Lipoprotein (a) level as a risk factor for stroke and its subtype: A systematic review and meta-analysis

The role of lipoprotein-A [Lp (a)] as a risk factor for stroke is less well documented than for coronary heart disease. Hence, we conducted a systematic review and meta-analysis for the published observational studies in order to investigate the association of Lp (a) levels with the risk of stroke and its subtypes. In our meta-analysis, 41 studies involving 7874 ischemic stroke (IS) patients and 32,138 controls; 13 studies for the IS subtypes based on TOAST classification and 7 studies with 871 Intracerebral hemorrhage (ICH) cases and 2865 control subjects were included. A significant association between increased levels of Lp (a) and risk of IS as compared to control subjects was observed (standardized mean difference (SMD) 0.76; 95% confidence interval (CIs) 0.53–0.99). Lp (a) levels were also found to be significantly associated with the risk of large artery atherosclerosis (LAA) subtype of IS (SMD 0.68; 95% CI 0.01–1.34) as well as significantly associated with the risk of ICH (SMD 0.65; 95% CI 0.13–1.17) as compared to controls. Increased Lp (a) levels could be considered as a predictive marker for identifying individuals who are at risk of developing IS, LAA and ICH.


Risk of bias in individual studies. The risk of bias was assessed by Newcastle-Ottawa Scale (NOS) for
quality assessment of all the included studies in the meta-analysis 30 . The assessment criteria involving NOS uses three broad criteria viz. selection, comparability and exposure. Selection criteria defines and analyses the cases and control subjects included in the study, comparability defines the matching or comparison of cases and control subjects for better empirical investigation and exposure determines whether the study was conducted in a blinded or unbiased manner along with the response of the subjects. Publication bias was assessed using Begg's and Egger's funnel plot analysis 31,32 . Data extraction. All relevant studies were analysed separately by two reviewers (PK and PS) based on the inclusion criteria listed above. The analysis was done first at the title and abstract level and then at the full-text level. Any disagreement was resolved by discussion with a third reviewer. Following data were extracted from the studies which included: First Author's Name, Published Year, Ethnicity, Country, Study Design, Number of Cases and Controls, Mean Age, mean and standard deviation values of biochemical parameters including Lp (a), methods of Lp (a) assay and follow up duration. Data was extracted independently by two authors (PK and PS) using a standardized extraction table. Lp (a) concentrations were reported with different units in the included studies and were converted to similar units for analysis purpose using online unit conversion tools (http:// units lab. com/ node/ 85).

Statistical analysis.
A random or fixed effect model was used to calculate the pooled Standardized Mean Difference (SMD) or Odds Ratio (OR) with 95% confidence interval (CI). Heterogeneity was calculated with the I 2 statistic and was adjusted by subgroup analysis followed by meta-regression using the quality score of the included studies. The heterogeneity was considered as significant in case of I 2 more than 50% for which randomeffects model was applied, on the other hand, if I 2 was less than 50%, then fixed-effect model was applied. A sensitivity analysis was performed by sequentially omitting a single study in each turn, to validate the pooled observed effect. Tests were considered statistically significant at a p-value less than 0.05. Data were analyzed using STATA, version 13.0 (Stata Statistical Software, Release 13; StataCorp LP, College Station, TX).

Statement of ethics.
Ethical approval was not required for this manuscript as it was a systematic review and meta-analysis done by using the existing published data and the research was not directly conducted in any human subjects. Figure 1 represents the PRISMA flow diagram listing the detailed reasons for exclusion and inclusion of studies in our systematic review and meta-analysis. Initially, a total of 322 studies were identified after searching in six different databases. PRISMA checklist has been provided in the supplementary table (Table S1). After removing duplicate articles, 98 articles were found and on further exclusion, a total of 56 full text articles were reviewed for eligibility and finally 45 studies were included in the systematic review and meta-analysis. The baseline characteristics of all the included 45 studies (41 studies for investigating the association of LP (a) with the risk of IS; 13 studies for IS subtypes and 7 studies for ICH) are given in Tables 1, 2 20,21, . Thirty-five studies were case-control, one was nested case-control and five were population-based cohort studies. The publication years of the studies included in our meta-analysis ranged from 1985 to 2020. The studies were divided into two groups of populations based on ethnicity; 25 studies were conducted in Caucasian population and 16 studies were in Asian population. The sample size for IS cases ranged from 31 to 1953. Twenty-two studies used hospital-based (HB) source of control and nineteen studies used populationbased (PB) source of control. The quality score was high in nine articles, medium in 22 articles and low in 10 articles. The detailed quality scores (NOS scores) of the included studies ranging from low to high are represented in supplementary tables (Table S2). ELISA technique was found to be most common methods for LPA assay and reported in 18 studies, follow-up duration and LPA timepoints were reported in limited studies. No information was available for the stroke patients undergoing any LPA treatments in all the included studies in our meta-analysis. Only  Seven studies were conducted in Caucasian population and six studies were in Asian population. The sample size for IS subtypes ranged from 09 to 1809. Ten studies were case-control, one was cross-sectional and two were population-based cohort studies. The quality score was medium in nine studies, high in two studies and low in two studies as represented in supplementary tables (Table S3).

