Eight thousand three hundred and twenty-three patients. A composite endpoint with four ways to fail. A drug that drove its molecular target lower than almost anything else in cardiology — and moved none of those endpoints.
On September 4, Novartis reported that pelacarsen, its lipoprotein(a)-lowering therapy, missed the primary endpoint of the Phase 3 Lp(a)HORIZON trial. Lp(a) levels fell sharply. Heart attacks, strokes, and cardiovascular deaths did not follow.
What Lp(a)HORIZON actually tested
Lipoprotein(a) — written Lp(a) — is a cholesterol-carrying particle with an extra protein tail that makes it unusually good at lodging inside artery walls. It is inherited, largely fixed for life, and elevated in roughly one in five people worldwide. Unlike LDL cholesterol, the target of the statin era, Lp(a) has never had an approved therapy aimed at it.
Pelacarsen is an antisense oligonucleotide: a short synthetic strand of genetic code designed to stop the liver from making apolipoprotein(a), the protein backbone of the particle. Ionis Pharmaceuticals invented the molecule. Novartis licensed development and commercial rights in 2019 and has carried it through late-stage testing ever since.
The commercial logic was straightforward. Statins crushed LDL cholesterol and, in doing so, built the largest preventive-medicine franchise in history. Lp(a) looked like the next frontier: a second inherited lipid particle, untouched by statins, sitting in a fifth of the adult population with no drug to offer them. If the mechanism held, the label would write itself.
Lp(a)HORIZON was built as the field's cleanest experiment. It randomized 8,323 people with established atherosclerotic cardiovascular disease (ASCVD) and elevated Lp(a), set at a threshold of 70 mg/dL or above. Every participant received guideline-directed care — statins, blood-pressure control, the full standard of prevention. The question was deliberately narrow. Strip out one inherited risk factor, leave everything else optimized, and see whether the event rate falls.
It did not. The trial missed its primary composite of cardiovascular death, non-fatal heart attack, non-fatal stroke, and urgent coronary revascularization requiring hospitalization. Pelacarsen lowered Lp(a) substantially, consistent with earlier studies. Its safety profile held up. Its efficacy readout did not.
The market's judgment came fast. Novartis shares slipped 4.4% in after-hours trading, to about $153. Ionis, which discovered the drug and still holds economic interests, barely moved — its investors had already priced the class's risk. The full dataset is scheduled for presentation at an upcoming medical congress.
That is the raw result. What follows is the argument — because the same numbers can be read as a verdict on one molecule, or as a verdict on an entire hypothesis.
The case that the biology still holds
The strongest evidence for Lp(a) has never come from a drug trial. It comes from human genetics. Over the past decade, Mendelian randomization studies — a method that treats inherited gene variants as a natural experiment — have repeatedly linked lifelong exposure to high Lp(a) with higher rates of heart attack and aortic stenosis. People born with genetically low Lp(a) suffer fewer events. That is causal evidence, gathered across large populations, and no single trial erases it.
If the target is real, the null result needs an explanation. Several are live.
The first is timing. Lp(a) does its damage over decades. The trial treated people who already had advanced disease and observed them for a few years. Lowering a risk factor after plaque has formed is not the same as preventing it from forming — the artery may already be committed to its trajectory. LDL-lowering works in secondary prevention, but its benefit scales with how early and how long it is applied.
The second is magnitude. Genetic estimates suggest Lp(a) must fall by roughly 65 to 100 mg/dL to cut cardiovascular risk by about a fifth over five years. A large percentage reduction can still leave a patient with an absolute Lp(a) burden that matters. Deep percentage cuts and deep absolute cuts are separate things.
The third is statistics. The trial ran in an era of optimized care, when event rates are lower than they were when older cardiovascular trials were powered. A lower event rate makes the same treatment effect harder to detect with the same sample size. The composite mixed four components with different biology, and a mechanism that changes plaque composition may not register on a count of urgent revascularizations.
