The prior NephJC discussion of finerenone versus spironolactone concluded that the comparative evidence base was asymmetric: explanatory trial design, not necessarily pharmacology, explained finerenone’s superior track record. Two papers published concurrently in June 2026 extended the evidence base beyond diabetic CKD and permitted a more direct test of that claim- a prespecified exploratory subgroup analysis of FIND-CKD restricted to glomerular disease (Neuen BL et al, JAMA, 2026) and INFINITY, a prespecified individual-participant-data pooled analysis of three phase 3 trials- FIDELIO-DKD, FIGARO-DKD, FIND-CKD (Neuen BL et al, Lancet, 2026). 

FIND-CKD (Heerspink HJL et al, NEJM, 2026 | NephJC Summary) randomized 1584 adults with non-diabetic CKD (eGFR 25- <60 ml/min/1.73 m² with UACR 200-500 mg/g, or eGFR 25-90 with UACR 500-3500 mg/g) to finerenone 10/20 mg or placebo on maximized RAS blockade. The primary outcome, total eGFR slope to month 32, favored finerenone by 0.7 ml/min/1.73 m²/year (95% CI, 0.3-1.1; P<0.001)- magnitude comparable to the effect sizes reported for RAS inhibitors and SGLT2i in similar populations. 

Resources: Brenner BM et al, NEJM, 2001, Holtkamp FA et al, Kidney Int,  2011| Heerspink HJL et al, Lancet Diabetes Endocrinol, 2021| EMPA-KIDNEY Collaborative Group, Lancet Diabetes Endocrinol, 2024| Heerspink HJL et al, NEJM, 2026

The prespecified hierarchical secondary composite (kidney failure, sustained ≥57% eGFR decline, heart failure hospitalization, or cardiovascular death) reached significance at HR 0.77 (95% CI, 0.60-0.99). 

JAMA. 2026 Jun 5:e269923. doi: 10.1001/jama.2026.9923. Online ahead of print.

Finerenone in Patients With Chronic Kidney Disease Due to Glomerular Diseases: A Randomized Clinical Trial

Brendon L Neuen, Vlado Perkovic, Rajiv Agarwal, David Z I Cherney, Carolyn S P Lam, Christoph Wanner, Katherine R Tuttle, Pantelis Sarafidis, Jonathan Barratt, Anna Burgner, Xiangmei Chen, Alfredo Chew-Wong, Vladimir A Dobronravov, Jürgen Floege, Kieran McCafferty, Masaomi Nangaku, David Packham, Evangelos Papachristou, Pablo E Pergola, Marijn M Speeckaert, Arunkumar Subbiah, Sydney C W Tang, Shuifu Tang, See Cheng Yeo, Niels Jongs, J David Smeijer, Mario Berger, Meike Brinker, Rania Dayoub, Jay Elliott, Na Li, Katharina Mueller, Nicole Rethemeier, Marina Yael Finkelsztein, Hiddo J L Heerspink; FIND-CKD Investigators

PMID: 42246414

DOI: 10.1001/jama.2026.9923

Why was the study done? 

Glomerular diseases account for a substantial proportion of global CKD-attributable kidney failure and disproportionately affects younger patients. For most glomerular disease subtypes, disease-specific therapeutic options remain limited, creating an unmet need for interventions that target shared mechanisms of progression independent of underlying etiology. While the selective, nonsteroidal mineralocorticoid  receptor agonist (nsMRA) finerenone is well-established in diabetic nephropathy (FIDELIO-DKD, NephJC summary| FIDELITY), clinical evidence evaluating its efficacy and safety within primary glomerular disease cohorts has historically been scarce.

The mechanistic basis for extending selective mineralocorticoid receptor (MR) blockade to podocytopathies- notably focal segmental glomerulosclerosis (FSGS)- is rooted in cell-specific receptor localization (Lazareth H et al, Kidney Int, 2025). The MR is functionally expressed on podocytes, which is the primary cellular injury site in FSGS. Pathological MR overactivation drives podocyte cytoskeletal disruption, slit diaphragm effacement, and the upregulation of pro-inflammatory and profibrotic pathways. Preclinical models demonstrate that selective MR antagonism halts this cascade, mitigating podocyte injury and slowing the progression of glomerulosclerosis (Shibata B et al, Nat Med, 2008, Tagaki N et al, Nephron Extra, 2012).

Methods 
This was a prespecified exploratory subgroup analysis of FIND-CKD (Heerspink HJL et al, NEJM, 2026| NephJC summary). Of 1584 randomized participants, 57% had investigator-reported glomerular disease, 712 (78.8%) biopsy-confirmed: IgA nephropathy 46.1%, FSGS 23.8%, membranous nephropathy 10%, MPGN 2.9%, and other 17.3%.

 Table1. Demographics of participants with glomerular diseases at baseline, from Neuen BL et al, JAMA, 2026

Mean baseline eGFR was 48 mL/min/1.73 m², and the median UACR was 839 mg/g. Randomization was stratified by screening UACR category and baseline SGLT2i use, but not by CKD etiology. Therefore, comparisons across glomerular disease subtypes don’t retain the protection afforded by randomization and may be affected by measured or unmeasured baseline imbalances. This limitation doesn’t compromise the internal validity of the trial but constrains causal interpretation of subtype-specific comparisons.

What do the results say?

The total GFR slope was -3.5 mL/min/1.73 m²/year with finerenone versus -4.23 with placebo; the between-group difference was 0.73 (95% CI, 0.22-1.24).

Figure 2. Change in eGFR over time allocation in participants with glomerular diseases, from Neuen BL et al, JAMA, 2026

Treatment effect was consistent across disease subtypes, with no evidence of heterogeneity (P for interaction = 0.52). The estimate was 1.34 (95% CI -0.29 to 2.38) in FSGS and 0.61 (95% CI, -0.14 to 1.36) in IgAN. Among biopsy-confirmed cases, the estimate was 0.89 (95% CI, 0.31-1.47). UACR decreased by 42% (95% CI 35-48%) at 12 months, similarly across subtypes.

Figure 3. Subgroup analysis according to glomerular disease etiology, from Neuen BL et al, JAMA, 2026

Kidney failure or ≥40% eGFR decline occurred less frequently with treatment (7.42 vs 9.60 events per 100 patient-years, HR 0.74, 95% CI 0.57-0.97). 

Figure 4A. Time to event analysis of key secondary and exploratory efficacy outcomes in participants with glomerular disease, from Neuen BL et al, JAMA, 2026

Results were consistent across subtypes and in biopsy-confirmed disease (HR 0.74, 95% CI, 0.55-0.99) (Figure 5). However, this composite was selected post hoc because it generated the most events and should be interpreted as exploratory. Serious adverse events were similar between groups, including serious hyperkalemia (0.9% in both).

Critical view

Three issues limit the strength of inference. First, surrogate endpoint validation: the CKD-EPI Clinical Trials Consortium meta-analysis (Inker A et al, Nat Med, 2023) established that a total eGFR slope effect of 0.75 mL/min/1.73 m²/year in a trial of comparable size (900 participants) confers a 99% probability of benefit on clinical kidney outcomes, with a median predicted HR of 0.74- a threshold the observed 0.73 mL/min/1.73 m²/year effect sits just under—and a predicted HR that matches the observed composite-outcome HR of 0.74 exactly. This concordance is reassuring but should be read as internal consistency between a surrogate and a clinical outcome measured in the same trial, not as an external replication. 

Second, the FSGS estimate was nominally significant but imprecise (N=215, 1.34 ml/min/1.73 m²/year; 95% CI 0.29-2.38). The subgroups are not powered for a definitive subtype-specific conclusion. Moreover, DUPLEX - the largest completed phase 3 FSGS trial- found no clear benefit of sparsentan over irbesartan on eGFR decline, increasing the evidentiary threshold for claiming efficacy in FSGS (Rheault MN et al, NEJM, 2023| NephJC summary). 

Third, generalizability was limited by the cohort’s demographics: mean age 51 years, 57-62% male, approximately 60-65% Asian, and few Black participants. About 20% lacked biopsy confirmation, while many available biopsies were historical rather than obtained at enrollment. Interaction tests were not adjusted for multiple comparisons, increasing the risk of chance-positive findings; their limited power also means that significant interactions do not establish uniform treatment effects across subtypes.

Lancet. 2026 Jun 13;407(10546):2375-2386.doi: 10.1016/S0140-6736(26)01009-3. Epub 2026 Jun 5.

Efficacy and safety of finerenone in patients with chronic kidney disease: an individual participant data pooled analysis (INFINITY)

Brendon L Neuen, Hiddo J L Heerspink, Vlado Perkovic, David Z I Cherney, Carolyn S P Lam, Katherine R Tuttle, Christoph Wanner, Pantelis Sarafidis, Stefan D Anker, Gerasimos Filippatos, Bertram Pitt, Peter Rossing, Luis M Ruilope, Niels Jongs, J David Smeijer, Meike Brinker, Christiane Ahlers, Andrea Lage, Andrea Horvat-Bröcker, Patrick Schloemer, Thomas Eissing, Rania Dayoub, Robert Lawatscheck, Rajiv Agarwal; FIND-CKD, FIDELIO-DKD and FIGARO-DKD Investigators

PMID: 42248158

DOI: 10.1016/S0140-6736(26)01009-3

Why was the study needed?

No individual finerenone trial was powered for kidney failure alone or all-cause mortality definitively. Pooling individual-participant data increased power for these hard time-to-event outcomes and enabled assessment of effect modification by diabetes status, which FIND-CKD could not address because it excluded patients with diabetes. 

How was the study done?

14,574 participants pooled across FIDELIO-CKD, FIGARO-DKD, and FIND-CKD.

Table 1. Baseline characteristics, from Neuen BL et al, Lancet, 2026

The protocol was prospectively registered to PROSPERO prior to FIND-CKD completion. Systematic search (PubMed, Embase, to March 31, 2026) identified no additional eligible trials; 301 records were screened to 3 included studies (figure S1), each assessed as low risk of bias across RoB2 domains (figure S2). 

The main kidney outcome was kidney failure or sustained ≥57% eGFR decline (matching the earlier FIDELITY definition). The main cardiovascular outcome was heart failure hospitalization or cardiovascular death; undetermined deaths were excluded. Important methodological differences exist across the trials: cardiovascular and mortality events were independently adjudicated in FIDELIO-DKD and FIGARO-DKD, but in FIND-CKD these events were investigator-reported via structured case-report forms without independent adjudication. The polled hazard ratios for cardiovascular and mortality outcomes therefore combine data collected under two different ascertainment standards.

What do the results say?

The composite kidney outcome occurred at 22.3 versus 28.8 events per 1000 patient-years (HR 0.76, 95% CI 0.68-0.86). Kidney failure alone was reduced more modestly (HR 0.85, 95% CI 0.74-0.99), with the upper confidence limit close to null. The HR for dialysis or transplantation alone was 0.77 (95% CI, 0.64-0.94), while sustained ≥57% eGFR decline alone had an HR of 0.71 (95% CI, 0.61-0.81- Table S5).

Figure 1. Cumulative incidence functions for key efficacy outcomes, from Neuen BL et al, Lancet, 2026

The cardiovascular composite occurred at 19.1 vs 23.9 events per 1000 patient-year (HR 0.80, 95% CI, 0.70-0.91). The HR was 0.78 for heart failure hospitalization (p= 0.0024) and 0.82 for cardiovascular death (95% CI 0.67-0.999). All-cause mortality was also lower (HR 0.88, 95% CI 0.79-0.99).

Figure 2. Kidney, cardiovascular, and mortality outcomes in the overall INFINITY population, from Neuen BL et al, Lancet, 2026

There was no evidence of effect modification by glycated hemoglobin (P for interaction = 0.51, figure 3), CKD etiology, eGFR, UACR, or SGLT2i use (figure 4), or trial origin (all P for interaction ≥0.40; figure S6).

Figure 4. Effect of finerenone on the composite kidney outcome across prespecified subgroups, from Neuen BL et al, Lancet, 2026

Hyperkalemia occurred in 14.3% of participants receiving finerenone and 7.6% receiving placebo (Table 2). Hospitalizations for hyperkalemia were uncommon (0.9% vs 0.2%), with no fatal events. The risk of serious hyperkalemia requiring hospitalization differed by diabetes status: HHR 6.21 (95% CI, 3.18-12.12) with diabetes versus 1.48 (95% CI, 0.47-4.67) without diabetes (P for interaction = 0.034). This might reflect the greater prevalence of type 4 renal tubular acidosis in diabetic kidney disease. 

The number needed to treat for the kidney-cardiovascular composite was 29.2, compared with a reported number needed to harm of -114.5 for hyperkalemia during hospitalizations (table S7), indicating an approximately fourfold benefit-to-harm ratio at the population level. Per 1000 patients treated for 3 years, finerenone was estimated to prevent 34 kidney-cardiovascular composite events, 21 kidney events, 14 heart failure hospitalizations or CV death events, and 10 deaths. Larger absolute cardiovascular and mortality benefit in participants with diabetes reflected their higher baseline risk rather than a greater relative treatment effect. NNT remains a dubious metric.

Figure 5. Estimated absolute effects of finerenone per 1000 patients treated over 3 years by diabetes status, from Neuen BL et al, Lancet, 2026

Critical view

FIND-CKD cardiovascular and mortality events were not independently adjudicated. Combining these with adjudicated events from FIDELIO-DKD and FIGARO-DKD introduces potential ascertainment bias (making some members of the study group more or less likely to be included in the study than others). No sensitivity analysis specifically excluding FIND-CKD from cardiovascular and mortality analyses was reported. 

No source trial enrolled participants with screening eGFR <25 mL/min/1.73 m² or non-albuminuric CKD. The findings do not support extrapolation to advanced or non-albuminuric CKD, as in “across the entire spectrum of CKD."

Subgroup interaction tests were not adjusted for multiple comparisons. A nonsignificant interaction test may also reflect limited power and should not be interpreted as proof of uniform benefit across all subgroups.

All three trials and the pooled analysis were funded by Bayer. Several authors were Bayer employees with disclosed equity interests, and Bayer funded medical writing support. Although the data were independently reanalyzed at University Medical Center Groningen, the evidence base remains entirely manufacturer-funded. 

External evidence provides partial support. FINEARTS-HF showed a heart failure benefit in HFpEF irrespective of diabetes status (Solomon SD et al, NEJM, 2024). Pooled FIDELIO/FIGARO analyses found heart failure benefits emerging within the first 6 months (Ostrominski JW et al, JACC, 2026). CONFIDENCE showed greater albuminuria reduction with combined finerenone and empagliflozin than with either drug alone, without increased acute kidney injury (Agarwal R et al, NEJM, 2025| NephJC summary). These findings support the biological and clinical rationale but don’t affect INFINITY’s outcome adjudication or eGFR-floor limitation.

Conclusion

FIND-CKD established that finerenone slows eGFR decline in non-diabetic CKD at an effect size consistent with established kidney-protective therapies; this finding is self-contained and doesn’t depend on the two subsequent analyses for validity. The JAMA subgroup analysis tests whether that effect generalizes across etiology and does not reject the null hypothesis of homogeneity- a result that is informative but constrained by the absence of etiology-based randomization, limited events in the rarer subtypes, and uncorrected multiplicity. INFINITY tests generalization across diabetes status and supplies, for the first time, adequately powered estimates for kidney failure and all-cause mortality- at the cost of pooling across trials with differing outcome ascertainment standards and a shared eGFR floor and albuminuria requirement that bound the population to which these estimates apply. 

Together, these analyses suggest that finerenone’s effect reflects mineralocorticoid receptor activity rather than a specific CKD etiology or glycemic state. This interpretation is supported by consistent findings across etiologic and glycemic subgroups and by external evidence from FINEARTS-HF and CONFIDENCE. However, efficacy remains untested in non-albuminuric CKD and at eGFR <25 mL/min/1.73 m², while cardiovascular outcomes were not uniformly adjudicated. These limitations define the population and outcomes to which the findings can be applied and the priorities for future trials.

Written by

Cristina Popa

Reviewed by

Brian Rifkin

2026 Jun 4. doi: 10.1056/NEJMoa2604625.

Finerenone in Persons with Chronic Kidney Disease without Diabetes

Hiddo J L Heerspink, Brendon L Neuen, Rajiv Agarwal, David Z I Cherney, Carolyn S P Lam,  Katherine R Tuttle, Christoph Wanner, Pantelis Sarafidis, Niels Jongs, J David Smeijer, Meike Brinker, Nicole Rethemeier, Patrick Schloemer, Paula Vesterinen, David Goldsbury, Sara Dizayee, Jon W Mares, Vlado Perkovic; FIND-CKD Investigators

PMID: 42246672

DOI: 10.1056/NEJMoa2604625

Introduction

Few drugs have moved up the ranks as quickly as finerenone in treatment of diabetic kidney disease. Following the landmark FIDELIO-DKD (Bakris DL et al, NEJM 2020 |NephJC Summary), FIGARO-DKD (Pitt B et al, New Engl J Med 2021) and the FIDELITY pooled analysis (Agarwal R et al, European Heart Journal 2022), it has become one of the pillars of management of diabetic kidney disease (DKD) in type 2 DM along with RASi and SGLT2 inhibitors aka flozins. FineONE is doing the same for it in type 1 DM (Heerspink H et al, NEJM 2026 |NephJC Short) with significant antiproteinuric effects, and FINEARTS in HFpEF (NephJC summary). Can the cardiorenal benefits seen in DKD be extended to the much larger population of patients with CKD without diabetes? 

