Optimal Phosphate Control Related to Coronary Artery Calcification in Dialysis Patients

Significance Statement

Hyperphosphatemia has been reported to be associated with severity of coronary artery calcification (CAC), a predictor of all-cause mortality in incident patients on hemodialysis. However, the optimal phosphate range in such patients remains unknown. The authors conducted a randomized study to compare the effects on CAC progression of two types of noncalcium-based phosphate binders (sucroferric oxyhydroxide or lanthanum carbonate) and of two different phosphate target ranges. The percentage change in CAC score in a strict phosphate control group (3.5–4.5 mg/dl) was significantly lower than that in a standard phosphate control group (5.0–6.0 mg/dl). The phosphate binders did not differ in their effects on CAC progression. Further study with a larger sample size is needed, but strict phosphate control shows promise for delaying CAC progression in patients on hemodialysis.

Visual Abstract

Abstract

Background

In patients on maintenance dialysis, cardiovascular mortality risk is remarkably high, which can be partly explained by severe coronary artery calcification (CAC). Hyperphosphatemia has been reported to be associated with the severity of CAC. However, the optimal phosphate range in patients on dialysis remains unknown. This study was planned to compare the effects on CAC progression of two types of noncalcium-based phosphate binders and of two different phosphate target ranges.

Methods

We conducted a randomized, open-label, multicenter, interventional trial with a two by two factorial design. A total of 160 adults on dialysis were enrolled and randomized to the sucroferric oxyhydroxide or lanthanum carbonate group, with the aim of reducing serum phosphate to two target levels (3.5–4.5 mg/dl in the strict group and 5.0–6.0 mg/dl in the standard group). The primary end point was percentage change in CAC scores during the 12-month treatment.

Results

The full analysis set included 115 patients. We observed no significant difference in percentage change in CAC scores between the lanthanum carbonate group and the sucroferric oxyhydroxide group. On the other hand, percentage change in CAC scores in the strict group (median of 8.52; interquartile range, −1.0–23.9) was significantly lower than that in the standard group (median of 21.8; interquartile range, 10.0–36.1; P=0.006). This effect was pronounced in older (aged 65–74 years) versus younger (aged 20–64 years) participants (P value for interaction =0.003). We observed a similar finding for the absolute change in CAC scores.

Conclusions

Further study with a larger sample size is needed, but strict phosphate control shows promise for delaying progression of CAC in patients undergoing maintenance hemodialysis.

Clinical Trial registry name and registration number:

Evaluate the New Phosphate Iron-Based Binder Sucroferric Oxyhydroxide in Dialysis Patients with the Goal of Advancing the Practice of EBM (EPISODE), jRCTs051180048

Patients on dialysis have a high mortality rate due to cardiovascular disease (CVD). Coronary artery calcification (CAC) contributes to CVD, and the involvement of hyperphosphatemia in CAC has been well recognized.¹ The severity of CAC was reported to be a predictor of all-cause mortality in incident patients on hemodialysis.² Several observational studies have demonstrated that phosphate binder use is associated with better survival in patients on dialysis.^3–5 Noncalcium-based phosphate binders, including lanthanum carbonate (LC)^6–8 or sevelamer,⁹ were found to be more effective in attenuating the progression of CAC than calcium-based phosphate binders, suggesting that noncalcium-based phosphate binders could slow the calcification and improve the prognosis in patients on dialysis. LC was shown to block gastrointestinal calcium channel (transient receptor potential vanilloid [TRPV] 5/6 channels), thereby decreasing gastrointestinal calcium absorption compared with sevelamer.¹⁰ Sucroferric oxyhydroxide (SO), a novel noncalcium-based phosphate binder, also has a potent serum phosphate–lowering effect; similarly, treatment with this drug is expected to slow the progression of vascular calcification.^11,12 Recently, SO, as an iron-based phosphate binder, showed a significant increase in transferrin saturation (TSAT) and hemoglobin (Hgb) levels, especially in patients on dialysis with low ferritin levels.¹³ The Japanese guidelines are still conservative in the prescription of iron, advocating lower target ranges of iron parameters than Western countries.¹⁴ As a result, median ferritin level was only 73 ng/ml in Japanese patients on hemodialysis.¹⁵ Thus, administration of SO was consistently reported to improve anemia along with the increase in ferritin and TSAT levels in Japanese patients on dialysis,^16,17 although the effect of supplementing iron by this agent is reportedly weak in Western patients on dialysis. It has been demonstrated that iron loading suppressed vascular calcification in the uremic rat model, reducing the expression of runt-related transcription factor 2 and phosphate transporter 1 in the aorta.¹⁸ Therefore, an iron-based phosphate binder might have a beneficial effect on CAC scores in Japanese patients on dialysis, independent of the phosphate-decreasing effect. However, there have been no randomized studies to determine whether the choice of noncalcium-based phosphate binder, LC or SO, affects clinical outcomes in patients on dialysis. Therefore, we compared the effects of these two types of noncalcium-based phosphate binders on CAC progression.

