Sustained potassium control and recurrent hyperkalemia during individualized potassium-guided sodium zirconium cyclosilicate titration in maintenance hemodialysis: a post hoc analysis of a randomized trial

Abstract

Background

A mean predialysis potassium value at a single visit does not show whether the same patient remains controlled across consecutive months or repeatedly crosses hyperkalemic thresholds. We evaluated sustained patient-level potassium control and recurrent hyperkalemia during individualized potassium-guided sodium zirconium cyclosilicate (SZC) therapy on non-dialysis days.

Methods

This exploratory post hoc analysis used data from a previously reported single-center, open-label randomized trial of 116 patients with chronic hyperkalemia (58 per group). The primary outcome was sustained response during months 3–6, which required all four scheduled predialysis potassium measurements after the long interdialytic interval to be available, at least three of the four values to be within 4.0–5.0 mmol/L, and no unplanned hemodialysis triggered by potassium ≥ 6.5 mmol/L during the evaluation period. Missing measurements were classified as nonresponse. Adjusted risk ratios (aRRs) were estimated using modified Poisson regression with robust standard errors.

Results

The primary outcome occurred in 25/58 (43.1%) SZC-treated participants and 8/58 (13.8%) controls (aRR 3.02, 95% confidence interval [CI] 1.49–6.14). Sustained response within 4.0–5.5 mmol/L occurred in 38/58 (65.5%) and 22/58 (37.9%), respectively (aRR 1.74, 95% CI 1.19–2.54). At least two scheduled potassium values ≥ 5.5 mmol/L occurred in 11/58 (19.0%) and 36/58 (62.1%) (aRR 0.31, 95% CI 0.18–0.55). The adjusted incidence rate ratio for visits with potassium ≥ 5.5 mmol/L was 0.36 (95% CI 0.24–0.54). Complete-case and conservative extreme-case missing-data analyses were directionally consistent with the main analysis.

Conclusions

Individualized potassium-guided SZC therapy was associated with a higher probability of meeting the sustained biochemical potassium-control criterion at scheduled monthly measurements and with fewer recurrent biochemical hyperkalemic excursions. Clinical outcomes were not evaluated in the present analysis; therefore, the observed biochemical improvements should not be interpreted as evidence of clinical benefit.

Trial registration

Chinese Clinical Trial Registry, ChiCTR2600119050. Registered on 14 February 2026; retrospectively registered.

Background

The present study draws on our previously reported randomized trial [1]. Hyperkalemia is common and recurrent in patients receiving maintenance hemodialysis. Serum potassium rebounds after dialysis and accumulates between sessions, making the predialysis measurement after the long interdialytic interval a clinically important exposure window [2–7]. Repeated or sustained elevations may be more informative than a single value because hyperkalemic excursions have been associated with hospitalization and mortality [3–6]. Accordingly, a group mean may not fully capture the longitudinal burden experienced by individual patients.

Sodium zirconium cyclosilicate is a nonabsorbed, selective oral potassium binder. In the phase 3b DIALIZE trial, SZC was given on non-dialysis days and titrated according to serum potassium; treatment increased the proportion of patients who achieved predialysis potassium of 4.0–5.0 mmol/L at three or more of four long-interdialytic-interval visits without rescue therapy [8]. Subsequent analyses showed efficacy across the 4.0–5.0 and 4.0–5.5 mmol/L target ranges and in patients with severe baseline hyperkalemia [9, 10]. DIALIZE China confirmed short-term efficacy in Chinese hemodialysis patients [11], and a small randomized trial of patiromer suggested that titrated potassium-binder therapy can reduce episodic hyperkalemia [12].

We previously conducted a single-center, prospective, open-label randomized controlled trial comparing the long-term effects of individualized, potassium-guided SZC titration on non-dialysis days plus usual care with usual care alone. In the original report, the primary endpoint was the between-group difference in mean predialysis serum potassium following the long interdialytic interval at month 6; the overall number of visits achieving the target range and exploratory clinical events were also reported [1]. However, mean potassium levels and aggregate visit counts could not directly answer whether the same patients maintained control over consecutive months or repeatedly crossed hyperkalemic thresholds. The present analysis therefore addresses the longitudinal durability and recurrence of potassium control rather than reassessing the established potassium-lowering effect of SZC.