Mineralocorticoid receptor overactivation not only causes fluid and sodium retention but it is also a well-recognized driver of inflammation and fibrosis, which are fundamental pathways underlying CKD progression, irrespective of the underlying cause. In a post hoc analysis of the FIDELIO-DKD, finerenone reduced albuminuria and improved kidney outcomes irrespective of baseline glycemic control or insulin use, suggesting that its kidney protective effects extend beyond glucose lowering.(Rossing P et al, Diabetes Care 2022) But a convincing biological hypothesis does not guarantee clinical benefit. Spironolactone, a steroidal MRA, improves albuminuria and hypertension, but its widespread use in CKD has been limited mostly by lack of robust evidence demonstrating improvement in clinical kidney outcomes. The hypothesis that finerenone, a highly specific, nonsteroidal antagonist of mineralocorticoid receptors (nsMRA) can improve renal outcomes in patients with CKD who do not have diabetes required prospective testing. So, let’s FIND out!

Methods

Study Design
The FIND-CKD study design (Heerspink et al, NDT 2025) was published in 2025. In this multicenter, randomized, double-blind, parallel-group, phase 3 trial, across 24 countries, investigators studied the clinical efficacy and safety of finerenone at doses of 10 mg and 20 mg daily versus placebo in slowing the progression of chronic kidney disease among participants without diabetes.

Study population
The study enrolled male and female patients older than 18 years of age with a clinical diagnosis of CKD without evidence of diabetes. The investigators chose patients with high risk of progression of CKD by eGFR and albuminuria. Accordingly, two categories of inclusion were made: 

• those with eGFR of ≥25 to 60 ml/min/1.73m² with low level albuminuria (UACR ≥ 200 to < 500 mg/g); total number of patients in this group was capped at 10% of total enrollment planned AND

• those between eGFR ≥25 to 90 ml/min/1.73m², if the UACR was ≥ 500 to < 3500 mg/g at the time of screening

The proteinuria / albuminuria of anytime in the prior 3 months could be used, and if not done in the prior 3 months, then the pre-screening investigation could be used to identify patients. The key inclusion and exclusion criteria are shown in figure below: 

Thus those with a clear indication (eg HF or DKD) were excluded, and so were those with lupus, ANCA vasculitis or those on active immunosuppression. Patients with prior organ transplantation, recent dialysis-requiring acute kidney injury, uncontrolled blood pressure or nephrotic-range albuminuria (UACR >3500 mg/g) were also excluded. 

Intervention
The protocol of Finerenone dosing and follow up is shown in the figure below. Notably, there was no run-in period unlike with the FIDELIO/FIGARO trials.

Participants were randomly assigned in a 1:1 ratio to receive oral finerenone or matching placebo in addition to standard-of-care therapy. Participants were randomized using an interactive response technology system, with stratification according to baseline flozin use (yes vs no) and screening UACR (≤1000 vs >1000 mg/g). 

Finerenone (or matching placebo) was initiated at 10 or 20 mg once daily according to baseline kidney function, with protocol-directed dose adjustments throughout follow-up. Participants receiving 10 mg were eligible for uptitration to the target dose of 20 mg after one month if serum potassium remained ≤4.8 mmol/L and eGFR had not declined by ≥30%. Hyperkalemia was managed using a predefined algorithm: treatment was temporarily withheld for serum potassium >5.5 mmol/L and restarted once potassium decreased to ≤5.0 mmol/L. The protocol allowed temporary interruption and rechallenge rather than routine permanent discontinuation. If hyperkalemia recurred despite resumption at the lower dose and no alternative cause was identified, permanent discontinuation of study treatment was recommended. Notably, the protocol did not mandate the use of potassium-binding agents as part of hyperkalemia management. 

An important methodological consideration was the handling of flozins, which had rapidly emerged as foundational therapy even for non-diabetic CKD during the conduct of the trial. The investigators permitted but did not require the co-administration of flozins. For those with a clear clinical indication, the protocol preferred that these agents be started and stabilized at a fixed dose for a minimum of four weeks prior to the screening process. Flozination following randomization was also allowed if deemed necessary by the clinician, though only after the participant had maintained a steady study drug regimen for at least one month. 

Outcomes
The primary efficacy outcome was the total eGFR slope, defined as the mean annual rate of change in the eGFR from baseline to month 32. To account for the expected early hemodynamic decline with finerenone, eGFR was modelled using a two-slope mixed-effects model that separated the acute (baseline to Month 3) and chronic (Month 3 onward) phases of treatment. Secondary outcomes were tested using a prespecified hierarchical strategy and included:

(i) Composite kidney or cardiovascular outcome: 

• Sustained ≥57% decline in eGFR (confirmed for ≥4 weeks; equivalent to doubling of serum creatinine), 

• Kidney failure (eGFR <15 mL/min/1.73 m² confirmed after ≥4 weeks, long-term dialysis for ≥30 days, or kidney transplantation), 

• Hospitalization for heart failure, or 

• Cardiovascular death.

(ii) Kidney-specific composite outcome:

• Sustained ≥57% decline in eGFR, or

• Kidney failure.

(iii) Cardiovascular composite outcome:

• Hospitalization for heart failure, or

• Cardiovascular death.

There were several exploratory outcomes as well as predefined safety (mostly related to hyperkalemia).

Sample size assumptions
The sample size was calculated to provide >90% power to detect a between-group difference of 0.7 mL/min/1.73 m²/year in the total eGFR slope over a 32-month period, requiring the enrollment of approximately 1,500 participants. The primary efficacy analysis followed the intention-to-treat principle and used a two-slope linear mixed-effects model to estimate the acute (baseline to Month 3) and chronic (thereafter) phases of eGFR decline. Secondary time-to-event outcomes were analyzed using stratified Cox proportional hazards models, with statistical significance assessed according to a prespecified hierarchical testing strategy to control for multiple comparisons.

Funding
FIND-CKD was funded and sponsored by Bayer. Seven co-authors were Bayer employees (clinical development, statistics, regulatory affairs, medical affairs, and related functions). Bayer co-designed the trial with the Steering Committee, performed the primary statistical analyses, and participated in manuscript submission. Analyses were independently verified by investigators at the University Medical Center Groningen, and safety oversight was provided by an independent Data Monitoring Committee. 

Results

A total of 1584 participants were randomized and included in the full analysis set, indicating that nearly half of those enrolled (perhaps meaning those screened?) met the eligibility criteria.

The median treatment duration was 36.6 months. Overall, 1566 participants completed the full study, including the month 32 visit. 

The mean age was 55 years; 35% were women, and the median UACR was 819 mg/g. The mean eGFR was 47 ± 16 mL/min/1.73 m², with approximately 80% of participants having an eGFR <60 mL/min/1.73 m². Hypertension was highly prevalent and RASi use was nearly universal (99.7% - as mandated by trial protocol) with most participants receiving an ARB, and 17% of participants were flozinated. Just over half (~ 57%) had chronic GN identified as their underlying cause of CKD.

Primary outcome 
Following the early hemodynamic dip, kidney function declined more slowly with finerenone than with placebo. This treatment effect became evident after the initial 3 months and was sustained throughout follow-up. The primary outcome of total eGFR change was -3.3 vs -4 ml/min/1.73m²/year in finerenone and placebo groups respectively (CI = 0.3-1.1, p<0.001).

Separating the acute and chronic eGFR slopes highlighted the biphasic response to finerenone. During the first 3 months, the acute eGFR slope with finerenone was 1.2 mL/min/1.73 m²/year greater than with placebo. Notably, after treatment discontinuation (in the washout phase), eGFR increased in the finerenone group (+1.2 mL/min/1.73 m²) but continued to decline in the placebo group (−0.5 mL/min/1.73 m²), supporting the reversible nature of the initial hemodynamic dip. 

Secondary outcomes
The secondary composite kidney–cardiovascular outcome (sustained ≥57% decline in eGFR, kidney failure, hospitalization for heart failure, or cardiovascular death) occurred less frequently with finerenone than with placebo (13.9% vs. 16.9%; HR 0.77, 95% CI 0.60–0.99; p=0.04). The next outcome missed the threshold for significance (see figure below for more).

The treatment effect was primarily driven by kidney outcomes, whereas cardiovascular events were infrequent, with no significant reduction in hospitalization for heart failure or cardiovascular death (HR 0.60, 95% CI 0.27–1.33).

With regard to the exploratory outcomes, finerenone was associated with a sustained reduction in albuminuria, with a 35.4% greater reduction in UACR than placebo at month 6, an effect that persisted throughout follow-up. More than half of the patients achieved a ≥30% reduction in UACR (56.0% vs. 24.4%; OR 3.99). Also, 4 weeks after treatment discontinuation, the decline in eGFR remained smaller with finerenone than with placebo (between-group difference, 2.4 mL/min/1.73 m²), supporting the reversible nature of the initial hemodynamic effect. 

Subgroup analysis of annual rate of change in eGFR from baseline to month 32 showed a consistent benefit of finerenone in all subgroups.

Safety outcome 
Overall adverse events and serious adverse events were similar between groups. As expected, hyperkalemia was more frequent with finerenone (17.0% vs. 13.3%); however, serious hyperkalemia events were uncommon (1.0% vs. 0.6%), no fatal hyperkalemia events occurred, and permanent treatment discontinuation due to hyperkalemia was infrequent (1.5% vs. 0.1%). 

Discussion

The FIND-CKD trial reports that the addition of finerenone in mostly proteinuric, but non-diabetic CKD, on background of RASi results in a meaningful reduction in albuminuria, and a significantly slower decline in GFR slope.

Strengths

This was properly designed, powered, blinded RCT. It included patients in whom we have little data beyond (low quality) RASi and (high quality) flozins to slow down GFR decline, and there remains a high residual risk of kidney failure. the RCT was appropriately enriched with patients who have proteinuria, and used validated outcomes (total GFR slope) while also demonstrating significant effect on UACR which is presumably in the mechanistic pathway. The risk of hyperkalemia was there, but low. Finerenone seems quite safe to use in this setting.

Limitations

• The study population was not fully representative of the overall non-diabetic CKD population. Most participants were men, had significant albuminuria thus limiting the generalizability of the findings. Thus, patients with polycystic kidney disease, or others with non-proteinuric CKD were excluded. It would be harder to show an effect in these as GFR decline is slower - as also the pathophysiology may not be MR related.

• Also excluded were those with lupus nephritis, ANCA-associated vasculitis, and other GNs on active immunosuppression. This is somewhat appropriate as the specific immune pathways need to be targeted rather than these non-specific supportive or palliative therapies. In GN, finerenone (possibly like flozins) might be quite useful in chronic or burnt out disease, not in the active state.

• Other excluded populations include those with GFR < 25 (as is common - but also perhaps too late, or at high risk of hyperkalemia) - but there were also very few Black patients (~ 5%). APOL1 and FSGS is quite common and it would be useful to have strong data in that population.

• Only 17% of participants were receiving an SGLT2 inhibitor, although flozins were allowed during the trial period.  Trial enrollment took place during a transitional period (2021–2023), before flozins became widely established in patients with non-diabetic CKD. Though table S2 includes changes in BB, CCB, statins etc they exclude flozins. We would not expect differential use of flozins, needless to say, and the effect size was consistent in those on flozins and those who were not.

• The relatively low number of cardiovascular events limited the statistical power to detect differences in cardiovascular outcomes. This was a patient population at much higher risk of kidney than CV outcomes.

• Finally, efficacy is only part of the equation. As CKD management increasingly shifts toward a multi-drug approach combining RAS inhibitors, SGLT2 inhibitors, and finerenone, an important unanswered question is whether these therapies will be equally accessible in routine practice. FIND-CKD was not designed to address cost-effectiveness or affordability, but these factors will inevitably influence how widely FIND-CKD findings can be translated into clinical care. 

How do we interpret these results?

Rather than relying on a traditional time-to-event kidney composite, FIND-CKD used total eGFR slope, an endpoint that has gained increasing acceptance as a surrogate marker of CKD progression and is accepted by FDA in DKD and non-DKD populations. This approach is supported by a meta-analysis of 66 randomized trials showing a strong association between treatment effects on eGFR slope and hard kidney outcomes, including sustained ≥57% eGFR decline and kidney failure (Inker LA et al, Nature medicine 2023). 

The GFR slope is very familiar to previous trials. Despite slight differences in the definition of the acute phase (4 months in FIDELIO versus 3 months in FIND-CKD), both studies showed the same biphasic pattern: an early hemodynamic eGFR dip followed by a slower chronic decline in kidney function. In FIND-CKD, finerenone improved the total eGFR slope by 0.7 mL/min/1.73 m²/year, similar to FIDELIO. Together, these findings suggest that this characteristic eGFR trajectory is an intrinsic effect of finerenone rather than a diabetes-specific phenomenon. The overall kidney effect was still far from 'remission in CKD' however.

One of the most consistent findings across studies of finerenone is the early and sustained reduction in albuminuria. In FIDELIO-DKD (Barkis GL et al, N Engl J Med 2020) and FIGARO-DKD (Pitt B et al, N Engl J Med 2021), finerenone reduced UACR by approximately 31% and 32% versus placebo, respectively. Similarly, FIND-CKD demonstrated a 35.4% placebo-corrected reduction in UACR at 6 months, indicating a similar antiproteinuric effect regardless of diabetes status. The clinical trials of finerenone across the spectrum of CKD are shown in the table below:  

Safety data was similar to what we have seen with other trials with finerenone. Hyperkalemia remained the main adverse event associated with finerenone (17.0% vs. 13.3%), but serious events were rare (1.0% vs. 0.6%), no fatal hyperkalemia occurred, and treatment discontinuation due to hyperkalemia remained uncommon (1.5%). 

FIND-CKD attempted to answer one important question—but immediately raised the next one.  If finerenone works in non-diabetic CKD (probably by targeting shared mechanisms of CKD progression), should we still think in terms of adding therapies one by one? Or is the future an upfront combination strategy with a flozins, similar to what is proposed in DKD? (CONFIDENCE trial, Agarwal R et al, New Engl J Med 2025 | NephJC Summary). Additionally, are we now heading into ‘GDMT’ like 4 pillars in non-DKD as well? Apart from RASi, flozins, now finerenone, we also have preliminary data from GLP1RAs on albuminuria (Apperloo et al Nat Med 2025) and GFR (Colhoun et al Nat Med 2024). To reiterate - this mostly applies to proteinuric CKD - in patients who are beyond the initial immunosuppression window. Also read the accompanying NephJC Short for discussion of the GN subgroup and INFINITY analysis.

Lastly - where does this leave spiro-stans? The authors disparagingly quote BARACK-D (Hobbs et al, Nat Med 2024 | NephJC commentary). There has also been the recent SPIRRIT-HF trial (Design: Lund et al EJHF 2025) which did not report a benefit. Though we await final publication, just like BARACK-D, most patients in this trial did not stay on spironolactone. This has been a frustrating few years in futile spironolactone trials, and it is possibly time to move on. Only the onset of ASIs may threaten to dethrone the expansion of finerenone indications across the CKM space.

Conclusion

Finerenone clearly slows progression of GFR decline in (mostly proteinuric) non-diabetic CKD, similar to effects in DKD, and is safe to use. Access and implementation will remain important barriers.

Summary by

Andreea Manolea, Nephrology resident, Clinical Hospital "Dr. C. I. Parhon" Romania

Smita Divyaveer,
Associate Professor,
Department of Nephrology,
PGIMER, Chandigarh, India

Reviewed by 

Akshaya Jayachandran, Brian Rifkin, Swapnil Hiremath,
Cristina Popa, Pallavi Prasad, Milagros Flores

Header created by AI from prompts by Brian Rifkin

Permalink

]]>Beyond Diabetic Kidney Disease: FIND-ing Finerenone's Next RoleVisual AbstractMilagros FloresFri, 17 Jul 2026 21:24:40 +0000http://www.nephjc.com/news/2026/7/17/find-ckd-the-visual-abstract535bcb2fe4b05fe61b320c51:535bd92ae4b0a78001c0e260:6a5a635204458c5f833eec7cIs finerenone ready to move beyond diabetes?

The FIND-CKD phase 3 arrived from the #ERA26 and is the first one to show kidney protection with finerenone in non-diabetic CKD.

The next challenge is no longer whether finerenone belongs in non-diabetic CKD, but how to best combine with the rest of the pillars to maximize protection.

Check out the newest VA created by our intern Raquel Barba

Low-Dose Rivaroxaban and Cardiovascular Events in Advanced Kidney Disease:

The TRACK Randomized Clinical Trial

Sunil V. Badve, PhD; Vlado Perkovic, PhD; Vivekanand Jha, MD; Raja Ramachandran, DM; Lily Mushahar, MMed; Jan Menne, MD; An S. De Vriese, PhD;Patrick Rossignol, PhD; Adrien Flahault, MD; Maha Al Ammari, MPH; Habib Skhiri, PhD; Michael Walsh, PhD; David Collister, PhD; Adrian Liew, MD;Laurent Billot, MRes; Severine Bompoint, BSc; Anthony Devaux, PhD; Min Jun, PhD; Enmoore Lin, PhD; Aline Ramos da Cruz, BBA; Jeffrey T. Ha, PhD;John W. Eikelboom, MBBS; Ahmed Shaman, PhD; Meg J. Jardine, PhD; Shilpanjali Jesudason, PhD; Muh Geot Wong, PhD; Craig S. Anderson, PhD;Amit X. Garg, PhD; Hiddo J. L. Heerspink, PhD; Helen Monaghan, BSc; Anushka Patel, PhD; Patrick B. Mark, PhD; David C. Wheeler, MD; Jicheng Lv, PhD;Li Zuo, PhD; Helen Pilmore, MD; Martin Gallagher, PhD; for the TRACK Trial Investigators

PMID: 42240165

DOI: 10.1001/jama.2026.9379

Introduction

The concept of cardiovascular nihilism in advanced kidney disease stems from a paradox: heart disease is the leading killer in end-stage kidney disease, yet standard cardiological interventions consistently fail. While statin therapy reduces events in the general population, the AURORA (Fellstrom BC et al, NEJM, 2007) and 4D (Wanner C et al, NEJM, 2005) trials proved that aggressively slashing cholesterol was associated with zero reduction in cardiovascular death or strokes on dialysis. Similarly, the ISCHEMIA-CKD trial shattered the belief that mechanical intervention could fix the pathophysiology, revealing that immediate invasive revascularization failed to lower death rates and instead triggered a four fold increase in strokes (Bangalore S et al, NEJM, 2019| NephJC summary). In the uremic milieu, traditional lipid-driven and macrovascular models of heart disease stop applying. 

