Sodium zirconium cyclosilicate compared with potassium restriction for hyperkalemia management in type 2 diabetes: The SILVERSTAR trial

ABSTRACT

Aims/Introduction

To demonstrate noninferiority of continued sodium zirconium cyclosilicate therapy versus a potassium‐restricted diet in achieving normokalemia at Visit 7 (Days 28–42) in individuals with type 2 diabetes and hyperkalemia, using a noninferiority margin of −10 percentage points (one‐sided α = 0.025).

Materials and Methods

This multicenter, open‐label, randomized trial assigned participants in the full analysis set to the sodium zirconium cyclosilicate (n = 37) or the diet group (n = 39) for the primary noninferiority analysis. Normokalemia was defined as serum potassium 3.5–<5.0 mEq/L at Visit 7.

Results

In the full analysis set, normokalemia occurred in 28/36 (77.8%) and 26/38 (68.4%) patients in the sodium zirconium cyclosilicate and diet groups, respectively (risk difference +9.4 percentage points; 95% confidence interval, −11.1 to +29.8; one‐sided P = 0.032), failing to demonstrate noninferiority. In the supportive per‐protocol set, normokalemia occurred in 26/31 (83.9%) versus 17/24 (70.8%) patients, and the noninferiority criterion was met (P = 0.020); this analysis is exploratory. Mean serum potassium levels at Visit 7 were lower in the sodium zirconium cyclosilicate group (adjusted difference −0.23 mEq/L; 95% confidence interval, −0.42 to −0.03; P = 0.026). Dietary fiber intake decreased in the diet group. No changes in quality of life were observed between the two groups during the study period. The frequency of adverse events was low; however, one sudden death occurred in the sodium zirconium cyclosilicate group.

Conclusions

The trial did not meet its primary endpoint: noninferiority of sodium zirconium cyclosilicate to the potassium‐restricted diet was not demonstrated in the full analysis set.

INTRODUCTION

Renin–angiotensin–aldosterone system inhibitors (RAASi) and mineralocorticoid receptor antagonists (MRAs)‐related cardiorenal therapies are foundational in contemporary care; however, hyperkalemia remains a major barrier to their initiation and long‐term maintenance, particularly in individuals with type 2 diabetes. Current guidelines (e.g., Kidney Disease: Improving Global Outcomes [KDIGO] 2024) recommend treating hyperkalemia and continuing RAASi whenever feasible through an individualized, stepwise approach that includes reassessment of dietary potassium intake and, when appropriate, the use of newer potassium binders ¹ . However, randomized head‐to‐head evidence comparing dietitian‐guided potassium restriction with pharmacologic potassium binding remains scarce, and strict restrictions may compromise overall diet quality.

To address this gap, we conducted the SILVERSTAR trial, a multicenter, open‐label, randomized, parallel‐group, noninferiority study in adults with type 2 diabetes and hyperkalemia. We tested whether maintenance therapy with sodium zirconium cyclosilicate (SZC) is noninferior to a potassium‐restricted diet for achieving normokalemia at Days 28–42 and assessed secondary biochemical outcomes. We also evaluated changes in quality of life and gastrointestinal symptoms as exploratory outcomes.

MATERIALS AND METHODS

Study design and oversight

SILVERSTAR is a multicenter, open‐label, randomized, parallel‐group, noninferiority trial conducted at sites across Japan (trial registration: jRCTs051230067). The study complied with Japan's Clinical Trials Act and the Declaration of Helsinki and was approved by the Kyoto Prefectural University of Medicine Clinical Research Review Board. All participants provided written informed consent. Funding was provided by AstraZeneca K.K., and the full protocol has been previously published ² .

Participants

Adults aged 20–<90 years with type 2 diabetes and screening serum potassium (sK) 5.0–6.0 mEq/L at consent were eligible. Principal exclusions included use of other hyperkalemia therapies within 7 days, acute kidney injury, estimated glomerular filtration rate (eGFR) <15 mL/min/1.73 m² or dialysis, advanced heart failure, clinically significant arrhythmias or marked QT prolongation, pregnancy, active malignancy under treatment, and anticipated poor adherence.

Interventions and follow‐up

Figure S1 shows the study design. All participants received open‐label SZC 10 g three times daily for 2–3 days (Visit 1 [Day 0] to Visit 2 [Day 3]) to achieve normokalemia prior to randomization. Participants with sK 3.5–<5.0 mEq/L at Visit 2 were randomized within 24 h; if sK remained 5.0–6.0 mEq/L, a repeat sample was obtained at Visit 3 (Day 4), and randomization proceeded if normokalemia was achieved.