Accordingly, we examined distinct longitudinal outcomes that were not evaluated in the original report: patient-level sustained response, recurrent hyperkalemia, and visit-level hyperkalemia burden. We assessed whether individualized SZC titration on non-dialysis days, compared with usual care alone, was associated with a higher probability of meeting the sustained-response definition during months 3 through 6 and with fewer recurrent hyperkalemic episodes.

Methods

Study design and participants

This study was an exploratory post hoc analysis of a previously published randomized controlled trial [1]. The parent publication and the present analysis used the same randomized cohort. The longitudinal outcomes reported here were defined after trial completion and were not reported in the parent article; no new participants, interventions, or follow-up data were added. The trial was conducted at our hemodialysis center. The first participant was enrolled on 9 September 2024 and the last on 14 April 2025. A total of 116 participants were randomly assigned 1:1 to the SZC or control group (58 per group). Treatment lasted 6 months, after which study-administered SZC was discontinued and participants were followed through month 9. This analysis used data collected from baseline through the 6-month treatment period. Reporting follows the CONSORT 2010 recommendations where applicable to this secondary analysis of a randomized trial.

Eligible participants were aged 18–80 years, had received maintenance hemodialysis for at least 3 months, and had chronic hyperkalemia, defined as at least two predialysis potassium values > 5.0 mmol/L after the long interdialytic interval during the preceding year, including the most recent measurement. Key exclusions were dialysis for acute kidney injury, hyperkalemia attributable to potassium supplementation or another clearly reversible medication factor, severe primary cardiac, cerebrovascular, hepatic, or hematologic disease, malignancy, or inability to complete follow-up. Block randomization (block size 4) was stratified by baseline potassium (> 6.0 vs. ≤ 6.0 mmol/L) and age (> 60 vs. ≤ 60 years), with allocation concealed in sequentially numbered, opaque, sealed envelopes [1].

Intervention and follow-up

Both groups received thrice-weekly hemodialysis with dialysate potassium 2.0 mmol/L, a target single-pool Kt/V > 1.2, and monthly standardized dietary counseling. A food-frequency questionnaire or dietary diary was not used; individual dietary potassium intake therefore could not be quantified. Recorded renin–angiotensin–aldosterone system inhibitor (RAASi) therapy comprised angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and angiotensin receptor–neprilysin inhibitors. No participant used a mineralocorticoid receptor antagonist, oral sodium bicarbonate, another potassium binder, a potassium supplement, or any other medication likely to materially affect serum potassium during the study. Serum bicarbonate, total carbon dioxide, and blood-gas pH were not systematically collected; metabolic acidosis status therefore could not be reliably classified.

Participants in the SZC group took SZC on non-dialysis days, with the dose adjusted according to the preceding long-interdialytic-interval predialysis potassium: 10 g for ≥ 6.0 mmol/L, 5 g for 5.0 to < 6.0 mmol/L, 2.5 g for 4.0 to < 5.0 mmol/L, and temporary withholding for < 4.0 mmol/L. The 2.5-g step was an exploratory maintenance dose used only after the target range had been reached [1]. Potassium ≥ 6.5 mmol/L triggered discontinuation of RAASi therapy and one additional hemodialysis session; two consecutive values ≥ 6.5 mmol/L met the protocol withdrawal criterion. All study potassium samples were collected before dialysis after the 72-h long interdialytic interval and analyzed by the hospital central laboratory.

Outcomes

The primary outcome of the present analysis was sustained response during the maintenance phase. Responders were required to have all four scheduled potassium measurements at months 3, 4, 5, and 6, with at least three values within 4.0–5.0 mmol/L and no unplanned hemodialysis triggered by potassium ≥ 6.5 mmol/L during the same period. Under the original protocol, each potassium value ≥ 6.5 mmol/L immediately triggered one additional dialysis session; the rescue event was therefore linked to the triggering measurement. The definition was informed by the patient-level response framework used in DIALIZE [8]. Because the assessment window was extended from 4 weeks to 4 months, however, it represents a longer-term application of that framework rather than a direct replication of the original definition.