Driven by these failures, the scientific narrative shifted toward targeting hypercoagulability. Landmark trials in broader populations offered hope: COMPASS (Eikelboom JW et al, NEJM, 2017) demonstrated that adding low-dose rivaroxaban to aspirin significantly reduced major adverse cardiovascular events (MACE), with subanalyses hinting at greater benefits in mild-to moderate CKD (Fox KAA et al, JACC, 2019). This was reinforced by the VOYAGER PAD trial (Bonaca MP et al, NEJM, 2020), which proved that dual pathway inhibition decreased ischemic limb and cardiovascular events after peripheral revascularization, and the ATLAS TIMI ACS 51 trial, which showed a reduction in composite of cardiovascular death, myocardial infarction and stroke along with a significant reduction in risk of secondary outcome of stent thrombosis post-acute coronary syndrome (Mega JL et al, NEJM, 2012). 

However, these practice-changing trials systematically excluded patients with stage 4 or 5 CKD and those on dialysis. Clinicians were caught in a dangerous blind spot: the KDIGO heat map definitively illustrates that as eGFR drops and albuminuria rises, CV risk skyrockets into the deepest red zone (KDIGO CKD guidelines 2024I NephJC Summary). 

Yet, advanced kidney disease simultaneously induces severe platelet dysfunction and elevates the bleeding risk(Genovesi S et al, NDT, 2024). Doctors were forced to extrapolate data from healthier cohorts, flying blind without knowing if the ischemic protection seen in COMPASS, VOYAGER, and ATLAS would be wiped out by devastating hemorrhagic complications. 

The evidence gap justifies the need for the TRACK trial, which puts low-dose rivaroxaban to the test in patients with stage  4 or 5 CKD, including those on dialysis. 

The Study 

Study design

Investigator-initiated, randomized, quadruple-blind, placebo-controlled trial

Methods

The study was conducted at 90 centers in 12 countries (Australia, Belgium, Canada, France, Germany, India, Malaysia, Nepal, Saudi Arabia, Singapore, Taiwan, and Tunisia). The enrollment of patients was from January 18, 2021, to July 31, 2025. The trial was stopped early on August 7, 2025, based on the recommendation of the data and safety monitoring board, and final follow-up occurred on October 30, 2025.

Funding

The major funding was provided by the National Health and Medical Research Council (NHMRC) Australia project grant, and additional funding was supplied by other participating centers. All funding was approved by the International Coordinating Centre for the George Institute of Global Health. Rivaroxaban and placebo tablets were provided by Bayer AG free of cost through the Investigator-Initiated Research Support Scheme. The funders had no role in the design and conduct of the study, the data collection, management, analysis, or interpretation of the data, the preparation or review of the manuscript, or the decision to submit for publication. 

Study population

Eligible participants were adults aged 18 or older with advanced CKD, defined as CKD stage 4 or 5, including those with dialysis-dependent kidney failure. All participants were required to have at least one additional cardiovascular risk factor, which could include established coronary artery disease, non-hemorrhagic non-lacunar stroke, peripheral artery disease, diabetes, or age 65 or older. Atrial fibrillation (AF) was not an inclusion criterion, but an exclusion criterion (see below). 

Key exclusion criteria included mechanical/prosthetic heart valve, anticoagulation indication/contraindication, high bleeding risk or active bleeding, P2Y12 or phosphodiesterase inhibitor use, prior hemorrhagic/lacunar stroke, severe heart failure (EF < 30% or NYHA III/IV), hemoglobin < 90 g/L, thrombocytopenia, uncontrolled hypertension or a functioning kidney transplant. Patients with AF were excluded unless anticoagulation was not indicated.

Run-In Phase and Randomisation 

Eligible patients were given a placebo for 21 days; the patients who were deemed to have < 80% adherence were excluded from the study (NephTrials on run-in periods for more). Randomization was performed through a web-based password-protected system, using a covariate-balancing adaptive allocation algorithm. The algorithm was designed to minimize between-group imbalances across four pre-specified stratification variables: diabetes, CKD stage, dialysis-dependency and aspirin use. Participants were allocated in a 1:1 ratio to receive either rivaroxaban 2.5 mg twice a day or matching placebo. All participants, investigators, study coordinators, site pharmacists, treating physicians, and outcome assessors were blinded to treatment allocation throughout the trial. 

Outcomes

The primary efficacy outcome was a MACE composite of cardiovascular death, nonfatal MI, nonfatal stroke, or nonfatal peripheral artery disease event. The outcome was assessed by two investigators who were blinded; one was an on-site investigator and another member of the independent clinical outcome review committee (CORC).

Secondary outcomes included all-cause mortality; individual components of the primary composite outcome; a composite of CV death, MI, or stroke; a composite of infarction, stroke, or peripheral artery disease event; and venous thromboembolism. All deaths were adjudicated similar to the primary outcome.

The net clinical benefit outcome, initially classified as a secondary outcome, (later converted to a non secondary outcome before data review) was defined as a composite of CV death,nonfatal MI,non fatal stroke, a nonfatal peripheral artery disease event, fatal bleeding, or symptomatic bleeding into a critical area or organ. 

Tertiary outcomes were thrombosis of dialysis vascular access among participants with AVF or AV graft and health-related quality of life (QOL) using EuroQ-5D-5L and cost-effectiveness. Cost effectiveness and EQ5D-5L were reported after getting complete data.

Safety outcomes were major bleeding (modified iSTH criteria: fatal bleeding, critical organ bleeding, bleeding requiring reoperation, or bleeding leading to hospitalization) and gastrointestinal bleeding. All bleeding was adjudicated by the CORC.

Additional data: Race and ethnicity were collected from the participants, as Asians are more prone to bleeding (Liang Y et al, JACC Asia 2024) and other comorbidity data were collected from medical records.

Statistical considerations

This trial planned to enroll 1900 participants over 3 years, with a follow-up period of 5 years. Assuming a control group primary event rate of 10 per 100 person-years, this sample size would provide a 90% power to detect a 25% relative risk reduction (hazard ratio 0.75) with a two-sided type I error rate of 5%. It was planned to conduct two equally spaced interim analyses with Haybittle-Peto efficacy boundaries, corresponding to a Z value of 3 standard deviations at each interim analysis. This yielded two-sided significance levels of 0.27% for each interim analysis and a final significance level of 4.82%. A total of 515 primary outcome events were required to achieve 90% power, and this number of events provided 80% power with a 22% relative risk reduction (HR 0.78).

All analyses were conducted according to the intention-to-treat principle. The primary outcome was analyzed as the time from randomization to the first event using a shared frailty Cox proportional hazard model with a random site effect, adjusted for the three stratification variables: 1) baseline aspirin use, 2) dialysis dependency and 3) diabetes. The final significance level was set at a 2-sided alpha of 4.86%, accounting for 1 interim analysis performed after 181 events. 

The secondary outcome was also analyzed using the same Cox model. To control the family-wise error rate at 5% across the five key secondary outcomes—all-cause death and the four individual components of the primary composite—a Holm-Šídák step-down approach was applied. For the remaining secondary outcomes, no formal hypothesis testing was performed; only point estimates and 95% confidence intervals were reported. The proportion of adverse events, such as bleeding and severe adverse events, was analyzed using the Fisher exact test.

Results

Of 5,979 prescreened patients, 1,777 met initial eligibility criteria. After the run-in phase, 1,463 patients were randomized. Median follow-up was 1.7 years, with 93% completing the follow-up assessment. Loss to follow-up was minimal and similar in both groups (4% lost contact, 2.3% withdrew consent).

The enrolled cohort was predominantly male (70%), had a mean age of 63 years, and nearly half (49%) had dialysis-dependent kidney failure. Most patients had diabetes (78%) and almost half (47%) were aged >65 years. Only 18% had established coronary artery disease and 6.6% had peripheral artery disease. Roughly two-thirds of the cohort met inclusion criteria based solely on diabetes or age, rather than prior CV disease. Aspirin use was identical between the arms (approximately 46%). Geographic representation was heavily concentrated in India and Malaysia (41.5% and 21%, respectively) (eTable 3 from Badve S et al, JAMA 2026)

The trial was stopped early (Aug 2025), after 254 primary outcome events had occurred (49.3% of the 515 anticipated events). The Data and Safety Board recommended termination due to concerns about net harm. 

Primary outcome

The primary composite outcome occurred in 164 patients (23%) receiving rivaroxaban vs 151 (21%) receiving placebo. This translated to 13 vs 11.8 events per 100 person-years, with a hazard ratio of 1.09 (95% CI 0.87-1.36). Rivaroxaban showed no benefit.

Subgroup analyses confirmed uniformity of the null effect. Across age, sex, country, aspirin use, diabetes, CKD stage, and smoking history, no subgroups showed benefit. Aspirin use at baseline did not modify the treatment effect (HR 1.06 among aspirin users vs 1.11 among non-users).

Mortality 

All-cause mortality showed no difference. In the rivaroxaban group, 25.6% died compared with 23% in placebo- 13.9 versus 12.4 events per 100 person-years (HR 1.14, 95% CI, 0.92-1.40). There were 18 more deaths in the rivaroxaban group overall. 

The competing risk analysis examined whether the high non-cardiovascular death burden distorted the primary outcome estimate. Accounting for the non-cardiovascular death as a competing event yielded a subdistribution HR of 1.08 (95% CI 9.86-1.34) for the primary outcome. As regards the cumulative mortality (eFigure 1) at 48 months, the rivaroxaban curve remained above placebo (51% vs 47%)

The breakdown of mortality causes (eTable 4 below) showed that 60% were cardiovascular deaths in the rivaroxaban arm. Sudden cardiac death, occurring outside the 30-day window after acute MI, dominated both groups (40% in the rivaroxaban arm vs 45% in the placebo arm). Infections were the leading non-CV cause, followed by renal failure and other organ-system failures. 

Secondary outcomes

Individual components of the primary outcome were comparable between the two arms. Additionally, when the events were recombined into secondary composite endpoints, rivaroxaban showed no advantage across any configuration (Table 2 below).

Venous thromboembolism

Fewer VTE events occurred on rivaroxaban than on placebo (0.6% vs 1.9%), with aHR 0.29, 95%CI 0.09-0.87.

Among patients on hemodialysis, access thrombosis occurred more frequently on rivaroxaban (7.6% vs 5.4%), but the result was not significant (HR 1.44, 95% CI, 0.75-2.74).

Safety outcomes: major bleedings

Major bleeding events occurred more in the rivaroxaban group (8.8% vs 6%), with a HR of 1.51 (95% CI 1.02-2.22).

The sites and characteristics of major bleeding are detailed in eTable5 (below). The gastrointestinal tract was the most common site. These events included overt bleeding confirmed by endoscopy, overt bleeding of unknown origin, and presumed occult bleeding with hemoglobin decrease.

Intracranial and intraocular bleeding were numerically higher in the rivaroxaban group. Fatal bleeding occurred in 1 patient on rivaroxaban and 2 patients in the placebo group.

Bleeding by age

Major bleeding risk showed significant age interaction (eFigure 5 below). In patients ≥65 years, major bleeding occurred in 10% on rivaroxaban vs 5.6% on placebo. In those <65 years, the rates were lower (rivaroxaban 7.3% vs 6.4% in placebo) and comparable.

Net clinical benefit 

The net clinical benefit outcome (composite of cardiovascular events and fatal/critical bleeding events) was comparable between the two arms  (24.2% rivaroxaban vs 21.3% placebo, HR 1.14, 95% CI 0.92-1.41- details in Table 3 above)

Serious adverse events 

They were similar between the groups. Infections were the most common (17.6% rivaroxaban vs 18.1% placebo). Overall, 50.5% of rivaroxaban patients and 45.8% of placebo patients experienced ≥ 1 serious adverse event.

Discussion

TRACK randomized more than 1400 patients with CKD stage 4, 5 or dialysis-dependent kidney failure, enriched for CAD, prior stroke, PAD, diabetes, or age 65 and older, to rivaroxaban 2.5 mg daily or placebo. The question was whether the “vascular dose” anticoagulation in COMPASS (Eikelboom JW et al, NEJM, 2017) and VOYAGER PAD (Bonaca MP et al, NEJM, 2020) would be of benefit to a population with advanced CKDs, and… it wasn’t. The primary MACE composite (CV death, MI, stroke, and PAD event) was not significantly different in rivaroxaban vs placebo (22.6% vs 20.7%). What’s even worse, major bleeding was higher on rivaroxaban, so this was not a benign intervention. The DSMB stopped the trial early after a post hoc conditional-power calculation put the chance of eventual benefit at 16% with a HR of 0.78 even if the trial continued till the pre-specified 515 primary events. Net clinical benefit, the composite meant to weigh efficacy against safety, was also negative. The study excluded patients with a clear indication for anticoagulation; rather, it tested whether across the board anticoagulation was worthwhile in high-risk CKD patients whose vasculature has already been modified by uremia. 

Context

The vascular-dose approach isn’t new. ATLAS TIMI 51 (Mega JL et al, NEJM, 2012) identified 2.5 mg twice daily as the dose that balanced antithrombotic benefit against bleeding, and a pooled analysis of seven trials across over 45,000 patients with CAD or PAD found low-dose rivaroxaban added to antiplatelet therapy reduced CV events (HR 0.86, 95% CI 0.78-0.94) and ischemic stroke (Bucci T et al, J Clin Med, 2024). Trials like COMPASS (Eikelboom JW et al, NEJM, 2017) and VOYAGER PAD (Bonaca MP et al, NEJM, 2020) enrolled patients with established atherosclerotic disease. TRACK’s population looked different: only ~19% had coronary disease, ~ 6% had prior stroke, ~6% had PAD, and 57% qualified on the strength of a single risk factor, usually age or diabetes alone (eTable 3). Either way, this looks closer to a primary-prevention population that the trials TRACK was built to extend, and it points to a precedent worth naming directly: aspirin’s own run of primary-prevention trials a decade earlier. Three trials tested low-dose aspirin for primary cardiovascular prevention in similarly enriched, disease-free populations. ASCEND, in adults with diabetes and no prior cardiovascular disease, found a 12% relative reduction in vascular events “largely counterbalanced” by a 29% increase in major bleeding, in the authors’ own phrasing (The ASCEND Study Collaborative Group, NEJM, 2018). ASPREE, in adults 70 and older without prior cardiovascular disease, found daily aspirin gave no cardiovascular benefit and significantly increased major hemorrhage (McNeil JJ et al, NEJM, 2018). ARRIVE, in moderate-risk adults without diabetes, found that aspirin had neither benefit nor safety (Gaziano JM et al, Lancet, 2018). None of these 3 trials used rivaroxaban or any other DOAC. But TRACK’s age and diabetes-heavy cohort sit roughly in the territory they mapped, and a different antithrombotic drug, tested there again, produced the same shape of result: no measurable protection and hemorrhagic side effects.

Guidelines

KDIGO’s 2024 CKD guideline draws a sharp line between two questions it treats differently. (KDIGO CKD Work group, Kidney Int, 2024| NephJC summary). For secondary prevention- aspirin in people with established CVD- it gives a strong graded recommendation (1C), built on Antithrombotic Trialists’ Collaboration meta-analysis: a roughly 19% relative reduction in serious vascular events (RR 0.82, 95% CI 0.75-0.87) (ATT collaboration, Lancet, 2009). For primary prevention, aspirin in people at high risk without established disease (the category TRACK mostly falls into), KDIGO issues no graded recommendation at all, only a call for more research. Its reasoning was that ASCEND, ASPREE, and ARRIVE all showed bleeding harm counterbalancing aspirin’s CV benefit. A CKD-specific Cochrane meta-analysis of more than 40,000 participants in antiplatelet versus placebo trials found similar results: MI reduced (RR 0.88, 95% CI 0.79-0.99), major bleeding increased (RR 1.35, 95% CI 1.10-1.65)- though that analysis didn’t separate primary from secondary prevention (Natale P et al, Cochrane Database Sys Rev, 2022).

The answer is “No”

Advanced CKD is often described as a state where clotting and bleeding risk rise together. The mechanism implies uremic toxins impair platelet adhesion and drive endothelial nitric oxide and prostacyclin production that further suppresses clotting, leaving CKD patients roughly 1.5 times more likely to bleed than people with normal kidney function, independent of antithrombotic use (Ocak G et al, J Thromb Haemost, 2018). Hemodialysis adds a further 1.5-fold risk over peritoneal dialysis (van Eck van der Sluijs A et al, NDT, 2021), and roughly one in seven older patients on dialysis has a major bleed within 3 years of initiation (Sood MM et al, Kidney Int, 2013). TRACK asked a fair question and got a clear answer. Rivaroxaban didn’t protect against clots, and it increased bleeding~ extra 1.7 events per 100 patients-years(Table 3). The only positive, VTE (HR 0.29), is more a pharmacological note- with what clinical significance? The bleeding risk fell hardest on patients 65 and older (interaction P = .03), the group most likely to get this drug in clinical practice.