Although CKD–mineral and bone disorder guidelines by Kidney Disease Improving Global Outcome (KDIGO) recommend lowering elevated phosphate levels toward the normal range in patients with CKD stages 3A to 5D, there have been no randomized trial data demonstrating that lowering serum phosphate levels will improve any outcomes in patients on dialysis.¹ Furthermore, the optimal target level of serum phosphate also remains uncertain. To answer these clinical questions, we conducted the Evaluate the New Phosphate Iron-Based Binder Sucroferric Oxyhydroxide in Dialysis Patients with the Goal of Advancing the Practice of EBM (EPISODE) trial, a randomized trial using a two by two factorial design that allows comprehensive investigation of the effects of both factors, type of noncalcium-based phosphate binder (LC versus SO) and different phosphate target ranges (standard treatment versus strict treatment for serum phosphate–level control), on CAC.

Methods

Study Design

The clinical trial EPISODE was a randomized, open-label, multicenter, interventional trial with a two by two factorial design (Japan Registry of Clinical Trials no. jRCTs051180048, UMIN identification: UMIN000023648).¹⁹ In this trial, patients on hemodialysis between the ages of 20 and 80 years who were stable on dialysis for at least 3 months with a predialysis serum phosphate level of at least 5.0 mg/dl or at least 6.1 mg/dl in those taking or not taking phosphate binders, respectively, as measured during the observation period were enrolled. Detailed inclusion and exclusion criteria are provided in Supplemental Table 1. Key exclusion criteria were severe hyperparathyroidism (a predialysis serum intact parathyroid hormone [PTH] level of >800 pg/ml) or iron excess (a predialysis serum ferritin level of >300 ng/ml or TSAT of >50%).

The trial was conducted in accordance with the Helsinki Declaration and was approved by the Institutional Review Board at Osaka University Graduate School of Medicine. Written informed consent was obtained from all enrolled patients. During the screening phase, patients on dialysis underwent multidetector computed tomography (MDCT) at a designated medical institution specialized in MDCT testing, as well as endocrine and special testing. Only patients with CAC scores >30 during the screening phase were enrolled. Enrolled patients received either LC or SO for 12 months according to the randomization result to reduce serum phosphate to the assigned target ranges. Enrolled patients also underwent MDCT, as well as endocrine and special testing, in the 12th month of the treatment period.

Randomization and Treatment Intervention

Using the Electronic Data Capture (Forum PLUS) software platform, randomization was performed within 1 week of enrollment. Enrolled patients were randomly assigned at a 1:1 ratio to receive LO or SC for 12 months (factor 1) to reduce serum phosphate levels to two target ranges (3.5 mg/dl≤ serum phosphate <4.5 mg/dl in the strict treatment arm or 5.0 mg/dl≤ serum phosphate <6.0 mg/dl in the standard treatment arm; factor 2) by the dynamic allocation (minimization method) for achieving overall balance across the covariates, including age at consent (≥20 and ≤64 years old and ≥65 and ≤79 years old), CAC scores measured during the screening period (≥30 and ≤400, ≥401 and ≤1000, or ≥1001), and investigational site. Enrolled patients received the allocated drug for 12 months starting within 1 week of registration. The use of phosphate binders other than LC and SO was prohibited during the treatment period. However, when the target serum phosphate level was not achieved, concomitant use of sevelamer hydrochloride or bixalomer was allowed. Switching between different types of erythropoiesis-stimulating agent (ESA) formulations was not allowed during the treatment period.

End Points

The primary end point was the percentage change in CAC scores (%CAC change). Secondary end points were (1) the absolute change in CAC scores; (2) changes in serum phosphate, calcium, and calcium-phosphate product and the ratio of patients who reached the target serum phosphate range at the end of the treatment period; (3) changes in the levels of intact PTH, intact fibroblast growth factor 23 (FGF-23), and C-terminal FGF-23; (4) effects on renal anemia assessed by TSAT, Hgb, and the changes in the dose of intravenous iron and ESA administration; (5) relationship between the change in serum phosphate level and that in CAC scores; (6) average doses of the phosphate binders required to achieve the target serum phosphate range; and (7) safety issues (adverse drug reactions). Imaging data scanned by MDCT with at least 16 rows were submitted as electronic data in Digital Imaging and Communication in Medicine format on either CD-R or DVD-R to the MDCT committee. The CAC scores were measured on the basis of the Agatston et al.²⁰ method at a center hospital independently by two radiologists of the MDCT committee who were blinded to drug/phosphate target assignment, and another two radiologists re-evaluated the CAC scores and decided which was correct in case of disagreement. The MDCT committee discussed the validity of the calculated CAC scores and determined the final CAC scores by consensus. The concordance rate assessed by two radiologists was quite high (Supplemental Figure 1).

Target Sample Size

The target number of enrolled patients was 200 (50 patients per arm ×4 arms). Rationale for the sample size¹⁹ was as follows. Our past research showed that the median %CAC change was 20.3% in incident patients on dialysis with serum phosphate level ˂4.5 mg/dl and 73.0% in those with serum phosphate level ≥5.0 mg/dl.⁶ The probability that %CAC change in the group with serum phosphate level <4.5 mg/dl would be less than that in the group with serum phosphate level ≥5.0 mg/dl was estimated to be 0.77. On the basis of the above results, the probability that %CAC change was lower in the strict group than in the standard group was conservatively estimated to be 0.70; therefore, when a two-sided Wilcoxon rank sum test with a significance level of 2.5% was used, the required sample size will be 63 patients per arm to achieve a desired power of at least 95%. On the other hand, no previous research data were available to define the expected %CAC change for the LC group and SO group; thus, we assumed a 0.675 probability that %CAC change was lower for the SO group than the LC group, and the required sample size was 83 patients per arm to have at least 95% power. Therefore, the larger sample size, 83 patients per group (166 patients for both arms), was the required sample size. Assuming that approximately 15% of all patients were excluded from full analysis set (FAS), the target number of registered patients would be 100 patients per factor arm (a total of 200 patients).

