Biomarkers · Clinical Literacy

    Lp(a) Fell 78 Percent. The Heart Attacks Didn't.

    The Full Lp(a)HORIZON Results Are In

    Reviewed by Dr. Sanjeev GoelNovember 9, 202612 min read
    A length of heavy chain with one open, disconnected link on a dark surface, rim-lit in sapphire blue

    In September, Novartis said only that Lp(a)HORIZON had missed. On 7 November the full data were presented in Chicago and published in the New England Journal of Medicine — and they are more decisive than the word "missed" suggests. This is what 8,350 patients and nearly five years bought us, and what it changes if your own Lp(a) is high.

    The 60-Second Answer

    Pelacarsen lowered lipoprotein(a) by 78.4% at twelve months and did not prevent a single cardiovascular event. Over a median 4.7 years, the primary endpoint occurred in 12.3% of patients on pelacarsen and 12.1% on placebo — hazard ratio 1.02 (95% CI 0.90 to 1.15).[1][3] Every re-cut of the data — on treatment, uninterrupted treatment, total events, biggest Lp(a) responders, highest baseline Lp(a) — landed on the same answer.[2] This does not mean a high Lp(a) is harmless; it remains one of the strongest genetic predictors of vascular disease we can measure.[7] It means that one particular drug, at the reduction it achieved, over the time it was given, did not convert that risk into benefit. Four larger trials of three other drugs are still running.[10][11][12]

    What the trial actually tested

    Lp(a)HORIZON randomised 8,350 adults with established cardiovascular disease and a lipoprotein(a) of at least 70 mg/dL (about 148 nmol/L) to monthly subcutaneous pelacarsen 80 mg or matching placebo, 1:1, across 797 sites in 42 countries, between December 2019 and July 2022.[1][5] Eight thousand three hundred and twenty-two entered the efficacy analysis. Median follow-up was 4.7 years. The primary endpoint was a four-part composite: death from cardiovascular causes, non-fatal myocardial infarction, non-fatal ischaemic stroke, or urgent coronary revascularisation.

    This was not a trial of mild risk in untreated people. Mean age was 59.7 and 27% were women; 76.6% had already had a heart attack. Median baseline Lp(a) was 235.6 nmol/L — roughly 108 mg/dL, about four times the level at which most guidelines start paying attention — and 78.6% of participants were above 90 mg/dL.[5] Crucially, the rest of their risk was already well managed: median LDL cholesterol was about 65 mg/dL, 77.6% were on a high-intensity statin, median systolic blood pressure was 125 mm Hg and median hsCRP was 0.8 mg/L. The trial's presenter, Steven Nissen, described the background therapy as among the best achieved in any cardiovascular outcome trial.[4]

    That design choice matters in both directions. It is the cleanest possible test of whether Lp(a) adds risk on top of well-treated LDL — and it is also a test in which any remaining Lp(a)-attributable risk is a small slice of a small absolute risk.

    The biomarker moved. Nothing else did.

    The drug worked on the thing it was designed to work on. Median Lp(a) fell from 235.6 nmol/L to 64.4 nmol/L at twelve months on pelacarsen, against 233.4 nmol/L on placebo — a 78.4% reduction, and a treatment difference the investigators put at 77.7 percentage points.[1][5] Averaged across the whole follow-up, Lp(a) was 161.2 nmol/L lower than baseline on pelacarsen and 17.1 nmol/L lower on placebo.

    The clinical endpoint did not move at all. Five hundred and eleven of 4,157 patients on pelacarsen (12.3%) and 503 of 4,165 on placebo (12.1%) had a primary endpoint event — hazard ratio 1.02, 95% confidence interval 0.90 to 1.15, not statistically significant.[1][3][5] Note what that confidence interval excludes. It does not merely fail to prove benefit; its lower bound sits at 0.90, which rules out anything larger than a 10% relative risk reduction. A near-miss looks like a hazard ratio of 0.88 with an interval brushing 1.00. This looks like nothing.

