Semaglutide protects the kidneys of people with type 2 diabetes. That much was settled in 2024 when the FLOW trial cut the risk of its main kidney outcome by 24%, according to the New England Journal of Medicine paper the new study cites. What nobody could explain was how. A trial published in Nature Medicine on Oct. 1 tried to look inside the kidney to find out, and its answer is more interesting than its scorecard.

The scorecard first: the study missed all three of its main goals.

What REMODEL Set Out to Do

The REMODEL trial, led by Katherine Tuttle, Petter Bjornstad and Matthias Kretzler with colleagues at sites in eight countries, randomized 106 adults with type 2 diabetes and chronic kidney disease to semaglutide 1 mg weekly or placebo for 52 weeks, two to one. Participants averaged 65 years old, and nearly all, 99.1%, were already taking drugs that block the renin-angiotensin system. About 40% also took an SGLT2 inhibitor.

It was built as a companion to FLOW. Instead of measuring kidney failure, it measured what semaglutide does inside the organ, using MRI scans, kidney biopsies in 33 participants, and gene-activity profiling of individual kidney cells before and after treatment.

The Main Endpoints Did Not Move

The three coprimary outcomes were kidney oxygenation, global blood flow and a scan marker of tissue inflammation. None differed significantly between semaglutide and placebo at 52 weeks, the authors report. Oxygenation and perfusion moved in the hoped-for direction numerically, but not enough to count.

The authors say why. Their sample size rested on expert guesses about how much oxygenation would change, because no earlier long-term trial had measured it in this population, and later evidence suggests those guesses were optimistic. They write that the study was probably underpowered for modest effects on that measure.

What Did Move

Several secondary and exploratory outcomes did. Renal artery resistive index, a measure of vessel stiffness, fell with semaglutide (ratio 0.96, P = 0.008). A diffusion scan marker that tracks fibrosis stayed stable instead of worsening, which the authors read as prevention of progression. Creatinine clearance in 24-hour urine rose by 12 ml per minute against placebo.

The exploratory results were striking. Albuminuria, protein leaking into urine, fell 40% against placebo. Fat around the kidney dropped 25% and fat in the kidney sinus 13%, which is more than weight loss alone would predict, since average body weight fell about 4.5 kilograms. The authors suggest less fat may reduce pressure on kidney vessels.

One caution applies to all of this. The trial had no prespecified testing hierarchy and made no adjustment for running many comparisons, so the nominal P values should be read loosely, as the authors themselves say.

Inside the Biopsies

The biopsy subgroup is where the paper earns its place. Among the most diseased arterioles, the fraction of the vessel wall thickened by disease fell by about 10 points with semaglutide (P = 0.032, exploratory). The tissue around glomeruli, the kidney's filters, showed less immune-cell inflammation.

Gene-activity analysis of 22 paired biopsies pointed to glomerular endothelial cells, the lining of the filter's small vessels, as the most responsive cell type. Genes tied to metabolic stress, inflammation and scarring were mostly turned down. In a smaller spatial analysis of 13 paired samples, fewer natural killer cells sat near those endothelial cells after treatment. The picture is of endothelium getting less stressed and less crowded by immune cells.

The numbers are small, and the authors call the links among imaging, tissue and clinical findings associative, not causal. That is the same distinction that shaped the debate in Blogerroom's piece on the cannabis and violence review, where an association looked bigger than it was once study design was accounted for.

The SGLT2 Hint and Other Limits

An intriguing post hoc result: in participants who also took an SGLT2 inhibitor, semaglutide was linked to a gain in eGFR, a standard measure of kidney filtration, of 8.1 ml per minute versus placebo. Without an SGLT2 inhibitor, the gain was 0.2. Subgroups this small can mislead, and the authors say the study was not powered to compare them, but it makes a case for testing combinations.

Other limits matter. Only 25 of 106 participants were women. Nine of 71 people on semaglutide stopped treatment because of side effects, against one of 35 on placebo. The paper did not directly measure filtration with gold-standard tracer tests. And data requests go through Novo Nordisk, the maker of semaglutide.

What It Means

My view: this trial should not be read as new proof that semaglutide works on the kidneys. FLOW did that. Its value is a testable map of how, running through vessels, fat and endothelial cells rather than just blood sugar and weight. A missed main endpoint in a mechanistic study is not a failed drug. It is a reminder that the biology is messier than the design assumed.

That fits a pattern in drug research, as in the mesothelioma drug that targets a tumor's own defenses, where understanding the mechanism often matters as much as the response rate. The next step is a larger mechanism trial with enough power, and a close look at semaglutide combined with SGLT2 inhibitors.

Anyone with diabetes and kidney disease should discuss treatment with their own doctor. This article is general information, not medical advice.