Association of Lipoprotein(a) With Major Adverse Cardiovascular Events Across hs-CRP

Protocol registration

This systematic review and meta-analysis adheres to the guidelines outlined in The Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement.¹⁵ The Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist is presented in Supplemental Table 1 The comprehensive protocol for this study was prospectively registered on PROSPERO (CRD4202345109), ensuring transparency and adherence to planned methodological processes before data collection and analysis began. The meta-analysis was conducted using data from previously published studies, so ethical approval and/or informed consent from patients was not required.

Eligibility criteria

Our study included original prospective and retrospective cohort studies as well as randomized clinical trials that investigated the association of lipoprotein(a) [Lp(a)] with incident MACE across varying levels of hs-CRP. Eligible studies needed to focus on adult populations (aged 18 years and above) and were required to report both Lp(a) and hs-CRP levels as concurrent exposures. Only studies that provided HRs for the association of Lp(a) with MACE, categorized by levels of hs-CRP, were considered. We excluded case reports, case-control, and cross-sectional studies to avoid design-related biases. Furthermore, studies that assessed Lp(a) and hs-CRP independently or only reported on one of these biomarkers were also excluded to ensure a consistent analytical focus on the association of Lp(a) with cardiovascular outcomes across hs-CRP levels.

Search strategy

We conducted a comprehensive electronic search of the following databases from their inception to February 28, 2024: MEDLINE (PubMed), Embase (Ovid), Cochrane CENTRAL (Wiley), and Web of Science (Clarivate). Our search strategy, detailed in the Supplemental Appendix, utilized a combination of controlled vocabulary and keywords targeting atherosclerotic cardiovascular disease, Lp(a), and hs-CRP. The complete search strategies for each database are presented in Supplemental Table 2. We imposed no restrictions on follow-up time, language, or publication format. To broaden our search and potentially reduce publication bias, we included conference abstracts and preprint articles. Additionally, we conducted a hand search of references from included articles to identify further studies. The search results and the study selection process are depicted in Figure 1. Where necessary, we contacted the authors of the included studies to request additional relevant data to ensure consistency in data analysis.

Data collection

The search results were managed using the Covidence systematic review platform, where screening was performed by 3 independent reviewers (P.A., C.Y.P., and G.K.). Any disagreements regarding study inclusion were resolved by a final reviewer (A.M.). The software automatically removed duplicate entries. Following initial screening, data extraction was independently conducted by 2 reviewers (P.A. and K.H.).

Data items

Data extracted included author names, publication year, study design, country of origin, follow-up duration, assays used for Lp(a) and hs-CRP, age, gender, and race/ethnicity. The primary endpoint was the time to occurrence of MACE, expressed as HRs. The definition of MACE varied across studies (Table 1). For studies presenting multiple statistical models, data from the most comprehensive model were extracted. We also noted whether the study was from a primary or secondary prevention cohort. Studies with both primary and secondary prevention cohorts were treated as separate entries to allow stratification in the analysis. The included studies utilized different assays for Lp(a) quantification and varied in their definitions of high and low Lp(a) levels, as detailed in Table 1. Data will be made available by the corresponding author upon reasonable request.

Quality assessment

The quality of each study was assessed using the Newcastle-Ottawa Scale for nonrandomized studies in meta-analyses.¹⁹ This scale includes 8 items divided among a selection of study groups, comparability of groups, and outcome assessment. Studies scoring ≥7 were considered of good quality, those scoring 2-6 were deemed fair, and scores ≤1 were rated as poor. Quality assessments were conducted by 2 investigators (PA and KH), with any discrepancies resolved through discussion to reach consensus.

Statistical analysis

The primary outcome, the association of elevated Lp(a) levels with MACE, was quantified using HRs derived from Cox proportional hazards regression models. These results were stratified by levels of hs-CRP, categorized as low (<2 mg/L) and high (≥2 mg/L). For studies that did not report HRs corresponding to the dichotomized Lp(a) level above or below 125 nmol/L, we converted the reported estimates to match this threshold. Details of the conversion method are provided in the supplementary material.

The meta-analysis employed a random-effects model using the Mantel-Haenszel method to accommodate the large study sizes and the relatively common nature of the outcome. Graphical representation of the results was achieved through forest plots. To quantify heterogeneity among the included studies, we used Tau², estimated via the DerSimonian–Laird method, along with the Cochran Q test and the I² statistic. Furthermore, the analysis was stratified by cohort type, distinguishing between primary and secondary prevention cohorts to elucidate differences in the impact of Lp(a) levels on cardiovascular outcomes. Publication bias was initially assessed visually using a funnel plot, while a more rigorous evaluation was conducted using Eger’s test within a mixed-effects meta-regression model to explore funnel plot asymmetry. Sensitivity analyses included the recalculated pooled estimates following the exclusion of influential studies and randomized controlled trials to test the robustness of our findings. All statistical analyses were conducted using R software, version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria), utilizing the Meta package. Significance was established at a 2-tailed P value of <0.05.

