The present study showed that both s-Mg and s-Ca are major determinants of ADPN levels in ESRD patients. ADPN was positively associated with s-Mg and negatively with s-Ca. In addition, a strong association was demonstrated between high ADPN levels and all-cause mortality, which persisted after multivariate adjustment for possible confounders. Our main finding was that the predictive value of the effect of ADPN levels on mortality was critically dependent on s-Mg and s-Ca concentrations, since high ADPN levels were not predictive of all-cause mortality in patients having high s-Mg and low s-Ca levels.
In this study, we confirm many of the metabolic associations reported previously with ADPN in non-renal 
and renal patients 
. Specifically, in our study, lower BMI, albumin, triglycerides and higher HDL cholesterol were associated with higher ADPN levels. In this regard, our findings are in accord with the literature and support the validity of our dataset. Most importantly, this study documents for first time the existence of strong positive and negative associations of ADPN with s-Mg and s-Ca, respectively, in ESRD patients. These associations were independent of each other and independent of body composition, nutritional and inflammatory status. Thus, our data confirm the results of previously reported associations of ADPN with s-Mg 
and s-Ca 
in non- renal populations and further extent these findings in the ESRD population, where s-Mg 
and s-Ca 
strongly impact on outcomes. The exact mechanisms underlying these associations are not clear, but the fact that common defects in Mg and Ca metabolism are reportedly 
related to glucose metabolism, provides a possible explanation for this. Indeed, there is enough evidence to indicate that both hypomagnesemia 
and hypercalcemia 
are closely associated with insulin resistance. These findings, in concert with the observation that ADPN levels are decreased in patients with type 2 diabetes and in insulin resistance states may at least partially explain the positive and negative associations of ADPN with s-Mg and s-Ca, respectively.
Our results indicated that high ADPN levels were an independent predictor of total mortality in ESRD patients. There was a significant 7% increased risk for death from any cause for each 1-µg/ml increment of ADPN. In addition, the survival rate was significantly lower in patients in the higher sex-specific tertiles compared to those in the lower tertile of ADPN. These data are consistent with recent studies, where there was a 3% to 10.3% increased risk for all-cause mortality for each 1-µg/ml increment of ADPN in CKD 
and ESRD patients 
. Since ADPN is presumed to possess antiatherogenic and cardioprotective properties, the association of high ADPN levels with adverse clinical outcomes may be explained by an increased counter-regulatory secretion of ADPN to mitigate inflammation, malnutrition and to protect against endothelial damage and atherogenesis. Although, nutritional and inflammatory statuses were independent determinants of ADPN at baseline, they did not affect the ADPN- mortality association in our study. Alternatively, the existence of a state of adiponectin resistance 
perhaps due to reduced ligand/receptor activities or down regulation of adiponectin receptors or both may trigger a counter-regulatory increase of ADPN secretion in high risk ESRD patients. Another consideration is that a higher adiponectin level may induce protein energy wasting, a condition associated with malnutrition and inflammation 
. Reportedly, ADPN may increase energy expenditure and induce weight loss through a direct effect on the brain 
, thus, linking increased ADPN levels to increased mortality in patients with ESRD. Conversely, due to the inverse relationship between adiponectin and fat mass or BMI, weight loss increases plasma adiponectin levels 
and thus, high ADPN levels in ESRD patients may be a marker of wasting processes and poor prognosis. However, in the present study, adjusting for body composition (BMI) did not alter the effect of high ADPN on mortality.
However, there are also studies, carried out in the general 
, CKD 
and ESRD 
populations, in which the lowest levels of ADPN had the worst outcome. Discrepancies among studies in ESRD patients might be explained by differences in the populations studied, inclusion criteria, method of dialysis, confounding influences of covariates, different retention of the different ADPN isoforms in kidney disease 
and post-translational modifications in the ADPN molecule 
. In the study by Diez et al 
, comprised of 98 HD and 86 PD patients, an inverse relationship between ADPN levels and all-cause and CVD mortality was reported. Beside a shorter mean follow-up period of 31.2 months, the dialysis vintage was 2.5 (1.7–11.5) months in PD and 12.2 (4.8–43) months in HD patients, whereas the corresponding figures in our study were 36 (18–54) and 80 (36–108) months in PD and HD patients, respectively. In addition PD patients had a mean residual renal function of 3.3 (0.5–6.9) ml/min. It cannot be excluded that the beneficial effect s of ADPN during the early period of renal replacement treatment become harmful over time, particularly when the compensatory increase of ADPN is overwhelming. This assumption is further supported by a population-based cohort of 2484 participants 
, aged 50–75 year, where a higher ADPN was associated with an increased risk of CVD mortality in people with prevalent CVD [HR 1.27 (0.98–1.63)] and with reduced risk in people without CVD [HR 0.90 (0.73–1.11)]. In addition, data regarding s-Mg and s-Ca levels and dialysis prescription were not reported. In contrast, the inverse relationship between ADPN levels and CVD events in a cohort of 227 HD patients 
can be potentially explained by the Mg and Ca dialysate concentrations used in concert with the findings of the present study, a topic which will be discussed later.
