You are on page 1of 8

Atherosclerosis 242 (2015) 496e503

Contents lists available at ScienceDirect

journal homepage:

Lipoprotein (a) as a risk factor for ischemic stroke: A meta-analysis

Alexander H. Nave a, b, c, *, Kristin S. Lange a, Christopher O. Leonards a, Bob Siegerink a,
Wolfram Doehner a, c, Ulf Landmesser c, f, Elisabeth Steinhagen-Thiessen e,
Matthias Endres a, b, c, d, Martin Ebinger a, b, c
Center for Stroke Research Berlin (CSB), Charit tsmedizin Berlin, Germany
e e Universita
Klinik und Hochschulambulanz fr Neurologie, Charit e e Universitatsmedizin Berlin, Germany
German Center for Cardiovascular Research (DZHK), Berlin, Germany
German Center for Neurodegenerative Diseases (DZNE), Berlin, Germany
Department of Geriatrics, Charit
e e Universitatsmedizin Berlin, Germany
Department of Cardiology, Charite e Universitatsmedizin Berlin, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Objective: Lipoprotein (a) [Lp(a)] harbors atherogenic potential but its role as a risk factor for ischemic
Received 31 March 2015 stroke remains controversial. We conducted a meta-analysis to determine the relative strength of the
Received in revised form association between Lp(a) and ischemic stroke and identify potential subgroup-specic risk differences.
11 July 2015
Methods: A systematic search using the MeSH terms lipoproteins OR lipoprotein a AND stroke was
Accepted 13 August 2015
Available online 15 August 2015
performed in PubMed and ScienceDirect for caseecontrol studies from June 2006 and prospective cohort
studies from April 2009 until December 20th 2014. Data from eligible papers published before these
dates were reviewed and extracted from previous meta-analyses. Studies that assessed the relationship
Ischemic stroke
between Lp(a) levels and ischemic stroke and reported generic datadi.e. odds ratio [OR], hazard ratio, or
Lipoprotein (a) risk ratio [RR]dwere eligible for inclusion. Studies that not distinguish between ischemic and hemor-
Meta-analysis rhagic stroke and transient ischemic attack were excluded. Random effects meta-analyses with mixed-
Systematic review effects meta-regression were performed by pooling adjusted OR or RR.
Lipoproteins Results: A total of 20 articles comprising 90,904 subjects and 5029 stroke events were eligible for the
Risk factor meta-analysis. Comparing high with low Lp(a) levels, the pooled estimated OR was 1.41 (95% CI, 1.26
Biomarker e1.57) for caseecontrol studies (n 11) and the pooled estimated RR was 1.29 (95% CI, 1.06e1.58) for
prospective studies (n 9). Sex-specic differences in RR were inconsistent between caseecontrol and
prospective studies. Study populations with a mean age of 55 years had an increased RR compared to
older study populations. Reported Lp(a) contrast levels and ischemic stroke subtype signicantly
contributed to the heterogeneity observed in the analyses.
Conclusion: Elevated Lp(a) is an independent risk factor for ischemic stroke and may be especially
relevant for young stroke patients. Sex-specic risk differences remain conicting. Further studies in
these subgroups may be warranted.
2015 Elsevier Ireland Ltd. All rights reserved.

1. Introduction classied as cryptogenic. In young stroke patients (<55 years),

cryptogenic strokes account for almost 40% of all ischemic stroke
Over 80% of stroke events are ischemic and correlate with cases emphasizing the potential importance of other, possibly un-
traditional cardiovascular risk factorsdnamely, age, male sex, known risk factors especially but not exclusively in this stroke
presence of hypertension, diabetes mellitus, atrial brillation, and subtype [2,3].
dyslipidemia [1]. However, a large proportion of strokes remain Lipoprotein (a) [Lp(a)] is a low density lipoprotein (LDL)-like
particle containing apolipoprotein(a) [apo(a)] and apolipoprotein
B100 which are covalently linked by a disulde bridge bond [4,5].
Lp(a) has been attributed to atherogenic, thrombogenic, vascular
* Corresponding author. Universita tsmedizin Berlin, Hochschulambulanz fr
inammatory, and antibrinolytic potential, and shares structural
platz 1, 10117 Berlin, Germany.
Neurologie, Charite
E-mail address: (A.H. Nave).
homology with plasminogen [5,6]. Serum Lp(a) levels are
0021-9150/ 2015 Elsevier Ireland Ltd. All rights reserved.
A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503 497