Characteristics of included studies for Intracerebral hemorrhage.
Only seven studies involving 871 Intracerebral hemorrhage (ICH) cases and 2865 control subjects were identified for the association of Lp (a) levels with the risk of ICH as compared to control subjects 38,45,58,63,64,67,69 . The publication years ranged from 1992 to 2020. Out of seven studies, three studies were conducted in Caucasian population and four studies were in Asian population. The sample size for ICH ranged from 06 to 499. Six studies were case-control and one was a nested case-control study. The quality score was medium in three studies, high in two studies and low in two studies as represented in supplementary tables (Table S4).

Association of Lp (a) levels with risk of ischemic stroke. A significant association between increased
levels of Lp (a) and risk of IS as compared to control subjects was observed (SMD 0.76; 95% CI 0.53-0.99). A subgroup analysis based on ethnicity also showed a significant association between increased levels of Lp (a)  (Fig. 2). Based on study design, further subgroup analysis also showed a significant association between increased levels of Lp (a) and the risk of IS as compared to control groups in 35 case-control studies (SMD 0.64; 95% CI 0.48-0.80) and one nested case-control study (SMD 3.30; 95% CI 3.00-3.60) (Table 4). However, we did not observe any significant association between Lp (a) levels and risk of IS in the subgroup consisting of five prospective cohort studies (SMD 0.96; 95% CI − 0.01 to 1.93).
On the basis of NOS quality grading, we also observed a significant association between increased levels of Lp (a) and the risk of IS as compared to control groups in high-quality studies (SMD 0.99; 95% CI 0.53-1.44); medium (SMD: 0.66; 95% CI 0.32-1.00) and low-quality studies (SMD 0.77; 95% CI 0.33-1.21) ( Table 4).
Based on extraction of reported calculated OR values directly from eleven studies, we observed an overall a significant association of increased levels of Lp (a) with the risk of IS as compared to control groups (OR 1.57; 95% CI 1.25-1.89). Based on ethnicity, a significant association of increased levels of Lp (a) with the risk of IS as compared to control groups was observed for Asian population (OR 1.97; 95% CI 1.67-2.26) but not for Caucasian population (OR 1.10; 95% CI 0.89-1.29). A significant association of increased levels of Lp (a) with the risk of IS as compared to control groups (OR 1.54; 95% CI 1.21-1.86) was observed for case-control studies but not for prospective cohort studies (OR 2.23; 95% CI 0.92-3.54). High-quality studies confirmed a significant association of increased levels of Lp (a) with the risk of IS as compared to control groups (OR 1.98; 95% CI 1.69-2.26) but not the medium (OR 1.01; 95% CI 0.95-1.08) and low-quality studies (OR 2.40; 95% CI 0.40-4.40) (Fig. 3).

Association of Lp (a) levels with the risk of large artery atherosclerosis. The association of Lp (a)
levels with the risk of LAA stroke subtype vs. control was investigated in eight studies and our findings reveal an overall significant association of increased levels of Lp (a) with the risk of LAA as compared to control groups (SMD 0.32; 95% CI 0.00-0.64). Based on ethnicity, no significant association for the increased levels of Lp (a) with the risk of LAA in Asian population (SMD 0.08; 95% CI − 0.22 to 0.39) as well as Caucasian population (SMD 0.45; 95% CI − 0.10 to 0.64) was observed. All included eight studies were of case-control design which showed a significant association and based on NOS quality grading, no association was observed for medium quality studies (SMD 0.99; 95% CI − 0.23 to 0.74). Single studies were found based on high and low NOS quality and reported significant association for the increased levels of Lp (a) with the risk of LAA as compared to control groups ( Table 4).