Statistical Analyses

Primary efficacy analyses were conducted for all enrolled patients who had available data for the second CAC measurements (FAS). Safety analyses were conducted in enrolled patients who took the investigational drug. The primary objective of this study was to investigate the superiority of SO to LC or of strict treatment to standard treatment in terms of the primary end point, %CAC change. We hypothesized that the superiority of SO to LC was due to a factor, which is independent of the phosphate-lowering effect, such as iron supplementation. Therefore, we assumed no interaction between the choice of a binder and phosphate target range; however, we assessed the significance of the interaction. The significance level used for the tests was 0.05 (two sided), and missing data were not imputed.

For the primary end point, the Wilcoxon rank-sum test was used to test null hypotheses that %CAC change was equivalent between factors or arms, whereas a two-sided significance level of 2.5% was used in testing to control the overall studywise type 1 error rate to ≤5%. Subgroup analyses were conducted by age at the time of consent acquisition, CAC scores as measured during the screening period, sex, and status of phosphate binder use during the observation period. For the secondary end points, a linear mixed effects model was used to compare the changes in serum phosphate, calcium, and calcium-phosphate product between arms, whereas the chi-squared test was used to compare the proportion of patients who achieved the target serum phosphate range between arms. The Wilcoxon rank-sum test was used to compare changes in intact PTH, intact FGF-23, and C-terminal FGF-23 between arms. Other investigations included the comparison of the effect on renal anemia, the relationship between change in serum phosphate level and change in CAC scores, and comparison of the average doses of phosphate binders required to achieve the target serum phosphate range. For safety, the total number of adverse events and the number of individual adverse events were calculated by arm, and serious adverse events were tabulated.

Because 41 patients were excluded from the primary analysis, we performed sensitivity analyses regarding the primary efficacy analysis for factor 2 (Wilcoxon two-sample test) after complementing the missing values by multiple imputation. Because %CAC change was found to be strongly dependent on baseline CAC scores, missing values for %CAC change as the primary outcome were supplemented from patient characteristic variables, including baseline CAC scores as an explanatory variable for multiple imputation for patients who did not have the second CAC scores at 12 months. As a specific method of multiple imputation of CAC scores at 12 months, we used predicted mean matching instead of regression because the values of CAC scores at 12 months were not normally distributed. We used fully conditional specification methods to impute missing values for all variables, assuming the existence of a joint distribution for these variables.²¹ The rank sums and their variances were calculated for each dataset, and they were integrated and tested according to Rubin rules.²² Four different imputed datasets of completing missing values of the second CAC scores were created according to the presence or absence of the following two conditions: (1)  logarithmic transformation of skewed variables  (CAC scores, ferritin, intact PTH, and intact FGF-23 at baseline and change in CAC scores as outcome) and (2) inclusion of the treatment allocation as the explanatory variables (factor 1 and factor 2) in the complementary model. Statistical analyses were performed using SAS version 9.4 (multiple imputation: PROC MI, FCS statement).

Ethical Approval

All procedures performed in the studies involving human participants were in accordance with the ethical standards of the Osaka University Hospital Clinical Research Committee (Institutional Review Board approval no. 16112-5), the Osaka University Clinical Research Review Committee (N18014), and the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Results

Study Population and Characteristics

Informed consent was obtained from 201 patients on dialysis who met the eligibility criteria.¹⁹ During the screening phase, 41 patients dropped out due to decreased serum phosphate levels below the criteria during the screening phase (n=6), baseline CAC scores ≤30 (n=27), past history of cancer (n=1), stent replacement (n=4), parathyroidectomy (n=2), and withdrawal of informed consent (n=1). Among patients with CACs≤30, 9% had diabetes as a comorbidity, whereas 43% of patients with CACs>30 had diabetes. A total of 160 patients were enrolled and randomized, but four patients were excluded due to not taking an assigned drug at all because of transfer to other hospitals or withdrawal by patients’ family just after enrollment, as shown in Figure 1. Thus, 156 patients were eventually assessed in the safety analysis set. Of the enrolled patients, 39 patients were excluded due to discontinuation of the assigned drug by adverse events (n=17), physicians’ judgements (n=4), or withdrawal (n=18). In addition, two patients were excluded because CAC scores could not be evaluated due to many artifacts in MDCT images at 12 months; thus, 115 patients were analyzed for efficacy as the FAS.

Figure 1.

Flow chart of enrollment and randomization. A total of 156 patients were randomly assigned to receive LC or SO (factor 1) in order to reduce serum phosphate levels to two target levels (strict or standard; factor 2). PPS, per protocol set.