On a less nihilistic view, a negative, well-done trial like TRACK is an answer. The mechanism was plausible, the drug worked in other settings, and the patients carried a real risk- reason enough to run the trial. The result, rigorously tested and honestly reported, is worth as much as a positive one: it tells clinicians where not to engage in speculative benefit when there is significant risk of harm. Aspirin’s primary prevention trials reached a similar answer a decade ago, in a different population, with a different drug. TRACK adds a boundary, drawn a second time in roughly the same place - here is where “thin the blood, protect the vessel" stops working.

Strengths

TRACK trial was a dedicated, investigator-initiated, randomized, double-blind, placebo-controlled trial across 90 centers in 12 countries, enrolling adults with CKD stage 4 or 5, including dialysis. Randomization was stratified and blinded at all relevant levels, with a covariate-adaptive algorithm balancing key prognostic factors. Prespecified outcomes were adjudicated by an independent clinical outcomes committee (essential when effects are modest and noise is abundant).

Limitations

The trial was powered for 1900 participants and 515 primary events, but it stopped early at 1463 randomized patients based on DSMB recommendation for futility (conditional power ~ 16% for HR 0.78), not a prespecified efficacy boundary. Follow-up was a median of 1.7 years, shorter than planned, and formal tablet counts were disrupted by COVID-19, though self-reported adherence was collected. Permanent discontinuation was 28%, which likely attenuated any true effect. The confidence interval for the primary outcome (0.87-1.36) rules out a large benefit but leaves modest effects uncertain. 

About generalizability: there were no US patients and almost no Black patients. Regional variation was observed (Australia/Europe/Canada- for primary outcomes: HR 1.82, 95% CI 1.02-3.22), but interaction tests were not significant, and this was an unadjusted subgroup comparison, hypothesis-generating only.  Nearly 40% of all-cause deaths in both arms were sudden cardiac death, a category that no anticoagulant was likely to prevent, which sets a biological ceiling on what this class could achieve. Minor bleeding was not collected, and quality-of-life data are deferred to a separate report. While 46% of patients were on aspirin at baseline, the trial did not mandate aspirin nor test the fixed combination- unlike COMPASS and VOYAGE PAD. 

Conclusion

TRACK doesn’t solve advanced CKD’s oldest problem- a body where treating one risk worsens another, with no drug built for that narrow ground. It settles the case against rivaroxaban. Unfortunately, push clotting down, and bleeding rises to meet it: the scales of hypercoagulability and excessive bleeding are never completely balanced. The boulder keeps returning to the bottom by a different route, and no antithrombotic strategy tried so far has gone over the hill instead of around it. That may be the shape of the answer for this population: something closer to a limit than a solution, since one antithrombotic therapy alone was always going to run up against the complex CKD pathophysiology. 

Summary by, 

Cristina Popa

and Sai Vani Yellampalli

Reviewed by Brian Rifkin, Pallavi Prasad, Swapnil Hiremath

Telitacicept for IgA Nephropathy - Interim Analysis of a Phase 3 Trial

Jicheng Lv, M.D., Lijun Liu, M.D., Wenxiang Wang, Ph.D., Xinyue Wang, Ph.D.,Qing Zuraw, M.D., Vlado Perkovic, M.D., Ph.D., Jianmin Fang, Ph.D.,and Hong Zhang, M.D., Ph.D., for the TELIGAN Investigators

PMID:  42127391

DOI:10.1056/nejmoa2514415

Why was this study needed?

We’re at quite an interesting juncture in IgAN therapeutics. Five years ago, we were still debating what dose of steroids should remain in the conversation after TESTING | (NephJC Summary). Today, the challenge is almost the opposite: there are so many promising therapies that reduce proteinuria, and the question is which drug to select and which pathway to target for which patient.

We’ve seen targeted-release budesonide (NefIgArd) | (NephJC Summary) focus on the gut mucosal immune system. Complement inhibition entered the scene with Iptacopan (APPLAUSE - IgAN) | (NephJC). More recently, APRIL-directed therapies such as sibeprenlimab (VISIONARY) | (NephJC Summary) and atacicept (which is dual BAFF-APRIL, ORIGIN) | (NephJC Summary | Podcast) have delivered striking reductions in proteinuria. Telitacicept now joins this increasingly crowded field,  sharing the distinct biological proposition of atacicept: simultaneous blockade of both BAFF and APRIL rather than targeting either pathway alone.  

With a slew of drugs already approved for IGAN, why does this trial matter?

BAFF and APRIL signaling sit in the space of B-cell survival, plasma-cell maturation, and immunoglobulin production. Elevated levels of both cytokines have been linked to IgAN activity. Telitacicept is essentially a soluble TACI-Fc fusion protein that acts as a decoy receptor, binding both BAFF and APRIL simultaneously. Theoretically, this should reduce production of pathogenic galactose-deficient IgA1 and downstream immune complex formation. The TELIGAN trial asks a direct question - exactly the same as with ORIGIN (atacicept): can dual BAFF/APRIL inhibition meaningfully alter disease activity (in the form of proteinuria and GFR stabilization) in high-risk IgAN? This question has been asked and answered for various groups of drugs in IgAN. After atacicept, do we need another -tacicept? Why not - the more the merrier! So, hit me baby one more time.

How was the study done and what did it find?

TELIGAN is a phase 3, multicentre, double-blind, placebo-controlled study conducted across 72 centres in China. They included adults with biopsy-proven IgAN, UPCR ≥0.5 g/g, eGFR ≥30 mL/min/1.73 m², and persistent proteinuria despite ≥12 weeks of maximally tolerated RAS blockade. This paper represents a prespecified interim analysis (stage A) focused on proteinuria at 39 weeks, while the ongoing stage B will evaluate longer-term kidney function outcomes through 104 weeks. The trial is being funded by Remegen (the company manufacturing telitacicept).  If you have read any of the IgAN trials linked above, feel free to skip the methods, they are almost a CTRL-C/CTRL-V of each other. 

A total of 318 adults with biopsy-proven IgAN were randomized 1:1 to weekly subcutaneous telitacicept 240 mg or placebo on a background of standard of care. Patients had persistent proteinuria despite optimized Renin Angiotensin System(RAS) blockade, eGFR ≥30 ml/min/1.73m², and were not receiving glucocorticoids or other immunosuppressive therapies. About one-third were already flozinated, reflecting contemporary practice. Unlike RAS blockade, flozination was allowed, but not required as part of standard of care. 

The primary endpoint for this analysis was reduction in 24-hour urinary protein-creatinine ratio at week 39.

What were the results?

The mean age was 38 years, with women comprising just over 50% of the study population. Mean eGFR was around 75 mL/min/1.73 m², and the median baseline 24-hour urinary protein-to-creatinine ratio was 1.26 g/g. Approximately 1/3rd of participants had eGFR < 60 mL/min/1.73 m². Histologic characteristics showed  M1 lesions present in ~79% and crescents (C1/C2) in ~45% of patients with available biopsy data. Around 70% had hematuria. All patients were receiving background RAS blockade at randomization. 

At 39 weeks, proteinuria fell by 58.9% with telitacicept compared with only 8.8% with placebo, with a relative treatment effect of −55% (95% CI −61.3 to −47.6; P<0.001). Separation from placebo appeared as early as week 4 and continued to widen over time. Equally striking was the proportion achieving a UPCR below 0.8 g/g: 61% with telitacicept versus only 19.5% with placebo.  

Mean eGFR was essentially preserved in the telitacicept group (−1.0%) while declining in the placebo arm (−7.7%) over 39 weeks. Likewise, fewer patients experienced a ≥30% eGFR decline (6.3% vs 27%).  Interestingly - in China they are allowed to present GFR data at this stage (you may note that the ORIGIN atacicept phase 3 trial suppressed that - the FDA only allows that to be analysed at 2 years).

The clinical response was also accompanied by pharmacodynamic effects. Circulating CD19-positive B cells fell by nearly 50%, while immunoglobulin levels declined substantially. Serum IgA levels decreased by approximately 60%, supporting the notion that the drug is engaging the intended disease pathway.

The overall adverse event rate was higher with telitacicept (89% vs 79%), largely driven by injection-site reactions and expected immunoglobulin reductions. Serious adverse events were actually less frequent with Telitacicept (2.5% vs 8.2%). Infection rates were broadly similar between groups. No deaths occurred. With regard to the HBV data, among 68 patients with latent hepatitis B infection, no reactivation occurred in those receiving Telitacicept during the 39-week study period. For clinicians practicing in Asia, where latent HBV is common, this is clinically relevant, although the sample size and follow up remains modest at this time.

WHAT ARE THE IMPLICATIONS?

TELIGAN shows that telitacicept produces one of the greatest reductions in proteinuria seen in contemporary IgAN therapeutics - similar to what was seen with sibeprenlimab and atacicept. The effect is consistent across subgroups, appears within four weeks, and is accompanied by reductions in circulating B cells, immunoglobulins and serum IgA, engaging a central pathogenic pathway in IgA nephropathy. This is also supported by the fall in hematuria (post-hoc analysis) from ~71% to ~21% with telitacicept and essentially unchanged with placebo. This matters because proteinuria can improve through hemodynamic mechanisms, but hematuria is often considered a marker of active glomerular inflammation, and since telitacicept improves both, it could be a disease-modifying effect. Given all these strengths, the enthusiasm generated by TELIGAN is understandable 

This remains a short term proteinuria reduction data report however (like all recent IgAN trials). The kidney function data are encouraging but shouldn’t be overinterpreted. The apparent eGFR preservation and lower frequency of a 30% eGFR fall favour telitacicept, but these were secondary analyses, the study was not powered for kidney outcomes and is just an interesting sidenote - we have to await 2 year GFR data (same as with other IgAN new therapeutics).

When we compare it with other phase 3 IgAN trials, we get an interesting context. Unlike ORIGIN, VISIONARY and APPLAUSE-IgAN, which enrolled multinational cohorts, TELIGAN was conducted exclusively in a Chinese population. Patients in TELIGAN were randomized a median of more than four years after biopsy, compared with around 2.5 years in ORIGIN and 1.5 years in VISIONARY and APPLAUSE-IgAN. 

Baseline proteinuria was among the lowest across these studies (median UPCR ~1.3 g/g in TELIGAN and VISIONARY,  ~1.9 g/g in APPLAUSE-IgAN, and ~2.2 g/g in ORIGIN). Also, this was also a well-treated cohort - with baseline standard of care similar to other IgAN trials (with supportive/palliative therapies). All patients received optimized RAS blockade and 35% were already receiving SGLT2 inhibitors, compared with 50% in ORIGIN, 40% in VISIONARY and 12% in APPLAUSE-IgAN. This suggests that the study population consisted of patients with established, persistent IgAN.

Again, with established disease and already on baseline RASi/ some SGLT2i,  the placebo arm still achieved an 8.8% fall in proteinuria. The numerically greater decrease in proteinuria with telitacicept than all other trials cannot be signal - comparing across studies with heterogeneous populations is fraught with uncertainty.

At present, it is difficult to know whether dual BAFF/APRIL blockade represents a genuine therapeutic advance over selective APRIL inhibition or simply another effective approach targeting the same biological pathway. As direct comparative trials are unlikely until several years from now, this question will probably be answered gradually through accumulating clinical experience and longer-term outcome data. While telitacicept may (or may not) make it to European or North American populations - would Chinese pharmaceutical companies be interested in making these drugs available in other parts of the world - especially across Asia where IgAN is so prevalent and where only steroids are available so far? Will they make it cheaper than sibeprenlimab, atacicept et al? One can hope. 

BOTTOMLINE?

TELIGAN is a well designed phase 3 study that delivers a remarkably consistent and biologically plausible treatment signal. The magnitude of proteinuria reduction is among the largest reported in IgAN, seems to have a convincing target engagement, and the short term safety profile appears acceptable

By

Akshaya Jayachandran

Reviewed by 

Brian Rifkin, Cristina Popa and Swapnil Hiremath

A secondary analysis of the TESTING trial predicted individual patient response to corticosteroid treatment in IgA nephropathy

Mark Canney 1, Sana Shan 2, Lee Er 3, Laurent Billot 2, Jialin Han 4, Muh Geot Wong 5, Helen Monaghan 2, Michelle Hladunewich 6, Lai Seong Hooi 7, Vivek Jha 8, Jicheng Lv 9, Vlado Perkovic 2, Hong Zhang 9, Daniel C Cattran 10, Sean J Barbour 11; TESTING trial steering committee

PMID: 41571096

Why was this study done? 

The TESTING trial (Lv et al, JAMA 2022| NephJC summary) is the largest RCT of corticosteroids in IgA nephropathy (IgA) with 503 adults with proteinuria ≥1 g/d on optimized RAS blockade. Oral methylprednisolone halved the risk of the composite kidney endpoint (HR 0.53, 95% CI 0.39–0.72), with an absolute risk reduction of 16.1% at 4 years. These results were quite impressive. However, the study also showed significant toxicity with corticosteroids, which led to a mid-trial protocol revision from full dose (0.6–0.8 mg/kg/d) to a reduced-dose regimen (0.4 mg/kg/d). The clinical conundrum is the effectiveness of a treatment in an IgAN population with heterogeneous risk and variable drug tolerability. The finding of 47% relative risk reduction is an average across many individuals who likely differ substantially in their clinical risk and disease progression patterns. 

The principal question this post hoc analysis of TESTING is: Can we predict, for a given patient, what their individual absolute risk reduction from methylprednisone would actually be?

Now you may ask, why does this matter when we have so many other options for treatment of IgAN, when even more treatment options are appearing rapidly? Not surprisingly, because in many parts of the world steroids are all you have (apart from RASi, flozins, and possibly MRAs). Even if new therapies become available, they will be very expensive and less accessible, at least in the near future. Second, these methods are cool - exploring the heterogeneity of effects to draw some tentative conclusions for shared decision making. They apply to other intervention in different settings, so worthwhile trying to understand this methodology. 

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

This was a secondary analysis of 483 of the 503 TESTING participants (20 excluded for missing MEST-C or covariate data). The analytic approach followed the PATH statement in four steps:

• Cox proportional hazards model with backward elimination (P < 0.2) to select main-effect treatment effect modifiers

• Selected variables entered with treatment exposure and all treatment × variable interaction terms (no selection on interactions)

• Ridge regression applied throughout

• For each patient, predicted 4-year absolute risk generated under both treatment scenarios; the difference = individual ARR

The primary outcome was the same as TESTING: ≥40% eGFR decline, kidney failure, or death from kidney disease.

Study population

After backward elimination, the following variables were selected as main effects and forced into the model with interaction terms: eGFR, age, proteinuria, RAAS blockade dose, ethnicity, time from biopsy, systolic blood pressure, sex, BMI, T-score, and C-score (MEST-C).

Results 

This figure shows the point estimates for each interaction term (treatment multiplied by each variable) on the hazard ratio scale. Variables to the right of the dashed line are associated with a better response to methylprednisolone; variables to the left with a worse response.

Better response to methylprednisolone:

• Higher eGFR

• Male sex

• MEST-C score: T1/T2 (vs T0)

• MEST-C score: C1/C2 (vs C0)

Worse response to methylprednisolone:

• Higher proteinuria

• Higher RASi dose

• Chinese ethnicity

Minimal impact: Age, SBP, BMI, time from biopsy to enrollment. 

Of note, time from biopsy to enrollment had minimal impact on modifying treatment response, likely in part because most patients enrolled relatively soon after biopsy (median 5 months), limiting variability in this predictor.

This is the distribution of individual-level predicted 4-year absolute risk reduction across the entire cohort. The colors represent tertiles - lightest blue is the lowest tertile, darkest blue is the highest. So, what are tertiles? The model generates a predicted absolute risk reduction (ARR) for every patient. Then all 483 patients are ranked from lowest to highest predicted benefit and divided into three equal groups (tertiles) of 161. The lowest tertile contains patients predicted to benefit least, some are predicted to have no benefit or even harm. The highest tertile contains patients predicted to benefit most.

The key finding here is the broad spread of the predicted ARR ranging from about −10% to +40%, compared to the average of 16.1% (in the trial). This illustrates exactly why the average treatment effect is insufficient for individual decision-making. Patients at the left tail may experience no benefit or even harm from treatment, while those at the right tail have substantial benefit. Wouldn’t that be nice to know before you prescribe a given treatment?

This is the key clinical finding. The model was used to split patients into two groups: those with predicted ARR of 10% or less, and those with predicted ARR greater than 10%. Among patients predicted to have ARR greater than 10%, the observed ARR was 24% (95% CI 13% to 36%). That's a large, statistically significant benefit. Among patients predicted to have an ARR of 10% or less, the observed ARR was −5% (95% CI −25% to 15%). That's no benefit and the point estimate actually suggests possible harm.

Overall, the model successfully identifies a subgroup with no observed benefit and a subgroup with substantial benefit. The 10% threshold was chosen as a clinically meaningful treatment threshold, but this can be adjusted based on individual patient values and risk tolerance. For example, a risk-averse patient might want a higher threshold, while a risk-accepting patient might accept a lower one.

Two key metric performances were used: 

• Restricted mean survival time (RMST): If only patients with predicted ARR >10% were treated (targeted treatment policy), the RMST was 1,194 event-free days, compared with 1,028 days under random allocation. This gives 166 additional event-free days with targeted treatment.

• C-statistic for benefit: the C-statistic for benefit was 0.63 (95% CI 0.56 to 0.70). As discussed earlier, this should not be compared to the 0.7–0.8 threshold for prognostic models. The C-for-benefit predicts an inherently unobservable quantity (the individual treatment effect) which introduces fundamental noise. Values of 0.55–0.65 are considered meaningful. For context, the SYNTAX Score II, one of the best-validated benefit prediction models in all of cardiology, achieved a C-for-benefit of about 0.59. So, 0.63 is actually quite good for this type of metric. 

