Skeletal and cardiovascular consequences of a positive calcium balance during hemodialysis

Abstract Patients on hemodialysis are exposed to calcium via the dialysate at least three times a week. Changes in serum calcium vary according to calcium mass transfer during dialysis, which is dependent on the gradient between serum and dialysate calcium concentration (d[Ca]) and the skeleton turnover status that alters the ability of bone to incorporate calcium. Although underappreciated, the d[Ca] can potentially cause positive calcium balance that leads to systemic organ damage, including associations with mortality, myocardial dysfunction, hemodynamic tolerability, vascular calcification, and arrhythmias. The pathophysiology of these adverse effects includes serum calcium changes, parathyroid hormone suppression, and vascular calcification through indirect and direct effects. Some organs are more susceptible to alterations in calcium homeostasis. In this review, we discuss the existing data and potential mechanisms linking the d[Ca] to calcium balance with consequent dysfunction of the skeleton, myocardium, and arteries.

Patients on hemodialysis are exposed to calcium via the dialysate at least three times a week. Changes in serum calcium vary according to calcium mass transfer during dialysis, which is dependent on the gradient between serum and dialysate calcium concentration (d [Ca]) and the skeleton turnover status that alters the ability of bone to incorporate calcium. Although underappreciated, the d [Ca] can potentially cause positive calcium balance that leads to systemic organ damage, including associations with mortality, myocardial dysfunction, hemodynamic tolerability, vascular calcification, and arrhythmias. The pathophysiology of these adverse effects includes serum calcium changes, parathyroid hormone suppression, and vascular calcification through indirect and direct effects. Some organs are more susceptible to alterations in calcium homeostasis. In this review, we discuss the existing data and potential mechanisms linking the d[Ca] to calcium balance with consequent dysfunction of the skeleton, myocardium, and arteries.

IntroductIon
Patients with end-stage kidney disease (ESKD) have increased morbidity and mortality 1 ; chronic kidney disease mineral bone disorder (CKD-MBD) is a consistent independent risk factor. Although calcium disturbances and treatments that alter serum calcium have been addressed in several reviews, the impact of a positive calcium balance during hemodialysis due to calcium dialysate content -d[Ca] -on this outcome is rarely considered.
Calcium is an essential mineral for the function of various organ systems through its impact on hormones and cell signaling. As a result, humans have a complex homeostatic system to maintain normal levels of calcium in the blood through regulation by multiple hormones acting on the intestine, parathyroid, kidneys, and bones. With evolution from fish to amphibians, the skeleton had to adapt because, compared to the ocean, the calcium content in land food and water is lower. Therefore, the skeleton became an organ that not only gave motility and strength to the body, but also served as a reservoir of calcium that can respond, even at its own expense, to maintain normal levels of ionized calcium. Calcium accounts for 1 to 2% of adult human weight, and over 99% is found in teeth and bones. The ratio of extracellular to intracellular calcium and the amount of calcium stored within the cells are tightly controlled 2 .
This homeostasis is disrupted in patients with chronic kidney disease (CKD) due to disruption of normal homeostatic loops with kidney function decline, with compensatory changes in several hormones (parathyroid hormone -PTH, vitamin D, fibroblast growth factor 23 -FGF23). The abnormalities of CKD-MBD begin in some patients as early as eGFR of 70 mL/min/1.73m 2 and are almost universal when eGFR is < 30mL/min/1.73m 2 3 . In the early 1970s, it was discovered that 1,25vitamin D levels were uniformly reduced, and calcium intestinal absorption was decreased in patients with CKD. The conclusion at that time was that patients with CKD were therefore in a negative calcium balance and thus would require calcitriol and/or calcium supplementation to enhance intestinal absorption and avoid hypocalcemia and secondary hyperparathyroidism. However, subsequent studies in patients with moderate CKD have shown that oral calcium carbonate supplementation 4 or calcium-rich diet 5 induce a positive overall calcium balance that might facilitate soft-tissue deposition of calcium salts. Furthermore, the discovery in the early 2000s that FGF23 inhibits calcitriol synthesis supports an attempt of the body to purposefully decrease intestinal calcium absorption, due to the marked reduction of calcium excretion with declining kidney function. These studies have led to the current concept that the decreased intestinal absorption is an appropriate adaptation to avoid positive calcium balance and its adverse consequences 6 . The need for hemodialysis adds a new variable to this broken regulatory system as patients are exposed to an external calcium-rich solution (dialysate) during each dialysis treatment. The acute change in ionized calcium during dialysis can alter the balance between bone and serum calcium, further altering overall calcium balance ( Figure 1). The purpose of this review is to examine the evidence supporting a pathologic role of a positive calcium balance on end organ damage in patients undergoing hemodialysis.