Baseline characteristics for the FAS were similar across the groups, and there were no significant differences between FAS-analyzed patients (n=115) and excluded patients (n=41) (Tables 1 and 2, Supplemental Table 2). The prescription of active vitamin D and calcimimetics at baseline was well balanced across the groups (Tables 1 and 2). In addition, there were no significant differences in the distribution of the investigational sites or the dialysate calcium concentration across the groups. The median (interquartile range [IQR]) CAC scores of the FAS patients were 680 (IQR, 340–2418) in the LC group (n=62), 824 (IQR, 224–1513) in the SO group (n=53), 891 (IQR, 257–1581) in the standard group (n=57), and 711 (IQR, 319–2270) in the strict group (n=58). The baseline phosphate levels (mean ± SD) were 5.98±1.16 mg/dl in the LC group, 6.00±0.75 mg/dl in the SO group, 6.04±1.16 mg/dl in the standard group, and 5.95±0.80 mg/dl in the strict group. In addition, the baseline characteristics, including intact PTH, intact FGF-23, TSAT, and ferritin levels, were similar across the groups.

Table 1.

Patients characteristics and baseline data (FAS) stratified by two factors

Characteristics	Factor 1		Factor 2
	LC, n=62	SO, n=53	Standard Control, n=57	Strict Control, n=58
No. (women)	62 (21)	53 (17)	57 (20)	58 (18)
Age, yr	62.9±10.1	60.7±11.0	61.3±10.6	62.4±10.5
CACs	680 [340–2418]	824 [224–1513]	891 [257–1581]	711 [319–2270]
Phosphate, mg/dl	5.98±1.16	6.00±0.75	6.04±1.16	5.95±0.80
Calcium, mg/dl	8.74±0.70	8.91±0.69	8.90±0.67	8.73±0.71
Magnesium, mg/dl	2.56±0.35	2.64±0.37	2.59±0.39	2.61±0.33
TSAT, %	27.4±12.4	22.6±9.0	23.8±10.9	26.6±11.5
Ferritin, ng/ml	71 [39–126]	59 [29–135]	45 [30–114]	89 [35–146]
Intact PTH, pg/ml	137 [99–187]	124 [88–178]	132 [105–216]	130 [84–179]
Intact FGF-23, pg/ml	5225 [2595–12,325]	5275 [2258–10,600]	6840 [2772–13,800]	4395 [2070–10,325]
Past history of CVD, %	25.8	43.4	35.1	32.8
Diabetes mellitus, %	38.7	50.9	38.6	50.8
Dyslipidemia, %	45.2	43.4	45.6	43.1
Hypertension, %	90.3	84.9	91.2	84.5
Active vitamin D prescription, %	66.1	66.0	68.4	64.0
Calcimimetics prescription, %	35.5	32.1	38.6	29.3

Data are on the basis of full set analysis. Continuous variables are presented as mean ± SD or median [IQR]. Active vitamin D included oral calcitriol, alfacalcidol, and intravenous maxacalcitol and calcitriol.

Table 2.

Patients characteristics and baseline data (FAS) stratified by four groups

Characteristics	LC		SO
	Standard	Strict	Standard	Strict
No. (women)	28 (12)	34 (9)	29 (8)	24 (9)
Age, yr	63.4±10.2	62.5±10.1	59.3±10.8	62.3±11.3
CACs	1184 [362–2588]	569 [322–1984]	711 [175–1437]	825 [313–2454]
Phosphate, mg/dl	6.08±1.40	5.91±0.93	6.00±0.88	6.00±0.57
Calcium, mg/dl	8.86±0.71	8.63±0.68	8.94±0.65	8.87±0.75
Magnesium, mg/dl	2.47±0.31 (miss =13)	2.61±0.37 (miss =11)	2.67±0.43 (miss =8)	2.60±0.27 (miss =9)
TSAT, %	26.6±10.9 (miss =6)	28.1±13.7 (miss =9)	20.3±10.0 (miss =11)	24.7±7.71 (miss =4)
Ferritin, ng/ml	45 [40–104] (miss =11)	99 [38–149] (miss =11)	46 [25–125] (miss =9)	79 [31–139] (miss =10)
Intact PTH, pg/ml	137 [108–224]	136 [78–189]	124 [93–216] (miss =1)	121 [86–174]
Intact FGF-23, pg/ml	4810 [1805–12,505]	6185 [3380–12,700]	8285 [3380–13,900] (miss =1)	2500 [1215–5825]
Past history of CVD, %	25.0	26.5	44.8	41.7
Diabetes mellitus, %	35.7	41.2	41.4	62.5
Dyslipidemia, %	57.1	35.3	34.5	54.2
Hypertension, %	96.4	85.3	86.2	83.3
Active vitamin D prescription, %	67.9	61.8	75.9	66.7
Calcimimetics prescription, %	35.7	35.3	41.4	20.8

Data are on the basis of full set analysis. Continuous variables are presented as mean ± SD or median [IQR]. Active vitamin D included oral calcitriol, alfacalcidol, and intravenous maxacalcitol and calcitriol. miss, missing.