genetically determined by an apo(a) size polymorphism encoded in cerebrovascular or carotid disease, or peripheral artery disease) or
the apolipoprotein(a) gene [5,7]. Lp(a)s role in the underlying with pre-existing hyperlipidemia; 6) reported at least one of the
pathophysiology of cardiovascular disease (CVD) has been highly following generic data for cases and controls with respective con-
investigated and studies have shown that Lp(a) is a causal risk dence intervals: odds ratio (OR), hazard ratio (HR), or risk ratio
factor for coronary heart disease (CHD) with improved risk pre- (RR). Studies analyzing distinct patient subgroups, e.g. diabetic
diction [8,9]. One observational study reported that cardiovascular patients or patients with renal insufciency were not included.
event rates decrease following effective Lp(a) lowering treatment in When several publications reported overlapping subjects, the most
patients with Lp(a)-hyperlipoproteinemia [10]. extensive, recent publication was included.
Recent reports have found a strong association between Lp(a)
and ischemic stroke in children [11,12], but others have failed to 2.3. Data extraction and synthesis
show equally homogeneous data in adult stroke patients, making
the current literature controversial [13]. Uncertainty regarding the The following data were extracted from relevant studies: rst
role of Lp(a) in adult ischemic stroke remains, because observed author's name, year of publication, study name; major inclusion
associations vary across populations and limited reports on age and and exclusion criteria for cases and controls or cohort; age and sex;
sex-specic Lp(a) associations exist [14,15]. Two previous meta- sample size of ischemic stroke and control patients; mean duration
analyses [16,17] analyzed 1903 ischemic stroke and >4609 all of follow-up (for prospective studies); sample storage time; tech-
stroke events. However, the latter study did not distinguish be- nique of Lp(a) measurement; Lp(a) levels for cases and controls, and
tween transient ischemic attack (TIA), ischemic or hemorrhagic Lp(a) baseline values for prospective studies, both converted into
stroke, and both reviews did not address subtypes of ischemic mg/dl where possible; the most adjusted risk for ischemic stroke as
stroke cases. Since the publication of these papers, evidence has OR, HR, or RR with 95% CI for high compared to low Lp(a); the
increased. Along with novel potential treatment options for Lp(a) dened Lp(a) contrast levels being taken as reported in the
lowering therapy, the high and sustaining interest underscores the contributing studies. We expected Lp(a) to be analyzed as cate-
relevance of an update on Lp(a) as a risk factor for ischemic stroke gories (e.g. RR for high vs. low with contrasts being obtained from
and qualies the attempt to identify subgroups at increased risk. dichotomous, tertile, quartile, or quintile categorization) as well as
Therefore, this meta-analysis summarizes the current evidence continuous (e.g. per unit or per SD increase of Lp(a)). If studies
regarding the association of Lp(a) and ischemic stroke and ad- reported multiple and equally adjusted effect sizes for different
dresses age and sex-specic risk differences of Lp(a). Lp(a) contrasts, the contrast with the highest risk ratio was used for
further analysis. Degrees of statistical adjustments were set as
2. Materials and methods follows: no adjustments; adjustment for age, sex,
smoking, alcohol abuse and conventional cardiovascular risk fac-
2.1. Data sources tors (hypertension, diabetes, dyslipidemia, atrial
brillation); additionally further adjustments (like lipid-