Participants were randomized 1:1 to SZC maintenance dosing (starting at 5 g once daily, titrated to 10 g or 15 g once daily if sK was 5.0–6.0 mEq/L) or to a dietitian‐guided potassium‐restricted diet (target <1.5 g/day) while maintaining preexisting prescriptions for calories, protein, and sodium. Study visits were scheduled at Visit 4 (Days 7–10), Visit 5 (Days 14–20), Visit 6 (Days 21–30, as needed), and Visit 7 (Days 28–42).

Non‐SZC potassium binders and other hyperkalemia drugs were prohibited. Post‐randomization SZC use was prohibited in the diet group. The use of RAASi and diuretics remained unchanged when feasible; unavoidable changes were recorded.

Randomization

Randomization used a secure web‐based minimization algorithm with strata for baseline eGFR (<30 vs. ≥30 mL/min/1.73 m²) and Visit 2 sK (<4.3 vs. ≥4.3 mEq/L).

Outcomes

The primary endpoint was the proportion of participants with normokalemia (sK 3.5–<5.0 mEq/L) at Visit 7 (Days 28–42). The prespecified noninferiority margin was −10 percentage points on the absolute risk‐difference scale. The secondary endpoint was sK at Visit 7 as a continuous measure and the proportion of normokalemic (3.5 mEq/L ≤ sK <5.0 mEq/L) participants at Visit 4. Prespecified key exploratory outcomes included the sK trajectory and changes in dietary intake estimated using the Brief‐type Self‐administered Diet History Questionnaire (BDHQ) ³ . These included potassium and dietary fiber levels, as well as adherence correlations and stratified analyses by baseline RAAS inhibitor use. Quality of life and gastrointestinal symptoms were evaluated using the Diabetes Diet‐Related Quality of Life Revised ⁴ and Gastrointestinal Symptom Rating Scale ⁵ , respectively.

Sample size

Assuming Visit 7 normokalemia rates of 58.6% for SZC and 34% for diet, a noninferiority margin of −10 percentage points, a one‐sided α = 0.025, and approximately 85% power, the target enrollment was 80 participants (40 per group), allowing for approximately 10% attrition ² .

Safety

Adverse events (AEs), serious AEs, and prespecified AEs of special interest were recorded. QT prolongation was evaluated using QTcF and assessed at Visits 2 and 7 using protocol‐defined stopping rules.

Statistical analysis

Noninferiority was tested on the absolute risk‐difference scale using the Farrington–Manning method (one‐sided α = 0.025). The full analysis set (FAS) was the prespecified primary population, and the per‐protocol set (PPS) was supportive. Participants who met the prespecified dyskalemia‐related stopping criteria and discontinued the intervention were classified as non‐achievers for the primary endpoint; participants without an evaluable Visit 7 outcome for other reasons were excluded from the primary analysis without imputation. As prespecified, a blinded data handling committee reviewed individual data before database lock; accordingly, in the per‐protocol analysis, Visit 7 serum potassium values recorded after a major protocol deviation or discontinuation were not used even when the participant remained in the PPS. For the continuous secondary endpoint, between‐group differences at Visit 7 were estimated using analysis of covariance (ancova) adjusted for baseline sK and eGFR. Secondary and exploratory analyses were considered descriptive without multiplicity adjustment.

RESULTS

Participants

Of 437 patients screened, 80 were enrolled, 76 were randomized to SZC (n = 37) or a potassium‐restricted diet (n = 39), and four registered participants were not assigned (Figure S2). The FAS comprised 37 (SZC) and 39 (diet) participants, whereas the PPS comprised 33 and 28 participants, respectively (Figure S2). The PPS was derived from the FAS by excluding participants with adherence below 60% (SZC, n = 4; diet, n = 11). The primary endpoint was analyzed in participants with an evaluable Visit 7 outcome (a Visit 7 measurement or a dyskalemia‐related discontinuation counted as non‐achievement): 36 (SZC) and 38 (diet) in the FAS and 31 (SZC) and 24 (diet) in the PPS. Baseline characteristics were balanced between the groups (Table 1). Recruitment occurred from November 10, 2023 to February 4, 2025, and follow‐up was completed on March 27, 2025. The trial ended as planned after the target sample size was reached (n = 80).