Changes in Laboratory Parameters

In FAS-analyzed patients, there were no significant differences in the decrease of serum phosphate levels between the LC group (0.57±1.45 mg/dl from baseline to 12 months) and the SO group (0.82±1.33 mg/dl) during the treatment period (P=0.56), and serum phosphate levels decreased to 5.13±1.44 mg/dl in the LC group and 5.07±1.10 mg/dl in the SO group at 12 months (Figure 2A). The mean (± SD) daily doses of LC and SO were 1590±694 and 1168±811 mg at 12 months, respectively, although the same doses of LC and SO were reported to have comparable phosphorus-lowering effects.^16,23 In factor 2, serum phosphate levels in the strict group decreased significantly to lower levels (4.68±1.26 mg/dl) than in the standard group (5.54±1.18 mg/dl) at 12 months, and the decrease of serum phosphate levels was significantly higher in the strict group (1.00±1.33 mg/dl) than in the standard group (0.37±1.41 mg/dl; P=0.05) (Figure 2B). The mean serum phosphate at 12 months in the strict group was above the upper limit (4.5 mg/dl) set by protocol because some patients could not tolerate the high doses of the assigned drug and/or add-on drugs due to digestive symptoms, etc. The proportions within the target range at 12 months (3.5 mg/dl≤ serum phosphate <4.5 mg/dl in the strict group and 5.0 mg/dl≤ serum phosphate <6.0 mg/dl in the standard group) were 46.7% in the LC group, 37.3% in the SO group, 37.5% in the standard group, and 30.0% in the strict group. The chi-squared test showed no differences in the proportions within the target range between the standard group and the strict group (P=0.39). There were no significant differences in the proportions of the concomitant use of sevelamer hydrochloride or bixalomer between the LC group (22 of 62 patients) and the SO group (18 of 53 patients), as well as between the standard group (15 of 57 patients) and the strict group (27 of 58 patients; P=0.94 and P=0.07, respectively). Patients using sevelamer/bixalomer received greater dosage of the assigned drug than those without concomitant use (Supplemental Table 3). In the strict group, 64% of patients received the maximum dose of the assigned drug, but others could not due to adverse effects, such as diarrhea.

Figure 2.

Changes in laboratory parameters. Shown are trends of the (A and B) serum phosphate levels, (C and D) serum calcium levels, and (E and F) calcium-phosphate concentration product in (A, C, and E) the LC versus SO group and (B, D, and F) the standard versus strict groups and (A) the dose of phosphate binder in the LC and SO groups. The error bars indicate SEMs.

There were no significant differences in serum calcium levels of FAS-analyzed patients during the treatment period between the LC group (8.80±0.72 mg/dl) and the SO group (8.73±0.49 mg/dl) at 12 months (P=0.12) (Figure 2C), as well as between the standard group (8.84±0.56 mg/dl) and the strict group (8.69±0.68 mg/dl; P=0.54) (Figure 2D). There was no significant difference in the calcium-phosphate product of FAS-analyzed patients between the LC group and the SO group during the treatment period (P=0.34) (Figure 2E), but the calcium-phosphate product was lower in the strict group than in the standard group at 12 months (P=0.17) (Figure 2F).

The changes in Hgb and TSAT levels of FAS-analyzed patients from baseline to 12 months were not significantly different between the standard group and the strict group (Supplemental Figures 1 and 2, A and B). However, although patients on dialysis treated with LC had slightly decreased Hgb (11.33±1.03 g/dl at baseline to 10.93±0.88 g/dl at 12 months, NS) and TSAT (27.42±12.36% to 24.01±13.13%, NS) levels, patients treated with SO showed increased Hgb (11.23±1.16 to 11.60±1.20 g/dl, P<0.001) and TSAT levels (22.61±9.04% to 30.03±10.45%, P<0.001). There were no significant differences in changes of Hgb or TSAT levels of FAS-analyzed patients from baseline to 12 months between the standard group and the strict group. The ratio of ESA doses (at 12 months-baseline) was not different between the LO group (1.32±1.20) and the SO group (1.37±0.85). There were no significant differences in changes of intact FGF-23 or intact PTH levels of FAS-analyzed patients from baseline to 12 months between the LC group and the SO group, as well as between the standard group and the strict group (Supplemental Figure 2, C and D). The prescription of calcimimetics at 12 months slightly increased compared with baseline, but no significant differences were observed across the groups (Supplemental Table 4).

End Points

There were no significant differences in %CAC change (assessed by [CAC scores at 12 months – baseline CAC scores]/baseline CAC scores ×100 [%]; median; IQR), as the primary end point between the LC group (13.7; IQR, 5.0–35.0) and the SO group (13.3; IQR, −2.2–26.7; P=0.48). On the other hand, %CAC change was significantly lower in the strict group (8.5; IQR, −1.0–23.9) than in the standard group (21.8; IQR, 10.0–36.1; P=0.006) (Figure 3A). Because 41 patients were excluded from the primary analysis, we performed sensitivity analyses with four different multiple imputed datasets regarding the primary efficacy analysis for factor 2. Sensitivity analyses with multiple imputation showed that %CAC change in the strict group was significantly lower under three of four conditions of multiple imputation than in the standard group (Supplemental Tables 5 and 6). The effect of strict phosphate control on %CAC change was not modified by factor 1 (LC or SO; P value for interaction =0.19). Similar to %CAC change, the absolute change in CAC scores (median; IQR) was significantly lower in the strict group (66.1; IQR, −3.8–220.1) than in the standard group (125.9; IQR, 66.6–321.0; P=0.01) (Figure 3B). Of interest is that the absolute change in CAC scores was significantly lower in the SO group (74.4; IQR, −9.8–173.3) than in the LC group (131.5; IQR, 26.9–314.5; P=0.03) (Figure 3B). The absolute change in CAC scores was positively correlated with the change in serum phosphate levels at 12 months in each group (Pearson correlation coefficients 0.245, 0.196, 0.223, and 0.173 in the LC, SO, standard, and strict groups, respectively) (Figure 3, C and D). Most patients whose CAC scores decreased were those with decreased phosphate levels. The proportions of patients who showed decreased CAC scores 12 months after treatment were 14.5%, 28.3%, 14.0%, and 27.6% in the LC, SO, standard, and strict groups, respectively.

