How to Read the SGLT2 Inhibitor Kidney Trials

The kidney trials of sodium-glucose cotransporter 2 inhibitors changed practice because they showed fewer major kidney events, not merely a small improvement in glucose. DAPA-CKD and EMPA-KIDNEY extended that evidence to chronic kidney disease populations that included many people without diabetes.

The headline is strong, but a patient-level decision still requires translation. What you have to know is who entered the trials, how kidney progression was defined, why estimated glomerular filtration rate can dip at first, and how a relative benefit becomes an absolute one at a particular baseline risk.

Key points#

• DAPA-CKD and EMPA-KIDNEY found fewer kidney-progression or cardiovascular-death events with an SGLT2 inhibitor than with placebo.
• Benefit was not limited to glucose lowering and extended to participants without diabetes.
• Composite outcomes combine events of different severity, so each component should be inspected.
• A small early eGFR dip is pharmacologically expected and differs from sustained chronic decline.
• Trial eligibility, albuminuria, eGFR, underlying diagnosis, concurrent therapy, and adverse-event risk shape applicability.

Start with the populations, not the drug names#

DAPA-CKD enrolled 4,304 adults with an eGFR of 25 to 75 mL/min/1.73 m² and a urinary albumin-to-creatinine ratio of 200 to 5,000 mg/g. Most participants were receiving a renin-angiotensin system inhibitor unless they could not tolerate one. About two thirds had type 2 diabetes, leaving a substantial prespecified group without diabetes.

EMPA-KIDNEY enrolled 6,609 adults. People with an eGFR from 20 to below 45 could enter regardless of albuminuria; those with an eGFR from 45 to below 90 needed a urinary albumin-to-creatinine ratio of at least 200 mg/g. More than half did not have diabetes. This broader structure added evidence for lower levels of albuminuria and a wider range of kidney disease.

These entry criteria are part of the result. They do not automatically cover type 1 diabetes, dialysis, kidney transplantation, pregnancy, acute kidney injury, or every uncommon glomerular disease. Both trials excluded polycystic kidney disease, and immunosuppression-related criteria limited some conditions. Applicability should be checked against the actual protocol, not inferred from the phrase “chronic kidney disease.”

Decode the primary composites#

DAPA-CKD's primary outcome combined a sustained eGFR decline of at least 50%, end-stage kidney disease, or death from a kidney or cardiovascular cause. Over a median 2.4 years, this occurred in 9.2% of participants assigned dapagliflozin and 14.5% assigned placebo. The hazard ratio was 0.61, and the absolute difference over the trial period was about 5.3 percentage points.

EMPA-KIDNEY combined kidney disease progression or cardiovascular death. Kidney progression included end-stage kidney disease, a sustained eGFR below 10, a sustained decline of at least 40%, or death from a kidney cause. Over a median 2.0 years, the primary outcome occurred in 13.1% with empagliflozin and 16.9% with placebo, a hazard ratio of 0.72 and an absolute difference of about 3.8 percentage points.

Those absolute differences are not universal promises. A hazard ratio compares event rates over time and is not identical to a risk ratio at a fixed date. The absolute benefit you can expect changes with follow-up duration and starting risk. A person with faster expected progression can have a larger absolute benefit from the same proportional effect than someone at low short-term risk.

Composite outcomes improve efficiency, but you should inspect the parts. In both trials kidney progression drove much of the primary contrast, and cardiovascular death alone was not clearly reduced in EMPA-KIDNEY, so a composite should not be summarized as though every component improved by the same amount.

Why results without diabetes matter#

SGLT2 inhibitors were developed as glucose-lowering drugs. The kidney findings in people without diabetes, and the relatively modest relationship between glucose change and kidney benefit, support mechanisms beyond glycemic control. Proposed pathways include altered tubuloglomerular feedback, lower intraglomerular pressure, natriuresis, and changes in energy handling and inflammation.

