Hu Hong*
1China Universal Asset Management
*Corresponding author; email: eeeqxxtg{at}pku.edu.cn
medRxiv preprint DOI: https://doi.org/10.1101/2025.10.29.25338763
Posted: November 10, 2025, Version 2
Copyright: This pre-print is available under a Creative Commons License (Attribution 4.0 International), CC BY 4.0, as described at http://creativecommons.org/licenses/by/4.0/
Abstract
Although Lp(a) is an established risk factor for ASCVD, our analysis indicates its importance remains substantially underestimated. Reanalyzing cardiovascular outcomes trial (CVOT) data for the PCSK9 antibody alirocumab stratified by Lp(a) quartiles, we find that approximately 70% of the observed benefit is attributable to absolute reductions in Lp(a), rather than to lowering of LDL-C. This result aligns with a prior post hoc analysis of the PCSK9 antibody evolocumab, which attributed 57% of the benefit to Lp(a) reduction. These findings challenge the prevailing assumption that the benefits of PCSK9 therapy are mediated primarily through LDL-C lowering.
Based on the observed relationship, we project that late-stage, Lp(a)-targeted therapies could reduce the risk of major adverse cardiovascular events (MACE) by roughly 50∼60% in phase 3 trials, which would be unprecedented in prior CVOT trials. Our projection also suggests that setting a therapeutic goal of a 15∼20% reduction in MACE would confer benefit to roughly 40% of secondary-prevention patients with elevated Lp(a), well beyond the current eligibility range (13∼21%). Further health-economic modeling suggests these therapies would have would have favorable health-economic value, as numbers-needed-to-treat would substantially lower than PCSK9 agents.
Background
Lipoprotein(a) [Lp(a)] is an independent risk factor for atherosclerotic cardiovascular disease (ASCVD) and aortic valve stenosis (Kronenberg F, 2022). Unlike low-density lipoprotein cholesterol (LDL-C), circulating Lp(a) concentrations are predominantly genetically determined and largely unaffected by lifestyle modification. Although present at substantially lower molar levels than LDL-C, Lp(a) particles are enriched in oxidized phospholipids (OxPL) that are highly pro-inflammatory and recognizable by innate immune cells, thereby promoting vascular inflammation, atherogenesis, and thrombosis (Tsimikas S, 2005). In addition, the apolipoprotein(a) [Apo(a)] moiety of Lp(a) shares high homology with plasminogen (McLean JW, 1987), potentially impairing fibrinolysis and further increasing thrombotic risk. Multiple Mendelian randomization studies have demonstrated a robust, likely causal association between Lp(a) concentration and ASCVD risk (Langsted, 2019; Madsen, 2020; Patel, 2021; Welsh, 2022; Berman, 2024).
The extent to which pharmacologic lowering of Lp(a) translates into reductions in major adverse cardiovascular events (MACE) remains uncertain. Several late-stage investigational therapies that directly target Apo(a) and reduce Lp(a) have achieved substantial decreases (≈70–100%) in early-phase trials (Tsimikas, 2020; O’Donoghue M. L., 2022; Nicholls, 2025; Nissen S. E., 2025; Nissen S. E., 2024), but cardiovascular outcomes trial (CVOT) results have not yet been reported. Prior attempts to infer expected MACE reduction from Lp(a) lowering (Burgess, 2018; Parish, 2018; Lamina & Lp(a)-GWAS-Consortium, 2019; Madsen, 2020), have often relied on a strong assumption—by analogy to LDL-C—that short-term exposure reductions capture only a fraction of lifetime, cumulative risk. Whether this assumption holds for Lp(a) biology remain unclear.
Certain drug classes, including PCSK9 monoclonal antibodies and CETP inhibitors, exert broad lipid effects, lowering both LDL-C and Lp(a). The prevailing view has been that CVOT benefits for these agents are mediated chiefly by LDL-C lowering.