"Lp(a)HORIZON was the first pivotal Phase 3 trial to address the question as to whether lowering Lp(a) could reduce cardiovascular risk in people with established cardiovascular disease who are already receiving optimal cardiovascular care."— Brett P. Monia, chief executive officer, Ionis Pharmaceuticals
Read that sentence as an argument rather than a consolation. The trial did not prove that lowering Lp(a) fails. It showed that lowering Lp(a) failed to produce a measurable benefit in this specific population, on this timeline, against this endpoint.
The case that the target is the wrong lever
Cardiovascular medicine is full of biomarkers that looked causal and turned out to be passengers. Raising HDL cholesterol looked protective until drugs that raised it failed to reduce events. Lowering homocysteine looked promising until folate supplementation did nothing to outcomes. CETP inhibitors cut a lipid marker and, for years, cut little else.
Lp(a) may belong to a different category. It may also belong to the same one. The uncomfortable version of this readout is that Lp(a) marks risk without being a practical lever once disease is established — a reliable signpost that resists being pushed.
There is a harsher reading still. Genetics captures a lifetime of exposure, but it runs in one direction: from birth. A drug given at seventy acts on a different system than a gene that shaped risk since childhood. If the benefit of low Lp(a) is cumulative, then testing it in patients with advanced disease and following them for a handful of years asks the molecule to undo decades of damage in a short window.
There is also the possibility that matters most to anyone holding the stock: that Lp(a) reduction is a valid surrogate for a trial that will never be run, and an invalid one for the trials that will.
Investors should sit with the asymmetry. The genetic case for Lp(a) remains among the most robust in cardiovascular research. The clinical translation now has a hole in it, and the next trial has to explain why it will land differently.
The readout punishes late-stage molecules and rewards earlier, longer, outcome-driven trials.
The class's next readouts carry the burden of proof that pelacarsen was meant to settle.
What the readout changes for the field
Lp(a) is not a niche target. With roughly one in five people affected and no approved drug, it is one of the largest untreated risk factors in cardiology — a market thesis that has drawn some of the biggest balance sheets in pharma.
Amgen's olpasiran, a small interfering RNA (siRNA) therapy, is in late-stage outcomes testing against the same question. Eli Lilly's lepodisiran is advancing with a different molecular format and a longer dosing interval. Royalty Pharma holds a royalty interest in pelacarsen, which converts a scientific disappointment into a line item on a public balance sheet.
The near-term read is that pelacarsen is a setback for one program, not an exoneration of the skeptics. The longer read is that Lp(a) just became harder to sell to the gatekeepers who fund and approve decade-long cardiovascular outcomes trials.
Lp(a)HORIZON trial size
Patients with established ASCVD and elevated Lp(a), all on guideline-directed care · Novartis, 2026
Earlier this month we looked at the detection side of this problem — voice biomarkers being tested to flag heart failure before it becomes an emergency. Lp(a) is the treatment side, and it just got a harder case to make.
| Program | Company | Format | Status |
|---|---|---|---|
| Pelacarsen | Novartis / Ionis | Antisense | ✗ Missed Phase 3 endpoint |
| Olpasiran | Amgen | siRNA | ◐ Late-stage outcomes testing |
| Lepodisiran | Eli Lilly | siRNA | ◐ In development |
Public company disclosures and trial registries, September 2026
Whether the full Lp(a)HORIZON dataset shows benefit in any prespecified subgroup
Amgen's olpasiran outcomes readout — the same question, a different molecule
Whether regulators still accept Lp(a) reduction as a surrogate endpoint in future filings
Pricing and reimbursement appetite for a lifelong injectable aimed at primary prevention
Where this leaves the Lp(a) trade
For investors, the lesson repeats across drug classes: a well-validated target and a well-behaved drug can still fail to produce an approvable outcome. Biology that looks causal in a population dataset still has to survive a controlled trial, a specific population, and a fixed clock.
That does not make Lp(a) a bad bet. It makes it an early one. The programs still standing now have to answer the two questions pelacarsen could not: treat earlier, and wait longer.