Kidney evidence arrived here the long way around. Nobody set out to test these drugs on kidneys; the signal kept appearing as a secondary endpoint in trials built to count heart attacks, and it kept pointing the same direction. FLOW was the trial that finally asked the question on purpose, in people who already had chronic kidney disease, and it was halted early because the answer was clear enough that continuing would have been hard to justify. The nine papers around it trace how the suspicion accumulated, from liraglutide slowing nephropathy in 2017, through a dedicated trial in advanced disease, to the analysis of whether the benefit survives on top of an SGLT2 inhibitor.
Ranked papers
#1FLOW
Perkovic V, Tuttle KR, Rossing P, et al. · N Engl J Med · 2024
What it measured: Kidney failure, losing half of kidney function, or dying of kidney or heart disease
The only kidney trial in the class that was designed as one. 3,533 patients whose type 2 diabetes had already produced established kidney disease, given semaglutide 1.0 mg against placebo and halted at a median 3.4 years, because the benefit was unambiguous. The kidney endpoint fell 24%, major cardiovascular events 18%, and death from any cause 20%. It produced the January 2025 kidney indication for Ozempic, the first kidney indication in this class.
PMID 38785209 ↗NCT03819153 ↗DOI 10.1056/NEJMoa2403347 ↗
#2LEADER (renal)
Mann JFE, Ørsted DD, Brown-Frandsen K, et al. · N Engl J Med · 2017
What it measured: Protein appearing in urine, kidney function halving, kidney failure, or kidney death
The pre-specified renal analysis of LEADER, and early randomized evidence that a drug in this class slows diabetic kidney disease. Across 9,340 patients over a median 3.84 years, the renal composite came in at 5.7% against 7.2%, a 22% reduction. Look closely at what drove it: almost all of it was less new macroalbuminuria, which is protein appearing in urine, not kidneys failing. It is a real finding, on a softer endpoint than the name suggests.
PMID 28854085 ↗NCT01179048 ↗DOI 10.1056/NEJMoa1616011 ↗
#3AWARD-7
Tuttle KR, Lakshmanan MC, Rayner B, et al. · Lancet Diabetes Endocrinol · 2018
What it measured: Blood sugar at 26 weeks, with the rate of kidney decline as the key secondary
The first trial to take a GLP-1 into moderate-to-severe kidney disease, where most diabetes drugs are restricted or contraindicated. 577 adults at stage 3 or 4 were given dulaglutide or insulin titrated to target. Blood sugar control matched, but kidney function declined far less on the drug (0.7 against 3.3 mL/min over 52 weeks). Its value is as much practical as scientific: it established that these drugs can be used at all in patients who have run out of options.
PMID 29910024 ↗NCT01621178 ↗DOI 10.1016/S2213-8587(18)30104-9 ↗
#4REWIND (renal)
Gerstein HC, Colhoun HM, Dagenais GR, et al. · Lancet · 2019
What it measured: Protein appearing in urine, a sustained 30% loss of kidney function, or needing dialysis
REWIND's renal analysis ran over a median 5.4 years, the longest follow-up available. The composite fell from 19.6% to 17.1%, a 15% reduction, with new macroalbuminuria down 23%. What makes it distinctive is the population: 69% had never had a cardiovascular event, so this is the strongest kidney evidence in the class for people who are not already sick. It was an exploratory analysis, though, which is weaker footing than a pre-specified one.
PMID 31189509 ↗NCT01394952 ↗DOI 10.1016/S0140-6736(19)31150-X ↗
#5AMPLITUDE-O
Gerstein HC, Sattar N, Rosenstock J, et al. · N Engl J Med · 2021
What it measured: First major cardiac event, with a kidney measure as secondary
4,076 patients with cardiovascular or kidney disease, where the pre-specified renal composite fell 32%. The key detail is that 15% were already on an SGLT2 inhibitor, and the kidney benefit held anyway. That was an early randomized sign that the two drug classes can work together. The drug itself was discontinued before reaching market, so the finding outlived the product.
