Revisa el resumen visual por Farah Wehbe

Check the VA by Farah Wehbe

N Engl J Med. 2026 Oct 14. doi: 10.1056/NEJMoa2608397. 

Intermittent or Continuous Renal-Replacement Therapy in Acute Kidney Injury

Stéphane Gaudry, Marouane Boubaya, Guillaume Louis, Khalil Chaïbi, Etienne de Montmollin, Julien Bohé, Maxime Desgrouas, Clément Devautour, Adrien Joseph, Bruno Mégarbane, Christophe Camus, Pierre Bailly, Alexis Lambour, Romain Arrestier, Samir Jaber, Adrien Robine, Nicolas Chudeau, Guillaume Chevrel, Béatrice La Combe, Jonathan Nicolas, Laurent Argaud, Côme Bureau, Antoine Kimmoun, Noémie Zucman, Jean-Damien Ricard, Guillaume Thiery, Stanislas Faguer, Mathieu Jozwiak, Nathan Ferrand, Benjamin Chousterman, Mehdi Marzouk, Julien Cadiet, Alain Combes, Arnaud Mari, Damien Contou, Roland Amathieu, Daniel Da Silva, Charlotte Salmon Gandonnière, Pierre-André Rodie-Talbere, Isabelle Durand-Zaleski, Patrick Rossignol, Jean-Pierre Quenot, Lucie Biard, Coralie Bloch, Didier Dreyfuss; ICRAKI Trial Group

Affiliations Expand

• PMID: 42873070

• DOI: 10.1056/NEJMoa2608397

Introduction

Severe AKI requiring renal-replacement therapy (RRT) in the ICU carries a higher risk of short-term mortality, prolonged hospitalization, and long-term chronic kidney disease. There are many potential RRTs used for AKI in the ICU including: intermittent hemodialysis (IHD),continuous hemodialysis, hemofiltration, or hemodiafiltration (CRRT), sustained low efficiency dialysis (SLED), and prolonged intermittent renal replacement therapy (PIRRT). Each “flavor” of dialysis has its own characteristics that affect volume and solute removal capacity. Which should be used first, and are there specific benefits or hazards associated with dialysis choice in critical patients, has been debated for decades. The KDIGO guidelines make a weak recommendation favoring CRRT in “hemodynamically unstable” patients. 

Randomized evidence is both neutral and dated. The largest previous trial, Hemodiafe (360 patients with multiorgan dysfunction), found no difference in 60-day survival between IHD and continuous venovenous hemodiafiltration (mean difference 1.1%, 95% CI −8.8 to 11.1) and no difference in renal recovery (Vinsonneau et al., Lancet, 2006). A Cochrane review of 15 RCTs (1,550 patients) found no difference in mortality, renal recovery, or hypotension between dialysis modalities in hemodynamically stable patients, although the dose of dialysis may have varied significantly in the included studies (Rabindranath, et al. Cochrane Database Syst Rev, 2007). Two post hoc analyses of RRT-timing trials reached opposite conclusions and reopened the debate.A pooled analysis of AKIKI and IDEAL-ICU trials linked first-line CRRT to shorter 60-day survival, particularly in less severely ill patients ( Grolleau et al. Crit Care, 2022). Antithetically, STARRT-AKI in 2023 linked first-line CRRT to a lower risk of death or RRT dependence at 90 days (OR 0.81, 95% CI 0.66 to 0.99), with most of the effect from decreased dialysis dependence (OR 0.61, 95% CI 0.39 to 0.94).

With conflicting evidence, and often significant heterogeneity of patients across multiple studies, does dialysis modality really matter to patients with AKI in the ICU? This brings us to ICRAKI, which tested whether IHD as the first modality is non-inferior to CRRT for major adverse kidney events at day 90 (MAKE90).

The Study

Methods

Trial Design: Investigator-initiated, open-label, randomized noninferiority trial in 38 French ICUs. All sites were experienced with both RRT modalities. An independent data and safety monitoring board reviewed day-28 mortality. Its interim analyses were Bayesian and descriptive, with no prespecified stopping boundaries. The statistician who ran them is a senior author on the paper. 

Funding

 The French Ministry of Health publicly funded the trial, while Assistance Publique – Hôpitaux de Paris (AP-HP) served as the sponsor.

Inclusion Criteria 

Adults in the ICU with AKI and at least one recognized RRT indication. Patients also had to be receiving, or have received, invasive mechanical ventilation and/or catecholamines. Both IHD and CRRT had to be available at the site.

The indications were:

• refractory hyperkalemia

• refractory metabolic acidosis

• refractory pulmonary edema

• KDIGO stage 3 AKI with urea >40 mmol/L

• KDIGO stage 3 AKI with oligoanuria for more than 3 days

Suggested thresholds for the metabolic indications were:

• potassium >6.5 mmol/L despite medical treatment

• pH <7.15 with PaCO2 <35 mmHg

• PaO2/FiO2 <200 with clinical volume overload

These thresholds were for guidance only. Investigators could start RRT below them if they chose. 

Exclusion Criteria 

• Moribund state

• eGFR <20 mL/min/1.73 m²

• Dialyzable toxin

• Suspected obstruction, rapidly progressive glomerulonephritis, vasculitis, thrombotic microangiopathy, or interstitial nephritis

• Brain injury or elevated intracranial pressure, 

• Fulminant hepatic failure

• Pregnancy

Randomization

1:1 through a central web-based system. Stratification factors were cumulative fluid balance (≥6 L vs <6 L), and norepinephrine/epinephrine dose (>0.5 vs ≤0.5 µg/kg/min). Vasopressin and angiotensin II were not stratified because their use is rare in France.

Interventions

The assigned RRT modality was mandatory from day 0 to day 3 unless the kidneys recovered. Switching was allowed from day 4.

IHD protocol recommended:

• At least 3 sessions per week, each 4 to 6 hours

• Blood flow >200 mL/min and dialysate flow >500 mL/min

• Dialysate sodium >145 mmol/L and dialysate temperature 35°C

• Bicarbonate dialysate

• Median session length was 4 hours, Kt/V was about 1.0, and median net ultrafiltration was about 375 mL/h (table s3)

CRRT was CVVHD, CVVH, or CVVHDF at clinician discretion. The protocol recommended:

• Delivered effluent of 20 to 25 mL/kg/h

• Filter change every 72 hours

• Regional citrate anticoagulation

Stopping RRT was to be considered when urine output exceeded 500 mL/24 h. It was strongly recommended above 1,000 mL/24 h without diuretics or 2,000 mL/24 h with diuretics.

Outcomes

• Primary: MAKE90, defined as all-cause death, RRT dependence, or serum creatinine >25% above baseline at day 90. When no measured baseline creatinine was available, it was back-calculated with MDRD assuming an eGFR of 75.

• Secondary: time-to-event mortality (days 28, 60, and 90), RRT dependence and creatinine elevation at days 60 and 90, organ-support-free days, length of stay, eGFR at ICU discharge, and prespecified complications.

• The protocol originally defined hypotension by blood pressure (systolic <80 mmHg or a fall >50 mmHg). This was later amended to an intervention-based definition: starting or increasing catecholamines, or giving more than 250 mL of IV fluid. The rationale for the change was that blood-pressure data was not systematically available.

Statistics

The trial was a non-inferiority design (see Manasi’s post for more on this design) between CRRT and IHD. The sample was intended to be 1,000 patients, assuming a MAKE90 incidence of 75%, for 80% power. The noninferiority margin was 7.5 percentage points absolute, or about 10% relative. This would mean that the 95% CI of MAKE90 had to be within this margin to establish non-inferiority. The main paper says the margin follows regulatory guidance, but the appendix says no such guidance exists and it was chosen pragmatically. The prespecified "intention-to-treat" population included randomized patients who received at least one RRT session. The authors acknowledge this as a modified ITT. The per-protocol population included eligible patients who received the assigned modality and did not switch between days 0 and 3. (note that a per protocol design might be superior - no pun intended - than an ITT design in a non-inferiority design; Sanchez and Chen Stat Med 2006). 

Results

The study assessed for eligibility 1268 participants from October 2023 to July 2025. 554 underwent randomization with 272 in the iHD and 282 in the CRRT group (as shown in figure S1).  The study had planned to enroll 1,000 patients but was terminated early. Two interim analyses were initially planned; The second planned look was brought forward from 600 to 450 patients. After it, enrollment was suspended and a third look was added.  The DSMB ultimately advised definitive termination because of a survival difference favoring CRRT. There was no pre-specified numerical boundary to terminate the trial. 

The baseline characteristics were generally balanced, although surgical admissions and pre-existing chronic kidney disease were more frequent in the IHD group. 

Table 1. Baseline characteristics of study participants, from S. Gaudry et al, NEJM 2026.

 The median duration of renal replacement therapy was 5 days (interquartile range [IQR], 2 to 15) in the IHD group and 7 days (IQR, 3 to 15) in the CRRT group. The RRT delivered over 28 days in the two study arms is shown in figure S2 below. Overall, 57 of 260 patients (22%) in the intermittent-hemodialysis group and 117 of 275 patients (43%) in the continuous renal-replacement-therapy group received the alternative modality during follow-up. Although the study required zero to 3 days of adherence to be assigned a dialysis modality, 10 patients (3.8%) assigned to intermittent hemodialysis and 9 patients (3.3%) assigned to continuous renal-replacement therapy switched before day 4. Authorized switches from CRRT to IHD occurred in 108 patients and also in 47 patients from IHD to CRRT.

 The proportion of patients receiving catecholamines on the day of switching was higher among those switched from IHD to CRRT than among those switched from CRRT to IHD (79% vs 47% overall; 77% vs 43% among protocol-authorized switches after day 3), suggesting that switching from IHD to CRRT was more often associated with ongoing hemodynamic instability and pressor support.

Outcomes

Data on survival were available for 99.3% of patients. The primary outcome, major adverse kidney events at day 90 (MAKE-90) occurred in 78% versus 72% in those assigned to IHD and CRRT, respectively. The adjusted absolute risk difference was 6.3%, (95% confidence interval [CI], −1.22 to 13.79) represented by the solid black box in figure S3, with an adjusted risk ratio of 1.09 (95% CI, 0.98 to 1.20; P=0.10). Because the upper boundary of the confidence interval exceeded the prespecified noninferiority margin of 7.5 percentage points (dotted red vertical line in figure S3), noninferiority of IHD was not established. The difference in the primary outcome also did not establish statistically significant superiority of CRRT. Results in the per-protocol population were similar. Mortality was higher in patients on IHD  with HR of 1.3 (95% CI, 1.04 to 1.63), with the Kaplan-Meier Curves of probability of death showing the difference from day 7 through day 90 (Figure 1). The details of outcomes are shown in Table 2. The mortality result needs careful reading. The trial set a separate noninferiority margin for death at day 90: an HR of 1.27. Noninferiority would have required the upper confidence bound to fall below this value, and it did not in any analysis. The lower bound, however, is close to 1. After adjustment for baseline imbalances, the HR is 1.26 (95% CI 1.01-1.58); in the per-protocol population, it is 1.27 (1.01-1.60), so both only just exclude no difference. In absolute terms, the margin of 1.27 allowed for 9 extra deaths per 100 patients with IHD, close to roughly 10 per 100 actually observed.

The subgroup analyses demonstrates numerically higher point estimates for CRRT over HD across pre-specified groups  (figure 2), however, in most subgroups confidence intervals were wide and included the null effect.

Discussion

The ICRAKI trial addresses a longstanding question in critical care nephrology: whether intermittent hemodialysis (IHD) or continuous renal-replacement therapy (CRRT) should be initiated in critically ill patients with severe acute kidney injury (AKI). The trial was stopped early, an ethically necessary decision, after interim monitoring showed higher mortality with IHD, a signal that persisted to day 90. Because of early termination, the trial was underpowered for its primary outcome: IHD was not shown to be non-inferior to CRRT for MAKE 90, but neither was it shown to be inferior.

The MAKE90 result is in line with earlier trials and a recent systematic review (Ma H et al, Evid Based Complement Alternat Med, 2023), none of which found a clear difference in mortality or kidney recovery between IHD and CRRT. Some trials reported numerically higher mortality with IHD that did not reach significance (CONVINT trial - Schefold et al, Crit Care, 2014). The only earlier trial with a significant mortality difference (Mehta et al, Kidney Int, 2001) favored IHD, and that result was attributed to baseline imbalance between groups. One plausible reason for the discordance is the IHD prescription. Hemodiafe (Vinsonneau C et al, Lancet, 2006), the largest earlier trial, used longer, near-daily. sessions with high sodium, cool dialysate, whereas ICRAKI used conventional 4-hour sessions.

The higher mortality in IHD as compared to CRRT found in ICRAKI  warrants further consideration, particularly given the early termination of the trial. The separation in Kaplan-Meier curves was apparent 7 days after randomization and persisted through 90 days. Although stopping a trial in response to a concerning mortality signal may be ethically justified, early termination can exaggerate the apparent magnitude of a treatment effect and limit the precision of its estimates. There is no group of patients in whom CRRT is the only option. Depending on local practice, unstable patients receive CRRT, SLED, or carefully prescribed IHD. The question ICRAKI raises is not which patients need CRRT, but how IHD was delivered. Here it was prescribed like chronic dialysis: 4-hour sessions aiming at a urea reduction target, with median hourly ultrafiltration of about 375 ml/h, in patients most of whom were on vasopressors. Not many clinicians would prescribe IHD this way for hemodynamically unstable patients. ICRAKI therefore compares CRRT with a particular IHD prescription, not with intermittent therapy as it is usually practiced. Mortality in ICRAKI (54-63% at day 90) appears higher than that reported in the delayed-treatment groups of previous AKI trials, such as AKIKI and STARRT-AKI. But this comparison is misleading: many patients in those arms never needed RRT. Among delayed-arm patients in AKIKI who did go on to receive RRT, 60-day mortality was around 60%, similar to ICRAKI. ICRAKI’s mortality is therefore typical of critically ill patients who actually require dialysis, rather than a marker of usually severe populations.

The substantial crossover between groups further complicates interpretation. Switching to the alternative modality occurred in approximately 22% of patients assigned to IHD and 43% assigned to CRRT, with a small number of switches occurring before the protocol-permitted period. Thus, ICRAKI is not a direct comparison of sustained IHD versus sustained CRRT throughout the course of AKI; rather, it evaluates the choice of initial RRT modality, with subsequent treatment decisions made according to the clinical course. The day-28 and day-90 outcomes remain valid for this initial-strategy comparison, but cannot establish the comparative effects of sustained treatment with either modality.

Catecholamine use on the day of randomization was similar in the two groups, although more IHD patients had received catecholamines at ICU admission (85% vs 78%). During RRT, patients on IHD more often needed catecholamine escalation (25% vs 21% of RRT days) and received higher doses (table S8). This is what one would expect from 4-hour sessions removing 375 mL of fluid per hour in patients on vasopressors, and it fits the argument that the IHD prescription, rather than intermittent therapy itself, is the problem. These analyses were post hoc, however, and the overall incidence of hypotension was similar (69% vs 67%). That figure is also hard to interpret: after a protocol amendment, hypotension was defined by clinician actions (starting or increasing catecholamines, or giving more than 250 mL of fluid) rather than by blood pressure, which is a subjective measure in an open-label trial.

From a practical standpoint, the choice is not always between conventional IHD and CRRT. Prolonged intermittent therapies such as sustained low-efficiency dialysis (SLED) are widely used in critically ill patients and may offer an alternative when haemodynamic tolerance, resource availability or cost is a concern. The findings of ICRAKI cannot be directly extrapolated to prolonged intermittent therapies. The availability and cost of CRRT, together with the expertise of the treating centre and the availability of trained personnel, also influence its effective delivery. Frequent circuit downtime and interruptions can compromise solute clearance and the control of fluid balance, electrolytes and acid–base status, particularly when manpower and technical support are limited.

Will ICRAKI change our practice? Probably not, due to substantial limitations from early termination. Although the trial raises concerns about initiating conventional IHD in critically ill patients with severe AKI, it does not establish the superiority of sustained CRRT over other modalities. However, if a centre is practicing IHD, there is little reason not to utilise SLED instead, which uses the same machines and staff and arguably is better tolerated hemodynamically, as elaborated in a narrative review by Clark EG et al. The choice of kidney-replacement therapy will continue to depend on the patient's clinical condition, haemodynamic status, local expertise, resource availability and the ability to deliver the chosen modality adequately.

Conclusion

ICRAKI attempted to determine the non-inferiority of IHD to CRRT as the initial dialysis modality in ICU patients with AKI. Due to an unexpected elevated rate of mortality in IHD patients, the study was stopped early and only enrolled half of the number of planned patients. Further conclusions about the primary outcome, MAKE90, are thus underpowered and hypothesis generating at best, and do not change the current practice of physician preference.

Summary by

David White FNP-C 

Smita Divyaveer,

Associate Professor,

Department of Nephrology,

PGIMER, Chandigarh, India 

Reviewed by

Swapnil Hiremath, Brian Rifkin, Cristian Popa

Header Image created by AI, based on prompts by Farah M Wehbe

For each study, we’ll unpack the findings, limitations, and consider what they could mean for kidney care. Whether you’re joining the meeting or following along online, pull up a chair and join the discussion.

We did this last year (link) with excellent feedback, so we are bringing it back. Swap and Cristina talked about the rationale (mostly FOMO!) and how this was done here.

Day 3 Studies (flozins dominate)

• Povetacicept in IgA (RAINIER AIS D08) Barratt et al NEJM 2026 

• Empagliflozin therapy in patients with ADPKD,  Selinger et al, JASN, 2026

• Flozins increase Hgb: Metanalysis, Yang et al JAMA 2026

• A post hoc analysis from CONFIDENCE on apparent treatment resistant hypertension, (Agarwal et al, JACC Advances 2026)

NEJM | Original article

Interim Analysis of a Phase 3 Trial of Povetacicept in IgA Nephropathy

Jonathan Barratt, Hong Zhang, Seung Hyeok Han, Yuzuke Suzuki, James Tumlin, Bobby Chacko,  Irene L. Noronha, Arvind Madan, Raphael Duivenvoorden, Jamil R. Azzi, Carla Nester, Cassie Dong, Neil Ahluwalia, Jiahua Li, Carmen Bozic, Ogo Egbuna, Muh Geot Wong, and Richard Lafayette, for the RAINIER AIS-D08 Trial Group

Published October 23, 2026 DOI: 10.1056/NEJMoa2608682

Why was the study done?

Still with us? IgA nephropathy is not leaving the stage just yet. After targeting a complement with ravulizumab on day 2, we move back upstream to the antibody supply. Enter BAFF and APRIL—two signals that help sustain the B cells and plasma cells involved in producing pathogenic IgA. If those names sound familiar,  atacicept, sibeprenlimab and telitacicept coverage might explain why! NephJC summary, TELIGAN 

Povetacicept is an engineered TACI–Fc fusion protein that inhibits both BAFF and APRIL. The idea is to reduce galactose-deficient IgA1 and the immune complexes that set kidney inflammation in motion. Its enhanced binding properties are an interesting design feature; whether they deliver better clinical outcomes than dual inhibition or APRIL alone is a separate question.

While earlier trials provided an encouraging opening act, demonstrating UPCR reduction (RUBY-3, Madan et al, KI Reports 2026). The phase 3 RAINIER trial evaluates monthly dual BAFF/APRIL blockade to deliver proteinuria reduction with placebo, on top of supportive care? A different target, but can say the same immediate focus as ravulizumab. 

What did they do?

This is a randomized, double-blind, placebo-controlled trial conducted at 224 sites in 29 countries. Adults had biopsy-confirmed IgA nephropathy, eGFR ≥30 mL/min/1.73 m², and persistent proteinuria ≥1 g/day or UPCR ≥0.75 g/g, despite stable background therapy. Participants were randomized 2:1 to subcutaneous povetacicept 80 mg or placebo every four weeks.

Figure S1, RAINER trial design, from Barratt, NEJM 2026

The main cohort included 557 participants,  but only the first 199 randomized participants contributed to this interim efficacy analysis. The efficacy population was defined by sufficient opportunity for follow-up, not simply by selecting treatment completers; participants who discontinued early could still qualify.

The primary endpoint was changed in 24-hour UPCR at week 36. The secondary endpoints were Gd-IgA1 change and hematuria resolution among participants with hematuria at baseline.

Figure S1, RAINER enrollment diagram, from Barratt, NEJM 2026

What do the results say?

From table 1, the GFR was higher in the povetacicept (74) versus placebo (68) though proteinuria was similar (~ 1.5 g/day). Likely chance - will see how final cohort shapes up.

At week 36, UPCR decreased by 52.0% with povetacicept versus 4.3% with placebo, corresponding to an approximately 49.8% lower adjusted UPCR relative to placebo (95% CI, 39.2–58.6; P<.001).

Gd-IgA1 decreased by 77.4% with povetacicept, compared with a 9.1% increase with placebo. Among participants with baseline hematuria, 85% versus 23% met the resolution definition, an adjusted absolute difference of 62 percentage points. Both secondary endpoints were statistically significant.

In the safety population, overall adverse events occurred in 75% versus 78%, and serious adverse events in 3% versus 4%. However, infections were numerically more frequent (46% vs 41%), as were injection-site reactions (15% vs 2%). No deaths occurred.

Figure 1, Percentage change in 24-hour UPCR and in spot UPCR, from Barratt, NEJM 2026

How should we interpret it?

Another IgAN trial, another substantial proteinuria reduction, but RAINIER adds a biologic coherence to the story. Proteinuria, galactose-deficient IgA1, and hematuria all improved, supporting activity against the disease’s underlying immune pathway. The design, the relevant add-on setting, the prespecified analysis and controlled testing strengthen these findings.

