counseling by a dietitian. Patients who had uremic symptoms were not targeted for this program. Patients were followed from the day of participation in the program to either the Kidney International
clinical investigation
Y Sakaguchi et al.: Role of Mg in hyperphosphatemic CKD patients
0.9
0.9
2.0
Mg (mmol/l) 0.75
1.5
0.75
Propidium iodide
P (mmol/l)
2.0 1.5
105
105
105
105
104
104
104
104
103
103
103
103
102
102 102
103
104
102
102
105
102
103
104
105
102
103
104
105
102
103
104
105
Annexin V
Apoptotic cells (%)
20
*
*
15
β-Actin NS
10
Cleaved caspase-3
5
0 P (mmol/l)
1 0.9
2 0.9
2.0 3
4 2.0
Mg (mmol/l) 0.75
1.5
0.75
1.5
Interaction P-value
