A GLP-1 Pill Against Insulin: What ACHIEVE-4 Proved for the Heart

In type 2 diabetes, when pills stop working, insulin has long been the next step. ACHIEVE-4, published in The Lancet on September 30, 2026 and presented the same day at the European diabetes meeting in Milan, asks a pointed question about that ladder: what happens to the heart if the next step is a daily pill instead?

The pill is orforglipron, which Eli Lilly sells in the United States as Foundayo. It is a small molecule GLP-1 drug, taken once a day with no rules about food or water. Its first job in this trial was not to win. It was to prove it does no harm to the heart.

Who was in the trial

The trial randomized 2,749 adults at 317 sites in 16 countries and territories between May 2023 and September 2024. Everyone had type 2 diabetes that was not controlled on up to three pills (metformin, a sulfonylurea, an SGLT2 inhibitor), a body mass index of at least 25, and established cardiovascular or chronic kidney disease. This was a sick, older group: mean age 63, mean HbA1c 8.2 percent, mean BMI 33, with almost 86 percent already living with cardiovascular disease.

Half were assigned to orforglipron, stepped up every four weeks to the highest dose each person tolerated. The other half got insulin glargine, a once daily basal insulin, adjusted over time to reach glucose targets. The design was open label, so patients and doctors knew who took what. The main outcome was the first of four events: cardiovascular death, nonfatal heart attack, nonfatal stroke, or a hospital stay for unstable angina. Median follow up was about two years.

The heart result, in absolute terms

The primary event struck 57 of 1,358 people on orforglipron (4.2 percent) and 67 of 1,343 on insulin (5.0 percent). That is a gap of 0.8 percentage points over two years, and a hazard ratio of 0.84, which Lilly's release translates as a 16 percent lower risk.

The confidence interval runs from 0.59 to 1.20. Read plainly, the data are compatible with anything from a 41 percent reduction in risk to a 20 percent increase. The trial was built to clear a bar of 1.8 on the upper end of that interval, and it cleared it easily. So the honest headline is the one the authors chose: cardiovascular safety confirmed. Not cardiovascular benefit proven.

The pill passed a safety exam. It did not pass a benefit exam, because it never sat one.

The secondary numbers lean the same way without settling anything. A narrower three part endpoint (cardiovascular death, heart attack, stroke) came in at a hazard ratio of 0.77, with an interval from 0.52 to 1.13. Neither was statistically significant as a superiority result (p values of 0.336 and 0.181).

The number that will travel: deaths

Here is the figure most likely to end up in a headline. During the study, 19 people died on orforglipron (1.4 percent) against 43 on insulin (3.2 percent). Lilly reports hazard ratios of 0.43 for death from any cause and 0.47 for cardiovascular death, both with p values well under 0.05.

Those are big numbers, and they come with two caveats Lilly itself prints in a footnote. The death analyses were prespecified but not controlled for multiplicity, meaning the trial did not protect against the extra chance of a false positive that comes from testing many outcomes. And the pattern is odd: a modest, uncertain effect on heart attacks and strokes sitting next to a halving of deaths. That can happen when a finding is real. It also happens when 62 deaths are split by chance. All but one death were judged unrelated to treatment by the investigators.

My guess is that part of the gap reflects the comparator rather than the pill. Insulin carries a well known risk of low blood sugar, and in this trial clinically significant or severe hypoglycemia hit 19.2 percent of the insulin group against 6.8 percent on orforglipron. Whether that explains any deaths is not something this paper can show, and I am flagging it as a hypothesis, not a finding.

Sugar, weight and the price in side effects

On glucose and weight, the comparison is lopsided. In Lilly's efficacy estimand, which estimates what would happen if everyone stayed on assigned treatment, HbA1c fell 1.6 points on orforglipron and 1.0 point on insulin at 52 weeks, from a starting average of 8.22 percent. Body weight dropped 8.8 percent on the pill and rose 1.7 percent on insulin, from about 91 kilograms. By week 104, 54.6 percent of the orforglipron group had an HbA1c of 6.5 percent or lower, against 23.8 percent on insulin, and 35.9 percent had lost at least a tenth of their weight, against 3.4 percent.

The bill arrives in the gut. Gastrointestinal side effects were reported by 62.1 percent of people on orforglipron and 14.2 percent on insulin. Over the minimum 52 week treatment period, 10.6 percent of the orforglipron group stopped treatment because of adverse events, mostly for stomach reasons. One in ten walking away matters for a drug sold on being easy to keep taking.

What comes next

Context matters for reading this. In the US, Foundayo's current label covers weight management in adults with obesity or overweight; type 2 diabetes is not on it. ACHIEVE-4, the largest and longest diabetes study of the drug so far, adds to the file regulators will weigh. The question it cannot answer, whether orforglipron actively reduces heart attacks and strokes, belongs to ATTAIN-OUTCOMES, an ongoing Lilly trial in people with obesity or overweight and established cardiovascular or kidney disease.

If you followed the earlier pill duel, this is a different kind of test. ACHIEVE-3 pitted orforglipron against oral semaglutide on blood sugar; ACHIEVE-4 measures it against the drug many patients would otherwise be handed next. And if the gap between a hazard ratio and an absolute risk still feels slippery, this five minute guide to reading a health study walks through it. None of this is a reason to change any treatment on your own; that conversation belongs with a doctor who knows your history.

What strikes me most is how low the bar was set and how much is being built on top of it. Choosing insulin as the comparator was clinically sensible, since insulin is a real alternative at this stage of the disease. It was also strategically comfortable: insulin adds weight and causes lows, so almost any modern GLP-1 drug was going to look good beside it on everything except the stomach. The trial delivered what it was designed to deliver, a clean safety pass, plus a mortality number it was never designed to prove. I would not dismiss that number, and I would not repeat it without its footnote. The real verdict on whether a pill can do for the heart what the injections have done is still out, sitting inside a trial that has not reported yet.