However, lowering proteinuria is not a surrogate of long term kidney protection. This interim efficacy analysis included 199 participants, and the effects on eGFR slope, disease progression, and longer-term safety await the 104-week analysis. Hematuria “resolution” also allowed small amounts of blood on dipstick, so it should not be read as complete disappearance in every patient. 

Monthly subcutaneous administration could be convenient, not sure if  comparative superiority or cost-effectiveness. RAINIER cannot tell us whether povetacicept is better than atacicept, telitacicept, or APRIL-only inhibition, or which combinations offer the best balance of benefit and risk. Immunoglobulin reductions and numerically more infections warrant continued surveillance, despite similar short-term serious adverse event rates. Extensive sponsor involvement also deserves consideration. For now, povetacicept offers convincing evidence of reduced disease activity; whether it delivers durable kidney protection remains the next test. Perhaps at #KidneyWk 2027?

Journal of the American Society of Nephrology Publish Ahead of Print, October 24, 2026.

Empagliflozin Therapy in Patients with Autosomal Dominant Polycystic Kidney Disease: A Feasibility Randomized Controlled Trial

Stephen L. Seliger, Berenice Gitomer, Zhiying You, Wei Wang, Anna Ostrow, Emily Andrews, Charalett Diggs, Diana George, Miranda West, Kristen L. Nowak, Terry Watnick, and Michel Chonchol

DOI: 10.1681/ASN.0000001271

Flozins appear now in almost every CKD algorithm except one. Patients with ADPKD, the most common inherited kidney disease, were excluded from DAPA-CKD (Heerspink et al, NEJM, 2017| NephJC summary) and EMPA-KIDNEY (EMPA-KIDNEY collab group, NEJM, 2022| NephJC summary). The US labels for dapagliflozin and empagliflozin still advise against using them for kidney protection in PKD. Tolvaptan remains the only disease-modifying drug (Torres et al, NEJM, 2017 | NephJC summary), and its aquaresis and liver signal keep many eligible patients off it. 

The physiology argues both ways. On the optimistic side, young adults with ADPKD already show a raised filtration fraction and reduced renal plasma flow while GFR is still preserved (Meijer et al, CJASN, 2010), which is precisely the lesion that restoring tubuloglomerular feedback should ease. Flozins also induce mild ketosis, and higher beta-hydroxybutyrate levels track with slower eGFR decline in ADPKD (Knol et al, NDT, 2024), and in a human organoid-on-chip, glucose uptake drove cystogenesis while SGLT inhibition limited it (Li et al, Nat Commun, 2022). 

The pessimist has an equally good story. The body answers an osmotic diuresis by conserving water, and copeptin rises on SGLT2i (Lytvyn et al, Diabetes Metab, 2020; Marton et al, JACC, 2024). In ADPKD that is the wrong hormone to provoke, since the vasopressin-V2R-cAMP axis is the very cyst engine that tolvaptan blocks. The animal data are a muddle (Naeem et al, Ren Replace Ther, 2026): the only clear win came from phlorizin, a nonselective SGLT1/2 inhibitor, in a non-orthologous rat model with proximal cysts (Wang et al, Kidney Int, 2013), while selective dapagliflozin raised cyst volume by 23% in PCK rats (Kapoor et al, PLoS One 2015), and canagliflozin did nothing useful in an adult-onset Pkd1 mouse (Leonhard et al, EbioMedicine, 2019). Empagliflozin also enlarges rodent kidneys simply by increasing cell size in the S3 segment and collecting duct (Sinha et al, Front Pharmacol, 2023), so on this drug class TKV is not automatically a cyst measure.

Table 1. Pre-clinical evidence of SGLT2i in PKD animal models, from Naeem et al, Ren Replace Ther, 2026

Human data are thin and contradictory. In an open-label cross-over of 27 tolvaptan-treated patients, dapagliflozin slowed kidney enlargement and chronic eGFR decline over 6 months (Uchiyama et al, Kidney Int Rep, 2025), whereas a retrospective series described TKV increases (Morioka et al, J Clin Med, 2023). Before committing to a 400-patient trial, the investigators asked a narrower question (Selinger et al, Am J Nephrol, 2026): can non-diabetic adults with rapidly progressive ADPKD take empagliflozin 25 mg for a year without trouble, and does anything worrying happen to kidney volume, eGFR, or copeptin along the way?
How was the study done, and what did it show?

This is a double-blind, placebo-controlled, parallel group feasibility trial at two US centers (Maryland and Colorado), with 1:1 randomization blocked by site at 12 months of treatment. This is the final analysis, and funding was academic (NIDDK, PKD Foundation, philanthropy). Patients started on empagliflozin 10 mg and moved to 25 mg at 1 month if they tolerated  and eGFR had not fallen by more than 30% or below 20. The 25 mg dose was chosen in 2021, before EMPA-KIDNEY settled on 10mg.

Eligible adults had ADPKD by modified Pei-Ravine criteria, a CKD-EPI 20221 eGFR of 30-90, Mayo class 1C-1E, and stable function over 3 months. Diabetes, HFrEF, loop diuretics, current urogenital infection, and tolvaptan were all exclusions. The co-primary outcomes were safety (all AEs and SAEs, plus AKI, UTI, and genital infection) and tolerability, defined as taking 25 mg at 12 months and answering yes to the TEMPO-era question: can the patients tolerate the drug for the rest of their life? At least 70% tolerability on empagliflozin was the prespecified signal to proceed to phase 3. Secondary outcomes at 3 and 12 months included hTKV by blinded 3T MRI, piecewise eGFR slopes, aPWV, copeptin, urinary KIM-1, and ADPKD Impact scale. 

50 of 104 screened patients were randomized, with no losses to follow-up. Mean age was 42, 96% were White, mean eGFR was 70 +/- 17; and 88% took an ACEI or ARB. Randomization left imbalances that matter for a volume endpoint: the empagliflozin arm had more men (60% vs 48%), higher eGFR (73 vs 67), larger kidneys (median htTKV 1,114 cs 1,030 mL/m) and more Mayo 1E patients (32 vs 20%).

Figure 1. CONSORT diagram: flow of the trial, from Selinger et al, JASN, 2026

Tolerability was 24 of 25 in both arms (96%; exact 95% CI 79.7-99.9), well clear of the 70% bar (P=0.002). One patient per arm stopped after a hospital admission (myocardial infarction on empagliflozin, cholecystitis on placebo). Any AE occurred in 12 vs18 patients (P=0.15) with one SAE per arm; empagliflozin brought two UTIs and one genital infection, and the only AKI was on placebo. 

Table 2. Safety and tolerability, by treatment assignment, from Selinger et al, JASN, 2026

The secondary outcomes held the surprise. htTKV grew 10.5% per year on empagliflozin (95% CI 8.2-12.7) against 5.1% per year on placebo (95%CI 3-7.3), a difference of 5% per year (95%CI 2.0-8.1, P=0.001 for time x treatment). Post hoc adjustment for Mayo class and sex did not change this. 

Figure 2. Median (IQR) values of htTKV per group over time,  from Selinger et al, JASN, 2026

eGFR showed the expected acute dip at 1 month (-6.2 vs +1; between-group -5.5, 95% CI -9.6 to -1.3) followed by a chronic slope of -0.77 vs-2.87 ml/min/1.73m2 per year (difference 1.4; 95%CI -3.3 to 6.2; P= .04). A PWV, copeptin, urinary KIM-1 and ADPKD-IS did not differ.

Table 3. Secondary efficacy outcomes, from Selinger et al, JASN, 2026

What does it all mean?

The feasibility question was answered convincingly: 96% tolerated empagliflozin 25mg for 12 months, with no excess of adverse events. The secondary finding is less reassuring: htTKV increased at about twice the placebo rate (10.5% vs 5.1% per year).

How much weight signal deserves is debatable. It is one of five unadjusted secondary outcomes in 50 patients; the empagliflozin arm started with bigger kidneys, and the analytic models differ from those in the published protocol. Yet the difference interval sits well clear of zero, adjustment did not erase it, and it matches the PCK rat. Placebo kidneys grew about as Mayo 1C_1E predicts, whereas 10.5% per year is well beyond the >6% that defines class 1E (Irazabal et al, JASN, 2015). Whether the extra volume is cyst is another matter. On our rough arithmetic from cross-sectional medians, most of the divergence happens in the first 3 months (=3.1% vs +0.6%) with near-parallel growth from month 3 to 12 (=5% vs 4.5). An early step like that looks more like tubular hypertrophy or another non-cystic shift than accelerated cystogenesis. Copepetin did not rise, but three fasting samples cannot exclude a meaningful increase, and there was no cyst-versus-parenchyma segmentation, no off-drug MRI, and albuminuria. The chronic eGFR slope leans toward empagliflozin, but its interval spans -3.3 to +6.2, and without a washout, hemodynamics and structure cannot be untangled.

The patients were mostly White, nondiabetic, tolvaptan-naive registry volunteers, given a dose no efficacy trial uses. If any harm runs through V2 receptors, flozins on top of tolvaptan might behave quite differently, which is what Uchiyama’s crossover hints at (Uchiyama et al, Kidney Int Rep, 2025).

The results also complicate the larger trials now underway: EMPA-PKD (Bahlmann-Kroll et al, BMJ Open, 2024), DAPA-PKD and STOP-PKD (Muller et al, NDT, 2025). If flozins enlarge kidneys without growing cysts, a TKV-primary trial could penalise a drug that preserves kidney function, while an eGFR-slope trial has to handle the dip. Until those trial reports, this pilot gives no reason to start SGLT2i in ADPKD outside established indications such as heart failure or DKD.

Table 2. Overview of clinical trials investigating SGLT2i in ADPKD, from Naeem et al, Ren Replace Ther, 2026

Sodium-Glucose Cotransporter 2 Inhibitor Use and Levels of Hemoglobin, Markers of Iron Metabolism, and Erythropoiesis: A Systematic Review and Meta-Analysis of Randomized Clinical Trials

Yaqin Yang, MD; Changyuan Yang, MD; Yanmei Wang, MD, Pien Rawee, MSc; Aleix Cases, MD, PhD; David E. Leaf, MD, MMSc Lucia del Vecchio, MD; Taulant Muka, MD, PhD; Hiddo J. L. Heerspink, PharmD, PhD; Ron T. Gansevoort, MD, PhD; Michele F. Eisenga, MD, PhD

JAMA Published Online: October 24, 2026 doi: 10.1001/jama.2026.19921

Why was the study done?

Flozins reduce rates of kidney failure, hospitalization with heart failure and cardiovascular as well as all-cause mortality in patients with diabetes, kidney disease or heart failure. Whilst flozinating these patients, one may notice the hemoglobin creeping up cometimes. The initial thought was that this might just be hemoconcentration from the osmotic natriuresis, however the natriuresis and volume contraction is rather mild - and the effect seems to be sustained. Since then, the mechanism is thought to be mediated by hypoxia induced erythropoietin stimulation, but even increased iron utilization is noted as seen by lower ferritin levels - and also lower hepcidin levels. This mechanistic study suggests that this might be mediated by the erythroferrone–transferrin-receptor-protein-1 driven reduction in hepcidin (Ferreira et al JACC 2025). Another hypothesis also suggests a difference between which HIF (1 or 2) are inhibited by flozins versus HIF-PHI agents (Packer, Am J Nephrol 2023). See image below for this mechanism. However, a proper quantitation of this hemoglobin rise - as well as iron indices, and in different populations has not been done from the RCT data. 

Figure: Possible mechanisms of Flozin (made with Nano banana)

How was it done and What did it show?

The systematic review included RCTs of flozins in diabetes, CKD, or heart failure populations, with a double-masked design, and reporting some hemoglobin index or iron metabolism numbers. The primary outcomes were change in hemoglobin and hematocrit with iron parameters and RBCs being secondary outcomes. The metanalytic component was fairly standard, with the usual assessment of publication bias, risk of bias, sensitivity and subgroup analysis which we will discuss in more detail with the actual results. 

Overall, of the 6000 odd studies screened, 85 RCTs with ~ 70,500 individuals made it to the final analysis. Empa (30 RCTs) followed by dapa (25) lead the pack, with cana (8) ertu (7) and then the other less well known ipra (6), luseo (4), sota (3) and bexagliflozin (1) contributing the rest. Most trials were just for DM, though notably 9 each were CKD and CV trials, but with an additional 14 HF trials. More details in the table below. 

Table 1 from Yang et al, JAMA 2026

Overall, there was a small, but significant and consistent increase in hemoglobin with flozination (~ 0.6 g/dL) with tight 95% CI given the large N. Only one small trial reported no change in hemoglobin.

Figure 2 from Yang et al, JAMA 2026

Flozination was also associated with a small increase in erythropoietin levels (N = 8 trials) with significant heterogeneity, a small increase in RBC number (pooled effect 0.2 x 1012/L). For iron indices, flozination was associated with a small decrease in ferritin, transferrin saturation, and hepcidin (figure 3 below). 

Figure 3 from Yang et al, JAMA 2026

The subgroups were pretty consistent when it came to use in different patient populations, age, geographic region. For different durations there was no significant interaction for hemoglobin, but there was one for hematocrit - with a greater increase noted over time.  Notably this was seen for both hemoglobin and hematocrit in the multivariable metaregression (eTables 9, 10, not shown here). . 

Figure 4 from Yang et al, JAMA 2026

What does it all mean? 

Flozins work for hard outcomes, what should we care about their effect on hemoglobin? For some reasons this is an important result - known to many of us, but good to see a properly conducted quantitative result. Could this be a partial explanation for the clinical outcomes? After all in DKD one can explain away flozin effects on the basis of TGF, but in non-diabetic CKD, in HF etc - just natriuresis seems insufficient - and is usually explained away as ‘pleiotropic’ mambo-jumbo. Flozins have several salutary metabolic effects - unlike RASi they reduce K a bit, unlike thiazides they lower blood sugars, uric acid and don’t cause hyponatemia. They may indeed help a bit in ameliorating hyponatremia or hypomagnesemia. Admittedly none of this metabolic effects are impressive by themselves but they do add up - and the hemoglobin and increased iron metabolism may be a similar contributor - especially in the heart failure population. In CKD - using flozination may help delay the need to use ESAs - or may help in using lower doses of ESAs? 

Another aspect to note is that sometimes this effect may be too much - case reports exist of the risk of polycythemia - and though the absolute risk is small, the relative risk was estimated to be about 6 times higher (Marques et al, Eur J Int Med 2026). This risk is slightly higher with COPD, smoking etc (perhaps in ADPKD?). As they note, it may be useful in patients with polycythemia to deflozinate before investigating for more esoteric causes.

Finerenone-Empagliflozin Combination Therapy in Apparent Resistant Hypertension: A CONFIDENCE Trial Post Hoc Analysis

Rajiv Agarwal MD, MS, Sankar D. Navaneethan MD, MS, MPH, Ricardo Correa-Rotter MD, Michelle M. Estrella MD, MHS, Hiddo J.L. Heerspink PhD, Janet B. McGill MD, Amy K. Mottl MD,  Masaomi Nangaku MD, PhD, Julio Rosenstock MD, Peter Rossing MD, Muthiah Vaduganathan, Javier Morales, Yehuda Handelsman MD, Jennifer B. Green MD, Johannes F.E. Mann MD, Charlie Scott MSc, Li Li MD, MBA, Meike Brinker MD, Matthew R. Weir MD

JACC Advancesdoi: 10.1016/j.jacadv.2026.103388 

Why was this study done?

In treatment resistant hypertension, based on the PATHWAY trial (Williams et al Lancet 2015), the first option should be spironolactone. In this patient population, it results in a blood pressure lowering of about 10 mmHg. This could be followed by beta blockade, alpha blockade or other agents. In this space, flozins have become very attractive, since they do lower blood pressure to an extent (Tran et al Am J HT 2020). In diabetes, the lowering of blood pressure may be anywhere from 5 to 10 mm, though in non-diabetic settings, it’s probably around 3 to 4 mm. On the other hand, finerenone itself has a very small effect on BP  lowering about 2 to 3 mm. Hence spiro has remained a fan favorite when BP lowering is important.
There is a significant overlap between patients who have resistant hypertension, and those who have diabetic kidney disease, and 40-50% of people living with CKD may have resistant hypertension, the proportion increasing as GFR decreases, and albuminuria increases. In someone with blood pressure that is uncontrolled despite three BP, lowering medications, and who have diabetic kidney disease, should you reach for spironolactone, or for a flozin or for finerenone? Maybe both? The individual flozin or finerenone trials cannot answer this question - but the CONFIDENCE trial could. 

How was it done and what did it show?

We have already discussed the trial methodology, so we will not belabor that. Briefly, patients with diabetic kidney disease received either empagliflozin, finerenone, or both for six months. The main outcome in the trial was proteinuria reduction, and safety outcomes, including GFR reduction, and potassium levels. In this analysis, the focus was on BP lowering. The population was optimized on the basis of baseline blood pressure. The subgroup of interest is those who BP was not controlled (SBP > 130) despite three or more BP lowering agents of any class, or any BP when or 4 or more BP lowering drugs. Since we need a 24 hour ABPM to truly classify patients as resistant hypertension which was not done in this trial, this patient population is labelled as ‘ apparent’ treatment resistant hypertension (aTRH). The outcomes of the present analysis were the change in SBP following randomization (days 14, 30, 90, 180, and 210) and the percentage of participants with a ≥10 mm Hg drop in systolic BP at day 180.

From table 1 (not shown here) - the interesting part is the drugs they were on. Though RASi was mandated, only 47% (no aTRH subgroup) to 86% (TRH) were on a CCB. The diuretic was even more stark, with little use in the ‘no aTRH’ group (18%) to 66% in the TRH group. Hence beta-blockers and other worse agents helped clinch the ‘apparent’ part of the TRH definition. For BP response (defined as > 10 mm Hg SBP drop) - the combination of empa + finerenone seemed to have a greater effect - and in the regression (figure 2) it is clear that this effect was more pronounced in the aTRH group (as well as with lower GFR). 

Figure 1 from Agarwal et al JACC Advances 2026

Figure 2 from Agarwal et al JACC Advances 2026

If you then examine the actual BP drop in the aTRH subgroup, it was about 11 mmHg by day 30 and 9 mm Hg at day 180. Empa and finerenone alone were about 7 to 5 mmHg at the same points, in the aTRH subgroup. The effect of empa was greater than the effect of finerenone alone in the no aTRH subgroup (data in Table S1, not shown here). Data broken down by various subgroups (figure 3) suggests, again, that the combination of empa-finerenone is more potent with decreasing GFR and in the aTRH group. 

Figure 3 from Agarwal et al JACC Advances 2026

For safety (table 3, not shown here) all adverse effects were more common in the TRH subgroup - which is a sicker population - but notably there was no more hyperkalemia than the no aTRH subgroup. 

What does it mean? 

The authors wish to suggest that this should correct the prevailing sentiment that spiro is better for hypertension, and that the empa-finerenone combination provides similar BP lowering in proteinuric DKD. However the spiro effect seen in PATHWAY was in all comers with resistant hypertension, and not proteinuric DKD. Would spiro effect be more or less in such a population? Naysayers would point out that PATHWAY was enriched with primary aldosteronism (not formally tested, but in additional analyses the spiro effect did correlate with lower renin values) which may have contributed to its seemingly outsize benefit. Nevertheless, we use empagliflozin (or any flozin) and finerenone not for BP lowering - but for their effect on clinical outcomes. This analyses serves to reassure us that the BP lowering effects that come along are not trivial.

Summaries by Cristina Popa,  Akshaya Jayachandran ,
Brian Rifkin, Milagros Flores, Swapnil Hiremath

For each study, we’ll unpack the findings, limitations, and consider what they could mean for kidney care. Whether you’re joining the meeting or following along online, pull up a chair and join the discussion.

We did this last year (link) with excellent feedback, so we are bringing it back. Swap and Cristina talked about the rationale (mostly FOMO!) and how this was done here. 

Day 2 Studies

• Ravulizumab in IgA (I CAN) Barratt et al, NEJM 2026

• Upacicalcet in Hemodialysis Patients with Secondary Hyperparathyroidism, Kendrik K et al, JASN, 2026

• Pentoxifylline for chronic kidney disease in people with diabetes: a meta-analysis of randomized trials, Agarwal et al, NDT, 2026

Studies we do not discuss:

• Povetacicept in IgA (RAINIER AIS D08) Barratt et al NEJM 2026 (though this was published it will only make it to KidneyWeek presentation on Saturday, so watch out for it in tomorrow’s edition)

• Pentoxiphylline trial baseline data, Leehey et al NDT 2026: the VA PTXRx trial was successfully enrolled with almost 2400 patients. Watch out for the main results to drop in the coming months. 

• A post hoc analysis from CONFIDENCE demonstrating ~ 40% patients did have apparent treatment resistant hypertension, and the combination empagliflozin-finerenone reduced BP by about 9-11 mm Hg (Agarwal et al, JACC Advances 2026): We changed our mind, see day 3 coverage for this study summary.

Ravulizumab in IgA Nephropathy — Interim Analysis of a Phase 3 Trial

J. Barratt, Y.M.K. Farag, K. Rice, T.J. Cammett, L.A. Inker, H. Trimarchi, A. Bruchfeld, H.J.L. Heerspink, A. Liew,K. Carroll, H. Zhang, S. Norouzi, S. Hou,S. Park, M... P.rez Valdivia, J. Radhakrishnan, Y. Suzuki, B. Schneider, J. Marsala, C. Carrillo‑Infante, S.L. Lake, A.K. Singh, S. Nolan, and R.A. Lafayette, for the I CAN Trial Group

DOI: 10.1056/NEJMoa2607544

Why was the study done?

