Chronic obstructive pulmonary disease and type 2 diabetes travel together often enough that the same person is frequently offered a drug for each, and a comparative effectiveness study set out to ask which glucose-lowering drug leaves that person in hospital less often. It emulated three target trials in US insurance claims, matching patients aged 40 and over with both conditions [1]. The result is genuinely interesting, and the part most likely to be quoted is not the part that matters — the same problem we keep finding in the claims sellers make about this class.
Two comparisons that favor the newer drugs
Against DPP-4 inhibitors, GLP-1 receptor agonists came in at 9.89 moderate or severe exacerbations per 100 person-years against 11.49, a hazard ratio of 0.86 (95% CI 0.81 to 0.91) across 32,107 matched pairs. SGLT2 inhibitors against the same comparator ran 9.26 against 11.4, a hazard ratio of 0.81 (95% CI 0.76 to 0.86) across 27,991 pairs. Both hold up across the sensitivity and subgroup analyses the authors report.
And one that finds almost nothing
The third comparison put the two newer classes against each other, in 36,218 matched pairs, and the difference nearly vanished: 9.47 against 10.00 per 100 person-years, a hazard ratio of 0.94 with a 95% confidence interval of 0.89 to 1.00. The upper bound touches no effect. The rate difference tells the same story, at −0.55 per 100 person-years with an interval running to −0.01.
So the honest summary is that both beat an older class and neither clearly beats the other. Quoting only the first two comparisons would imply a ranking this study does not support, which is the shape of argument we have taken apart before where two meta-analyses of the same six trials disagreed.
Who was in it
The cohorts were old and sick by the standards of this market: a mean age near 70, type 2 diabetes, and COPD active enough to be coded as such. Median follow-up was 145 days, with an interquartile range of 61 to 355 — short, for a chronic lung condition. A person buying compounded semaglutide from a telehealth page at 45 with no lung disease is outside the population these numbers describe, and nothing here is a reason to choose a seller. It is a reason to make sure whoever writes the prescription knows the rest of your history, which is the gap we count in what sellers do not publish and put into questions they leave open.