HARMONY OUTCOMES: A Win for a Withdrawn Drug

Name: HARMONY OUTCOMES: A Win for a Withdrawn Drug
Creator: Ruth Calder
Published: 2026-11-07

Last verified November 2026 · Phase 3 in substance, logged on the registry as Phase 4 because it ran after approval as a cardiovascular outcomes study · Finished and published October 2018, three months after GSK had already pulled the drug from every market worldwide for commercial reasons rather than anything to do with safety · NCT02465515 ↗

By Ruth Calder · Enforcement Editor

Editorially reviewed. Not medical advice · How we verify contentLast reviewed November 2026

Albiglutide never sold well. It was less potent than its competitors on both glucose and weight, and by 2017 GSK had decided it was not worth continuing. The outcomes trial was already running and completed anyway, which is how a drug nobody can buy came to hold one of the larger cardiovascular results in its class. The median follow-up of 1.6 years is short by the standards of these trials, which usually makes a positive finding harder to obtain.

Enrollment: 9,463
Duration: A median 1.6 years. The trial ran until enough events accumulated: the committee judged the threshold met in November 2017, and the last participant visit was March 12, 2018
Drug: Albiglutide 30-50 mg weekly (subcutaneous)
Population: Adults 40 and over whose type 2 diabetes came with existing cardiovascular disease (an event in the heart, brain or limb arteries, or a procedure to repair one) and an HbA1c of at least 7.0% on stable treatment. Everyone was in secondary prevention, which makes this trial easier to read than most in its class. Average age 64.1, 30.6% women, average BMI 32.3, average HbA1c 8.7%, and nearly 14 years since diagnosis. Roughly 70% had had a heart attack or a coronary procedure, about 18% a stroke.

Primary endpoint

Cardiac death, a nonfatal heart attack, or a nonfatal stroke — whichever came first

Treatment arm

338 of 4,731 (7%) — incidence rate 4.6 events per 100 person-years

Comparator

428 of 4,732 (9%) — incidence rate 5.9 events per 100 person-years

Treatment difference: Hazard ratio 0.78 (95% CI 0.68–0.90); P<0.0001 for non-inferiority; P=0.0006 for superiority

A 22% reduction, among the strongest in this class, achieved in a shorter trial than most of its peers. Short follow-up normally makes a positive result harder to reach. In absolute terms the gap was roughly 2 percentage points.

Secondary endpoints

HARMONY OUTCOMES secondary endpoints, treatment against comparator. Last verified November 2026.
Endpoint	Treatment	Comparator	Difference
Death from a cardiac cause Lower, but not significantly. At 1.6 years there is not enough time for deaths to separate, and that is the main cost of how quickly this trial finished.	Incidence rate 1.61 events per 100 person-years on albiglutide	Incidence rate 1.72 events per 100 person-years on placebo	Hazard ratio approximately 0.93 (95% CI crossed 1.0); no significant reduction
Heart attacks that were not fatal Down 25%, and the biggest single contributor to the headline. In the semaglutide trial heart attacks fell by a similar amount without reaching significance, and strokes carried the benefit. Both drugs belong to the same class, yet the significant result landed in different places.	Incidence rate 2.43 events per 100 person-years on albiglutide	Incidence rate 3.26 events per 100 person-years on placebo	Hazard ratio 0.75 (95% CI 0.61–0.90); statistically significant
Strokes that were not fatal Fewer, without reaching significance. That is the inverse of semaglutide's pattern, and part of why claims about this class need to name a drug.	Incidence rate 1.25 events per 100 person-years on albiglutide	Incidence rate 1.45 events per 100 person-years on placebo	Hazard ratio approximately 0.86; confidence interval crossed 1.0
The same count plus emergency procedures to reopen an artery Tracked the headline closely, suggesting the benefit on heart attacks extended to the emergencies that precede them.	Incidence rate 5.06 events per 100 person-years on albiglutide	Incidence rate 6.45 events per 100 person-years on placebo	Hazard ratio 0.78 (95% CI 0.69–0.90); statistically significant
Cardiac death or admission for heart failure Pointing the right way and not significant. Heart failure has been a persistent null across this whole class, and it is where SGLT2 inhibitors clearly outperform.	Incidence rate 2.49 events per 100 person-years on albiglutide	Incidence rate 2.92 events per 100 person-years on placebo	Hazard ratio approximately 0.85; confidence interval crossed 1.0
Serious small-vessel damage: needing dialysis or a transplant, going blind from diabetes, or needing eye injections or laser Numerically fewer. The trial was never sized to prove this, but the direction matches the kidney and eye signals seen with other drugs in the class.	Incidence rate 0.46 events per 100 person-years on albiglutide	Incidence rate 0.69 events per 100 person-years on placebo	Hazard ratio approximately 0.66; confidence interval crossed 1.0
Three-month average blood sugar at 16 months This row is the most revealing. The drug was a weak glucose-lowerer, about half a point better than placebo from a starting HbA1c of 8.7%, and it still produced one of the larger cardiac reductions in the class. Whatever protects the heart, it is not the blood sugar.	−0.83 percentage points (least-squares mean) on albiglutide	−0.31 percentage points (least-squares mean) on placebo	Estimated treatment difference approximately −0.5 percentage points vs placebo
Weight change at 16 months Under 1 kg against placebo. This molecule is fused to albumin and barely reaches the brain, which likely explains both the negligible weight effect and the weak appetite suppression. Nothing here is relevant to the weight drugs sold today.	−1.36 kg (least-squares mean) on albiglutide	−0.53 kg (least-squares mean) on placebo	Estimated treatment difference approximately −0.8 kg vs placebo
How long before someone not already on insulin had to start it The rate of starting insulin was less than half that on placebo. For a patient, avoiding insulin is a concrete quality-of-life outcome, yet it got no attention at publication because the cardiac headline took the room.	Incidence rate 3.56 events per 100 person-years on albiglutide	Incidence rate 8.58 events per 100 person-years on placebo	Hazard ratio approximately 0.42; statistically significant
Hitting an HbA1c of 7% or under without a severe low and with weight gain held under 5% 26.0% managed all three at once against 15.1% on placebo. Older diabetes drugs could rarely deliver target blood sugar, no dangerous lows and little or no weight gain together, and that combination is the everyday case for this class, separate from anything about hearts.	26.0% of participants on albiglutide	15.1% of participants on placebo	Treatment difference approximately +11 percentage points; statistically significant