IgA takes the center stage on day 2. So, what haven't we discussed about IgA yet? To open the show, we turn to complement, a key driver of inflammation and kidney injury on IgAN. Ravulizumab binds complement protein C5, preventing formation of the inflammatory mediator 5a and the membrane attack complex C5-9.  Since everyone remembers the complement pathway so well (!) we have a helpful figure which shows how and where ravulizumab might help in IgAN. 

Figure from Barratt et al Kidney Int 2023

Building on promising findings in the phase 2 SANCTUARY trial (JASN 2025), and with molecular support from sustained reductions in urinary biomarkers (Cammett et al CJASN 2026); the phase 3 I CAN  trial (NEJM 2026) asked: can terminal complement blockade reduce proteinuria in patients with persistent disease despite standard care? 

How was the study done?

They conducted a phase 3, multinational, randomized, double-blind, placebo-controlled trial enrolled adults with biopsy-confirmed IgA nephropathy, eGFR >30 ml/min/1.73 m², and persistent proteinuria: UPCR > 0.75 g/g or protein excretion > 1 g/day. All participants required meningococcal vaccination. Patients received intravenous ravulizumab or placebo every eight weeks alongside stable background therapy.

The primary endpoint was changed in 24-hour UPCR at week 34. All participants required meningococcal vaccination. 

Figure S1, study design, from Barratt, NEJM 2026

What did it show?

The interim analysis included 352 patients, 176 per group. At week 34, UPCR decreased by 46.6% with ravulizumab versus 5.6% with placebo (95% CI, -53.2 to -39.0), corresponding to an adjusted UPCR 43.4% lower than placebo (99% CI, 30.1–54.2; P<.001). 

The effect was apparent by week 10. An intention-to-treat sensitivity analysis produced an almost identical result, strengthening confidence that the primary finding was not dependent on the main analysis assumptions.

Urinary complement and injury biomarkers also decreased. In the safety population, adverse events occurred in 69.5% versus 75.8%, and serious adverse events in 4.6% versus 9.0%. Serious infections occurred in four patients per group. No meningococcal infections or deaths were reported.

Figure 1, Changes from Baseline in 24-hour UPCR, from Barratt, NEJM 2026

How should we interpret it?

The double-blind randomized design, prespecified interim analysis, international recruitment, contemporary background therapy, and low discontinuation strengthen the proteinuria finding. An intention-to-treat sensitivity analysis produced consistent results.

However, proteinuria reduction does not yet establish preservation of kidney function. Durability, eGFR effects, and longer-term safety require further follow-up. The absence of meningococcal infections in vaccinated participants does not exclude this known risk of C5 inhibition.

The trial cannot establish comparative effectiveness against other IgAN therapies or the benefits of combining them. Results for patients with eGFR 20–29 were not included, and applicability to children, transplant recipients, and patients with IgA vasculitis remains uncertain. Extensive sponsor involvement in trial conduct, analysis, and reporting also warrants consideration.

However, based on proteinuria, ravulizumab is already approved in the US (brand name Ultomiris), and it is already available in many places for its previous indications (aHUS, PNH, myasthenia, NMOSD). Based on its mechanism of action far distally to BAFF/APRIL, we are not confident that its GFR lowering effect will equal the GFR stabilization effect we see with BAFF/APRIL inhibition. While we are glad there are more options, it is not clear where ravulizumab (like iptacopan) will settle into in our battle against the most common glomerular disease. 

What can we conclude?

Ravulizumab substantially reduced proteinuria in adults with IgA nephropathy receiving supportive therapy, with no major short-term safety imbalance observed. The conclusion remains unproven in long term kidney protection.  The present list price is about an annual $500-600,000.

Efficacy and Safety of Upacicalcet in Hemodialysis Patients with Secondary Hyperparathyroidism

Two Randomized Clinical Trials

• Jessica Kendrick, Markus Kettele, Myles Wolf,Snezana Pesic, Biserka Tirmenstajn-Jankovic, Steven Zeig, Aaron Camp, Ellie Copeland, BT Slingsby, Bastian Dehmel, Geoffrey Block

DOI: 10.1681/ASN.0000001285

Why was the study done?

Calcimimetics are well-trodden ground. Cinacalcet lowered PTH more than 20 years ago (Block et al, NEJM, 2004), then missed its hard endpoint in EVOLVE (ITT HR 0.93, 95% CI 0.85-1.02) (EVOLVE trial investigators, NEJM 2012). Etelcalcetide moved the drug into the dialysis circuit and solved adherence. It did not solve nausea, and it brought more hypocalcemia than cinacalcet (Block et al, JAMA, 2017| Block et al, JAMA, 2017).

Upacicalcet is a small molecule, positive allosteric modulator of the calcium-sensing receptor, given IV after each session. Preclinical data suggest it loses agonist activity when extracellular calcium falls below the physiologic range, and a single session removes 79-100% of it. Japan approved it in 2021 (Hoy, Drugs, 2021) on the strength of a 24- week placebo-controlled phase 3 trial (Shigematsu et al, CJASN, 2023). PATH-1 and PATH-2 tested whether upacicalcet lowers iPTH more than placebo in a predominantly US hemodialysis population with moderate to severe secondary hyperparathyroidism. The protocol restricted changes to calcium supplements and active vitamin D sterols, so that the effect of upacicalcet could be estimated on its own. 

How was the study done and what did it show?

PATH-1 and PATH-2 were identical phase 3, double blind trials run in parallel, each randomizing 2:1 to upacicalceet or placebo at 131 in the US, Bulgaria, Spain, Poland, Portugal, and Serbia (May 2023 to January 2025). Treatment lasted 27 weeks: 19 weeks of titration (25-300 mcg) followed by 8 weeks of stable dosing. The efficacy window was weeks 22-27. Eligible adults had iPTH 400-1500 pg/mL, cCA ≥ 8.3 mg/dL, and dialysate calcium ≥  2.5 mEq/L. Cinacalcet users needed a 4-week washout. Calcium supplements, active vitamin D, and binders could not be changed unless medically necessary. 

Figure 1. Participant disposition in PATH-1 and PATH-2, from Kendrick et al, JASN, 2026

The full analysis set was 768 patients (510 upacicalcet, 258 placebo). Mean age was 59 years, and median iPTH was 649-712 pg/mL. PATH-1 enrolled 82% of its patients in the US, and 38% were Black (20% in PATH-2). About 88% of PATH-1 and 70% of PATH-2 patients were on active vitamin D. 

Patients who discontinued treatment counted as nonresponders, and the analysis was a stratified Cochran-Mantel-Haenszel test. A ≥ 30% fall in mean iPTH occurred in 42.1% vs 16.8% in PATH-1 (difference 25.3 percentage points, 95% CI 16.2-34.5) and 43.5% vs 17.4% in PATH-2 (26.1 points, 95% CI 17.4-34.8; both P<0.001). 

Figure2. Strata-adjusted proportion of participants with ≥ 30% decrease from baseline in mean iPTH during the efficacy assessment period 922-27) in PATH-1 and PATH-2, from Kendrick et al, JASN, 2026

The key secondaries were consistent but modest. A ≥ 50% fall occurred in 27.2% vs 7.5% and 25.1% vs 4.3%, and mean iPTH of 150-300 pg/ml in 18.8% vs 8.3% and 17.3% vs 4.3%. Mean iPTH change was -32.7% vs -1.5% and -35% vs -0.9%. Effects were consistent across subgroups, and FGF23 fell (exploratory endpoint).

Supplemental Table 4. Key secondary iPTH endpoints, weeks 22-27, from Kendrick et al, JASN, 2026

Serious AEs were balanced  (26 vs 24%; 23% vs 22%), and so were nausea (3% vs 2%; 7% vs 5%). Calcium was a different story. By central laboratory, cCa fell below 7.5 mg/dl in 31% vs 12% (PATH-1) and 45% vs 19% (PATH-2). It fell below 8.3 mg/dl in 86% of upacicalcet patients in both trials. Dose interruptions occurred in 44% vs 24% and 52% vs 27%. Investigators recorded symptomatic hypocalcemia in 0.8% and 2% of upacicalcet patients and in no placebo patients. 

What does it mean?

Upacicalcet lowers PTH more than placebo does, same as in the Japanese study (Shigematsu et al, CJASN, 2023). What PATH adds is a less flattering effect size. Fewer than half of the patients reached a 30% reduction, and fewer than one in five reached 150-200 pg/ml. In the Japanese trial, 81% reached a ≥30% reduction (baseline IPTH roughly half as high; calcium carbonate allowed to creep up). With etelcalcetide vs placebo, the figure was 74-75%. 

The authors blame the calcium-restrictive protocol, and they have a point: nearly 30% of patients could not up-titrate because of low calcium. That argument cuts both ways. If upacicalcet backs off on a low calcium, the patients with the worst disease may get the least conditions on a post-randomization variable that itself suppresses PTH. Across the same cCa quartiles, the placebo response climbs from 0% to 34.8%. This gradient therefore likely reflects, at least in part, the direct suppressive effect of serum calcium on PTH secretion rather than an upacicalcet effect hidden by the protocol. Background therapy also did not stay entirely constant: in PATH-2, the proportion of upacicalcet-treated patients using calcium supplements or calcium-containing binders rose during treatment, while it remained stable with placebo.

Supplemental figure 4. Use of calcium supplements or calcium-containing phosphate binders over time in PATH-1 and PATH-2, from Kendrick et al, JASN, 2026

The “asymptomatic hypocalcemia” label is carrying a lot of weight: nearly half of PATH-2 upacicalcet patients dropped below 7.5 mg/dl at least once. KDIGO 2017 tolerates mild, asymptomatic hypocalcemia (KDIGO, Kidney Int Suppl, 2017). A cCa below 7.5 mg/dl is not mild hypocalcemia. Symptoms were also judged by investigators, which makes them a safe safety measure. The authors note that comparing these rates with the roughly 7% symptomatic hypocalcemia seen with etelcalcetide means comparing across trials. 

Beyond that, the trials ran only 27 weeks and measured biochemistry, not fractures, parathyroidectomy, CV events, or death. Between 22% and 36% of patients per arm stopped the treatment. Comparing with placebo in 2026 sidesteps the question prescribers will ask. Cinacalcet is generic, and since 2021 calcimimetics have been paid from the US dialysis base rate (CMS, 2020), so a new IV agent has to beat etelcalcetide on something. The sponsor funded the trials, employed two authors, and paid for medical writing (Note: a “129%” share on the top dose and a twice-listed PATH-2 analysis plan suggested the proofreading was rushed, too).

A calcimimetic that steps back when calcium falls sounds elegant. In PATH, it stepped back often enough that the real question is not whether it is safer than etelcalcitide, but whether, at equal safety, it does enough.

Pentoxifylline for Chronic Kidney Disease in People with Diabetes: A Meta-Analysis of Randomized Trials

Rajiv Agarwal, David J. Leehey, Sankar D. Navaneethan, Kristina M. Utzschneider, Bin Xing, Kimberly Carlson

DOI: 10.1093/ndt/gfag244

Why was the study done?

We are almost at remission (see story 8 from 2025 top stories)  of Diabetic Kidney Disease (DKD) so do we need more agents? Especially with the newer GLP1RA related drugs coming online. However, even with RAS blockade, SGLT2 inhibitors, finerenone and semaglutide, DKD patients do still progress, and cost limits access to these drugs in many places. Also, everyone does not tolerate every drug - so more options are always better. Pentoxifylline is a cheap generic oral methylxanthine with anti-inflammatory and antifibrotic effects, tested in DKD for almost 40 years. We discussed the PREDIAN trial on NephJC back in 2016! Seven earlier meta-analyses reported lower albuminuria, however the authors say none of these previous ones reported prediction intervals, This one asks whether that result is actually reliable, whether enough evidence had accrued to settle the question, or formally looked at hard outcomes. 

How was it done?

It’s a prospectively registered systematic review of RCTs of oral pentoxifylline added to standard care in adults with DKD. Comparators were placebo, no additional treatment, or an ACEi/ARB (kept as a separate stratum). Seven databases plus the WHO trials registry were searched to June 2026, with no language restriction. One reviewer screened and extracted, and a second verified every value. Risk of bias was rated per result with RoB 2, and certainty with GRADE.

Figure 1. PRISMA 2020 flow diagram of study selection. from Agarwal et al, NDT 2026

The co-primary surrogate outcomes were albuminuria (pooled as a ratio of geometric means across ACR, albumin excretion, protein excretion and PCR) and GFR change from baseline (primary analysis limited to trials of 12 months or longer). The primary hard outcome was ESKD (dialysis or transplant) or all-cause death. Safety outcomes were GI adverse events, withdrawal for adverse events, and serious adverse events. They used random-effects models with Hartung-Knapp adjustment, and added two things to every estimate: a 95% prediction interval (where a new trial would be expected to land) and a trial sequential analysis (TSA), which shows how much of the required sample has accrued.

What did it show?

34 studies were included and 32 contributed data. Most were small, single-center, short (3 to 6 months) and often unblinded, usually using 400 mg two or three times daily. Risk of bias was mostly high, driven by selective reporting.

Table 1. Characteristics of the trials contributing to the review from Agarwal et al, NDT 2026

Figure R1. Risk of bias, albuminuria and proteinuria results from Agarwal et al, NDT 2026

Figure R2. Risk of bias, kidney function results from Agarwal et al, NDT 2026

Albuminuria: 22 trials gave a ratio of means of 0.69 (95% CI 0.50 to 0.94), about a 31% reduction. Heterogeneity was extreme (I² 97%) and the prediction interval was 0.15 to 3.10, so a new trial could show an 85% reduction or a threefold increase. It depended on the comparator: against an ACEi/ARB the ratio was 0.94 (0.87 to 1.03), while as add-on therapy it was 0.62 (0.41 to 0.91). Removing four trials with data-quality concerns gave 0.80 (0.61 to 1.07). TSA showed 53% of the required information, with no boundary crossed.

Figure 2. Forest plot of the ratio of means for albuminuria from Agarwal et al, NDT 2026

GFR: Five trials and 375 patients (12 months or longer) gave +4.49 mL/min/1.73 m² (0.84 to 8.14), with a prediction interval of −4.25 to +13.23. Restricting to the two low or some-concerns risk of bias trials gave +2.96 (−15.34 to 21.26). The estimate rose with risk of bias, from +1.52 where concerns were minor to +7.52 where they were serious. TSA showed 11% of the required information.

Figure 3. Forest plot of the mean difference in glomerular filtration rate change from baseline  from Agarwal et al, NDT 2026

Hard outcomes: 29 events in 593 patients (14 on pentoxifylline, 15 on control), with a risk ratio of 0.90 (0.46 to 1.77). That is 3% of the information needed.

Safety: GI adverse events were three times as common (RR 3.00; 1.94 to 4.64). Withdrawal for adverse events trended higher (1.58; 0.96 to 2.60). Serious adverse events were similar (0.83; 0.56 to 1.22). GRADE certainty was low for albuminuria and GFR, very low for ESKD or death, and moderate for GI tolerability.

What does it mean?

Pentoxifylline is a drug in search of an indication. Here, it  probably lowers albuminuria on average, but the estimate is really fragile. The benefit mainly comes from some add-on trials that disagree with each other for reasons nobody could explain, and it shrinks as trial quality improves. The GFR signal is also similar,  with a few small trials, and the biggest effect in the weakest ones. For kidney failure and death there is almost no data, so this is absence of evidence, not evidence of no effect. The GI side effects are the one firm finding. The authors call for an adequately powered, long-term hard-outcome trial, and the VA PTXRx trial is ongoing, with enrolment complete (Leehey et al NDT 2026: for baseline data). It will be interesting to see how it performs on a baseline of highish flozination (~ 50%) as well as a third on GLP1RAs. A few years ago, a derivative (lisofylline) was created with supposedly better tolerability, but that went nowhere. 

A few things to keep in mind while reading this meta-analysis. , firstly, the evidence is quite old: 24 of 32 trials predate 2019, so we can’t tell whether it adds anything on top of SGLT2i, finerenone or GLP-1 RA. It is quite possible that with these agents (particularly flozins) on board, the effect of pentoxifylline may not be as huge - and be difficult to demonstrate. Secondly, mixed measures were used and ACR, albumin and protein excretion and PCR were pooled together. The log-scale values were reconstructed because no trial reported geometric means, and the GFR pool mixed MDRD, creatinine-cystatin C eGFR and creatinine clearance. Thirdly, 5 of the six trials were excluded for data-integrity concerns, identified after inspecting extracted data. Four more with data-quality concerns were kept, and removing them moves albuminuria from 0.69 to 0.80. Also, hard events were incidental, came from flow diagrams and adverse-event tables, were likely undercounted, and weren’t assessed for risk of bias.

Summaries by Akshaya Jayachandran Brian Rifkin,
Cristina Popa,  Milagros Flores, Swapnil Hiremath

For each study, we’ll unpack the findings, limitations, and consider what they could mean for kidney care. Whether you’re joining the meeting or following along online, pull up a chair and join the discussion.

We did this last year (link) with excellent feedback, so we are bringing it back. Swap and Cristina talked about the rationale (mostly FOMO!) and how this was done here. 

Day 1 Studies:

• VISIONARY: Sibeprenlimab in IgAN, Rizk et al, NEJM 2026

• ORIGIN: Atacicept in IgAN, Barratt et al, NEJM 2026

• VIDEO-KD: Video Decision aids in CKD, Volandes et al, JAMA IM 2026

• Pregnancy in kidney donors; Wang et al, JAMA 2026

• Protein restriction in CKD is worse, Suh et al, JAMA Network Open 2026

• Vascular Access outcomes over time, Roetker et al, JASN 2026

A couple of other studies were published which are not getting into:

• Longer term data from APPLAUSE-IgAN (iptacopan) Rovin et al, Kidney Int 2026. After VISIONARY and ORIGIN, these look pretty tame results. They are important nevertheless as access to these drugs varies. Read the studies - it does suggest ongoing benefit. 

• Urinary biomarkers from the SANCTUARY trial (Ravulizumab in IgA) Cammett et al CJASN 2026 - and it seems this molecule was approved by the FDA today.

Double bonanza for IgAN therapeutics today: atacicept and sibeprenlimab both report their final phase 3 results in NEJM.

We’ve spent a lot of time on both of these trials. We covered the proteinuria interims (ORIGIN Interim,  NephJC summary) and (VISIONARY Interim, NephJC summary), and both times we said the same thing: lovely proteinuria data, but we’ll hold the Hallelujah until the eGFR slope arrives. Well, the slopes are here.

Sibeprenlimab: and the kidneys followed

Sibeprenlimab in IgA Nephropathy — Final Phase 3 Trial Analysis

Authors: Dana V. Rizk, M.D., Jonathan Barratt, Ph.D., Yusuke Suzuki, M.D., Ph.D., Hong Zhang, M.D., Ph.D., Muh Geot Wong, M.B., B.S., Ph.D., Adrian Liew, M.B., B.S., M.R.C.P., Richard Lafayette, M.D., Vladimír Tesař, M.D., Ph.D., Hernán Trimarchi, M.D., Ph.D., Kevin Carroll, Ph.D., Ian S.D. Roberts, M.B., Ch.B., Vivekanand Jha, M.D., Alejandra Quevedo, M.D., Seung Hyeok Han, M.D., Ph.D., Rosnawati Yahya, M.B., B.Ch., M.R.C.P., Chee Kay Cheung, M.B., Ch.B., Ph.D., Raja Ramachandran, D.M., M.D., Michael Walsh, Ph.D., Bobby Chacko, M.D., D.M., Manuel Praga, Ph.D., Jin-Bor Chen, M.D., José Suassuna, M.D., Ph.D., Jing Xia, Ph.D., Lokesh Shah, M.D., and Vlado Perkovic, M.B., B.S., Ph.D., for the VISIONARY Trial Investigators Group*

DOI: 10.1056/NEJMoa2609127

Why was the study done?

We all know Sibe! Sibeprenlimab is a monoclonal antibody that neutralizes APRIL selectively. For more on APRIL (and BAFF) see our previous summaries. In the interim analysis, 24-hr UPCR at 9 months was reduced by 51.2% Vs placebo, which supported accelerated approval. When we covered it, our question was whether the kidneys would follow, and whether blocking APRIL alone is enough.

How was it done?

This is the same double-blind VISIONARY design: sibeprenlimab 400 mg subcutaneously every 4 weeks vs placebo for 100 weeks. Like all IGAN trials, these are biopsy proven IGAN, with proteinuria > 1g, eGFR > 30, and excluding sedonary IgAN. Everyone got RASi, flozins were allowed, not mandated. Patients with T2 or C2 on MEST-C, nephrotic syndrome, or heavy fibrosis were excluded. The key secondary endpoint was the annualized eGFR slope over 24 months.

Figure S1: VISIONARY Trial design from Rizk et al, NEJM 2026

What did it show?

The cohort was 510 adults (259 vs 251) with biopsy-proven IgAN, proteinuria ≥0.75 g/g (or ≥1 g/day), and stable maximal RAS blockade, with a mean eGFR of 64 and 45% on SGLT2i.  The slope was +0.3 with sibeprenlimab vs −4.2 with placebo (difference 4.5; 95% CI 3.6 to 5.4), with consistent results across sensitivity analyses and subgroups. The 24-month eGFR change was +1.3 vs −7.9 (difference 9.2; 95% CI 7.1 to 11.4). The composite of sustained 40% eGFR decline, creatinine doubling, eGFR <15, or kidney replacement therapy occurred in 1 patient (0.4%) vs 28 (11.2%) (HR 0.029; 95% CI 0.004 to 0.214), with KRT in 5 placebo patients and none on sibeprenlimab. The antiproteinuric effect held at 24 months, with UPCR down 61.3% vs 1.3%. Proteinuria below 0.5 g/day was reached by 36.7% vs 9.1%, and hematuria resolved in 78.6% vs 20.2%. IgA fell about 75%, IgM 77%, and IgG 37%. The subgroup analysis (not shown here) is pretty impressive for consistent effect. 