In this study, s-Mg levels were directly correlated with nutritional factors and inversely with pulse pressure, a gross estimate of arterial stiffness, inflammatory markers and age. Furthermore, s-Mg levels predicted total mortality, but this association was largely dependent on age. These findings confirm the results of previous studies supporting a link between low s-Mg levels and atherogenesis 
or arterial calcification 
, malnourishment and increased risk of death in HD patients 
Elevated s-Ca levels and treatment with HdCa, both associated with an increased risk of Ca overload, have also been linked with morbidity and mortality 
. Our data agree with these reports showing that both increased s-Ca levels and the use of HdCa are associated with adverse clinical outcomes. Indeed, elevated s-Ca levels were associated with a more disadvantageous metabolic risk profile, in terms of increased pulse pressure and IL-8 and lower transferrin, while treatment with a HdCa of 1.75 mmol/l was associated with increased CRP. Both increased s-Ca levels and HdCa predicted independently total mortality. Thus, we provide solid evidence suggesting that Mg deficiency and Ca overload may contribute significantly to malnutrition, inflammation, arterial stiffening and increased CVD death in ESRD patients 
The most important finding of this study is that the association between ADPN and mortality varied among subgroups of patients stratified by s-Mg and s-Ca (and/or dCa). In contrast to low s-Mg and high s-Ca (and/or HdCa) groups, ADPN levels were not predictive of death in the high s-Mg and low s-Ca (and/or LdCa) groups. We speculated that the presence of ADPN resistance could be more pronounced in the former groups, due to a worse CVD risk profile, as discussed above. Alternatively, ADPN may not directly affect death risk, but may be a marker of other risks. Another possibility is that s-Mg and s-Ca may impact directly on the bioactivity of ADPN isomers in uremia. ADPN circulates in plasma as a low-molecular-weight (LMW) adiponectin (trimer), middle-molecular-weight (MMW) adiponectin (hexamer) and a high-molecular- weight (HMW) adiponectin (multimer). Although HMW is the most abundant isoform in ESRD patients 
, the distribution and role of each isoform in CKD remains largely unknown. However, emerging evidence suggest that LMW isoforms are associated with better clinical outcomes in both non-uremic and uremic populations, compared to the other isoforms. LMW isoforms were associated with lower CVD risk in children with CKD stage 2–4 
, and as opposed to HMW isoforms, appear to exert a protective role in older adults with previous coronary heart disease 
and lead to a reduction of liver cancer risk 
. Most importantly, a recent study clearly demonstrated that the formation of the fully developed complex HMW structure of ADPN is influenced by the presence of Ca 
. In both human and mice adipocyte cells, the presence of Ca led to a substantial increased formation of HMW adiponectin, with a corresponding decrease in MMW and LMW isomers, whereas the absence of Ca had the opposite result. These data indicate that low s-Ca and/or potentially high s-Mg levels may be associated with increased LMW isoforms and better outcomes, whereas high s-Ca and/or potentially low s-Mg levels may be associated with HMW isoforms and poor prognosis. This intriguing hypothesis needs to be confirmed in future studies.
