SGLT2 inhibitors as a strategy for volume management in peritoneal dialysis: a case series

Introduction

Peritoneal dialysis (PD) is a home-based renal replacement therapy that offers flexibility and a slower initial decline in residual kidney function compared with haemodialysis. However, fluid overload remains a major complication, affecting up to 25–50% of PD patients and independently predicting cardiovascular mortality (1, 2). Adequate peritoneal ultrafiltration (UF) is essential for volume homeostasis, yet chronic exposure to glucose-based dialysate paradoxically damages the peritoneal membrane through glucose degradation products, oxidative stress and fibrosis, ultimately leading to UF failure (3).

Sodium–glucose cotransporter 2 inhibitors (SGLT2i) block glucose and sodium reabsorption in the proximal tubule, inducing glucosuria, osmotic diuresis and tubuloglomerular feedback–mediated reduction in intraglomerular pressure (4). Landmark trials have demonstrated benefits across a broad range of kidney function (5–7). Additionally, SGLT2i stimulate erythropoiesis via hypoxia-inducible factor 2α upregulation, consistently increasing haemoglobin in randomized trials (8).

Translational studies have shown that SGLT2 is expressed in the human peritoneum (9), that its inhibition reduces dialysate glucose absorption in animal models (10) and that it mitigates peritoneal fibrosis (11). However, clinical data in PD remain scarce (12–16): no placebo-controlled trial has reported results, although the EMPOWERED trial protocol has been published (17), and a recent review confirmed the paucity of outcome data (18). The rationale for using these agents in PD is threefold. Their cardiovascular and metabolic actions do not depend on renal clearance and need not be forfeited when residual function is exhausted; the peritoneum expresses SGLT2, which offers a plausible target for reducing dialysate glucose absorption and preserving the osmotic gradient (9, 10); and for patients with ultrafiltration failure and no alternative modality there is at present no other pharmacological option.

We therefore describe the systematic use of these agents in an unselected, predominantly non-diabetic PD cohort. Given the sample size, the purpose of this report is descriptive: to document what happened to seven patients, individually and without inference, in clinical scenarios for which no evidence currently exists.

Methods

Design and setting

Retrospective, single-centre, descriptive case series comprising all consecutive patients receiving PD at Hospital Universitario Ramón y Cajal in whom an SGLT2i was prescribed between February 2023 and March 2025. No patient meeting this criterion was excluded.

Ethics

The study was approved by the Clinical Research Ethics Committee of Hospital Universitario Ramón y Cajal (reference 188/26) and conducted in accordance with the Declaration of Helsinki and Spanish Law 14/2007 on Biomedical Research. Given the retrospective design and the use of fully anonymized data derived from routine care, the Committee waived the requirement for informed consent for participation.

Treatment decision and off-label framework

All prescriptions were made outside the authorised conditions of use (19, 20), under the Spanish regulatory framework for medicines used in special situations (Royal Decree 1015/2009). Each decision was taken collectively in the Peritoneal Dialysis Unit clinical session. The reasoning differed by indication: recurrent congestive admissions in two patients with type 2 cardiorenal syndrome, in whom the heart failure indication specifies no lower renal threshold for continuation; insufficient peritoneal ultrafiltration with no conventional alternative in three; preservation of residual kidney function in one; and adjuvant therapy in one patient with histologically confirmed encapsulating peritoneal sclerosis (11). In the anuric patients no renal effect could be expected and the rationale was exclusively peritoneal (9).

Timing of treatment

The interval between the start of PD and the start of the SGLT2i is reported for every patient in Table 1. Four patients began therapy on established PD, two began before PD initiation and one began after prolonged PD vintage. One patient had received the drug in advanced CKD, had it withdrawn following an episode of acute kidney injury, and resumed it before starting PD.

Table 1. Baseline characteristics, timing of treatment and peritoneal transport, by patient (N = 7).

Icodextrin was used in all patients except Patient 5. Patient 6 had received dapagliflozin in advanced CKD from 17/05/2022 to 10/04/2023, withdrawn for acute kidney injury and restarted on 05/02/2024. UF, 24-hour peritoneal ultrafiltration (mL/24h); EPS, encapsulating peritoneal sclerosis; n.a., not available.

Peritoneal dialysis prescription

The PD prescription — modality, number and duration of dwells, glucose concentration and use of icodextrin — was not modified during the treatment period in any patient. Ultrafiltration refers to the total 24-hour drained volume minus the volume infused, recorded by the patient at home. Icodextrin-based solutions were used in six patients.

