The prespecified protocol for this systematic review and meta-analysis has been registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration ID: CRD42024603931) and adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [7] and the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines [8].

A systematic literature search was conducted to identify eligible studies (RCTs, observational cohort studies, or individual patient data meta-analysis of studies) including adult patients with ischemic stroke or transient ischemic attack for whom Lp(a) levels were measured at baseline. Patients with elevated levels of Lp(a) were compared to those with low levels of Lp(a). Reporting of any of the outcomes of interest as outlined below was required for studies to be considered eligible for inclusion. The literature search was performed independently by three reviewers (LP, KM, MIS). The electronic databases MEDLINE and Scopus were searched, using search strings that included the terms “lipoprotein(a)” and “stroke recurrence.” Studies published in English were considered eligible. Our search spanned from the inception of each database to October 20, 2024.

Studies without a control arm, case series, and case reports were excluded. Commentaries, editorials, and narrative reviews were also discarded. In case of studies with duplicate data, the most recent study, further accounting for the number of included patients and outcome reporting, was retained, while the rest were excluded. All retrieved studies were independently assessed by three reviewers (LP, KM, MIS), resolving any disagreements through discussion with the corresponding author (GT).

Two reviewers (LP and KM) independently assessed the quality control and bias assessment among eligible studies employing the revised Cochrane Risk-of-Bias tool (RoB 2) for RCTs [9] and the Risk Of Bias In Non-randomized Studies of Exposures (ROBINS-E) tool for observational studies [10]. Any disagreements were settled by consensus after discussion with the corresponding author (GT). Structured forms, which included trial names, year of publication, period of enrollment, patient sample, patients’ characteristics, and outcomes of interest, were used for data extraction.

The primary outcome of interest was stroke recurrence at follow-up, including ischemic stroke, hemorrhagic stroke, any stroke, or transient ischemic attack, as defined by each study. Poor functional outcome, as defined by each study based on dichotomized modified Rankin Scale (mRS) scores (0-2 vs. 3-6 & 0-1 vs. 2-6), all-cause mortality, and recurrent vascular events at follow-up were evaluated as secondary outcomes of interest.

For the pairwise meta-analysis, we calculated the corresponding odds ratios (OR) with 95% confidence intervals (CI) for each dichotomous outcome of interest, for the comparison of outcome events among patients with high Lp(a) levels versus those with low Lp(a) levels. Sensitivity analyses were conducted to address variations in outcome definitions by excluding studies with differing definitions. Furthermore, a subgroup analysis was performed for the primary outcome after stratification according to trial setting and population (i.e., Chinese vs. non-Chinese). Baseline characteristics were described using pooled proportions (after the implementation of the variance-stabilizing double arcsine transformation) for categorical variables and mean values for continuous variables. For studies reporting continuous outcomes in median values and corresponding interquartile ranges, the sample mean and standard deviation were estimated using the quantile estimation method [11]. The random-effects model (DerSimonian and Laird 1986) was used to calculate the pooled estimates [12]. The threshold for statistical significance for the above analyses was set at a two-sided P-value of <0.05. Heterogeneity was assessed with the I² and Cochran Q statistics. For the qualitative interpretation of heterogeneity, I² values <25%, between 25%-50%, and >50% were considered to represent low, moderate, and significant heterogeneity, respectively. The significance level for the Q statistic was set at 0.1. The above statistical analyses were performed using the R software version 3.5.0 (package: meta; R Foundation for Statistical Computing, Vienna, Austria) [13].

All data generated or analyzed during this study are included in this article and its supplementary information files.

The flow diagram for the selection and inclusion of studies in this systematic review is presented in Figure 1. After excluding duplicates, the systematic literature database search yielded a total of 1,028 records. Following the initial screening process, the full texts of 36 records were retrieved. After reading the full-text articles, 24 records were further excluded. Finally, 12 eligible studies (one post hoc analysis of an RCT and 11 observational studies) (Table 1) [14-25] were included in the systematic review and meta-analysis, comprising a total of 17,903 ischemic stroke patients for whom Lp(a) was measured at baseline. Included patients had a mean age of 63 years (Supplementary Figure 1), and 38% were of female sex (Supplementary Figure 2). The majority of the cases presented moderate stroke (mean National Institutes of Health Stroke Scale [NIHSS] score: 7) (Supplementary Figure 3), of which 25% received intravenous thrombolysis and 13% endovascular treatment (Supplementary Figure 4). Twenty-nine percent of the cases were attributed to large-artery atherosclerosis (Supplementary Figure 5), and the mean baseline Lp(a) levels were 34.4 mg/dL (Supplementary Figure 6).

There was no significant bias detected in any of the quality domains assessed for the study of Chemello et al.,16 which was the only post hoc analysis of an RCT included in this systematic review (Supplementary Figure 7). The majority of the observational studies presented a moderate risk of bias (Supplementary Figure 8), mainly due to unblinded outcome measurements and losses to follow-up, with a lesser contribution from selection bias (inclusion of ischemic stroke patients with diabetes or under statin treatment) and reporting bias (incomplete reporting of outcome data). Only two of the observational studies were graded as “low risk of bias” in all domains.

Regarding the overall effect for the primary outcome of interest, increased levels of Lp(a) were associated with stroke recurrence (OR: 1.69; 95% CI: 1.09-2.63; P=0.020; 5 studies; I²=87%; P for Cochran Q<0.01) (Figure 2). A sensitivity analysis was further conducted to ensure homogeneity of stroke recurrence definition, by excluding the study of Chemello et al.16 that had a more inclusive definition for stroke recurrence taking into account not only ischemic stroke, hemorrhagic stroke, or transient ischemic attack, but also cases with carotid revascularization. In the sensitivity analysis, high Lp(a) levels were again significantly associated with stroke recurrence at follow-up (OR: 2.28; 95% CI: 1.23-4.21; P=0.008; 4 studies; I²=84%; P for Cochran Q<0.01) (Supplementary Figure 9). After stratification for trial setting and population, there were no significant subgroup differences for the primary outcome among studies that included Chinese versus non-Chinese patients (P for subgroup differences =0.82) (Supplementary Figure 10).

Regarding the secondary outcomes of interest, high Lp(a) levels were associated with increased odds of poor functional outcome at follow-up (OR: 2.09; 95% CI: 1.40-3.11; P<0.001; 5 studies; I²=77%; P for Cochran Q<0.01) (Figure 3). During sensitivity analysis, the study of Chakraborty et al.15 that defined poor functional outcome as mRS scores of 2-6 was excluded to ensure the homogeneity of definition (i.e., the rest of the studies used the dichotomized mRS score of 3-6 to define poor functional outcome). This sensitivity analysis confirmed similar results (OR: 1.96; 95% CI: 1.29-2.97; P=0.002; 4 studies; I²=79%; P for Cochran Q<0.01) (Supplementary Figure 11).

Furthermore, high Lp(a) levels were not associated with mortality (OR: 2.20; 95% CI: 0.89-5.43; P=0.088; 3 studies; I²=66%; P for Cochran Q=0.05) (Figure 4). Finally, there was no association between Lp(a) levels and recurrent vascular events (OR: 2.66; 95% CI: 0.95-7.44; P=0.063; 3 studies; I²=86%; P for Cochran Q<0.01) (Figure 5).