PMID 34215025 ↗NCT03496298 ↗DOI 10.1056/NEJMoa2108269 ↗
#6SURPASS-4 (kidney)
Heerspink HJL, Sattar N, Pavo I, et al. · Lancet Diabetes Endocrinol · 2022
What it measured: Losing 40% of kidney function, kidney failure or death, or new protein in urine
The kidney analysis from SURPASS-4, and the only such evidence for a dual GIP/GLP-1 agonist. 1,995 high-risk patients on tirzepatide against titrated insulin over a median 85 weeks, with the kidney composite 42% lower (hazard ratio 0.58), and kidney function declined at 1.4 rather than 3.6 mL/min a year. The comparator is insulin, not placebo, so this shows benefit relative to another treatment, not against none.
PMID 36152639 ↗NCT03730662 ↗DOI 10.1016/S2213-8587(22)00243-1 ↗
#7SUSTAIN-6 (nephropathy composite)
Marso SP, Bain SC, Consoli A, et al. · N Engl J Med · 2016
What it measured: First major cardiac event, with kidney disease planned as a secondary measure
The nephropathy composite inside SUSTAIN-6, eight years before anyone ran FLOW. Across 3,297 patients over 104 weeks it came in at 3.8% against 6.1%, a 36% reduction. This was the earliest randomized hint that semaglutide specifically does something for kidneys, and in retrospect it is the observation that made the dedicated trial worth funding.
PMID 27633186 ↗NCT01720446 ↗DOI 10.1056/NEJMoa1607141 ↗
#8ELIXA (renal)
Muskiet MHA, Tonneijck L, Huang Y, et al. · Lancet Diabetes Endocrinol · 2018
What it measured: Protein leaking into the urine, measured at just over two years
The renal analysis of the trial that found nothing for the heart. Among patients who already had macroalbuminuria, lixisenatide reduced urinary protein 39% more than placebo, but kidney function decline did not differ. Its importance is as a floor: even a short-acting GLP-1 with no cardiovascular benefit whatsoever still moved albuminuria. That tells you albuminuria is the easiest endpoint in this field to move, and should temper how much weight it carries elsewhere.
PMID 30292589 ↗NCT01147250 ↗DOI 10.1016/S2213-8587(18)30268-7 ↗
#9FLOW (SGLT2 interaction)
Mann JFE, Rossing P, Bakris G, et al. · Nat Med · 2024
What it measured: The same kidney measure, split by whether patients were already on an SGLT2 inhibitor
The subgroup analysis built around the question every nephrologist asks: does this still work if the patient is already on an SGLT2 inhibitor? Of FLOW's 3,533 participants, 550 were. The hazard ratio for the main kidney outcome was 0.73 without an SGLT2 inhibitor and 1.07 with one, but the SGLT2 subgroup was small and the difference between the two was not statistically significant. The authors read the kidney benefit as consistent across both groups while noting the analysis had limited power, so it does not show that the two effects add.
PMID 38914124 ↗NCT03819153 ↗DOI 10.1038/s41591-024-03133-0 ↗
#10GLP-1 class meta-analysis
Sattar N, Lee MMY, Kristensen SL, et al. · Lancet Diabetes Endocrinol · 2021
What it measured: Cardiac and kidney outcomes pooled across eight cardiovascular trials
The class-level meta-analysis that set the pre-FLOW baseline, pooling eight cardiovascular trials and 60,080 patients. The kidney composite fell 21%. The headline hides a caveat: the albuminuria component drove almost all of it, with much weaker signals for actual loss of kidney function or progression to dialysis. It is useful as a benchmark, but it is not the same claim as preventing kidney failure.
PMID 34425083 ↗DOI 10.1016/S2213-8587(21)00203-5 ↗
About this list
Obesity medicine throws up the same handful of questions over and over, and we answer them with ranked lists anchored to the papers themselves. On 2026-10-30 every citation below was put back through PubMed. Where a trial has a registry entry, the card carries that link too, alongside the PubMed one.
Browse our full index of research lists or our long-form research articles.