Semaglutide kidney trial misses its main MRI goals, but biopsies offer clues

Why a proven kidney drug still needed explaining

Chronic kidney disease develops in about 40% of people with type 2 diabetes, and diabetes is the leading cause of kidney failure worldwide.[1] In the FLOW trial, published in 2024, 3,533 people with both conditions were randomized to weekly semaglutide or placebo. Over a median 3.4 years, the risk of a major kidney disease event was 24% lower with semaglutide, 331 first events versus 410.[4]

Those events were kidney failure, death from kidney or cardiovascular causes, or a fall of at least 50% in estimated glomerular filtration rate (eGFR), the standard blood-test estimate of how well the kidneys filter.[4] Based on FLOW, the FDA approved Ozempic in January 2025 to reduce the risk of sustained eGFR decline, end-stage kidney disease and cardiovascular death in adults with type 2 diabetes and chronic kidney disease.[5][7]

What FLOW could not show is how. Semaglutide lowers blood sugar and body weight, but analyses of GLP-1 drug trials suggest those effects only partly explain the kidney benefit.[1] GLP-1 receptors, the docking sites the drug acts on, have been found in the small arteries and arterioles of human and animal kidneys, which leaves room for a direct effect on the kidney itself.[1]

What REMODEL measured

REMODEL was sponsored by Novo Nordisk, which makes semaglutide, and built as a mechanistic companion to FLOW.[1][2] Between April 2021 and October 2023, 106 adults with type 2 diabetes and chronic kidney disease at 28 sites in 8 countries were randomized two to one: 71 to weekly semaglutide injections and 35 to a matching placebo, for 52 weeks. The trial was double-blind, so neither participants nor investigators knew who got which.[1][2][3]

Their average age was 65 and their average eGFR was 51 ml/min per 1.73 m². The median urine albumin-to-creatinine ratio (UACR), a measure of protein leaking into urine, was 187 mg/g. Only 25 of the 106 were women.[1]

The trial scheduled multiparametric MRI, several kidney scans in one session, at baseline and the end of treatment. Thirty-three participants (22 on semaglutide, 11 on placebo) also joined a biopsy group, scheduled to give kidney samples at the start and end of treatment. Joining required an eGFR of at least 40.[1][3]

That tissue went through two newer techniques. Single-nucleus RNA sequencing (snRNAseq) reads which genes are switched on inside individual cell nuclei, so researchers can see how each cell type responds. Spatial transcriptomics reads genes in an intact tissue slice, preserving which cells sit next to which.[1]

The main MRI tests came up empty

The week-52 MRI comparisons below are model estimates that assume participants stayed on their assigned treatment and did not start medications the trial prohibited. Missing follow-up measurements, and values excluded under those rules, were filled in using observed data from the same treatment group, a statistical process called multiple imputation.[1]

The three coprimary outcomes were kidney oxygenation, read from an MRI signal called R2*, global kidney perfusion, meaning blood flow relative to kidney size, and T1 mapping, used as an indicator of inflammation. None differed significantly between semaglutide and placebo after 52 weeks.[1]

R2* in the kidney's outer layer was 2% lower with semaglutide, which points toward better oxygenation, and perfusion was about 10% higher. In both cases the 95% confidence interval included no difference (70 versus 35 and 65 versus 33 people analyzed). T1 did not change.[1][2]

The authors offer reasons. They sized the trial from short-term studies that showed large oxygenation swings and now say those assumptions "may have been optimistic." R2* is also sensitive to hydration and sodium balance, and T1 blends inflammation, perfusion and scarring, so improvements in one could cancel out another.[1]

Where the signals showed up

Two secondary MRI measures moved. The renal artery resistive index, a measure of resistance to blood flow in the kidney's vessels, fell 4% relative to placebo (95% confidence interval 1% to 7% lower, 71 versus 34 analyzed).[1]

The apparent diffusion coefficient (ADC) tracks how freely water moves through tissue. Scarring, or fibrosis, restricts that movement, so a falling ADC signals more scarring. Cortical ADC ended 5% higher with semaglutide than with placebo (interval 1% to 9%, 66 versus 32 analyzed), which the authors interpret as fibrosis that stopped advancing.[1]

Kidney fat changed far more than body weight. In an exploratory analysis of 34 people (22 semaglutide, 12 placebo), fat around the kidney was 25% lower and fat in the renal sinus 13% lower than with placebo. Body weight fell about 5% more than with placebo, or 4.45 kg.[1] The authors suggest losing organ fat could ease pressure on blood vessels, a link the trial did not test directly.[1]

Familiar kidney numbers went the expected way. In an exploratory analysis, UACR fell 40% versus placebo (71 versus 35 analyzed). Creatinine clearance, a filtering measure taken from a full day's urine collection, was 12 ml/min higher (67 versus 34 analyzed). The eGFR difference, 3.4 ml/min per 1.73 m² in semaglutide's favor, was not statistically significant.[1]

What the biopsies showed

Under the microscope, the inner layer of the most diseased arterioles made up a smaller share of the vessel with semaglutide than with placebo after 52 weeks, an exploratory structural finding.[1]