Mechanism can strengthen plausibility, but the randomized clinical result should carry the claim. DAPA-CKD showed a broadly consistent effect by diabetes status. EMPA-KIDNEY also found benefit across diabetes categories. That does not imply equal precision in every underlying kidney diagnosis or every albuminuria subgroup. Subgroups contain fewer events, and an apparently smaller or larger estimate may reflect chance. So the useful subgroup question is whether there is credible evidence of interaction, supported by biology and replicated data, not whether one subgroup's confidence interval crosses 1.0 while another's does not.

The eGFR curve has two phases#

Soon after an SGLT2 inhibitor starts, eGFR often falls modestly. This acute hemodynamic change reflects altered pressure within the glomerulus and is conceptually different from structural kidney injury. After that initial phase, the chronic eGFR slope is generally flatter than with placebo, indicating slower long-term loss of filtration.

A single early laboratory value can therefore be misleading if read without timing and clinical context. Trial reports often separate the total slope, which includes the initial dip, from the chronic slope after the early period. The chronic slope can reveal benefit even when few participants have yet reached a hard kidney outcome, but it remains a surrogate summary. Dialysis, transplantation, sustained major eGFR loss, hospitalization, symptoms, and survival are more directly consequential.

An expected pattern is not permission to ignore a large or progressive change. Volume depletion, acute illness, interacting medicines, urinary obstruction, and other causes of acute kidney injury still require assessment. Trial-level education cannot determine what a particular laboratory result means.

Limits of cross-trial comparison#

CREDENCE studied canagliflozin in people with type 2 diabetes, albuminuric chronic kidney disease, and background renin-angiotensin blockade. It established kidney benefit in a high-risk diabetic population before DAPA-CKD and EMPA-KIDNEY broadened the picture.

The three trials differ in eGFR thresholds, albuminuria, diabetes status, primary definitions, follow-up, and trial drug. Similar direction across these programs supports a class-related kidney effect, but direct comparisons of hazard ratios are unreliable. A smaller hazard ratio in one trial does not prove its drug is superior. Different baseline risks, outcome definitions, and event ascertainment can create different estimates without a true drug-to-drug difference.

Read harms with the same denominator#

SGLT2 inhibitors can increase genital mycotic infections and contribute to volume depletion. Diabetic ketoacidosis can occur, sometimes with glucose levels lower than expected, particularly around fasting, acute illness, surgery, or reduced insulin. Risk differs greatly by population, which is one reason trial eligibility matters.

In DAPA-CKD, serious adverse events and discontinuation due to adverse events were not higher overall with dapagliflozin. In EMPA-KIDNEY, serious urinary infection, hyperkalemia, acute kidney injury, dehydration, and bone fracture were broadly similar between groups; ketoacidosis was uncommon but numerically more frequent with empagliflozin. Rare-event estimates remain imprecise even in large trials. What you want is absolute events in the randomized groups over the same follow-up, not a list of possible effects from one group alone, and a check on whether the events were actively solicited, adjudicated, and reported after treatment stopped.

A practical appraisal sequence#

When you read an SGLT2 kidney trial, ask:

1. Do the person's eGFR, albuminuria, diagnosis, diabetes status, and concurrent therapy resemble trial eligibility?
2. What exactly counts in the primary composite, and which components account for the difference?
3. What are the absolute event rates and follow-up duration?
4. Is the eGFR result an acute change, chronic slope, or sustained threshold event?
5. Were subgroup findings prespecified, adequately powered, and supported by interaction tests?
6. Which harms were adjudicated, and which higher-risk groups were excluded?
7. Did early stopping or incomplete follow-up limit long-term precision?

DAPA-CKD stopped early on a monitoring recommendation after a clear efficacy boundary was met. That supports a valid finding under the design, but shorter follow-up leaves less information about durability and late harms. EMPA-KIDNEY also ended as planned under its event-driven design with about two years of median follow-up. Continuing observational follow-up can add information, but it no longer preserves every advantage of the randomized comparison once assigned treatment changes.