Novo's Study Favored Novo, Lilly's Study Favored Lilly: How to Read a Drug-Company Comparison

On September 29, 2026, the second day of the EASD diabetes meeting in Milan, Novo and Lilly each put out a press release comparing its own drug with the other company's. Novo's analysis found that people who stayed on its semaglutide and moved up to the Ozempic 2 mg dose had fewer deaths, heart attacks, and strokes than people who switched to Lilly's tirzepatide (Mounjaro). The same day, Lilly's analysis found that its daily pill Foundayo beat Novo's oral semaglutide 25 mg on both weight loss and blood sugar.

So who won? Each company, according to each company. This is less a scandal than standard practice in drug marketing, and both releases carry the fine print that tells you how much weight to put on the headline. The trick is knowing what kind of study you are reading before you look at the score.

Novo's study, called COMPETE SWITCH CV, did not give anyone a drug. It looked back through U.S. insurance claims gathered by the data company Komodo Health from January 2018 to September 2025. The researchers found adults with type 2 diabetes who had been taking semaglutide 1 mg and then made one of two moves: 185,705 went up to semaglutide 2 mg, and 23,104 switched to tirzepatide. Over up to 720 days, the switchers had a slightly higher rate of death, heart attack, or stroke. The adjusted hazard ratio was 1.06 (95% CI 1.04 to 1.08), which Novo's headline turns around into a 6% lower risk for its own drug.

The weak spot is the word "switched," because doctors do not flip coins when they change a prescription. A patient might switch because their blood sugar is not under control, because they want more weight loss, or because their insurance plan changed, and some of those reasons could also affect heart risk in ways claims records do not capture. Researchers call this confounding, and statistical adjustment can correct only for the factors someone measured. Novo says as much in its own release: the study shows an association rather than cause and effect, unmeasured confounding may remain, and safety outcomes were not assessed.

Lilly's analysis did not test the two pills against each other either. It lined up results from two separate trials. One was Lilly's ACHIEVE-3, which compared Foundayo with oral semaglutide at 7 mg and 14 mg. The other was Novo's PIONEER PLUS, which tested oral semaglutide 25 mg and 50 mg against 14 mg. Both trials happened to include a 14 mg semaglutide group, which gives an analysis like this a shared reference point, and Lilly says it adjusted for sex, age, baseline A1C, and baseline weight. The result: at 52 weeks, Foundayo 17.2 mg was estimated to give 1.5 percentage points more weight loss and a 0.3-point larger drop in A1C than oral semaglutide 25 mg.

The two trials were not run the same way. PIONEER PLUS enrolled its patients in 2021, kept the dose hidden from patients and staff, and started at an average A1C of 9.0%. ACHIEVE-3 enrolled from 2023 to 2025, was open-label, so patients knew what they were taking, and started at an average A1C of 8.3%. Adjustment can shrink gaps like these but cannot erase them. The 95% confidence interval for the weight difference ran from 0.2 to 2.7 points, a wide range around a 1.5-point headline. Lilly's Rachel Batterham put the caveat plainly: "Head-to-head trials ultimately provide the best comparison of two medicines, but when we don't yet have those data, then other approaches can provide informative data points."

One note on doses: the trials used an investigational formulation, and Foundayo's FDA label restates those doses as equivalent tablet strengths. The ACHIEVE-3 paper calls its top dose 36 mg; Lilly's releases call the same arm Foundayo 17.2 mg.

The strongest comparison is a randomized head-to-head trial: one study, one set of rules, and a computer deciding who gets which drug. Randomization spreads the things nobody measured, like motivation or how sick someone is, evenly between the groups. That is the job claims data and cross-trial comparisons cannot do.

Both companies run these trials, and they do not always come out the way the sponsor hopes. In SURMOUNT-5, funded by Lilly and published in the New England Journal of Medicine in 2025, 751 adults with obesity but without diabetes were randomly assigned to weekly tirzepatide or semaglutide injections. At 72 weeks, weight had fallen 20.2% with tirzepatide and 13.7% with semaglutide. In ACHIEVE-3, Foundayo beat oral semaglutide 7 mg and 14 mg on A1C in 1,698 adults with type 2 diabetes. In REDEFINE 4, Novo's own trial, Novo lost: after 84 weeks CagriSema produced 23.0% weight loss against 25.5% for tirzepatide and missed its goal of showing it was no worse. Novo announced the result in February 2026, as a listed company has to.

ACHIEVE-3 did not test Novo's 25 mg oral semaglutide dose. That gap is exactly where Tuesday's indirect comparison lives, and only another head-to-head trial can close it.

Novo's claims study covered about 209,000 people, and its confidence interval, 1.04 to 1.08, looks impressively tight. Compare it with SURPASS-CVOT, Lilly's randomized, double-blind heart-outcomes trial of tirzepatide against the older GLP-1 drug dulaglutide in 13,299 people with type 2 diabetes and heart disease, published in December 2025. Its hazard ratio for death, heart attack, or stroke was 0.92, with a confidence interval of 0.83 to 1.01. That was enough to show tirzepatide was no worse than dulaglutide, but not enough to prove it was better.

A narrow interval tells you how little random noise is left, and huge datasets shrink noise. They do nothing about bias. If the people who switched drugs differed from the people who stayed, a million records would give you a very precise estimate of the wrong number. The smaller randomized trial is less precise and more trustworthy, which is why drug labels rest mainly on controlled trials.

A 2017 Cochrane review of 75 papers on funding and results found that drug and device studies sponsored by the manufacturer more often reported favorable efficacy results (risk ratio 1.27) and favorable conclusions (risk ratio 1.34) than studies with other funding. Industry studies did not score worse on the standard checklist for bias; they were more often well blinded. The authors concluded that the industry effect "cannot be explained by standard 'Risk of bias' assessments." They also found that in industry studies, the conclusions agreed with the results less often.

That is the useful lesson for this week. Lilly paid for SURMOUNT-5 and SURPASS-CVOT too, and both are solid trials. The logo on a study matters less than the choices behind it. When a company picks the comparator dose, as in ACHIEVE-3, or the patients who get compared, as in a switching study, those choices deserve the first look.

• Were patients randomly assigned to the two drugs in the same trial? If not, treat the result as a lead worth following.
• Which doses were compared, and are they the doses people take today?
• Who was compared: everyone who started a drug, or only people who chose to switch?
• How wide is the confidence interval, and does it cross 1.0 for a hazard ratio, or 0 for a difference?
• Who paid, and has the work been published in a peer-reviewed journal yet?
• What does the release's own fine print say? Both of Tuesday's releases included caveats, just further down the page than the headline.

For weekly weight-loss shots, SURMOUNT-5 gives a direct answer: tirzepatide beat semaglutide by about 6.5 points at 72 weeks. Our review of the tirzepatide-versus-semaglutide evidence covers the safety side of that trade. For pills in type 2 diabetes, ACHIEVE-3 shows Foundayo beat oral semaglutide at 7 mg and 14 mg; whether it beats 25 mg is a question for a trial that has not been run. For heart protection, both companies have randomized outcome trials, semaglutide in SELECT and tirzepatide in SURPASS-CVOT, but neither compared the two drugs with each other, so heart comparisons between them still come from observational studies like Tuesday's.

The best choice for any one person also turns on side effects, cost, insurance coverage, and whether they would rather take a pill or a shot. That is a conversation for a clinician. See our pieces on Foundayo versus oral Wegovy and on CagriSema's REDEFINE 4 loss, and the September 29 digest for the day's full EASD coverage.