This study may also have important clinical implications. Indeed, if s-Mg and s-Ca levels prove to be true effect modifiers of the association between ADPN and mortality, then these findings may impact on clinical practice in the management of ESRD patients, through modifications of dialysate prescriptions, particularly with regard to Mg and Ca and lead to improved guidelines for better outcomes in our high-risk patients. The median s-Mg concentration of 2.45 (2.3–2.7) mg/dl, above which a survival benefit was observed in this study, remained within normal range (1.7 to 2.67 mg/dl). Also, in a previous study 
using the same dialysate Mg concentration of 0.5 mmol/l, survival was significantly higher in patients with a mean s-Mg concentration above 2.77 mg/dl, a value considered indicative of mild hypermagnesemia. It is possible that if higher Mg dialysate levels had been used, the ensuing higher degree of hypermagnesemia could have resulted in an even better outcome. Since dialysate Mg concentration is an important determinant of Mg balance in both HD and PD patients, a higher s-Mg can be achieved by using a higher dialysate Mg concentration (0.75 mmol/l) than the currently used (0.5 mmol/l) in most countries. We have previously reported 
that after a four-week treatment with a dialysate Mg concentration of 0.75, 0.5 and 0.25 mmol/l, mean s-Mg concentrations were 2.94, 2.57 and 2.21 mg/dl, respectively. Major guidelines do not comment on dialysate Mg concentrations and trials on this topic with morbidity and/or mortality end points are lacking. A recent review 
of Mg in dialysis patients indicated that a Mg dialysate of 0.75 mmol/l is likely to cause mild hypermagnesemia, whereas s-Mg levels were mostly normal to low when 0.2 and 0.25 mmol/l Mg concentrations were used. Results were inconsistent (normomagnesemia in most studies) with regard to Mg dialysate of 0.5 mmol/l. A higher survival rate was also observed in patients with a s-Ca concentration below the median 9.3 (8.8–9.7) mg/dl and/or using a LdCa of 1.25 mmol/l. Current guidelines 
recommend the use of a dCa concentration of 1.25 to 1.5 mmol/l in both HD and PD patients. However a recent study 
showed that the intradialytic Ca mass balance was nearly neutral using a dCa of 1.25 mmol/l, whereas treatment with a dCa of 1.50 mmol/l resulted in gain of Ca during HD. dCa concentrations as high as 1.75 mmol/l should be avoided to prevent calcium overload and the induction of adynamic bone disease. However, most studies 
showed a positive effect of HdCa on haemodynamic stability during dialysis compared with LdCa concentrations. Taken all these data together, one could speculate that by increasing dialysate Mg concentration up to 0.75 mmol/l and decreasing dCa concentration from 1.75 or 1.50 to 1.25 mmol/l, the increased ADPN levels in uremia would have rather a beneficial effect on outcomes. Unfortunately, dialysate Mg and Ca levels are not reported in the relevant studies. In the study of Zocalli et al 
, where these were reported, the use of a high Mg dialysate of 0.75 mmol/l and a LdCa of 1.25 mmol/l were associated with a 3% CV risk reduction for each 1-µg/ml increase in plasma ADPN levels.
Thus, we recommend that s-Mg and s-Ca levels should be taken into consideration when assessing the role of ADPN on outcomes in ESRD and the optimal s-Mg and s-Ca levels required for a survival advantage in relation to ADPN be established.
This study has several limitations. First, due to the small number of patients who died, specific mortality risk (i.e. CVD) could not be determined and generalizability of study results might have been compromised. Generalizability might also have been jeopardized by the low percentage of diabetics, low number of comorbid conditions, lack of other ethnic groups and the fact that a single center participated in the study. Nevertheless, this study enabled us to detect a strong ADPN-mortality association in ESRD patients, the magnitude and direction of which were comparable to those previously reported in relevant studies of the same 
or larger populations 
. In the study of Ohashi et al 
, with a sample size (n
75), number of deaths (n
15) and a threshold for assessing mortality (15 µg/ml) quite similar to ours, the magnitude of association between ADPN and total mortality was comparable to ours (10.3% vs. 7% adjusted risk increment for each 1-µg/ml increase in ADPN). The robustness of this association did not decrease after adjusting for potential confounder and/or mediators in both pooled and subgroup analyses. Second, we measured total ADPN and not its various isoforms, the reason being that the relevant methodology at the time of our measurements was not available. Notwithstanding, since this was generating hypothesis study, further assessment of ADPN isomers will be necessary to elucidate the difference in the effect of each ADPN isomer on clinical outcomes. In this regard, the first step in testing this intriguing hypothesis is to confirm the presumed positive and negative associations of LMW isoforms with s-Mg and s-Ca, respectively and the corresponding opposite associations regarding HMW isoforms in large cross-sectional studies, and b) then prospectively verify the favorable and unfavorable effects of LMW and HMW isoforms, respectively, on outcomes in relation to targeted s-Mg and s-Ca concentration, through appropriate use and manipulation of Mg and Ca concentration in the dialysis bath. Third, the use of a single baseline measurement to predict events several years in the future. However, serum concentrations of adiponectin seem stable during a period of 1 yr, with minimal short-term variation and high degree of reproducibility 
In conclusion, we showed that s-Mg and s-Ca levels can modify the effect of ADPN on all-cause mortality, aiding in unraveling the controversy which surround this association in the existing literature. High ADPN was an independent predictor of death risk only in patients with low s-Mg and high s-Ca levels, respectively, conditions highly associated with a worse CVD risk profile and possibly a marked increase in ADPN resistance. Conversely, the better survival rates seen with high s-Mg and low s-Ca may be caused by altered ADPN bioactivities associated with death risk reduction. Future studies are needed to elucidate the exact roles of s-Mg and s-Ca on ADPN bioactivity in relation to clinical outcomes in ESRD.