Table 1.

Baseline characteristics (full analysis set)

Characteristic	SZC group (n = 37)	Potassium‐restricted diet group (n = 39)
Age, years	73.5 ± 7.3	72.5 ± 8.6
Height, cm	162.4 ± 9.2	160.9 ± 9.4
Weight, kg	61.3 ± 11.0 (n = 37)	60.1 ± 13.5 (n = 38)
Body mass index, kg/m2	23.1 ± 2.9 (n = 37)	23.0 ± 3.6 (n = 38)
Systolic BP, mmHg	136.6 ± 14.9 (n = 37)	134.5 ± 17.3 (n = 38)
Diastolic BP, mmHg	69.2 ± 11.6 (n = 37)	72.3 ± 12.1 (n = 38)
Heart rate, bpm	77.6 ± 12.3 (n = 36)	78.9 ± 10.7 (n = 36)
Diabetes duration, years	22.3 ± 12.6 (n = 35)	19.0 ± 12.7 (n = 37)
Serum potassium at baseline, mEq/L	5.30 ± 0.30	5.35 ± 0.31
eGFR, mL/min/1.73 m2	50.0 ± 16.1	47.9 ± 15.1
QTc(f), ms	410.8 ± 24.7	406.5 ± 22.0
Sex: male, n (%)	25 (67.6%)	23 (59.0%)
Sex: female, n (%)	12 (32.4%)	16 (41.0%)
Prior cardiovascular/cerebrovascular disease, n (%)	4 (10.8%)	8 (20.5%)
Hypertension, n (%)	31 (83.8%)	27 (69.2%)
Dyslipidemia, n (%)	24 (64.9%)	23 (59.0%)
Microvascular complications (any), n (%)	27/35 (77.1%)	32/39 (82.1%)
Antidiabetic medications
Sulfonylurea, n (%)	10 (27.0%)	4 (10.3%)
Biguanide (metformin), n (%)	20 (54.1%)	11 (28.2%)
DPP‐4 inhibitor, n (%)	19 (51.4%)	16 (41.0%)
SGLT2 inhibitor, n (%)	20 (54.1%)	21 (53.8%)
GLP‐1 receptor agonist, n (%)	9 (24.3%)	9 (23.1%)
Insulin, n (%)	10 (27.0%)	18 (46.2%)
Antihypertensive medications
ACE inhibitor, n (%)	3 (8.1%)	3 (7.7%)
ARB, n (%)	22 (59.5%)	18 (46.2%)
ARNI, n (%)	1 (2.7%)	2 (5.1%)
MRA, n (%)	4 (10.8%)	2 (5.1%)
Beta‐blocker, n (%)	3 (8.1%)	4 (10.3%)
Calcium‐channel blocker, n (%)	20 (54.1%)	15 (38.5%)
Other medications
Statin, n (%)	20 (54.1%)	18 (46.2%)
Antiplatelet agent, n (%)	5 (13.5%)	13 (33.3%)

Values are mean ± SD or n (%) unless shown; denominators for some rows reflect available data at baseline. ACE, angiotensin‐converting enzyme; ARB, angiotensin receptor blocker; ARNI, angiotensin receptor–neprilysin inhibitor; BP, blood pressure; DPP‐4, dipeptidyl peptidase‐4; eGFR, estimated glomerular filtration rate; GLP‐1, glucagon‐like peptide‐1; MRA, mineralocorticoid receptor antagonist; QTc(f), Fridericia‐corrected QT; SGLT2, sodium‐glucose cotransporter 2; SZC, sodium zirconium cyclosilicate.

Primary outcome

In the prespecified primary FAS analysis, normokalemia at Visit 7 (Days 28–42; sK 3.5–<5.0 mEq/L) occurred in 28/36 (77.8%) and 26/38 (68.4%) participants in the SZC and diet groups, respectively, for a risk difference of +9.4 percentage points (95% confidence interval [CI], −11.1 to +29.8). The one‐sided noninferiority P value was 0.032, which did not meet the α = 0.025 threshold; therefore, noninferiority was not demonstrated. According to the prespecified hierarchy, superiority was not formally tested because FAS noninferiority was not achieved (Table 2; Figure 1). In the supportive PPS, normokalemia occurred in 26/31 (83.9%) and 17/24 (70.8%) participants in the SZC and diet groups, respectively (risk difference, +13.0 percentage points; 95% CI, −9.0 to +35.1; noninferiority P = 0.020). Superiority was not demonstrated.