cAlcium tRAnspoRt duRing hemodiAlysis
Calcium balance during dialysis is defined as the net amount of calcium that is obtained (or lost) during a treatment. Briefly, this balance is determined by two factors: the calcium gradient and the ultrafiltration volume. Calcium gradient is defined by the difference between dialysate and serum calcium. The blood calcium is composed of three fractions: protein bound, complexed, and ionized 7 . The ionized calcium is the active fraction representing ~50% of the total calcium. Only the ionized calcium and calcium complexed to small anions are transported across the dialyzer membrane, and the latter is only estimated and not usually measured. Albumin is the primary calcium-binding protein, but unfortunately, formulas estimating ionized calcium based on total calcium and albumin are inaccurate, particularly in dialysis patients 8 . Calcium can also bind to phosphate and bicarbonate, and concentrations of both can change markedly during dialysis. Therefore, the difference between dialysate and the blood-free calcium times the total dialysate volume is the main component of the calcium mass balance during the dialysis and could be either negative or positive. Conversely, there will be always a negative balance of calcium due to the ultrafiltration volume. In other words, the patients will always loose a small amount of calcium when we prescribe a negative balance of water. However, this is a small component of the calcium balance during dialysis, and the calcium gradient predominates.
Usually, the concentration of total calcium in the extracellular compartment increases during dialysis and decreases between the sessions. In most studies, the use of d[Ca] of 1.5 or 1.75 mmol/L (3.0 or 3.5 mEq/L) leads to post-dialysis hypercalcemia and positive calcium balance. However, the actual calcium mass transfer is unpredictable due to 1) differences in the ionized calcium blood level, 2) differences in convective loss or calcium that varies depending on ultrafiltration, and 3) imprecise calculations of the delivered dialysate calcium 9,10 . Finally, our work has shown that the change in ionized calcium is greatest in the first 30 to 60 minutes and also depends on bone remodeling 9 . Thus, not all patients will gain or lose calcium, or develop hyper or hypocalcemia on the same d[Ca] concentration. However, as shown in Table 1, many studies based their conclusion on intradialytic calcium balance using only the serum calcium variation. Moreover, some studies have shown that changing sodium concentration in the dialysate will decrease the calcium concentration affecting the intradialytic calcium mass transfer 11 .
Interestingly, the majority of the studies on intradialytic calcium balance was done on hemodialysis (Table 1). However, hemodiafiltration has been adopted by several centers, sometimes as the main modality of renal replacement therapy. In this case, water and fluid change through convection might reach 30 liters and certainly gains importance in comparison with the maximum 3-4 liters that are extracted in hemodialysis. Conversely, the diffusion process has a lower impact on the calcium balance when compared to hemodialysis. Studies on post-dilutional hemodiafiltration done in the 1990s show similar results on calcium balance to those on hemodialysis 12 .
However, the use of pre-dilutional hemodiafiltration can be associated with a negative intradialytic calcium balance. Pieces of evidence of long-term effects of hemodiafiltration, however, are scarce in the literature. Argiles et al. 12 in 1993 compared the effects of lowering d[Ca] to 1.25 mmol/L in 7 patients vs. 6 control patients using d[Ca] 1.5 mmol/L. Calcium carbonate oral intake was more than doubled in the low d[Ca] group. Total calcium, phosphate, and alkaline phosphatase were similar in both groups over the year. Basile et al. 13 showed in 2001 that neither pre-nor post-dialysis systolic and diastolic blood pressures, predialysis serum bicarbonate and pH, and pre-dialysis serum sodium, potassium, calcium, or phosphorus were significantly different when comparing hemodialysis and hemodiafiltration.
Another potential confounding factor for the intradialytic calcium balance is the use of citrate instead of acetate in the dialysate. It has been demonstrated that PTH decreases and ionized calcium increases using bicarbonate buffer, whereas acetate does the opposite, increasing PTH and decreasing serum calcium 14 .