Secondary outcomes were: sustained response within a wider range of 4.0–5.5 mmol/L; at least four of six scheduled values within 4.0–5.0 mmol/L; any scheduled value ≥ 6.0 or ≥ 6.5 mmol/L; recurrent hyperkalemia, defined separately as at least two scheduled values ≥ 5.5 or ≥ 6.0 mmol/L; and the total number of observed visits with potassium ≥ 5.5 or ≥ 6.0 mmol/L. All outcomes were defined for this analysis and were not confirmatory endpoints of the original trial.

Statistical analysis

All 116 randomized participants were analyzed according to assigned group. Categorical outcomes are reported as counts and percentages. For sustained-response outcomes, any missing month 3–6 potassium measurement was classified as nonresponse. A complete-case sensitivity analysis included only participants with all four month 3–6 measurements. For the main threshold-exceedance analyses, an unobserved visit was not counted as an event, and count models used each participant’s number of observed scheduled visits as the exposure. To examine the effect of differential missingness, we performed a conservative extreme-case sensitivity analysis separately for each potassium threshold. All missing scheduled measurements in the SZC group were classified as hyperkalemic events, whereas all missing measurements in the control group were classified as non-events. Recurrent hyperkalemia was recalculated using the resulting six-visit event histories. For the extreme-case visit-level analyses, each participant’s six planned visits were used as the exposure, with the natural logarithm of six included as the offset.

Binary outcomes were analyzed using modified Poisson regression with the heteroskedasticity-consistent type 0 (HC0) robust sandwich variance to estimate risk ratios and 95% confidence intervals [13]. Models adjusted for the two original randomization strata: baseline potassium > 6.0 mmol/L and age > 60 years. Fisher’s exact test was used descriptively for sparse outcomes. Hyperkalemic visit counts were analyzed using Poisson regression with an HC0 robust sandwich variance, the natural logarithm of the relevant number of observed or planned visits as an offset, and adjustment for the same stratification factors; adjusted incidence rate ratios (aIRRs) are reported. Tests were two-sided. Because numerous exploratory endpoints and thresholds were examined, no multiplicity correction was applied; all secondary findings were considered hypothesis-generating and were interpreted according to effect estimates, confidence intervals, and consistency rather than statistical significance alone. All analyses were performed using R version 4.6.0 (R Foundation for Statistical Computing, Vienna, Austria).

Ethics and trial registration

The original trial was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of our hospital (approval No. 2023-27). All participants provided written informed consent. The present analysis used de-identified data collected under the approved protocol and involved no additional intervention or data collection. The trial was retrospectively registered in the Chinese Clinical Trial Registry (ChiCTR2600119050) on 14 February 2026. Because registration occurred after trial completion, all outcomes in the present analysis were considered exploratory.

Results

Analysis population

All 116 randomized participants were included (58 per group). Baseline demographic and clinical characteristics were generally similar (Table 1). Potassium was complete in both groups at months 1 and 2. Missing measurements at months 3, 4, 5, and 6 numbered 1, 1, 2, and 3 in the SZC group and 0, 3, 4, and 9 in the control group, respectively. Complete month 3–6 data were available for 55 SZC-treated participants and 49 controls.

Table 1.

Baseline characteristics of the study population

Variable	SZC group (n = 58)	Control group (n = 58)	P	SMD
Age, years	61.47 ± 11.97	58.84 ± 10.67	0.235	0.226
Sex	33 male, 25 female	37 male, 21 female	0.448	0.141
spKt/V	1.49 ± 0.29	1.44 ± 0.25	0.363	0.185
Hypertension	45 (77.6%)	46 (79.3%)	0.821	0.043
Diabetes	25 (43.1%)	25 (43.1%)	1.000	0
Dialysis vintage, months	27.50 (10.00, 75.00)	30.50 (10.75, 73.25)	0.875	0.031
Serum potassium, mmol/L	5.71 ± 0.46	5.66 ± 0.44	0.526	0.112
IDWG, kg	2.85 (2.20, 3.55)	2.95 (2.20, 3.80)	0.843	0.081
Dry weight, kg	57.51 ± 13.76	59.89 ± 12.15	0.326	0.177
RAASi use (ACEI, ARB, or ARNI)	24 (41.4%)	26 (44.8%)	0.708	0.068

