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. [1]
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 — 0.73 and 1.57 — and 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, and a paper reporting both has told you something a paper reporting only the first would not — much as a ratio and an absolute risk describe one finding in two sizes.
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, and 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, and the reason is usually that the two groups differed in ways nobody recorded — as a hazard ratio of 0.24 for death shows at the extreme.
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, and 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, which is easier to run than most of the questions this field keeps answering observationally.
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.