Association of Lp (a) levels with the risk of small vessel disease. No significant association for
the increased Lp (a) levels with the risk of SVD subtype vs. control (SMD − 0.06; 95% CI − 0.46 to 0.34) was observed. Moreover, based on ethnicity, study design and NOS quality grading, similar non-significant association was observed except for high quality study which included only a single study for SVD vs. control subjects (SMD 0.14; 95% CI 0.04 to 0.24) ( Table 4).  www.nature.com/scientificreports/ was observed. Subgroup analysis based on ethnicity, study design and NOS quality grading also revealed a nonsignificant association for the increased Lp (a) levels with the risk of CE subtype vs. control subjects ( Table 4).

Association of Lp (a) levels with the risk of intracerebral hemorrhage.
Overall, a significant association for the association of increased Lp (a) levels with the risk of ICH as compared to control subjects (SMD 0.65; 95% CI 0.13-1.17) was observed (Fig. 4). Non-significant association for the increased Lp (a) levels with the risk of ICH vs. control was observed based on ethnicity and study design. However, after conducting a subgroup analysis based on NOS quality grading, a significant association for the increased Lp (a) levels with the risk of ICH as compared to control subjects (SMD 0.27; 95% CI 0.03-0.52) was only observed in medium quality studies (Table 4).
Publication bias analysis. The shape of the funnel plots indicated the presence of publication bias while analysing the Lp (a) levels with the overall risk of IS. After conducting the Begg's test, we observed that a significant publication bias was present in the included studies for Lp (a) levels with the overall risk of IS (p-value: 0.002) (Figure S-1a). The shape of other funnel plots for the included studies of IS subtypes and ICH in the metaanalysis did not indicate the presence of any publication bias ( Figure S-1b-e).

Meta-regression analysis.
To further explore the amount of heterogeneity present in our meta-analysis, we conducted meta-regression analysis based on NOS quality score, study design and ethnicity for determin-

Discussion
The present systematic review and meta-analysis of 45 studies analysed the potential role of Lp (a) levels and its association with the risk of IS, IS subtypes based on TOAST classification and ICH compared to control subjects.
To the best of our knowledge, this is the most robust and the largest meta-analysis conducted till date which comprised of both Asian and Caucasian ethnicities to ascertain the risk of IS, IS subtypes and HS with increased levels of Lp (a). Our meta-analysis revealed that increased levels of Lp (a) are significantly associated with the risk of IS in Asian as well as Caucasian population. Also a significant association for the increased level of Lp   www.nature.com/scientificreports/ (a) with the risk of LAA and ICH as compared to control subjects was observed. Lp (a) levels were found to be greater in Asian population as compared to Caucasian population confirming greater risk of stroke as compared to control group. However, on the basis of subtypes, no significant association was observed in Asian as well as Caucasian population separately for increased levels of Lp (a) in LAA, SVD and CE subtypes when compared to control groups. Two previous meta-analyses had also established the association of elevated Lp (a) levels and risk of stroke by pooling data from case-control, prospective cohort and nested case-control studies 27,28 . A total of thirty-one studies were included in the meta-analysis by Smolders et al. 2007 27 in which the association of overall stroke was found to be statistically significant with Lp (a) increment levels (SMD 0.39; 95% CI 0.23-0.54) which is in agreement with the findings of our meta-analysis. Another meta-analysis by Nave et al. 2015 28 observed a significant association between Lp (a) and IS with OR of 1.41 (95% CI 1.26-1.57) for case-control studies (n = 11) and the pooled estimated risk ratio was 1.29 (95% CI 1.06-1.58) for prospective studies (n = 9).
Despite the fact that this systematic review and meta-analysis was undertaken comprehensively with defined inclusion and exclusion criteria along with uniform measured-effect across all analyses, the study has some following limitations: (1)included studies had a wide range of incorporated variables like age, ethnicity, sample size, study-design; (2) mean and standard deviations of Lp (a) levels obtained from few studies were converted from either the actual reported median values or the inter-quartile range values, inferring that they did not actually represent the original mean and standard deviation values of Lp (a) levels. (c) Subgroup analysis based on cutoff values of Lp (a) levels was not performed owing to non-availability of cut-off values of Lp (a) in majority of the included articles as represented in Tables 1, 2 and 3. (d) A random-effects model was used to account for the significant heterogeneity arising out of the studies. Therefore, large scale population based observational studies with defined clinical characteristics of the stroke-affected subjects and healthy controls are needed to ascertain the association of Lp (a) with either IS, subtypes of IS or ICH in a statistically significant manner.

Conclusion
Increased Lp (a) levels could be considered as a predictive marker for identifying individuals who are at risk of developing IS, LAA and ICH.