Baseline data, P values, and SMDs were reported previously [1]. Recorded RAASi therapy included angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, or angiotensin receptor–neprilysin inhibitors. SMD, standardized mean difference; IDWG, interdialytic weight gain; RAASi, renin–angiotensin–aldosterone system inhibitor; SZC, sodium zirconium cyclosilicate

Sustained potassium control

The primary sustained-response endpoint occurred in 25/58 (43.1%) participants assigned to SZC and 8/58 (13.8%) controls. After adjustment for the randomization strata, participants assigned to SZC were about three times as likely as controls to achieve sustained response (aRR 3.02, 95% CI 1.49–6.14). The wider 4.0–5.5 mmol/L sustained-response endpoint occurred in 38/58 (65.5%) and 22/58 (37.9%), respectively (aRR 1.74, 95% CI 1.19–2.54). Across the full 6-month window, at least four of six values were within 4.0–5.0 mmol/L in 30/58 (51.7%) and 7/58 (12.1%) participants (aRR 4.29, 95% CI 2.03–9.06; Table 2).

Table 2.

Patient-level sustained potassium control and recurrent hyperkalemia

Outcome	SZC	Control	Adjusted effect (95% CI)
Primary sustained response: ≥3/4 values 4.0–5.0 mmol/L during months 3–6 and no rescue dialysis in the evaluation period	25/58 (43.1%)	8/58 (13.8%)	aRR 3.02 (1.49–6.14)
Wider sustained response: ≥3/4 values 4.0–5.5 mmol/L during months 3–6 and no rescue dialysis in the evaluation period	38/58 (65.5%)	22/58 (37.9%)	aRR 1.74 (1.19–2.54)
At least 4/6 values 4.0–5.0 mmol/L over 6 months	30/58 (51.7%)	7/58 (12.1%)	aRR 4.29 (2.03–9.06)
Any scheduled potassium ≥ 6.0 mmol/L	9/58 (15.5%)	32/58 (55.2%)	aRR 0.28 (0.15–0.53)
Any scheduled potassium ≥ 6.5 mmol/L	1/58 (1.7%)	14/58 (24.1%)	aRR 0.07 (0.01–0.53)
At least two scheduled values ≥ 5.5 mmol/L	11/58 (19.0%)	36/58 (62.1%)	aRR 0.31 (0.18–0.55)
At least two scheduled values ≥ 6.0 mmol/L	2/58 (3.4%)	14/58 (24.1%)	aRR 0.14 (0.04–0.56)

Adjusted risk ratios were estimated using modified Poisson regression with robust standard errors and adjustment for baseline potassium > 6.0 mmol/L and age > 60 years. Sustained-response endpoints required all four month 3–6 measurements; missing values were classified as nonresponse. All endpoints were defined post hoc. aRR, adjusted risk ratio; CI, confidence interval; SZC, sodium zirconium cyclosilicate

Hyperkalemic excursions and recurrence

Any scheduled potassium value ≥ 6.0 mmol/L occurred in 9/58 (15.5%) SZC-treated participants and 32/58 (55.2%) controls (aRR 0.28, 95% CI 0.15–0.53). Values ≥ 6.5 mmol/L occurred in 1/58 (1.7%) and 14/58 (24.1%), respectively (aRR 0.07, 95% CI 0.01–0.53); this sparse estimate was imprecise. At least two values ≥ 5.5 mmol/L occurred in 11/58 (19.0%) and 36/58 (62.1%) (aRR 0.31, 95% CI 0.18–0.55), while at least two values ≥ 6.0 mmol/L occurred in 2/58 (3.4%) and 14/58 (24.1%) (aRR 0.14, 95% CI 0.04–0.56; Table 2).