Lastly, calibration showed good agreement between predicted and observed ARR, with mild overestimation at the low end and underestimation at the high end. Sensitivity analyses across reduced and full-dose subgroups were consistent. eGFR slope analyses selected similar predictors. Notably, adverse events did not increase across predicted benefit tertiles, and patients predicted to benefit most did not experience more toxicity.

The PRED-IgAN tool

A web-based calculator is publicly available at www.gnpredict.com. Inputs include eGFR, age, proteinuria, RAAS blockade dose, ethnicity, time from biopsy, systolic BP, sex, BMI, T-score, and C-score. Output is a predicted 4-year ARR from methylprednisolone, which can be used with any clinician-chosen threshold in shared decision-making.

Here are two examples on PRED-IgAN in which the first patient is unlikely to benefit (low ARR) and the second patient is likely to benefit (high ARR).

What are the implications? 

This is the first individual treatment effect prediction tool for immunosuppression in IgAN, and it addresses a real clinical need. PRED-IgAN could help frame the risk-benefit discussion for patients, particularly relevant given that most patients likely fall in a zone of genuine equipoise for corticosteroids.

That said, important caveats deserve emphasis. The absence of external validation is the most significant limitation and, as the authors acknowledge, a second large corticosteroid RCT in IgAN is unlikely. The cohort is 74% Chinese, limiting generalizability to other populations. Overfitting risk persists despite ridge regression, given 483 patients and 174 events with many interaction terms. MEST-C scoring relied on local pathology reports without central review. Finally, the model predicts benefit, not individual-level harm so adverse event analysis remains at the group level.

Contextually, KDIGO 2025 now positions targeted-release budesonide (Nefecon) as first-line immunosuppression in IgAN, with systemic corticosteroids reserved for settings where budesonide is unavailable (grade 2B).

Yet corticosteroids remain the only accessible immunosuppressive option across much of the world.

For those contexts, PRED-IgAN offers a practical tool to optimize who receives treatment and who might reasonably be spared. One should remember that claims of Nefecon’s superiority or safety over prednisone have been made without any direct comparisons. Nefecon withdrawal can result in adrenal insufficiency (thus there is clear, significant systemic effects) as described by Laxamana et al Clin Kid J 2026. Even ordinary budesonide might be just as good as ‘targeted release’ budesonide for IgAN patients ( Jhaveri and Reich, KI Reports 2026). Even in places where it is available, the cost of Nefecon is prohibitively high, and it is always good to have cheaper and effective options.  

The question of whether a C-statistic for benefit of 0.63 is sufficient to change practice is fair, however perhaps the right framing is whether it improves on ‘doing nothing’. In a disease where the average ARR is 16% but individual responses are heterogeneous and toxicity is real, even modest discrimination has value. The model is best used not as an absolute gatekeeper, but as a conversation starter for shared-decision making. The treatment threshold, which is the minimum predicted benefit that justifies accepting the risks of corticosteroids, really depends on individual patient values. This is what personalized evidence-based medicine looks like in practice. The tool doesn't replace clinical judgment; it informs it with individualized data. 

Summary

ARR = absolute risk reduction; IgAN = IgA nephropathy; PATH = Predictive Approaches to Treatment effect Heterogeneity; RAAS = renin–angiotensin–aldosterone system; RMST = restricted mean survival time; RCT = randomized controlled trial.

Summary by Sumaiya Ahmed
Nephrology Fellow, University of Ottawa

Reviewed by Swapnil Hiremath, Brian Rifkin

Tuesday, June 23rd 2026, 9 pm Eastern on X and Bluesky

N Engl J Med. 2026 Jun 5., doi: 10.1056/NEJMoa2602678. Online ahead of print.

Obinutuzumab or Tacrolimus in Primary Membranous Nephropathy

Fernando C Fervenza, Fan Fan Hou, Chuan-Ming Hao, Gianna M Kirsztajn, Loreto Gesualdo, Tomasz Hryszko, Antonio Pisani, Dario Roccatello, Andrew S Bomback, Julie Rae, Farima Barmaki, Eriola Berisha, Thomas Schindler, Theodore A Omachi, Jay P Garg, Ana Malvar; MAJESTY Trial Investigators

PMID: : 42246654

DOI: 10.1056/NEJMoa2602678

Introduction

If you’re a fan of trying to shoot fish in barrels, then maybe a career in designing obinutuzumab trials is for you. Simply ask ‘Does rituximab work for this condition?’, and then show that obinutuzumab can get the job too - hence the ongoing trials in ANCA vasculitis, frequently relapsing nephrotic syndrome, and this feature - the MAJESTY trial in primary membranous nephropathy (pMN).

We are well aware that rituximab (RTX) has efficacy in pMN; remember the NephJC summary of the MENTOR trial, in which investigators compared RTX to 12 months of the calcineurin inhibitor (CNI) ciclosporin, and looked at a primary outcome of complete remission (CR) at 24 months. Partly as large numbers of patients relapsed once discontinuing the short-acting CNI, and didn’t relapse when treated with the long-acting RTX, the trial was a significant success for RTX. Even though the sponsor didn’t apply for a label for the indication, ever since RTX has been used off-label by nephrologists with access.

Now the same first author is back for MAJESTY, though the B-cell depletor is swapped to obinutuzumab (OBI) and the CNI is swapped to tacrolimus (TAC). OBI is a type II anti-CD20 monoclonal antibody (whereas RTX is type I), with a glycoengineered Fc region that confers greater antibody-dependent cellular cytotoxicity and direct B-cell apoptosis.  RTX, by contrast, relies more on complement-dependent cytotoxicity and is more prone to CD20 internalization, a potential resistance mechanism. These differences translate into more profound and sustained B-cells depletion with OBI, including superior clearance from peripheral sites like lymph nodes and spleen (Mossner E et al, Blood, 2010| Reddy V et al, Rheumatology, 2017).  OBI’s first trip to NephJC got it off to a very good start, as though LUNAR was negative (with caveats) for adding rituximab to standard of care in lupus nephritis, OBI in the REGENCY trial was actually able to show nice additional benefit. It has been around for over a decade for the treatment of chronic lymphocytic leukeamia (CLL), but then excitingly also recently gained a license for lupus nephritis, and case reports about OBI efficacy in treatment-resistant cases of almost all of our immune-mediated kidney diseases continue to grow.  

CNIs reduce proteinuria partly via hemodynamic effects (afferent arteriolar vasoconstriction) without necessarily inducing immunologic remission, contributing to high relapse rates after withdrawal - unlike B-cell depleters, that target the autoantibody source (Faul C et al, Nat Med, 2008). 

In pMN we have already had case series such as this one of 20 ‘RTX-resistant’ patients with pMN who were given OBI, with encouraging results (Su X et al, KI reports, 2024). There is also an ongoing need in first line therapies in pMN, as currently a substantial proportion of patients fail to achieve complete response, remission occurs slowly (especially in those with very high starting PLA2R titres), and relapses occur - this has set the stage to see what would happen when more potent B-cell depletion strategies are used up front rather than just in a resistant cohort.

The Study

Methods

MAJESTY was a multinational, randomised, open-label, phase 3 trial of obinutuzumab against tacrolimus, conducted at 49 sites across 11 countries.

Inclusion criteria

Adults aged 18–75 years with all of:

• Biopsy-proven, primary membranous nephropathy

• Persistent nephrotic-range proteinuria despite optimized supportive therapy

• uPCR ≥5 g/g (>565mg/mmol) for at least 3 months, or

• uPCR ≥4 g/g (>450mg/mmol) for at least 6 months

• eGFR ≥40 mL/min/1.73 m²

Patients previously treated with immunosuppression could be included, provided at least 6 months has passed since they’d had CNI or cyclophosphamide, or 9 months for RTX, and they’d not demonstrated resistance to RTX or CNI in the past. Other exclusion criteria included evidence of spontaneous remission, diabetes, or suspected secondary MN.

Randomisation

Patients were randomised 1:1 to receive open label:

Obinutuzumab

• 1000 mg IV at day 1, week 2, week 24, and week 26

or

Tacrolimus

• Starting dose 0.05 mg/kg/day, in 2 divided doses

• Trough target 5–7 ng/mL

• Maintained through week 52 (unless complete proteinuria response, in which case tapering could start from week 36)

• Tapered over 8 weeks

Randomisation was stratified by region, and anti-PLA2R antibody level (≥175 U/mL vs lower). 

If patients met pre-defined failure criteria in either arm there were 4 time points at which their treatment would change, as below, which they called ‘escape criteria’:

Primary endpoint

Complete remission (CR) at week 104, defined as::

• uPCR ≤0.3 g/g (<34mg/mmol) on 24 hour urine collection, AND

• Stable kidney function (eGFR ≥85% of baseline), AND

• No intercurrent events (escape therapy, treatment failure, early study withdrawal)

Key secondary endpoints (tested in hierarchical order)

Partial remission (PR) was defined as a 50% reduction in uPCR from baseline but still in the 0.3 - 3.5g/g range (35 - 396mg/mmol) AND with stable eGFR.

Specifically based on MENTOR, they estimated that a sample size of 140 patients would provide the trial at least 90% power to detect a difference between trial groups in the percentage of patients who had complete remission (25% with OBI vs. 5% with TAC) at a two-sided alpha level of 0.05. The efficacy analyses were performed in the intention-to-treat population, which included all the patients who had undergone randomisation, with patients grouped according to the treatment assigned rather than the treatment received. 

As seen in Figure S1 above, there is a study of longer term safety planned for a further 2 years, the results of which are not yet available.

Funding

The sponsor (Roche) designed the trial, participated in the collection, analysis, and interpretation of the data, and contributed to manuscript preparation - with their employees amongst the authors.

Results

The consort diagram is found in figure S2:

The baseline characteristics are shown in Table 1:

Of note, the population had a mean age of 50 years with male and white pre-dominance, and around 78% PLA2R positivity, all of which is representative of pMN. Mean eGFR was >80, and mean 24 hour urinary protein level was around 10g/day despite great uptake of RAASi, with mean serum albumin of 29. Around 30% had received prior immunosuppressive therapy.  Rates of supportive care with baseline SGLT2i were not reported (and their addition was not permitted during the trial).

Slightly confusingly, the eGFR range seems to go down to 24 mL/min/1.73 m² in Table 1, and the uPCR range down to 1.7g/day, making it unclear how these patients were recruited, as these values fall outside the inclusion criteria (our assumption being their lab values changed between screening and first visit).

Primary Endpoint

The trial met its primary endpoint convincingly - at week 104, 37% of patients in the OBI group had reached complete remission (CR), versus only 6% in the TAC arm, which (as you’d guess) was highly statistically significant.

Analysis showed consistent results across all sub-groups.

The peak remission rate with obinutuzumab occurred by week 76, and was sustained thereafter. For key secondary endpoints: overall remission at week 104 was 51% vs 13% (P<0.001), and complete remission at week 76 was also significant (44% vs 14%).  Patients on TAC who had achieved CR/PR relapsed very frequently after drug weaning at 52 weeks, as shown in Figure 2.

Sequentially next, the sustained eGFR reduction endpoint showed no difference (6% in both arms; HR 1.10, 95% CI 0.18-6.61), at which point hierarchical testing stopped. Consequently, the duration of CR and PROMIS fatigue score changes were not formally tested and their CIs are descriptive only (though the fatigue scores don’t appear to have a meaningful clinical difference).

Escape Therapy

A vast majority (61%) of the TAC arm met escape criteria and therefore received OBI, with what appears to be a reasonably even split between meeting escape criteria during the first or second year of the study (with 10 patients at week 24, 3 at week 52, 17 due to relapse after week 52, and 9 due to increased serum creatinine while on TAC).  In the OBI arm 28% met escape criteria (meaning that 44% of the entire trial population met escape criteria at some stage). Data on total OBI doses given across the original arms and then in the escape groups was not reported. Only 2/70 patients in the OBI arm had to start an additional non-glucocorticoid immunosuppressant due to treatment failure.

Immunologic remission and B-cell depletion

Immunological remission (defined as a change in anti-PLA2R autoantibody status from positive to negative) can of course only be explored in PLA2R positive patients. As can be seen in Figure S6(A), OBI gave faster, more effective, and more sustained PLA2R suppression when compared to TAC, with consistent results seen even when looking at patients with high starting PLA2R titres (defined here as ≥175 U per milliliter).

Fig S6 A/B. Immunologic remission of anti-PLA2R. Fervenza FC, et al. N Engl J Med, 2026.

OBI certainly does what it says on the label - depletion of CD19+ B cells seen in Figure S8(A).

The pattern of complete B cell depletion (levels <10 cells per microliter) is interesting, tailing off a little before the second course of OBI was administered at 6 months, getting to 94% of the patients with complete depletion at week 52, which gradually dropped to only 14% by week 104. Notably, the protocol defined depletion as ≤5 cells/ microliter, and no justification for this threshold change compared to protocol was offered.

Safety

The analysis of any safety signals gets a bit hard to interpret and needs some background knowledge to read - Table 3 reports the randomised safety data for the 104 week trial, but not if they left, as directed for escape therapy. Therefore, what you are seeing in Table 3 is NOT safety data for all the patients who received OBI (remembering that 61% of the original TAC group met escape criteria and then received OBI). In fact, no systematic escape period safety data are reported anywhere. At a glance, you might think no deaths occurred, but actually, as is explained in the legend, two deaths did occur during escape therapy with OBI (one from each original treatment arm), attributed to covid-19 pneumonia and cardiac arrest.  The main article text states that investigators did not think the pneumonia death was related to the OBI received (though how exactly one adjudicates a death due to infection as unrelated to the potent long-acting B-cell depleter you have given them is not justified further, especially given deaths due to covid-19 were much less common by the time recruitment opened in summer 2021).

Table 3. Adverse events. Fervenza FC, et al. N Engl J Med, 2026.

Infections occurred at a rate of 61 events per 100 patient years in the OBI arm and 57 per 100 patient years with TAC. Serious adverse event rates per 100 patient-years were 11 (OBI) vs 14 (TAC). Table 3 reports a 4% neutropenia rate with OBI, though the relevance of this is less clear when overall the infection comparison data is reassuring.

There was a 38% infusion reaction rate with OBI, but they specifically note that these were manageable with infusion-rate adjustments and symptomatic treatment. I think we’d interpret Table 3 and it’s legend as saying two subclinical HepB re-activation events occurred, one in the OBI arm and maybe one after OBI was given for escape criteria (as if re-activation occurred on TAC it would be recorded in Table 3, which it is not) - it is unclear, and it’s also not stated whether the re-activation occurred despite antiviral prophylaxis or not.

The number of patients in both arms who had a sustained eGFR decrease of >30% at week 76 was low, at 6% in both arms.

Serum IgG levels actually increased over the course of the trial in both arms, likely due to decreased IgG losses in the urine as nephrosis improved, and not indicating less IgG recovery with OBI over TAC. 

Discussion

MAJESTY results analysis

No one is surprised to see the pattern of MENTOR’s results reproduced here 7 years later - patients who stop their CNI then frequently relapse over the next year, whereas the patients who instead received long-acting B cell depletors are much more likely to achieve and stay in remission.  

It is nice to see not only that OBI works in a large cohort, but in particular the immunological remission data looked impressive. There was no difference in renal function between groups, as would be expected over a 2-year follow-up of a group with relatively preserved kidney function at baseline.

The catchphrase of seemingly all Phase 3 trials presented at ERA was ‘there were no new or unexpected safety signals’. Equal infection rates between arms looked reassuring, as did the IgG data, but given there is a black box warning regarding hepatitis B reactivation on OBI’s FDA prescribing label, I don’t think it is too much to ask for them to be more explicit on whether the 2 patients who had subclinical HepB reactivation during the trial had both had OBI, as the text leaves you guessing. Rates of neutropenia of 4% with OBI were lower than the 14% seen in REGENCY (Furie RA et al, NEJM, 2025), probably because these patients with SLE were also on MMF, and in keeping with this, the placebo control arm in REGENCY also had high neutropenia rates at 6%. They do make it clear in MAJESTY that both patients who died during the trial had received OBI, even mentioning this in their abstract. As in REGENCY, a death due to COVID-19 after OBI in this trial could be seen as a marker of increased viral morbidity, though somehow here the investigators judged the COVID-19 death and having received OBI as unrelated to each other, which, at face value, is baffling. The supplement in REGENCY gave further details about deaths, but no further details were given here. It also appears the local investigators have flagged up more ‘adverse events related to the study drug’ in the TAC arm than with OBI, with 34 events in the OBI arm - though with 27 infusion reactions and 36 patients having infections in the OBI arm, it isn’t clear how that maths stacks up, and it’s just hard to see how an investigator could confidently judge an infusion reaction or infection to be unrelated to their open label B-cell depletor immunosuppression. Very disappointingly, there is also a major safety reporting gap of adverse events during escape therapy, with no mention of infusion reaction or infection rates for the large crossover population, which you’d have thought would have been common sense to include.

Comparison with MENTOR and rituximab

OBI isn’t trying to inherit the throne from CNI therapy in countries that will be able to afford OBI access - it’s trying to succeed rituximab. While MAJESTY was obviously not designed to test whether OBI is superior to rituximab, and comparing different trials must be done with caution, nephrologists will still be tempted to compare results here with those from MENTOR (Fervenza FC et al, NEJM, 2019).