Subgroup analysis

In the primary prevention population, the pooled HR for the low and high hs-CRP groups were 1.33 (95% CI: 1.06-1.66) and 1.43 (95% CI: 1.13-1.82), respectively, with a nonsignificant subgroup difference (P = 0.65). The corresponding HR for the secondary prevention population was 1.13 (95% CI: 1.00-1.27) and 1.31 (95% CI: 1.12-1.52), respectively, with a nonsignificant subgroup difference (P = 0.13) (Figure 3).

Sensitivity analysis

Due to the high heterogeneity observed (I² = 85.0% and 81.0% in the low and high hs-CRP groups, respectively), further sensitivity analysis excluding the influential Thomas et al.¹⁰ study recalibrated the pooled HRs to 1.16 (95% CI: 1.07-1.23; I² = 52.9%) for low hs-CRP and 1.28 (95% CI: 1.16-1.41; I² = 73.9%) for high hs-CRP group.

Some of the included studies were randomized controlled trials (RCTs)^8,12 and given the potential for trial interventions to influence MACE outcomes and biomarkers such as Lp(a) and hs-CRP, we performed a sensitivity analysis excluding all RCTs. The pooled HRs remained consistent with those from the primary analysis, which included all studies. The results of this sensitivity analysis are presented in Supplemental Figure 2.

Clinical implications

The identification of Lp(a) as a significant risk factor for MACE, independent of hs-CRP has important clinical implications for potential Lp(a) therapeutics, as these results suggest that they may be effective in all patients with elevated Lp(a), not only those with high Lp(a) and high hs-CRP. Elevated Lp(a) level (≥50 mg/dL or ≥125 nmol/L) and high hs-CRP (≥2 mg/L) are both recognized as risk-enhancing factors by the ACC/AHA primary prevention guidelines.²⁸ As Lp(a) becomes integrated into clinical practice for cardiovascular risk assessment, our findings support its independent association with MACE in primary prevention regardless of hs-CRP status.

Multiple emerging therapies can significantly reduce Lp(a) synthesis in the liver by employing distinct pathways including antisense oligonucleotide (pelacarsen),²⁹ small interfering RNA (olpasiran, lepodisiran, zerlasiran),30, 31, 32 and stereotactic inhibition of apolipoprotein B-100 and apolipoprotein(a) interaction (muvalaplin).³³ Among these therapies, pelacarsen and olpasiran are currently being studied in phase III randomized clinical trials and might offer a novel preventive and risk mitigation strategy in high-risk secondary prevention populations. Given the consistent association of elevated Lp(a) with MACE risk across hs-CRP levels, our study indicates that patients may benefit from these therapies regardless of their “residual inflammatory” risk.³⁴

Strengths and limitations

Our study benefits from the large sample size, a rigorous methodological approach, including a comprehensive search strategy and robust statistical analysis, lending confidence to our findings. However, our findings should be interpreted in the context of several limitations. First, Lp(a) was measured using different assays across studies and there was variability in the cutoff values used for defining elevated Lp(a), in addition, the definition of MACE varied across studies, these limit direct comparisons across studies. Second, in 4 cohorts^8,17,18 where HRs were reported for Lp(a) as a continuous variable, we converted them to dichotomized values, which could have biased our estimates towards the null. Third, we could not assess the impact of race/ethnicity on the differential association of Lp(a) risk across hs-CRP levels. It is well established that individuals of African ancestry typically exhibit higher levels of Lp(a) compared to other racial groups. However, the papers included in the meta-analysis lacked racial diversity, which may limit the generalizability of our findings. Fourth, while there was consistency in defining hs-CRP categories, some studies did not indicate if individuals with extremely high hs-CRP levels were excluded, which often indicates acute illnesses or an active rheumatologic condition. Also, the distribution of Lp(a) is rightly skewed, hence the majority of participants in the studies included in our meta-analysis have low Lp(a) levels. Consequently, the analysis of pooled HRs may obscure any potential effects in the clinically significant high Lp(a) subgroup. This limitation suggests that a potential modification of Lp(a) function by an inflammatory milieu might not be adequately captured in our pooled analysis. Furthermore, IL-1β, IL-6, and oxidized phospholipids, which are closely linked to Lp(a) in the inflammatory cascade, were not assessed in these studies. Fifth, the use of study-level data rather than individual-level data limits our ability to perform subgroup analysis and may introduce potential bias. Sixth, although RCTs provide superior internal validity, their interventions may confound the association between Lp(a) and MACE, potentially reducing the generalizability of our findings. To address this, we conducted a sensitivity analysis excluding RCT’s and the results remained consistent with the primary analysis. Seventh, the high heterogeneity of studies indicates substantial variability, which potentially reduces the reliability in the pooled estimates.

Finally, there is a potential for publication bias, as suggested by the asymmetry in the funnel plot for the high hs-CRP group. This may potentially lead to overestimation of the true effect size due to the disproportionate influence of published studies.