Notably, a post hoc analysis of phase 3 FOURIER trial of PCSK9 antibody evolocumab (O’Donoghue, 2019) reported that Lp(a) lowering accounted for 57% of the variability in clinical benefit. However, the observation that has received limited attention.
For another PCSK9 antibody, alirocumab, there is also a post hoc analysis concluded that only 13% of clinical benefit was attributable to Lp(a) reduction (Szarek, 2020). Upon reanalysis, we identify major methodological limitations in that work and find that the contribution of Lp(a) lowering to clinical benefit was substantially underestimated; our estimates suggest a contribution closer to ∼70%, broadly consistent with the evolocumab analysis. Collectively, these findings imply that the impact of Lp(a) on MACE risk has been materially underestimated in the current literature.
Data Re-analysis for Lp(a) Lowering Attribution in CVOT Benefit of PCSK antibody
The prior post hoc analysis of the alirocumab ODYSSEY OUTCOMES trial (Szarek, 2020) reported event rates and biomarker changes stratified by baseline Lp(a) quartiles for the active and placebo arms, including cardiovascular event incidence, baseline Lp(a) and its absolute reduction, and LDL-C reduction (Figure2, Table1).
In the lowest baseline Lp(a) quartile, the median change in Lp(a) was ∼0, whereas the median LDL-C reduction was the largest across strata. Yet the clinical benefit in this subgroup was minimal (hazard ratio [HR] ≈ 0.95; ∼5% risk reduction). Conversely, in the highest baseline Lp(a) quartile—the subgroup that also experienced the greatest absolute Lp(a) reduction—the LDL-C reduction was slightly smaller, but the observed benefit was much greater (HR ≈ 0.75; ∼25% risk reduction).
Figure 1.
There was a significant relationship with a 15% lower risk per 25 nmol/L reduction in Lp(a) after adjusting for the change in LDL-C (model accounts for 57% of total variability of clinical benefit). (O’Donoghue M. L.-B., 2019)
Figure 2.
Relative and absolute treatment effects on total cardiovascular events, overall and by quartile of baseline lipoprotein(a). (Szarek, 2020)
Table 1.Baseline Lp(a), median Lp(a) lowering and median LDL-C lowering in the alirocumab group, overall and by quartile of baseline lipoprotein(a). (Szarek, 2020)
We identified a key methodological misuse in the original analysis
The authors correlated on-treatment Lp(a) change with event counts within the treatment arm. In the unadjusted model, the correlation appeared weak; after adjusting for baseline Lp(a), a clear inverse association emerged, with little further change after additional adjustments. However, this approach primarily compares treated patients with similar baselines but different pharmacodynamic responses. What is needed for causal attribution is the relationship between absolute Lp(a) lowering and treatment effect versus placebo at a given baseline—i.e., how much of the randomized treatment benefit (active vs. placebo) is explained by Lp(a) reduction. As a result, the reported “effect of Lp(a) change” in the original work does not validly represent the proportion of cardiovascular benefit mediated by Lp(a) lowering under PCSK9 inhibition.
We therefore re-estimated the contribution of Lp(a) lowering using a simple mediation-style decomposition based on the quartile-level summaries, under three assumptions:
- In the lowest baseline Lp(a) quartile, the observed HR reflects benefit attributable entirely to LDL-C lowering (i.e., negligible contribution from Lp(a) change given median ΔLp(a) ≈ 0).
- Additivity on the log-hazard scale: The effects of LDL-C lowering and Lp(a) lowering on MACE risk are independent and additive in log space, such that for quartiles 2–4,
3. Approximate linearity for LDL-C: log(HRLDL–C) is proportional to the median LDL-C reduction across quartiles. (This assumption is not pivotal here because median LDL-C reductions are similar across quartiles.)
Operationally, we take the HR observed in quartile 1 as HRLDL–C. For quartiles 2–4, we compute the implied HRLp(a)as
and then derive the fraction of benefit attributable to Lp(a) lowering asThe result is shown in Table 2.