    Figure 1

    The drug hit its target and missed its purpose

    PlaceboPelacarsen

    Median Lp(a) at 12 months (nmol/L)

    Primary endpoint over 4.7 years (% of patients)

    Read down each panel, not across — the two panels are in different units and answer different questions. The left panel is the pharmacology; the right panel is the medicine.

    Source: Lp(a)HORIZON, New England Journal of Medicine, 7 November 2026; figures as reported in conference coverage of the presentation.

    Every way they cut it, the answer was the same

    A companion paper in NEJM Evidence reports the analyses that usually rescue a neutral trial, and none of them did.[2] Restricted to time on treatment, the hazard ratio was 0.98 (0.85 to 1.12). Restricted to patients whose treatment was never interrupted for more than 60 days, it was 1.09 (0.89 to 1.34). Counting every event rather than only the first, the rate ratio was 1.00 (0.87 to 1.14) across 644 events on pelacarsen and 646 on placebo.

    Then the two analyses that should have found a signal if one existed. Among patients still alive and event-free at three months whose Lp(a) had fallen by at least the median 170 nmol/L, the hazard ratio was 1.09 (0.92 to 1.29). Among those who got their Lp(a) down to at or below the median achieved level of 69.5 nmol/L, it was 1.10 (0.93 to 1.29).[2] The biggest responders did no better than the smallest. Nor did results vary by baseline Lp(a), baseline LDL cholesterol, baseline hsCRP, or how recently the qualifying cardiovascular event had happened.

    One honest caveat about the on-treatment numbers, because they look stranger than they are. The raw counts in that analysis favour pelacarsen — 380 events (9.1%) versus 434 (10.4%) on placebo — yet the hazard ratio is 0.98.[2] That is because far more patients stopped pelacarsen than stopped placebo, so mean exposure was 3.5 years against 4.0 years, and an on-treatment analysis simply stops counting people after they quit.[5] Fewer events accrue because less time is observed, not because less disease occurred. In a trial where a third of the treated arm came off drug, "on treatment" is the least trustworthy lens available, and the hazard ratio is the honest number.

    Figure 2

    Ten ways of asking the same question

    Hazard ratios for pelacarsen versus placebo, with 95% confidence intervals. A value of 1.0 means no difference. Every interval crosses it.

    Sources: Lp(a)HORIZON primary report, NEJM, and extended analyses, NEJM Evidence, both 7 November 2026.

    Four explanations, and how good each one is

    Not enough absolute reduction. This is the most serious argument, and it rests on genetics. A Mendelian randomisation analysis of 20,793 coronary cases with individual participant data and 62,240 more in validation found that the benefit of lowering Lp(a) is proportional to the absolute fall, not the percentage — and estimated that a 101.5 mg/dL reduction (95% CI 71.0 to 137.0) would be needed to match the coronary benefit of lowering LDL cholesterol by 1 mmol/L, "a change that has been shown to produce a clinically meaningful reduction in the risk of CHD."[6] Lp(a)HORIZON's time-averaged reduction was 161.2 nmol/L. Using the trial's own paired units (235.6 nmol/L reported as 108.4 mg/dL), that is roughly 74 mg/dL, and roughly 66 mg/dL net of the placebo arm's drift — at or below the bottom of that confidence interval. That conversion is our arithmetic from two published figures, not an analysis the investigators published. At the meeting, discussant Jeffrey Berger made the same point from the genetic data, suggesting something closer to a 90 mg/dL reduction might be required.[4] If this explanation is right, a 78% reduction was impressive pharmacology and insufficient medicine, and drugs that lower Lp(a) by more than 90% have a real shot.

