Pelacarsen Miss Hands the Lp(a) Bet to Amgen

Novartis said on September 4, 2026 that pelacarsen failed to cut heart attacks, strokes, and related crises in Lp(a)HORIZON. The antisense shot did lower lipoprotein(a) in people already on guideline heart therapy, and that drop did not reduce events versus placebo.

The first large test of an Lp(a) drug is now a loss for Novartis and Ionis Pharmaceuticals. The live question sits with later RNA shots that cut the particle harder and enroll a higher starting level.

The Biochemical Win Did Not Move Events

Pelacarsen is an experimental antisense oligonucleotide built to shut down apolipoprotein(a) production in the liver. Novartis holds global rights under a license from Ionis, which discovered the drug and ran its early studies. On September 4 the Swiss firm said the phase 3 outcomes study did not meet its primary endpoint against placebo.

That endpoint was a four-part mix of cardiovascular death, nonfatal heart attack, nonfatal stroke, and urgent coronary revascularization that required a hospital stay. It was to be judged in the full cohort with Lp(a) of at least 70 mg/dL and again in a prespecified slice at 90 mg/dL or higher. The company said the miss was in the overall study population and did not describe the 90 mg/dL slice.

Although lower Lp(a) levels were observed with pelacarsen, the findings did not demonstrate that this translated into reduced cardiovascular risk in the overall study population. These are not the results we hoped for, but they provide important evidence that advances scientific understanding of the relationship between Lp(a) lowering and cardiovascular outcomes and may help inform future approaches to cardiovascular risk management.

Shreeram Aradhye, President of Development and Chief Medical Officer, Novartis

Lp(a)HORIZON randomized 8,323 patients with established CVD and high Lp(a) to pelacarsen 80 mg under the skin once a month or matching placebo. The registry lists a start date of December 12, 2019, completion on July 16, 2026, ages 18 to 80, and 902 sites. A 2025 design paper in American Heart Journal by Leslie Cho and colleagues said the study was event-driven, aiming for 993 confirmed primary events, with a minimum 2.5 years of follow-up.

HORIZON AT A GLANCE

• The dose: Pelacarsen 80 mg subcutaneous once a month, chosen so time-averaged Lp(a) would fall by about 80%, in line with the 20 mg weekly arm from phase 2.
• The entry bar: Lp(a) of at least 70 mg/dL, about 149 nmol/L in the design paper, on top of a prior heart attack, ischemic stroke, or symptomatic peripheral artery disease.
• The background care: Guideline lipid-lowering and blood-pressure drugs in both arms, which is the setting Novartis called residual risk beyond optimized care.
• The earlier lab result: In a 286-person phase 2 study, 20 mg weekly cut Lp(a) by 80% versus 6% on placebo, and 60 mg monthly cut it by 72%.

Ionis said the new trial again produced substantially lower Lp(a), consistent with those earlier studies, and that the safety profile was acceptable. Full tables, hazard ratios, and p-values were held for a medical meeting. William Blair analysts wrote that past work averaged about a 72% drop and that Ionis indicated the phase 3 lowering looked similar. That figure is still not in the public topline.

Who Takes the Loss on Pelacarsen

Novartis loses the first chance to sell a targeted Lp(a) drug in a market the companies had framed around an inherited risk that hits about one in five people worldwide, and nearly one-third of people with premature cardiovascular disease. Levels are about 90% genetic and barely move with diet. There is still no approved drug aimed at the particle itself.

Ionis loses the late-stage royalty path it sold in February 2019, when Novartis exercised its option to develop and commercialize the medicine. Chief executive Brett P. Monia said the firm was disappointed that the Lp(a) drop did not translate into fewer events, then pointed investors back to wholly owned products, including Tryngolza, Dawnzera, and Zanvastro, the last of which the FDA cleared on September 3, 2026 for Alexander disease.

THE PELACARSEN CLOCK

1. February 2019: Novartis exercises its option on pelacarsen for targeted cardiovascular therapy.
2. December 12, 2019: Lp(a)HORIZON begins randomizing patients to monthly pelacarsen 80 mg or placebo.
3. January 2023: Royalty Pharma closes a $500 million funding deal with Ionis that puts $150 million of value on pelacarsen royalties and $350 million on Spinraza.
4. March 13, 2026: U.S. lipid guidelines tell clinicians to measure Lp(a) once in every adult.
5. July 16, 2026: The outcomes trial completes, according to the public registry.
6. September 4, 2026: Novartis and Ionis report that the primary endpoint missed in the overall population.

Royalty Pharma, which bought a slice of Ionis cash flows in that January 2023 deal, moved the same day to ring-fence the miss. Of the $500 million, $150 million had been assigned to pelacarsen royalties and $350 million to Spinraza. The fund said it does not anticipate making any milestone payments to Ionis, expects to recoup the whole stake plus a modest gain, and still aims for 2030 portfolio receipts of $4.7 billion or more.