Table 2.

Primary endpoint and serum potassium at Visit 7

(A) Primary endpoint (noninferiority)
Analysis population	Normokalemia at Visit 7, n/N (%)	Risk difference, % (95% CI)	Non‐inferiority test	NI conclusion	Superiority P	Superiority conclusion
FAS (prespecified primary)	28/36 (77.8) vs. 26/38 (68.4)	+9.4 (−11.1 to +29.8)	P = 0.032	Not demonstrated	—*	Not formally tested*
PPS (supportive)	26/31 (83.9) vs. 17/24 (70.8)	+13.0 (−9.0 to +35.1)	P = 0.020	Demonstrated (supportive)	0.12	Not demonstrated

(B) Serum potassium levels at Visit 7 (FAS)
Metric	SZC group	Diet group	P value	Adjusted mean difference (SZC − Diet)
n	36	38
Mean ± SD	4.61 ± 0.42	4.86 ± 0.42	0.016
Median [Q1, Q3]	4.60 [4.30, 4.90]	4.85 [4.50, 5.20]
Min, Max	3.70, 5.60	4.10, 5.80
Adjusted mean (SE)	4.62 (0.07)	4.85 (0.07)	0.026	−0.23 (95% CI −0.42 to −0.03)

Risk difference (SZC − Diet). Noninferiority test (one‐sided α = 0.025; margin −10 pp). Superiority P (two‐sided).

^*Superiority formally tested only if FAS noninferiority met. ancova adjusted for baseline serum potassium and baseline eGFR. NI = noninferiority; normokalemia = serum K 3.5–<5.0 mEq/L at Visit 7; Farrington–Manning test. Hierarchical testing prespecified: superiority was to be tested only if NI is met in the primary FAS; because NI was not met in FAS, superiority was not formally tested (any P values shown are descriptive). Risk differences are percentage‐point differences (SZC minus diet). Entries marked with an asterisk indicate that superiority was not formally tested per the prespecified hierarchy because noninferiority was not met in the primary FAS; “—” denotes not applicable.

Figure 1.

The risk differences of primary endpoint and prespecified subgroup by baseline RAAS inhibitor use at Visit 7 and serum potassium levels over time. Left panel (a) indicates effect shown as risk difference (SZC − diet) with 95% CIs. Vertical dashed line indicates the NI margin (−10 percentage points). Noninferiority tested with Farrington–Manning (one‐sided α = 0.025). Proportions with normokalemia at Visit 7 by baseline RAAS inhibitor use (no/yes). Observed proportions: RAAS(−) 7/11 (63.6%) versus 13/15 (86.7%); RAAS(+) 21/25 (84.0%) versus 13/23 (56.5%). Subgroup analyses are exploratory; no formal interaction test was prespecified in the SAP. Right panel (b) indicates serum potassium levels over time.

Secondary endpoints

At Visit 7, the mean sK was lower with SZC than with diet. In the FAS, the adjusted mean difference (ancova adjusted for baseline sK and eGFR) was −0.23 mEq/L (95% CI, −0.42 to −0.03; P = 0.026); the unadjusted two‐sample t‐test was also significant (P = 0.016) (Table 2). These secondary analyses were prespecified without multiplicity adjustment and should be interpreted as exploratory. In the FAS, normokalemia at Visit 4 (Days 7–10; sK 3.5–<5.0 mEq/L) occurred in 30/37 (81.1%) and 27/38 (71.1%) participants in the SZC and diet groups, respectively; the chi‐squared test was not significant (P = 0.31).

Exploratory outcomes

Following the shared loading phase, sK decreased to approximately 4.2 mEq/L at Visit 2 and remained within the protocol‐defined normal range thereafter. The increase from Visit 2 to Visit 7 (Δ[V7 − V2]) was smaller with SZC than with diet (mean [SD], 0.41 [0.52] vs. 0.66 [0.52] mEq/L; P = 0.048) (Table S1; Figure 1).

From Visits 1 to 7, the BDHQ‐estimated potassium intake did not change significantly within or between the groups (Table S2). In contrast, total, soluble, and insoluble fiber decreased in the diet group but not in the SZC group.

No changes in quality of life scores evaluated using the Diabetes Diet‐Related Quality of Life Revised were observed between the two groups during the study period (Table S3). The Gastrointestinal Symptom Rating Scale showed that the diarrhea score increased in the diet group (P = 0.037; Table S4), which, together with the reduction in dietary fiber, may reflect a tolerability trade‐off of the potassium‐restricted diet. These analyses were prespecified but exploratory without multiplicity adjustment.

