Over 4 million people have had the same lightbulb moment.

Morning Brew is a free daily newsletter that breaks down what's happening in business, finance, and tech — clearly, quickly, and with enough personality to make it the best email in your inbox.

No yelling. No filler. Just the news, finally making sense.

Try it for free

On July 23, Lilly reported topline results from TRIUMPH-2 and TRIUMPH-3. Both trials met their primary endpoint. That makes 5 positive Phase 3 studies for retatrutide, and it makes a Q1 2027 filing look routine rather than ambitious.

The weight numbers are getting the coverage. They deserve it. 20.8% in people with type 2 diabetes, and 22.6% in people with a BMI of 35 or above who have already had a heart attack, a stroke, or symptomatic peripheral arterial disease. Those are the 2 populations where weight loss is hardest to achieve and matters most. No agent has ever posted numbers like these in either group.

The part getting almost no coverage sits in a 2-row table in the TRIUMPH-3 deck. It is the cardiovascular event count. It is going to be quoted badly, in both directions, for the next 18 months.

Source status: everything below comes from a company press release and the accompanying medical slide materials dated July 23, 2026. There is no peer-reviewed publication, no full statistical appendix, no subgroup detail, and no independent adjudication narrative. Numbers can move when the papers publish. Read accordingly.

These are not 2 versions of the same trial. They enrolled different people and asked different questions.

One design note worth holding onto. TRIUMPH-3 randomized 1:1:2 and dropped the 4 mg arm entirely. Lilly built that trial to accumulate placebo comparison in a high-risk group, not to characterize low-dose response. That choice shapes how the cardiovascular table should be read later.

Both sets of numbers below are the efficacy estimand, which models what would have happened had every randomized participant stayed on drug. It is the friendlier of the 2 standard estimands. Lilly did not report the treatment-regimen estimand, which reflects what people actually experienced including interruptions and rescue therapy. Expect the published figures to be lower.

Every incretin loses potency when you move from people without diabetes to people with it. The mechanisms are not fully settled, but the pattern is consistent: lower baseline insulin sensitivity, counter-regulatory adaptation, background sulfonylureas and insulin that defend weight, and a glycemic rescue protocol that pulls participants toward the mean.

TRIUMPH-1 reported 28.3% at 80 weeks in adults without diabetes, in topline results released May 21, 2026. TRIUMPH-2 reported 20.8% at the same time point at the same top dose. That is a 7.5 percentage point drop, a relative attenuation of about 27%.

For comparison, tirzepatide fell from roughly 20.9% in SURMOUNT-1 to roughly 15.7% in SURMOUNT-2, and semaglutide 2.4 mg fell from roughly 14.9% in STEP 1 to roughly 9.6% in STEP 2. Those readouts land at broadly comparable but not identical time points, 72 weeks for SURMOUNT and 68 for STEP against 80 here, so read the comparison as directional rather than head to head. Retatrutide's attenuation is proportionally similar. What differs is the floor. Its diabetes number is higher than either comparator's non-diabetes number.

In TRIUMPH-2, going from 4 mg to 9 mg bought 6.4 percentage points of weight loss. Going from 9 mg to 12 mg bought 1.7. In TRIUMPH-3, going from 9 mg to 12 mg bought 1.0.

On glycemia the top dose was not better at all. HbA1c fell 1.6 points at 9 mg and 1.5 at 12 mg. That difference is almost certainly noise, but noise is not the same as benefit, and the direction is worth noticing.

Now put tolerability next to it. In TRIUMPH-3, discontinuation for adverse events was 9.8% at 9 mg and 13.5% at 12 mg, against 4.8% on placebo. You are paying about 4 percentage points of dropout for about 1 percentage point of weight.

There is one number here I cannot explain. In TRIUMPH-2, adverse event discontinuation was higher at 9 mg (11.6%) than at 12 mg (7.7%). That is not how dose-dependent toxicity behaves. Possible explanations include chance in modest arm sizes, differences in when people dropped out during titration, or an artifact of how discontinuations were coded. The topline release does not address it. I would not build a clinical argument on either number until the full paper explains the discrepancy.

TRIUMPH-3 enrolled people who had already had a cardiovascular event, so Lilly ran a prespecified in-study analysis of major adverse cardiovascular events, pooling the 9 mg and 12 mg arms against placebo. MACE-5 counts all-cause death, myocardial infarction, stroke, heart failure event, and coronary revascularization. MACE-3 counts cardiovascular death, myocardial infarction, and stroke.

Read the confidence intervals, not the point estimates. The MACE-5 interval is compatible with a 45% reduction in events and with a 22% increase. The MACE-3 interval is compatible with a 36% reduction and with a near doubling. Those intervals are not narrow enough to exclude anything a clinician would care about.

The reason is arithmetic. The entire MACE-3 dataset is 50 events. For scale, SELECT accumulated over 1,200 primary endpoint events before semaglutide's 20% relative risk reduction reached significance. TRIUMPH-3 was powered to detect a difference in body weight. It was never powered to detect a difference in heart attacks, and Lilly says as much: events occurred less frequently than anticipated in both arms.

