Citric Acid-Based Bicarbonate Dialysate Attenuates Aortic Arch Calci�cation in Maintenance Haemodialysis Patients: An Observational Study

Introduction: The progression of aortic calci�cation is associated with mortality in haemodialysis patients. Blood calciprotein particle (CPP) levels are associated with coronary artery calci�cation, and was reported to be inhibited when using citric acid-based bicarbonate dialysate (CD). Therefore, we examined the effect of CD on the progression of the aortic arch calci�cation score (AoACS) and blood CPP levels in haemodialysis patients. Methods: A 12-month retrospective observational study of 262 haemodialysis patients who met the eligibility criteria was conducted at Joban Hospital. Patients taking warfarin or bisphosphonates and those with baseline AoACS of 100% were excluded. Progression, de�ned as ΔAoACS (12 months – baseline) > 0%, was compared between the CD and acetic acid-based bicarbonate dialysate (AD) groups. Results: The CD group had signi�cantly lower AoACS progression than the AD group (P = 0.037). ΔCPP and ΔAoACS were not correlated in the AD group (R 2 = 0.030, P = 0.098), but were negatively correlated in the CD group (R 2 = 0.065, P = 0.022). Multivariate logistic analysis showed that the CD (odds ratio [OR] 0.51, 95% con�dence interval [CI] 0.27 ‒ 0.97, P = 0.042) was signi�cantly associated with the AoACS progression. Conclusion: CD may suppress the progression of vascular calci�cation in haemodialysis patients. low density lipoprotein cholesterol; HDL-c, high density lipoprotein cholesterol; TG, triglycerides; AoACS, aortic arch calci�cation score; RAS inhibitors, renin ‒ angiotensin system inhibitors; MAP, mean arterial blood pressure; I, intermittent infusion hemodia�ltration; O, online hemodia�ltration; HD, haemodialysis; Cre, creatinine; BMI, body mass index; Hb, haemoglobin; FGF23, �broblast growth factor-23; CPP, calciprotein particle. albumin;


Introduction
The main cause of death in patients with chronic kidney disease (CKD), particularly in those undergoing maintenance haemodialysis, is heart failure and cardiovascular disease due to a high degree of systemic vascular calci cation and atherosclerosis. In fact, the mortality rate in patients with CKD is extremely high compared to that for the general population. 1 The aortic arch calci cation score (AoACS), based on a simple chest radiography, correlates well with the aortic arch calci cation volume determined by a multidetector computed tomography (R = 0.635, P < 0.001). 2 Thus, this is a useful tool as it can be easily evaluated in the clinical setting. Moreover, the presence of aortic 3 or coronary artery calci cation 4 and progression of aortic arch calci cation 5,6 are reported to be associated with all-cause and cardiovascular mortality. In contrast, clinical studies have reported an association of blood calciprotein particle (CPP) with coronary artery calci cation, 7 atherosclerosis, and markers of in ammation in blood. 8 CPP in the blood has a physiological effect of transporting calcium-phosphorus, the raw material for teeth and bones; however, in patients with CKD, the amount of primary CPP in the blood is excessive or the dynamics of CPP are abnormally changed, resulting in the formation of secondary CPP, which has a pathological effect. 9,10 Our in vitro experiments show that citric acid inhibited CPP production as measured by the gel-ltration method. 11,12 We therefore hypothesised that the use of citric acid-based bicarbonate dialysate (CD) would inhibit the progression of vascular calci cation by decreasing blood CPP levels in patients with CKD. Here, we aimed to examine the effect of CD on the progression of AoACS and blood CPP levels in patients on maintenance haemodialysis.