Adverse events were similar (90.7% vs 90.0%). Serious adverse events were 8.1% vs 12.7%, and serious infections 1.9% vs 4.0%. Twenty-four patients dropped below an IgG of 400 mg/dl, but none had a serious or severe infection. There were no deaths.

Figure 1: Composite Kidney End Point (Full Analysis Population) from Rizk et al, NEJM 2026

Table 2: Efficacy Outcomes from Rizk et al, NEJM 2026

What does it mean?

The kidneys followed. An APRIL-only strategy can flatten the slope, and the Gd-IgA1 and hematuria result fits a real change in disease activity. A few methodological notes of caution, the striking hazard ratio (0.029) rests on a single event in the treated arm, so the absolute difference (0.4% vs 11.2%) is the better number to quote. They back it up with logistic and other sensitivity analyses that don’t rely on that assumption. The slope was estimated from week 4 onward, and set aside data after rescue therapy but a post hoc analysis from baseline agreed (difference 4.7). It’s another sponsor-run trial (Otsuka), and follow-up ends at 24 months, with the open-label extension ongoing.

Atacicept: the slope we were waiting for…

Atacicept in IgA Nephropathy — Phase 3 Trial Final Efficacy Analysis

Authors: Jonathan Barratt, Ph.D., Robert M. Brenner, M.D., Sean J. Barbour, M.D., Kirk N. Campbell, M.D., Tom Doan, Necmi Eren, M.D., Jürgen Floege, M.D., Marshall W. Fordyce, M.D., Vivekanand Jha,  Zeeshan Khawaja, M.D., Beom Seok Kim, M.D., Ph.D. https://orcid.org/0000-0002-5732-2583, Adrian Liew, M.D., Bart Maes, M.D., Ph.D., Atanu Pal, M.D., D.M., Roberto Pecoits-Filho, M.D., Ph.D., Richard K.S. Phoon, M.D., Dana V. Rizk, M.D., Hitoshi Suzuki, M.D., Ph.D., Vladimír Tesař, M.D., Ph.D., Hernán Trimarchi, M.D., Ph.D., Xuelian Wei, Ph.D., Lixi Yu, Ph.D., Hong Zhang, M.D., Ph.D., and Richard Lafayette, M.D., for the ORIGIN Phase 3 Trial Investigators

DOI: 10.1056/NEJMoa2610103 

Why was the study done?

IgA nephropathy is the most common primary glomerular disease (yes we all know that, but it seems disrespectful to not state this crowdpleaser cliche at least once). Atacicept is a TACI-Fc fusion protein that blocks both BAFF and APRIL. The ORIGIN 3 interim (NephJC summary | Podcast) showed a 45.7% vs 6.8% proteinuria reduction at 36 weeks, which led to FDA accelerated approval in July 2026. Approval on a surrogate endpoint comes with a promise to show the hard eGFR data will follow, so this is that report.

How was it done?

We’ve discussed this many times before - the IgAN trials follow a cookie cutter established design. It is the same double-blind, placebo-controlled design: adults with biopsy-proven IgAN on maximal supportive care, randomized 1:1 to atacicept 150 mg weekly subcutaneously or placebo. Proteinuria > 1g/day and eGFR > 30 with a BP < 150/90 and excluding secondary IgA, nephrotic syndrome or RPGN. Everyone gets background RASi, flozin use is measured, not mandated. Finerenone is not even mentioned (since FIND-CKD [Heerspink et al NEJM 2026 | NephJC summary | podcast] was still ongoing at the time this was designed). 

Figure S1, Trial design, from Barratt J et al, NEJM 2026,

The analysis plan changed in May 2026, after alignment after the sponsor agreed this with the FDA, so the final analysis looks at eGFR at week 52 (instead of week 104) plus the annualized slope through week 104. That lets patients move to open-label drug sooner, again, a sign of how hard a prolonged placebo is becoming to justify in this rapidly changing times. 

What did it show?

Of 431 randomized, 428 were treated (214 per arm), with a median age of 41, about 55% Asian, and an impressive 61% on flozins at baseline. Mean eGFR was 64 ml/min/1.73 m2 and UPCR was about 1.6 g/g.

At week 52 the eGFR change was −0.11 with atacicept vs −5.69 with placebo (difference 5.58; 95% CI 3.68 to 7.48). The annualized slope through week 104 was −0.60 vs −5.60 (difference 5.00; 95% CI 3.55 to 6.45). That atacicept slope sits under the KDIGO 2025 target of losing less than 1 ml/min/yr - and practically implies hardly any loss of GFR at all. The composite of a sustained 30% eGFR decline, eGFR <15, dialysis, transplant, or death occurred in 11 patients (5.1%) vs 38 (17.8%) (HR 0.24; 95% CI 0.12 to 0.48). Nobody on atacicept reached dialysis, transplant or death, compared with 7 patients on placebo. The rest of the hierarchy fell in line too: UPCR down 53% vs 5%, Gd-IgA1 down 66%, and hematuria resolved in 81% vs 21%. Subgroups, including by baseline eGFR and flozin use, pointed the same way, pretty consistently (not shown here).

Figure 1, Mean Change in Estimated GFR, Annualized Estimated GFR Slope, and Composite Kidney-Disease Progression, from Barratt J et al, NEJM 2026,

Adverse events were similar (74.8% vs 78.0%), and serious events were fewer with atacicept (7.0% vs 15.4%). Infections were balanced, and there were no opportunistic infections or clinically relevant hypogammaglobulinemia, despite the expected drop in immunoglobulins. Injection-site reactions were more common (30.4% vs 7.5%). Antidrug antibodies appeared in nearly half of atacicept patients, with no apparent effect on efficacy or safety.

Table 2. Key Secondary Efficacy End Points from Barratt J et al, NEJM 2026

What does it mean?

This is the confirmation we asked for. Atacicept stabilized eGFR on top of contemporary care, and the proteinuria signal was based on actual biology, not just a surrogate. Some caution still applies. The composite was driven entirely by the 30% eGFR decline, and hard outcomes were few. The placebo slope (−5.6) is steep, and 21% of placebo patients discontinued vs 7% on atacicept, though the post hoc ITT analysis still held (slope difference 4.15). The composite was driven entirely by a sustained ≥30% eGFR decline, a lower bar than the 40% decline used in VISIONARY. Only 7 patients (all on placebo) reached dialysis, transplant, or death, so the hard-outcome data are still sparse. Only 50 atacicept and 37 placebo patients contributed at week 104, because of open-label crossover and the June 2026 data cutoff. The slope through 104 is therefore largely extrapolated from earlier follow-up. It’s a sponsor-run trial with no active comparator, and newer IgAN drugs were barred during the blinded period. Two years of follow-up won’t tell us about durability, so we’ll watch the open-label extension. Even with the caveats, the eGFR slope on atacicept is the one we’ve been waiting for, so a restrained “Hallelujah” feels earned. 

So who wins: dual BAFF/APRIL or APRIL alone?

We asked this in the interim blogs, and these papers give the first answer on eGFR. Atacicept’s slope difference was 5.0 and sibeprenlimab’s was 4.5, with hematuria resolution at about 80% for both and similar safety. Honestly, it’s a draw. Cross-trial comparisons are risky: the composite endpoints differ (30% vs 40% decline, plus creatinine doubling or KRT), the placebo slopes differ (−5.6 vs −4.2), and the follow-up timepoints differ. Only a head-to-head trial or longer follow-up will separate them. For now, upstream B-cell and plasma cell modulation holds up as a real disease modifying strategy in IgAN, and that is real good news for us. Whether it is APRIL alone or BAFF/APRIL combined as a mechanism doesn’t seem to matter with these empiric data. 

What about access and cost?

Sibeprenlimab (Voyxact) was approved by the FDA under accelerated approval on November 25, 2025, with an estimated annual net price of $292,500. Atacicept (Trutakna) followed on July 7, 2026, with a list price (wholesale acquisition cost) of $32,700 for a 28-day supply, roughly $425,000 a year. Net and list prices aren’t comparable, so the real gap is smaller than it looks. There are also telitacicept (China) and povetacicept on the way, so this space is likely to get crowded, though unlikely any cheaper. 

Neither drug is cheap, and both are weekly or monthly injectables that need specialist prescribing and payer approval. As NephJC noted last year, most IgAN patients live in Asia, where access to these agents is far more limited than in the US. The data on palliativbe therapies (aka supportive management) has also improved with data on flozins (DAPA-CKD), finerenone (FIND-CKD). Perhaps atrsentan and sparsentan also belong here? There are also other intermediate agents with effect sizews in between, such as iptacopan, and ravulizumab. And let’s not forget good old steroids - and fancier steroids (targeted release budesonide) as well. See table below to think about all of them together.

Telitacicept (TELIGAN): Lv et al, NEJM 2026 | NephJC Shorts
Povetacicept (RAINIER): Barratt et al NEJM 2026 | NephJC day 3
Iptacopan (APPLAUSE): Barratt et al, NEJM 2026 | NephJC Shorts
Ravulizumab (I CAN): Barratt et al, NEJM 2026 | NephJC Day 3
Sparsentan (PROTECT): Rovin et al Lancet 2023 | NephJC Summary
Atrasentan (ALIGN): Heerspink et al Lancet 2026
Finerenone (FIND CKD): Neuen et al JAMA 2026 | NephJC summary
Nefecon (NEFIGAN): Lafayette et al Lancet 2023 | NephJC summary
Steroids (TESTING): Lv et al JAMA 2022 | NephJC Summary
Flozins (DAPA CKD): Wheeler et al Kidney Int 2021 | NephJC Summary

• It may seem that the BAFF/APRIL agents really seem to slow things down - but before we use words like remission - note that the treatment was never stopped for the period of study. They keep the disease under control - what will happen if the drugs are stopped? (Or is the expectation is that we pay half a million for the rest of our lives - the aHUS model infamously set up by expensivumab eculizumab?)

• In NEFIGAN and TESTING, treatment was stopped at ~ 9 months (and the GFR slope in NEFIGAN at 9 months is a flat line too). Would they be just as effective as BAFF/APRIL if the drug were continued?

• Cost of course is the other consideration. Underlying differences in population, flozin use also matter making inter-trial comparisons difficult - however, the ERAs aka -santans don’t look as impressive - statistically significant as the effect sizes are. However access is the key. Most patients with IgAN are in Asia, and the cost will be a big deal. Often patients from these geographical areas constituted the majority of patient populations in these RCTs - and once the observational and extended periods of follow up are done - are we going to say, thanks for your service, no more of these fancy expensive drugs for you?

JAMA | Original Investigation 

Video Decision Support and Advance Care Planning in Older Adults With Advanced Kidney Disease

The VIDEO-KD Randomized Clinical Trial

Angelo E. Volandes; Joshua R. Lakin; Yuchiao Chang; Charlotta Lindvall, MD; Lisa M. Quintiliani; Andrew L. Lundquist; Annie Liu, MD; Manjula Kurella Tamura; Mark Unruh, MD; Susan P.Y. Wong; Aretha Delight Davis,; Cynthia Garde; Lori Henault,; Upeka Samarakoon; Amresh Hanchate; Ernest Mandel; Jamie A. Green; Kyla Lawson; Seth N. Randa; Michael K. Paasche-Orlow.

Published Online: October 22, 2026 doi: 10.1001/jamainternmed.2026.5989

Why was the study done?

VIDEO-KD was done to address a gap of older adults facing kidney failure: dialysis vs conservative kidney management. The authors describe a practical problem on advanced CKD care: conservative management is not consistently discussed, goals and preferences often remain undocumented, and dialysis can become the default during an acute illness rather than an informed choice.

So far, previous kidney decision-support trials have shown promising but varied results. Education improved knowledge, but video was not significantly better than the verbal script, and treatment preferences did not differ between groups (Eneanya et al, Am J Nephrol 2020). Other interventions improved decision quality (Landin et al, Ann Intern Med 2023) and advance care planning (ACP) engagement and increased advance directive documentation , Lupu et al, AJKD 2022). Thus, earlier trials had demonstrated improvements in decision quality and formal ACP documentation, while VIDEO-KD extends this evidence to documentation of kidney-treatment discussions.

The intervention therefore targeted two steps: understanding the options, through a video decision aid, and communicating preferences, through a patient-created video, audio, or written declaration shared with the clinician. 

How was the study done, and what did it show?

This was a randomized trial at 12 outpatient clinics including 453 adults > 65 years or older with advanced CKD (eGFR ≤30 mL/min/1.73 m²) who were not yet receiving dialysis. 

Intervention participants combined a 12-minute video explaining HD, PD, and conservative kidney management. They then created a video, audio, or written statement describing their values and treatment preferences, which was sent to their treating clinician. Controls received usual care. 

The actual video is not available (link describes it, but requires fee to access). The description from the methods paper is:
’The video begins with a physician introducing the viewer to the concept of ACP as well as a review of advanced CKD and kidney failure. The narrator explores each of the options for kidney failure, reviewing the risks and benefits of each, and then discussing the trade-offs among the three options. For each option, visual images illustrate the therapy while discussing risks and benefits. The visual images illustrating the first option, haemodialysis, include a patient receiving in-centre haemodialysis, nurses attending to the person on haemodialysis, and images of caregivers. The second option, peritoneal dialysis, includes visual images of a person at home on peritoneal dialysis with the assistance of caregivers, and the daily activities around peritoneal dialysis equipment care. The third option, medical management without dialysis, is introduced by the narrator as a potential option for those persons who wish to not pursue dialysis or any of its associated burdens and who would prefer to focus on their quality of life. The narrator explains that medical management without dialysis focuses on clinicians, patients and caregivers working together to treat symptoms through medical management, and other core principles of palliative care. The narrator then begins to describe CPR and the option of whether to receive CPR or not. The images include CPR on a mannequin and the likelihood of success for older patients with CKD. Visual images include physicians, nurses, patients and caregivers in clinic, at home, and in the hospital.’

The primary outcome was ACP documentation within 12 months, or before death, whichever came first. Outcome reviewers were masked to treatment assignment.

Figure1, flowchart of the study population, from Volandes, JAMA, 2026

ACP documentation was higher in the intervention: 56.6% versus 40.0%, an absolute increase of 16.6 percentage points (95% CI, 7.5-25.7; p<0.001). Documented discussions of kidney-treatment options also increased: 46.1% versus 31.6%.

Immediately after the video, the proportion of intervention participants who were undecided fell from 39.6% before the video to 13.2%. Conservative management was preferred by 50.7%, compared with 31.8% of controls at enrollment. These were stated preferences, not treatment subsequently received. 

Table 2, primary and secondary outcomes at 12 months, from Volandes, JAMA, 2026

How should we interpret it?

The design: randomization, multicenter recruitment and masked outcome, strengthen the primary finding. However, the trial tested the whole package: education, plus a personal statement shared with the clinician, so the contribution of each cannot be separated. 

The preference findings deserve more caution: intervention patients were reassessed immediately after the video, whereas controls were assessed only at enrollment. The study therefore did not establish whether preferences remained stable over time. It also did not measure whether clinicians viewed the declarations, whether documentation was comprehensive, or whether subsequent care matched patients’ wishes.

The practical takeaway: this approach improved the recording of treatment conversations and helped patients articulate an immediate preference. Whether it leads to better real-world treatment decisions remains unanswered.

Renal reserve, or reserved judgement? Preeclampsia after kidney donation

Pregnancy Outcomes After Living Kidney Donation

Carol Wang; Kyla L. Naylor; Eric McArthur; James A. King; Jessica M. Sontrop; Joseph Beyene; Dervla Connaughton; Elizabeth Hendren; Michelle Hladunewich; Ngan N. Lam; Krista L. Lentine; Rhiannon Lyons; Sarah D. McDonald; Pavel S. Roshanov; Erik Youngson; Amit X. Garg.

JAMA Published Online: October 22, 2026
doi: 10.1001/jama.2026.21737

Why was the study done?

About two-thirds of living kidney donors are female, and most are of childbearing age. That makes "Can I still have a baby?” a standard question at donor evaluation. The 2004 Amsterdam forum said yes, with no expected penalty (Delmonico F, Transplantation, 2005). Three later studies disagreed. One compared pregnancies before and after donation in Minnesota (Ibrahim NH et al, Am J Transplant, 2009). One used Norwegian registry data (Reisaeter AV et al, Am J Transplant, 2009). The third was an Ontario-matched cohort of 131 donor pregnancies (Garg AX, NEJM, 2015). All three suggested more gestational hypertension or preeclampsia after donation.
KDIGO’s 2017 donor guideline responded by recommending that candidates be counseled about the risk (Lentine KL et al, Transplantation, 2017). Systematic reviews still graded the evidence as low to moderate certainty, because the studies had few postdonation pregnancies and wide CIs (Pippias M et al, Am J Transplant, 2022; Bellos and Pergialitois, Eur J Obstet Gynecol Reprod Biol, 2022).
The biology is plausible. Pregnancy requires a large rise in renal blood flow and GFR and a solitary kidney has already used much of its reserve. In uninephrectomized mice, pregnancy produced impaired renal adaptation, placental hypoperfusion, and an angiogenic state (Dupont V et al, JCI, 2022).
The question was this: in a larger cohort drawn from three Canadian provinces, after donation, more often complicated by hypertensive disorders than pregnancies in matched healthy donors?

How was the study done, and what did it show?

This was a matched cohort study using link administrative data from Ontario, British Columbia, and Alberta, with a protocol published in advance (Wang C et al, CJKHD, 2021). Donors were females who donated between July 1992 and August 2021 at 10 centers and later had a singleton pregnancy of at least 20 weeks. Each donor pregnancy was matched 1:4 to nondonor pregnancies on delivery date, maternal age, time from cohort entry to delivery, parity, and neighborhood income quintile. The authors excluded anyone with any of 35 conditions that would preclude donation, required recent primary care contact, and assigned each non-donor a simulated nephrectomy date. Estimates were calculated within each province and pooled by fixed-effect meta-analysis. Follow-up ran to March 2023, and the unit of analysis was the pregnancy, not the person.
The final sample was 385 pregnancies in 353 donors and 1540 pregnancies in 1535 non-donors. Mean age at delivery was 32.7 vs 32.6 years, and about a third of mothers were 35 or older. Delivery came a mean of 5.6 years after donation. Roughly a third of the donor pregnancies were already in the 2015 NEJM report (Garg AX, NEJM, 2015). 

Figure- flowchart of inclusion and exclusion criteria for selecting donors and non-donors, from Wang C, JAMA, 2026

Gestational hypertension (RR 1.51, 95% CI 1.10 to 2.08) was again significantly higher in pregnancies after kidney donation. So was preclampsia or eclampsia (RR 1.52, 95% CI 1.10 to 2.29). More striking was  small for gestational age (SGA, below 1oth percentile) which stood out: 13.8% vs 7.9% (RR 1.86; 1.37-2.52, risk difference 5.4 points). The post hoc cutoffs pointed the same way, with RR 1.84 below the 5th percentile and 2.47 below the 3rd. Early onset disease before 34 weeks was rare (1.6% vs 1.3%). Preterm birth, low birth weight, NICU admission, and cesarean delivery did not differ significantly, severe maternal morbidity was 4.4% vs 6.5%, and there were no stillbirths in either group. 

Table 3. Maternal and fetal outcomes of kidney donor and nondonor pregnancies, from Wang C, JAMA, 2026

The primary result held when pregnancies from the 2015  study (Ontario deliveries before April 2013) were excluded (RR 1.58; 1.05-2.36). It was similar in nulliparous and parous pregnancies. It was larger at maternal age 35 or older (RR 1.99, 1.18-3.33) than under 35 (RR 1.41; 0.93-2.15). The E-value for the lower bound of the primary CI was 1.40 (VandeWeele and Ding, Ann Intern Med, 2017).

What does it mean?
Hypertensive disorders of pregnancy appear modestly more common after donation. The crude difference is 3.6 percentage points, about one extra hospitalized case per 28 donor pregnancies. The modelled difference is 2.7 points, about one per 37, and its CI includes zero. Most donor pregnancies (88%) were unaffected, and early-onset disease was not increased.

The data cannot tell us why. An RR of 1.51 on a code-based composite, with an E value of only 1.40, supports counseling but does not prove the nephrectomy causes preeclampsia. Start with what’s missing: there is no blood pressure, BMI, aspirin use, or postdonation eGFR. The “renal reserve” mechanism is therefore never measured. The outcome itself is blunt, putting gestational hypertension and severe preeclampsia in the same bin. 

Ascertainment bias is possible. Donors have slightly more surveillance than non-donors in Ontario, with a median of 4 vs 3 ultrasounds, 11 vs 10 prenatal visits, and 12 vs 11 urinalyses. The difference is small, but it runs in the direction that would inflate a coded diagnosis. Unmeasured confounding is plausible. In Ontario, 64% of donors were first-degree relatives of their recipients. A shared hereditary tendency to vascular or kidney disease could easily produce an association of 1.4, and the relatedness analysis was too small to settle the question. Multiplicity is another concern. With 12 secondary and 8 post hoc outcomes, the SGA signal could be chance, as the authors acknowledge, although its consistency across percentile cutoffs is hard to dismiss. 

Finally, generalizability is limited. Race and ancestry were not available, most donors were White, and care was delivered in publicly funded systems. Donors of African ancestry, including APOL1 high-risk carriers, remained unstudied (Hsu et al, JAMA Internal Med, 2026).

Fou counseling, this is now the most precise estimate available, and it moves the field from “maybe” to “modest but real”. Whether donors should be managed as a high-risk group, for example with aspirin prophylaxis and closer blood pressure follow-up, is a question this dataset cannot answer. Has nephrectomy dented the placenta’s supply line, or are we seeing the family history that the donor shares with the recipient?

