Biomarkers · Clinical Literacy

    The Lp(a) Drug Worked. The Hypothesis Didn't.

    What Lp(a)HORIZON actually showed when the numbers finally arrived

    Reviewed by Dr. Sanjeev GoelNovember 12, 20268 min read
    A single coiled ribbon form on a dark surface, lit in dusty orchid

    Two weeks ago I wrote that there were no numbers attached to the Lp(a)HORIZON trial, that anyone quoting one was making it up, and that I would write the results up when they existed. They exist. They were presented at the American Heart Association meeting on Saturday 7 November and published in the New England Journal of Medicine the same day. The short version is stranger, and more interesting, than "the drug failed."

    The 60-Second Answer

    Pelacarsen lowered lipoprotein(a) by 78.4% at twelve months. Over a median of 4.7 years in 8,322 people who had already had a heart attack, an ischaemic stroke or symptomatic peripheral arterial disease, it prevented nothing. The primary composite of cardiovascular death, non-fatal myocardial infarction, non-fatal ischaemic stroke and urgent coronary revascularisation occurred in 12.3% on the drug and 12.1% on placebo — hazard ratio 1.02, 95% CI 0.90 to 1.15, P = 0.76.[1] That is not a near miss. There is no signal anywhere in it: every component, every key secondary endpoint and every prespecified subgroup sits on top of the null. The drug did exactly what it was designed to do, to a degree that would have been science fiction ten years ago, and it made no difference to anyone.

    Why ‘the drug failed’ is the wrong summary

    A drug failing usually means it didn't work. This drug worked beautifully. Median Lp(a) fell from around 236 nmol/L at baseline to 64.4 nmol/L at twelve months on treatment, against 233.4 nmol/L on placebo — a 78.4% reduction on drug versus 3.2% on placebo, a placebo-adjusted difference of 77.7 percentage points (95% CI 77.0–78.4).[1] The pharmacology is not in question.

    What failed is the inference — the step from Lp(a) predicts events to therefore removing Lp(a) prevents events, in this population, at this stage of disease.

    That distinction matters because the two claims have very different consequences for you, and only one of them has been tested here.

    Figure 1

    The dissociation

    Lp(a) fell by 78%. The event rate did not move.

    The left panel is what the drug did. The right panel is what it changed. A 78% reduction in the target and a hazard ratio of 1.02 is not a failed drug — it is a successful drug and a failed hypothesis. Both y-axes start at zero.

    Source: Cho L, Nicholls SJ, Frndak S, et al., N Engl J Med, 7 November 2026. 8,322 patients analysed, median follow-up 4.7 years.

    Three questions that keep getting collapsed

    Does Lp(a) mark risk? Yes, and nothing in this trial touches that. The epidemiology and the Mendelian randomisation work are exactly where they were the day before.

    Is Lp(a) causal? Probably, on genetic grounds. One negative trial of one drug in late disease does not overturn that, and anyone telling you it does is moving faster than the evidence.

    Does lowering Lp(a) late, in someone who already has established disease and well-controlled LDL-C, prevent the next event? This is the question Lp(a)HORIZON asked. The answer is no.

    Everyone in this trial had already had the event. The trial says nothing — not one thing — about a 45-year-old with an Lp(a) of 200 nmol/L and clean coronary arteries. That person was not studied.

    Steven Nissen, who is not a man given to softening negative findings, put it this way: "A single trial does not provide definitive evidence of a lack of benefit."[3]

    Figure 2

    Every endpoint sits on the null

    Hazard ratios with 95% confidence intervals

    Hazard ratio, 95% CI — all cross 1No difference (1.0)

    Seven analyses, seven confidence intervals crossing 1.0. The highest-Lp(a) subgroup — the group where an effect should have been easiest to see — returned a hazard ratio of exactly 1.00. Axis is truncated and does not start at zero.

    Source: Cho L, Nicholls SJ, Frndak S, et al., N Engl J Med, 7 November 2026, and AHA Scientific Sessions 2026 late-breaking presentation.

    The part most of the coverage skipped

    The efficacy result led every headline. The safety data did not, and it is the part a patient would actually want to know.