At the visit level, potassium was ≥ 5.5 mmol/L at 47 of 341 observed visits in the SZC group and 128 of 332 visits in the control group (aIRR 0.36, 95% CI 0.24–0.54). Potassium was ≥ 6.0 mmol/L at 11/341 and 52/332 visits, respectively (aIRR 0.20, 95% CI 0.10–0.40; Table 3).

Table 3.

Hyperkalemic visit burden and sensitivity analysis

Analysis	SZC	Control	Adjusted effect (95% CI)
Scheduled visits with potassium ≥ 5.5 mmol/L	47/341	128/332	aIRR 0.36 (0.24–0.54)
Scheduled visits with potassium ≥ 6.0 mmol/L	11/341	52/332	aIRR 0.20 (0.10–0.40)
Primary endpoint, complete-case analysis	25/55 (45.5%)	8/49 (16.3%)	aRR 2.62 (1.30–5.28)
Recurrent potassium ≥ 5.5 mmol/L, (extreme-case missing-data analysis)	12/58(20.7%)	36/58 (62.1%)	aRR 0.34 (0.20–0.58)
Recurrent potassium ≥ 6.0 mmol/L, ( extreme-case missing-data analysis)	3/58 (5.2%)	14/58 (24.1%)	aRR 0.21 (0.07–0.63)
Planned visits with potassium ≥ 5.5 mmol/L, (extreme-case missing-data analysis)	54/348(15.5%)	128/348(36.8%)	aIRR 0.42 (0.28–0.63)
Planned visits with potassium ≥ 6.0 mmol/L, ( extreme-case missing-data analysis)	18/348 (5.2%)	52/348 (14.9%)	aIRR 0.34 (0.17–0.68)

Main visit-count models used the natural logarithm of each participant’s observed visits as an offset. In the conservative extreme-case analysis, all missing visits in the SZC group were classified as threshold events and all missing visits in the control group as non-events; six planned visits per participant were used as the exposure. Models used HC0 robust variance estimates and adjusted for baseline potassium > 6.0 mmol/L and age > 60 years. aIRR, adjusted incidence rate ratio; aRR, adjusted risk ratio; CI, confidence interval; SZC, sodium zirconium cyclosilicate

Sensitivity analyses and data verification

Among participants with all four month 3–6 measurements, the primary endpoint occurred in 25/55 (45.5%) SZC-treated participants and 8/49 (16.3%) controls (aRR 2.62, 95% CI 1.30–5.28), consistent with the main analysis. Results were directionally consistent in the conservative extreme-case analysis that classified all missing visits in the SZC group as hyperkalemic events and all missing visits in the control group as non-events. Recurrent potassium ≥ 5.5 mmol/L occurred in 12/58 (20.7%) SZC-treated participants and 36/58 (62.1%) controls (aRR 0.34, 95% CI 0.20–0.58), while recurrent potassium ≥ 6.0 mmol/L occurred in 3/58 (5.2%) and 14/58 (24.1%), respectively (aRR 0.21, 95% CI 0.07–0.63). At the visit level, the corresponding event counts were 54/348 versus 128/348 for potassium ≥ 5.5 mmol/L (aIRR 0.42, 95% CI 0.28–0.63) and 18/348 versus 52/348 for potassium ≥ 6.0 mmol/L (aIRR 0.34, 95% CI 0.17–0.68; Table 3). Data verification showed that, for two participants in the control group, the cumulative number of unplanned dialysis sessions over 6 months exceeded the number of monthly scheduled visits with serum potassium ≥ 6.5 mmol/L by one event each. However, neither participant’s scheduled serum potassium pattern from months 3 through 6 met the criteria for sustained response; therefore, the timing of the additional events did not affect the primary endpoint classification.

Discussion

The present findings extend the parent report by evaluating longitudinal, patient-level potassium control. From months 3 through 6, participants assigned to individualized SZC titration were about three times as likely as controls to meet the strict sustained-response definition. Recurrent hyperkalemia and the visit-level hyperkalemia burden were also lower at both the ≥ 5.5 and ≥ 6.0 mmol/L thresholds. The complete-case analysis and a deliberately conservative extreme-case analysis of missing visits showed the same direction of effect.