Figure 3.

Progression of CAC scores. Shown are (A) %CAC change and (B) absolute change of CAC scores in the LC, SO, standard, and strict groups. The effects on %CAC change or absolute change of CAC scores are compared between groups using the Wilcoxon rank-sum test. Data are from the FAS population. (C and D) The change in CAC scores is associated with the change in phosphate levels.

The effect on %CAC change was modified by age (20–64 and 65–79 years old; P value for interaction =0.003) but not baseline CAC score (30–400, 401–1000, and 1001+; P value for interaction =0.53) or sex (P value for interaction =0.69). In other words, stratified analyses of %CAC change by baseline CAC score and by sex showed a similar tendency, but the effect of strict phosphate control on %CAC change was more prominent in aged patients on dialysis. Although %CAC change was significantly lower in the strict group (6.5; IQR, −1.0–13.8) than in the standard group (21.4; IQR, 7.6–33.0; P=0.01) in elderly (65–79 years old) patients on dialysis, %CAC change was not significantly different between the strict group (12.46; IQR, 1.08–43.47) and the standard group (23.3; IQR, 10.4–44.2; P=0.27) in younger (20–64 years old) patients on dialysis.

Side Effects and Serious Adverse Events

During the treatment period, the incidence of side effects was not different between the LC group (eight of 78 patients; 10.3%) and the SO group (17 of 78 patients; 21.8%; P=0.08), as well as between the standard group (17 of 78 patients; 21.8%) and the strict group (eight of 78 patients; 10.3%; P=0.08) (Table 3). Although diarrhea was reported more often in the SO group, there was no significant differences between the LC group and the SO group. The incidence of serious adverse events was also not different across the groups. Among the FAS-analyzed patients, two patients at 3 months, six patients at 6 months, six patients at 9 months, and eight patients at 12 months had low phosphate levels <3.5 mg/dl, but they exhibited no significant adverse events.

Table 3.

Adverse events (safety analysis set)

Side effects and adverse events	No. of patients (%)
	LC, n=78	SO, n=78	Standard, n=78	Strict, n=78
Any side effects	8 (10.3)	17 (21.8)	17 (21.8)	8 (10.3)
Diarrhea	5 (6.4)	9 (11.5)	10 (12.8)	4 (5.1)
Constipation	1 (1.3)	3 (3.8)	2 (2.6)	2 (2.6)
Vomiting	3 (3.8)	0	3 (3.8)	0
Abdominal distention	0	2 (2.6)	2 (2.6)	0
Any serious adverse events	14 (17.9)	17 (21.8)	15 (19.2)	16 (20.5)
Coronary artery diseases	3 (3.8)	4 (5.1)	3 (3.8)	4 (5.1)
Cerebral infarction	1 (1.3)	1 (1.3)	2 (2.6)	0
Shunt disorder	1 (1.3)	2 (2.6)	1 (1.3)	2 (2.6)
Aortic valve stenosis	0	1 (1.3)	0	1 (1.3)
Arthritis	1 (1.3)	0	1 (1.3)	0
Arrhythmia	2 (2.6)	2 (2.6)	2 (2.6)	2 (2.6)
Increase of PTH	0	1 (1.3)	0	1 (1.3)
Acute heart failure	1 (1.3)	0	1 (1.3)	0
Death	1 (1.3)	0	1 (1.3)	0
Femoral fracture	2 (2.6)	2 (2.6)	3 (3.8)	1 (1.3)
Spinal canal stenosis	1 (1.3)	1 (1.3)	0	2 (2.6)
Lymphadenitis	0	1 (1.3)	0	1 (1.3)
Lymph node metastasis	1 (1.3)	0	0	1 (1.3)
Myelopathy	1 (1.3)	0	1 (1.3)	0
Esophageal bleeding	0	1 (1.3)	1 (1.3)	0
Pneumonia	0	1 (1.3)	1 (1.3)	1 (1.3)
Colon polyp	1 (1.3)	0	1 (1.3)	0
Hypotension	0	1 (1.3)	1 (1.3)	0

Discussion

In this EPISODE study, 160 stable patients on dialysis were randomly assigned to receive LC or SO to reduce serum phosphate levels to two target (standard or strict) ranges for 12 months. The primary analysis showed that %CAC change was significantly lower in the strict group (3.5 mg/dl≤ serum phosphate <4.5 mg/dl) than in the standard group (5.0 mg/dl≤ serum phosphate <6.0 mg/dl). Similarly, the absolute change in CAC scores was also significantly lower in the strict group than in the standard group. The absolute change in CAC scores was positively correlated with the change in serum phosphate levels (from baseline to 12 months) in each group, although one-time phosphate level, but not cumulative phosphate exposure, is less predictable for change in CAC scores. Although the KDIGO guidelines recommend lowering elevated phosphate levels toward the normal range in patients with CKD, including patients on dialysis, there have been no randomized trial data providing the evidence that lowering serum phosphate levels could improve any outcomes.¹ We were the first to demonstrate that decreasing serum phosphate levels toward the normal range in patients on dialysis could attenuate the exacerbation of CAC, an important predictor of all-cause mortality in patients on dialysis.² This study suggests that interventions designed to decrease phosphate levels toward the normal range could lead to improved patient outcomes by slowing or reversing the progression of CAC.