Despite the intradialytic effect on calcium balance, it is important to recognize the potential acute effect of dialysis on serum calcium. It is plausible that acute changes in extracellular calcium may affect intracellular calcium concentration, which is a key signaling pathway for nearly every cell. In animal models of CKD, there is an increase in intracellular calcium concentration in multiple cell types. In cardiomyocytes, the increase is mediated by PTH through a process that includes both increased entry of calcium into cardiac myocytes and decreased exit of this ion from these cells 15 . There is also an increase of basal intracellular calcium in vascular smooth muscle cells regardless of the PTH 16 . Thus, acute extracellular changes in ionized calcium may alter cellular function by changing the gradient across the cell membrane, which in turn can alter intracellular calcium. For example, in cardiomyocytes and vascular smooth muscle cells, contraction occurs through the increase in cytoplasmic calcium that comes from the extracellular calcium and/or release from the sarcoplasmic reticulum and mitochondria. Conversely, cytosol calcium efflux leads to relaxation by energy-dependent with adenosine triphosphate (ATP) generation. Thus, acute changes in extracellular ionized calcium may alter the gradient across the cell membrane resulting in unwanted intracellular changes in calcium and cell dysfunction.   One of the confounding factors for the poor accuracy of serum calcium to predict calcium balance is the skeleton. Bone is a reservoir of an exchangeable 300 mg/day of calcium 20 . Almost 20 years ago, Talmage et al. 21 hypothesized that PTH not only stimulates osteoclast-mediated bone resorption, but also increases the content of some bone surface calcium binding proteins, such as osteocalcin, that could act as calcium buffers. In other words, PTH would increase the amount of osteocalcin in the bone surface, increasing the capacity of the skeleton to acutely donate or retain calcium during dialysis, depending on acute changes in blood calcium. More than 20 years ago, Kurz et al. 22 , using double radiolabeled calcium, showed that the acute bone calcium uptake was higher in high turnover bone disease compared to either mixed uremic osteodystrophy or low turnover bone on a non-dialysis day. Therefore, the skeleton determines not only the exchangeable calcium pool during dialysis but also the pre-dialysis serum calcium through its response to PTH. Indeed, recent studies have proven that calcium balance also varies according to bone turnover status 9,10 . We and others have shown that calcium balance varies from negative 1500 to positive 800 mg using the same d[Ca] of 1.25 mmol/L 9, 10 .
Our group has demonstrated the association between bone remodeling markers and calcium mass transfer during a conventional hemodialysis 10 . We studied 23 patients dialyzed using a d[Ca] of 1.0, 1.25, 1.5, and 1.75 mmol/L, in which the mean ± SD and range of calcium removal was -578, -468, 46, and 405 mg, respectively. Multivariate analysis showed that calcium balance was dependent on calcium gradient, PTH, and osteocalcin. Bone remodeling, however, is hard to predict when based only on biomarkers. The ideal study design to evaluate the influence of bone remodeling state on calcium balance would be a repeated analysis in the same patient with different bone turnovers, and maintaining other biases unchanged, i.e., ultrafiltration volume and d [Ca]. We conducted such a study, examining calcium mass transfer in the same patients before parathyroidectomy (PTX), during the first month after surgery in the hungry bone phase, and after surgery when blood calcium stabilized 9 . We confirmed a wide variation of calcium mass transfer during hemodialysis according to each phase (before PTX, hungry bone, and late after PTX) and each d[Ca] used (1.25 vs. 1.5 vs. 1.75 mmol/L). Even with no difference in the ultrafiltration volume, calcium mass transfer varied among phases and among d [Ca] used, supporting our hypothesis of the importance of bone turnover. (Figure 2) Therefore, the skeleton is a key determinant in net calcium gain or loss during dialysis, and is an important consideration when prescribing the optimal dialysate calcium.

effects of cAlcium bAlAnce on myocARdium
Cardiovascular (CV) morbidity and mortality rates in patients with ESKD on hemodialysis are alarming with a 5-year survival rate of only 45%. The leading cause of death is cardiovascular events 23 . Alterations in the d[Ca] may be a plausible mechanism to stabilize blood pressure/hemodynamics, myocardial function, hemodynamics, and reduce arrhythmias, especially during and immediately after the dialysis treatment.