Nevertheless - while we do not want to dissuade living kidney donation, these are potential future risks (and notably absolute risks remain small despite the significant relative risks) that should be discussed with potential donors, particularly if pregnancy is in their future. There was a similar notable finding in male kidney donors (Garg et al, Annals IM 2025; a RR of about 39 for hydrocele and scrotal surgery on the side of the laparoscopic kidney donation) which is again worth discussing with male donors as well.

Low Daily Protein Intake and Risk of Kidney Failure and Mortality in Chronic Kidney Disease

Sang Heon Suh, MD, PhD1,2; Hong Sang Choi, MD, PhD1,2; Chang Seong Kim, MD, PhD1,2

Eun Hui Bae, MD, PhD1,2; Jayoun Kim, PhD3; Ji Yong Jung, MD, PhD4,5; Jong Cheol Jeong, MD, PhD6; Kyu-Beck Lee, MD, PhD7; Kook-Hwan Oh, MD, PhD8,9; Seong Kwon Ma, MD, PhD1,2; Soo Wan Kim, MD, PhD1,2; for the Korean Cohort Study for Outcomes in Patients With Chronic Kidney Disease (KNOW-CKD) Investigators

JAMA Netw Open Published Online: October 22, 2026 2026;9;(10):e2634568. doi:10.1001/jamanetworkopen.2026.34568

Why was the study done? 

Low protein diet was the OG intervention for CKD, to slow down the inexorable decline….predating RAS inhibition. And an ancient intervention like that should remain there, in the cobwebs of nephrology history. After all, the large, randomized trials were emphatically negative (eg MDRD, Klahr et al NEJM). However, the KDOQI guidelines gave low protein diets a grade 1A recommendation - and KDIGO still has it in there, though significantly downgraded as a practice recommendation in select patients - and the main recommendation (2C) is for a higher protein intake of 0.8-1 g/kg/day. But this topic - though entirely non-controversial, remains something that is practiced in parts of the world (coupled with the $$$ ketoanalogues - despite this negative trial: Bellizzi et al, Am J Clin Nutr 2022) based on biased and poorly conducted and analysed observational studies. On NephJC we have discussed this issue tangentially (NephJC short on cost analysis; NephJC summary on protein supplements). The major issue here of course is that a low protein diet will lower the serum creatinine - by causing muscle wasting and malnutrition. This is how the MDRD formula was derived in the first place from the MDRD trial - all patients underwent nuclear GFR estimation and not serum creatinine based measurements. Practitioners still prescribing and using low protein diets are thus either ignorant, illinformed, or charlatans (or some combination). Nevertheless - changing diet is always a popular method which patients invariably ask us about diet as an intervention (see this entertaining blog about one such story from the Last Month team). Possibly this is still prevalent in Korea - hence the relevance of this study. 

How was the study done, and what did it show?

This was an observational study using the KNOW-CKD cohort, which included patients ranging from CKD stage 1 (!) to advanced CKD stage 5 not on dialysis, from 2011 to 2016. The study observation ended 2024. In this analysis, they used 5 groups of dietary protein intake (DPI) - calculated from everyone having 24 hour urines and nitrogen measurement - at baseline, and not with poor quality food diaries or food frequency questionnaires, or unreliable diet recalls. The 5 groups were DPI of less than 0.6, 0.6 to less than 0.8, 0.8 to less than 1.0, 1.0 to less than 1.2, or at least 1.2 g/kg/day. Over the period of follow up, they noted the occurrence of kidney failure (aka dialysis or kidney transplant). All cause mortality and MACE were secondary outcomes, ascertained from medical records. Subgroup and sensitivity analysis including a propensity score matched analysis was done - more on that with results. 

Overall, from 2238 participants in KNOW-CKD, 1861 were included (figure 1) with a median follow up of 9.3 years. Table 1 below shows the baseline characteristics - and several features jump out. An impressive 80+ % were on RASi. Low protein intake was more common in females, with higher Charlson score (aka more comorbidities), diabetes, worse CKD status, and higher SBP. Though median ACR was lower in this group, the proportion with ACR > 300 mg/g was similar. Thus overall slightly higher risk of CKD even at baseline in the low protein diet group.

Table 1 from Suh et al, JAMA NO, 2026

After a median of 9.3 years follow up, 345 ended up with kidney failure, and there was a gradient of higher risk with lower DPI (see table 2 and figure 2). This was after adjustment for the covariates mentioned in table 1 above. All cause mortality (but not MACE) were also higher in the low DPI group. 

Figure 2 from Suh et al, JAMA NO, 2026

Table 2 from Suh et al, JAMA NO, 2026

The results were robust in all analysis, including restriction to those with normal serum albumin, diabetes status, sx, and after propensity score matching.

What does it mean?

Low protein diets should be dead, but like zombies they remain practiced in many places. This is another stake in their body (but not a decapitation which is what zombies deserve). The main limitation is that table 1 gives us a lot of concern. The association of worse CKD status with low protein intake might mean someone may have recognized them as higher risk and advised low protein intake. That is a form of selection bias which we cannot adjust away. The higher mortality is a slightly more robust signal - and the authors talk a low about protein energy wasting in the discussion as a cause despite not showing any data to support that in their study. Malnutrition does occur - famously in the long term outcomes of MDRD too higher mortality was noted, presumably from similar mechanisms. We have written about these topics as well in the past for unconvinced readers (Bawazir et al, J Bras Nefrol 2025 and Obeid et al Kidney360 2022). 

Trends in Mortality of Patients with Central Venous Catheters versus Arteriovenous Accesses for Hemodialysis

Nicholas S. Roetker,  James B. Wetmore, Jiannong Liu, David T. Gilbertson, Kirsten L. Johansen

Journal of the American Society of Nephrology Publish Ahead of Print, October 22, 2026. | DOI: 10.1681/ASN.0000001293 

Why was the study done?

Another comparative vascular access study! The fistula first zealots believe in every hemodialysis patient getting an arteriovenous fistula aka AVF (or AV Graft) and the infidel non-believers do not (and even dare to attempt doing a clinical trial comparing the two; Quinn et al, JASN 2026). A central venous catheter (CVC) aka the line of death - are associated with higher infection rates, and many observational studies suggest a significantly higher mortality rate (eg Xue et al AJKD 2003). OTOH, the heretic Canadians point out that such comparisons are inherently biased, as sicker patients, those who are not good AVF surgical candidates, and those whose AVFs fail get into the CVC control group which represents a huge selection bias. See figure below from Quinn and Ravani, NDT 2014. To belabor this point, in this prospective observational study, they demonstrated that even an attempt to create a fistula predialysis (regardless if that attempt was successful)l was associated with a lower mortality - however only 2.3% of the deaths were access related. Hence the vast majority of this ‘excess’ mortality probably captures underlying higher risk features not related to CVC or AVF interventions.  

Figure from Quinn and Ravani, NDT 2014. 

Since the pilot AVF vs CVC trial (Quinn et al, JASN 2026) was unable to recruit successfully, this observational trial takes a similar approach in a much larger number of patients. 

How was the study done, and what did it show?

This study was done with USRDS data - and the figure below shows the methods in more detail. The patient population was incident kidney failure from 2013 to 2022. Notable aspects are:

• Patients with an AVF (or maturing AV access) at dialysis initiation were excluded, which removes some of the immortal time bias that would be accrued by these patients. 

• a clone-censor-weighting approachwas used to evaluate two treatment patterns

• beginning at hemodialysis initiation: continuous CVC use versus successful transition to an AV access

• inverse-probability-of-censoring weighting was then used to account for selection bias from artificial censoring and maintain the covariate balance over follow-up. This approach 

Figure 2 from Roetker et al, JASN 2026

The base cohort was followed until the earliest occurrence of death, treatment modality change, a 3-month EQRS data gap, loss to follow-up, or 18 months of follow-up.The primary outcome was all-cause death, and the secondary outcome was bloodstream infection (BSI) related death.Specifically, the authors wanted to measure the difference between CVCs and AVF, but also evaluate the change over time. 

Figure 1 from Roetker et al, JASN 2026

The overall study included ~ half a million dialysis patients, as shown in figure 1. This excludes about 350,000 with functioning or mature AV accesses. The median age was 63 years, with 47% non-Hispanic Whites, representing the usual mix of the US dialysis population. At hemodialysis initiation, 46% of patients had received no prior nephrology care (reflecting this select group with CVCs). 

Table 2 from Roetker et al, JASN 2026

Among patients initiating HD with a CVC in 2013 and remaining on a CVC, the 18-month cumulative risk of death was 26.8%, ~ 5.9% absolute risk higher than in those with AVF. By 2022, this absolute risk had decreased to 4.2%, as seen in table 2. The BSI-related eaths were a much smaller component of this, at 2.19 versus 1.49, and demonstrate a similar shrinkage of differences over time (from 0.72 to 0.37) over time as shown in figure 3. 

Table 3  from Roetker et al, JASN 2026

Figure 3 from Roetker et al, JASN 2026

What does it mean?

Things are getting better - which is a hopeful note regardless of the CVC/AVF fights. Notable this is despite including the 2020-22 pandemic years in the analysis. A lot has been done to prevent blood stream infections (with CDC-led initiatives and tracking and measurement) along with the underlying secular improvement over time.  Secondly, the difference in outcomes between AVF and CVC mortality is clearly decreasing, especially more marked in the absolute scale. Lastly, though the same also applies not just to all-cause mortality but to blood stream infection related mortality - the absolute numbers here are about a tenth of the all-cause mortality. Unless one believes that CVC kills dialysis patients by another mechanism such as inflammation (and we ignore the Zeus non-effect)  - this just demonstrates the sheer magnitude of underlying bias that cannot be adjusted away despite the sophisticated methodology. Patients who get AVF surgery are inherently different. Databases may also under-ascertain underlying comorbid conditions as the authors point out, which limits the ability to adjust for these confounders. 

Overall this is a good news story, with greater methodological strength - albeit not perfection -  than the average observational study in this area. Though another AVF vs CVC trial may drop soon, (the OASIS trial of ~ 150 patients from Netherlands was presented today showing no difference between randomization to CVC vs AVF, final publication awaited) - this is high quality data that we have now on the relative outcomes and change over time. 

Summaries by Cristina Popa,  Akshaya Jayachandran ,
Brian Rifkin, Milagros Flores, Swapnil Hiremath

Kidney Week 2026 Run Club

Rise, Shine, and Run with the Bear

We are continuing a new tradition of building community through movement at Kidney Week, and we want you out there with us! Join us on Friday, October 23rd at 6:30 AM for a casual, friendly 3-mile run. We are meeting right in front of the giant blue bear in front of the Denver Convention Center. 

Yes, it is early, and yes, it might be chilly, but we promise the camaraderie—and the mutual survival of an extra early alarm—will be entirely worth it. Lace up your sneakers, bring your warm layers, and let’s move together!

If Friday doesn’t work for you, run with Joel (@kidneyboy) on Saturday. This one will be at 6 AM, starting from the Hyatt Regency Denver

Friday October 23rd, 6:30 AM. Meet at the giant blue bear in front of the convention center

NephJC KidneyWk NephRun

Saturday October 24th, 6 AM, Meet at the Hyatt Regency

If you can’t make Friday, or one run is just not enough for you, join NephJC for the second annual ASN NephRun. We have a 5k loop mapped out that starts and finishes at the Hyatt Regency.

Read more to discover how these findings can inform transplant discussions.

Check out VA by Daniel Ramirez

Lee más para descubrir cómo estos hallazgos pueden orientar las conversaciones sobre el trasplante en edades avanzadas. Revisa la versión en español por por Daniel Ramirez

#NephJC Ten POsts Chat

Tuesday, October 13 2026, 9 pm Eastern

Comparative Study BMJ. 2026 Sep 18:394:e100624. doi: 10.1136/bmj-2026-100624.

Survival benefit of deceased donor kidney transplantation versus continued dialysis: international target trial emulation

Rachel Hellemans, Nicholas C Chesnaye, Anneke Kramer, Edouard L Fu, Miha Arnol, Anders Åsberg, Claus Bistrup, Pietro Manuel Ferraro, Kristian Heldal, Jaakko Helve, Camille Legeai, James Medcalf, Dorothea Nitsch, Laia Oliveras, María José Pérez-Sáez, Belén Ponte, Emilie Savoye, Mårten Segelmark, Shaminie Shanmugaranjan, Søren S Sørensen, Roser Torra, Alberto Ortiz, Kitty J Jager, Vianda S Stel 

PMID: 42759976

DOI: 10.1136/bmj-2026-100624

Introduction

Consider a conversation at a transplant clinic. A 70-year-old patient with diabetes who is receiving dialysis has been offered a kidney from a deceased older donor. After hearing about the operation, medications, and follow-up care, the patient asks, “Will this kidney help me live longer?”

There is a lot riding on that little word “this”. The patient wants to understand the likely benefit of kidney transplantation, given their age and medical history. Individualized medical recommendations are the goal, but they are often difficult in the context of evidence based medicine, where study populations often don’t reflect the patient in front of you.

The evidence supporting transplantation, in most patients with ESRD, is substantial. A 2022 systematic review by Chaudhry et al. (BMJ 2022), with 48 observational studies (but no RCTs) and 1,245,850 patients found a long-term survival advantage associated with transplantation in 44 studies (92%). Across the 18 studies included in the meta-analysis, the pooled hazard ratio for mortality was 0.45 (95% CI 0.39–0.54). However, results varied markedly between studies (I²=95.3%), and 11 studies identified subgroups without a statistically significant benefit, implying that there could be “some subgroups that may lack a survival benefit.” 

The transplant population has also been getting older. In an ERA Registry study of 99,543 first kidney transplant recipients across 12 European countries, the proportion aged 65 or older increased from 18% in 2010 to 28% in 2019. That analysis included both living and deceased donor transplantation (Boenink et al, NDT 2023). The donor population was changing as well: among 24,177 deceased donor transplants in seven European countries or regions, the median donor age increased from 50 to 55 years between 2005 and 2015, while at the same time the proportion involving donation after circulatory death increased from 19% to 37%. These trends make the combination of recipient characteristics and donor characteristics increasingly relevant to the survival discussion (Pippias et al., NDT 2021).

Dialysis outcomes have also improved in recent years. In DOPPS, international data covering 1999–2015, adjusted modelled survival time increased by 13% per decade in four European countries, 12% in Japan, and 47% in the United States. These estimates suggest that hemodialysis patients were living longer than those treated in earlier eras, reinforcing the need to revisit transplant-versus-dialysis comparisons based on older data (McCullough et al., AJKD 2025). However, these advantages don’t necessarily extend to the extreme elderly. Another recent study showed that in patients 75 years and older, on average, only lived an additional 77 days versus those who chose conservative (non-dialysis) management of advanced kidney disease (Chou et al, Kidney Med, 2022). 

There remains uncertainty about the size and timing of the benefit for particular recipient-donor combinations. How does the comparison change with advancing age, diabetes, cardiovascular disease, donor quality, and donation after brain versus circulatory death? Answering these questions needs thoughtful study design. Patients have to survive long enough to receive a transplant; an analysis that mistakenly credits their earlier dialysis survival to transplantation can introduce immortal time bias. Target trial emulation helps address this problem by first specifying the hypothetical randomised trial and then aligning eligibility, treatment assignment, and follow-up in the observational data. In this way, it applies the logic of a clinical trial to registry research, while still depending on how completely the available data captures clinically important differences between patients (Hernán and Robins, American Journal of Epidemiology, 2016; Fu et al., BMJ 2026).

In a sea of data supporting the survival benefits of kidney transplantation, clinicians still need clearer directions when navigating decisions for higher-risk patients. Hellemans and colleagues, using international registry data, set out to address that uncertainty by examining recipient age, comorbidities, and donor characteristics together. Their question was specific: for a patient with a specific risk profile, receiving a specific type of donor kidney, how much survival benefit could transplantation offer compared with continuing dialysis?

The Study

Methods

Data collection

The study used donor and transplant-related data from the European Renal Association (ERA) Registry, comprising five national registries: Catalonia (Spain), Denmark, France, Norway, and England/Wales/Northern Ireland (UK).

Recipient variables were: age, sex, cause of kidney failure, dates of waitlisting, dialysis initiation, kidney transplantation, and date and cause of death. 

Diabetes was defined as diabetes either as a primary renal disease or as a comorbidity. Cardiovascular disease was defined as a history of ischaemic heart disease, peripheral vascular disease, or cerebrovascular disease. Both were recorded at waitlisting only. Donors were classified as expanded criteria donors (ECD) if aged ≥60 years, or 50–59 years with at least two of the following: arterial hypertension, cerebrovascular cause of death, or pre-retrieval serum creatinine >1.5 mg/dL before retrieval. All others were standard criteria donors (SCD). 

The retrieval type was donation after brain death (DBD) or donation after circulatory death (DCD). Variables required to calculate the Kidney Donor Risk Index (KDRI) were collected. KDRI refers to the relative risk of kidney graft failure for a specific donor compared to the median (average) donor. Calculation includes age, height, weight, hypertension and diabetes history. Denmark and Norway were excluded from all DCD analyses because DCD transplants were absent or rare.

Target trial specification and emulation

Target trial emulation applies the design of a randomized trial to observational data. The investigators first write the protocol of the trial they would ideally conduct, then reproduce each element of it with registry data (Hernán and Robins, Am J Epidemiol, 2016; Fu et al, BMJ, 2026). The ideal trial here would randomize waitlisted patients on dialysis. Each time a kidney became available the authors would then compare survival depending upon whether the patient proceeded to transplantation or continued dialysis. The table below sets the target trial against its emulation; each row in which the two differ is a potential source of bias.

The question. Among patients who have spent the same time on waitlists while on dialysis, does receiving a deceased donor kidney this month, rather than remaining on dialysis, change survival? 

Step 1- entry. Patients enter the dataset when they are both waitlisted and receiving dialysis. Time is counted in months from that point. 

Step 2- one trial per month. For each of the 60 months after waitlisting, a new trial is started among patients still eligible: alive, waitlisted, on dialysis, and not yet transplanted. Patients were transplanted during that month from the transplant arm; all other eligible patients form the continued dialysis arm. A patient re-enters every monthly trial until transplantation, death, or censoring. A patient transplanted in month 11 therefore contributed ten observations to the dialysis arm (months 1-10) and one to the transplant arm. The 60 trials were then pooled for analysis; the 64,036 patients generated 1,491,537 person-trials.

Step 3- a common time zero. In each trial, eligibility is assessed, treatment is assigned, and follow-up starts at the same moment. This prevents three time-related biases that affect conventional comparison of transplanted and non-transplanted patients and gives a fourth property:

• No immortality bias. If follow-up started at waitlisting and patients were classified by eventual transplantation, the wait would be counted as a transplant follow-up, although a patient cannot die during it and still be counted as a recipient. With a median wait of 441 days (IQR 181-853), recipients are not credited with the survival they accrued while waiting. 

• No lead time bias. If the arms started follow-up at different points in the disease course, survival would differ for that reason alone. In the emulation, both arms start at the same point after waitlisting.

• No depletion-of-susceptible bias. Recipients are, by definition, survivors of the wait, whereas the frailest candidates die early on dialysis. In the emulation, only patients alive at month X are compared, so both arms have survived equally long.

• Exact matching on waiting time. Patients in the two arms have identical time on the waiting list. 

The design does not address unmeasured confounding due to the fact that healthier candidates remain more likely to be transplanted. 

Step 4- emulating randomization. Assignment was not random. Receipt of a kidney depends on allocation rules, blood group, sensitization, and clinical judgment. Treatment was therefore taken as observed (“pseudo-randomization”), and differences in measured baseline characteristics between arms were removed by weighting. Differences in unmeasured characteristics, such as frailty, are not removed.

Step 5- one emulation per recipient-donor combination. The whole procedure was repeated for each of four donor categories (SCD/DBD, SCD/DCD, ECD/DBD, ECD/DCD) in each of four patient groups (all patients; diabetes; CVD; neither), giving (four by four) 16 separate target trials. Each compares recipients of one donor category with dialysis patients from the same patient group. Because each trial has its own comparator, results cannot be compared directly between donor categories. Further emulation stratified donors by KDRI and by donor age.

 Reporting. The study followed the Transparent Reporting of Observational Studies Emulating a Target Trial (TARGET) guideline.

Study Population

Inclusion Criteria:

• Adults ≥20 years 

• Receiving dialysis and waitlisted for a first deceased donor kidney-only transplant between 1 January 2000 and 31 December 2019.

Exclusion Criteria: 

• Preemptive kidney transplant

• On the waiting list for more than 10 years without receiving a kidney transplant

• Started dialysis more than 10 years before waitlisting

• Permanently delisted

• Missing diabetes data

• For the analyses stratified by cardiovascular disease status, patients with missing data on cardiovascular disease. 

Treatment strategies and treatment assignment

Transplantation was a point treatment strategy, a one-time intervention, with the transplant received at the start of that monthly trial. Continued dialysis was a sustained treatment strategy, requiring the patients to be on dialysis at the start of the trial and remain on it throughout follow-up. 

This setup was an observational analogue of a per-protocol comparison, estimating the survival outcomes associated with transplant, including perioperative and early post-transplant risks, while isolating the effect of continued dialysis in the comparison arm.

Follow-up 

Follow-up started at treatment assignment and ended at death, recovery of kidney function, or loss to follow-up, or administrative censoring, whichever occurred first.  Administrative censoring occurred 5 years after waitlisting (dialysis arm) or 5 years after transplantation (transplant arm), and in all patients on December 31, 2019.

The two groups were handled differently if the treatment changed. In the transplant group, patients remained under follow-up even if their graft later failed and they returned to dialysis, fitting the point treatment strategy. In the dialysis group, patients were censored if they received a deceased or living donor transplant, as continuing dialysis was a sustained treatment strategy.