Table 2.Analysis result indicating the attribution of Lp(a) and LDL-C on HR, overall and by quartile of baseline lipoprotein(a).
Applying this decomposition to the quartile-stratified ODYSSEY OUTCOMES data yields a consistent pattern: the majority of treatment benefit in the higher Lp(a) strata is captured by the Lp(a)-lowering component, whereas the LDL-C–lowering component remains relatively constant across quartiles. Aggregating across quartiles, we estimate that approximately ∼70% of the overall benefit of alirocumab is attributable to Lp(a) lowering, with the remaining ∼30% attributable to LDL-C lowering. These results align closely with the independent post hoc findings for evolocumab highlighted a dominant role (57%) for Lp(a) reduction in explaining variability in clinical benefit.
Projected MACE Benefit of Lp(a)-Targeted Therapies
If our inference is correct, reductions in Lp(a) exert a decisive influence on residual risk in patients already receiving intensive statin therapy. As noted above, the empirical relationship between Lp(a) lowering and changes in log (HRLp(a))exhibits similar slopes in quartiles 3 and 4, with a smaller slope in quartile 2. This pattern implies diminishing marginal benefit when baseline Lp(a) is low, whereas an approximately linear relationship holds when baseline Lp(a) is sufficiently high. Conservatively, we assume linearity for patients whose on-treatment Lp(a) remains ≥30 mg/dL (≈72 nmol/L).
Using a working slope of Δlog (HRLp(a)) 0.025 per 2 ng/dL (≈5 nmol/L) decrement in Lp(a), the implied magnitudes of MACE risk reduction (1-HR) are shown in Table 3.
Table 3.Calculated magnitudes of MACE risk reduction related to Lp(a) Lowering.
Given that several Lp(a)-targeted agents in late-stage development produce on the order of 150–200 nmol/L reductions in Lp(a), these estimates imply approximately 50–60% relative reductions in MACE—an effect size unprecedented in prior CVOTs
For the Novartis/Ionis Apo(a) antisense oligonucleotide (ASO) program now in phase 3 (dose regimen: 80 mg Q4W (Cho, 2025)), the anticipated degree of Lp(a) lowering likely lies between the phase 2 results for 20 mg QW (−80%) and 60 mg Q4W (−72%) (Tsimikas, 2020). In the phase 2 population, the median baseline Lp(a) was 224.3 nmol/L. Assuming a similar baseline in phase 3 and an average 75% reduction, the absolute decrease would be ∼168 nmol/L. As above, we conservatively assume no incremental benefit below 72 nmol/L; thus, the *effective* decrease is ∼152 nmol/L, corresponding to an estimated ∼53% reduction in MACE risk.
For Amgen/Arrowhead Apo(a) siRNA olpasiran (also in phase 3), the median baseline Lp(a) was 260.3 nmol/L and on-treatment levels approached zero in its phase 2 (O’Donoghue M. L., 2022). Applying the same conservative threshold (no benefit <72 nmol/L), the effective decrease is ∼188 nmol/L, corresponding to an estimated ∼60% reduction in MACE risk.
Together, these projections suggest that dedicated Lp(a)-lowering agents could deliver clinically transformative—and potentially category-defining—reductions in ASCVD events, particularly among patients with substantially elevated baseline Lp(a).
Potential of Broadening Eligible Population
If a therapeutic target were set at a modest 15%–20% reduction in MACE, and we retain the conservative assumption that on-treatment Lp(a) values below 72 nmol/L (≈30 mg/dL) do not confer additional benefit, one can back-calculate the baseline Lp(a) required to achieve such effects. Under the slope estimated above, the necessary baseline concentration would be approximately 105–117 nmol/L (≈44–49 mg/dL). All currently investigational Lp(a)-targeted modalities are capable of lowering Lp(a) from these baselines to at or below the assumed risk floor.