The effect of d[Ca] on myocardial perfusion has been studied by a number of methods. Myocardial stunning is common during dialysis, and it can be minimized by the use of a cool dialysate 24 and the reduction of the ultrafiltration rate, which can be achieved by changing to short daily or long nocturnal hemodialysis sessions 25 . Diastolic dysfunction, an independent predictor of mortality in patients on dialysis, involves abnormal left ventricular relaxation, filling, and distensibility. As coronary blood flow is greater during diastole, diastolic dysfunction may lead to the reduction of coronary perfusion, which may lead to subendocardial ischemia and systolic dysfunction.
Using echocardiography, one study measured diastolic function in hemodialysis with no ultrafiltration on a 1.75 mmol/L d[Ca], finding no impairment in diastolic function despite an increase in ionized calcium 26 . However, two other studies using calcium gluconate infusion 27 or a high d[Ca] of 1.75 mmol/L 28 both found impaired diastolic relaxation. Therefore, studies suggest that higher calcium dialysate may worsen ventricular relaxation assessed by echocardiography.
Recently, our group used a more sensitive method, two dimensional speckle imaging with strain analysis, to evaluate the impact of d[Ca] on myocardial performance during hemodialysis 29 . We found hypercalcemia (11.5 ± 0.8 mg/dL) after hemodialysis using d[Ca] of 1.75 mmol/L and improved hemodynamic stability in terms of blood pressure. However, the global longitudinal strain (GLS) was worse during the last hour of hemodialysis compared to baseline (p < 0.001). In addition, the GLS was worse with d[Ca] of 1.75 than 1.25 mmol/L (−16.1 ± 2.6% vs. −17.3 ± 2.9%, respectively; p < 0.001). Multiple linear regression showed that independent risk factors for GLS were transferrin, c-reactive protein, baseline GLS, weight loss during hemodialysis, and post dialysis serum calcium. Figure 3 shows echocardiogram images illustrating GLS at baseline and at the peak hemodialysis using d Previous studies have shown an increased calciuminduced myocardial contractility when a d[Ca] of 1.75 mmol/L was employed 30,31 . However, one of these studies was performed in 1984 30 and included only eight patients using three different d[Ca] to evaluate the left ventricular contractility by two-dimensional echocardiography. Authors concluded that the increase of ionized calcium after dialysis was associated with an improvement of contractility. Other authors four years later 31 included seven patients and tested 3 different d [Ca]. Left ventricular contractility was assessed using the relation between left ventricular end-systolic wall stress and myocardial systolic performance. They concluded that high d[Ca] had a positive impact of myocardial performance. A recent study has evaluated cardiac function in a Langendorff-like system of a zebrafish 32 . By manipulating the calcium concentration of the perfusion buffer, authors surprisingly found that the ejection fraction initially increased along with the increase in calcium concentration, similarly to previously mentioned studies, and then decreased. Although the experimental scenario is unlike clinical practice, this finding should raise alert about the deleterious effect of high d [Ca]. There is no doubt that calcium is imperative to ventricular contraction. These inconsistent findings suggest that calcium is indeed important for effective myocardial contraction in an almost direct relationship, although hypercalcemia can impair ventricular function. The most common cause of cardiovascular death in patients undergoing dialysis is arrhythmias. In an analysis of a large dialysis provider dataset, low calcium dialysate (< 1.25 mmol/L), and increased dialysate calcium gradient were associated with a 2 and 1.4 odds ratio, respectively, of sudden cardiac arrest 33 . Arrhythmias may be more common with low versus high d[Ca], but increased sympathetic activity that may trigger arrhythmias may be more common with high, compared to low, d [Ca]. The effect of long term d[Ca] concentration on cardiovascular outcomes is not well studied. However, a recent secondary analyses of the EVOLVE trial found that baseline dialysate calcium concentration, which was generally higher in Europe and Latin America, did not alter the outcomes of the trial that compared cinacalcet to placebo on cardiovascular composite outcome 34 .

effects of cAlcium bAlAnce on vAsculAR cAlcificAtion
Arterial calcification is common as over 70% of patients starting dialysis have significant coronary artery calcification on CT scanning 35,36 . The pathology of arterial calcification includes a mix of atherosclerotic calcification and medial calcification. In dialysis patients who died of a cardiovascular event, atherosclerotic plaque was more calcified than age-matched controls with similar cause of death 37 .