Results
Patient and baseline characteristics Figure 1 shows the ow of participants through the study. Of the 262 patients who met the eligibility criteria, 19 patients in the AD group (bisphosphonate: 1, warfarin: 17, AoACS = 100%: 1) and 8 patients in the CD group (bisphosphonate: 1, warfarin: 7) were excluded, resulting in the analysis of 130 patients in the AD group and 105 patients in the CD group. Table 1 shows the characteristics at baseline along with the results of comparing the respective characteristics between the two groups. Table 1 Baseline (0 month) characteristics of the AD and CD groups (n = 235). Pi (mg/dL) 5.0 (4.1-6.1) 4.5 (3.9-5.5) 5.6 (4.5-6.7) < 0.001 cCa (mg/dL) 9.0 (8.5-9.4) 9.0 (8.7-9.5) 8. 9  Outcomes Table 2 shows the characteristics after 12 months, along with the results of comparing the respective characteristics between the two groups. At the end of the 12-month observation period, AoACS had progressed in 71 of the 130 patients in the AD group (54.6%) and 43 of the 105 patients in the CD group (41.0%) (P = 0.037) (  (Table 3).   Propensity score matching Subgroup analysis showed an interaction only in the comparison of the main effect between AoACS less than 19% and 19% or more. Based on previous report 5 and our results, the patients were divided into the following groups according to their baseline AoACS: Grade 0; AoACS = 0%, grade 1; AoACS 0% and ≦ 25%, grade 2; AoACS 25% and ≦ 50%, grade 3; AoACS 50%. Of these, the mild to moderate calci cation group with grade 2 or less was included in the study, and a comparison using propensity score matching between the two groups of AD and CD was performed. As a result, 15 patients in the AD group and 4 patients in the CD group with AoACS higher than 50% at baseline were excluded, and 115 patients in the AD group and 101 patients in the CD group were selected (Fig. 1.).
After 1:1 propensity score matching, 37 patients were extracted from each group (Fig. 1). We then compared the characteristics at baseline (Table 4) and after 12 months (Table 5). When examining the presence or absence of AoACS progression between the two groups of 115 patients in the AD group and 101 patients in the CD group, which consisted of the mild to moderate calci cation group with AoACS of 50% or less, there was signi cantly less progression in the CD group, with 66 patients (57.4%) in the AD group and 42 patients (41.6%) in the CD group (P = 0.020) ( Table 5). Analysis of the presence of AoACS progression after propensity score matching between the two groups showed that AoACS progressed in 27 of the 37 patients in the AD group (73.0%) and 14 of the 37 patients in the CD group (37.8%) (P = 0.002) ( Table 5). Table 4 Baseline (0 month) characteristics of the AD and CD groups before and after propensity score matching (n = 216).  Although the conventional level of statistical signi cance was not reached, the simple subtraction 12-month AoACS  (Table 5). Unexpectedly, the blood phosphorus levels were signi cantly higher in the CD group at 12 months, both before (P = 0.001) and after (P = 0.030) propensity score matching (Table 5).
Relationship between AoACS and blood CPP (total n = 235) Correlation coe cients for ΔCPP and ΔAoACS were analyzed for 130 patients in the AD group and 105 patients in the CD group after excluding patients with AoACS = 100%, warfarin, and bisphosphonate use. The correlation between ΔCPP and ΔAoACS was examined in the AD (R 2 = 0.030, P = 0.079) and CD groups (R 2 = 0.030, P = 0.079), respectively, but there was no correlation in both groups (Fig. 2a, b).

Relationship between AoACS and blood CPP (total n = 216)
Correlation coe cients for ΔCPP and ΔAoACS were analyzed for 115 patients in the AD group and 101 patients in the CD group after excluding patients with AoACS > 50%, warfarin, and bisphosphonate use. The AD group showed a non-signi cant but a weak positive correlation between ∆AoACS and ∆CPP (R 2 = 0.030, P = 0.098) (Fig. 3a), while the CD group showed a signi cant negative correlation between these factors (R 2 = 0.065, P = 0.022) (Fig. 3b).

