Lancet Diabetes Endocrinol. 2025 Oct 10:S2213-8587(25)00222-0. doi: 10.1016/S2213-8587(25)00222-0. Online ahead of print.
Effects of empagliflozin on conventional and exploratory acute and chronic kidney outcomes: an individual participant-level meta-analysis
William G Herrington, Zhaojing J Che, Rebecca Sardell, Alistair J Roddick, Parminder K Judge, Christoph Wanner, Sibylle J Hauske, Stefan Anker, Javed Butler, Gerasimos Filippatos, Milton Packer, Faiez Zanna, Dominik Steubl, Martina Brueckmann, Jennifer B Green, Jonathan R Emberson, Martin J Landray, Colin Baigent, Richard Haynes, Natalie Staplin
PMID: 41082289
It’s rare in nephrology to watch a paradigm form in real time. Empagliflozin debuted as a glucose-lowering agent and quietly began rewriting renal physiology. Each trial completion revealed another fragment of the story: fewer deaths, heart failure hospitalizations, slower eGFR loss (EMPAREG, Zinman et al, NEJM 2015; EMPEROR-Reduced Packer et al, NEJM 2020), and finally fewer dialysis starts (EMPA-KIDNEY Herrington et al, NEJM, 2023 | NephJC Summary). The data are compelling, and added sequentially larger swathes of population that received empa-benefits: diabetes with high cardiovascular risk, HFrEF with or without diabetes, HFpEF with or without diabetes, and finally CKD with or without diabetes or with or without proteinuria, akin to different dialects of the same disease. (and yes, this benefit transcends empa - the flozination seems quite clearly a class effect). For a decade, we argued about eGFR dip, albuminuria thresholds, and who “deserved” flozination. Science moved faster than our understanding. The individual participant data meta-analysis (IPD-MA) is not the conclusion of flozins in the CKD book… it is the inspiration for yet another sequel. It pulled every patient from 4 empa trials, forcing the question we still cannot answer: how do flozins preserve the kidney beyond what we can measure? (Herrington WG, et al, 2025)
Why is this individual participant data meta-analysis (IPD-MA) necessary?
A meta-analysis, including 13 RCTs, showed a 38% relative risk reduction of CKD progression with SGLT2i compared with placebo (relative risk 0.63, 95% CI 0.58–0.69). (Nuffield Department of Population Health, Renal Studies Group; SMART-C. Lancet. 2022) The main findings were based on a composite categorical outcome combining kidney failure with ≥40% or ≥50% sustained declines in eGFR. However, the benefit remains uncertain in patients without diabetes, low albuminuria, or slowly progressing CKD. This, in part, is due to the fact that some of the included trials were stopped earlier, mostly due to meeting efficacy endpoints overwhelmingly at interim analyses (DAPA-CKD, EMPA-KIDNEY, CREDENCE), or occasionally due to lack of funding (SOLOIST-WHF).
Metaphorically speaking, each empagliflozin trial was a continent on its own. They differed in inclusion criteria, definitions, and follow-up. An aggregate meta-analysis, built from the averages, can blur important details. The IPD-MA achieved this by re-analyzing 23,340 participants with harmonized definitions of acute kidney injury, chronic kidney disease progression, and kidney failure, thereby eliminating inter-trial variability and clarifying the underlying biological effect.
Table 1. Baseline characteristics of participants from the 4 included trials, from Herrington WG, et al, 2025
Access to participant level data allowed a precise assessment of whether diabetes status, albuminuria, baseline eGFR, or the early eGFR dip modified response, and whether empagliflozin’s renoprotective action was consistent across kidney disease phenotypes.
How was the individual participant data meta-analysis built?
Harmonization of endpoints
AKI: ≥50% serum creatinine increase confirmed in consecutive samples.
CKD progression: sustained ≥40% decline in eGFR or onset of kidney failure or renal death.
Kidney failure: chronic dialysis >90 days, kidney transplantation, or sustained eGFR <10 mL/min/1.73m².
Adapted from web figure 2. Additional details on kidney outcome definitions, from Herrington WG, et al, 2025
Modeling kidney function over time: eGFR slopes
Chronic slope: eGFR decline from end of early dip (≈4–8 weeks) to final follow-up - primary measure.
Total slope: eGFR change from randomization to final visit- secondary measure.
Off-treatment (dip-free) slope: eGFR change from baseline to post-discontinuation value - exploratory, tests reversibility.
Figure from supplemental methods, illustrating eGFR slope definitions, from Herrington WG, et al, 2025
Joint longitudinal survival models
These statistical models analyzed two related processes simultaneously. First: the longitudinal trajectory of the eGFR over time (how quickly kidney function declines), and secondly, the time-to-event outcome (when a participant reaches kidney failure or dies). By modeling them together, the analysis recognizes that faster eGFR decline increases the probability of kidney failure, and both reflect the underlying disease activity. This approach provides a more accurate estimate of treatment effect than analyzing eGFR slopes and failure events separately.
Pooling and weighting
Each trial yielded its own log-hazard ratio, combined with inverse-variance weighting: larger, event-rich trials (EMPA-KIDNEY, EMPA-REG) carried greater weight. This minimizes variance and produces a most stable pooled hazard ratio.
Robustness checks
Bootstrapping: repeatedly resampling the data confirmed result stability
Leave-one-out analysis: removing one trial at a time showed no single study drove the findings
Proportional-hazard testing: verified model assumptions.
Developing the acute dip prediction model
One persistent clinical question was whether the magnitude of the early eGFR dip modifies long-term benefit. A prediction model was built using only baseline data from participants randomized to empagliflozin.
The average predicted acute eGFR dip was ~5%, with modest variability (SD ≈2.4).
