Most claims studies report an association and leave the reader to guess how much unmeasured confounding would erase it. This one tests that directly, which makes it worth reading for its method as much as its result [1]. Other comparisons between the same two drugs are in the semaglutide and tirzepatide comparison.
Two instruments were used. A negative control outcome is something the drug could not plausibly affect: if the analysis finds an effect there, the method is manufacturing associations. An E-value asks how strong an unmeasured confounder would have to be, in association with both treatment and outcome, to explain away the result entirely. Together they convert “this might be confounded” from an objection into a quantity.
The finding itself is modest. Venous thromboembolism over one year gave a hazard ratio of 0.90 with a confidence interval of 0.83 to 0.98. An upper bound that close to 1 across 701,374 people describes a small effect measured precisely — the precision comes from the sample size, not from the size of the difference.
All-cause mortality also fell, at 0.90 — the same magnitude as the primary outcome. In a one-year study of blood clots, a mortality difference of equal size suggests the two arms differ in general health rather than specifically in clotting risk, which is the pattern described at greater length in six unrelated complications all halved. That the authors ran negative controls makes this less likely than it otherwise would be, without eliminating it.
For a reader the practical content is small and the methodological content is not. A 10% relative reduction in clots between two drugs in the same class is not a reason to choose between them, and it sits alongside the other small tirzepatide advantages this site has covered in the glaucoma comparison and the carpal tunnel cohort. What is genuinely useful is that a large claims study can quantify its own fragility, and that most do not.