Discussion
In this study, we examined the effect of CD on the progression of the AoACS (AoACS progression as de ned by ΔAoACS > 0%) and blood CPP levels and showed that the use of CD attenuated blood CPP and suppressed the progression of vascular calci cation in patients on maintenance haemodialysis.
The physiological role of primary CPP is to transport calcium-phosphorus, a raw material for teeth and bone, to these tissues, without it precipitating in soft tissue. 9,10 The uremic environment may cause transformation of primary CPP to secondary CPP to be phagocytosed by the macrophages in the vessel wall, which may contribute to atherosclerosis 13 and vascular calci cation in the tunica media. 14 In fact, it has been reported that CPP levels measured by the gel-ltration method were positively correlated with coronary plaque assessed by intravascular ultrasound, even after adjusting for a number of confounding factors. 15 Due to this positive association between blood CPP and arterial calci cation in CKD, the Ca-chelating effect of citric acid in CD may have inhibited the formation of blood secondary CPP, which in turn inhibited the progression of vascular calci cation. In the present study, ΔCPP and ΔAoACS were negatively correlated in the CD group (Fig. 2b), suggesting that citric acid may have inhibited the formation of primary to secondary CPP and promoted the incorporation of primary CPP into bone.
There is a lack of consensus on whether FGF23 in blood is associated with vascular calci cation in CKD patients. 16,17 In our study, we did not nd any signi cant association between blood FGF23 and AoACS (Fig. 2c, 2d). From the viewpoint that FGF23 is a phosphaturic hormone, in the early stage of CKD, when renal function is preserved, FGF23 acts on the kidney to produce Pi-diuretic effect and inhibition of vitamin D activation, and thus FGF23 is thought to have a protective effect against vascular calci cation. However, in CKD patients undergoing maintenance hemodialysis, renal function is abolished, so the effects of Pi-diuresis and vitamin D inhibition in the kidney are not obtained and are not associated with vascular calci cation. In the present study, we did nd a weak but signi cant positive correlation between ΔCPP and ΔFGF23 in blood (Fig. 2e, 2f). FGF23 in blood is probably produced by osteoblasts and osteocytes upon primary CPP stimulation, while secondary CPP acts on vascular calci cation because it cannot pass through bone sinuses due to its large size. 12,18 As compared to AD, CD has the potential to improve prognosis in patients on maintenance haemodialysis. 19 This may be due to the inhibition of secondary CPP production by citrate. 12 CPP is thought to induce FGF23 secretion, 12,18 and FGF23 is associated with mortality and cardiovascular disease in patients on maintenance haemodialysis. 20 An in vivo rat study has reported that recombinant FGF23 directly causes cardiac hypertrophy and that there was a signi cant positive correlation between blood FGF23 and cardiac hypertrophy assessed by echocardiography in patients with CKD. 21 Similar to our ndings, the use of CD has previously been reported to inhibit CPP ripening, as indicated by T50 (T50 is 1/2 of the time it takes for the turbidity of primary CPP in the blood to rapidly increase to secondary CPP). 22,23 Based on previous reports, primary CPP excess, in other words, excess of both Pi and Ca in the blood is thought to be associated with worse prognosis via FGF23 secretion, and secondary CPP is thought to be associated with worse prognosis via vascular calci cation.