Subgroup analysis by baseline RAAS inhibitor use (exploratory)

Findings by baseline RAAS inhibitor use were directionally heterogeneous (RAAS(+): 21/25 vs. 13/23; Fisher P = 0.060; RAAS(−): 7/11 vs. 13/15; P = 0.35). No treatment‐by‐RAAS interaction was prespecified; these results are hypothesis‐generating only (Figure 1).

Adherence correlations

Within the diet group, self‐reported adherence (%) did not significantly correlate with the proximal outcomes (Table S5). For sK change (ΔV1 → V7), correlations were nonsignificant (Pearson r = −0.26 (95% CI, −0.53 to 0.07; P = 0.12); Spearman ρ = −0.20 (95% CI, −0.49 to 0.13; P = 0.22)) in 38 participants. For estimated urinary potassium change, correlations were near null (Pearson r = 0.01 (95% CI, −0.52 to 0.54; P = 0.96); Spearman ρ = 0.20 (95% CI, −0.38 to 0.65; P = 0.51)) in 14 participants, and CIs were wide, indicating limited precision. These findings should be interpreted as exploratory and hypothesis‐generating.

Safety

Across the entire study period, AEs occurred in 2/37 (5.4%) and 0/39 (0.0%) participants in the SZC and diet groups, respectively. Events in the SZC group comprised one nonserious case of nausea and one serious AE (sudden death) during the randomized treatment period; no events were reported in the diet group (Table S6). The fatal event occurred in a 76‐year‐old man with an approximately 35‐year history of type 2 diabetes, dyslipidemia, and hypertension, without baseline QT prolongation or a documented arrhythmia. He had been randomized to SZC maintenance and remained on the 5‐g once‐daily starting dose without up‐titration; serum potassium was controlled throughout (4.5 mEq/L at Visit 2, 5.1 mEq/L at Visit 4 [within the protocol allowance], and 4.1 mEq/L at Visit 5), with no SZC‐related adverse events at any visit. While awaiting the scheduled Visit 7, he was found dead at home; no autopsy was performed and the cause of death could not be determined. After review of the available clinical information, it was judged unlikely to be related to SZC and was reported to the independent clinical research review board. Prespecified AEs of special interest (edema/fluid retention, worsening heart failure, QTc prolongation, metabolic alkalosis, hypomagnesemia, and other gastrointestinal symptoms) were not observed, except for nausea. Longitudinal body weight, office blood pressure, physical edema assessments, and diuretic dose trajectories were not systematically captured after baseline.

DISCUSSION

This randomized, open‐label, parallel‐group, noninferiority trial did not demonstrate noninferiority of SZC to a potassium‐restricted diet for the primary endpoint in the prespecified FAS. The per‐protocol findings were supportive but cannot override the primary FAS conclusion by design.

Among secondary outcomes prespecified without multiplicity adjustment, mean sK at Visit 7 was lower with SZC, and in exploratory outcomes, the rebound from the shared post‐loading nadir (Δ[V7 − V2]) was smaller with SZC than with diet, consistent with more durable biochemical control during maintenance. However, this Δ(V7 − V2) comparison rests on sparse intervening visits and a borderline P value and is therefore fragile and hypothesis‐generating.

In the diet group, the BDHQ‐estimated potassium intake did not show a clear reduction over Visits 1–7, whereas total, soluble, and insoluble fiber decreased, suggesting a trade‐off that may limit real‐world sustainability. Over a short maintenance window, no decline in quality of life was observed; however, achieving strict potassium restriction without compromising overall diet quality may be challenging. Although these findings are exploratory and should be interpreted cautiously in light of the primary noninferiority result, they support the evaluation of potassium control along with nutritional consequences when selecting a management strategy.

An important caveat is that the diet comparator may not have produced the intended reduction in potassium intake: BDHQ‐estimated potassium did not change significantly within the diet group or between groups. The diet arm nonetheless achieved 68.4% normokalemia, far exceeding the 34% assumed at design. This pattern suggests that normokalemia in both arms may have been driven less by the assigned maintenance strategy than by regression to the mean from the enrollment hyperkalemia threshold and by residual effects of the shared pre‐randomization SZC loading. Accordingly, the head‐to‐head contrast should be interpreted with this limitation in mind.

