This is a retrospective study to assess survival in a cohort of prevalent HD patients from 2014, followed for 2 years until the end of 2016. Mortality during follow-up has been evaluated in relation to epidemiological parameters, comorbidities, hemodialysis procedure with special emphasis on the type of dialysis fluid and, laboratory parameters with special emphasis on Mg.

Inclusion criteria: prevalent HD patients from the Infanta Leonor University Hospital Unit during 2014. Patients were on Chronic HD, over 18 years of age and signed an informed consent to allow the use of their clinical and laboratory information.

Exclusion criteria: patients with follow-up of less than 3 months or with less than 3 sessions of HD per week with significant residual renal function.

The dialyzers used wereAK200us® and Artis® (Baxter) and ST5008 (Fresenius, FMC, Bad Homburg, Germany). All machines have ionic dialysance (Diascan® or OCM®). The machines were connected to the TSS® computer application, automatically uploading the data for each session. Ultra-pure Dialysis fluid was used with levels of colony-forming units/ml below 0.1 and endotoxin levels below 0.03 EU/ml (SEN Guides). Patients were dialyzed 3 days per week for minimum of 12 hours a week, except those with significant residual renal function. There were high permeability dialyzers with synthetic membranes and a surface between 1.8 and 2.1 m2. Patients with adverse reactions to these membranes were dialyzed with cellulose triacetate. The blood flow was the maximum allowed by vascular access without a drop of arterial line pressure below

-220 mmHg. Dialysis techniques were: online hemodiafiltration (OL-HDF) and high-flow HD (HF-HD).

During follow-up, 5 formulas of dialysis fluid (LD) were used, which have been grouped into three types: type 1 (Mg of 0.5 mmol/ L) and type 3 (Mg 0.37 mmol/L) both with acetate and type 2 (Mg 0.5 mmol/L) with citrate. The full composition is specified in the Table 1.

The following parameters were analyzed: age, sex, underlying disease, comorbidity, dialysis technique HDF-OL/ HD-HF and Kt.

Biochemical parameters: sodium (Na), potassium (K), magnesium (Mg), total calcium (Ca t) and ionized Calcium (Ca++), phosphorus (P) and parathyroid hormone (PTH). The pH was measured by potentiometry, the pCO2 by Sveringhaus electrode, the pO2 by amperometry and the Ca++ by ion selective electrode (ISE). Biochemical determinations were obtained with an autoanalyzer (ADVIA® 2400 Chemistry System, Bayer). PTH determinations were made by chemiluminescence, using the ADVIA CENTAUR system from Bayer.

The serum concentration of Mg was was measured every 6 months. The initial Mg and the mean of the determinations in each patient were used for analysis.

Information was collected on whether or not the patient was on proton pump inhibitors (PPIs).

The patients were followed until the end of 2016. Information recorded included whether the patients was still active, had received a transplant, was lost to follow-up or died.

Variables with normal distribution are expressed as the mean and the standard deviation. The only variable with a non-normal distribution was the C-reactive protein (CRP), Kolmogorov-Smirnov Z, and with Log transformation it was possible to transform it into normal distribution.

For statistical analyzes, the serum Mg concentration was considered the first predialysis serum Mg concentration determined in each patient upon inclusion in the study. The mean of the Mg determinations in each patient was also evaluated.

The variable predialysis serum Mg concentration has been categorized as low and high, using the mean value of the population, 2.1 mg/dl, as the cut-off point.

The differences of the analytical variables according to the types of dialysis fluid was analyzed by one-step analysis of variance after performing the Levene test.

Test of Pearson was used to analyze bivariate correlations between variables with normal distribution.

Kaplan–Meier survival curves have been calculated and the factors were compared using the log rank test (Mantel-Cox). Stepwise Cox regression has been used, to elucidate the variables that independently influence mortality.

A p < 0.05 has been considered significant. SPSS 15.0 (Chicago. Illinois. USA) has been used for statistical analysis.

The present study shows that hemodialysis patients with serum Mg concentrations greater than mean of the population have a better survival than patients with values of Mg below the mean. Similar result has been observed in previous studies.16–22 In these studies, patients separated into two or three categories were compared according to the Mg concentrations. The patients with the best prognosis were those with high serum Mg concentration. In the work by Ishimura et al16 the cut-off point for Mg concentration favoring survival was 2.77 mg/dl of Mg. In another study in non-diabetic patients, the cut-off point was 2.5 mg/dl,20 which is higher than the value of 2.1 mg/dl obtained in our study. In the present study, the mean serum Mg concentration is lower than in others, probably due to the use in some patients of dialysis fluid with low Mg or with citrate. After the analysis of our results, our patients have not been dialyzed with a low Mg concentration in the dialysis fluid and higher Mg has been requested for some patients with acetate and for all patients on citrate.

The better survival of patients with high Mg in HD has been associated with a better nutritional status, with higher concentrations of albumin, triglycerides, phosphorus and lower values of CRP19 and with a lower degree of inflammation.18,22,31 Yu et al.22 observed that serum concentrations of albumin, urea, creatinine and uric acid were higher in patients with hypermagnesemia and the CRP and Lipoprotein A were lower.

In the work of Mizuiri et al31 the independent predictive value of Mg concentration on mortality is lost if albumin is included in the analysis which would mean for the authors that its role would be held in the nutritional status of the patient. In our study, both Mg and Albumin concentration maintains its independent predictive value for mortality, suggesting some other beneficial effect of Mg apart from the nutritional status, such as vascular calcifications and cardiac arrhythmias.