Statistical Analysis

Matching. Because a new trial started each month, transplanted patients were compared with patients who had spent the same time on the waiting list while on dialysis, giving exact matching on waiting time at assignment.

Confounding. Inverse probability of treatment weights (IPTW) were estimated by logistic regression of treatment on age at trial start, sex, primary renal disease, calendar year of waitlisting, time on dialysis before waitlisting, diabetes, and country. Continuous variables were modeled with restricted cubic splines. Covariate balance before and after weighting was assessed with absolute standardized mean differences, with 0.1 as the threshold for acceptable balance. CVD was not included in the main models; a sensitivity analysis added it in patients with available data. 

Informative censoring. Censoring of dialysis patients at transplantation may be informative, because transplantation depends on health status. Inverse probability of censoring weights (IPCW) were estimated by logistic regression of censoring on the same baseline covariates. 

Weights. IPTW and IPCW were stabilized using the marginal probabilities of transplant or censoring, multiplied into a single weight, and truncated at the 99th centile to limit the influence of extreme values. 

Survival model. A weighted flexible parametric (Royston-Parmar) survival model was fitted, with a restricted cubic spline for the baseline hazard and time-varying coefficient for treatment, allowing the hazard ratio to change over follow-up and capture early post-transplant mortality.  Age at trial start was also modelled with a restricted cubic spline and an age x treatment interaction, allowing the treatment effect to vary non-linearly with age. The model yielded marginal 5-year survival probabilities, time-varying hazard ratios, and restricted mean survival time (RMST) at 5 years, the mean time alive during the first 5 years. Results are reported at 5-year age intervals.  Confidence intervals for all survival estimates were based on the 2.5th and 97.5th centiles from the distribution of 1000 parametric bootstrap resamples. 

Subgroup analyses were restricted to adults aged 40 or older because there were relatively few younger transplant recipients.

Sensitivity analyses. Crude (unweighted) analyses of waitlisting in 2000-2009 vs 2010-2019 (pre-2010 DCD analyses include UK patients only); men vs women; patients with both diabetes and CVD; and waitlist duration up to 1, 1-2, and 2-5 years; additional adjustment for CVD.

Results

Baseline Demographics:

Of the 64,036 patients included in the trial (mean age 55, with 64.1% male), most of the participants were from the UK and France, with the leading causes of renal failure being glomerulonephritis (20.9%) and diabetes (18.3%).  

Within 5 years, 61.3% received a deceased-donor kidney transplant after a median waiting period of 441 days: SCD/DBD 43.1%, SCD/DCD 11%, ECD/DBD 37.7%, ECD/DCD 8.2%. Most elderly recipients received ECD transplants; among the 1580 recipients ≥75 years, 1264 received ECD/DBD, 184 ECD/DCD, and 132 SCD kidneys; the median KDRI was 2.6 (Tables S5, S7).

Transplantation was associated with a survival advantage across many recipient ages and donor categories, but the picture became less certain in some older patient groups. 

SCD/DBD

Standard criteria donor kidney transplantation was associated with a significant five-year survival benefit over the dialysis across all recipient ages. Even at the age 75, transplantation significantly reduced mortality risk (HR 0.68; 95% CI 0.59–0.79). This benefit persisted among recipients with diabetes (HR 0.58; 95% CI 0.39–0.85). In patients with cardiovascular disease, a similar trend was observed towards the benefit, though the hazard ratio did not reach statistical significance (HR 0.71; 95% CI 0.47–1.06)

SCD/DCD

Standard criteria donor kidney transplantation after circulatory death was associated with improved survival over continued dialysis across recipient ages. At age 75, mortality was 39% lower (HR 0.61, 95% CI 0.44–0.84), and five-year survival probability was 17% higher (95% CI 10%–23%): 70% (95% CI 64%–76%) versus 53% (95% CI 53%–54%). Results for recipients aged 75 with diabetes (HR 0.69, 95% CI 0.31–1.40) or cardiovascular disease (HR 0.75, 95% CI 0.34–1.75), point estimates suggested a survival benefit, but the findings were inconclusive because of small subgroup numbers (n=45) and confidence intervals that crossed the null value of 1.

ECD/DBD

Expanded criteria donor kidney transplantation after brain death was associated with improved survival over continued dialysis, although the benefit decreased and statistical uncertainty increased with age. At age 75, mortality was 26% lower (HR 0.74, 95% CI 0.66–0.86), and five-year survival probability was 12% higher (95% CI 9%–15%): 66% (95% CI 63%–68%) versus 53% (95% CI 53%–54%). The benefit remained significant in recipients with diabetes (HR 0.70, 95% CI 0.55–0.86), but was inconclusive in those with cardiovascular disease (HR 0.82, 95% CI 0.65–1.03), where CI crossed 1.

ECD/DCD

For expanded criteria donor kidney transplantation after circulatory death, the survival benefit declined more sharply, and statistical uncertainty increased with age. At age 75, five-year survival did not differ significantly from that with continued dialysis: 59% (95% CI 51%–66%) versus 55% (95% CI 54%–55%). The mortality estimate was close to the null value of 1 (HR 0.97, 95% CI 0.70–1.33). These findings were consistent across diabetes and cardiovascular disease subgroups.

Older candidates with cardiovascular disease, therefore, emerge as a group in whom the survival advantage warrants particularly careful interpretation. These estimates do not demonstrate consistently poor transplant outcomes; several point estimates favour transplantation, while their confidence intervals remain compatible with both benefit and harm. Therefore, any survival advantage depends on the recipient’s clinical profile and the donor kidney being considered.

Donor Age

The transplant survival benefit decreased with advancing donor age across all recipient categories, with transplant and dialysis survival curves converging for donor ages above 70 years overall and from 60 years among recipients with cardiovascular disease.

Paradox: Risk vs Gain- The Early Post-Transplant Mortality

During the first months after transplantation, mortality risk was higher than with continued dialysis, particularly in older recipients and those receiving expanded criteria donor kidneys. At age 75, peak early post-transplant mortality hazard ratios, derived from modelled curves, were 2.0 for SCD/DBD, 1.7 for SCD/DCD, 2.1 for ECD/DBD, and 3.0 for ECD/DCD.

Time gained: months of life gained within 5 years (RMST)

At age 75, compared with continued dialysis, the adjusted average survival time gained over five years was 2.7 months (95% CI 1.6–3.8) for standard criteria donors after brain death, 4.1 months (95% CI 1.7–6.0) for standard criteria donors after circulatory death, and 1.8 months (95% CI 0.8–2.8) for expanded criteria donors after brain death. Expanded-criteria donors after circulatory death showed no clear advantage (−1.4 months; 95% CI, −4.4 to 1.2).

KDRI

The transplant survival benefit decreased, and became non-significant in the oldest recipients, with kidneys in the highest KDRI quintile; in all patients, the HR at recipient age was 0.76 (0.55-1.04) for the fourth of five quintiles (Fig S4). In patients with CVD, no KDRI fifth quintile reached significance at age 75; HRs rose from 0.63 (0.33-1.14) in the lowest to 0.97 (0.63-2.45) in the highest fifth. With donors aged ≥ 70, the HR was 0.81 (0.66-0.99) at recipient age 75 and 1.13 (0.82-1.57) at 80; in patients with diabetes, HRs at 75 ranged from 0.61 to 0.84 across donor age groups, without a consistent trend (Fig S5, Table S12). 

Sensitivity analyses

Findings were consistent across transplant periods (2000–09 versus 2010–19) and largely comparable between men and women. Longer waitlist duration was associated with a modestly greater transplant survival benefit, which could potentially reflect poorer survival on dialysis. Adjustment for cardiovascular disease had little effect on the results.

Discussion

Deceased donor kidney transplantation (DDKT) generally is associated with survival advantage over dialysis in the majority of clinical scenarios, but the trial reveals a critical gradient, the benefit diminishes as the donor quality declines and recipient age increases. Standard criteria donor transplantation, including either brain death or circulatory death, consistently confers a 5-year survival benefit across all age groups; for SCD/DCD kidneys in the oldest patients with diabetes or CVD, estimates were imprecise because such transplants were rare. Conversely, the survival advantage of expanded criteria donor declines progressively with advancing recipient age, most steeply for ECD/DCD kidneys. 

The over-75 tail

Loss of benefit follows a combination of recipient age, cardiovascular disease, and donor quality, not an age threshold. At 75, no benefit could be shown for ECD/DCD kidneys (HR 0.97, 0.70-1.33), for ECD/DBD kidneys in CVD (0.82, 0.65-1.03), or for any KDRI fifth in patients with CVD. With donors aged 70 or older, the benefit seen at recipient age 75 (0.81, 0.66-0.99) was absent at 80 (1.13, 0.82-1.57). The ECD/DCD interval is compatible with 30% lower or 33% higher mortality: it shows no detectable benefit, no absence of benefit. 

These tail estimates are hypothesis-generating. Recipients aged 75 or older made up 4% of transplants (1,580 of 39,244); only 184 received ECD/DCD kidneys, with about ten deaths in each comorbidity subgroup, so homogeneity across subgroups is not demonstrated. Estimates at 75-85 years come from an age spline that borrows information from younger recipients. Selection is strongest at these ages and acts through unrecorded factors such as frailty, which doubles post-transplant mortality (McAdams-DeMarco MA et al, Am J Transplant, 2015). The ECD/DBD estimate at 75, based on 1,264 recipients, is more robust (HR 0.74, 0.66-0.86).

Early mortality explains why gains in time are small. Peak early hazards at 75 were 1.7 to 3 times those on dialysis (Fig S7), so a 15-point gain in 5-year survival with SCD/DBD kidneys equals 2.7 months of life. With 5-year dialysis survival of 54% at 75, the 5-year window covers much of the remaining life expectancy: differences of -1.4 to + 4.1 months are close to the whole survival effect. Outcomes that matter more at this age, such as hospital admissions, infections, functional decline, time at home and quality of life, were not measured. 

Comparison with other studies
This is the largest trial emulation of the question. Earlier emulation from France (7-month gain in life expectancy at 10 years; Lenain R et al, Epidemiology, 2021), Austria (benefit at all ages; Strohmaier S et al, JAMA Netw Open, 2022), and the US (5-year benefit at ≥ 75 years; Leeaphorn N et al, AJKD, 2026) did not stratify by donor quality. An Australian study found a 10-month gain over 10 years with the lowest-quality 10% of kidneys in patients aged ≥60, but pooled all older patients (Zhu L et al, JAMA Intern Med, 2025). Gill et al. found long-term benefit even with ECD kidneys in 25,468 US candidates aged over 65 years, but measured benefit from listing and could not separate DCD donors (Gill JS et al, Am J Transplant, 2013). Studies that ignored immortal time overestimated the benefit (Chaudhry et al, BMJ, 2022).

Strengths

The major strength is that this was a multinational study including a vast dataset of 64000 patients over 5 European registries. The design removes immortal time, lead time, and depletion-of-susceptibles bias and matches arms exactly on waiting time. Results are stratified by single year of age, comorbidity, donor quality, and retrieval type. Reporting follows TARGET, and the code is public. 

Limitations acknowledged by the authors

 Confounders were measured at baseline only; frailty, temporary inactivation, hospital admissions, and evolving CV risk were not captured, and residual confounding probably overestimates benefit. Temporary delisting was not accounted for. Donor categories are coarse, and HLA sensitisation, preservation techniques, and immunosuppression were unavailable. Some combinations were rare. Follow-up was limited to 5 years without multistate modelling. Results may vary by period, region, and centre, and apply only to waitlisted patients. Donor types cannot be compared with each other, because each has its own comparator.

Critical appraisal

First, the contrast in outcomes is measured in transplantation versus never being transplanted (Table S1). The editorial describes it as accepting an organ at a decision point (Kaewput and Patel, BMJ, 2026), but declining an offer to wait for a better one. This scenario describes a completely different strategy, estimated by neither (Husain SA et al, JAMA Netw Open, 2019| Jay CL et al, Transplantation, 2017). 

Second, confounding was incompletely controlled. After weighting, age and country remained imbalanced in several subgroups, and transplant-arm weights reached 21.7-34/7 before truncation (Fig. S3). DCD transplants were 82-87% British (Table S4A), so the DBD-DCD contrast partly reflects countries. CVD, missing 19.5%, was not in the main models. An unmeasured confounder with a risk ratio of about 1.5-1.6 could explain the lower confidence limits of the SCD and ECD/DBD benefits (E-values; reviewers’ calculation, VanderWeele TJ, Ding P, Ann Intern Med, 2017). For marginal kidneys, bias could also run against transplantation if they are offered to candidates with a poor waitlist prognosis.

Third, the “threefold” early risk is the peak of a modelled curve without a confidence interval. Several hundred interval estimates are reported, and variance estimation across 1.49 million correlated person-trials is not described. 

Fourth, the supplement contains errors. The table S11 repeats the fourth KDRI fifth as the fifth, the ECD/DBD HR is 0.74 in the text and 0.76 in table S8, and the methods and legends name different calendar-year confounders. 

Overall, the direction and gradient of the findings are credible; the ECD/DCD result at ≥ 75 years should be reported as uncertain. 

Implications for practice

Older candidates do receive deceased donor kidneys: all 1580 recipients aged ≥ 75 years here are considered “old”, and the Eurotransplant Senior Program allocates kidneys from donors ≥ 65 to recipients ≥ 65 (Frei U et al, Am J Transplant, 2008). But longevity matching keeps the best kidneys from them: only 132 (8.4%) recipients aged ≥ 75 received SCD kidneys. In the US, about half of candidates aged ≥ 70 were projected to die before receiving a deceased donor kidney (Schold JD et al, CJASN, 2009). The realistic choice is a living donor kidney, with outcomes comparable to or better than SCD transplantation (Patel K et al, Transpl Int, 2024| Toapanta N et al, Transpl Int, 2024), or a lower-quality deceased donor kidney versus continued waiting (Jay CL et al, Transplantation, 2017| Wey A et al, Am J Transplant, 2018). Age alone should not decide eligibility, as chronologic age and physiologic age can vary significantly (Kaewput and Patel, BMJ, 2026| Chadban SJ et al, Transplantation, 2020).

For the 75-year-old with diabetes offered an ECD/DCD kidney: mortality is higher in the first months, and 5-year survival is similar to dialysis, with uncertainty in both directions. A better deceased donor kidney or a living donor kidney offers a clearer benefit. Finally, quality of life, which improves after transplantation even in older recipients (Tsarpali V et al, Kidney Med, 2021| deBoer SE et al, Transpl Int, 2024) but wasn’t measured here, should carry substantial weight and merit significant discussion. 

Conclusion

Deceased donor kidney transplantation was associated with a survival benefit compared with continued dialysis in most settings. However, the survival benefit associated with ECD transplantation decreased with advancing recipient age; for ECD/DCD kidneys at 75 or older it could not be demonstrated, and the estimate is imprecise. Where survival gain is small or uncertain, the decision rests on the patient’s priorities. 

Summary by

Urvashi Khan,

Associate Consultant (Nephrology),

Max Superspeciality Hospital,Noida,India

S.M.Assad,

Assistant Professor (Nephrology)

Christian Medical College,Vellore,India

Reviewed by

Cristina Popa, Brian Rifkin, Akshaya Jayachandran

Header Image created by AI, based on prompts by S.M.Assad

Increasing plasma sodium with Tolvaptan under regulated water intake: comparison with hypertonic saline

Nabil William Sweis, Jennifer Tuazon, Yusra Cheema, Robert Rosa, Daniel Batlle, Gary Robertson

Why is the present study needed? 

Hyponatremia is highly prevalent and affects approximately 35% of hospitalized patients (Adrogue HJ et al, JAMA, 2022). Although there is still a great amount of debate, most clinicians believe the rate of correction is the primary issue in the treatment of hyponatremia. Slower correction increases the risk of brain herniation, while faster correction increases the risk of osmotic demyelination (ODS) in high risk persons. The serum osmolality is tightly regulated between 275 and 295 mOsm/kg by vasopressin, which acts on the V2 receptor in the collecting duct. This mechanism triggers the transport of aquaporins from the basolateral to the apical membrane, thereby increasing water absorption and concentrating urine. (Martinez-Sanchez FD, et al. World J Nephrol, 2026) .

Vaptans increase serum sodium by blocking V₂ receptors, which causes aquaresis and improves serum sodium levels in a more physiological way; however, some articles report a risk of ODS with vaptans (Morris JH, et al. AJKD, 2018; Scalla AP, et al. J of Pharm Pract, 2024). A meta-analysis with 1840 patients with 3 RCTs and 2 cohorts, claimed a higher rate of correction with the vaptans compared to placebo and fluid restriction with an OR of 5.72 for overcorrection, with a pooled overcorrection incidence of 13.1% vs 3.3%, yet zero observed ODS events. (Krisanapan P et al. J Clin Med, 2023). Can we control the rate of correction with vaptans and use them to correct hyponatremia while avoiding overcorrection?

How was the study done?

The study included healthy volunteers by way of advertisement and was conducted at Northwestern Memorial Hospital in Chicago. Out of 17 volunteers, the first 6 received 60 mg of tolvaptan, and of the next 11 volunteers, 6 received 30 mg and completed the study; the remaining 5 were disqualified due to difficulty in voiding, dizziness while voiding, and alcoholic intoxication. The study was funded by Otsuka American Pharmaceuticals, the manufacturer of vaptans.

The selected subjects were admitted to the hospital at 7:00 am; their vitals, height, weight, and time of recent void were recorded. At 9:00 am, they were instructed to void into a preweighed plastic container, and urine volume was determined. The sample was measured for urine osmolality, and blood was collected for plasma osmolality and serum sodium. Another sample was stored at -20°C for vasopressin measurement by radioimmunoassay. After collecting basal data, the first six participants received 60mg of Tolvaptan and the next received 30 mg of tolvaptan they were instructed to take tolvaptan with 150 ml of water. Every hour, weight, urine osmolality, plasma osmolality, serum sodium, and vasopressin samples were collected. They were given room-temperature water to drink equal to the amount of urine volume (UV) in the previous hour, minus 5 ml/kg of body weight (underreplaced). After the 6 hours of regulated water intake, the subjects were advised to eat, drink, exercise, and sleep as they wished until the next day at 7 AM. The same measurements were taken at 2, 2, 4, and 8 hours.

Hypertonic saline infusion: After completing the tolvaptan study, the same participants were admitted 2 weeks later, and the same basic measures were taken as in the tolvaptan study. They were infused with hypertonic saline at 0.033 ml/kg body weight (approx. 2 ml/kg/h cumulative) over six hours with complete water restriction, and the same measurements were taken for the next 4 hours.

What did they find? 

Tolvaptan 60mg group: 

Initial 6 hrs of the regulated water intake phase:

• Urine osmolality, urine volume, and free water clearance were monitored hourly.

• Urine osmolality decreased from baseline 478 to 59 at the 3rd hour and gradually increased to 154 at the 6th hour.

• Urine volume increased from 72 to 749 at the 3rd hour, then gradually decreased to 451 by the end of the 6th hour.

Thirst: Thirst increased progressively from basal to the 6th hour of regulated water intake On a scale of 0 to 10, the individual ratings (mean ± SD) rose from a basal level of 2.0 ± 1.5 to a maximum of 5.8 ± 2.9 after 6 h of regulated water intake, and mean thirst rates were correlated with the plasma osmolality during the first 6 h of regulated water intake. (R = 0.8687, P < 0.05)

Water intake: At zero hour, all six consumed 150 ml of water, and after that, they consumed an amount equal to UOP for that hour, lessened by 5 ml/kg as calculated by the nurse in charge (controlled water intake).

Body weight: Since the amount of water intake was smaller than the amount excreted, the participants went into a negative water balance, which was also reflected in their body weight, decreasing from 71.8 ± 13.5 kg to 69.8 ± 13.2 kg after 6 h of regulated water intake. When adjusted for differences in body weights, individual losses over the 6 h of regulated intake were very similar, ranging from 2.7 to 3.4% of basal body weight. 

.

Plasma osmolality and sodium: The decrease in body weight is closely correlated with the increase in plasma osmolality and plasma sodium during regulated water intake. Osmolality increased from 287 ± 1.6 to 300 ± 3.0 mosmol/kgH₂O, and sodium concentration increased from 138 ± 0.6 to 144 ± 2.0 mEq/L, respectively.

Vasopressin (AVP): The progressive rise in plasma osmolality and plasma sodium resulted in a progressive rise in vasopressin, and the osmotic threshold for vasopressin release was similar, ranging from 286 to 290 mosm/l.

Ad libitum water intake phase (1500 h - 700 h): 

After 6 h of regulated water intake, all volunteers were permitted to eat and drink at will; this phase was divided into two parts, the first 8 h part and the second 8 h part.

First 8 h part: Urine osmolality and urine volume were maintained at the same level. But body weight increased, and plasma osmolality, sodium, and vasopressin declined about halfway back to basal pretreatment levels. 

Last 8 h part: The subjects slept, water intake decreased dramatically, urine osmolality increased, and urine output decreased; however, plasma osmolality and plasma sodium continued to fall but remained above basal levels, and vasopressin rose slightly but remained well above basal levels the next morning.

Tolvaptan 30 mg group: Despite comparable peak PNa (~144 mEq/L in both groups), the percentage rise in POsm/PNa was smaller with 30 mg (3.3-3.6%) than 60 mg (4.1-4.3%), even though weight loss relative to body mass was nearly identical between groups (~2.8% in both). Using the paper’s own free-water model (TBW-57% of body weight), the observed weight loss in both groups should have produced a ~4.9% rise in POsm/PNa; the actual rise fell short in both, and more so in the 30 mg group.  