    Not enough time. Lp(a) is set at birth and acts across a lifetime; the trial intervened for under five years in people in their sixties who had already had their heart attack. Nissen raised exactly this: "it may be that a person needs to have lifelong Lp(a) reduction."[5] The difficulty is that this argument is close to unfalsifiable by any trial anyone will ever run, and the same objection was raised before statin and PCSK9 trials that nonetheless produced benefit in five years or less.

    The wrong population. Everyone enrolled had established disease and excellent LDL control. Nissen noted the trial says nothing about primary prevention.[3] But the honest counterweight is that secondary prevention is where event rates are highest and effects are easiest to detect. A drug that cannot show benefit here faces a harder, not easier, task in lower-risk people.

    Lp(a) is causal but not reversible in the way we assumed. The uncomfortable possibility. Lp(a) particles may do their damage early and structurally — through decades of retention in the arterial wall, or through the oxidised phospholipids they carry — such that clearing them from plasma late does not undo the plaque. Nothing in this trial distinguishes that from the dose and duration explanations. Nissen was blunt about the risk of over-reading: "A single trial does not provide definitive evidence of a lack of benefit," and "I am worried that people... will bail out on the Lp(a) hypothesis. That would be a tragic outcome."[3][4] Discussant Jared Spitz was blunt in the other direction: "This was about as neutral a trial as a trial can get."[4]

    What the drug cost the people who took it

    A neutral efficacy result still has a safety column, and here it is not neutral. Drug-related adverse events occurred in 56.8% on pelacarsen against 18.0% on placebo.[5] Flu-like constitutional symptoms affected 31.9% versus 5.8%, injection-site reactions 25.6% versus 2.9%, and investigator-reported hypersensitivity 17.1% versus 5.4%.[1][5] Nineteen patients had adjudicated anaphylaxis against none on placebo; eight hypersensitivity reactions led to hospitalisation, and none were fatal. After a suspected anaphylaxis in March 2023 the trial moved injections into supervised settings for six months before resuming at-home dosing.[1]

    The consequence shows up in adherence. Twenty-point-six percent of the pelacarsen arm stopped the drug because of an adverse event, against 7.5% on placebo, and 35.1% stopped for any reason against 18.3%.[5] The one number that runs the other way is worth printing: serious adverse events were less common on pelacarsen, 37.3% versus 41.1% — which is what you would expect when the two arms have the same event rate and the treated arm is observed for less time.

    Figure 3

    Tolerability, in percentages of each arm

    PlaceboPelacarsen

    The last row runs the other way: serious adverse events were less frequent on pelacarsen, consistent with equal event rates and shorter time on drug in the treated arm.

    Source: Lp(a)HORIZON safety data as reported in conference coverage of the 7 November 2026 presentation.

    What did not change: Lp(a) is still a risk marker

    It is worth being precise about what a failed drug trial does and does not overturn. Lp(a)HORIZON tested a treatment. It did not test whether elevated Lp(a) predicts disease, and nothing in it weakens that evidence.

    The most recent large analysis makes the point. In 237,729 UK Biobank participants free of cardiovascular disease at baseline and followed a median 14.1 years, an Lp(a) at or above the ESC/EAS threshold of 105 nmol/L was associated with a 50% higher rate of revascularisation procedures (hazard ratio 1.50, 1.42 to 1.60) and a 55% higher procedure count over follow-up (incidence rate ratio 1.55, 1.45 to 1.65).[7] The gap widened with repeat procedures: 4.36% versus 2.96% had at least one, 1.53% versus 0.95% had two or more, 0.36% versus 0.18% had three or more. The authors grade their own finding carefully — robust for coronary revascularisation, supportive for peripheral, and inconclusive for carotid for want of statistical power.