Another reason for the greater e cacy of CD than AD in suppressing arterial calci cation may be related to the following: An ex vivo model showed that citric acid, at concentrations of 733 µmol/L or higher, inhibited rat aortic ring calci cation. 24 The post-dialysis blood citrate concentration in haemodialysis patients using CD has been reported to be 771.6 ± 184.3 µmol/L, which was approximately ve times higher than that of AD. 24 Moreover, bicarbonate promoted rat aortic ring calci cation in a concentration-dependent manner. 24 The use of AD in haemodialysis patients was reported to result in a higher post-dialysis blood bicarbonate concentrations and lower citric acid concentrations than those obtained with CD. 24 Moreover, it was reported that no excess bicarbonate correction was observed after dialysis with CD. 25 Switching from AD to CD was reported to signi cantly reduce malondialdehyde-modi ed LDL(MDA-LDL) (an indicator of lipid oxidation) and pentosidine (an indicator of glycoxidation) level in the blood. 26 Therefore, it is likely that CD inhibits lipid oxidation, which is associated with coronary artery calci cation, 27 as well as glycation, which causes atherosclerosis 28 in patients on maintenance haemodialysis, thereby exerting anti-atherosclerotic effects. In subgroup analysis, CPP dissolution from bone was enhanced in patients with secondary hyperparathyroidism, 29 suggesting a signi cant bene t of CD use in patients with controlled PTH (iPTH 132 pg/mL) (P = 0.036) (Fig. 3). If Pi is controlled (Pi ≦ 6 mg/dL), PTH will also be suppressed, and the anti-calci cation effect of CD use will be more effective (P = 0.008). Typically, patients on maintenance haemodialysis have chronic in ammation due to the uremic environment, low nutrition, the dialysis equipment used, and the use of AD. 30 In vitro analysis showed that CD signi cantly inhibited the induction of oxidative stress in the blood of these patients, as compared to AD. 31 Thus, CD is less in ammatory than AD and may inhibit the transition to a vicious cycle of atherosclerosis (i.e. malnutrition-in ammation-atherosclerosis syndrome). Therefore, if there is persistent in ammation in hemodialysis patients, AD may further induce in ammation, and CD may reduce in ammation. Subgroup analysis showed that CD was signi cantly more favorable than AD in inhibiting the progression of vascular calci cation when CRP levels were ≧ 0.1 mg/dL (P = 0.041). Under the conditions of low Ca levels and no CaCO3 use, the chelating effect of citric acid on Ca ions was more effective, and CD was thought to have a more anti-calci cation effect than AD (P = 0.044 under cCa < 9 mg/dL, P = 0.036 without CaCO3 use). CD had a signi cant anti-calci cation effect compared to AD when the AoACS was < 19 (%). The reason for this is that it has been reported that when calci cation is already highly advanced, treatment is unlikely to have an inhibitory effect on progression. 32 Warfarin induces vascular calci cation by inhibiting matrix Gla protein, an inhibitor of calci cation in arteries, in both models of CKD 33 and in clinical trials of patients on haemodialysis. 34 Therefore, patients on warfarin were excluded from our study. Patients using bisphosphonates were also excluded as bisphosphonate use reportedly reduced the coronary artery calci cation scores in patients on maintenance haemodialysis 35 and inhibited rat aortic calci cation by reducing bone-derived CPP. 29 There are several limitations in the present study. First, we did not measure blood Mg levels. Administration of magnesium oxide (MgO) signi cantly inhibited the progression of coronary artery calci cation in patients with CKD, 36 and basic experiments in vascular smooth muscle cells showed that Mg dose-dependently inhibited calci cation, 37 Mg citrate treatment reduced the carotid intima-media thickening in haemodialysis patients, 38 and elevation of Mg concentration in the dialysate prolonged the T50. 39 Thus, blood Mg levels and oral medications containing Mg are associated with vascular calci cation and atherosclerosis in dialysis patients. Nevertheless, none of the patients in this study were receiving oral medications containing Mg. Second, blood bicarbonate and citrate levels may be associated with vascular calci cation, but these were not measured in this study. Third, observational study is not of high quality as a research method, and the number of patients included in this study was small.
In conclusion, for patients on maintenance haemodialysis, CD is more effective than AD in inhibiting the progression of vascular calci cation as assessed by ΔAoACS. Larger randomised controlled trials are warranted to validate the anti-atherosclerotic results obtained with the use of CD, and to elucidate the mechanism of this therapeutic effect.