Figure 5. Urine osmolality and volume, water intake, and plasma osmolality and sodium following Tolvaptan (30 mg) from Sweis N, et al. Am J Physiol Renal Physiol, 2026

3% Hypertonic saline infusion:

After 10-14 days following the tolvaptan study, the same subjects were infused with IV hypertonic saline (3%) for 6 h at a rate of 0.033 ml/kg/hr; during the infusion they were not allowed to drink, and the same variables were measured.

Urine osmolality and body weight increased, and urine volume also increased, but not to the same level as after tolvaptan. Plasma osmolality, plasma sodium, and thirst also rose similarly to tolvaptan. Even vasopressin levels rose, but not to the same extent as in the tolvaptan study. 

After hypertonic infusion stoppage, the individuals were allowed to eat and drink at will for 4 hours. At that time, urine osmolality and volume did not change, but plasma osmolality and plasma sodium were decreased by half of the values during the infusion. Vasopressin levels decreased to baseline levels.

Comparison of the Effects of Tolvaptan and Hypertonic (3%) Saline Infusion

The slopes of changes in plasma sodium and plasma osmolality after each dose of Tolvaptan (60 or 30 mg) are 2.1 to 2.4, while those after hypertonic (3%) saline infusion are 1.8 to 2.6, and both are similar. However, the slope of the rise in plasma AVP after Tolvaptan 30 and 60 mg is twice that of the rise with 3% saline.

There was a high correlation between plasma osmolality and plasma AVP levels for all three treatments: tolvaptan 60 mg, tolvaptan 30 mg, and 3% NS in all 3 groups, with an r value above 0.9 and a p value < 0.001, how ever in each of the 6 subjects in whom plasma AVP was measured, it rose in close correlation with plasma osmolality during treatment with tolvaptan when fluid intake was regulated (r=0.9736) as well as when it was not (r=0.9360). A similar correlation was observed during the infusion of hypertonic saline when fluid intake was not permitted (r=0.9637) as well as when it was (r=0.9360). However, the slopes of the relationships between plasma AVP and plasma osmolality differed appreciably not only as a result of the two different treatments but also between subjects during each treatment.

Critical Review

How can this study impact our practice? Can we use tolvaptan in chronic hyponatremia without the fear of ODS?

Although a recent meta-analysis (Ayus J, et al. JAMA Intern Med, 2025) suggested that slow and very slow sodium correction are associated with increased risk of mortality and longer length of hospital stay compared to rapid correction, guidelines have not changed regarding the rate of correction, as the meta-analysis was based on observational cohorts and not RCTs. 

The present study supports the hypothesis that the rise in plasma osmolality and sodium under regulated water intake can be controlled with tolvaptan intake. Though tolvaptan produced thirst, it did not correlate with the magnitude of increase in urine output, but showed a strong correlation between body weight decline and rise in plasma osmolality and sodium where water intake was not regulated by thirst but by under-replaced urine output.

 In the initial 8 hrs of unrestricted water intake, plasma osmolality and sodium continued to decline about halfway back to basal pretreatment levels, even though they continued to excrete abnormally large volumes of dilute urine, suggesting that vaptans with normal water intake may not increase plasma sodium levels. In the next 8 hrs, when the subjects were sleeping, their intake was reduced, and the urine osmolality increased. 

Strengths: 

The study explained the impact of vaptans with regulated fluid intake and when given ad libitum fluid intake. The study showed that regulated fluid intake in healthy males resulted in a similar correction of plasma osmolality as per the guidelines. 

Limitations: 

1) The study was performed only in males. 

2) The study was conducted only in healthy subjects whose kidneys and water regulatory systems would not allow them to experience serious side effects, but how practical is it for patients with already damaged water regulatory systems? This is especially true for patients who are more prone to ODS during correction, such as those with chronic alcoholism, malnutrition, or hypokalemia. 

3) This treatment cannot be given on an outpatient basis because hourly measurement of urine output and adequately replacing that output is practically impossible, so it requires ICU management. 

4) It was not a safety study: no ODS-relevant endpoints were assessed, and the absence of adverse events beyond thirst does not constitute a safety evaluation. 

5) Duration mismatch with clinical practice: the regulated phase was 6 h; the Na rise it produced (5-6 mEq/l) largely reversed overnight once ad lib intake resumed. Real hyponatremia correction is a multi-day process. What’s untested is whether a fixed per-hour deficit sustained over 24-48 h (vs a single 6 h burst) produces the same predictability or compounds error.

6) The unresolved TBW-model shortfall (predicted 4.9% rise vs. observed 3.3-4.4%) means the underlying free-water deficit arithmetic (the basis of the 5 mL/kg/h prescription itself) is an approximation, not a validated dosing formula. 

Conclusions:

 The authors have demonstrated that tolvaptan, when combined with regulated water intake, has a rate of sodium rise similar to 3% NS infusion in normal healthy males. Further studies in a broader population of hospitalized patients with hyponatremia are needed to prove the efficacy of tolvaptan at the bedside and to support its wider use in hyponatremia.

Summary by

Sai Vani Yellampalli

Reviewed by

Brian Rifkin, Cristina Popa

#NephJC Ten POsts Chat

Tuesday, September 29th 2026, 9 pm Eastern

Kidney International Reports, 2026 

Randomized Controlled Trial of Corticosteroid Tapering in Steroid-Sensitive Podocytopathy 

Arun Prabhahar, Joyita Bharati, Tamanna Sharma, Vivek Sood, Pavitra Manu Dogra, Thakur Sain, Vinod Kumar, Ritambhra Nada, Smita Subhash Divyaveer, Manish Rathi, Harbir Singh Kohli, Vivekanand Jha, and Raja Ramachandran
doi: 10.1016/j.ekir.2026.107004 

Introduction

The fewer steroids, the better. This has been the catchphrase in glomerulonephritis in the past decade, with the caveat that we are not paying for it with more relapses. Adults with steroid-sensitive podocytopathy can respond to glucocorticoids, but relapse is common, and repeated courses can quickly add up to a substantial steroid burden. That matters because prolonged glucocorticoid therapy brings familiar baggage: metabolic complications, weight gain, risk of infection, bone disease, and other dose-related toxicities. The catch, of course, is that steroids work as a systemic, but non-specific, modifier of immunity and inflammation. Once remission is achieved, the practical question becomes: over what period should we taper steroids before enough is enough?

For a question we face so often in patients with nephrotic syndrome, the evidence in adults is surprisingly thin. Relapse rate in adult MCD has been reported between 61- 73% (Fenton et al, BMC Nephrology 2018; Waldman et al, CJASN 2007), with 28% frequently relapsing in the Waldman Cohort. Data from the Toronto Glomerulonephritis Registry (Troyanov et al, KI 2023) showed a 57% relapse rate among patients with FSGS and hypoalbuminemia, with a median time to relapse of 7 months. 

KDIGO Guidelines recommend high-dose glucocorticoids as first-line therapy for adult MCD (1C), and suggested tapering after remission for up to about 24 weeks of total glucocorticoids exposure. For primary FSGS, glucocorticoids are recommended first line (1D), with a Practice Point suggesting a total duration of ≥6 months in responders. Infrequent MCD relapses and relapses of previously steroid-sensitive FSGS can be retreated with a shorter glucocorticoid course (tapering over 1-2 months). Much of the current practice is shaped by observational experience rather than randomized data.

Studies in children have given us some clues as to how to effectively decrease steroid exposure in nephrotic syndrome (one of the few instances where the peds literature surpasses the adult side). Randomized trials (Sinha et al, KI 2015 and Yoshikawa et al, KI 2015; Webb et al, Health Tech Assess 2019  | NephJC summary) ) found no clear advantage in extending the initial steroid treatment compared with shorter courses. Whether we can carry that lesson into adult practice has yet to be supported by RCT data.

NephJC has wrestled with the  question of steroid minimisation in adults before while discussing tacrolimus as first line therapy in adult MCD (Thomas et al, CJASN 2020| NephJC summary). Despite all the hullabaloo around steroids, they remain the first line therapy for both childhood and adult MCD and FSGS considering their cost, availability and the vast literature supporting their use.

Prabhahar et al set out to address the evidence gap regarding length of steroid taper in adults with podocytopathy (MCD/FSGS). In the current study adults with biopsy-proven steroid-sensitive MCD or FSGS who had already achieved remission with prednisolone were randomized to either a 2-month or 4-month taper. The trial was not asking whether the shorter taper was better, but whether it was noninferior at preventing relapses. If the taper could be cut in half without meaningfully increasing relapses, then the steroid-sparing payoff could be substantial. The question they addressed: can we taper steroids sooner without paying for it later?

The Study

Methods

Study Design

The START (Steroid Tapering protocol in Adult Nephrotic Syndrome due to Podocytopathy) was a randomized, parallel-group, active-controlled, open-label noninferiority trial conducted at two centers in India. Enrollment took place from May 2021 to May 2023, with follow-up continuing through December 2025. Participants were centrally randomized using a computer-generated sequence, with allocation concealed. Neither participants nor treating nephrologists were blinded to treatment assignment; however, the person generating the randomization sequence was not involved in patient care or outcome assessment.

Study Population

Inclusion Criteria:

• Adults >18 years

• Biopsy-proven MCD or FSGS

• Newly diagnosed or infrequently relapsing nephrotic syndrome

• Achieved remission with oral prednisolone alone, defined as proteinuria <0.5 g/day at randomization

Exclusion Criteria: 

• Hepatitis B, hepatitis C or HIV infection

• Corticosteroid exposure within the previous 12 months or other immunosuppressive therapy within the previous 12 months.

• Persistent kidney dysfunction despite proteinuria remission

• Diabetes mellitus

• Active urinary sediment (>5 RBC/hpf) or ANA positivity 

• Suspected secondary cause of nephrotic syndrome

• Steroid intolerance

• Abnormal liver function tests

Intervention

Patients initially treated with oral prednisolone 1 mg/kg/day for up to 8 weeks were randomized after remission to the following arms

• Short-course arm: prednisolone tapered over 2 months

• Standard-course arm: prednisolone tapered over 4 months

Both tapering protocols used stepwise dose reductions based on the starting prednisolone dose, with dose changes every 7 days in the 2 month arm and every 14 days in the 4 month arm.

All participants also received an ACE inhibitor or ARB (Did they really need that?), along with calcium and vitamin D, until completion of steroid therapy.

After steroid completion, participants were followed regularly through 12 months (monthly for 2 months, then with decreasing frequency of follow up till month 12) and then every 6 months for long term follow up. Serial assessment of 24-hour urine protein, serum creatinine, and serum albumin was done in all patients.

Outcomes

The primary outcome was refreshingly simple: relapse free-remission at 12 months after randomization. The study-specific definition of remission at randomization was proteinuria <0.5g/day (slightly higher than the KDIGO definition of 0.3 mg/day), and relapse was defined as recurrence of nephrotic syndrome after remission.

The secondary outcomes examined some interesting endpoints as well:

• Time to first relapse

• Relapse-free remission by MCD vs FSGS

• Relapse-free remission in treatment-naive vs infrequently relapsing patients

• Relapse-free remission at last follow-up

• Cumulative prednisolone exposure

• Safety and adverse events

Sample Size

The investigators assumed that 70% of patients in both groups would remain relapse-free at 12 months. Because this was a noninferiority trial, they also had to decide how much worse the shorter taper could be before it would no longer be considered an acceptable alternative. They set that margin at 25 percentage points meaning that the short taper could not be more than 25 percentage points worse than the standard taper. The authors based this on the lack of prior adult randomized data, the high relapse rates reported even with prolonged steroid courses in MCD and steroid-responsive FSGS, and the potential benefit of roughly 50% less steroid exposure. For more on non-inferiority trials, see this NephTrials summary.

With these assumptions, 106 patients were needed to provide 80% power using a one sided alpha of 0.025; allowing for a 5% attrition increased the target enrollment to 114 patients. 

Statistical Analysis

The primary analysis was intention-to-treat (ITT), supported by a per-protocol (PP) analysis. Noninferiority had to be demonstrated in both. Noninferiority was met if the lower bound of the one-sided 97.5% confidence interval for the difference in relapse-free remission remained above -25 percentage points. A sensitivity analysis treated missing 12-month outcomes as relapses, and all other outcomes were considered exploratory. Relapse-free remission according to histology (MCD vs FSGS) and clinical phenotype (treatment-naive vs infrequently relapsing) was included among the secondary outcomes. Separately, the broader subgroup analysis (shown in Figure 3), including sex, age, histology, and phenotype, was described as nonprespecified and should therefore be considered exploratory.

Funding

The study was funded by the Indian Council of Medical Research, New Delhi. The manuscript does not describe a specific role for the funder in study design, data analysis, interpretation, or manuscript preparation. Vivekanand Jha reported consulting fees and honoraria from several pharmaceutical companies, with payments made to his institution; all other authors declared no competing interests.

Results

Of 118 patients with MCD or FSGS, 94 achieved complete remission within 8 weeks and were randomized to short- or standard-course taper. The target sample size of 114 could not be met in the specified time interval.

Eighty-eight patients completed follow-up, with only six lost to follow-up. All randomized patients were included in the ITT analysis, while 80 were included in the PP analysis.

There is an error in the consort diagram above (did you spot it?), the 24 excluded patients should appear above the 94 remission patients who were then randomized.

The study population was relatively young (mean age 31.9 years; 60.6% male), with MCD and FSGS almost equally represented (51.1% vs. 48.9%). The standard treatment group had a slightly higher proportion of patients with FSGS and treatment naive participants compared to the short course group. Just over half the patients (55.3%) were treatment-naïve. Median steroid exposure before randomization was comparable. At presentation, median proteinuria was 6.0 g/day, decreasing to 0.24 g/day at randomization, with serum albumin increasing from 2.1 to 3.7 g/dL. Baseline characteristics were comparable between treatment groups.

Primary outcome

At 12 months, relapse-free remission was 63.3% with the short-course vs. 57.8% with the standard-course regimen in the ITT analysis. The risk difference was 5.5% (lower CI bound −14.3%), remaining above the prespecified −25% noninferiority margin (P=0.001). Noninferiority was also confirmed in the PP analysis (66.7% vs. 68.6%; risk difference −1.9%, lower CI bound −22.6%; P=0.014).

Secondary Outcomes 

Regarding the type of disease, outcomes differed numerically between MCD and FSGS. In patients with MCD, relapse-free remission was more frequent with the short-course regimen (75% vs. 55%; risk difference +20.0%), whereas in FSGS, it was less frequent (47.6% vs. 60%; risk difference −12.4%). However, neither comparison reached statistical significance. 

Relapse-free remission at 12 months was similar between regimens in both treatment-naïve patients (60.7% vs. 54.2%; P=0.63) and those with relapsing disease (66.7% vs. 62%; P=0.75).

At 36 months, relapse-free remission remained comparable between the short- and standard-course groups (38.8% vs. 44.4%; P=0.58), with no significant difference in time to relapse. However, the short-course regimen substantially reduced cumulative prednisolone exposure, with a median post-randomization dose of 1750 vs. 3430 mg in the ITT analysis (P<0.001). This reduction was also observed in the PP analysis (2310 vs. 4620 mg; P<0.001).

No significant differences in treatment efficacy were observed according to underlying disease (MCD 51.1%; FSGS 48.9%) or disease phenotype (treatment-naïve 55.3%; infrequently relapsing disease 44.7%). 

Safety profile

Adverse events were significantly less frequent with the short-course regimen (24.5% vs. 53.3%; RR 0.46, 95% CI 0.26–0.81; P=0.003), with similar findings in the PP analysis.

Steroid-related adverse events were consistently more frequent with standard therapy, including dysglycemia (8.9% vs. 0%), cosmetic effects (15.5% vs. 4.1%), and arthralgia/myalgia (13.3% vs. 2.0%). Serious adverse events, as defined by the study protocol, were numerically less frequent with the short course (8.2% vs. 17.8%). However, the SAE definition also included events necessitating treatment interruption or protocol deviation, which is broader than conventional SAE definitions and may have contributed to the apparent difference between groups. 

Two serious infections requiring hospitalization occurred in the short-course group (1 case of severe COVID-19 infection and 1 case of renal abscess with pyelonephritis). In the standard-course group, serious adverse events included steroid-induced myopathy/arthralgia requiring protocol deviation (n=3), psychiatric disturbances (n=2),hemorrhoidal bleeding requiring blood transfusion (n=1), and avascular necrosis of the hip (n=1).

Discussion

The optimal duration of corticosteroid therapy in adult steroid-sensitive nephrotic syndrome remains uncertain. Current KDIGO guidance (KI 2021) recommends glucocorticoids as first-line therapy for adult MCD/primary FSGS, with tapering starting 2 weeks after complete remission, with a total exposure steroid exposure period of 6 months.  These are Practice Points based on small observational studies or extrapolated from childhood NS, largely reflecting limited adult evidence and real concerns regarding relapse with shorter treatment courses.

Attempts to reduce corticosteroid exposure in adults have produced mixed results. In a prospective study of 35 adults with steroid-sensitive MCD, a 2-month steroid regimen significantly reduced cumulative corticosteroid exposure compared with conventional treatment (Ozeki T et al, Am J Nephrol 2019). However, the shorter regimen was associated with an earlier occurrence of relapse (adjusted HR 2.45, CI 1.51-3.91), although the risk of frequent relapse was not significantly different. Importantly, this study was an observational study with a historical control rather than a randomized trial.

More recently, the ADAPT randomized noninferiority trial explored a steroid-minimizing strategy in adult MCD, this time by reducing the corticosteroid dose rather than treatment duration, comparing prednisolone 0.5 mg/kg/day plus alfacalcidol with conventional prednisolone 1 mg/kg/day (Kristensen et al, Kidney Int 2026). Remission (88% vs. 91%) and relapse rates (35% vs. 32%) were similar between groups, with no significant differences in time to remission or relapse. The lower-dose strategy substantially reduced cumulative prednisolone exposure (3.7 vs. 8.0 g) and glucocorticoid toxicity. However, formal noninferiority was not established. 

Against this background, the present trial adds randomized evidence for adults with podocytopathies. Unlike previous studies focused predominantly on MCD, it included both MCD and steroid-sensitive FSGS, directly evaluating whether shortening the taper, after remission has already been achieved, could preserve relapse-free remission. Noninferiority at 12 months was demonstrated in both ITT and PP analyses, while cumulative prednisolone exposure was approximately halved.

The clinical trade-off therefore appears favorable: similar relapse control with substantially less corticosteroid exposure. This was further reflected in the safety findings, with adverse events occurring more than twice as often in the slow taper group. The reduction was particularly apparent for typical steroid-related metabolic, cosmetic, and musculoskeletal complications.

Protocol deviations were also more frequent in the standard-course group (6 vs. 2), with most deviations in this group related to steroid-associated adverse events, further supporting the tolerability advantage of reducing corticosteroid exposure. However, the reduction in overall adverse events should not be interpreted as evidence that the shorter taper reduces serious infections, hospitalization, or long-term complications. Notably, the two serious infections requiring hospitalization occurred in the short-course group. 

The findings are also reassuring over longer follow-up. At 36 months, relapse-free remission remained similar between groups, strengthening the conclusion that shortening steroid therapy did not simply postpone relapses beyond the 12-month primary endpoint. Notably, the persistence of comparable outcomes at 36 months provides important longer-term support for the durability of the short-course strategy. Nevertheless, fewer than half of patients in either group remained relapse-free, highlighting that long-term relapse remains an important clinical challenge. This highlights that long-term relapse remains an important unresolved issue irrespective of taper duration, but that simply prolonging corticosteroid exposure may not substantially modify the underlying tendency to relapse. 

Limitations

So, should we just taper  steroids quickly for adult MCD/FSGS?! Well, there are important reasons for caution. The study was smaller than planned, enrolling 94 of the targeted 114 participants, which reduced the statistical power to approximately 75% and resulted in relatively wide confidence intervals. The −25% noninferiority margin was also wide, meaning that noninferiority should not be interpreted as equivalence and a clinically meaningful difference cannot be completely excluded. Although the point estimate favored the shorter taper, the lower confidence interval bound of −14.3% means that the results remain compatible with approximately 14 additional relapses per 100 patients treated with the shorter regimen. Whether this potential loss of efficacy represents an acceptable trade-off for substantially lower corticosteroid exposure may differ between patients, particularly when previous relapses have been severe or complicated by acute kidney injury, infection, or thrombosis. Additionally, the relapse free remission was much lower than the assumed 70% used to calculate sample size, further questioning the evidence this sample size could generate.

The study was open-label and conducted at only two centers, potentially limiting generalizability. The findings, therefore, cannot be directly extrapolated to older or more comorbid patients, steroid-resistant FSGS, or patients in whom induction therapy is limited by toxicity.  

Importantly, the trial was not powered to detect histology-specific differences. At baseline, a higher percentage of patients in the short course group were MCD rather than FSGS. Additionally, microscopic hematuria, which is more common in FSGS compared to MCD, and a predictor of proteinuria remission (Yoshida et al, Kidney Medicine 2026), was excluded in the present study. Interpretation of the FSGS subgroup is further complicated by the heterogeneity of FSGS itself. FSGS represents a histological pattern encompassing primary, genetic, and secondary forms of podocyte injury. Although nephrotic presentation and steroid responsiveness increase the likelihood of primary FSGS, electron microscopy was not mandatory and systematic genetic testing was not reported, leaving some uncertainty regarding the classification of the FSGS population. 

Although relapse-free remission numerically favored the short-course regimen in MCD and the standard-course regimen in FSGS, these subgroup findings should be considered hypothesis-generating rather than definitive. This is particularly relevant for FSGS, where evidence supporting shorter steroid courses remains much more limited and current KDIGO guidance does not support extrapolating shorter corticosteroid regimens from MCD to FSGS. 

Whether steroid-sensitive MCD and FSGS should therefore be managed with the same corticosteroid tapering strategy remains an important unanswered question and warrants adequately powered, histology-specific studies. 