    The guidelines had already anticipated this result. The 2025 focused update of the ESC/EAS dyslipidaemia guidelines recommends measuring Lp(a) at least once in every adult's lifetime (Class IIa, Level B), treats a level above 50 mg/dL or 105 nmol/L as a risk modifier that can move a moderate-risk patient into the high-risk category — and states plainly that "it has not yet been demonstrated whether Lp(a) reduction reduces the risk of ASCVD."[9] That sentence was written before these data and is unchanged by them. One practical refinement: because Lp(a) drifts down slightly over the course of a day — by about 6 to 7% by evening in a study of 34 patients sampled at seven timepoints — a borderline result is worth repeating rather than treated as final.[8]

    Elevated Lp(a) is a reason to be more aggressive about everything you can change.

    What is still coming and when

    Three other drugs are in phase 3 outcome trials, and between them they will enrol more than 45,000 people. Two are small interfering RNAs that silence apolipoprotein(a) production and have reduced Lp(a) by more than 90% in earlier studies; one is an oral small molecule that blocks assembly of the Lp(a) particle. None of them reports before 2028.

    DrugTypeTrialParticipantsPrimary completion
    OlpasiransiRNAOCEAN(a) Outcomes (NCT05581303)7,297March 2028
    LepodisiransiRNAACCLAIM-Lp(a) (NCT06292013)17,300March 2029
    Muvalaplinoral small moleculeNCT0715777410,450March 2031
    OlpasiransiRNAprimary prevention (NCT07136012)11,000October 2031

    Planned enrolment and primary completion dates as listed on ClinicalTrials.gov, accurate as of November 2026. Completion dates on trials of this size commonly move.

    Two smaller phase 3 imaging trials are also running, testing whether lepodisiran (252 participants, completing December 2029) or olpasiran (406 participants, completing June 2028) changes coronary plaque on CT.[10][11] Those will report a mechanism, not an outcome — a useful distinction to hold on to when the headlines arrive.

    The sequencing matters for how much weight to put on Lp(a)HORIZON. If the dose explanation is correct, OCEAN(a) Outcomes — the same secondary-prevention question with a drug that lowers Lp(a) by more than 90% — is the trial that tests it, and it reports in about sixteen months.

    The practical bottom line

    • Keep the number. Lose the expectation of a pill. A high Lp(a) still identifies real, measurable risk. What changed on 7 November is the near-term prospect of lowering it as treatment.
    • Nothing about your current treatment should change because of this trial. Pelacarsen was never available outside the study.
    • Spend the risk information on the things that are modifiable. In practice this means LDL cholesterol and apolipoprotein B driven lower than you would otherwise target, blood pressure, smoking, glucose and fitness — the levers that have outcome trials behind them. A high Lp(a) is a reason to treat those harder, not a reason to wait for an Lp(a) drug.
    • One measurement, repeated once if borderline. Lp(a) is largely genetic and is a once-in-a-lifetime test for most people, but same-day variation of 6 to 7% means a result sitting right on a threshold deserves a second sample.
    • Family matters. Because Lp(a) is inherited, an elevated level in you is information about your first-degree relatives, and that is worth acting on even though no drug is yet available.
    • This is not the last word. Four outcome trials of three drugs are running, the first reporting in 2028. On 25 October we wrote about what the topline announcement did and did not say.

    Frequently asked questions

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

    No. The trial tested a drug, not the risk marker. In 237,729 UK Biobank participants, an Lp(a) at or above 105 nmol/L was associated with a 50% higher rate of arterial revascularisation over 14 years. What the trial showed is that lowering Lp(a) by 78% for under five years, in people who had already had a cardiovascular event and whose LDL was already well controlled, did not reduce further events.

    Why didn't a 78% reduction help?

    The leading explanation is that the benefit of lowering Lp(a) tracks the absolute fall, not the percentage. Genetic analyses estimate that a reduction of roughly 100 mg/dL may be needed to produce a clinically meaningful coronary benefit; Lp(a)HORIZON's time-averaged reduction works out to roughly 66 to 74 mg/dL using the trial's own units. Duration, population and the possibility that late Lp(a) lowering simply cannot undo established plaque are the other candidates, and this trial cannot separate them.

    Should I still get my Lp(a) measured?