Definitions and measurements

Residual kidney function was determined from 24-hour urine collections and is expressed as the arithmetic mean of the renal urea and creatinine clearances, normalized to body surface area. No creatinine-based estimating equation was used at any point, and the term estimated glomerular filtration rate is not applicable to this cohort. Anuria was defined as a urine output below 200 mL/24h. Peritoneal transport status was categorised by peritoneal equilibration test. Doses of erythropoiesis-stimulating agents were unified using the conversion 1 µg darbepoetin alfa = 200 IU epoetin alfa, and the erythropoietin resistance index was calculated as the weekly equivalent dose divided by haemoglobin.

Analysis

Analysis is descriptive only: continuous variables as median and range, categorical variables as counts and percentages. No hypothesis testing was performed, being neither justified nor informative at this sample size; every outcome is presented as an individual trajectory.

Results

Cohort

Seven patients on PD treated with SGLT2i were enrolled at a single university center between February 2023 and March 2025 (Table 1). The median age was 69 years (range 43–73). Five patients (71.4%) were male. Only one (14.3%) had type 2 diabetes mellitus, and this patient was also the only one treated with empagliflozin. CKD etiologies were heterogeneous: cardiorenal syndrome type 2 (n = 2), long-standing CKD with subsequently acquired cirrhosis and cardio-hepato-renal syndrome (n = 1), lupus nephritis (n = 1), membranous nephropathy (n = 1), chronic tubulointerstitial nephritis (n = 1), and cortical necrosis (n = 1). Four patients started the SGLT2i on established PD, at a median of 5.8 months (range 0.8–16.1) after starting the technique; one started after 30 months on PD; and two started before PD initiation, 11.9 and 3.0 months respectively. Six patients received dapagliflozin 10 mg/day and one received empagliflozin 10 mg/day. Six patients (85.7%) used icodextrin-based solutions. Two patients (28.6%) were anuric at baseline by the prespecified definition.

Peritoneal ultrafiltration

Paired baseline and follow-up UF data were available for four patients (Patients 1–4), who are precisely the four in whom the drug was introduced onto an established and unchanged PD prescription. Patients 5 and 6 have no baseline value because they were already receiving the drug when the technique began, and Patient 7 has no ultrafiltration data at all. Median UF rose from 659 mL/24h (range 500–800) at baseline to 988 (758–1139) at one month and 1250 (721–1596) at three months (Figure 1A). Three of the four patients increased their ultrafiltration, by 1096, 842 and 340 mL/24h at three months; one decreased it by 79 mL/24h. Patient 2 was the sole non-responder, and was also the only low-average transporter in the group; he was transferred to haemodiafiltration three months after starting therapy because of pulmonary hypertension with intractable volume management. In Patients 3 and 4 (near-anuric and anuric, respectively), SGLT2i-enhanced UF was critical for sustaining PD viability. Peritoneal transport category was unchanged between paired equilibration tests in all four patients in whom these were available. Clinically, peripheral oedema was present in four patients at baseline and in two at six months. Thus, patient number 1, who had suffered from frequent episodes of cardiac failure, notably improved following UF increase after SGLT2i. In addition, peritoneal dialysis was the only possible treatment for patient number 3. All vascular accesses were exhausted, the inferior vena cava was filiform, and kidney transplantation was not an option. Increasing ultrafiltration with SGLT2 inhibitors was essential to maintain euvolemia in this anuric patient.

Residual kidney function and diuresis

Residual kidney function was available in five patients at baseline and in four throughout. In the four with complete data the median value was 10.6 mL/min at baseline, 10.5 at one month and 14.5 at six months; two showed a transient fall and recovery, one remained stable and one rose progressively. The fall in one patient coincided with a simultaneous fall in Kt/V and reflected an intercurrent clinical event. Both patients anuric at six months were already anuric at baseline, and no patient developed anuria de novo. Among the five with preserved diuresis, urine output decreased in three (by 200, 210 and 400 mL/24h), was unchanged in one and increased in one. Three individual Kt/V values fell below 1.7 during follow-up (Table 2). The urinary protein-to-creatinine ratio, available in five patients, decreased in three (by 24%, 42% and 58%) and increased in two.

Table 2. Volume status and kidney function by patient.

UF, 24-hour peritoneal ultrafiltration (mL/24h); RKF, residual kidney function from 24-hour urine collection (mL/min/1.73 m²); n.a., not available. The PD prescription was unchanged throughout. Anuria was defined as urine output below 200 mL/24h.