It has been demonstrated that hyperphosphatemia is independently associated with mortality and CVDs in patients with CKD, especially in patients on dialysis.^24–26 It has also been demonstrated that arterial medial calcification linked to hyperphosphatemia increased arterial stiffness and hence, left ventricular hypertrophy, both of which are independent mortality risk factors.²⁷ The most likely mechanism underlying the association between hyperphosphatemia and mortality is vascular calcification.² The process of hyperphosphatemia-induced vascular calcification is not merely a passive accumulation of calcium-phosphate particle but an osteogenesis-like process.²⁸ Hyperphosphatemia was reported to induce apoptosis of vascular smooth muscle cells, consequently leading to vascular calcification.²⁹ In this study, it was demonstrated that the changes in serum phosphate levels were associated with the absolute changes in CAC score levels. The benefit of lowering phosphate levels seemed consistent across the groups (LC group versus SO group or standard control versus strict control). In addition, 27.6% of patients had decreased absolute CAC scores, suggesting that decreased serum phosphate could inhibit the phenotypic change of vascular smooth muscle cells and delay or reverse the progression of vascular calcification, as shown in an animal model.³⁰

In this study, there was no difference between the two types of noncalcium-based phosphate binders in %CAC change as primary end point, possibly due to the lack of statistical power driven by the lower number of enrolled patients than we had planned. We demonstrated that the absolute change in CAC scores, one of secondary end points, was significantly lower in the SO group than in the LC group. Moreover, the proportions of patients who showed decreased absolute CAC scores were higher in the SO group (28.3%) than in the LC group (14.5%). The inconsistent results between %CAC change and absolute change in CAC scores may result from the difference in the baseline CAC scores; the SO group had a slightly higher median CAC score (824) than the LC group (680). This is partly because the SO group were more likely to be diabetic than the LC group. Previous reports demonstrated that the progression of calcification was greater in patients with a higher level of arterial calcification at baseline,⁹ but %CAC change might be underestimated in patients with higher CAC scores at baseline. Although there was no significant difference in the decrease of serum phosphate levels between the LC group and the SO group during the treatment period, further studies are necessary to compare the effect of LC and SO on CAC. In addition, there were no differences in the concomitant use of sevelamer hydrochloride or bixalomer between the LC group and the SO group. Patients treated with SO, however, showed increased Hgb and TSAT levels as previously shown in Japanese patients on dialysis,^16,17 suggesting that iron contained in SO was absorbed in the intestine.

The effect of strict phosphate control on %CAC change was more prominent in older patients on dialysis (P value for interaction =0.003). This finding in older subjects is compatible with previous reports documenting the beneficial effect of a noncalcium-based phosphate binder, sevelamer, on mortality only in patients over 65 years of age.³¹ Similarly, cinacalcet, a calcimimetic agent, significantly reduced the risk of death or major cardiovascular events in elderly patients (>65 years) but not in younger patients on dialysis (<65 years).³² Although the progression of calcification was reported to be greater in patients with higher levels of arterial calcification at baseline,⁹ the effect of strict phosphate control on %CAC change was not modified by the baseline CAC score (P value for interaction =0.53). Even in patients on dialysis with baseline CAC scores >1000, %CAC change was significantly lower in the strict group than in the standard group (P=0.04). It has been reported that vascular calcification increases with age,³³ and the burden of artery calcification on mortality might occur in older patients on dialysis.^33,34 This study is the first to demonstrate that lowering elevated phosphate levels toward the normal range would have a beneficial effect on the progression of CAC even in elderly patients with more calcification; thus, strict phosphate control might improve mortality in this superaging society.

There were no significant differences in any side effects or severe adverse events between the LC group and the SO group. The frequency of diarrhea as a side effect was higher in the SO group than in the LC group. In this study, the mean daily dose of SO was lower than that of LC, although there were no significant differences in the decrease of serum phosphate levels between the LC group and the SO group during the treatment period. It has been reported that the phosphate binding activity of SO was similar to that of LC.³⁵ SO-induced diarrhea or improvement of constipation might affect the phosphate-lowering effect of this agent. Observational studies demonstrated that a low serum phosphate was associated with a high mortality risk,^24,26 but there were no significant differences in any side effects or severe adverse events between the standard group and the strict group, suggesting that intervention to decrease serum phosphate levels to the normal range would not increase the mortality risk.

