In one retrospective cohort, yes. Contrast-induced kidney injury occurred in 8.7% of GLP-1 users against 25.7% of non-users [1]. The adjusted odds ratio was 0.290, 95% CI 0.119 to 0.708. That outcome is defined by a creatinine rise. Creatinine itself did not differ significantly, at p = 0.084. Nobody was randomized, and the cohort is 336 people at one center.
A kidney injury here is not something anybody felt. It is a definition: creatinine up by at least 0.5 mg/dL, or by at least a quarter, within seventy-two hours. Hold that next to the fact that creatinine did not differ significantly between the two groups. The randomized kidney evidence for this drug class is a separate body of work, summarized in the FLOW trial.
Both numbers, side by side
The event rate was 8.7% among GLP-1 users and 25.7% among non-users. Adjusted, the odds ratio came to 0.290 with an interval of 0.119 to 0.708. At the same timepoint, mean creatinine was 1.02 mg/dL in users and 1.17 in non-users, p = 0.084. Those standard deviations are wide, at 0.73 and 1.57.
A difference in group means can be absent while a difference in how many people cross a threshold is large. That is not a contradiction and it is not nothing either. A threshold outcome and a mean are different questions about the same measurement. A paper reporting both has told you something a paper reporting only the first would not. That is how a ratio and an absolute risk describe one finding in two sizes.
That last point is not theoretical. A propensity-matched study of semaglutide in polycystic kidney disease used eGFR as its outcome [2]. The same dependency applies wherever a drug changes lean mass. The fuller reading of that trap is in the eGFR and muscle problem.
What the design can carry
336 people at one center, looked at afterwards. Users were compared with non-users by regression rather than by matching. Nobody chose who got what.
An odds ratio of 0.29 is a large effect. Large effects from small retrospective cohorts are the ones that most often shrink when somebody randomizes. The usual reason is that the two groups differed in ways nobody recorded. A hazard ratio its own authors distrust shows that at the extreme.
What the randomized evidence says about kidneys
The strongest kidney result for this drug class comes from a trial that stopped early. FLOW randomized people with type 2 diabetes and chronic kidney disease to semaglutide or placebo [3]. It was halted for efficacy on a composite of kidney failure, sustained eGFR decline and death. That is chronic disease progression over years, not a creatinine spike in seventy-two hours.
The two questions are related and they are not the same. Nothing in the chronic evidence predicts what happens around a scheduled dose of contrast dye. Nothing in a 336-person cohort establishes it either.
Why it is still worth reading
Contrast injury is common, it is expensive, and it happens on a schedule somebody controls. A drug patients are already taking that reduced it would matter. There is a plausible route: better volume status, less inflammation, improved endothelial function.
So this is a hypothesis with a mechanism, measured in a way that leaves room for doubt. The useful next step is a trial that randomizes around a scheduled procedure. Nothing here is a reason to start, stop or time a dose around an angiogram. Tell the team doing the procedure what you take and let them decide.