Adverse events

HARMONY OUTCOMES adverse events, rate on treatment against rate on comparator. Last verified November 2026.
Event	Treatment rate	Comparator rate
Pancreatitis, confirmed by independent review No meaningful excess, on small numbers, matching the other large trials in this class.	0.2% (10/4,731)	0.1% (7/4,732)
Pancreatic cancer No imbalance between the arms over the trial period.	0.1% (6/4,731)	0.1% (5/4,732)
Medullary thyroid cancer Not one case among 9,463 people. This is the cancer behind the boxed warning on the whole class, and that warning comes from rodent studies rather than any human trial.	0 of 4,731	0 of 4,732
Reactions where the injection went in The most common problem attributable to the drug, and worse than with its competitors. Prescribers cited it as a reason the product never caught on.	Higher on albiglutide than on placebo (numerically the most common drug-attributable adverse event)	Lower on placebo
Deaths judged related to treatment Two on the drug against three on placebo, assessed by investigators who did not know which arm anyone was in.	2 of 4,731 (<1%)	3 of 4,732 (<1%)
Stopped the drug over a side effect Comparable to placebo. That made it an easier drug to stay on than semaglutide or liraglutide, for the same reason it was weaker: less pharmacological punch means less nausea.	Comparable to placebo overall	Comparable to albiglutide overall

Subgroup analyses

People who had already had a heart attack or a coronary procedure — about 70%: In line with the trial overall
The dominant group, and the one that defines who this evidence applies to.
People who had already had a stroke — about 18%: The same direction, with wider uncertainty on smaller numbers
Not enough participants to reach a conclusion on its own.
Participants 65 and over — roughly half the trial: The same benefit as everyone else
No age effect across any of the pre-planned cut points.
People whose blood sugar started higher or lower than the trial median: The same benefit either way
More evidence that the cardiac protection does not run through glucose. A drug with a small blood-sugar effect helped people whose blood sugar was already lower just as much.

Clinical significance

Its value lies entirely in the class argument. Another structurally distinct GLP-1 reducing cardiovascular events strengthens the case that this is a mechanism effect, not a property of one or two molecules. It also complicates the simple story, because EXSCEL tested exenatide in a larger and longer trial and came back neutral. Any claim that GLP-1s protect the heart should be made about specific molecules with specific evidence, not about the initials.

Frequently Asked Questions

On its primary endpoint — Cardiac death, a nonfatal heart attack, or a nonfatal stroke — whichever came first — the treatment arm recorded 338 of 4,731 (7%) — incidence rate 4.6 events per 100 person-years against 428 of 4,732 (9%) — incidence rate 5.9 events per 100 person-years for the comparator. Hazard ratio 0.78 (95% CI 0.68–0.90); P<0.0001 for non-inferiority; P=0.0006 for superiority. A 22% reduction, among the strongest in this class, achieved in a shorter trial than most of its peers. Short follow-up normally makes a positive result harder to reach. In absolute terms the gap was roughly 2 percentage points.