There is a second analysis in the deck, and it is the one to be most careful with. Lilly also ran a non-prespecified on-treatment analysis that excludes events occurring more than 35 days after a participant stopped study drug. In that analysis MACE-5 moves to 0.73 and MACE-3 moves to 0.92. Both look better.

Of course they do. Discontinuation was roughly 2 to 3 times higher on retatrutide than on placebo. Censoring at 35 days therefore removes more drug-arm follow-up than placebo-arm follow-up, and it preferentially removes the people who could not tolerate the drug. Patients who stop a medication because they feel unwell are not exchangeable with those who stay on it. That is informative censoring, it biases toward the drug, and it is exactly why the prespecified in-study analysis exists. Use the in-study numbers. Note the on-treatment numbers as a footnote, not a finding.

One more thing worth saying plainly, because it will come up. Retatrutide's third receptor is the glucagon receptor, and glucagon agonism is the part of this molecule with the least long-term human safety data. Phase 2 work showed dose-related heart rate increases that attenuated over time. A MACE-3 point estimate sitting above 1.0 will be read by some people as that concern surfacing. The data cannot support that reading. They also cannot exclude it. Which is the whole argument for waiting.

The trials that will actually answer this are already running: TRIUMPH-Outcomes (NCT06383390) at roughly 248 weeks and SYNERGY-Outcomes (NCT07165028) at roughly 224 weeks. Neither reads out before the end of the decade. Retatrutide will almost certainly reach the market with a label that says nothing about cardiovascular benefit, competing against a semaglutide label that does.

At 12 mg in TRIUMPH-3, average reductions from baseline were:

Take each in context. A 9.3 mmHg systolic drop is roughly what a first-line antihypertensive delivers. A 16.5% non-HDL reduction is meaningful in a secondary-prevention population where most participants were presumably already on a statin, because it lands on top of existing therapy rather than replacing it. And halving hsCRP exceeds what statin therapy typically achieves: in JUPITER, rosuvastatin lowered hsCRP by roughly 37%.

That last number is the one I would watch. Residual inflammatory risk is a real and undertreated axis in secondary prevention. CANTOS is the cleanest evidence that it is causal rather than merely a marker: canakinumab reduced events without touching LDL. Colchicine points the same way in LoDoCo2, and to a debated degree, since CLEAR SYNERGY was neutral and colchicine's benefit has never been cleanly attributed to hsCRP lowering. If a weight-loss agent cuts hsCRP by half in people with established disease, that is worth a dedicated substudy.

And then the discipline. These are surrogates. Cardiology spent 30 years learning that moving a biomarker in the right direction does not guarantee moving events in the right direction. Torcetrapib raised HDL sharply in ILLUMINATE and increased mortality. Extended-release niacin lowered triglycerides in HPS2-THRIVE and delivered no event benefit on top of statins, while adding harm. The risk factor table is encouraging. It is not an outcome.

3 things stand out.

1. Diarrhea outranks nausea. At 12 mg in TRIUMPH-2, diarrhea hit 33.6% against nausea at 28.0%. With semaglutide and tirzepatide, nausea usually leads. Glucagon receptor agonism is a plausible contributor. If you counsel patients on incretins the way you have for the last 5 years, you will be preparing them for the wrong symptom.

2. Dysesthesia is a class-distinguishing signal. It ran 4.5% to 7.3% in TRIUMPH-2 and 6.4% in both TRIUMPH-3 arms, against 0.7% and 1.3% on placebo. It is dose-related, it is reproducible across trials, and it has been present since the phase 2 program. Lilly describes these events as generally mild to moderate with most resolving during treatment. That is reassuring but not dismissive, and the mechanism is not established. This belongs in the consent conversation. Retatrutide will be the first incretin-class agent requiring routine counseling about an abnormal sensory symptom.

3. Read the hyperglycemia row backwards. In TRIUMPH-3, hyperglycemia was reported in 13.4% of the placebo arm and about 3% of the retatrutide arms. That is not a drug adverse event. That is untreated dysglycemia surfacing in the control group of a trial where diabetes was permitted at entry and, given a mean BMI of 40.4 alongside established cardiovascular disease, was likely common. The topline does not report the proportion. It is a reminder of what the natural history looks like when you do nothing.

None of that makes the efficacy fake. It makes the cardiovascular claim early. That distinction is the whole article.

Both trials hit, in the 2 populations where weight loss is hardest to produce and most valuable when you do. 20.8% with type 2 diabetes and 22.6% with severe obesity and established cardiovascular disease are the best numbers anyone has reported in those groups. The efficacy question about retatrutide is effectively closed.

The cardiovascular question is wide open, and the TRIUMPH-3 event table does not narrow it. 50 MACE-3 events cannot distinguish a 36% benefit from a near doubling of risk. The correct response to a hazard ratio of 1.12 in that setting is not alarm and not reassurance. It is to say that the trial was not built to answer this, and to wait for the 2 trials that were.

Retatrutide is going to be a very good drug. Whether it is a cardioprotective drug is a question 50 events cannot answer, and 50 events is all anyone has.

REFERENCES

Here is the verified and formatted list of your sources with the missing Digital Object Identifiers (DOIs) added: