Semaglutide and kidney disease: what FLOW’s 24% reduction means

The question

Semaglutide has evidence for reducing important kidney-related outcomes in adults with type 2 diabetes and chronic kidney disease. The central FLOW finding is a reduction in a combined outcome, rather than proof that every participant avoided kidney failure or recovered lost kidney function.

That distinction matters when a headline compresses a clinical trial into “24% lower risk.” To understand the result, keep the population, the events counted and the type of percentage together.

The question FLOW was designed to answer

FLOW randomized 3,533 adults with type 2 diabetes and chronic kidney disease to injectable semaglutide or placebo. Eligibility included specified kidney-filtration and urine-albumin ranges. Median follow-up was 3.4 years; the trial stopped early after a prespecified interim analysis. Its primary outcome combined kidney failure, a large sustained decline in estimated filtration, and death from kidney or cardiovascular causes. Novo Nordisk funded the trial. Original FLOW report.

The official FLOW registry records the trial's design and outcome definitions. “Kidney disease” here is not shorthand for anyone worried about future kidney health. The eligibility criteria defined a particular clinical population, and the findings need to be read in that setting.

Imagine two hypothetical readers: one has the diagnoses and risk profile studied in FLOW; the other has normal kidney tests and no diabetes. Giving both the same risk-reduction promise would discard the very information the trial used to decide whom to study. Evidence in the first setting can motivate a question about the second, but it does not answer it automatically.

Relative reduction and percentage points are different

The FDA's Ozempic label reports a first primary-outcome event in 331 of 1,767 semaglutide participants (18.7%) and 410 of 1,766 placebo participants (23.2%). The hazard ratio was 0.76, with a 95% confidence interval of 0.66–0.88. FDA label, kidney-outcomes table.

The widely quoted 24% comes from subtracting the hazard ratio from one. A hazard ratio incorporates when events occur during follow-up. It is not the same statistic as subtracting the two percentages.

That subtraction gives 4.5 percentage points: 23.2 minus 18.7. This is a descriptive difference between the reported proportions over the observed trial, not a fixed-year survival estimate or a personalized prediction. It should not be restated as “24 fewer kidney failures per 100 people.” That sentence changes both the arithmetic and the endpoint.

Nor should the separate components simply be added together. A composite based on time to first qualifying event counts a person when the first such event occurs; a person may subsequently experience another component. Reading the component definitions prevents double-counting and helps explain what the combined number represents.

“Kidney benefit” does not mean every component changed equally

A combined endpoint allows investigators to analyze a defined set of clinically important events together. It does not establish that each event occurs equally often, has the same importance to every patient, or responds to treatment by exactly the same proportion.

For a hypothetical reader chiefly concerned about dialysis, the combined result is relevant background, but the component-level evidence is the more specific question. For another reader, cardiovascular death may be especially relevant. Neither interest can be fully answered by replacing the detailed endpoint with a slogan such as “kidney protection.”

This is why a good trial summary retains cardiovascular death in the definition even when discussing a kidney-focused study. Leaving it out would change what the headline result appears to measure.

New mechanism research asks a different question

On October 1, 2026, investigators published REMODEL, a 52-week randomized study of 106 participants with type 2 diabetes and chronic kidney disease. Its coprimary MRI measures did not differ significantly between semaglutide and placebo. Secondary imaging and molecular findings suggested possible pathways of benefit; the analyses did not apply multiplicity adjustments. Original REMODEL publication.

The contrast is useful. FLOW asked about clinical events; REMODEL investigated how an effect might arise. Interesting secondary findings do not turn nonsignificant primary measurements into positive ones. Equally, an unresolved mechanism does not erase a separately measured clinical-outcome result.

For readers following new studies, label each result before combining it with the older evidence: clinical outcome, primary mechanistic measure, or secondary hypothesis-generating observation. The labels keep an evolving research story from becoming a claim that all questions are settled.

The approved product and the safety context

As reviewed October 4, 2026, Ozempic's US labeling includes reducing sustained eGFR decline, end-stage kidney disease and cardiovascular death in adults with type 2 diabetes and chronic kidney disease. The same label warns about acute kidney injury associated with volume depletion, including in the setting of gastrointestinal reactions. A long-term benefit in a defined population and a treatment-related safety risk can coexist. FDA prescribing information.

This article does not determine whether a particular person should use a medicine or change treatment. It also does not transfer the studied formulation's evidence to a research-material vial. Our compounded-semaglutide explainer addresses a separate product-identity question; our weight-regain article concerns what happens after treatment withdrawal.