Try Factor, America's #1 ready-to-eat meal delivery service.

Made from ingredients you recognize. Whole food. Nothing unnecessary. Let’s eat real.

Get 50% off your first Factor box + Free Breakfast for a year *1 free breakfast item per box for 1 year while subscription active.

Try Factor Now

For 3 years I have written some version of the same 3 sentences about lipoprotein(a). One in 5 carry it. Statins can't touch it. Diet can't move it. Those sentences always came with an implicit promise: the first real answer on whether lowering it prevents heart attacks was coming, and the trial that would deliver it had a name. Lp(a)HORIZON.

The answer arrived Friday afternoon in a 5-paragraph press release from Basel, and the answer was no.

Novartis announced that pelacarsen, an antisense oligonucleotide that shuts down apolipoprotein(a) production in the liver, did not meet the primary endpoint of Lp(a)HORIZON: a composite of cardiovascular death, nonfatal MI, nonfatal stroke, and urgent coronary revascularization requiring hospitalization, in 8,323 patients with established cardiovascular disease and Lp(a) of at least 70 mg/dL, all on guideline-directed therapy. The drug lowered Lp(a). The sponsors say the reduction was consistent with previous studies, which in phase 2 meant 35 to 80 percent depending on dose. Events did not follow.

Here is what makes this result different from an ordinary drug failure. Pelacarsen did exactly what it was designed to do, pharmacologically. The molecule is not the thing on trial. The hypothesis is: that a genetically determined particle we have measured, feared, and been unable to treat effectively for 60 years is a target whose lowering prevents disease. This was the first randomized outcomes test of that hypothesis in human history, and it came back negative.

Before anyone writes the obituary, look at the math the genetics gave us 8 years ago. It matters more today than the day it was published.

Start with what the release establishes, because a topline press release is a blunt instrument and the temptation to over-read it is enormous.

Lp(a)HORIZON randomized 8,323 patients with prior MI, ischemic stroke, or symptomatic peripheral artery disease and screening Lp(a) of at least 70 mg/dL (roughly 149 nmol/L) to pelacarsen 80 mg subcutaneously monthly or placebo, on top of optimized standard of care. The trial was event-driven, designed to run until 993 adjudicated primary events accrued, and the endpoint was pre-specified for testing both in the overall population and in a subpopulation with Lp(a) of at least 90 mg/dL. The primary endpoint was not met. Lp(a) fell. Safety was described by Ionis as acceptable. Full data go to an upcoming congress, which almost certainly means AHA in November.

That is the entire evidentiary record. No hazard ratio, no confidence interval, no achieved-Lp(a) distribution, and no word on the ≥90 mg/dL subpopulation, a silence I would caution against reading in either direction. Today's job is to frame the question correctly; the framing gets stress-tested when the data get presented.

In 2018, Burgess and colleagues published a Mendelian randomization analysis in JAMA Cardiology that should have been stapled to every commentary written about this trial. Using genetic data from more than 80,000 coronary disease cases, they asked how much Lp(a) lowering would be required to produce the same risk reduction as lowering LDL-C by 1 mmol/L, the workhorse unit of statin benefit. Their answer: roughly 101.5 mg/dL, because the genetic association with coronary disease tracked the absolute change in Lp(a) concentration, not the percentage. A 10 mg/dL drop bought about a 5.8 percent lower risk. Nothing more.

Now run HORIZON's numbers against that yardstick. Entry required Lp(a) of 70 mg/dL or higher, so a meaningful share of the trial plausibly sat in the 70s and 80s. Cut a patient at 80 mg/dL by even the full 80 percent seen at the top phase 2 dose regimen and you have removed 64 mg/dL, short of the genetic benchmark for one statin-unit of benefit, delivered over roughly 5 years rather than a lifetime. The genetics never promised this trial would win. It promised a tight race, and it pre-registered the explanation before the trial finished enrolling: percentage reductions flatter the intervention, while absolute reductions pay the bills.

I want to be honest about the limits of this argument, because it is currently doing a lot of work for people who want the hypothesis to survive, and I am one of them. We do not yet know the trial's baseline Lp(a) distribution or the achieved absolute reduction, and Mendelian randomization estimates the effect of lifelong exposure, which a 5-year drug trial was never guaranteed to reproduce in either direction. The tight-math story is the best available explanation. It is not yet an evidenced one. AHA will tell us more.

What it already does, though, is explain why the rest of the field built different trials.

Deeper percentage cuts from higher starting values. If the absolute-reduction hypothesis is right, OCEAN(a)-Outcomes is a far stronger test of it than HORIZON ever was, and lepodisiran's 17,000-patient ACCLAIM program, which reaches into primary prevention, is stronger still. If those fail too, the hypothesis is dead and the genetics community owes us an explanation. That is not where we are today.

1. Keep measuring Lp(a). Once, in everyone. The guideline recommendation to check Lp(a) once in a lifetime was never premised on having a drug for it. A high value reclassifies risk and changes how hard you treat everything else. HORIZON does not touch that logic.

2. The response to a high Lp(a) is unchanged, because it never was an Lp(a) drug. It is aggressive apoB and LDL-C lowering, blood pressure control, and the unglamorous rest. The one lever that reliably moves events in these patients is the one we already own.

3. Do not treat the number for its own sake. Niacin lowers Lp(a) modestly and failed its outcomes trials. PCSK9 inhibitors lower it a similar amount, and while exploratory analyses have associated that reduction with benefit, the causal credit plausibly belongs to the LDL lowering that comes along for the ride. HORIZON is now the strongest single argument in medicine against prescribing to a surrogate because the surrogate scares you.

4. Tell patients the honest thing. For the roughly 1 in 5 who carry elevated Lp(a): the first drug built for you did its pharmacologic job and did not prevent events; drugs that lower it much further are in trials designed with this possibility in mind; and your risk, today, is best managed with tools that work. Patients tolerate honest uncertainty far better than we assume.

The trial failed, and the drug likely dies with it. The hypothesis is wounded, not buried; its sharpest test reads out next, on a design built for exactly the weakness that may have sunk this one. Your Monday changes almost nothing: measure Lp(a) once, let a high value make you more aggressive with the levers that work, and refuse to treat a lab value on faith.

And notice what happened here, because it is the reason this newsletter has the name it does. A company ran a 6-year, 8,323-patient experiment that could end a multibillion-dollar program, got the answer no one wanted, and published it the same day. That is not the system failing. That is the only part of it working exactly as designed.

REFERENCES