Long-term effects of low calcium dialysates on the serum calcium levels during maintenance hemodialysis treatments: A systematic review and meta-analysis

Hypercalcemia and hyperparathyroidism in patients receiving maintenance hemodialysis (MHD) can cause the progression of cardiovascular diseases (CVD) and mineral bone disorders (MBD). The KDIGO recommends the dialysates with a calcium (Ca) concentration of 1.25–1.5 mmol/L for MHD treatments, but the optimal concentration remains controversial. Here, we conducted a systematic review and a meta-analysis of seven randomized controlled trials examining a total of 622 patients to investigate the optimal concentration for MHD for 6 months or longer. The dialysates with a low Ca concentration (1.125 or 1.25 mmol/L) significantly lowered the serum Ca and raised the intact parathyroid hormone levels by 0.52 mg/dL (95% confidence interval, 0.20–0.85) and 39.59 pg/mL (14.80–64.38), respectively, compared with a high Ca concentration (1.50 or 1.75 mmol/L). Three studies showed that a low concentration was preferred for lowering arterial calcifications or atherosclerosis in different arteries, but one study showed that coronary arterial calcifications increased with a low concentration. Two studies showed contradictory outcomes in terms of MBD. Our meta-analysis showed that a dialysate with a low Ca concentration lowered the serum Ca levels in patients receiving long-term MHD, but further studies are needed to determine the optimal Ca concentration in terms of CVD and MBD.

be achieved. Here, we conducted a systematic review and a meta-analysis of randomized controlled trials (RCTs) to investigate the optimal dCa to prevent the progression of CVD and CKD-MBD during long-term conventional maintenance HD.

Results
Database search and profiles of studies. We searched for studies using the PubMed and CENTRAL databases and identified a total of 513 articles. After removing duplicates, we reviewed the titles and/or abstracts and excluded 446 articles. We then assessed the full texts of the remaining 16 articles and excluded 9 articles. Finally, we adopted seven RCTs [10][11][12][13][14][15][16] in which researchers compared the effects of a low Ca dialysate (LCD) of 1.25 or 1.125 mmol/L with those of a high Ca dialysate (HCD) of 1.5 or 1.75 mmol/L when used for 6 months or longer in patients undergoing conventional maintenance HD. A flow diagram 17 showing our search strategy and process is presented in Fig. 1. The profiles of the seven included studies are shown in Table 1. A summary of the risk of bias for each of the included studies was determined using the Cochrane Collaboration's tool 18 , as shown in Fig. 2. A meta-analysis of the levels of serum Ca, P and intact PTH. First, we aimed to conduct a meta-analysis of the levels of serum Ca, P and intact PTH (iPTH) among patients in all seven RCTs. However, data on the serum Ca and P levels were not available for one RCT 16 , though the data was presented graphically. Another study 12 only reported the values as time-averaged data during the 24-month study. We excluded these two studies from our meta-analysis of the serum Ca levels. A forest plot of the mean difference (MD) in the predialysis serum Ca levels between patients in the LCD and HCD groups in the other five RCTs is shown in Fig. 3A. This meta-analysis showed that the use of LCD significantly lowered the serum Ca levels, compared with the use of HCD, by 0.52 mg/ dL (95% confidence interval [CI], 0.20-0.85; I² = 57%). A funnel plot, Begg's test 19 and Egger's test 20 showed that a publication bias was not present in this meta-analysis (p > 0.10, Fig. 3B). We could not combine the serum P levels obtained in the five RCTs together for a meta-analysis because of the high heterogeneity (I² = 88%), nor could we observe any tendencies (data not shown). Regarding the serum iPTH levels, three of the seven RCTs [10][11][12] showed their values as geometric means (95% CIs), medians (interquartile ranges), and time-averaged data during the 24-month study, respectively. A forest plot of the MD of the serum iPTH levels between patients in the LCD and HCD groups in the other four RCTs is shown in Fig. 3C. This meta-analysis showed that the use of LCD significantly increased the serum iPTH levels, compared with the use of HCD, by 39.59 pg/mL (95% CI, 14.80-64.38; I² = 67%). As a matter of fact, the use of LCD increased the serum iPTH levels in most of the RCTs despite the heterogeneity in the clinical backgrounds, as shown in Table 1. Review of the effects on CVD and CKD-MBD. Next, we attempted to review the effects of dCa on CVD and CKD-MBD in patients undergoing maintenance HD. However, the outcomes were shown using different indexes in each study, preventing the data from being combined in a meta-analysis. Instead, we have summarized the study outcomes in Table 2. For CVD, three study groups independently showed that the use of LCD was favorable for preventing the progression of arterial calcifications and/or atherosclerosis in different arteries. He et al. 11 reported that the carotid intima-media thickness (cIMT) and the carotid-femoral pulse wave velocity (cf-PWV), parameters of atherosclerosis and aortic stiffness respectively, were significantly lower in an LCD  Table 1. Characteristics and Ca, P, iPTH outcomes of included RCTs. Abbreviations: ABD = adynamic bone disease; Ca = calcium; CI = confidence interval; HCD = high calcium dialysate; iPTH = intact parathyroid hormone; IR = interquartile ranges; LCD = low calcium dialysate; n.s. = not significant; P = phosphate; RCT = randomized controlled trial; SD = standard deviation; SE = standard error. *Mean ± SE.  12 showed that the progression rate of the coronary artery calcification scores (CACS) was significantly slower in an LCD group than in an HCD group (p = 0.03). Lu et al. 13 showed that the cIMT, the resistance index and the abdominal aortic calcification scores were notably decreased after 6 and 12 months of study in an LCD group (p < 0.05). However, Kim et al. 10 reported a contradictory outcome, with the CACS increasing in both groups but with the increase in the LCD group being especially significant. As for CKD-MBD, two study groups reported contradictory outcomes. Ok et al. 12 showed that both the bone turnovers and volumes were significantly higher in an LCD group than in an HCD group at 24 months, based on the results of iliac crest bone biopsies. On the other hand, Sánchez et al. 16 reported that the bone mineral density of the lumbar spine was significantly reduced after one year of HD with LCD, but not with HCD, as shown using quantitative computed tomography.
A meta-analysis of the mortality. Finally, we investigated the mortality rate from any cause during each study period in all seven RCTs. As shown in Fig. 3D, the use of LCD did not increase the mortality rate of patients, compared with the use of HCD (risk ratio [RR], 0.83; 95% CI, 0.58-1.19; I² = 19%).