Conclusion

START shifts the balance toward shorter steroid exposure, with a 2-month taper meeting the prespecified noninferiority criterion while substantially reducing cumulative prednisolone exposure. The inability to meet sample size, wide noninferiority margin and uncertainty around FSGS keep this from being a one size fits all taper. 

Summary by

Andreea Manolea,

Nephrology resident,

Clinical Hospital "Dr. C. I. Parhon" Romania

Farah Wehbe,

Renal hypertension fellow,

University of Ottawa, Canada

Reviewed by

Pallavi Prasad, Brian Rifkin

Header Image created by AI, based on prompts by Brian Rifkin

The START trial compared 2-month and 4-month prednisolone tapers in adults with minimal change disease or FSGS. Explore the study design, findings, and clinical implications on the next #NephJC

El ensayo START comparó pautas de 2 y 4 meses de prednisolona en adultos con enfermedad de cambios mínimos o GEFS. Conoce el diseño, los hallazgos y sus implicaciones clínicas en el próximo #NephJC

This year we will be at The Denver Press Club, a cool venue, a mere 4 minute stroll from the Colorado Convention Center.

We will have an open bar, dinner, and entertainment including another live recording of Freely Filtered.

If you want to go to the party you need to donate to NephJC. All donations for the fall fund drive include a complimentary ticket to NephJC Night 2026. Otherwise, you can just buy a ticket here.

…and, as extra enticement, some images from NephJC Night, Houston Texas, 2025!

Thoughtful critical appraisal of the 26 most impactful manuscripts every year, and even more with shorts. Half of those are additionally covered by our podcast, Freely Filtered.

No paywalls. No subscriptions. No industry support.

NephJC exists because our readers and listeners choose to support it.

For 12 years, you’ve made this possible.

Help us keep it going for the next 12.

DONATE

We suggest attendees pay $600 per year; this will keep the lights on.

All donations include a ticket to the NephJC party.

If you just want to go to NephJC Night at Kidney Week, and don’t want to make any additional donation, click here

Where does the money go?

NephJC is a 501(c)(3) nonprofit. Nobody makes a dime running the place—but running the place isn’t free.

We have an absurd number of domains to register and maintain: NephJC.com, Flozinator.com, NephRun.com, NSMC.blog, FreelyFiltered.com, Rosebook.club, VisualAbstract.org, and, yes, NephMadness.com. Cha-ching.

NephJC, NSMC, Channel Your Enthusiasm, and The Skeleton Key all have websites that need to be hosted and maintained. Those fees add up.

We pay for Zoom, Auphonic, Descript, stock photography, and the other tools we use to create and distribute FOAMed.

And then there are the NephJC trophies. Those things are surprisingly expensive—especially when we ship them halfway around the world.

None of these expenses is enormous. But together, they add up.

We also believe an online community should occasionally become a real-life community. So once or twice a year, we get people together. We’ve bowled at KIDNEYCon, taken over a bar at NKF Spring Clinical, and every year we throw the best party at #KidneyWk.

And yes, we’re doing it again this year. Keep Friday, October 23rd open.

That’s what your donation pays for: the infrastructure, tools, and community that keep NephJC independent, open, and free.

A good time was had by all at the 2026 NephJC #KidneyWk Party

How have we paid for all of this?

Over the years, we’ve funded NephJC in a lot of different ways.

The first year, we literally passed the hat among the NephJC workgroup. Satellite Healthcare has been a longtime and generous supporter. In 2019, Arkana gave us an unrestricted three-year grant. That was awesome. And in 2022, sales of the Flozinator Pin helped carry us through the pandemic.

But the till is getting low, and we need your help.

Our bylaws prohibit NephJC from accepting money or support from the pharmaceutical industry. Given our mission—independent, critical appraisal of the medical literature—we believe that would create an unacceptable conflict of interest.

That’s a hill we are proud to die on.

It also means we’ve cut ourselves off from the deepest pockets in medicine.

Our dream is for NephJC and everything associated with it to be supported by the people who actually use it. Think Wikipedia or NPR for nephrology education: everything remains free to everyone, but the people who value it and can afford to support it help keep it going.

No one will ever encounter a paywall. No one will ever be turned away because they can’t pay. The podcasts, journal clubs, visual abstracts, websites, and everything else remain free and open.

That’s why we’re holding the annual NephJC Fall Pledge Drive.

If NephJC has taught you something, helped you take better care of a patient, prepared you for journal club, made you laugh, or just made nephrology a little more fun, please pony up.

Help keep NephJC independent.
Help keep NephJC free.
Help keep NephJC going.

Choose a level below and make a donation. All donations are tax-deductible in the United States. Any merchandise offered with a donation is a thank-you gift; you are not purchasing those items.

September 1st, 2026, 9 pm EST

Kidney Int Rep,  2026 Jun 23;11(9):106673., doi: 10.1016/j.ekir.2026.106673.eCollection 2026 Sep.

Randomized Controlled Trial of Large Language Model–Assisted Diagnostic Accuracy in Nephrology 

Raphaël Bentegeac, Philippe Amouyel, Bastien Le Guellec, Wisit Cheungpasitport, Mehdi Maanaoui, Aghiles Hamroun

PMID: 42502668
DOI:  10.1016/j.ekir.2026.106673

Introduction

Modern Artificial Intelligence (AI) encompasses a spectrum of computational approaches, from traditional machine learning to deep neural networks capable of processing increasingly complex data (see NephMadness 2026, AI region). Among the most rapidly evolving applications of deep learning are large language models (LLMs) that are trained on vast amounts of information and can interpret, synthesize, and generate natural language.  Since the release of ChatGPT in late 2022,  medicine has been grappling with a collective identity crisis.  In the last few years, there has been an expanding body of research evaluating the performance of different LLMs across a wide range of medical tasks, from answering clinical questions and interpreting complex case vignettes to diagnostic reasoning and medical decision-making. A recent systematic review (Chen SF et al, Nat Med, 2026)  found that LLMs outperformed human comparators in 33% of studies, with better relative performance in knowledge-based assessments than in real-world clinical settings, highlighting the persistent gap between benchmark performance and clinical utility. Essentially, AI programs are very good at multiple-choice exam questions.

Nephrology has not been an exception to this trend. In an evaluation of 975 questions from the Nephrology Self-Assessment Program (NephSAP) and Kidney-Self-Assessment Program (KSAP), GPT-4 achieved 74% accuracy, although remaining below the 77% passing threshold and the average performance of nephrology examinees (Miao J et al, Clin J Am Soc Nephrol, 2024). Moreover, beyond knowledge-based assessments, the potential applications of LLMs in nephrology have rapidly increased to include patient management, clinical documentation, decision support, and medical training. Evidence supporting the safety and effectiveness of LLMs in real-world nephrology settings remains limited (Yongzheng Hu et al, Ren Fail, 2025). 

Against this vast background, the REFINe (Reasoning Enhancement With Feedback From a Generative AI in Nephrology) was conducted. Rather than asking whether a LLM could independently solve nephrology questions, REFINe addressed a more clinically relevant question: Can access to a high-reasoning LLM improve physicians diagnostic accuracy when faced with complex, open-ended nephrology cases?

REFINe tested whether AI can better diagnosticians of physicians- a more relevant question for the future of AI-assisted medicine. 

Methods

REFINe was a prospective, randomized, open-label, parallel-group superiority trial conducted entirely online between November 2025 and March 2026. English or French-speaking residents and board-certified physicians were randomized 1:1 to diagnose complex nephrology cases with or without LLM assistance. Each case evaluation was one unit of analysis. The review board of lille University Hospital approved the study. The study group planned to enroll 100 participants and recruited 97. They report no change to the criteria, the outcomes or the analyses after the trial started. 
The cases came from the Make Your Diagnosis series in Kidney International, published between August 2007 and January 2025. Two nephrologists (N.F. and A.H.) rewrote 245 of these cases. The aim was to stop the LLMs from recognizing cases used during training. Study designers changed every sentence and every number but kept the clinical logic and the final diagnosis. Ten cases were then chosen at random for the trial. 
It is worth noting what those ten cases detailed, because this shapes how we should interpret percentages in the paper. There were four glomerular and vascular cases, two tubular, interstitial and cystic disorders, two electrolyte and acid disorders, one case of acute kidney injury in critical care, and one transplant case. Seven of the ten included kidney or allograft pathology. One included urine microscopy. Two were text only. One was pediatric. This question inventory tells us more about the difficulty than any adjectives or descriptors could. 

Participants were residents or board-certified doctors who spoke English or French. They were recruited through professional societies and conferences. The main exclusion criteria was previous contact with the specific cases. The team used the Pocock and Simon minimization algorithm (80% fixed and 20% random) to balance five features: certification status, speciality, years of experience, how often the doctor used LLMs, and academic status. The process was automatic, and the allocation sequence was hidden from the study staff. This is a stronger allocation method than most similar AI studies have used. 

Each participant received the 10 cases, in random order, on a dedicated platform. They could not go back to review a previous case. In the control group, doctors gave up to three diagnoses and a confidence score, then moved directly to the next case- no AI input at any point. In the AI group, participants completed the same first step, and the platform showed them a fixed GPT-5 answer with short, structured reasoning: they could revise their diagnoses and confidence once before moving on. The AI answer came from a two step prompt with no examples given to the model. The first step returned exactly three ranked diagnoses. The second step turned them into case specific reasoning with probabilities and a short summary of the key differences. All GPT-5 answers were produced in October 2025 and then frozen (this was a fixed ranked list with probabilities, not a conversation). 

Then two board-certified nephrologists scored the main outcome. They worked independently and did not know from which group each answer was obtained. They agreed on about 97% of answers before any discussion, and the rest were solved by consensus. For emphasis, the whole result of this trial rested upon the “yes” or “no” judgments about free text answers by the adjudicators. 

The sample size came from a simulation. It assumed 50 participants per arm, 10 cases each, 50% accuracy in the control group, and 70% in the AI group. With 100 participants, the power was calculated to be above 99%. The main analysis used mixed-effects logistic regression, with a fixed effect for the AI study arm and random intercepts for the participant cases. It was adjusted for gender, country, speciality, academic status, experience, confidence in AI, how often the doctor used LLMs, and how many AI tools they used. Every randomized participant was analysed in the assigned group. The subgroup analyses were planned in advance. The analyses by number of cases completed were done afterwards. The R 4.3.2 software environment was used for data analysis, statistical modelling, and graphics.

REFINe Population

Results

The 97 participants produced 556 case evaluations. The median age was 38 yo (IQR 34 to 45). Three out of four said nephrology was their speciality, and 69% did not work in academic centres. Nine countries were represented, but 87% to 93% of participants were French, depending on the arm. The two groups were balanced at baseline. About four in ten participants had never used an LLM. 

Primary Outcome

Secondary Outcomes

Doctors only took half of the good advice!

There were 178 answers that were wrong at first where the AI suggestion prompted the correct answer. Of those, 85 (47.8%) were corrected, and just under half were not. In the AI arm, doctors changed their top-3 in 47.4% of answers (from 41.3% to 53.6%).

Harm was rare

Only 1 of 69 answers that were right at first became wrong after a bad AI suggestion (1.5%; 95% CI: 0.04 to 7.9). In this format, doctors were not simply following the machine. 

No subgroup effect reached significance (all interaction p > 0.05). The analyses by number of cases completed gave similar effect sizes, but they also show something the main paper does not discuss.

One pattern stands out despite the null result: monthly LLM users showed by far the largest AI benefit (OR 10.51, 95% CI 3.18-24.67), well above never users or daily users, and the LLM-usage interaction (p=0.008) was the closest to significance of any subgroup tested (supp figure 3).

The adjusted odds ratios were 4.3 (1.99 to 9.28) for doctors who completed at least five cases, and 3.94 (1.70 to 9.13) for those who completed all ten. 

Before the trial, the team tested eleven models on all 245 cases. This took about three hours and cost around EUR 300. GPT-5 was the best performer in the pre-trial benchmark (77.1%), so it was the model used in the AI arm.

DeepSeek V3.2 and MedGemma 27B were tested with text-only inputs because those setups could not accept images. Their scores therefore, mix model ability with a missing input. 

Funding: The authors received no specific funding for this work. 

Data availability statement: Participant consent did not include permission for public data sharing. The analysis code will be made publicly available in a Git repository.

Discussion

From AI versus nephrologists to AI with nephrologists

LLMs have already shown that they can readily solve board-style questions. This study has moved the field of study from multiple-choice questions to complex, integrated clinical vignettes requiring the synthesis of history, laboratory data, pathology, urine microscopy and images. This trial addresses a very practical and pressing question: Does giving clinicians access to an LLM actually improve diagnostic reasoning in nephrology? In this RCT of fixed scenarios, the answer appears to be yes. While the results seem like a slam dunk endorsement of our machine overlords (we will add your biological and technological distinctiveness to our own; resistance is futile), we need to read in between the lines. Top-3 and Top-1 diagnostic accuracy improved by around 20% with AI, and the effect was preserved among participants who completed at least 5 cases, but it could have been an even greater improvement if the physicians had trusted the AI more. 
Clinical nephrology is open-ended, and new clinical information is integrated in ways particular to our specialty. Our diagnoses rely on the incorporation of heterogeneous data types, including a urinary sediment, light microscopy and immunofluorescence pattern, blood gases, and a serology panel that half fits multiple etiologies (diseases rarely read textbooks and don’t deal in absolutes). There is rarely a diagnosis that relies solely on an individual, decisive test. Answers may depend on trajectory rather than on a point value, since the same creatinine of 2.4 mg/dL means one thing over three days and something entirely different over three years. Finally, nephrologists draw on a long tail of low-prevalence entities, precisely the region in which a probabilistic model may either shine or possibly exaggerate the exotic. 

The human - AI gap may be the real story

There were 178 occasions where the clinician's initial answer was wrong, but GPT-5 gave the correct diagnosis. Yet only 85 (47%) of those answers were eventually corrected. This is probably one of the most important findings of the study. The problem is not always whether AI can correctly determine the diagnosis, but rather whether physicians are willing and/or able to recognize their mistakes. When it comes to correct diagnosis, humans are subject to errors in logic, including anchoring, premature closing, and confirmation bias (Mutlack Z, et al. J Clin Med 2025). AI assistance brought clinicians substantially closer to the apparent performance ceiling of the model, but nowhere near all the way (maybe the real trial was about us, not LLMs). That is where this paper becomes less about how clever GPT-5 is and more about how we, as clinicians, respond when AI disagrees with us. Do we trust it, challenge it, ignore it, or simply refuse to change despite the evidence? 

Why does a tool capable of very high diagnostic accuracy not automatically transfer that accuracy to the clinician sitting in front of it? The study workflow and the AI interaction here were very artificial and may have contributed to the gap. GPT-5 produced exactly three ranked diagnoses with concise reasoning and probability estimates. The answer was pre-generated and frozen. There was no chat, no follow-up, no chance to ask, “Why not my diagnosis?” or “Which one feature in this case makes you think that?” That matters because most of us do not use AI like a static answer key. We question it, push back, ask for discriminating features, ask it to argue against itself, and sometimes keep asking until either we are convinced or the model starts hallucinating creatively enough that we close the window. So, the question is not simply whether GPT is accurate. It is how much of that accuracy can be transferred to a clinician in a safe and useful way.

What has been examined previously

Other studies with the use of LLMs and physicians diagnosing complex cases have had mixed results. Goh and colleagues found no benefit at all from giving doctors an LLM, while the model on its own beat both groups (physicians alone and AI assisted physicians) by 16 points, similar to the current study (Goh E, et al. JAMA Netw Open, 2024). Everett and colleagues did improve accuracy with collaborative workflows (Everett SS, et al. NPJ Digit Med, 2026). We know LLMs keep evolving (quickly), but human integration may take longer due to trust and fear of machine hallucinations. Presentation of source materials might help physicians be more accepting of AI prodding changes, especially if the logic is solid and the primary references could be reviewed. Again, a conversation versus a mandate seems to work better in altering a doctor’s clinical reasoning.

Many previous studies scored the quality of the reasoning against a rubric (Goh E et al, JAMA Netw Open, 2024| Everett SS et al, npj Digit Med, 2026| Qazi IA, et al. Nature Health, 2026). REFINe scored a binary “yes” or “no”: was the right diagnosis somewhere in the list. The nephrology study mixed a percentage with a completeness rating. 

We don't want doctors to blindly accept whatever an AI says. If AI gives a diagnosis that doesn't fit the patient, it should be questioned, because currently AI has its own set of logistical errors (hallucinations, prejudice due to training errors, and context loss). There’s another side to this; even if the AI is actually right, our own clinical experience and preconceived notions can make us dismiss it. The trial results show that the ability to decide when to listen to AI may be just as important as the AI's accuracy. The authors make a similar point, suggesting that AI literacy and how the interface feels to us may be important steps in human/AI integration to achieve the best results.

Considering the reverse scenario, among the responses that were initially correct, very few became wrong after viewing the AI suggestion. This means the participating clinicians didn’t simply abandon a correct diagnosis whenever the AI disagreed with them. That is reassuring, yet we should be cautious, because the study was small and the AI was presented in a very controlled format somewhat dissimilar from real world clinical reasoning.

It is important to note, the control group in this study did not have access to the usual clinical resources. That is quite different from trials in which doctors could use Google, UpToDate, or other conventional resources. In addition, LLMs are evolving at an incredible rate, making what’s true in machine logic yesterday almost obsolete in just a few weeks. We cannot simply say that REFINe proves that LLMs are better than physicians (this sounds like a headline AI would write). It just shows that adding this particular AI workflow to unaided clinical reasoning improved diagnostic performance. In clinical practice, AI is being integrated in a somewhat haphazard way, slowing potential real world improvements. However, AI will be a useful tool for clinicians who choose to enhance diagnosis and treatment strategies with an eye on confirming machine-generated recommendations.

The number we are highlighting is the distance between 57.5% (physician enhanced with AI) and 77.1% (AI alone), and the fact that half of the correct suggestions were simply not used. That gap is not an AI model problem. It is a problem about how we read, trust, and wrestle with these tools. Qazi and colleagues showed just how much this can be moved: an AI-literacy curriculum alone lifted diagnostic assistance accuracy from 42.6% to 71.4%, a 27.5 point gain - arguably a bigger lever than the model itself (Qazi IA et al, Nature Health, 2026). It is becoming quite apparent that AI literacy is no longer optional for the modern physician. Mastery of AI (a potentially daily consulted assistant) sits closer than the urine sediment and biopsy interpretation in core kidney education than most of us are comfortable admitting. Similar to the acceptance of the electronic medical record, AI will be transformational once everyone gets on board.

Strengths

Methodologically, there is quite a lot to like. Group allocation used the Pocock-Simon minimization algorithm, with 80% deterministic and 20% random assignment. This balanced certification status, specialty, years of experience, LLM usage frequency and academic status. Randomization was automated and allocation remained concealed from study personnel during enrollment. The primary analysis used mixed-effects logistic regression with random intercepts for both participant and vignette, which is important because each clinician contributed multiple cases and each vignette had a different intrinsic level of difficulty. The model also adjusted for gender, country, specialty, academic status, experience and AI-use characteristics. Although the intervention itself could never be blinded, the diagnostic outcome adjudication was. This is a stronger allocation method than most similar AI studies have used. 

Weaknesses

The sensitivity analysis is also reassuring as the benefit did not disappear when the analysis was restricted to people who completed more of the task. At the same time, 97 participants generated only 556 completed evaluations, despite being assigned up to 10 cases each, and that’s before counting the 46 signups (32% of all registrants) who never completed a single case and were excluded from analysis all together. That raises a question: were difficult cases more likely to be abandoned or did participants simply decide that ten consecutive renal diagnostic puzzles were enough nephrology for one evening? Completion was voluntary and uncompensated, so informative non-completion remains worth thinking about even though the sensitivity analyses are reassuring.

Baseline confidence was similar between control and AI groups, but confidence increased significantly after exposure to AI (p<0.001). That makes intuitive sense if someone or something has just agreed with you, produced three polished diagnoses, and assigned probabilities to them. You naturally feel more certain. But there is an important caveat here. The confidence has to move in the correct direction, not just increase. If AI makes a correct clinician more confident, that is useful; if it makes a wrong clinician more confident, that may be dangerous.

The control arm also had no access to conventional resources. That makes the trial clean experimentally, but it is not quite how nephrology is practiced. If we come across a rare genetic tubulopathy, an unusual biopsy or a toxic alcohol case, or a lot many numbers, we may look something up. That is not cheating; that is how nephrology is currently practiced. The clinically relevant future comparison is therefore probably usual resources versus usual resources plus LLM. 

Conclusion

This is the first randomized trial in nephrology to show that a structured LLM workflow improves diagnostic accuracy, with very few errors introduced by the AI. Although this study may not currently have real world implications, the growing breadth and speed of LLM evolution will make it an essential tool for all future physicians.

One last detail, buried in the disclosures: the authors used Claude Sonnet 4.5 and GPT-5 to help write and edit this manuscript. A paper about human-AI diagnostic collaboration was itself a product of human-AI collaboration, which feels about right.   

Summary by

Anca Elena Stefan
Nephrology specialist, Assistant Lecturer
Romania

Assad S M

Assistant Professor, Nephrology

Christian Medical College, India

Srinivasavaradan Govindarajan
Assistant Professor (Ped Neph)
VMMC and Safdarjung Hospital, India

Luis Daniel Ramírez-Calvillo

Nephrology Fellow
Instituto Nacional de Cardiología Ignacio Chavez, México

Reviewed by

Akshaya Jayachandran, Brian Rifkin, Cristina Popa

Special thanks to guest reviewer & author

Wisit Cheungpasitporn (The AI guy)

Header created by AI and prompts from Assad SM