For decades, cardiologists had a name for the problem they couldn't fix. Elevated lipoprotein(a) runs in families, barely responds to diet or exercise, and turns up again and again in patients who have heart attacks despite normal cholesterol and clean habits. Novartis's own press release calls it an "inherited, independent and causal cardiovascular disease risk factor."
Then the company reported what happened when it lowered that factor by a lot.
In a trial of 8,323 people followed for seven years, pelacarsen reduced Lp(a) as intended and still did not reduce cardiovascular death, nonfatal heart attacks, nonfatal strokes, or urgent coronary revascularizations requiring hospitalization compared with placebo. The drug hit its biological target. The patients did not do better.
The drug hit its biological target. The patients did not do better.
Until that Friday, September 4, the field had been bracing for a triumph. Three companies were running large trials of Lp(a)-lowering drugs, and doctors hoped the first result would be "as important as a statin," as Gina Kolata reported in The New York Times. Novartis released the news late in the afternoon, after the markets closed.
The trial, called Lp(a)HORIZON, enrolled adults who already had cardiovascular disease — a prior heart attack, an ischemic stroke, or symptomatic peripheral artery disease — and an Lp(a) level of at least 70 mg/dL. Participants injected 80 milligrams of pelacarsen or a placebo once a month, on top of guideline-directed care including lipid-lowering and blood-pressure drugs.
The primary endpoint combined four outcomes, and the design called for testing it both in the overall group and in a prespecified subgroup with starting levels of at least 90 mg/dL. Novartis said lower Lp(a) levels were achieved. The events did not follow.
Cardiologists reacted with something closer to bewilderment than disappointment. "There's something we don't understand here," said Dr. John Alexander, a cardiologist at Duke. "This is a huge blow," Dr. David Maron, president of the American Society for Preventive Cardiology, told The Times. Dr. Harlan Krumholz of Yale drew the general lesson: even strong evidence that two things occur together is never proof that one causes the other.
The case for Lp(a) had always rested on more than correlation. Roughly 20 percent of people worldwide have elevated levels, and nearly a third of those with premature cardiovascular disease do. About 90 percent of a person's level is set by their genes.
The risk is continuous rather than neatly divided into safe and dangerous. Levels at or above 50 mg/dL, or 125 nmol/L, are treated as a risk-enhancing factor, and levels around 250 nmol/L may double risk. Statins don't lower Lp(a); if anything, they raise it — an "inconvenient truth," as one cardiologist put it.
That is why the archetypal patient matters. A 44-year-old masters swimmer with a total cholesterol of 160, a low-fat diet, and no obvious risk factors had a heart attack caused by plaque and a clot blocking a coronary artery. Testing found an HDL of 25 and an Lp(a) of 80. Nothing about his daily life explained the event. One inherited particle did.
So what went wrong? The topline disclosure cannot say, and several explanations remain live. Pelacarsen may not have lowered Lp(a) far enough, early enough, or long enough to reverse risk accumulated over decades of exposure. The particle may do harm through properties a concentration measurement doesn't capture.
Any incremental benefit may also have been too small to separate from chance in patients already on intensive cholesterol, blood-pressure, and antithrombotic therapy. And as routine care improves, event rates fall, which is good for patients and brutal for trials trying to prove an added benefit.
Crucially, Novartis released almost nothing quantitative. No treatment effect, no confidence interval, no event counts, no result in the highest-baseline subgroup. Barclays noted that Novartis had said even a 13 percent benefit would have been statistically significant in the overall population, and that reading between the lines, pelacarsen appears to have fallen materially short.
Jefferies analysts said they would be watching for any correlation between starting Lp(a) levels, the magnitude of reduction, and clinical benefit — and for the statistical impact of patients who didn't complete the trial. Full data are promised at a medical meeting in November.
The failure weakens the hypothesis without killing it, at least for now. "The Lp(a) hypothesis is weakened, but not disproven," Citi analysts wrote, noting that the miss could reflect the drug's mechanism, the trial's design, or a genuine challenge to the whole idea.
Other programs are further behind and, on paper, more powerful. Amgen's olpasiran reduced Lp(a) by more than 95 percent in mid-stage testing, with outcomes due in 2028. Eli Lilly's lepodisiran produced a nearly 94 percent reduction and is being tested in a broader population, including people without established disease, with data expected in 2029. Merck paid nearly $2 billion in biobucks to partner on another Lp(a) inhibitor.
Deeper lowering is the obvious next bet. BMO Capital Markets said the Novartis result "raises the burden of proof" for those rivals. William Blair told investors it saw "meaningful risk to a potential future in Lp(a)-driven CVD trials."
The stakes were never purely scientific. Analysts had modeled peak annual sales of roughly $4 billion to $5 billion for pelacarsen; William Blair estimated $6 billion in U.S. sales alone. Novartis shares fell more than 7 percent in postmarket trading, Amgen nearly 7 percent, Ionis 10 percent.
This is a familiar shape in heart medicine. A drug meant to raise "good" cholesterol failed years ago. An anti-inflammatory heart drug failed just last month. Each time, the biology that looked causal from a distance refused to behave causally when interrupted.
Lp(a) remains a real explanation for a real kind of misfortune: the heart attack that shouldn't have happened, in the patient who did everything asked of them. Guidelines now recommend testing every adult once in a lifetime, and that advice stands.
What has changed is the confidence that measuring a risk factor and moving it are the same act. For patients with high Lp(a), the number still means something. What it no longer promises is that anyone can take it away and hand back the years.