    Current guidance is unchanged. The 2025 ESC/EAS focused update recommends measuring Lp(a) at least once in every adult's lifetime and treating an elevated level as a risk modifier. A borderline result is worth repeating, because levels drift down about 6 to 7% over the course of a day.

    Will the other Lp(a) drugs work?

    Unknown, and the honest position is that this result lowers the prior. Olpasiran, lepodisiran and muvalaplin lower Lp(a) more than pelacarsen did, which is the specific reason to think they might still succeed if the absolute-reduction explanation is right. The first outcome data are due in 2028.

    Was pelacarsen safe?

    It was tolerable for most but demanding. Drug-related adverse events affected 56.8% versus 18.0% on placebo, flu-like symptoms 31.9% versus 5.8%, injection-site reactions 25.6% versus 2.9%. Nineteen patients had adjudicated anaphylaxis and none died. Just over a third of the treated arm stopped the drug for some reason.

    What should I actually do if my Lp(a) is high?

    Discuss it with your physician as a reason to treat your modifiable risk more aggressively than your other numbers alone would suggest — particularly LDL cholesterol and apolipoprotein B, blood pressure and smoking — and to consider testing first-degree relatives, since Lp(a) is inherited. There is no approved Lp(a)-lowering therapy.

    References

    1. [1]Cho L, Nissen SE, et al. Pelacarsen and Cardiovascular Outcomes in Patients with High Lipoprotein(a). N Engl J Med. 7 November 2026. doi:10.1056/NEJMoa2610159
    2. [2]Nissen SE, Cho L, Frndak S, et al. Extended Analyses of Pelacarsen and Cardiovascular Outcomes in Patients with High Lipoprotein(a). NEJM Evid. 7 November 2026. doi:10.1056/EVIDoa2600373
    3. [3]American College of Cardiology. Lp(a) HORIZON: Pelacarsen vs. Placebo in Patients With High Lp(a). Latest in Cardiology, 6 November 2026
    4. [4]Healio Cardiology. Pelacarsen no better than placebo in patients with high lipoprotein(a), heart disease. AHA Scientific Sessions coverage, 7 November 2026
    5. [5]Cleveland Clinic Consult QD. Lipoprotein(a) Reduction With Pelacarsen Has No Impact on Cardiac Events, Lp(a)HORIZON Finds. 7 November 2026
    6. [6]Burgess S, Ference BA, Staley JR, et al. Association of LPA Variants With Risk of Coronary Disease and the Implications for Lipoprotein(a)-Lowering Therapies: A Mendelian Randomization Analysis. JAMA Cardiol. 2018;3(7):619–627. doi:10.1001/jamacardio.2018.1470
    7. [7]Cesaro A, Calabrò P, Jamialahmadi O, et al. Lipoprotein(a) and long-term revascularization burden: a recurrent-event analysis of the UK Biobank. Eur J Prev Cardiol. 2026. doi:10.1093/eurjpc/zwag499
    8. [8]Song S, Hirayama S, Inaba H, et al. Physiological within-day Variation in Lipoprotein(a). J Atheroscler Thromb. 2026. doi:10.5551/jat.66290
    9. [9]European Society of Cardiology, Council for Cardiology Practice. What is new in the 2025 focused update of the 2019 ESC/EAS guidelines for the management of dyslipidaemias
    10. [10]ClinicalTrials.gov. Olpasiran: OCEAN(a) Outcomes, NCT05581303
    11. [11]ClinicalTrials.gov. Lepodisiran: ACCLAIM-Lp(a), NCT06292013
    12. [12]ClinicalTrials.gov. Muvalaplin cardiovascular outcomes trial, NCT07157774
    Disclaimer: This article provides educational, evidence-based information and is not medical advice. It does not establish a physician–patient relationship. Do not start, stop or change any medication, including lipid-lowering therapy, on the basis of this article. Discuss your own Lp(a) result and cardiovascular risk with your physician.