Using representative distributions of Lp(a) in secondary-prevention ASCVD cohorts (Figure 3.1&3.2), these thresholds imply a substantial expansion of the treatable population. Specifically, the proportion of secondary-prevention patients who could plausibly benefit from Lp(a)-lowering therapy would increase from the oft-cited Figure 3.2 (O’Donoghue M. L.-B., 2019) ∼13%–21% to roughly ∼40%, based on our calculations. This shift suggests that even conservative efficacy goals could materially broaden clinical applicability and public-health impact.
Figure 3.1 & 3.2
Lp(a) Distribution in secondary-prevention ASCVD cohorts
In-Depth Discussion of Potential Objections
Q1. Why would LDL-C lowering contribute little to CVOT benefit in PCSK9 trials, given LDL-C lowering is significant?
A. We posit that the risk slope for LDL-C flattens at very low levels
Participants in PCSK9 outcomes trials were largely on moderate- or high-intensity statins at baseline, with median LDL-C ≈92 mg/dL (≈2.4 mmol/L)—already low. Even lower baselines were only seen with CETP inhibitors (e.g., anacetrapib ≈61 mg/dL; evacetrapib ≈81 mg/dL) (REVEAL Collaborative Group, 2017; Lincoff, 2017), which helps explain why their CVOT effects underperformed expectations. Moreover, CETP programs enrolled populations with lower baseline Lp(a) and achieved minimal Lp(a) reductions, further limiting the potential benefit attributable to Lp(a) lowering.
A useful comparator is the TNT trial (high-vs. standard-dose atorvastatin; baseline LDL-C ≈98 mg/dL; on-treatment ≈77 vs. 101 mg/dL) (LaRosa, 2005), which reported HR ≈0.78. We interpret this larger effect, relative to PCSK9 trials at comparable LDL-C ranges, as reflecting statins’ pleiotropic effects—including reduced vascular inflammation, improved endothelial function, and plaque stabilization—beyond LDL-C lowering per se. Analogously, in statin-treated patients whose LDL-C–mediated risk has already been substantially suppressed, non–LDL-C pathways (e.g., Lp(a)-related mechanisms) may dominate the residual benefit signal under PCSK9 inhibition.
Q2. Under your assumptions, could high–Lp(a) patients treated with Lp(a)-lowering agents end up with lower risk than low–Lp(a) patients? For example, using FOURIER strata, patients in the highest Lp(a) quartile (>165 nmol/L) had only ∼33% higher risk than the lowest quartile; doesn’t that cap the attainable risk reduction at ∼25% (0.33/1.33)?
A. This apparent paradox arises from collider bias
In the lowest Lp(a) quartile, the randomized HR was ≈0.95, indicating only ∼5% of risk was modifiable by PCSK9 (via either LDL-C or Lp(a))—consistent with very low baseline Lp(a) and already very low on-treatment LDL-C. Inevitably, a third category of risk (neither LDL-C– nor Lp(a)–related; e.g., inflammatory, thrombotic, microvascular, or other mechanisms) accounts for the remaining ≈95% of events in that quartile. This third-risk component is independent of LDL-C and Lp(a) in the general population but jointly influences overall MACE risk.
PCSK9 CVOTs enroll secondary-prevention patients on statin therapy—i.e., individuals pre-selected for high absolute risk. When multiple independent determinants (e.g., Lp(a) and the third-risk component) influence a downstream variable (risk) that is conditioned upon by study design (selecting high-risk patients), spurious correlations (often negative) can emerge within the selected sample—the hallmark of collider bias.
We infer that, due to this bias, Lp(a) and the third-risk component become negatively correlated in PCSK9 trial populations
in the lowest Lp(a) quartile, the third-risk component explains ∼95% of events, whereas in the highest Lp(a) quartile its share is much less than 0.95/1.33, and the Lp(a)-related contribution is much greater than 0.33/1.33. Consequently, the theoretical ceiling on treatment benefit from Lp(a) lowering is not constrained by 0.33/1.33 ≈ 25%. In high–Lp(a) strata where Lp(a) contributes a larger share of total risk, substantial (>25%) risk reductions remain biologically and statistically plausible when Lp(a) is effectively lowered.