A further design consideration is that all participants underwent an identical open‐label SZC loading phase before randomization, reaching approximately 4.2 mEq/L at Visit 2. Over the short 28‐ to 42‐day maintenance window, residual effects of this loading plausibly contributed to normokalemia in both arms. The trial may therefore, to some extent, contrast continued SZC with SZC discontinuation followed by dietary restriction, rather than comparing two newly initiated maintenance strategies—a feature that should be considered when interpreting the diet arm and the short follow‐up.

Interpretation in context

Current cardiorenal practice emphasizes preserving RAASi‐ and MRA–based regimens (the “Fantastic Four” in heart failure; the “four‐pillar” strategy in chronic kidney disease [CKD]), with KDIGO 2024 recommending that clinicians treat hyperkalemia promptly while maintaining RAASi whenever feasible through individualized, stepwise use of dietary measures and modern potassium binders ⁶ , ⁷ , ⁸ , ⁹ . This is supported by evidence showing that discontinuation of RAASi due to hyperkalemia in patients with CKD is associated with increased mortality and cardiovascular events ¹⁰ .

Prior placebo‐controlled studies established that SZC rapidly normalizes potassium levels and maintains normokalemia for several weeks ¹¹ . Our head‐to‐head design addresses a persistent evidence gap by directly comparing a dietitian‐guided potassium restriction strategy with pharmacologic binding under frequent monitoring and pragmatic titration.

Although the primary FAS analysis did not meet the noninferiority criterion, the point estimates numerically favored SZC, and the supportive PPS, which can be informative but is also susceptible to post‐randomization selection bias, was consistent with noninferiority. These results, however, are exploratory and cannot override the primary FAS conclusion ¹² , ¹³ . The FAS–PPS discrepancy is likely attributable primarily to medication nonadherence in the SZC group. Because SZC requires consistent dosing to exert its potassium‐lowering effect, nonadherence directly reduced the normokalemia achievement rate in the SZC group. In the PPS, normokalemia was higher after excluding participants with major protocol deviations, including SZC nonadherence. However, because such exclusions occur after randomization, the PPS estimate can be biased in favor of SZC and does not establish intrinsic efficacy; it should therefore be interpreted only as supportive and hypothesis‐generating. A higher‐than‐expected diet success rate may also have contributed to the failure to demonstrate noninferiority in the FAS. These findings underscore the importance of presenting both FAS and PPS results in noninferiority trials and carefully examining the sources of any discrepancy between them ¹² , ¹³ , ¹⁴ .

The realized effect sizes also differed substantially from the design assumptions. The diet‐arm normokalemia rate (68.4%) greatly exceeded the assumed 34%, and the SZC rate (77.8%) exceeded the assumed 58.6%. With both arms performing better than expected and a smaller‐than‐anticipated between‐group difference, the trial was effectively underpowered for the observed contrast, yielding a wide risk‐difference confidence interval (−11.1 to +29.8 percentage points). In addition, only 76 of the planned 80 participants were randomized, and the primary analysis included 36 and 38 participants. Firm conclusions are difficult to draw from confidence intervals of this width.

Dietary performance and nutrition tradeoffs

In the supportive per‐protocol set, the noninferiority criterion was met for the primary endpoint; because this is a post‐randomization analysis, it is supportive and exploratory and does not override the primary full analysis set result. sK levels at Visit 7, a secondary endpoint, were significantly higher in the diet group than in the SZC group. Coupled with a significant decrease in dietary fiber, this aligns with contemporary observations that rigid “low‐potassium food lists” can unintentionally degrade diet quality while producing modest and variable effects on sK levels ⁸ , ¹⁵ , ¹⁶ . This may be particularly relevant in Japanese settings, where habitual potassium intake is relatively low; in the INTERMAP study, dietary potassium intake was lower in East Asian samples (particularly Chinese) than in the United Kingdom and United States ¹⁷ . In such contexts, the “headroom” for further potassium restriction may be limited, and diet‐only strategies may disproportionately reduce fiber‐rich plant foods without materially lowering total potassium exposure.