Erythropoietin resistance and anaemia

Six patients were receiving ESA at baseline (four epoetin alfa, two darbepoetin alfa). Doses were unified using the conversion 1 µg darbepoetin = 200 IU epoetin. The ERI decreased in all six patients (Figure 1B) although the magnitude of change varied widely and was clinically relevant in only three. Responses ranged from complete ESA withdrawal (Patient 1, whose haemoglobin rose from 11.8 to 15.3 g/dL) to changes below three per cent in two patients. In one patient the index improved only because the dose was reduced by 60% while haemoglobin fell from 9.4 to 8.6 g/dL, which does not represent improved responsiveness. In another the direction of change depends entirely on the conversion factor, because the agent was switched during follow-up. The clearest response was that of Patient 7, whose haemoglobin rose from 8.6 to 9.9 g/dL on an unchanged dose. Median haemoglobin for the cohort rose from 10.4 to 10.8 g/dL, and individual values are given in Table 3.

Table 3. Anaemia, metabolic parameters and clinical outcome by patient.

Hb, haemoglobin (g/dL); ESA, erythropoiesis-stimulating agent; ERI, erythropoietin resistance index (weekly epoetin-equivalent IU per g/dL), calculated with the conversion 1 µg darbepoetin alfa = 200 IU epoetin alfa; urate in mg/dL; P/Cr, urinary protein-to-creatinine ratio (mg/g); HDF, haemodiafiltration; RRT, renal replacement therapy; n.a., not available.

Serum urate

Serum urate fell in the five patients with preserved diuresis, by between 0.9 and 6.6 mg/dL. It was unchanged in the patient who became anuric during the first month and rose, from 6.6 to 7.0 mg/dL, in the only patient anuric throughout the entire follow-up. That same patient showed the largest sustained increase in ultrafiltration of the series.

Safety and clinical outcomes

No hypoglycaemia, euglycaemic ketoacidosis or genital mycotic infection was recorded. One patient required temporary withdrawal for acute kidney injury, with uneventful resumption; this episode occurred before PD initiation. Three patients died during follow-up: two from end-stage cardiac disease in the context of type 2 cardiorenal syndrome, and the third on 31 May 2025 from fungal peritonitis, after approximately six months of dapagliflozin. In this patient, for whom PD was the only available modality, the peritonitis was treated according to the fungogram, but he was persistently non-adherent, subsequently declined treatment and did not attend appointments, and the infection recurred in that setting.

Encapsulating peritoneal sclerosis (patient 7)

This patient started PD in October 2021 for long-standing chronic kidney disease and developed cirrhosis subsequently; the terminal picture was a cardio-hepato-renal syndrome rather than hepatorenal syndrome as the cause of kidney failure. Encapsulating peritoneal sclerosis was diagnosed in April 2024, after approximately 30 months on the technique, and confirmed histologically on surgical specimens. Dapagliflozin was started as adjuvant therapy at that time. Residual kidney function improved progressively and PD was discontinued in January 2025, after nine months of treatment. This did not represent recovery from acute kidney injury nor resolution of chronic kidney disease: the patient remained in stage 4 and has continued without renal replacement therapy since. No ultrafiltration data are reported for him, since in cirrhosis with ascites the drained volume does not equate to ultrafiltration.

Discussion

This case series provides real-worldobservations on the use of SGLT2i in an unselected, predominantly non-diabetic PD population. Its value is clinical and mechanistic rather than statistical: it documents outcomes in situations — anuria without alternative modality, recurrent congestive admissions, encapsulating peritoneal sclerosis — in which the PD community currently improvises without evidence.

Ultrafiltration increased substantially in three of the four patients in whom the drug was added to an unchanged PD prescription, consistent with the documented expression of SGLT2 in human peritoneal biopsies (9). Because the prescription was not modified, the obvious alternative explanation — a higher dialysate glucose concentration or additional dwells — is excluded by design. We make no claim of statistical significance: with four patients none could be supported, and the sample-size assumptions of the ongoing EMPOWERED trial (17) indicate that a larger cohort will be needed to quantify this effect. Our observations are directionally consistent with previous reports in PD (12–14) and with real-world data on anaemia and residual kidney function (15).

The strongest mechanistic argument in this series arises from a dissociation observed within a single patient

Serum urate fell only in those with preserved diuresis and rose in the patient anuric throughout, as the physiology predicts, since the uricosuric action of these agents requires urine flow. The anuric patients therefore act as an internal control confirming that the renal route behaved as expected. Yet that same patient, in whom no renal drug effect was demonstrable, showed the largest sustained ultrafiltration gain of the series. A response present in the peritoneum and absent in the kidney, in the same individual at the same time, argues for a direct peritoneal mechanism. That peritoneal transport category did not change in any patient with paired equilibration tests points the same way, suggesting improved osmotic conduction — slower dissipation of the glucose gradient (3, 10) — rather than altered small-solute permeability.