Inclusion criteria were similar enough, and baseline characteristics of the population were also broadly similar. MENTOR and MAJESTY differed in their exact outcome definitions, in that MAJESTY more stringently also required eGFR preservation ≥85% of baseline in both their CR and PR definitions on top of proteinuria remission, whereas MENTOR had proteinuria criteria only. In that context, looking at CR at week 104, it was 37% in MAJESTY with OBI versus 35% in MENTOR with RTX.  In MAJESTY at week 104 CR and PR combined was only at 51% with OBI, whereas it was up at 60% with RTX. However, if the eGFR criteria were removed from the MAJESTY definition and instead just the proteinuria definitions were applied to both, then at 52 weeks, OBI CR rates were 26% versus 14% in MENTOR, and by week 104 OBI got 49% to complete proteinuric response versus the 35% in with RTX.

‘Reading the tea leaves’ of the above, you could say that OBI’s results look better than the RTX results, if not dramatically so.  It’s not clear why the eGFR criteria for PR/CR were added between studies.  

What about the patients with high starting PLA2R titres? To remind you of table S13 from MENTOR, when they split their patients into tertiles of PLA2R level, the immunological response rate did tail off a lot in their highest PLA2R cohort, unlike OBI’s results here (accepting the number of patients in each group by that stage of subdivision is relatively small).

In terms of safety, MENTOR did not include neutropenia or IgG data, and did report that RTX had numerically fewer infection events than ciclosporin. They did not have any deaths during the trial, which is the norm across pMN trials, and contrasts with MAJESTY. Infusion reactions to RTX were at 25% (which certainly seems higher than I’d have expected), versus the 38% rate with OBI.

MAJESTY limitations

The open-label design certainly risks influencing more subjective outcomes like ‘AEs related to study drug’ and fatigue score, if less so the hard ‘number based’ outcomes of the trial, and blinding would have been difficult (given the high rates of transfusion reaction and need to adjust TAC doses based on levels) and time-consuming. Treatment durability and safety beyond 2 years is currently unknown, and the further 2 years of observation is underway.

However, the big talking point of MAJESTY will be the decision to have the comparator as 52 weeks of TAC, rather than against RTX. In recent years, we’ve been inundated with trial designs that aren’t exactly what patients or their doctors would have chosen - voclosporin not being compared to TAC for lupus, or finerenone against other MRAs, to name a few - and here finding that B cell depletion therapy beats 52 weeks of CNI at a 104 week endpoint was certainly not providing the world with new scientific information.  

Supporting the choice of TAC in MAJESTY is that 7 years ago a study of US prescribing patterns still put TAC use at 40% (O'Shaughnessy MM et al, Kidney Int Rep. 2019), though you’d expect this to have shifted significantly since MENTOR, and the fact that TAC is on the KDIGO GN guidelines (at least for patients at moderate risk, though you’d guess actually many MAJESTY patients actually fell into the high risk group, in whom KDIGO would not have recommended CNI - I’m looking at that 3043 U/ml PLA2R titre in Table 1…). But most importantly, the increase in sample size that would have been required for a superiority trial of OBI versus RTX would have been entirely prohibitive, put at n=2500 patients if assuming 45% versus 35% remission rates, which can’t be done in rare disease. The ‘need’ to have a feasible trial done has been said to be that it is hard to access RTX in the US given the lack of license for the pMN indication, whereas now hopefully OBI will soon be licensed - though anecdotally plenty of US physicians do seem to access RTX without any problems as things stand.

The escape criteria were enacted to standardise the implementation of an effective rescue therapy - interestingly, the initial protocol had TAC as the switch option for patients failing on OBI, but this was replaced in later iterations by a second course of OBI, which seems reasonable. The large breadth of the escape criteria was important, as investigators knew patients randomised to the TAC arm were going to relapse like wildfire when they discontinued the TAC at 52 weeks in the same way that occurred many years prior in MENTOR, but the huge 61% of patients on TAC meeting escape criteria did make this trial feel almost closer to a cohort study of OBI rather than an RCT of two effective therapies. The primary estimand’s composite strategy (counting escape as non-response) correctly penalizes the TAC arm, preserving intent-to-treat validity.  However, the treatment policy sensitivity analysis (counting escape responses) showed a similar effect (44% vs 14%), confirming robustness (table S5).

MAJESTY’s TAC monotherapy arm contrasts with STARMEN’s (Fernandez-Juarez G et al, Kidney Int, 2021) sequential TAC-RTX regimen (58% CR/PR at 24 months). The 20% CR/PR rate in MAJESTY’s TAC arm at 24 months is far lower, confirming that adding RTX at TAC withdrawal (as in STARMEN) prevents relapse, but still likely underperforms OBI monotherapy (accepting that in clinical practise most would add the RTX earlier than in the STARMEN protocol in the hope of improved outcomes).

The question is, would you let your family member be recruited into a trial in which, by their own numbers, they (accurately) predicted a primary outcome achievement rate in one arm of 5%, when a much more effective therapy exists? To be fair, relapse is not a disaster in pMN when the patient is closely watched, but would you risk randomisation to TAC if their eGFR was 40 and proteinuria >10g/day? I’m really not so sure. However, if you were hoping this is the last pMN trial of ‘new agent versus CNI’ then you haven’t heard of the Phase 3 PROMINENT trial of felzartamab versus TAC with completion expected in 2029, or the Phase 2/3 ALMOND trial of zanubrutinib versus TAC also on the horizon - so it looks like we’re getting everything compared to TAC (except, of course, for voclosporin) for a while yet. How we’ll sort out which agent is best in 5 years when they have never been directly compared to one another is anyone's guess for now, but the industry interest in pMN will hopefully be a great thing for patients going forward.

Will clinicians now use OBI first line?

It is accepted that OBI has superior speed and depth of B-cell depletion compared to RTX, particularly in lymph nodes and spleen; for the simple-minded like me, it can be thought of as ‘super-rituximab’, much like sparsentan for IgAN sits in my brain as ‘super-ACEi’, with increased efficacy but also concern regarding an uptick in side-effects that goes along with that. While waiting to see if OBI gets approval for pMN in their healthcare system, nephrologists will meanwhile gain experience with OBI from the lupus nephritis indication, and we’ll be eager to see more and more trials in disease areas where RTX is good but there’s room for improvement, for example, in this phase 2 in AAV, which, fantastically, they have finally called OBI-WAN.

Now, on top of the reports that OBI can overcome the pMN disease that is resistant to RTX, we have this MAJESTY data, which indicates an impressive immunological response even in patients with high starting PLA2R titres (which may be lacking with RTX), and likely favourable proteinuria outcomes at 2 years with OBI compared to what RTX achieved in MENTOR.  If access to both were equal, you can imagine many reaching for OBI first line now if their patient is in agreement, though the safety data outlined above may put some off. If having a license for the pMN indication makes a huge difference to access in an insurance-based system, then first-line OBI will become the norm in those areas, whereas in systems which (forgive me) sensibly recognise the rituximab efficacy data and prioritise value over formal application for indication, then it seems possible that OBI will be rationed in some healthcare systems only for resistant patients, especially with generic RTX now available. Of course, some clinicians will want to reach for OBI in patients with high PLA2R titers, but others might make do with RTX in combination with a few early months of CNI for their higher-risk patients, or give more doses of RTX than usual in the hope of deeper B-cell depletion. Overall, it will be good for patients to have increasing options, and the toxicity of cyclical cyclophosphamide/steroid protocols will be seen less as treatment resistance is increasingly consigned to the past.

Conclusion

MAJESTY had no surprises by showing obinutuzumab beats 52 weeks of tacrolimus in primary membranous, with markedly fewer relapses and escape-therapy requirements.

No doubt a label application will come and be successful, and then we’ll constantly hear how obinutuzumab is “the only licensed treatment for membranous nephropathy”.  For the future it’s great that patients will have increasing options with additional efficacy, but if where you work still has you reaching for rituximab rather than obinutuzumab for now, on balancing the available efficacy and safety data it seems unlikely your patients are missing out too much.

Written by Jamie Willows

Renal and GIM Consultant

South Tyneside and Sunderland Foundation Trust

UK

Reviewed by 

Joel Topf, Cristina Popa, Brian Rifkin, Pallavi Prasad

Audiosummary prompted by Nikhil Shah

Header designed by AI and prompts from Brian Rifkin

Measured and Estimated Glomerular Filtration Rates and Risk of Adverse Health Outcomes

Edouard L Fu, Antoine Créon, Morgan E Grams, Josef Coresh, Arvid Sjölander, Anne-Laure Faucon, Michelle M Estrella, Friedo W Dekker, Michael G Shlipak, Lesley A Inker, Andrew S Levey, Juan-Jesus Carrero

PMID:  42240159

DOI: 10.1001/jama.2026.9639

Why was this study needed?

Let us begin with an uncomfortable truth. The entire edifice of CKD classification, the thresholds of 60, 45, 30, and 15 mL/min/1.73 m2, rests on associations between estimated GFR and adverse outcomes (Levey AS et al, Kidney Int, 2011). This is a circular argument. We use eGFR to define disease, then validate that definition by showing that eGFR predicts outcomes. The implicit assumption has always been that eGFR is a faithful proxy for true GFR, differing only by random, non-systematic error. But we have known for decades that this assumption is false.

Creatinine, the workhorse biomarker, is a byproduct of muscle metabolism influenced by a patient’s frailty, high-protein intake and/or tubular secretion. (Stevens LA et al, NEJM, 2006). Cystatin C, while more independent of muscle mass, is an acute-phase reactant potentially influenced by inflammation, obesity, smoking, and glucocorticoids (Stevens LA et al, AJKD, 2008| Russel WA et al, JASN, 2025 | also see NephJC discussions here and here). The non-GFR determinants of these markers are systematic biases that correlate with the very outcomes we are trying to predict. A patient with sarcopenia has both a falsely elevated eGFRcr and an increased risk of death, but from different causes. A patient with chronic inflammation has both a falsely depressed eGFRcys and an increased cardiovascular risk. The central question, then, is not whether eGFRcr predicts outcomes - we know it does, redundantly - but whether the risk signal we are measuring actually originates from GFR itself.

The study by Fu et al, simultaneously published during ERA26, breaks this circularity by anchoring risk to measured GFR (mGFR) (Fu EL et al, JAMA, 2026). Using iohexol plasma clearance - a marker that is freely filtered, neither secreted nor reabsorbed, with very low extra-renal clearance (Delanaye P et al, Clin Biochem Rev, 2016) - the authors provide the first large-scale, direct quantification of how true mGFR associates with adverse outcomes. They then ask a second, equally important question: which of our common eGFR equations most faithfully captures these mGFR-based risk associations?

How was the study done?

The investigators used the Stockholm CREAtinine Measurement (SCREAM, an homage to Edvard Munch?) project to identify 6174 adults who underwent clinical mGFR using single-sample plasma iohexol clearance. The protocol was rigorous: a 5 mL iohexol injection, timed blood sampling tailored to expected GFR (approximately 4 hours if GFR >40, 6-8 hours if GFR 15-40, and 24 hours if GFR <15), quantification via ultra-high-performance liquid chromatography with UV detection, and the Jacobsson equation for clearance calculation (eMethods- supplement 1). Importantly, 78% of patients had mGFR, creatinine, and cystatin C measured on the same day, eliminating temporal confounding between the gold standard and estimators. The analytical coefficient of variation for the iohexol method was 2.31% for a control sample at 32 mg/L and 2.04% at 65 mg/L- acceptable but not negligible. 

We must be clear about what mGFR is and is not. While iohexol clearance is the best available reference method, it is not error-free. The within-person biological coefficient of variation for mGFR has been estimated at approximately 5-10% across studies (Rowe C et al, Kidney Int, 2019| Delanaye P et al, Clin Biochem Rev, 2016). The single-sample method, while practical, is less accurate than multi-sample protocols, particularly at extremes of GFR and in patients with altered extracellular fluid volumes. The supplementary materials note that patients with extensive edema or ascites were not excluded; in such patients, plasma clearance protocols can be inaccurate, and urinary clearance would be preferred. This is a limitation the authors acknowledge but cannot fully address. 

The key design feature was the simultaneous measurement of serum creatinine and cystatin C, allowing direct comparison of mGFR with 3 eGFR equations: eGFRcr (CKD-EPI 2021), eGFRcys (CKD-EPI 2012), and eGFR-cys (CKD-EPI 2021). The study population (eTable2) shows that 28% had missing BMI and 21% had missing UACR, requiring multiple imputation. To note, patients with complete data had substantially lower eGFR (median 68 vs 91 mL/min/1.73 m2) and higher cystatin C (1.36 vs 1.05 mg/L) than those with missing data. They also had higher prevalence of hypertension (53% vs 25%), diabetes (27% vs 12%), and heart failure (14% vs 6%). “Missingness” was clearly not random, and patients with complete data were sicker with more advanced CKD. The imputation model, which included event indicators and the Nelson-Aalen estimator, is a best practice, but it assumes data are missing at random, a strong assumption that may not hold.

eTable 2. Baseline characteristics of persons undergoing mGFR testing using plasma clearance of iohexol in Stockholm with simultaneous creatinine and cystatin C testing during 2011-2021, with versus without missing data of UACR and body mass index, from Fu EL et al, JAMA, 2026

The authors employed Cox proportional hazard regression with restricted cubic splines to model GFR as a continuous, non-linear variable, adjusting for the covariates detailed in eTable 1. To avoid index event bias - a form of selection bias that occurs when individuals with a prior history of the outcome are included in the analysis of incident events - they correctly excluded such individuals from each outcome-specific model (eFigure1)- only applied to secondary outcomes. Additionally, to quantify the statistical uncertainty around the hazard ratios comparing eGFR, the authors used bootstrap resampling with 500 iterations, a robust and methodologically appropriate technique for constructing confidence intervals in this context.

Results

First finding: the current GFR threshold of 60 mL/min/1.73m2 was associated with significantly higher risks of both all-cause mortality (HR 1.21, 95% CI 1.14-1.28) and kidney failure requiering replacement therapy (HR 2.85, 95% CI 2.06-2.94). As measured GFR declined further, these risks increased steeply: at an mGFR of 30 mL/min/1.73m2, the hazard ratio was 38.5; at an mGFR of 15, it was 200.3. The graded, monotonic relationship between lower true GFR and higher risk is unambiguous. The heatmaps from the CKD Prognosis Consortium are not mere artifacts of estimation (Grams ME et al, JAMA, 2023). The threshold of 60 mL/min/1.73m2 holds.

Below 90 mL/min/1.73m2, eGFRcr systematically underestimates mortality risk (ratio of HR at 60, 0.87, 95% CI 0.79-0.95), while eGFRcys systematically overestimates it (ratio of HR at 60, 95%CI 1.08-1.27). In contrast, eGFRcr-cys showed no significant deviation from mGFR-based risk across the entire GFR range (eFigure 4, panel A).

The supplementary materials add critical nuance. For kidney failure with replacement therapy, the confidence intervals for the HRs were extremely wide- for eGFRcr at 60 mL/min/1.73m2, the 95% CI ranged from 0.00 to 1.17 (eFigure 4, panel B). The authors cite limited power, but an alternative explanation is that kidney failure is a kidney-specific outcome, while mortality is influenced by systemic factors. The non-GFR determinants of creatinine and cystatin C may be less correlated with progression to kidney failure than with death. 

In addition, an interesting finding can be seen in eTable 6. In the general population of 1.58 million Stockholm residents, the HR for mortality at an eGFRcr of 120 (vs 90) was 4.43 (95% CI, 4.34-4.52), more than double that observed in the mGFR cohort (2.10, 95% CI 1.76-2.50). The well-described U-shaped association between high eGFRcr and mortality was markedly attenuated in the selected mGFR cohort. Patients referred for mGFR testing- typically for drug dosing, liver cirrhosis, transplant evaluation, or eGFRcr-eGFRcys discordance- have a different risk profile than the general population. These results may not generalize to patients who would not ordinarily undergo mGFR testing. 

eTable 11 and eFigure 10 show that including individuals with a history of heart failure substantially attenuated the HRs for eGFRcys and eGFRcr-cys. At an mGFR of 15, the HR for eGFRcys dropped from 4.25 (excluding prevalent cases, eTable 7) to 2.96 (including them, eTable 11). This attenuation suggests that the overestimation of risk by eGFRcys is most pronounced in patients without pre-existing heart failure, where cystatin C is less confounded by non-GFR determinants.

eTables 7 and 8 show consistent results using the older CKD-EPI 2009/2012 equations and the European Kidney Function Consortium (EKFC) equations. The combined equation outperformed single-marker equations across all GFR ranges, strengthening the conclusion that the principle of combining markers- not the specific coefficients of any one equation- is what matters.

How would the study change the practice?

First, eGFRcr alone is insufficient for risk stratification in patients with muscle wasting, chronic inflammation, or corticosteroid use. The bias is quantifiable: at an mGFR of 60, the mortality HR based on eGFRcr is 13% lower than the true HR. The 2024 KDIGO guidelines recommend cystatin C when greater accuracy is required (KDIGO CKD Work Group, KI, 2024). The study provides empirical justification: eGFRcr-cys is the only estimator whose risk associations are statistically indistinguishable from those of mGFR. 

Second, the study reframes the race-free equation controversy. The 2021 CKD-EPI equation removed the Black race coefficient because race is a social, not biological, construct (Inker LA et al, NEJM, 2021). That decision was ethically correct. But the more fundamental problem is the use of a single, non-specific biomarker. eGFRcr-cys, which requires no race coefficient, provides the most accurate risk estimation regardless of race.

Third, the study raises a subtle question about the nature of risk. If eGFRcys overestimates risk because it captures inflammation, and inflammation is itself a true risk factor, is the overestimation truly a “bias”? For the question, “what is the patient’s prognosis?”, the discrepancy between eGFRcr and eGFRcys-low eGFRcr with normal eGFRcys suggesting muscle-wasting, or low eGFRcys with normal eGFRcr suggesting inflammation, may be prognostic in its own right. 