Data Availability
All data produced in the present work are contained in the manuscript
Conclusion
Our reanalysis of PCSK9 outcomes data indicates that the clinical benefit traditionally attributed to LDL-C lowering has been materially overestimated in statin-treated, secondary-prevention populations, and that Lp(a) reduction accounts for the majority (∼70%) of observed benefit, consistent with independent findings for evolocumab (∼57%). Methodologically, reframing effect attribution on the log-hazard scale and comparing treatment vs. placebo within baseline strata resolves contradictions in prior post-hoc work and highlights a coherent, biologically plausible role for Lp(a) as a decisive driver of residual ASCVD risk. Leveraging the empirically derived slope between absolute Lp(a) lowering and log (HRLp(a)), we project that late-stage, Lp(a)-targeted agents can deliver ∼50–60% reductions in MACE among patients who remain ≥72 nmol/L on treatment— magnitudes not seen in previous CVOTs. Even conservative targets (15–20% MACE reduction) would broaden eligibility to ∼40% of secondary-prevention patients with elevated Lp(a), substantially expanding the potential population beyond current expectations. Taken together, these results argue for re-prioritizing Lp(a) in risk assessment and trial design,and they support the clinical and health-economic rationale for dedicated Lp(a)–lowering therapies.
Footnotes
- an unclear statement in the abstract section is revised
References
Welsh, P. W.-M. (2022). Lipoprotein(a) and cardiovascular disease: prediction, attributable risk fraction, and estimating benefits from novel interventions. uropean Journal of Preventive Cardiology, 28(18), 1991–2000. Retrieved from doi:10.1093/eurjpc/zwaa063
Berman, A. N. (2024). Lipoprotein(a) and major adverse cardiovascular events in patients with or without baseline atherosclerotic cardiovascular disease. Journal of the American College of Cardiology, 83(9), 873–886. Retrieved from doi:10.1016/j.jacc.2023.12.031
Burgess, S. F. (2018). Association of LPA variants with risk of coronary disease and the implications for lipoprotein(a)-lowering therapies: A Mendelian randomization analysis. JAMA Cardiology, 3(7), 619–627. Retrieved from doi:10.1001/jamacardio.2018.1470
Cho, L. N. (2025). Design and rationale of Lp(a)HORIZON trial: Assessing the effect of lipoprotein(a) lowering with pelacarsen on major cardiovascular events in patients with CVD and elevated Lp(a). American Heart Journal, 287, 1–9. Retrieved from doi:10.1016/j.ahj.2025.03.019
Kronenberg F. M. S. (2022). Lipoprotein(a) in atherosclerotic cardiovascular disease and aortic valve stenosis: a European Atherosclerosis Society consensus statement. European Heart Journal, 43(39), 3925–3946. Retrieved from doi:10.1093/eurheartj/ehac361
Lamina, C. &., & Lp(a)-GWAS-Consortium, f. t. (2019). Estimation of the required lipoprotein(a)-lowering therapeutic effect size for reduction in coronary heart disease outcomes: A Mendelian randomization analysis. JAMA Cardiology, 4(6), 575–579. Retrieved from doi:10.1001/jamacardio.2019.1041
Langsted, A. K. (2019). High lipoprotein(a) and high risk of mortality. European Heart Journal, 40(33), 2760–2770. Retrieved from doi:10.1093/eurheartj/ehy902
