Your Lp(a) Is High and the First Drug to Lower It Just Failed

Lp(a)HORIZON randomised 8,323 patients with established cardiovascular disease and elevated Lp(a) across more than 900 sites, to monthly injections of pelacarsen or placebo.^[1][2] Background therapy was well controlled — mean LDL-C has been reported in commentary on the trial at around 66 mg/dL, which matters, and I will come back to it.^[6]

What Novartis released on 4 September was a top-line statement: lower Lp(a) levels were achieved, and the primary composite endpoint was not reduced. The company's chief medical officer said, "These are not the results we hoped for," adding that the findings did not demonstrate that lower Lp(a) translated into reduced cardiovascular risk in the overall study population.^[1]

What has not been released is everything you would need to interpret it. There is no hazard ratio. There are no event counts, no confidence interval, no p-value — and, remarkably, not even a figure for how much Lp(a) actually fell.^[2] If you have seen a number attached to this trial in the last seven weeks, it did not come from the trial.

That is worth saying plainly, because this is exactly the gap into which confident commentary flows. Three journal commentaries on the result appeared in a single week in October. All three are opinion. None of them has the data either.

One genuinely new detail did emerge, from JAMA's news desk on 16 October: at two European sites — Geneva and Basel — the principal investigator reported that roughly half the participants withdrew.^[2] Because the trial was double-blinded, nobody yet knows whether those dropouts concentrated in the drug arm or the placebo arm. A retention problem of that size at some sites is a real interpretive question, and it is better raised now, before the data land, than afterwards when it will look like special pleading. Per-protocol analyses, which exclude dropouts, are expected alongside the primary results.^[2]

On 14 October, a recurrent-event analysis of the UK Biobank was published in the European Journal of Preventive Cardiology.^[4] It asked a different question from the trial: not "does lowering Lp(a) help?" but "what does a high Lp(a) actually cost you over a lifetime?"

237,729 participants of European ancestry, free of atherosclerotic cardiovascular disease at baseline, followed for a median of 14.1 years, during which 10,731 coronary, peripheral and carotid revascularisation procedures were performed. Comparing Lp(a) at or above 105 nmol/L — the ESC/EAS 2025 threshold — against below it:

• At least one procedure: 4.36% vs 2.96%
• At least two: 1.53% vs 0.95%
• At least three: 0.36% vs 0.18%

Overall, the recurrent-event hazard ratio was 1.50 (95% CI 1.42–1.60), and the incidence rate ratio 1.55 (1.45–1.65). Broken down by which procedure it was: first HR 1.48 (1.40–1.56), second 1.28 (1.09–1.50), third 1.23 (0.94–1.62) — and note that the third crosses 1, so the signal thins as the procedures accumulate. The authors describe the association as robust for coronary revascularisation, supportive for peripheral, and inconclusive for carotid, where there was too little statistical power.^[4]

The honest limitation here is a good one to understand, because it is not the usual hand-wave. A revascularisation is a clinical decision, not purely a biological event. Whether a stent happens depends on who gets sent for angiography, on local practice patterns, and — awkwardly for this analysis — potentially on whether the treating team knew the patient's Lp(a) was high. UK Biobank is also European-ancestry only and healthier than the general population. This is an association, not a causal dose of risk.

But it is a large, long, consistent association, and it is measuring something the trial was not designed to measure at all.

The two findings sit together without strain once you separate three questions that get collapsed into one.

Does Lp(a) mark risk? Yes. That is what the October paper measures, and it is consistent with Mendelian randomisation work and decades of epidemiology.

Is Lp(a) causal? Probably, on genetic grounds — but this is a live argument among cardiologists right now and one failed trial of one drug does not settle it. I am not going to settle it on a blog.

Does lowering Lp(a) in someone who already has established disease, on top of well-controlled LDL-C, prevent events? That is the question Lp(a)HORIZON asked, and the answer appears to be no — or at least, not detectably, in that population, over that duration, with that background therapy.

That third question is narrower than it sounds. A mean LDL-C around 66 mg/dL means these patients were already being treated aggressively. If most of the modifiable risk has already been removed by other means, the remaining room for an Lp(a) drug to show benefit is smaller — which is a plausible reason a trial can be underpowered against its own control arm doing well. Three other Lp(a)-lowering agents — olpasiran, lepodisiran and zerlasiran — are still in outcome trials. The pelacarsen result does not predict theirs.

This is the part that hasn't changed, and it was never dependent on a drug existing.

Measure it once. Lp(a) is largely genetically determined, and major guidance recommends measuring it at least once in adults. A small study published the same week as the paper above looked at within-day variation in 34 people across seven timepoints: Lp(a) ran about 6.2% below fasting at 17:30 and 7.0% lower at midnight, absolute differences of roughly 3 nmol/L.^[5] Practically: you don't need to fast for it, and a single measurement is generally sufficient for the decision it informs — though those authors suggest re-checking a result that sits borderline. (Whether it drifts over years is a different question that study did not address, and the "measure once and never again" version of the advice is firmer than the evidence strictly supports.)

If it's high, treat everything else harder. ApoB, blood pressure, smoking, cardiorespiratory fitness, glucose. This is not a consolation prize. It is the only part of the picture that currently has outcome evidence behind it, and an elevated Lp(a) is a reason to be more aggressive about all of it, not less.

Don't let the trial talk you out of knowing your number. The result that a drug failed is not a result that a risk marker is meaningless. Knowing you carry a lifetime elevation in risk changes how hard you and your physician work on everything that is modifiable — which is the whole point of a long-term risk strategy.

The full presentation is scheduled for the AHA Scientific Sessions in Chicago, which run 6–9 November 2026 at McCormick Place. Lp(a)HORIZON appears in the late-breaking session "Translating Science into Prevention: Lessons from Landmark Trials" on Saturday 7 November, 8:30–9:45 am Central, with slots for the primary results and extended analyses of cardiovascular outcomes.^[3] The same session carries ZEUS, the ziltivekimab trial, which also missed.

Three things will be worth looking for: the actual hazard ratio and confidence interval; the achieved Lp(a) reduction, which tells you whether the drug did its pharmacological job; and the dropout and adherence figures, given what JAMA reported about those two European sites.

I will write that up when there are numbers. There are no numbers yet.