Higher fiber intake has been associated with improved outcomes in CKD, suggesting that indiscriminate restriction of plant‐forward patterns may be counterproductive ¹⁸ . Moreover, food frequency instruments such as the BDHQ, although validated and useful at the group level, may undercapture additive potassium from processed foods and other high‐bioavailability sources, potentially biasing between‐group contrasts toward the null ¹⁹ , ²⁰ . A recent feasibility trial involving 26 patients with hyperkalemic CKD stage 4–5 receiving SZC showed that transitioning to a higher potassium plant‐based diet improved dietary quality and fiber intake, whereas plasma potassium remained stable without severe dyskalemia under protocolized SZC titration ²¹ . These considerations support reframing dietary counseling toward minimizing highly absorbable additive potassium while preserving overall diet quality ¹⁵ , ¹⁶ , with pharmacologic binders deployed to safeguard RAASi when needed ⁶ , ⁸ .

Relation to prior trials and practice

Our maintenance‐phase findings are congruent with those of placebo‐controlled SZC trials ¹¹ . They are also consistent with broader evidence that potassium binders can enable and sustain RAASi or MRA use across CKD and heart failure with reduced ejection fraction populations ²² , ²³ , ²⁴ , ²⁵ . Similarly, observational data suggest that outpatient SZC initiation may facilitate RAASi optimization in routine care ²⁶ . In heart failure, emerging randomized evidence from REALIZE‐K indicates that SZC can improve the probability of maintaining guideline‐directed spironolactone doses without loss of normokalemia, although signals for more heart failure events in SZC recipients in that program warrant careful clinical vigilance and confirmatory evaluation ²⁷ , ²⁸ . Collectively, these data, alongside our exploratory advantages in achieved potassium in RAASi and rebound control, reinforce the stepwise, individualized approach of KDIGO 2024 that treats hyperkalemia to maintain disease‐modifying therapy.

Safety

Overall, AEs were infrequent. One serious AE (sudden death) occurred in the SZC group, and no prespecified AEs of special interest were recorded beyond a single case of nausea. These safety findings should be interpreted cautiously given the small sample size and short study duration. The single sudden death warrants careful appraisal. Although the independent clinical research review board judged a relationship between this event and SZC to be unlikely, definitive conclusions regarding the safety profile of SZC cannot be drawn from this trial alone. In the broader context of safety evaluation for newer potassium binders, 15 deaths were reported in the Patiromer AMETHYST‐DN trial without an established causal relationship to treatment, and exploratory analyses from REALIZE‐K suggested a potential signal for increased heart failure events in the SZC group ²⁷ , ²⁸ , ²⁹ . Larger and longer term studies with independent event adjudication are warranted.

Strengths and limitations

Strengths include a randomized, multicenter design; an objective biochemical primary endpoint with frequent monitoring; a pragmatic SZC titration algorithm; and prespecified analytic safeguards (no imputation; discontinuations due to dyskalemia counted as non‐achievers).

The limitations of this study include its short follow‐up period, open‐label design, and reliance on the BDHQ for dietary quantification, all of which constrain inferences regarding clinical outcomes and nutritional trajectories. Additionally, the statistical power of the subgroup analyses (e.g., baseline RAASi use) was limited. Generalizability may also be constrained because habitual potassium intake and dietary sources differ across regions, with lower potassium intake reported in East Asian populations than in Western populations ¹⁷ . Moreover, the shared pre‐randomization SZC loading and the short maintenance window mean that the comparison may reflect continuation versus withdrawal of SZC rather than two independently initiated strategies, which may have narrowed the between‐group difference. The population was elderly (mean age approximately 73 years) and exclusively Japanese, in whom habitual potassium intake is comparatively low, further limiting extrapolation of the diet‐arm effects to other settings. In addition, the prespecified −10‐percentage‐point noninferiority margin is relatively liberal for a biochemical endpoint, which reduces the stringency of the noninferiority criterion.

Clinical implications

Because the primary FAS analysis did not meet noninferiority, these results should be regarded as exploratory and hypothesis‐generating. In PPS, SZC achieved potassium control that appeared comparable to the potassium‐restricted diet, and the lower adjusted mean serum potassium with SZC was a supportive secondary finding.

In an exploratory subgroup analysis among participants receiving RAASi at baseline, the proportion achieving normokalemia at Visit 7 was numerically higher with SZC; however, because multiplicity was not controlled and CIs remained wide, these findings should be regarded as hypothesis‐generating. Nevertheless, the possibility that SZC may be particularly beneficial in patients for whom continuation of RAASi‐ or MRA‐based regimens is a therapeutic priority represents a clinically meaningful finding.