In this study, there were no significant differences in changes of intact FGF-23 or intact PTH levels between groups, despite the significant decrease in serum phosphate levels in the strict group. This is partly because titration of calcimimetics and/or active vitamin D doses for the control of intact PTH might have affected the intact FGF-23 levels because these agents have opposing effects on intact FGF-23 levels.^36–38 However, we did not collect the dose data of these medications.

This study has several limitations. First, the effect of the serum LDL cholesterol (LDL-C) level on %CAC change was not examined, although the baseline total cholesterol level was not different across the groups (162.9±30.0, 162.8±36.1, 158.4±32.3, and 167.4±33.1 mg/dl in the LC, SO, standard, and strict groups, respectively). In addition, there was no significant difference in the prevalence of dyslipidemia at baseline across the groups (Tables 1 and 2). Although sevelamer attenuated CAC compared with a calcium-based phosphate binder,⁹ less calcium load by this agent might not explain this effect. In fact, sevelamer decreased serum LDL-C in patients on dialysis,³⁹ and the progression of CAC was similar in patients on dialysis treated with calcium acetate or sevelamer when the LDL-C level was decreased by a statin.⁴⁰ In this study, there were no significant differences in the proportion of concomitant use of sevelamer hydrochloride or bixalomer between the LC group and the SO group (P=0.94), as well as between the standard group and the strict group (P=0.07). Second, patient outcomes, including death or CVDs, were not assessed because of the relatively short duration of the follow-up period. The mortality risk of Japanese patients on dialysis is lower than that of European and United States patients,⁴¹ suggesting that a longer period is needed to examine the effects of proactive phosphate control on mortality risk or CVD events. In fact, we observed few patients who died or had cardiovascular events in this study (Table 3). Thus, we are now planning an extension study to observe the events of death or CVD. Third, the number of enrolled patients was low, which might have weakened the statistical power. An increased number of patients may resolve the inconsistent results between %CAC change and absolute change in CAC scores between the LC group and the SO group. Data on nutritional parameters, including dietary phosphate intake, the normalized protein catabolism rate, and the body mass index, were not available. We excluded patients with stents and patients with atrial fibrillation to assess CAC scores. Therefore, the observation in this study cannot be extrapolated to patients with such complications. Twenty-six percent of the randomized patients (41 of 156) dropped out from the study, and our post hoc analyses with the multiple imputation method, which included all randomized patients, showed a borderline significant difference in %CAC change between the strict and standard groups, when treatment allocation was not included as an explanatory variable in the multiple imputation analyses. This can be attributed to the fact that the missing %CAC change in the strict control group was imputed by the complete dataset including the standard control group data, resulting in more conservative analyses. Thus, it may be possible that our primary result on the basis of FAS was somewhat affected by the missing data due to patient dropout from the study; however, the results were similar to those of the main analysis. Finally, it is possible that uneven randomized allocation within strata occurred by chance. However, non-negligible differences were not observed in the baseline clinical characteristics between the groups within each stratification factor (e.g., within CACs>1000).

In conclusion, this randomized trial showed for the first time that strict phosphate control toward the normal range delayed the progression of CAC in patients on dialysis. In this study, there was no difference in the effect of two types of noncalcium-based phosphate binders on CAC, but clinical meaningful differences cannot be excluded due to the sample size limitations. A future study should address whether strict phosphate control, as well as an iron-based phosphate binder, could improve the hard outcomes, such as cardiovascular events and mortality, in patients on dialysis.

Disclosures

H. Fujii and Y. Sakaguchi received lecture fees and grants from Bayer Yakuhin Ltd. and Kissei Pharmaceutical Co. Ltd. T. Hamano received lecture fees and grants from Bayer Yakuhin Ltd. and Kissei Pharmaceutical Co. Ltd. Y. Isaka received lecture fees and grants from Kissei Pharmaceutical Co. Ltd. and fees for the chairperson from Bayer Yakuhin Ltd. F. Koiwa received lecture fees and grants from Kissei Pharmaceutical Co. Ltd., and honoraria from and is part of the speakers bureau for Kissei Pharmaceutical Co. Ltd. and Kyowa Kirin Co., Ltd. S. Teramukai received lecture fees from Bayer Yakuhin Ltd.; grants from Nippon Boehringer Ingelheim; and personal fees from Chugai Pharmaceutical, Daiichi Sankyo, Sanofi, Solasia Pharma, Sysmex, and Takeda Pharmaceutical, outside the submitted work. Y. Tsujimoto received lecture fees and fees for the chairperson from Bayer Yakuhin Ltd. and Kissei Pharmaceutical Co. Ltd. All remaining authors have nothing to disclose.

Funding

This research was supported by a grant from Kissei.

Supplemental Material

This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020050598/-/DCSupplemental.

Supplemental Table 1. Inclusion and exclusion criteria.

Supplemental Table 2. Patients characteristics and baseline data (safety analysis set).

Supplemental Table 3. Dosage of sevelamer hydrochloride, bixalomer, or assigned drug.

Supplemental Table 4. Percentages of active vitamin D and calcimimetics users at baseline and 12 months.

Supplemental Table 5. Patient characteristics and baseline data (safety assessment; N=156).

Supplemental Table 6. Sensitivity analyses with multiple imputation.

Supplemental Figure 1. Evaluation of the concordance of the CAC scores by using intraclass correlation coefficients.

Supplementary Figure 2. Change is laboratory parameters.