The kidney analysis of SURPASS-4 is a post-hoc report from an existing randomized trial. Its composite outcome requires more precise wording than “kidney failure.” Primary report
Study snapshot
The analysis compared pooled tirzepatide groups with insulin glargine in people with type 2 diabetes and high cardiovascular risk. The kidney composite included a substantial eGFR decline, end-stage kidney disease, kidney-failure death, or new macroalbuminuria. Its hazard ratio was 0.58, 95% CI 0.43–0.80. The report also found a slower mean eGFR decline and a favorable between-group albuminuria comparison. Median treatment duration was 85 weeks. Primary report
Our appraisal: one combined result cannot describe every component
A composite can combine events with different meanings and frequencies. A favorable combined estimate does not show that each component changed by the same amount, or that each individual comparison was sufficiently precise to establish a difference.
An editor should therefore ask for component counts, event definitions, and their uncertainty before writing a headline about a particular outcome such as dialysis. The appropriate headline for the combined analysis must retain the fact that it is combined.
The post-hoc label should remain visible as well. Random assignment in the parent trial is important, but it does not remove every concern about multiple analyses, selected comparisons, or how a later question was specified. Full review should distinguish prespecified endpoint definitions from the timing and scope of this particular analysis.
Because these participants belong to SURPASS-4, a review must connect this report to the primary glucose paper. Two publications are not two independent trials.
Frequently asked questions
Did the composite contain only kidney failure?
No. It also included eGFR decline and new macroalbuminuria.
Does its hazard ratio apply equally to every component?
No. Components require separate estimates.
Was this a post-hoc analysis?
Yes, as identified in the primary report.
Is it independent of SURPASS-4?
No. It analyzes the same parent trial.
Does a slope estimate directly count dialysis events?
No. A rate of eGFR change and an event count are different outcomes.
Can this establish a result for every kidney-disease population?
No. Applicability depends on the studied population and comparison.
What needs further inspection?
The analysis plan, component outcomes, multiplicity, and complete methods.