In a setting where guideline‐directed cardiorenal therapies are often limited by hyperkalemia, these data support a practical, individualized strategy that promptly addresses hyperkalemia to preserve RAASi, favor dietary counseling that does not compromise overall diet quality, and consider pharmacologic potassium binding when needed to maintain treatment continuity.

Longer, blinded, and adequately powered studies incorporating patient‐centered outcomes are warranted to define long‐term effectiveness and safety.

CONCLUSION

Noninferiority of SZC maintenance therapy to a dietitian‐guided potassium‐restricted diet was not demonstrated in the primary FAS analysis.

DISCLOSURE

Masahide Hamaguchi has received grants from Kowa Pharma Co. Ltd, Ono Pharma Co. Ltd, and AstraZeneca K.K. Michiaki Fukui has received grants from Ono Pharma Co. Ltd, Oishi Kenko Inc., Yamada Bee Farm, Nippon Boehringer Ingelheim Co. Ltd, Kissei Pharma Co. Ltd, Mitsubishi Tanabe Pharma Corp., Daiichi Sankyo Co. Ltd, Sanofi K.K., Takeda Pharma Co. Ltd, Astellas Pharma Inc., MSD K.K., Kyowa Kirin Co. Ltd, Sumitomo Dainippon Pharma Co. Ltd, Kowa Pharma Co. Ltd, Novo Nordisk Pharma Ltd, Sanwa Kagaku Kenkyusho Co. Ltd, Eli Lilly Japan K.K., Taisho Pharma Co. Ltd, Terumo Corp., Teijin Pharma Ltd, Nippon Chemiphar Co. Ltd, Abbott Japan Co. Ltd, and Johnson & Johnson K.K. Medical Co. The remaining authors declare no conflicts of interest.

Approval of the research protocol: The study protocol was approved by the Kyoto Prefectural University of Medicine Clinical Research Review Board and will be conducted in accordance with the Declaration of Helsinki.

Informed consent: Written informed consent will be obtained from all participants before commencing the study.

Approval data of Registry and the Registration no. of the study/trial: Japan Registry of Clinical Trials (jRCT): jRCTs051230067.

Animal studies: N/A.

Supporting information

Figure S1. Study design. sK, serum potassium; SZC, sodium zirconium cyclosilicate.

Figure S2. Trial profile. Flow of participants from screening to analysis populations. Numbers in parentheses denote participants at each step. Reasons for discontinuation and dropout are shown per protocol stopping criteria; for example, SZC arm discontinuations included criteria 4 (n = 2) and 10 (n = 2), with one dropout; diet arm discontinuations included 10 (n = 3), 10 & 12 (n = 1), 12 (n = 6) with one dropout; four registered but unassigned participants discontinued due to 12. The PPS was derived from the FAS by excluding participants with drug adherence or diet compliance below 60% (SZC, n = 4; diet, n = 11), a criterion independent of intervention discontinuation or dropout. The primary endpoint was analyzed without imputation in participants with an evaluable Visit 7 serum potassium (FAS: 36 and 38; PPS: 31 and 24, respectively). For the PPS, a blinded data‐handling committee determined, according to a prespecified rule and before database lock, that Visit 7 values recorded after a protocol deviation or intervention discontinuation should be excluded. Intervention Discontinuation Criteria 4: At Observation Point 5 (Days 14–20), if the serum potassium level is 5.0 mEq/L or higher but 6.0 mEq/L or lower, and if there is no decrease in the serum potassium level from Observation Point 4 (Days 7–10) to Observation Point 5 (Days 14–20). Intervention Discontinuation Criteria 9: When administration of the study drug or continuation of the potassium‐restricted diet is difficult due to disease, adverse events, or other reasons. Intervention Discontinuation Criteria 10: When adherence to the study drug or compliance with the potassium‐restricted diet is significantly poor (determined to be less than 75% of the total scheduled doses or more than 120% of the total scheduled intake). Intervention Discontinuation Criteria 12: When the principal investigator or sub‐investigator deems it appropriate to discontinue the study for other reasons.

Table S1. Serum potassium over time by visit (FAS; exploratory).

Table S2. Changes in dietary composition including dietary potassium and fiber intakes by BDHQ(FAS).

Table S3. Changes in diabetes diet‐related quality of life revised subscale scores (FAS).

Table S4. Changes in gastrointestinal symptom rating scale scores (FAS).

Table S5. The correlations between adherence and Δ serum and estimated urinary potassium (FAS).

Table S6. Adverse events by randomized group (safety set, overall study period).

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.