Fourth, the study has implications for trial design. If eGFRcr underestimates risk in frail patients and eGFRcys overestimates it in inflamed patients, using either single marker as an inclusion criterion or endpoint could introduce bias. Trials using eGFRcr-cys for enrollment may achieve more homogeneous risk populations and greater statistical power.

Limitations

 Several limitations temper our enthusiasm. Missing data were substantial: patients with complete data had a standardized mean difference of 0.63 for eGFRcr and 0.56 for mGFR compared to those with missing data (eTable 2). The imputation model may have been inadequate if “missingness” was related to unmeasured factors such as frailty. 

The generalizability concern from eTable 6 is not trivial. The mGFR cohort was selected by clinicians. The attenuated U-shaped mortality curve at high eGFRcr suggests that healthy individuals with high eGFRcr were systematically excluded. These results may apply primarily to patients with established CKD or complex comorbidities.

eTable 10 shows that excluding kidney transplant recipients (approx. 2.7% of the cohort) did not materially change the results, which is reassuring for generalizability to non-transplant CKD populations. 

Finally, the study had insufficient power to assess mGFR-related risks across albuminuria categories. Given that albuminuria is an independent and synergistic risk factor- central to the KDIGO heatmaps- this is an important gap.

Conclusion

Fu and colleagues have confirmed that our current GFR threshold of 60- derived from eGFR- holds up when tested against measured GFR. They have also shown that eGFR underestimates risk, eGFRcys overestimates it, and eGFRcr-cys corrects most of this bias. The supplement adds important caveats: the biases differ by outcome, weaken in some sensitivity analyses, and may not apply to unselected populations. 

None of this means abandoning creatinine. It remains an excellent screening tool. But when the stakes are high, when a patient’s eGFR falls on the border of a major clinical decision, we owe it to them to be honest about the limits of our estimates. eGFRcr-cys is a better map, though not a perfect one. When even that is not enough for dosing highly toxic chemotherapy, evaluating living donors with discordant markers, or confirming CKD in frail patients with borderline creatinine, measured GFR remains the territory we must be willing to visit.  

By Cristina Popa

Reviewed by 

Brian Rifkin and Swapnil Hiremath

Tuesday, June 2nd 2026, 9 pm Eastern on X and Bluesky

Kidney Int. 2026 Apr 23:S0085-2538(26)00305-4. doi: 10.1016/j.kint.2026.03.015

Clinical and histologic predictors of non-diabetic kidney disease in patients with diabetes mellitus

TN Caza, V Charu, DF Dai, VG Davis, F Boyd, L Spenst, PD Walker 

PMID: 42034202 

Audio Abstract: https://audioscholar.cc/pdf/q2brtw2u7b?t=0

Introduction

Diabetic nephropathy (DN) has long been recognized as the leading cause of chronic kidney disease (CKD) and end-stage kidney disease (ESKD) worldwide, correlating with the escalating global prevalence of diabetes mellitus (Zouh et al, Lancet, 2024). Historically, the clinical paradigm assumed an inevitable progression of renal decline in diabetic patients, characterized by a predictable transition from hyperfiltration to overt proteinuria and eventually ESKD. However, the advent of modern nephroprotective therapies has altered this natural history (Afkarian et al, JAMA, 2016), making atypical clinical courses more common and questioning the assumption that all kidney dysfunction in these patients is solely attributable to diabetes. It is increasingly recognized that patients with diabetes frequently develop non-diabetic kidney disease (NDKD), either as an isolated pathology or superimposed on underlying DN (Fiorentino et al, Nephrol Dial Transplant, 2017).

Recent global literature highlights that the prevalence of NDKD among diabetic patients undergoing kidney biopsy is remarkably high, frequently ranging from 18% to over 62%, depending on the specific cohort and biopsy thresholds utilized (Zeng et al, BMC Endocr Discord, 2022). The accurate identification of NDKD is critical because it dramatically alters a patient's prognosis and therapeutic trajectory. Many secondary diagnoses, such as acute interstitial nephritis, IgA nephropathy, or crescentic glomerulonephritis, are potentially reversible with targeted interventions like immunosuppression. Consequently, identifying and treating these conditions can significantly delay or entirely prevent the onset of ESKD.

Despite the high prevalence of NDKD, routine kidney biopsy is not universally performed in this population, and clinicians often face a diagnostic dilemma when evaluating kidney function impairment in diabetics. While emerging non-invasive diagnostic models and clinical predictors such as a short duration of diabetes (e.g., less than five years), the absence of diabetic retinopathy, active urinary sediment (hematuria), and acute, rapidly progressive renal failure provide valuable guidance, they cannot replace the definitive diagnostic utility of a tissue biopsy.

Currently, there are no established, universal guidelines dictating precisely when to perform a renal biopsy in a patient with diabetes for the identification of underlying non diabetic kidney disease. To address this critical knowledge gap Caza et al (Kidney Int, 2026) present the largest retrospective cohort study till date of patients with diabetes who have undergone native kidney biopsies, comprising 49,075 cases. By leveraging this unprecedented dataset, this study aims to robustly quantify the true frequency of NDKD, map the distribution of specific histopathological diagnoses, and identify the clinical indications and demographic parameters that significantly increase the odds of discovering NDKD.

The Study

Methods

Study Design

This study represents the largest retrospective analysis till date of native renal biopsies from patients with diabetes. The research was conducted following institutional review board (IRB) approval and utilized a vast database of biopsies received at Arkana Laboratories (Little Rock, Arkansas, USA) between 2001 and 2024.

Cohort Development and Characterization

The study population was derived from a primary database of 229,026 native kidney biopsies. From this pool, 78,886 biopsies (34.4%) were identified as originating from patients with a clinical history of diabetes mellitus.

To ensure diagnostic accuracy and specimen adequacy, cases with fewer than 10 glomeruli by light microscopy were excluded. This resulted in the finalization of two primary cohorts:

• Cohort 1 (n = 49,075): This large-scale cohort was utilized to determine the overall frequency of non-diabetic kidney disease (NDKD), the distribution of specific renal diagnoses, and the clinical morbidities present at the time of biopsy. 

• Cohort 2 (n = 13,995): A subset of Cohort 1, this group consisted of patients with biopsy-proven diabetic nephropathy (DN) where the specific clinical indication for the biopsy was available. This indication was determined through a meticulous manual review of clinical data, including age, gender, ethnicity, and the specific diagnosis the clinician intended to rule out.All kidney biopsies in diabetic patients were included from 2001-2013. From 2013 to 2024, the first 1000 biopsies in patients with DM were included every year. 

• Clinical Subset (n = 400): A subset of consecutively diagnosed patients was further analyzed for granular clinical parameters, including serum creatinine, quantitative proteinuria at presentation, and inpatient versus outpatient status.

Histopathological Evaluation

All renal tissue samples underwent comprehensive processing for light microscopy, immunofluorescence, and electron microscopy using standard diagnostic techniques. The pathologic evaluation included:

• RPS Classification: Diabetic nephropathy was graded according to the Renal Pathology Society (RPS) classification system by a single pathologist

• Chronicity Indices: Quantitative assessment of global glomerulosclerosis, segmental glomerulosclerosis, interstitial fibrosis and tubular atrophy (IFTA), arteriosclerosis, and arteriolar hyalinosis

• NDKD Identification: All non-diabetic renal lesions were recorded, whether they occurred in isolation or concurrently with diabetic nephropathy

Clinical Indications for Biopsy

Clinical indications were self-reported by the treating clinicians and categorized for Cohort 2. These indications included:

1. Acute Kidney Injury (AKI)

2. Acute Nephritic Syndrome

3. Rapidly Progressive Renal Failure

4. Isolated Hematuria

5. Suspected NDKD

6. Proteinuria significantly greater than expected for the degree of DN

7. Chronic Kidney Disease (CKD) of unclear etiology

Analysis of Outcomes and ESKD Progression

To evaluate the long-term prognostic impact of NDKD, biopsy results were integrated with the United States Renal Data Service (USRDS) database (data available through 2022). Renal survival and progression to end-stage kidney disease (ESKD) were compared across three groups:

• DN alone

• DN with concurrent NDKD

• NDKD alone

Kaplan-Meier survival analysis was performed for cases collected between 2001 and 2019, ensuring a minimum of 3 years of follow-up for outcome evaluation.

Statistical Analysis

Descriptive statistics (means, SDs, counts, and percentages) were used to summarize the data. The association between clinical exposures and outcomes was evaluated using:

• Odds Ratios (ORs): With 95% confidence intervals (CIs) to assess the likelihood of NDKD based on clinical indications.

• Breslow-Day Test: To evaluate heterogeneity among ORs.

• Cochran-Armitage Trend Test: To assess linear associations between ordinal measures (e.g., RPS class) and binary outcomes.

• Survival Analysis: Kaplan-Meier curves and Cox regression models were used to determine hazard ratios (HR) for ESKD progression based on NDKD status.

• Multiple Testing Correction: X² tests utilized the Benjamini-Hochberg correction for categorical variables, while continuous variables were analyzed via t-tests or Kruskal-Wallis tests as appropriate.

Results

The cohort included 49,075 kidney biopsies from patients with diabetes analyzed between 2001 and 2024. Among biopsied patients, DN alone was present in 41.2%, 22.9% had DN and concurrent NDKD, and 35.9% had isolated NDKD(not 18.3% as erroneously mentioned at one place in the article describing cohort characteristics), resulting in an overall NDKD prevalence of 58.8% of all biopsies. Patients with NDKD were more frequently biopsied in the inpatient setting and more commonly had positive autoimmune serologies or concurrent infections. A second cohort of 13,995 patients with biopsy-proven DN and documented biopsy indications was subsequently analyzed to assess predictive variables of concurrent NDKD. Mean age was 58.5 years, with male predominance in all three diagnostic categories. Among patients with known diabetes type, 83.2% had T2DM. Hypertension and obesity were significantly more common in patients with DN alone. 

Overall the most prevalent NDKD diagnoses included acute tubular injury (ATI), acute interstitial nephritis (AIN), proliferative glomerulonephritis, IgA nephropathy (IgAN), and crescentic glomerulonephritis. ATI was the most common  diagnosis in both the cohort with DKD + NDKD and those with NDKD, the most frequent concurrent diagnoses were ATI, IgAN, and AIN, while in isolated NDKD, the most common findings included ATI, arterionephrosclerosis, membranoproliferative glomerulonephritis (MPGN), crescentic glomerulonephritis, focal segmental glomerulosclerosis (FSGS), and IgA nephropathy.   The biopsy incidence of DN has steadily increased over time, from 7% of native renal biopsies between 2001–2007 to 17% between 2015–2022.

The highest odds of concurrent NDKD diagnosis was associated with acute nephritic syndrome in a 59% (OR 4.01, 95% CI 2.99 to 5.38), followed by rapidly progressive renal failure (OR 2.38, 95% CI 1.66 to 3.4), acute kidney injury (OR 3.01, 95% CI 2.79 to 3.26), and hematuria (OR 1.60, 95% CI 1.03 to 2.48). Meanwhile, proteinuria (17%) as well as CKD (9%) were associated with the lowest diagnostic yield, making DN alone the most probable diagnosis in these clinical settings. 

Common second diagnoses in patients with DN as per clinical indication were as follows: 

• AKI: ATI (46%), Infection associated GN, Acute interstitial nephritis

• RPRF- crescentic GN (49%), ATI 

• Acute nephritis- infection associated GN(51%), ATI, crescentic GN

• Unexpected increase in proteinuria- ATI(19%), MN , IgAN

• Hematuria- IgAN (44%), Infection associated GN

• CKD- ATI (32%), IgAN, AIN

• Suspicion of NDKD- ATI (18%), Infection associated GN

Kidney biopsies were performed to rule out a specific diagnosis for 19.8% of cases, most commonly for monoclonal gammopathy of renal significance confirmed in only 7% of these patients. Among rule-out biopsies, minimal change disease had the highest diagnostic accuracy (45%), followed by ATI (33%) and membranous nephropathy (30%).

Patients younger than 30 years and older than 60 years had the highest prevalence of NDKD, increasing steadily over 80 years, particularly when diagnosing AKI or acute nephritic syndrome. Black and Hispanic patients were less likely to have an NDKD diagnosis (OR 0.75 and 0.76, respectively), but more frequently demonstrated severe DN on biopsy.

Earlier diabetic nephropathy lesions (RPS class I–II) were associated with higher rates of NDKD compared with advanced diabetic lesions (RPS class III–IV), with a prevalence of 51% versus 23%, respectively. This was a nearly 2-fold difference across almost all clinical indications. 

Among the 13,061 patients who progressed to end-stage kidney disease (ESKD), patients with an NDKD had significantly better kidney outcomes compared with those with DN alone. The presence of DN alone  was associated with a markedly higher risk of progression to ESKD compared to those without DN (HR 2.56; 95% CI 2.45 to 2.68), and this association held across sex, diabetes type, and age groups.

Discussion

This large retrospective cohort of 49,075 kidney biopsies represents the most comprehensive analysis to date examining the prevalence and predictors of NDKD in diabetic patients. The authors report that 58.8% of biopsied patients harbored NDKD either alone or concurrent with diabetic nephropathy (DN), broadly consistent with the prior study, in which 36% of patients had isolated NDKD and an additional 27% had concurrent DN and NDKD (Sharma et al,,CJASN, 2013). 

How should we interpret this finding? Do these data suggest that most diabetic patients with kidney disease require biopsy, or do they instead demonstrate that biopsy selection strongly determines diagnostic yield? The striking heterogeneity in NDKD prevalence across studies, ranging from 3% to 82.9% —is likely explained by differences in patient selection and biopsy thresholds (Fiorentino M et al, NDT 2017). For instance, those who prospectively biopsied consecutive patients with moderate CKD and/or significant proteinuria regardless of clinical suspicion, reported only 33.6% NDKD (18.2% NDKD alone +15.4% NDKD+DKD) (Basu et al, BMJ Open Diabetes Res Care, 2022); while those who biopsied based on multiple atypical indications, reported a higher NDKD prevalence of 74.9% (64% NDKD +10.1% DKD+NDKD)(Sakaci T et al, Clin Nephrol, 2024).

One of the most clinically relevant findings of this study is the multiple clinical parameters examined as potential predictors of NDKD. Patients presenting with AKI, rapidly progressive kidney dysfunction, or hematuria had markedly higher odds of harboring NDKD. These findings reinforce the evidence from prior studies and meta-analyses suggesting that acute or inflammatory presentations should prompt consideration of kidney biopsy in diabetic patients, particularly when the clinical trajectory appears inconsistent with classic diabetic nephropathy (Sreedharan S et al, Cureus, 2026 | Sakaci T et al, Clin Nephrol, 2024 | Bermejo et al, Nefrologia, 2024| Vignesh S et al, Indian Journal of Kidney Diseases, 2024 | Chemouny et al, AJN, 2021| Bermejo et al, Nefrologia, 2016). 

The histologic findings further emphasize the heterogeneity of kidney disease in diabetes. Notably, ATI and AIN emerged among the highest scores of NDKD, findings that were less prominently reported in earlier biopsy cohorts (Sharma et al, CJASN, 2013). Interestingly, this distribution differs from prior studies in which glomerular diseases such as FSGS predominated among NDKD diagnoses (Sreedharan S et al, Cureu, 2026 | Sakaci T et al, Clin Nephrol, 2024 | Vignesh S et al, Indian Journal of Kidney Diseases, 2024| Bermejo et al, Nefrologia, 2016). This discrepancy may reflect differences in biopsy practices, including the broader inclusion of inpatient biopsies and patients presenting with acute kidney injury or inflammatory syndromes in the current study. Taken together, these findings suggest that contemporary diabetic biopsy populations may increasingly capture acute and potentially treatable tubulointerstitial or infection-related pathologies rather than exclusively chronic glomerular diseases.

It is important to note that despite its size and clinical relevance, there are several limitations.

Since this study is a retrospective pathology-based cohort, it is inherently subject to referral and selection bias, such that only patients appropriate for biopsy were included.

Baseline kidney function, longitudinal clinical follow-up, and standardized biopsy criteria were unavailable for many patients. Moreover, the evolution of diabetes therapies and changes in biopsy practice over the two-decade study period may also have impacted both biopsy indications and histologic findings.

Although the KDIGO 2020 Diabetes in CKD guideline recognizes uncertainty of atypical presentations, it does not provide specific recommendations regarding when kidney biopsy should be pursued in diabetic patients. This study highlights the need for systematic and integrated biopsy frameworks that combine clinical features, rather than isolated predictors alone (Gesualdo et al, CKJ 2023). 

This study does offer compelling evidence that kidney disease in diabetics is more heterogeneous than commonly appreciated. These findings challenge the traditional assumption that progressive renal dysfunction in diabetes reflects DN alone, and calls for a more individualized approach to kidney biopsy consideration. In particular, the observation that patients with NDKD demonstrated significantly lower risk of progression to ESKD reinforces the clinical value of obtaining a tissue diagnosis when atypical features are present.

Conclusion

In a time of precision nephrology, this study suggests that kidney biopsy is still considered one of the most useful diagnostic techniques applicable to diagnose the existence and the presence of unexplained or non-classic kidney disease in diabetics.