Discussion
To our knowledge, this is the first systematic review and meta-analysis of the long-term effects of LCD on the serum Ca, P and iPTH levels and on the risk of CVD and CKD-MBD during conventional maintenance HD treatments. We compared the results of seven RCTs with regard to these issues and found that the use of LCD (1.125 or 1.25 mmol/L) significantly lowered the serum Ca levels and raised the serum iPTH levels by 0.52 mg/ dL and 39.59 pg/mL, respectively, compared with HCD (1.50 or 1.75 mmol/L) in patients undergoing maintenance HD for 6 months or longer periods. As shown in Table 1, these changes in most cases typically fell within a range which the KDIGO guideline 8,9 and/or The Japanese Society for Dialysis Therapy guideline recommend 21 , although cinacalcet was not used in any of the seven included RCTs and vitamin D was not used either in the two RCTs where the baseline serum iPTH levels were below 100 or 120 pg/mL. Consistent with our meta-analysis, a meta-analysis 22 comparing the long-term (12-24 months) clinical effects in patients undergoing peritoneal dialysis (PD) using a dialysate with a dCa of 1.25 or 1.75 mmol/L showed that the serum total Ca levels were lower in the LCD group (MD, 0.09 mmol/L; 95% CI, 0.05-0.13; p < 0.0001) and that the serum iPTH levels were significantly higher in the LCD group. In addition, three of the included RCTs 11-13 showed favorable long-term effects from the use of LCD on the incidence of CVD, as shown in Table 2. One study group 10 , however, unexpectedly showed a contradictory outcome; in this study, the authors inferred that the increased serum iPTH levels in the LCD group might have influenced the progression of coronary artery calcifications. In another RCT that was not included in the present meta-analysis because the patients were randomized between groups receiving LCD with a dCa of 1.12 or 1.37 mmol/L (both lower than 1.50 mmol/L), LeBoeuf et al. 23 showed that the cf-PWV was higher in the higher dCa group, compared with that in the lower dCa group, supporting the idea that a high dCa might be a risk factor for the progression of aortic stiffness in patients receiving maintenance HD. As for CKD-MBD, two study groups reported contradictory outcomes, as shown in Table 2. We suggest, however, that the use of LCD has also favorable effects on CKD-MBD for the following reasons. First, the serum alkaline phosphatase levels, which were regarded as a marker of the bone turnovers in the included RCTs, tended to be higher in the LCD groups than in the HCD groups, as shown in Table 2. Second, Haris et al. 24 conducted a study in which 51 patients treated with maintenance PD and biopsy-proven adynamic bone disease (ABD) were randomized to treatment with control Ca (1.62 mmol/L) or low Ca (1.0 mmol/L) dialysate over a 16-month period. The study showed that the LCD group experienced a decrease in the serum Ca levels, an increase in the serum PTH levels and a rise in bone formation rates into normal range significantly, but that the control group did not. The KDIGO Guideline considered that LCD would yield neutral Ca balance 8,9 . Positive Ca balance from the dialysate to the body by use of HCD could lead to low turnover bone disease which is associated with a decreased capacity to buffer extra Ca load 25 , so that not only ABD but also ectopic calcification such as atherosclerosis can be caused 26 . In a meta-analysis 27 , one randomized study 28 of HD patients (although the patients were not randomized according to dCa) showed that the serum Ca levels did not change significantly in patients who underwent long-frequent nocturnal HD (six times per week, ≥ six hours per session) with a dCa ≥ 1.5 mmol/L for 12 months, compared with those who underwent conventional HD with a dCa < 1.5 mmol/L. This result is probably because in such intensive HD treatments, the serum Ca is lost from the body to the dialysate even if a higher dCa is used.
The serum Ca, P and iPTH levels are important therapeutic targets requiring proper control during maintenance HD treatments, but it is often difficult to achieve this because these parameters interact with each other. Hypercalcemia can cause the vascular calcifications that effect the risk of death 29 , while mild hypocalcemia without clinical symptoms is thought to be acceptable 9 . Severe HPT can increase the risk of bone fractures 30 , but the relatively low serum iPTH levels can lead to the low bone turnovers or ABD 31 . Although non-Ca-based P-binders and cinacalcet have been available recently, these medications have also demerits (e.g., side effects such as gastrointestinal distress and binding of essential nutrients) 9 . Moreover, EVOLVE trial 32 failed to show that cinacalcet reduced the risk of death or major cardiovascular events in patients undergoing HD with moderate-to-severe secondary HPT. Therefore, a consensus remains to be achieved as to which of vitamin D or cinacalcet should be used to prevent secondary HPT, and both of these medications are often used simultaneously 9 . Although additional randomized studies might be needed to determine which dialysate is optimal for preventing the progression of CVD and/or CKD-MBD, our systematic review and meta-analysis suggested that the use of LCD could prevent vascular calcifications by lowering the serum Ca levels, maintain bone turnovers by raising the serum iPTH levels properly and might improve the vital prognosis, as shown in Fig. 3D, in patients undergoing maintenance HD. cIMT (p = 0.029) and cf-PWV (p = 0.024), parameters of atherosclerosis and aortic stiffness, respectively, were significantly lower in the LCD group than in the HCD group.
No data.
5 of 64 died in the LCD group, 13 of 64 died in the HCD group. Kaplan-Meier curve revealed significancy. p = 0.041.