Similarly, the prevalence of medial calcification of large arteries increases with progression of kidney disease and is higher in patients on dialysis than age matched controls 38 . Aorta calcification can lead to increased pulse pressure and increased cardiac afterload resulting in increased cardiovascular events and mortality 39 .
The mechanism of medial calcification is a multistep process initiated by dedifferentiation of vascular smooth muscle cells (VSMC) to become osteo-chondrocytic-like cells via upregulation of the transcription factor RUNX2. These transformed cells then mineralize in a manner similar to bone, with production of collagen and non-collagenous proteins on which secreted matrix vesicles (containing multiple proteins and calcium and phosphate) mineralize. Vascular calcification is regulated in part by inhibitors of calcification including fetuin-A, a circulating reverse acute phase reactant protein that acts to bind circulating 'calcioproteins' containing calcium and phosphate for clearance, and matrix-gla protein, an inhibitor that is upregulated in vascular smooth muscle cells 40 .
Positive calcium balance and/or hypercalcemia are involved in the pathogenesis of arterial calcification. VSMC move from a contractile to synthetic phenotype, a prerequisite for de-differentiation. In rats with CKD, freshly isolated VSMC from aorta have increased intracellular calcium concentration, indicative of the synthetic state 16 . These cells then upregulate RUNX2 to become osteo-chondrocytic-like cells in the presence of uremic serum 41 . Shanahan and colleagues demonstrated that calcium induces release of matrix vesicles and increased calcification independently and synergistically with phosphate in cultured VSMC 42 . Giachelli's group also demonstrated that calcium-induced calcification was synergistic with hyperphosphatemia. Furthermore, they demonstrated that incubating VSMC with high calcium media led to upregulation of Pit-1, a sodiumphosphate transporter important in the upregulation of RUNX 43 . In vivo, calcium containing phosphate binders induce arterial calcification in 5/6 th nephrectomy rats 44 , Cy/+ model of CKD 45 , and the LDLR-/-highfat-fed mice with CKD 46 despite lower levels of serum phosphate. In our Cy/+ rat model of progressive kidney disease, we treated rats with advanced CKD with calcium administered in drinking water, calcimimetic R-568, and R-568 plus calcium versus no treatment 47 . Treatment with calcium in the drinking water led to increased thoracic aorta, heart, and aortic valve calcification regardless of the serum level of calcium, indicating positive calcium exposure/balance can induce arterial calcification regardless of calcium blood levels. The calcium treatment led to an even greater calcification than observed with hyperphosphatemia and normal calcium levels 47 . Thus, hypercalcemia, or positive calcium balance even without hypercalcemia, can directly induce calcification in vitro and in vivo.
In patients receiving hemodialysis, most randomized trials comparing calcium-based phosphate binders, compared to non-calcium binders, show greater progression of coronary artery calcification 48 . However, studies examining the role of calcium load from dialysate calcium are limited. A small study compared dialyzed patients against three acute variations of calcium concentrations and found increased carotid-femoral and carotid-radial pulse wave velocity (PWV; a measure of increased stiffness) with higher dialysate calcium 49 . Another small study randomized patients on nocturnal dialysis to low calcium dialysate (1.3 mmol/L, n = 24) or high calcium dialysate (1.6 or 1.75 mmol/L, n = 26) and found no difference in abdominal aorta calcification by CT over one year 50 . However, Ok and colleagues (17) conducted a large randomized trial in patients on thrice weekly HD with intact parathyroid hormone levels ≤ 300 randomized to 1.25-mmol/L Ca dialysate (n = 212) or 1.75-mmol/L Ca dialysate (n = 213). The results showed a significant increase in coronary artery calcification in the patients randomized to 1.75 mmol/L Ca dialysate compared to the lower dialysate calcium. Importantly, hyperphosphatemia also increased coronary artery calcification, and the combination of hyperphosphatemia and high calcium dialysate was additive in inducing increased coronary artery calcification. Thus, increased calcium dialysate, especially in the setting of hyperphosphatemia, appears to increase arterial calcification in hemodialysis patients, similar to observations from in vitro VSMC cultures and in vivo in rodent models of CKD.