Study design and population
This was a retrospective observational study. We used a 12-month observation period, from October 2017 to October 2018. Eligibility criteria were as follows: 1) having undergone maintenance haemodialysis at Joban Hospital (Fukushima, Japan); 2) patients who, upon introduction of dialysis treatment, have given written informed consent to save the remaining plasma collected for routine testing and subsequently use it for research; 3) those on continuous outpatient maintenance dialysis for 12 months; 4) those who had available data; 5) ≥ 18 years of age; 6) no active cancer; and 7) those stable for at least 3 months after dialysis induction at baseline. Patients using warfarin or bisphosphonates and those with a baseline AoACS of 100% were excluded.
We have obtained written informed consent for renal biopsy from all patients, and clinical data were obtained from all recently recruited patients, who had the opportunity to opt-out from this study.

Treatment interventions
Page 16/23 The dialysis centre had two oors; originally, all patients on maintenance haemodialysis were on AD. Patients on the 1st oor remained on AD, while patients on the 2nd oor were assigned to start CD from October 2017. AoACS was compared between two groups: one in which AD (Kindary 4E®; Fuso Pharmaceutical Co., LTD, Osaka, Japan) remained unchanged and the other in which AD was changed to CD (Carbostar®; Ajinomoto Pharmaceutical Co., LTD, Tokyo, Japan). All relevant investigators were blinded to the treatment groups.

Comparison method and outcomes
AoACS progression was de ned as ΔAoACS (12-month AoACSbaseline [0-month] AoACS) > 0% over the 12month observation period and was compared between the AD and CD groups. Thus, ΔAoACS = 0 was interpreted as no AoACS progression during the 12-month observation period, and ΔAoACS < 0 was interpreted as improvement in AoACS.

Evaluation scale and evaluator
Chest X-ray imaging was performed after dialysis, and the X-ray image was divided into 16 sections using the previously described method. 2 The presence of one calci cation was calculated as 1 point divided by 16 sections × 100 (%) = 6.25 (%); thus, 1 point = 6.25%. A clinical engineering technician at Joban Hospital performed the AoACS measurements and was available for consultations at the Department of Radiology at Tokyo Women's Medical University as needed.

Sample collection
In the non-fasting state, blood samples were taken at the beginning of the week, 2 days after the last dialysis, and were obtained before beginning the dialysis procedure. Isolated plasma samples were promptly stored at − 80°C

Ethics
This study adhered to the tenets of the Declaration of Helsinki and was approved by the Medical Ethics Committee at the Joban Hospital, Iwaki-city, Fukushima, Japan (# 5104). Since this was an observational study and no invasive procedures were performed, patient consent was obtained in an opt-out format for the use of residual plasma in regular practice.

Statistical analyses
Data were expressed as means ± standard deviations if normally distributed; otherwise, they were reported as median and interquartile range (25-75 percentile) for non-normally distributed data. The baseline characteristics of the two groups were compared using analysis of variance as a parametric or a nonparametric test. Comparisons between proportions were made using chi-square tests. In case of two-group comparisons of normally distributed data, equal variance tests were performed for quantitative variables, by Student's t-test for equal variances, and Welch's t-test for unequal variances. For two-group comparisons of non-normally distributed data, Wilcoxon's signed-rank test was used for quantitative variables. The baseline characteristics of the two groups were subjected to univariate and multivariate logistic regression analyses to assess the impact of factors on AoACS progression.
A subgroup analysis was used to examine the characteristics involved in the suppressive effect of CD on AoACS progression. As this was an observational study and prone to various confounders, we performed 1:1 propensity score matching to adjust for covariates and estimate causal association. We created a model that adjusted for all baseline characteristics (Table 4) as covariates, for the outcome of AoACS progression due to AD and CD treatment interventions. The model was used to perform propensity score matching. The analyses were performed with JMP version 15 (SAS Institute Inc., Cary, NC).

Figure 2
Relationship between blood CPP and AoACS and blood FGF23 (n = 235). a) Correlation of ΔCPP and ΔAoACS in the AD group. b) Correlation of ΔCPP and ΔAoACS in the CD group. c) Correlation of ΔFGF23 and ΔAoACS in the AD group. d) Correlation of ΔFGF23 and ΔAoACS in the CD group. e) Correlation of ΔFGF23 and ΔCPP in the AD group.
Note: Statistically signi cant P-values are in bold.