Transport phenotype may condition the response

The three responders were average or high-average transporters, whereas the single non-responder was the only low-average transporter, the phenotype with the least baseline peritoneal glucose absorption and therefore the least to gain from blocking it. With four patients this is a hypothesis, but one that deserves prespecified stratification in prospective trials.

The erythropoietin resistance index fell in every treated patient, which is consistent with the HIF-2α–mediated erythropoietic effect of these agents (8). The magnitude, however, was clinically relevant in a minority, and the index must be interpreted alongside the haemoglobin trajectory: a fall driven by dose reduction in a patient whose anaemia worsened is not improved responsiveness. The absence of systematic iron and inflammatory data is a material limitation for any statement about erythropoietin responsiveness.

A safety observation requires emphasis

One patient died of fungal peritonitis while receiving dapagliflozin. Causality cannot be inferred from a single case in a non-adherent patient who declined treatment, and fungal peritonitis in PD is classically associated with prior antibiotic exposure. Nevertheless, the mechanism we propose for the ultrafiltration benefit — reduced peritoneal glucose absorption through inhibition of mesothelial SGLT2 — necessarily implies a higher residual intraperitoneal glucose concentration for a longer dwell period. This is a biologically plausible route to increased fungal risk, and it has not to our knowledge been raised previously. In a field where this off-label use is spreading ahead of the evidence, we consider it necessary to state the hypothesis and to recommend explicit surveillance for fungal peritonitis in prospective studies.

We deliberately refrain from drawing an analogy with the haemodynamic pattern described in DAPA-CKD and EMPA-KIDNEY. On restricted analysis of the patients with complete paired data, the apparent early fall in residual kidney function was generated by a single patient with an intercurrent clinical event, and by median values no such fall existed. Extrapolating a tubuloglomerular feedback pattern to patients with minimal residual function and concurrent peritoneal clearance is not supportable with these data.

The absence of drug-attributable adverse events than the event discussed above is consistent with pharmacokinetic data confirming no dapagliflozin accumulation in PD (16).

Limitations

This study has several important limitations. The very small sample size (n = 7) precludes any inferential conclusion, and no hypothesis testing was therefore attempted; the observations are hypothesis-generating only. Paired ultrafiltration data were available in four patients. The retrospective, single-center design without a control group prevents causal inference. ESA heterogeneity necessitated dose conversion with a standard factor that has known inter-individual variability, and in one patient the direction of change in the resistance index depends on that factor. Follow-up duration and data completeness varied across patients. One patient began therapy three weeks after starting PD, so his baseline ultrafiltration is that of an incident patient. Finally, the single diabetic patient is also the only one given empagliflozin, the largest ultrafiltration responder and the only case of ESA withdrawal, so both principal observations derive disproportionately from the least representative patient; in a diabetic, glycosuria-driven osmotic diuresis and improved glycaemic control confound both outcomes. The mortality observed, although largely attributable to underlying disease, limits long-term outcome assessment.

Conclusion

SGLT2 inhibition augmented peritoneal ultrafiltration in patients receiving an unchanged dialysis prescription, and did so in the absence of any demonstrable renal effect, which places the site of action in the peritoneum rather than the kidney. In patients without residual function and without alternative modality, this gain was the difference between a viable technique and none. Erythropoietin requirements fell consistently in direction though not uniformly in magnitude, an effect that does not appear to require preserved renal parenchyma. Against this stands a fungal peritonitis whose plausibility follows from the very process that confers the ultrafiltration benefit, and which must be actively sought rather than assumed absent. Prospective controlled evaluation should stratify by peritoneal transport phenotype and monitor infectious outcomes explicitly.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The research results were visually processed with the assistance of an AI tool. This technology was exclusively employed for the generation and rendering of the figures (author’s own production using AI based on the author’s data and presentations). The author conducted an exhaustive quality control review of all visual material to ensure accurate representation of the research data and to correct any technical inaccuracies before its final inclusion in this document.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fneph.2026.1942462/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Anil Agarwal, United States Department of Veterans Affairs, United States

Reviewed by: Shreepriya Mangalgi, University of Missouri, United States

Shogo Kuwagata, Shiga University of Medical Science, Japan