These findings complement the main results of the original report. The original study showed that, at month 6, predialysis serum potassium following the long interdialytic interval was 0.363 mmol/L lower in the SZC group than in the control group, and that a greater proportion of visits achieved the target range of 4.0–5.0 mmol/L [1]. Neither summary measure directly shows whether an individual patient remained within the target range over consecutive months or experienced repeated hyperkalemic excursions. The present analysis addressed both questions: whether individual patients maintained the target range over consecutive months and whether the same patients repeatedly crossed hyperkalemia thresholds. These findings therefore provide complementary patient-level information about long-term potassium stability.

The sustained-response endpoint extended the patient-level framework used in DIALIZE from a 4-week evaluation period to four consecutive months [8–10]. This longer assessment window allowed durability to be evaluated across consecutive monthly measurements obtained after the 72-h interdialytic interval. The consistent findings across the strict and wider target ranges suggest that the between-group difference did not depend on a single potassium boundary. A small randomized trial of patiromer also reported fewer episodic potassium elevations with titrated binder therapy [12], further supporting recurrent hyperkalemia as a clinically relevant outcome across potassium-lowering strategies.

Recurrent hyperkalemia is clinically relevant because serum potassium often fluctuates in patients receiving maintenance hemodialysis, and hyperkalemic exposure after the long interdialytic interval is associated with higher risks of hospitalization and mortality [2–5]. Previous SZC studies also support flexible maintenance treatment. In DIALIZE, potassium-guided titration resulted in maintenance doses of 5, 10, or 15 g on non-dialysis days without clinically meaningful dose-dependent worsening of short-term fluid balance [14]. Our potassium-guided algorithm also included an exploratory 2.5-g maintenance step after target attainment. Because the dose selected at each visit depended on the preceding potassium measurement rather than random assignment, the independent efficacy of the 2.5-g dose cannot be determined. A longer-term open-label study in non-dialysis outpatients with hyperkalemia found that individualized SZC dosing could maintain normokalemia for up to 12 months [15].

Neither study, however, quantified recurrent hyperkalemia within the same patient over consecutive months. In the present analysis, the SZC group had an approximately 69% lower risk of experiencing at least two potassium values ≥ 5.5 mmol/L and an approximately 64% lower visit-level rate at the same threshold. Reductions were even greater at ≥ 6.0 mmol/L. The consistency across thresholds and levels of analysis suggests that potassium-guided titration may reduce both recurrence and overall hyperkalemic exposure.

These biochemical findings should be interpreted cautiously. Although the parent trial reported exploratory clinical events, clinical outcomes, including arrhythmia, hospitalization, and mortality, were not evaluated in the present analysis. The observed reduction in recurrent hyperkalemia therefore cannot be interpreted as evidence of clinical benefit. In DIALIZE-Outcomes, which included 2,690 patients receiving hemodialysis, SZC improved potassium control but did not reduce the primary arrhythmia-related composite endpoint; early trial termination and treatment discontinuation complicated interpretation [16]. The smaller crossover ADAPT trial reported fewer episodes of atrial fibrillation and clinically significant arrhythmia with higher dialysate potassium combined with SZC on non-dialysis days than with lower dialysate potassium without SZC [17]. Because the intervention changed both dialysate potassium and SZC use, the independent effect of SZC cannot be isolated. Findings from the Dialysis Outcomes and Practice Patterns Study (DOPPS) likewise suggest that serum potassium, dialysate potassium, and arrhythmia risk are closely related [18]. Whether sustained biochemical control translates into fewer arrhythmias, hospitalizations, or deaths therefore remains uncertain.