9,463 participants, over A median 1.6 years. The trial ran until enough events accumulated: the committee judged the threshold met in November 2017, and the last participant visit was March 12, 2018. It was a Phase 3 in substance, logged on the registry as Phase 4 because it ran after approval as a cardiovascular outcomes study trial, finished and published October 2018, three months after GSK had already pulled the drug from every market worldwide for commercial reasons rather than anything to do with safety. Registered as NCT02465515, which you can look up yourself rather than taking anyone's summary for it.

Adults 40 and over whose type 2 diabetes came with existing cardiovascular disease (an event in the heart, brain or limb arteries, or a procedure to repair one) and an HbA1c of at least 7.0% on stable treatment. Everyone was in secondary prevention, which makes this trial easier to read than most in its class. Average age 64.1, 30.6% women, average BMI 32.3, average HbA1c 8.7%, and nearly 14 years since diagnosis. Roughly 70% had had a heart attack or a coronary procedure, about 18% a stroke. ⚠ A trial result describes the population it was measured in. If you differ from it materially — in age, in baseline weight, in whether you have diabetes — the number is context rather than a prediction.

The most commonly reported were pancreatitis, confirmed by independent review (0.2% (10/4,731) against 0.1% (7/4,732) on comparator), pancreatic cancer (0.1% (6/4,731) against 0.1% (5/4,732) on comparator), medullary thyroid cancer (0 of 4,731 against 0 of 4,732 on comparator). No meaningful excess, on small numbers, matching the other large trials in this class. Rates come from this trial's population under its own dose-escalation schedule; a faster or slower titration changes what people experience.

Secondary endpoints included death from a cardiac cause (Incidence rate 1.61 events per 100 person-years on albiglutide against Incidence rate 1.72 events per 100 person-years on placebo); heart attacks that were not fatal (Incidence rate 2.43 events per 100 person-years on albiglutide against Incidence rate 3.26 events per 100 person-years on placebo); strokes that were not fatal (Incidence rate 1.25 events per 100 person-years on albiglutide against Incidence rate 1.45 events per 100 person-years on placebo). ⚠ Secondary endpoints are weaker evidence than the primary one — a trial is powered for its primary outcome, and the rest are read as supporting rather than as findings in their own right.

It cannot tell you that. A trial reports an average across a defined population under a fixed protocol, with monitoring and adherence support most people do not get. It is the best evidence available and it is still a distribution rather than a promise — some participants lost far more than the mean and some lost nothing. What it is genuinely good for is comparing options against each other, which is what the rest of this section is for.

References

1.Hernandez AF, Green JB, Janmohamed S, D'Agostino RB Sr, Granger CB, Jones NP, Leiter LA, Rosenberg AE, Sigmon KN, Somerville MC, Thorpe KM, McMurray JJV, Del Prato S; Harmony Outcomes committees and investigators. Albiglutide and cardiovascular outcomes in patients with type 2 diabetes and cardiovascular disease (Harmony Outcomes): a double-blind, randomised placebo-controlled trial. Lancet. 2018. PMID: 30291013.
2.GlaxoSmithKline. Effect of Albiglutide, When Added to Standard Blood Glucose Lowering Therapies, on Major Cardiovascular Events in Subjects With Type 2 Diabetes Mellitus (HARMONY OUTCOMES) — Study Results. ClinicalTrials.gov, NCT02465515. 2019. https://clinicaltrials.gov/study/NCT02465515
3.U.S. Food and Drug Administration. Drugs@FDA: Tanzeum (albiglutide) for injection — BLA 125431, status discontinued. FDA Drugs@FDA database. 2018. https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview.process&ApplNo=125431
4.Marso SP, Daniels GH, Brown-Frandsen K, et al.; LEADER Steering Committee; LEADER Trial Investigators. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes. N Engl J Med. 2016. PMID: 27295427.
5.Marso SP, Bain SC, Consoli A, et al.; SUSTAIN-6 Investigators. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes. N Engl J Med. 2016. PMID: 27633186.
6.Gerstein HC, Colhoun HM, Dagenais GR, et al.; REWIND Investigators. Dulaglutide and cardiovascular outcomes in type 2 diabetes (REWIND): a double-blind, randomised placebo-controlled trial. Lancet. 2019. PMID: 31189511.
7.Gerstein HC, Sattar N, Rosenstock J, et al.; AMPLITUDE-O Trial Investigators. Cardiovascular and Renal Outcomes with Efpeglenatide in Type 2 Diabetes. N Engl J Med. 2021. PMID: 34215025.