In some of the previous studies it was deduced that a better nutritional state is related to higher levels of Mg. The intake of Mg is associated with foods recommended in a healthy diet.11,12 The intake of Mg is currently decreasing and may not be sufficient to maintain the magnesium levels required to prevent diseases.11 The diet that is usually recommended for chronic kidney disease (CKD) patients, before or after initiation of regular HD is focused on K, P and proteins and not on Mg, and its content is often low. Adequate and sufficient intake will contribute to higher levels of Mg and better nutritional status.

In other studies, a higher concentration of serum Mg was due to treatment with MgO.19 Treatment with P binders containing Mg could contribute to higher levels of Mg.15,32 A reduced concentration of Mg is associated with treatment with PPI drugs.33–35 In our case, it was not observed the association of low Mg with PPI drugs, probably because it is masked by the lower concentrations of Mg induced by some of the dialysis fluids.

As observed in our study in HD patients, the concentration of Mg in the dialysis fluids and the presence of acetate or citrate would be determinants of the serum Mg concentration. Others have shown that increasing Mg concentration in the dialysis fluid results in an elevation of serum Mg concentration 36.

In dialysis patients, the ionic Mg / total Mg ratio is decreased. That is, “normal” magnesium values ​​in the general population may be indicative of true hypomagnesemia in HD patients 37. This would be another argument to raise the content of magnesium in the Dialysis fluid. Alterations in the acid-base balance could influence the free fraction of magnesium.

Symptoms due to hypermagnesemia develop with concentration above 1.5 mmol/l (3.6 mg/dl); with values above 2.5 mmol/l there is areflexia and above 4.5 mmol/l respiratory or cardiac arrest may occur. These extremely high levels should be prevented, and generally caused by acute poisoning due to dialysate fluid contamination 38, or the massive intake of Epson's salt 39. For the treatment of eclampsia, the therapeutic limits would be between 2.06 and 3.7 mmol/l.40

As in the general population, the association of serum Mg with mortality has also been described in patients with (CKD) not on dialysis.41,42 Low serum Mg levels and low dietary magnesium intake would be associated with an increase in the incidence of CKD and its progression to advanced CKD.42 This could be due to the association of Mg with other factors that favor CKD progression and it may also be due to direct effects through endothelial dysfunction and vascular calcifications.

With respect to the HD treatment there are other factors that should be commented:

A relationship between the number of hypotensions and hypomagnesemia has been described.43 Hypotension on dialysis is a determinant of morbidity and mortality.

In animal models it has been demonstrated the role of Mg on inflammation.44 In HD patients there is an inverse relationship between CRP and Mg suggesting a role of low Mg on inflammation.18 Thus, hypomagnesemia would be another of the factors involved in the chronic inflammatory status of HD patients.

In uremic rats, increasing Mg intake reduces mortality and reduce vascular calcifications.45 The initiation of calcification by the complex Ca-PL-PO4 is dependent on the Mg / Ca ratio present in the calcifying tissue.46 Mg, by preventing the maturation of the calciprotein particles, would inhibit the calcification of the smooth muscle cells of the vessels induced by high P.42 The risk of increased mortality associated with hyperphosphatemia would be attenuated by high Mg concentrations.47 Serum Mg would be an independent predictor of the non-progression of vascular calcifications.48 Increasing Mg in the dialysis fluid from 1.0 to 2.0 mEq / L decreases the rate of calcification in HD patients.49

Mg is capable of modulating PTH but with a lower potency than Ca. Hypermagnesemia could be associated with lower levels of PTH.50 Mild hypermagnesemia would be associated with a lower risk of fractures in HD patients.51 Mg has always been the great forgotten player of bone and mineral metabolism alterations.52

Changes in Mg along with modifications of K and Ca are associated with QT prolongation and the occurrence of cardiac arrhythmias.53,54 Severe cardiac arrhythmias are a major cause of mortality in HD.

High concentrations of serum Mg would be related to a better response to erythropoietin (EPO).55 The response index to erythropoietin is a prognostic marker.

Muscle function would be altered in HD patients with hypomagnesemia.56

All these effects of a low concentration of Mg could explain the relationship between Mg and mortality in HD patients.

The concentrations of Acetate in dialysis fluid commonly used in clinical practice, 3-4 mmol / l, increases the oxidative stress, the production of pro-inflammatory cytokines and synthesis of nitric oxide and may act as an adjunct to the other pro-inflammatory stimuli of uremic patients on HD.57 The elimination of acetate from HD could improve survival of HD patients, at least in those older than 70 years.58 Dialysis fluid with Citrate (CDF) do not produce this activation, so it could be an alternative in clinical practice. Citrate dialysis produces less acute post-dialysis alkalemia and modifies significantly serum levels of Ca, Mg and PTH.30,59 CDF has a positive impact on hemodynamic tolerance.30 Citrate in the dialysis fluid produces a negative balance of Ca and Mg26–29 so it could induce hypomagnesemia in some patients.39 The beneficial effect of a change of acetate by citrate could be counteracted by the development of hypomagnesemia. In the present work, the three types of dialysis fluids do not significantly influence mortality, but when the citrate dialysis fluid is associated with other variables and corrected for magnesium, it is associated with better survival. In the CDF, CX 265 G (Table 1) the concentration of Ca has been increased to 1.65 mmol/l to counteract the lower balance of Ca with citrate, the same should be done with Mg, which should at least be increased to 0.75 mmol/l. This would also be valid for some patients on dialysis fluid containing acetate; Schmaderer et al.36 showed that by increasing the Mg concentration in the dialysis fluid to 0.75 mmol/l there is an improvement of cardiovascular mortality. The routine use of 0.5 mmol/l of Mg in dialysate with acetate would cause a slow decrease in serum Mg in most patients.60

Being a single center study the total number of patients recruited resulted to be low. This small number of patients prevented the inclusion of more variables in the regression model, although all patients being treated with the same guidelines reduces of this limitation.