Several features strengthen this analysis. The data arose from a randomized trial, so analyses of sustained control and recurrence retained the original randomized comparison. All potassium samples were collected before dialysis after the standardized 72-h long interdialytic interval, limiting variation in sampling time. Results were consistent across the strict and wider target ranges, hyperkalemic excursions, recurrent events, and visit-level rates. Missing month 3–6 measurements were classified as nonresponse in the primary analysis. The complete-case analysis showed a similar direction of effect, and the extreme-case analysis remained directionally consistent under assumptions deliberately unfavorable to SZC.

Several limitations warrant consideration. Because this was a post hoc analysis, the outcomes were defined after trial completion and the original sample size was not calculated for sustained response or recurrence; all findings should therefore be regarded as exploratory and hypothesis-generating. Numerous endpoints and potassium thresholds were examined without adjustment for multiple comparisons, increasing the possibility of false-positive findings. None of the secondary estimates should therefore be regarded as independently confirmatory, even when the corresponding confidence interval excluded the null. The retrospective registration of the original trial further limits external verification of which analyses were prespecified [1].

The sustained-control endpoint was based on scheduled monthly predialysis potassium measurements. It therefore indicates that participants met the prespecified criterion at those visits, not that potassium remained continuously controlled between visits; intervening hyperkalemic episodes may have gone undetected.

The single-center, open-label design may have influenced clinician-dependent decisions, including the implementation and timing of unplanned rescue dialysis after the protocol potassium threshold was reached. The modest sample size also resulted in limited precision for infrequent outcomes, as reflected by the wide confidence interval for potassium ≥ 6.5 mmol/L. Missing data became increasingly imbalanced between groups later in follow-up. Although unobserved visits were not counted as events in the main threshold analyses, the estimates remained consistent in the deliberately conservative extreme-case analysis; this does not establish the actual potassium status at unobserved visits. Individual dietary potassium intake was not quantified, and metabolic acidosis could not be characterized because bicarbonate, total carbon dioxide, and blood-gas pH were not systematically collected. Residual effects of dietary intake or acid–base status therefore cannot be excluded.

These limitations highlight several priorities for future research. Prospective multicenter studies should prespecify sustained-response and recurrence endpoints, use blinded adjudication for clinically determined events, and incorporate systematic rhythm monitoring. Integrating serial potassium measurements with treatment exposure and subsequent clinical events may also help identify patients most likely to benefit from continued maintenance therapy or more intensive dose titration.

Conclusions

In the present analysis, individualized potassium-guided SZC titration on non-dialysis days was associated with a higher probability of meeting the sustained biochemical potassium-control criterion at scheduled monthly measurements and a lower biochemical burden of recurrent hyperkalemia. Clinical outcomes, including arrhythmia, hospitalization, and mortality, were not evaluated; clinical benefit therefore cannot be inferred from these biochemical findings. These exploratory findings require confirmation in prospectively designed multicenter studies.

Abbreviations

ACEI

Angiotensin-Converting Enzyme Inhibitor

ARB

Angiotensin II Receptor Blocker

ARNI

Angiotensin Receptor–Neprilysin Inhibitor

aIRR

adjusted Incidence Rate Ratio

aRR

adjusted Risk Ratio

CI

Confidence Interval

DOPPS

Dialysis Outcomes and Practice Patterns Study

HC0

Heteroskedasticity-Consistent type 0

IDWG

InterDialytic Weight Gain

RAASi

Renin–Angiotensin–Aldosterone System inhibitor

SMD

Standardized Mean Difference

spKt/V

single-pool Kt/V

SZC

Sodium Zirconium Cyclosilicate

Author contributions

HW performed the statistical analyses and drafted the manuscript. WZ supervised the parent trial and the present analysis. YJ collected the parent-trial data and contributed to data interpretation. All authors critically revised the manuscript, read and approved the final version, and agreed to be accountable for the work.

Funding

This study was supported by the Lin’an Sci-Tech Guidance Project (Agriculture & Social Development), grant No. 2023Y07.

Data availability

The de-identified dataset supporting this analysis is available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The parent trial was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of The First People’s Hospital of Lin’an District (approval No. 2023-27). All participants provided written informed consent before enrollment. The present analysis used de-identified data collected under the approved protocol and involved no additional intervention or data collection.00

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.