    Adjudicated anaphylaxis occurred in 19 patients on pelacarsen and none on placebo. Eight hypersensitivity reactions led to hospitalisation. There were no allergy-related deaths. After a suspected anaphylaxis case in March 2023, the trial moved injections into medically supervised settings for six months before resuming at-home dosing.[1]

    MeasurePelacarsenPlacebo
    Influenza-like reactions31.9%5.8%
    Injection-site reactions25.6%2.9%
    Permanently stopped for adverse events20.6%7.5%
    Stopped the regimen early, any reason35.1%18.3%

    One in five people on the drug stopped it because of side effects. In my 25 October piece I flagged a JAMA report that roughly half the participants at two European sites had withdrawn, and said the dropout figures would be worth watching. They were. The trial-wide numbers are now the better ones to quote.

    One oddity to resist: serious adverse events were lower on pelacarsen, 37.3% against 41.1%. That is not a benefit. It is the kind of imbalance that turns up in a trial this size by chance, and nothing was powered to detect it. I mention it because somebody is going to post it as though it meant something.

    So what do you do with a high Lp(a)?

    This has not changed, and it never depended on a drug existing.

    Know your number. Lp(a) is largely genetically set, and major guidance recommends measuring it at least once in adults. You do not need to fast for it.

    If it is 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 with outcome evidence behind it, and an elevated Lp(a) is a reason to be more aggressive about all of it.

    Don't let a negative drug trial talk you out of a useful measurement. A drug that doesn't prevent events in people who already have disease tells you very little about whether a lifelong elevation matters. Those are different time horizons and different patients.

    And if your result sits near a threshold, the measurement itself deserves a second look. Agreement between commercial Lp(a) assays is good in general but imperfect close to the clinical cut-offs — which matters more now that the number's job is purely to stratify risk rather than to select a treatment.

    What I'd watch next

    The other three agents. Olpasiran, lepodisiran and zerlasiran are still in outcome trials, with different mechanisms, different dosing and in some cases different populations. This result does not predict theirs.

    If those also come back flat, the causal story gets much harder to hold. If one of them is positive, the explanation for Saturday's result will probably turn out to be about when you intervene rather than whether Lp(a) matters.

    There is a version of this where the answer was never the drug but the timing — and we will not know which for several years yet.

    The practical bottom line

    • Pelacarsen lowered Lp(a) by 78.4% and changed no clinical outcome: hazard ratio 1.02, P = 0.76.
    • Everyone studied had already had a heart attack, stroke or peripheral arterial disease. Primary prevention was not tested.
    • Nineteen cases of anaphylaxis on drug versus none on placebo, and one in five stopped treatment for side effects.
    • None of this changes the case for knowing your Lp(a) once, or for treating every other risk factor harder if it is high.
    • Three other Lp(a)-lowering drugs are still in outcome trials. This result does not predict theirs.
    • Nothing here is a reason to start, stop or change a medication. That is a conversation with your own physician.

    Frequently asked questions

    Did the Lp(a) drug fail?

    It depends what you mean. Pelacarsen lowered lipoprotein(a) by 78.4%, which is exactly what it was designed to do. It did not reduce cardiovascular events: 12.3% versus 12.1% on placebo, hazard ratio 1.02. The drug worked; the hypothesis that lowering Lp(a) late in established disease prevents events did not.

    Does this mean Lp(a) doesn't matter?

    No. Lp(a)HORIZON enrolled only people who had already had a heart attack, stroke or peripheral arterial disease, with well-controlled LDL-C. It tested whether lowering Lp(a) at that point helps. It did not test, and cannot answer, whether a lifelong elevation raises risk — and the genetic and epidemiological evidence that it does is unaffected.

    Should I still have my Lp(a) measured?

    Yes. Major guidance recommends measuring it at least once in adults, and no fasting is required. Knowing the number changes how aggressively every other modifiable risk factor should be managed, whether or not a drug exists.

    Were there safety problems with pelacarsen?

    Adjudicated anaphylaxis occurred in 19 patients on pelacarsen and none on placebo, with eight hospitalisations for hypersensitivity and no deaths. Flu-like and injection-site reactions were several times more common on the drug, and 20.6% of patients stopped it permanently because of adverse effects, against 7.5% on placebo.

    Are other Lp(a) drugs still being tested?

    Yes. Olpasiran, lepodisiran and zerlasiran are in ongoing cardiovascular outcome trials. They use different mechanisms and dosing, and the pelacarsen result does not predict theirs.

    Disclaimer: This article is for education and is not medical advice. Lp(a) results should be interpreted by a clinician who knows your history. Do not start, stop or change any medication based on this article.