Ok, 2016
The progression rate of CACS was significantly slower in the LCD group than in the HCD group.
Iliac crest bone biopsies showed that both bone turnovers and volumes were significantly higher in the LCD group than in the HCD group at 24 months. P < 0.05. ALP (U/L) shown as time-averaged data during 24 months (mean ± SD): LCD = 139 ± 72; HCD = 101 ± 52; p = 0.2217. 31 of 212 died in the LCD group, 37 of 213 died in the HCD group.

Lu, 2016
The average IMT of the carotid artery, RI and AACS scores were notably decreased at 6 and 12 months in the LCD group. After 1 year of hemodialysis, the BMD of the lumbar spine assessed by QCT was significantly reduced in the LCD group, but not in the HCD group. ALP (IU/L) at 1 year (mean ± SD): LCD = 152 ± 38; HCD = 138 ± 38; p value not available.
Nobody died in either group.  Our study had several major limitations. First, the number of RCTs was small, and important data were unavailable in some of the RCTs, reducing the validity of the outcomes of the present meta-analysis. Second, the risks of bias of the included studies were judged to be mostly high or unclear, as shown in Fig. 2. For example, it may be difficult to blind medical doctors to knowledge regarding which dCa their patients received during long-term HD treatment when evaluating performance bias in these kinds of RCTs. Third, the serum Ca levels were corrected according to the serum albumin levels in some RCTs, but not in others, as shown in Table 1. Finally, the present meta-analysis had a moderate or high level of heterogeneity probably because of the variety in clinical backgrounds of the patients included in each RCT, as shown in Table 1. This factor might have led to heterogeneities in the present meta-analysis that cannot be ignored.
In conclusion, the present meta-analysis shows that the use of a low dCa (1.125 or 1.25 mmol/L) significantly reduces the serum Ca levels and raises the serum iPTH levels, compared with the use of a high dCa (1.50 or 1.75 mmol/L), in patients undergoing conventional maintenance HD for 6 months or longer. The use of a low dCa seems to have favorable effects on the risk of CVD in these patients, but additional RCTs are needed to determine which dialysate is optimal for preventing the progression of CVD and/or CKD-MBD among patients receiving long-term conventional maintenance HD treatments.

Methods
Patients, interventions and outcomes. We compared the clinical outcomes of RCTs examining patients undergoing conventional maintenance HD for at least 6 months using either a dCa of 1.25 mmol/L or lower or a dCa of 1.50 mmol/L or higher. A total of 622 patients (n = 322 in the LCD group, n = 300 in the HCD group) were included in the present systematic review. Our primary outcomes were the mean differences in the serum Ca, P and iPTH levels at the end of each trial, and our secondary outcomes were the occurrences of CVD, CKD-MBD and deaths from any causes during each trial period.
Search strategy and eligible criteria. We searched for eligible studies among all papers published prior to July 31, 2017, without any language restrictions using the PubMed and Cochrane Central Register of Controlled Trials (CENTRAL) databases, in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines 33 . Two authors (M.Y. and K.H.) performed the database searches independently. Each discrepancy was discussed until a consensus was reached. Our eligibility criteria were as follows: (1) the study was designed as an RCT; (2) the participants were randomized between an LCD group (dCa of 1.125-1.25 mmol/L) and an HCD group (dCa of 1.50-1.75 mmol/L); (3) the participants were undergoing conventional maintenance HD; and (4) the study period was 6 months or longer. Our terms for the PubMed database search were as follows: ("hemodialysis"

Data extraction.
From the included RCTs, we extracted the following necessary information: first author's name; publication year; dCa; number of patients; baseline characteristics of patients; study period; medications used during the study period; levels of serum Ca, P and iPTH at the end of the study; surrogate markers (if any) for CVD and CKD-MBD; and mortality during the study period.
Risk of bias. The risk of bias for each eligible RCT was assessed using the Cochrane Collaboration's tool 18 and was categorized as a high, low or unclear bias by two authors (M.Y. and K.H.) acting independently, as shown in Fig. 2. Each discrepancy was discussed until a consensus was reached.
Data analysis and statistics. The meta-analyses were conducted using random-effects models because the presence of heterogeneity across the RCTs was expected. The heterogeneity was estimated using the I² statistic 33 and was categorized as low if I² was 0-25%, moderate if I² was 25-75%, or high if I² was 75-100% 34 . The MD and RR with 95% CI data were pooled, and forest plots were drawn using Review Manager, version 5.3. The Begg's and Egger's tests 19,20 were conducted, and a funnel plot was drawn using R software, version 3.4.0 (using the reiv function in the metaphor package), to assess the publication bias. A p value below 0.1 was considered statistically significant in both the Begg's and Egger's tests. The units for the serum Ca and P levels were unified using the following formulas: mg/dL × 0.2495 = mmol/L for Ca; mg/dL × 0.3229 = mmol/L for P. The standard error (SE) was converted to the standard deviation (SD) using the following formula: = SE SD/ n , where n means the sample size of the study. Data availability statement. All data analyzed during this study are available on reasonable request to the corresponding author.