Under the contract, Royalty Pharma takes 25% of Ionis Spinraza royalties through 2027 and 45% in 2028, on up to $1.5 billion in annual sales. After the HORIZON result, that Spinraza interest reverts to Ionis once payments hit $550 million, a 1.1x return on the $500 million funded. The structure left Spinraza to carry the deal after the Lp(a) option went to zero.

Geoff Meacham at Citi said the field still needs the full dataset to separate four explanations: the antisense mechanism, too little Lp(a) reduction, trial design, or a deeper problem with the idea that lowering Lp(a) cuts disease. That list is now the scorecard for whoever still has a running outcomes study.

Amgen Enrolled a Higher Lp(a) Floor

Amgen’s olpasiran is a GalNAc-conjugated small interfering RNA given every 12 weeks. In the phase 2 OCEAN(a)-DOSE study of 281 people, a 75 mg dose every 12 weeks produced a placebo-adjusted mean change of -97.4% at 36 weeks. The 225 mg twelve-week dose reached -101.1%. Those cuts sit well above the 72% to 80% range pelacarsen showed in its own dose-finding work.

The outcomes follow-on, OCEAN(a)-Outcomes, is a different experiment from HORIZON on three counts that now matter. It enrolled 7,297 patients and Lp(a) of at least 200 nmol/L, started December 14, 2022, and lists an estimated primary completion of March 31, 2028. Two hundred nmol/L is about 94 mg/dL on the conversion used in the HORIZON design paper, which means Amgen’s floor sits just above Novartis’s unpublished 90 mg/dL subgroup, not at the 70 mg/dL line that defined the missed overall analysis.

The primary endpoint is also narrower: coronary heart disease death, heart attack, or urgent coronary revascularization. Stroke is not in that composite. HORIZON counted stroke. If Lp(a) biology is more coronary than cerebral, Amgen’s endpoint is the friendlier of the two. If the problem is simply that a few years of lowering after a first event cannot match lifelong genetic exposure, both trials can still miss.

THE LIVE LP(A) OUTCOMES RACE

Amgen has a second, larger bet that HORIZON never placed. OCEAN(a)-PreEvent is recruiting about 11,000 people aged 50 and up with Lp(a) of at least 200 nmol/L and no prior event. That study exists because the genetic case for Lp(a) was built in people who had not yet had a crisis, and HORIZON only treated people who already had.

Lilly Still Has Time on the Clock

Eli Lilly is running ACCLAIM-Lp(a) with lepodisiran, another siRNA, in people with established disease and in a high-risk group that has not had an event. Dosing in earlier work was spaced at six to twelve months. Lilly also has muvalaplin, an oral assembly blocker, in the MOVE-Lp(a) outcomes program. Those readouts sit years behind OCEAN(a).

Silence Therapeutics has zerlasiran, an siRNA that cut Lp(a) more than 80% in phase 2, and has said it will not start a cardiovascular outcomes trial without a partner. After HORIZON, that partner is harder to find. The first-mover loss is Novartis and Ionis. The delayed-mover risk is everyone still writing a protocol that looks like HORIZON.

Guidelines Got Ahead of the Drug

U.S. and European lipid guidance already tells clinicians to test Lp(a) once. The 2026 ACC/AHA multisociety dyslipidemia guideline, published online March 13, 2026, says Lp(a) should be measured at least once in adulthood because lifestyle barely changes it. The writing committee treats values of 125 nmol/L, or 50 mg/dL, as a risk-enhancing factor tied to about 1.4-fold higher long-term risk of heart attack or stroke, and 250 nmol/L, or 100 mg/dL, as at least a two-fold rise. The prescribed response, for now, is harder LDL-C lowering and tighter control of everything else, not an Lp(a) shot.

HORIZON enrolled above 70 mg/dL, which is already past that 50 mg/dL risk flag, and still did not move the event curve in the overall group. People who followed the new testing advice now hold a number that marks inherited risk, with no targeted drug that has been shown to change outcomes. Cascade testing of first-degree relatives is also a class 1 recommendation when Lp(a) is high, familial hypercholesterolemia is present, or disease came early. That family conversation just got more awkward, not less necessary, because the genetic association has not been withdrawn.

WHAT THE 2026 U.S. GUIDELINE ACTUALLY SAYS TO DO

• Test once: Measure Lp(a) at least once in every adult for risk assessment; repeat draws are usually pointless because the level is stable.
• Treat the rest harder: High Lp(a) is a reason to push LDL-C lower and to manage blood pressure, diabetes, and smoking with more urgency.
• Check the family: If Lp(a) is high, or if disease came early, first-degree relatives should be offered the same test.