About half the participants had minimal or no dip (≥−5%), while only ~9% experienced a dip of >8%.
The largest dips occurred in EMPA-KIDNEY, reflecting inclusion of advanced CKD patients with lower baseline eGFR and higher albuminuria.
No evidence suggested these larger dips were harmful or reduced long-term benefit.
Table 1, continuation, from Herrington WG, et al, 2025
Predictors of higher dip: higher age, higher eGFR, diuretic use, RASi use, and BMI.
Slides from Natalie Staplin’s presentation- Predicting acute eGFR dip, Late-Breaking Clinical trial session, 62nd European Renal Association Congress
Results
In this IPD meta-analysis, outcomes were analyzed on two complementary scales: categorical events (yes/ no endpoints like AKI or kidney failure) and continuous trajectories (the chronic eGFR slope)
Chronic eGFR slope as a dynamic measure of CKD progression
After the early hemodynamic dip (first 4-8 weeks), the chronic slope reflects the true rate of renal decline over the years.
A slower chronic slope indicates slower structural CKD progression
In the empa group, the annual eGFR loss was reduced by 1.4-1.7 mL/min//1.73m² per year vs placebo: a 64% slower decline.
This slope-based effect parallels and predicts the reduction seen in categorical outcomes such as CKD progression and kidney failure, both capturing the same biological process at different resolutions.
Table 2. Kidney outcomes assessed and the effect of empagliflozin in the four combined trials, from Herrington WG, et al, 2025
Categorical outcomes are fewer but clinically concrete
Events like “≥50% creatinine rise’, “CKD progression”, and “kidney failure” are binary and less frequent. They define milestones in the continuum of disease captured continuously by the slope. The beneficial effects persisted even with/ no diabetes, irrespective of the degree of albuminuria.
Figure 2. Effects of empa on chronic kidney outcomes, from Herrington WG, et al, 2025
Relative vs absolute effects
Relative effects (ie, HR 0.70= 30% lower risk) express proportional risk reduction and remain stable across populations.
Absolute benefit depends on baseline event risk: patients with more advanced CKD or higher albuminuria experience greater absolute gains because more events are prevented per unit of time. Thus, while the relative benefit was uniform, the clinical impact (absolute benefit) was greatest in high-risk groups.
Empagliflozin slowed chronic eGFR decline by two-thirds and proportionally reduced the risk of AKI, CKD progression, and kidney failure by 20–34%. Early AKI events were rare and similar between groups (0.1% each).
Figure 1. Effects of empa on acute kidney outcomes, from Herrington WG, et al, 2025
The observational BMI study: a reminder, not a rebuttal
A small retrospective study in non-diabetic glomerulopathies suggested SGLT2i lowers proteinuria only in overweight patients (BMI ≥25, but only 25% of the included patients had a BMI >25). (Vargas-Brochero MJ, et al, Nephrol Dial Transplant. 2025) Follow-up: six months. Endpoint: proteinuria. No control group, no adjudicated events, multiple uncontrolled confounders.
Its findings describe physiological variability, not necessarily efficacy. The IPD-MA has already proven event-level benefit independent of albuminuria, BMI, or dip size. The observational study illustrates the problem IPD meta-analysis solved: fragmentary, surrogate-based data can mislead, while harmonized, patient-level data reveal the real pattern.
Mechanism and meaning
After treatment withdrawal, eGFR trajectories in both arms ran parallel, with no evidence of rebound or catch-up loss in the former empagliflozin group (see also: longer term data from Herrington et al NEJM 2024 | NephJC Summary). The benefit in kidney function, therefore, persisted beyond the pharmacologic exposure, confirming that empagliflozin does not merely induce a transient hemodynamic shift but modifies long-term disease trajectory.
Figure 3. Effects of empa on chronic and off-treatment dip-free eGFR slope, from Herrington WG, et al, 2025
Empagliflozin’s renal protection cannot be confined to a hemodynamic frame. The drug reduces AKI, slows chronic eGFR decline, and benefits even low-albuminuric CKD. These effects most probably imply deeper biological integration: reduced tubular workload, improved oxygen use, and dampened inflammation. (Malijan et al, KI reports, 2025)
The hemodynamic dip is the visible surface of the unseen process- maybe a global resetting of nephron energetics (?).
Why does the methodology matter?
For us clinicians, the statistical terms have real-world translation:
Inverse-variance weighting: bigger trials speak louder, smaller ones whisper; no one distorts the message.
Joint modeling: measures disease trajectory, not just endpoints.
Chronic slope: filters out the early dip to show long-term truth.
Predicted-dip model: prevents bias from post-randomization sorting.
Together, these methods turn heterogeneous trials into a coherent clinical statement: empagliflozin protects kidneys broadly, predictably, and safely.
The unfinished story
The IPD-MPA is not the closing argument. It is the beginning of intellectual honesty in how we read SGLT2i data. It tells us the therapy works across phenotypes, but it also tells us we don’t completely understand why. The traditional lens (pressure, proteinuria, filtration) no longer suffices. Something else, perhaps metabolic or cellular, binds the signal together. Until its underlying mechanism is fully elucidated, this meta-analysis remains a precise quantification of this phenomenon that exceeds the explanatory reach of the current renal pathophysiology. Rather than closing the evidentiary loop, the IPD-MA redefines its limits, leaving one, central, unresolved question: What exactly does the kidney perceive in SGLT2 inhibition that we do not? There are limits to what clinical trials can reveal - sometimes we need to go from the bedside to the bench.
(*inspired by RADIANT session on
systematic reviews and meta-analyses)
Reviewed by Brian Rifkin and Swapnil Hiremath