Summary by

Priya John, Milagros Flores

Reviewed by 

Brian Rifkin, Pallavi Prasad

Header designed by AI and prompts from Brian Rifkin

The NephJC internship is built around a simple but powerful idea: that the best way to learn medicine is to engage with its evidence directly, critically, and in community. Our interns don’t just read papers; they work through the full editorial process, learning to appraise methodology, interrogate study design, challenge conclusions, and translate complex research into a clinical context. Under close mentorship, they develop the kind of rigurous, evidence-based thinking that separates good clinicians from great ones, and that stays with them for most of their career.
Good critical appraisal, it turns out, also has a way of teaching us something about ourselves. It was never really about finding fault- it is about asking honest questions of the evidence we depend on to care for patients. Our interns quickly discover that a well-designed study and a poorly-designed one can look remarkably similar on first read, and that spotting the difference is a skill that takes practice, patience, and more than a little humility. The kind of thinking that makes you a better reader of the literature has a funny way of making you a better clinician, too.

But the internship is only as rich as the community surrounding it. Or interns will be showing up on social media feeds, contributing to journal club discussions, and putting their critical appraisal skills to work, alongside you. That kind of learning takes courage and thrives when met with generosity. So, give our interns a follow on Bluesky/X, and bring your questions and experience to the table. The best critical appraisal happens in dialogue, and that dialogue has always been what makes NephJC worth coming back to.

The NephJC 2026 Interns:

#NephJC Chat

Tuesday, May 19th, 2026, 9 pm Eastern on Bluesky and Twitter

N Engl J Med. 2026 Apr 9;394(14):1376-1387. doi: 10.1056/NEJMoa2504068.

Multifaceted Strategies for Hypertension Control in Low-Income Patients

Katherine T Mills, Marie Krousel-Wood, Erin M Peacock, Jing Chen, Farah Allouch, Amy K Carreras, Siyi Geng, Alecia Cyprian, Gerrelda Davis, Sonja R Fuqua, Darie Gilliam, Angelique Greer, Tammy Mitchell, Wylea Gray-Winfrey, Shondra Williams, Gary M Wiltz, Keith L Winfrey, Hua He, Paul K Whelton, Jiang He

PMID: 41950472 
DOI: 10.1056/NEJMoa2504068

Introduction

Hypertension affects over 1.3 billion adults globally, with prevalence rising fastest in low- and middle-income countries (NCD Risk Factor Collaboration, Lancet 2021| Mills KT et al, Nature Review Nephrology, 2020). Elevated systolic blood pressure (SBP) contributed to nearly 11 million deaths in 2021 (Martin SS et al, Circulation, 2025). For every 10 mmHg reduction in SBP, there is a reduction in major cardiovascular events by 20%, stroke by 27%, and all-cause mortality by 13%- evidence that makes the persistence of uncontrolled hypertension not merely a clinical frustration, but a preventable human tragedy (Ettehad D et al, Lancet, 2016).

Yet hypertension doesn’t hurt; its silence is its cruel feature. Nonadherence to therapy is not a character flaw but rather a consequence of poverty, an architecture of chaos where daily survival consumes cognitive bandwidth. Across the globe, social determinants consistently predict worse outcomes: income-related inequalities persist even within wealthy nations like Japan, where hypertension prevalence is nearly double in low-income vs high-income groups (Aida J et al, Hypertens Res, 2024). Neighborhood deprivation drives undiagnosed hypertension in the US (River CA et al, Hypertension, 2024), and social determinants mediate one-third of the Black-White difference in uncontrolled BP (Akinyelure OP et al, Hypertension, 2023). “Clinical inertia”- failure to intensify therapy despite uncontrolled BP- compounds the problem, accounting for 40% of treatment failures in some real-world datasets (Satoh M et al, Hypertens Res, 2024).

Single-level strategies, when employed, have failed. A meta-analysis of 121 trials found that provider training alone yielded no significant BP reductions, while team-based care with nonphysician medication titration, health coaching, and home monitoring reduced SBP by 7.1, 3.9, and 2.7 mmHg respectively (Mills KT, Ann Intern Med, 2018). Task sharing with community health workers proved effective, even in resource-poor settings: a trial among uninsured patients in Argentina achieved a 6.6 mmHg reduction over 18 months (He J et al, JAMA, 2017). Yet Argentina’s national system bears little resemblance to the landscape of US federally qualified health centers, where insurance instability and mistrust add up to impair care. The IMPACTS-BP trial (Mills KT et al, NEJM  2026) was designed to find out whether these strategies could be employed into that uniquely challenging context- and whether intensive BP targets could be achieved safely in a low-income population.

The Study

Methods

IMPACTS-BP was a cluster-randomized, effectiveness-implementation hybrid type 2 trial (Landes SJ et al, Psychiatry Res, 2019) conducted in 36 federally qualified health center (FQHC) clinics across Louisiana and Mississippi. The investigators aimed to answer 2 questions simultaneously: whether intensive hypertension treatment improves blood pressure control in underserved populations and whether resource-limited primary care clinics can realistically adopt and sustain a complex implementation strategy. 

Randomization was done at the clinic level because the intervention targeted the entire care system, including providers, nurses, workflows, audit systems, and patient coaching. Clinics were randomized 1:1 to either the intervention or enhanced usual care arm, with stratification by FQHC organization to minimize imbalance between health systems. 

The investigators used the Consolidated Framework for Implementation Research (CFIR) to guide both intervention development and implementation evaluation. 

Study population

Eligible participants were adults aged ≥40 years with uncontrolled hypertension and high cardiovascular risk. Participants had to have SBP ≥140 mmHg if untreated or ≥130 mmHg while receiving antihypertensive therapy. High cardiovascular risk included prior cardiovascular disease, chronic kidney disease, diabetes, estimated 10-year ACVD risk ≥10%, or age ≥65 years. Pregnant women and patients with end-stage kidney disease were excluded. Eligibility BP was confirmed using the average of six standardized measurements obtained during two screening visits.

Intervention strategy

The intervention consisted of a multifaceted, team-based implementation strategy targeting barriers to hypertension control at the system, provider, and patient levels. 

The core clinical component was protocol-based intensive blood pressure management adapted from the SPRINT trial (SPRINT Research group, NEJM, 2015). The target BP was SBP <120mmHg and DBP <80 mmHg. A stepped-care algorithm guided medication intensification: if BP remained above target, providers were expected to uptitrate therapy or add another antihypertensive agent unless contraindicated. 

Follow-up was intentionally intensive to reduce therapeutic inertia. Patients attended monthly visits during the first 3 months and then every 3 months thereafter; additional monthly visits continued until BP targets were achieved.

The implementation strategy incorporated team-based care involving physicians, nurses, medical assistants, and health coaches. Staff were trained in standardized BP measurement, lifestyle counseling, and medication adherence support. All intervention patients received home BP monitors and were instructed to measure BP three days per week. Providers incorporated both clinic and home BP readings into treatment decisions. Health coaching addressed medication adherence, lifestyle modification, reminder systems, low-cost medication planning, and social barriers to care. Quarterly BP audit and feedback sessions reviewed patients with uncontrolled hypertension. Providers also underwent initial and annual retraining on intensive BP management and the SPRINT-based treatment protocol.
Control group

Enhanced usual care clinics continued routine hypertension management. Providers attended a webinar reviewing the ACC/AHA hypertension guidelines and SPRINT findings and received training in standardized BP measurement, but no additional implementation strategies were introduced.

Outcome assessment

The primary effectiveness outcome was the change in SBP from baseline to 18 months. Secondary clinical outcomes included achievement of SBP <120 mmHg, SBP <130 mmHg, reduction in SBP >30 mmHg, change in diastolic BP, quality-of-life measures, and adverse events such as hypotension, falls, and kidney function decline. 

The primary implementation outcome was an adherence summary score ranging from 0 to 4. This composite incorporated four domains: high medication adherence, treatment intensification, home BP monitoring, and receipt of health education. Higher scores reflected better implementation fidelity and patient engagement. 

Additional implementation outcomes included acceptability, feasibility, adoption, fidelity, sustainability, provider experience, and organizational readiness. These were assessed through surveys, electronic health records, provider assessments, and coaching completion data.

Blood pressure measurement

Blood pressure measurement was rigorously standardized. Measurements were obtained using an automated Omron HEM-907XL device with appropriate cuff sizing. Patients rested in the seated position for 5 minutes before measurements and were instructed to avoid caffeine, smoking, alcohol, and exercise for at least 30 minutes beforehand. Three BP measurements were obtained at each visit. 

Statistical methods

The trial was powered to detect a 5 mmHg between-group difference in SBP, assuming an intraclass correlation coefficient of 0.063 and 20% attrition.

Missing data were addressed using multiple imputation by chained equations, generating 40 imputed datasets. The imputation model incorporated demographic variables, baseline BP, BMI, hypertension history, and cardiovascular comorbidities.

Sensitivity analyses included tipping-point analyses in which progressively worse systolic BP values were imputed into the intervention arm to determine the threshold at which statistical significance would be lost.

Funding and oversight

The study was primarily funded by the National Heart, Lung, and Blood Institute, with additional support from other NIH institutes. Oversight was provided by an independent Data and Safety Monitoring Board, and the protocol was approved by the Tulane University Institutional Review Board. 

Results

A total of 1,272 participants from 36 Federally Qualified Health Center clinics across Louisiana and Mississippi underwent randomization. 642 were assigned to the multifaceted intervention strategy and 630 to enhanced usual care. 

The study population reflected the communities these clinics were designed to serve - the mean age was 58.8 years, nearly two-thirds of participants were African American, over three-quarters were unemployed, and almost three-quarters reported an annual family income below $25,000. More than 90% were already receiving antihypertensive medications at baseline, and many had lived with hypertension for over a decade.

Primary efficacy outcome 

Over 18 months, systolic blood pressure fell substantially in both groups, although the decline was greater in the intervention arm. Mean systolic blood pressure decreased by 15.5 mmHg in the intervention group compared with 9.1 mmHg in the enhanced usual care group, yielding a net between-group difference of 6.4 mmHg. Most of this separation emerged within the first six months and was then largely sustained through follow-up. Diastolic blood pressure showed a similar pattern, with a net reduction of 3.4 mmHg favoring the intervention.

Secondary effectiveness outcomes

Secondary effectiveness outcomes also favored the intervention strategy. At 18 months, around 22% of patients in the intervention group and 15% in the enhanced usual care group achieved a systolic BP <120 mm Hg. Systolic BP <130 mm Hg was achieved in 48% and 36% of intervention and control patients, respectively. 18.7% of the intervention group versus 10.9% of the control group had a systolic BP reduction of >30 mm Hg from baseline. The mean reduction in diastolic blood pressure was −8.7 mm Hg in the intervention arm compared with −5.3 mm Hg in the control arm, with a between-group difference of −3.4 mm Hg. Changes in physical and mental SF-12 quality-of-life scores were similar between groups. 

The results were robust across sensitivity analyses. The complete case analysis, which included only patients with complete data, showed an even larger effect: -16.1 mmHg in the intervention group vs -9 mmHg in controls, yielding a between-group difference of -7.1 mmHg (95% CI, -9.7 to -4.4). The tipping point analysis (table S8) demonstrated that missing data would need to be inflated by +6.5 mmHg in the intervention group before the result lost statistical significance, confirming the robustness of the primary finding.

In the subgroup analysis, the intervention appeared to work similarly across age groups, in both men and women, among Black and non-Black participants, and across different levels of income, education, and insurance status. It also showed similar benefits in participants with and without cardiovascular disease, diabetes, or chronic kidney disease. The magnitude of reduction varied slightly between groups, however there was no clear evidence that the intervention worked substantially better or worse in any particular subgroup. Sensitivity analyses, including complete-case and tipping-point analyses, did not materially alter the primary findings.

Implementation outcomes

Implementation outcomes favored the intervention group throughout the trial. The trial showed that the fidelity scores (which reflect adherence to major intervention components like medication intensification, home BP monitoring, and health coaching) were consistently higher in the intervention clinics. Treatment intensification occurred in more than 90% of intervention participants compared with 71.5% in the control arm, and home blood pressure monitoring was nearly universal in the intervention group. Interestingly, self-reported medication adherence did not significantly differ between groups despite the greater blood pressure reduction observed in the intervention arm.

Another noteworthy point is that, over the course of follow-up, the rates of both provider and health coaching visit completion decreased. Health coaching visit completion reduced from 94.1% at baseline to 53.4% at 3 months and then fluctuated between 43% and 66% during later follow-up visits. A similar pattern was observed for provider visits, with completion rates falling from 87.1% at baseline to 52.2% at 3 months and remaining between 40% and 56% thereafter. 

Serious adverse events were seen in both groups with similar frequencies of around 20%. There wasn’t a significantly increased frequency of hypotension, syncope, falls, or kidney-related adverse events in the intervention arm.

Discussion

At first glance, this trial appears to be about blood pressure reduction, but the more one sits with the paper, the clearer it becomes that the intervention was not simply the intensification of drugs. It was about the restructuring of care. The discussion of any trial usually focuses on the trial methodology, the results, and if the “significant” values were actually significant. However, this trial gives us an entirely different direction of discussion. In a single word: implementation.

The investigators did not try out a novel drug or some breakthrough biologic pathway. They simply tried to redesign/reframe the rules surrounding hypertension management. BP was just measured more carefully, reviewed more frequently, discussed more consistently, and acted upon more aggressively. They coached, followed, and reminded the patients, and the patients were drawn into the process through home monitoring and repeated engagement. The study asked a question which seems deceptively simple: what happens when vulnerable patients receive the kind of organized hypertension care that guidelines assume already exists?

The answer? A 6.4 mmHg greater reduction in systolic blood pressure. This is not a trivial finding, more so because the population was already heavily treated at baseline. It’s also notable that the benefit was achieved in an underserved population frequently excluded from tightly controlled efficacy trials: predominantly low-income patients who receive care in resource-limited settings. These are places where factors such as therapeutic inertia, fragmented follow-up, and multiple other barriers dominate everyday practice more than drugs and doctors do.

Notably, IMPACT-BP’s impact isn't an isolated one. Previous community-based hypertension studies have shown that even relatively simple, culturally grounded interventions can produce meaningful improvements in BP control when it is accessible and persistent. The landmark “barbershop trial” among Black men (Victor RG et al, NEJM 2018) is perhaps the best example of this, where blood pressure reduction was achieved by combining trusted community spaces with pharmacist-led medication management. 

So we are led to believe that a multipronged strategy = BP lowering in low-resource settings. Case closed. And yet, the study becomes more interesting as we think further and deeper.

The control group also had nearly 9 mmHg reduction in systolic BP, and this points us towards a simple fact. Medical attention, by itself, is therapeutic. Standardized BP measurement, education of the healthcare providers, repeated follow-up, and participation in a trial most likely improved care even in those who didn't get the full intervention package. What does this mean? This makes the between-group difference more credible, but it also highlights how poorly measured BP, in conjunction with irregular follow-up, may contribute to apparent “treatment-resistant” hypertension in routine practice.
A key observation is that the intervention arm had meaningful BP reduction even though the medication adherence wasn’t significantly better. This disconnect raises huge questions. Was the scale used (Morisky Adherence Scale) simply too blunt to capture the real behavioral change? Did the medication intensification protocol overcome the not-so-good adherence? Or is self-reported adherence fundamentally unreliable? These are the socially vulnerable populations where medication-taking behaviour is fluid. Does this factor make adherence hard to measure? The trial does not fully answer this, but it shows that there are many factors apart from patient-reported drug adherence that lead to BP control.

This trial also exposes the persistent discord between evidence and implementation. Although the intervention was modeled after SPRINT-style intensive BP control, fewer than 1/4th of participants ultimately achieved a systolic BP below 120 mmHg. The authors suggest that providers may have hesitated to pursue such aggressive targets because there are quality benchmarks in Federally Qualified Health Centers that still emphasize less stringent thresholds. This, in itself, shows that there is a wide gap between what guidelines recommend, what doctors practice, and what is best for any individual patient. While trials increasingly support lower BP targets, individualized BP goals will always eclipse population-based goals.

Although the results seem overwhelmingly favorable towards the intensive strategy, we have to be cautious about certain points. This was a bundled intervention, making it impossible to determine which component carried the greatest weight. Was it home BP monitoring or the health coaching? Audit and feedback? Medication intensification? Or simply more human contact? The intervention succeeded as a package, but the ability to scale it may depend on knowing which pieces are essential and which are merely supportive. At the end of the day a more selective approach may be more cost-effective and generalizable.

From a nephrologist’s perspective, there was a small proportion of participants with CKD in the trial. This limits direct extrapolation to higher-risk CKD populations.  Also, hypertension management in CKD is often complicated by volume overload, polypharmacy, fluctuating kidney function, and competing cardiovascular risks. Although the subgroup analyses were consistent, the study was not specifically designed to answer whether these implementation strategies perform similarly in patients with advanced kidney disease.

Whatever is said and done, the crux of hypertension management is the drugs we use to control BP. Interestingly, there are no data regarding the choice and sequence of drugs used and how many percent of the trial population that received each drug, and at what dose. That would have added more scientific rigor and guidance from this implementation study.

Another issue is with the sustainability of this package. This intervention was done within the structured plan of a fully funded clinical trial, and yet the provider and health coach visit completion rates significantly declined over time. The average implementation cost approached $760 per participant, which may appear reasonable from a public health perspective, but  is pretty challenging for under-resourced primary care systems already operating under financial strain. Finally, this study was only 18 months, what kind of intensity and expense would be expected over potentially decades (a lifetime) of treatment? The trial proves that intensive hypertension management can work in low-income populations. Whether health systems are willing to invest in maintaining that intensity is a separate question entirely.

Conclusion

IMPACT-BP  showed that meaningful blood pressure reduction is possible in underserved populations when provided with structured, consistent, and team-based care. At the same time, it reminds us that sustaining this level of intensity of care outside a clinical trial remains challenging. 

Summary by
Akshaya Jayachandran
Cristina Popa

Reviewed by
Brian Rifkin, Swapnil Hiremath

Header Image created by AI, based on prompts by Cristina Popa