The gene data pointed in a similar direction. The analysis used 22 matched before-and-after biopsy pairs (14 semaglutide, 8 placebo). Semaglutide changed gene activity in 7 of the 31 cell clusters analyzed. The most responsive were two populations of glomerular endothelial cells, which line the glomeruli, the kidney's tiny filtering units. Most of their affected genes tied to metabolic stress, inflammation and fibrosis were turned down.[1]

The cell count did not change, which the authors read as a change in gene activity rather than numbers. But when the team checked whether the changed genes formed recognized biological pathways, none passed the statistical bar after correction for multiple testing.[1]

An exploratory spatial analysis of 13 paired samples (8 semaglutide, 5 placebo) found a smaller share of natural killer and natural killer T immune cells near those endothelial cells after treatment. It treated placebo and pre-treatment biopsies as one untreated group, so it was not a clean randomized comparison.[1]

How far the findings stretch

No formal hypothesis testing, multiplicity adjustment or testing hierarchy was planned for the three coprimary MRI outcomes. The P values for the many secondary and exploratory outcomes are nominal, and the authors write that they "should therefore be interpreted with caution." With so many outcomes examined, some could look positive by chance.[1]

The links among imaging, tissue and clinical results are associations, because no mediation analysis was done. Filtration was not measured directly with a tracer such as iohexol, and the small number of women limits what the trial can say about them. REMODEL tracked mechanisms for a year and was never designed to show fewer kidney failures. That evidence comes from FLOW.[1][4]

Is semaglutide hard on the kidneys?

REMODEL's safety reporting covered up to 57 weeks, with 52 weeks of planned treatment and five weeks of follow-up.[1][2] Its safety numbers are informative but too small for firm comparisons. Adverse events occurred in 36 of 71 people on semaglutide (50.7%) and 17 of 35 on placebo (48.6%). Serious adverse events affected 14 of 71 (19.7%) versus 5 of 35 (14.3%), which the authors describe as similar rates. Treatment was permanently stopped because of adverse events by 9 of 71 on semaglutide (12.7%) versus 1 of 35 (2.9%).[1]

ClinicalTrials.gov reports treatment-emergent adverse events among randomized participants who received at least one dose, defining these as events beginning from the first dose through permanent treatment discontinuation. Within its reporting window of up to 57 weeks, serious acute kidney injury affected 4 of 71 semaglutide participants and none of 35 on placebo, plus one case of end-stage kidney disease among the 71 on semaglutide.[2] During the trial's treatment and follow-up period, serious events judged possibly related to treatment affected one semaglutide participant, and no biopsy-related serious adverse events were reported.[1]

FLOW remains the larger reference. Its posted registry results assessed safety in all randomized participants during their in-trial observation, from randomization to follow-up, withdrawal, last contact or death, over a reporting window of up to 238 weeks. Serious adverse events affected 877 of 1,767 semaglutide participants (49.6%) and 950 of 1,766 placebo participants (53.8%).[8] The Ozempic label warns of postmarketing reports of acute kidney injury, mostly in patients who became dehydrated from nausea, vomiting or diarrhea.[5]

Who paid for the trial

Novo Nordisk funded REMODEL. Seven of the 15 authors are current or former Novo employees who own company shares, Novo staff provided editorial writing help, and several academic authors report consulting or grant ties to Novo and other drugmakers.[1] Outside researchers can request patient-level data, but access needs approval from both Novo Nordisk and what the paper calls an internal independent review panel.[1]

What would settle the question

The authors write that mechanisms of kidney protection "may include reduced vascular resistance, prevention of fibrosis and improved underlying molecular programs promoting endothelial cell health."[1] Testing that, they say, requires adequately powered mediation analyses, possibly using FLOW's larger dataset. Neither the paper nor the trial record gives a timetable.[1][2]

The authors also identify combination therapy with SGLT2 inhibitors, another diabetes drug class studied for kidney protection, as a research direction. Their small post hoc analysis was not powered to establish a subgroup difference.[1]

FLOW established that semaglutide reduces major kidney events in people with type 2 diabetes and chronic kidney disease. REMODEL points the search for the reason toward the kidney's blood vessels and the cells lining its filters. Whether those changes produce the benefit is a question for a trial powered to test it.[1][4]

Frequently asked questions

Sources and further reading

1. Effects of semaglutide on kidney disease in type 2 diabetes: a randomized placebo-controlled trial (Nature Medicine, 2026). DOI: 10.1038/s41591-026-04674-2
2. REMODEL trial record and posted results, NCT04865770 (ClinicalTrials.gov)
3. Rationale, design and baseline characteristics of REMODEL (Nephrology Dialysis Transplantation, 2025). DOI: 10.1093/ndt/gfaf114
4. Effects of Semaglutide on Chronic Kidney Disease in Patients with Type 2 Diabetes (FLOW, NEJM 2024). DOI: 10.1056/NEJMoa2403347
5. Ozempic (semaglutide) injection prescribing information (DailyMed)
6. Wegovy (semaglutide) injection and tablets prescribing information (DailyMed)
7. Novo Nordisk: FDA approves Ozempic to reduce kidney disease risk in type 2 diabetes and CKD (28 January 2025)
8. FLOW trial record and posted safety results, NCT03819153 (ClinicalTrials.gov)

Related reading

• FDA moves to permanently close bulk compounding of GLP-1 drugs: what the proposed rule means
• Roche reports Phase 2 enicepatide results in adults with type 2 diabetes