European guidance in the 2025 focused update of the ESC/EAS dyslipidaemia document points in the same direction. The testing push was built on genetics and epidemiology, not on a completed outcomes trial. HORIZON is the first of those trials, and it did not give the treatment half of that pairing a win.

LDL-C Was Already Near 65 mg/dL

Sam Tsimikas, a University of California San Diego cardiologist who has spent decades on Lp(a), wrote on September 5, 2026 that the miss was shocking and that HORIZON may have been run in some of the best-treated patients in any recent heart-outcomes study. He put baseline LDL-C at about 65 mg/dL and noted that a standard LDL-C reading includes cholesterol sitting on Lp(a) particles, so the LDL-particle burden could have been 15 to 20 points lower, near 45 to 50 mg/dL in people with very high Lp(a).

I think the story of Lp(a) therapy is beginning, not ending.

Sam Tsimikas, MD, University of California San Diego, on X

His point is practical. If the remaining apoB load is already that low, and blood pressure, diabetes, and antiplatelet therapy are already in place, a further cut in one inherited particle may not move a four-part event score over a few years. That is a different claim from saying Lp(a) never caused disease. The genetic and epidemiologic file Tsimikas cites is large, and most of it comes from community cohorts enrolled before this intensity of secondary prevention.

Jason Fung, a Toronto nephrologist, took the harsher line the same weekend, writing that genetic and epidemiology data never prove causality. Pablo Corral, past president of the Argentine Lipid Society, listed a long set of reasons not to bury the target, including treatment that starts too late, a possible benefit only in the extreme tail, oxidized phospholipid cargo that concentration may not capture, and event rates that crawled because modern background care had already eaten the residual risk.

WHERE EXPERTS DISAGREE

• The hypothesis is unproven: Jason Fung argues that association studies never established that lowering Lp(a) would prevent events, so a null outcomes trial is the expected result, not a surprise.
• The hypothesis is untested in the right window: Tsimikas and Corral argue that lifelong exposure is not the same experiment as a few years of drug after a first heart attack, stroke, or graft of peripheral disease, especially when LDL-C is already very low.
• The next drugs can still differ: Analysts at William Blair have kept room for deeper knockdown and for patients whose baseline Lp(a) sits higher than HORIZON’s overall 70 mg/dL line, which is the exact design Amgen already chose.

Both arms of HORIZON were on optimized lipid-lowering therapy. The trial cannot speak to LDL-C, because LDL-C was not the thing it randomized. It randomized pelacarsen. Reading the miss as a verdict on statins, or on the lipid hypothesis as a whole, ignores the protocol.

What Full HORIZON Data Still Must Show

Novartis and Ionis have not released event rates, hazard ratios, or p-values. Until those numbers are on a slide, every explanation is a guess. The dual primary structure makes one gap larger than the rest: the 90 mg/dL subgroup was written into the protocol and then left out of the September 4 note. If that slice is also flat, Amgen’s 200 nmol/L floor looks less protective. If it bends, HORIZON becomes a story about who was enrolled, not about whether the particle matters.

WHAT THE CONGRESS DROP MUST ANSWER

• The overall hazard ratio: A near-miss with a confidence interval that skims 1.0 is a different object from a hazard ratio sitting on 1.0 with no hint of benefit.
• The 90 mg/dL analysis: That prespecified population is the closest published match to OCEAN(a)’s entry rule, and it is still dark.
• Achieved Lp(a): Time-averaged percent and absolute drops, and how many patients crossed 50 mg/dL or 125 nmol/L, will show whether the 80 mg monthly shot did the job the model promised.
• Which events moved: Coronary death and stent crises can diverge from stroke inside a four-part composite, and Amgen’s primary already bets on that split.
• How low LDL-C really was: A corrected LDL-C, stripping out Lp(a) cholesterol, would test Tsimikas’s 45 to 50 mg/dL sketch against the actual trial file.

Pelacarsen did the laboratory job it was designed to do, in the first outcomes trial the Lp(a) field was built to wait for, and it did not buy fewer events in the overall 8,323-person file. Novartis and Ionis are out of the first-mover chair. Amgen now has to show that a deeper cut, a higher starting level, and a coronary-heavy endpoint can find a benefit that monthly antisense did not.

Disclaimer: This article is news reporting and analysis of a clinical-trial announcement and related company statements. It is informational only and is not medical advice, a treatment recommendation, or investment advice. Readers should consult a licensed physician before changing lipid testing, heart medicines, or prevention plans, and a licensed financial adviser before making decisions about Novartis, Ionis, Amgen, Lilly, Royalty Pharma, or any other security named here. Figures, trial statuses, and guidance summaries reflect the company, registry, and society sources dated through September 4, 2026 and may change when full HORIZON data or later trials are released.