LaRosa, J. C. (2005). Intensive Lipid Lowering with Atorvastatin in Patients with Stable Coronary Disease. New England Journal of Medicine, 352(14), 1425–1435. Retrieved from doi:10.1056/NEJMoa050461
Lincoff, A. M. (2017). Evacetrapib and Cardiovascular Outcomes in High-Risk Vascular Disease. New England Journal of Medicine, 376, 1933–1942. Retrieved from doi:10.1056/NEJMoa1609581
Madsen, C. M. (2020). Lipoprotein(a)-Lowering by 50 mg/dL (105 nmol/L) may be needed to reduce cardiovascular disease 20% in secondary prevention: a population-based stud. Arteriosclerosis, Thrombosis, and Vascular Biology, 40(1), 255–266. Retrieved from doi:10.1161/ATVBAHA.119.312951
McLean JW, T. J. (1987). cDNA sequence of human apolipoprotein(a) is homologous to plasminogen. Nature, 330, 132–137. doi:10.1038/330132a0
Nicholls, S. J. (2025). Oral Muvalaplin for Lowering of Lipoprotein(a). JAMA, 333(3), 222–231. Retrieved from doi:10.1001/jama.2024.24017
Nissen, S. E. (2022). Lipoprotein(a) levels in a global population with established atherosclerotic cardiovascular disease. Open Heart, 9(2), e002060. Retrieved from doi:10.1136/openhrt-2022-002060
Nissen, S. E. (2024). Zerlasiran A Small-Interfering RNA Targeting Lipoprotein(a). JAMA, 332(23), 1992–2002. Retrieved from doi:10.1136/openhrt-2022-002060
Nissen, S. E. (2025). Lepodisiran—A long-duration siRNA targeting Lp(a). New England Journal of Medicine, 392, 1673–1683. Retrieved from doi:10.1056/NEJMoa2415818
O’Donoghue, M. L. (2022). Small Interfering RNA to Reduce Lipoprotein(a) in Cardiovascular Disease. New England Journal of Medicine, 387, 1855–1864. Retrieved from doi:10.1056/NEJMoa2211023
O’Donoghue L., Fazio, S., Giugliano, R. P., Stroes, E. S. G., Kanevsky, E., Gouni-Berthold, I., … Sabatine, M. S.M. (2019). Lipoprotein(a), PCSK9 inhibition, and cardiovascular risk: Insights from the FOURIER trial. Circulation, 139(12), 1483–1492.
Parish, S. H.-M. (2018). Impact of apolipoprotein(a) isoform size on lipoprotein(a) lowering in the HPS2-THRIVE study. 11(2), e001696. Retrieved from doi:10.1161/CIRCGEN.117.001696
Patel, A. P. (2021). Lp(a) (lipoprotein[a]) concentrations and incident atherosclerotic cardiovascular disease: New insights from a large national biobank. Arteriosclerosis, Thrombosis, and Vascular Biology, 41(1), 465–474. Retrieved from https://www.ahajournals.org/doi/10.1161/atvbaha.120.315291
REVEAL Collaborative Group, B. L. (2017). Effects of Anacetrapib in Patients with Atherosclerotic Vascular Disease. New England Journal of Medicine, 377, 1217–1227. Retrieved from doi:10.1056/NEJMoa1706444
SzarekBittner, V. A., Aylward, P., Baccara-Dinet, M., Bhatt, D. L., Diaz, R., … Schwartz, G. GM., (2020). Lipoprotein(a) lowering by alirocumab reduces the total burden of cardiovascular events independent of low-density lipoprotein cholesterol lowering: ODYSSEY OUTCOMES trial. European Heart Journal, 41(44), 4245–4255.
Tsimikas S. B. E. (2005). Oxidized phospholipids, Lp(a) lipoprotein, and coronary artery disease. New England Journal of Medicine, 353(1), 46–57. Retrieved from doi:10.1056/NEJMoa043175
Tsimikas, S. K.-P.-B.-C.-T. (2020). Lipoprotein(a) reduction in persons with cardiovascular disease. New England Journal of Medicine, 382(3), 244–255. Retrieved from doi:10.1056/NEJMoa1905239