The Weekly Dose — Issue #13

One story this week, but it is the story the entire oral GLP-1 category has been waiting for. ACHIEVE-4, published in The Lancet on 30 September, is the first cardiovascular outcomes trial of a once-daily oral small-molecule GLP-1 receptor agonist. Lilly put a 57% mortality reduction in the headline. The footnotes say something considerably more cautious, and the footnotes are the part worth your time.

For a decade the deal was simple: if you wanted demonstrated cardiovascular protection from incretin therapy, you accepted a subcutaneous injection. Semaglutide earned that claim in SELECT and SUSTAIN-6. Tirzepatide is building its own case. Pills promised adherence, scale, no cold chain, no needles — but nobody could say whether a non-peptide molecule docking the same receptor would reproduce any of the cardioprotection. Mechanistically that is not a given. ACHIEVE-4 is the first real test, and it clears the bar it was designed to clear while leaving the interesting questions open.

We have the full breakdown below: what the trial actually measured, why the comparator choice matters more than it looks, which numbers are controlled for multiplicity and which are not, and what the tolerability ledger means if you are personally weighing an oral against an injectable GLP-1. One compound this week — orforglipron — but it carries the entire non-peptide incretin class on its back, and it is benchmarked throughout against the two molecules you already know: semaglutide and tirzepatide.

Worth stating plainly why a small molecule is a different animal. Orforglipron, discovered by Chugai and licensed by Lilly in 2018, is not a peptide at all. Peptides like semaglutide are large, fragile chains that get chewed up in the gut — oral Rybelsus only works with a SNAC absorption enhancer and strict empty-stomach dosing rules. A small molecule survives digestion on its own terms, costs far less to manufacture, needs no cold chain, and can be dosed without the fasting ritual. If a molecule with those logistics can also protect hearts, the access implications dwarf the clinical ones. That is the bet ACHIEVE-4 was built to test — and the cardiac half of the bet is still unresolved.

An Oral GLP-1 Pill Just Passed Its First Cardiovascular Test. Read the Death Numbers Carefully.

The trial design first, because it frames everything. ACHIEVE-4 randomised 2,749 adults with type 2 diabetes and established cardiovascular disease or chronic kidney disease — 85.9% had established CVD — across 317 sites in 16 countries. The comparator was titrated insulin glargine, not placebo, and the non-inferiority margin was generous: an upper confidence bound below 1.8. Orforglipron met it comfortably. MACE-4 occurred in 4.2% of the orforglipron arm versus 5.0% on glargine, hazard ratio 0.84 (95% CI 0.59–1.20), after a median 104 weeks. On the tighter MACE-3 endpoint the HR was 0.77 with p=0.181 — superiority was not established. The pill did not lose to basal insulin on cardiovascular events. It also did not prove it wins.

Now the number everyone is quoting. Pre-planned secondary analyses showed 53% lower cardiovascular death (HR 0.47) and 57% lower all-cause death (HR 0.43; 19 deaths versus 43), with heart-failure hospitalisation at 0.7% versus 1.5%. Those analyses were not controlled for family-wise type 1 error — Lilly says so in its own release footnotes — which makes them hypothesis-generating, not findings. And there is a second wrinkle: completion on treatment was 69.1% in the orforglipron arm against 79.5% on glargine. A ten-point differential dropout gap sits underneath the mortality claim. None of this means the signal is fake. It means it is exactly the kind of signal that needs a dedicated superiority trial before anyone acts on it.

The tolerability ledger is where the honest trade-off lives. Gastrointestinal adverse events hit 62.1% on orforglipron versus 14.2% on glargine — nausea 31.9%, vomiting 21.0%, diarrhoea 20.4% — and 15.4% discontinued due to adverse events versus 5.4%. Most of it was mild to moderate and clustered during dose escalation, the standard incretin profile. But hypoglycaemia ran strongly the other way: clinically significant events at 6.8% versus 19.2%, which matters a great deal if you have chronic kidney disease (37.6% of this population did) or take a sulfonylurea. For someone running a GLP-1 protocol, the practical read is this: the oral option trades a heavier early GI burden and higher attrition for needle-free dosing and markedly less hypoglycaemia. The cardiovascular claim is not yet on the table for either side of that trade.

The efficacy numbers, for completeness, moved exactly as expected. From a baseline HbA1c of 8.22%, orforglipron dropped it 1.6 percentage points versus 1.0 on glargine at 52 weeks; body weight fell 8.8% while the insulin arm gained 1.7%. At 104 weeks, 54.6% reached HbA1c at or below 6.5% and 35.9% lost at least 10% of body weight, against 23.8% and 3.4% on insulin. Lilly also reports waist circumference down 7.0 cm, systolic BP down 4.4 mmHg, triglycerides down 18.9%, hsCRP down 38.9%, and a slower eGFR decline with 27.1% greater albuminuria reduction — none of it multiplicity-controlled either, and all of it the kind of broad cardiometabolic movement that makes the mortality signal plausible without making it proven. One structural caveat nobody should skip: this trial never compared orforglipron to semaglutide or tirzepatide. Beating insulin glargine on hypoglycaemia while tying it on events is a solid result. It is not evidence the pill matches the injectables.

[@portabletext/react] Unknown block type "span", specify a component for it in the `components.types` prop

The Week Ahead

What to watch next: the regulatory clock. Lilly's orforglipron submissions rest on the ACHIEVE-1 and ACHIEVE-3 glycaemic and weight data, and approval decisions for the oral obesity indication will land while the mortality signal from ACHIEVE-4 is still officially a hypothesis. Expect the injectable franchise — semaglutide, tirzepatide — to keep its demonstrated-CV-benefit moat until a dedicated superiority trial reads out. If you are choosing between oral and injectable GLP-1s in the meantime, you are choosing on tolerability, adherence and access, not on cardiac outcomes. Anyone who tells you otherwise is reading the press release instead of the paper.

The other thread to follow is ATTAIN-OUTCOMES, the ongoing trial in adults with obesity or overweight, with or without diabetes. That is where the mortality question gets a second, cleaner look, and where the obesity-only population that orforglipron is actually approved for gets studied. Remember ACHIEVE-4's limits before anyone extrapolates: open-label design, a type 2 diabetes population, and a median two years of follow-up — short by cardiovascular-outcomes standards, since SUSTAIN-6 and SELECT both ran considerably longer. Our evidence grade for the trial lands at B: rigorous enough to trust the safety and non-inferiority conclusions, too preliminary to bank the mortality signal.

Practically, if you are weighing an oral incretin against daily basal insulin, the usable takeaways are already here. Expect meaningful gastrointestinal effects concentrated in the first months of titration. Expect that roughly one in seven people stops because of them. Expect markedly fewer hypoglycaemic events unless a sulfonylurea is also in the mix. That trade is worth making for a lot of people — it just is not the trade the headline numbers imply.

Model This Week's Compounds in Your Own Stack

This week's analysis lives entirely on one pathway: the GLP-1 receptor. Orforglipron is the new small-molecule entrant, and the compounds it is implicitly benchmarked against every single day — semaglutide and tirzepatide — are already in our interaction database. If you are running a GLP-1, considering the switch from injectable to oral, or stacking one alongside anything else, the BioStack Generator models those interactions against your specific goals, dosing schedule and titration plan. Two minutes, and you will know whether your stack has a conflict before your next dose, not after.

[@portabletext/react] Unknown block type "span", specify a component for it in the `components.types` prop