This review was performed in accordance with Preferred lowering therapy, socioeconomic status, family history of cardio-
Reporting Items for Systematic Reviews and Meta-Analyses vascular diseases, or others).
(PRISMA) and Meta-Analysis of Observational Studies in Epidemi-
ology (MOOSE) guidelines [18,19]. A systematic literature search of 2.4. Statistical analysis
fully accessible studies published in the English language was
performed by two independent readers (K.S.L. and A.H.N.) using Log RRs (prospective cohort studies), log ORs (caseecontrol
Medline via PubMed and ScienceDirect using a combination of the studies), 95% condence intervals (95% CI), and prediction intervals
MeSH terms lipoproteins OR lipoprotein a AND stroke. for studies that assessed the relationship between Lp(a) and
Regarding inclusion periods of the latest Lp(a) and ischemic ischemic stroke, dened by the respective study, were calculated
stroke meta-analyses [16,17], we chose the following specic in- using a random effects model a priori as we anticipated heteroge-
clusion periods according to study type: prospective cohort studies neity to be present. Random effects modeling is better suited for
from April 2009, caseecontrol studies from June 2006, and inter- higher degrees of heterogeneity, and prediction intervals help es-
ventional studies from 1960, to December 20th, 2014. If a study was timate a range for the true treatment effect in a heterogeneous
cited in one of the above mentioned meta-analyses but had been setting (like reference ranges for laboratory values) [20]. Hetero-
republished with a more complete dataset or longer follow-up geneity was tested using Cochrane's Q and the I2 statistics [21]. If
period, the latter was included. Reference lists of relevant publi- signicant heterogeneity was present for a given analysis (dened
cations (including review articles) as well as related citations pro- as a p < 0.05, or I2>50%), we generated funnel plots and performed
posed by the databases were hand searched and reviewed for mixed effects meta-regression (moderator variables e age, sample
relevance. Further, studies of both previously published meta- size, number of cases, Lp(a) levels, Lp(a) contrast levels, and stroke
analyses [16,17] were screened for eligibility according to the subtype) to assess potential sources of heterogeneity [22]. Potential
stringent inclusion and exclusion criteria of this meta-analysis. publication bias was additionally assessed using Egger's regression
Eligible studies were included in the nal meta-analyses. test. All meta-analysis procedures were performed using the soft-
ware package MIX 2.0 Pro [23] and the R version 3.1.1. Metafor
2.2. Study selection package [24]. Caseecontrol studies and prospective cohort studies
were analyzed separately; nested caseecontrol studies were
Studies were eligible for inclusion if they 1) examined the analyzed together with prospective cohort studies.
relationship between Lp(a) levels and ischemic stroke, diagnosed
by computed tomography or magnetic resonance imaging, and 3. Results
clearly distinguished ischemic from hemorrhagic stroke and tran-
sient ischemic attack (TIA); 2) reported data from a control group; 3.1. Search results
3) included subjects aged >18 years; 4) had 50 subjects per group;
5) included either subjects chosen from the general population, The study selection process is displayed in Fig. 1. The comput-
subjects with established cardiovascular disease (CHD, erized literature search in Medline via Pubmed and in ScienceDirect
498 A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503

Fig. 1. Meta-Analysis Flow Chart. Study selection process ow diagram, adapted from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) group
statement. MeSH terms: lipoproteins OR lipoprotein a AND stroke. Eligible studies were additionally extracted from previous meta-analyses. Abbreviations: IS, ischemic

yielded 531 and 515 citations. After reviewing titles, abstracts, and eligible for inclusion in this meta-analysis yielding a total of 20
hand searching related citations, 36 potentially relevant articles studies (11 caseecontrol-, 3 nested caseecontrol-, and 6 prospec-
were identied. Of those, after review of the full-text articles, 24 tive cohort studies) comprising 90,904 subjects and 5029 ischemic
studies were excluded for the following reasons: being duplicates stroke events. Study characteristics of included studies that origi-
(n 7) [25e31]; lack of ischemic stroke-specic data (n 8) nate from previous meta-analyses are listed in eTable 3e4 in the
[31e38]; lack of control group (n 3) [28,39,40]; insufcient sta- supplement.
tistic data (reporting OR without 95% CI or lacking OR) (n 2) Lp(a) detection methods, time between stroke onset and phle-
[29,41]; was included in the meta-analysis of 2006 (n 1) [30]; or botomy, sample storage time, and Lp(a) contrast values differed
partially overlapping study populations (n 3) [42e44]. between studies (for details please see eTable 5e6 in the
A total of 12 relevant studies published after June 2006 (for supplement).
caseecontrol studies) and after April 2009 (for prospective studies)
comprising 56,961 subjects and 3159 stroke events were identied 3.3. Caseecontrol studies
for qualitative and quantitative analysis: 6 caseecontrol studies
[27,45e49], 1 nested caseecontrol study [25], and 5 prospective The meta-analysis of 11 caseecontrol studies (2749 cases and
cohort studies [13,26,50e52]. The meta-analysis of these new 5328 controls) resulted in an increased risk of ischemic stroke with
studies is depicted in eFigure 1 in the supplement, and eTables 1e2 elevated Lp(a) with a pooled adjusted OR of 1.41 (95% CI, 1.26e1.57;
provide a detailed description of these studies. prediction interval, 1.07e1.86), as displayed in Fig. 2A. High het-
The aforementioned search criteria yielded only one interven- erogeneity was present in the analysis (Q 114.49, p < 0.01,
tional study [53] which, however, did not address stroke as an I2 89.52%). Funnel plots and Egger's regression intercept of 2.73
outcome parameter. We therefore decided to hand search for (p < 0.01) indicated the presence of potential publication bias in
relevant interventional studies addressing stroke and Lp(a) treat- this meta-analysis. Stroke subtypes signicantly contributed to
ment and dedicated a separate paragraph in the discussion. heterogeneity assessed by meta-regression; stroke due to large
artery atherosclerosis and stroke due to undetermined etiology
3.2. Merged meta-analyses reported signicantly higher ORs (QM 6.89; p 0.03;
beta 0.008; se 0.003). Four studies [47,48,54,56] dichotomized
After adopting the inclusion and exclusion criteria of this study Lp(a) levels at 30 mg/dl (clinical cut-off) or 14 mg/dl, respectively,
to all relevant studies of the previous meta-analyses [16,17], ve three studies [27,49,58] compared highest vs. lowest tertile or
caseecontrol studies [54e58], two nested caseecontrol studies quartile, respectively, and three studies [45,55,57] reported risk
[59,60], and one prospective cohort study [61] were additionally differences for continuous increase in Lp(a), see eTable 6 in the
A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503 499

Fig. 2. Pooled Analysis of all eligible Studies. Forest plots depicting study sample size with number of ischemic stroke patients (n) and total study participants (N), pooled adjusted
risk ratio with 95% condence interval (95% CI), and p value (P) for all included caseecontrol studies (A) and prospective studies (B) after eligible studies of previous meta-analyses
were added. Studies are arranged according to study sample size. Data on subgroups were reported separately. Box sizes represent the individual study weight. aLarge artery
atherosclerosis patients; bSmall vessel occlusion patients; cMale patients; dFemale patients.

supplement. Four studies reported sex-specic ORs [27,47,48,56]. data in women yielded a lower RR (pooled estimated OR 1.49;
95% CI 1.09e2.04; Q 3.44; p 0.39; I2 12.72%; see eFigure 3A
3.4. Nested caseecontrol and prospective cohort studies in the supplement).
Subgroup analysis of six prospective studies [25,50,51,59,60,62]
Nine prospective studies with a total of 80,527 participants and that reported sex-specic data resulted in a pooled adjusted RR of
2280 ischemic stroke cases were eligible for the nal analysis. The 1.49 (95% CI 1.10e2.03; Q 12.18; p 0.02; I2 67.15%) for
studies differed both in design and study population. Four studies women. There was moderate heterogeneity. For men, the pooled
reported a positive association between Lp(a) levels and ischemic adjusted RR was 1.05 (95% CI 0.80e1.38; Q 7.76; p 0.05;
stroke risk. The meta-analysis revealed a signicantly increased I2 61.34%). There was not strong evidence for heterogeneity. A
ischemic stroke risk with high Lp(a) levels and a pooled adjusted RR higher proportion of studies reporting linear risk calculations were
of 1.29 (95% CI, 1.06e1.58, prediction interval, 0.70e2.49), as indi- present among males, as depicted in eFigure 2 and 3B in the
cated in Fig. 2B. Funnel plots and Egger's regression intercept of supplement.
3.50 (p < 0.01) indicated the presence of potential publication bias
in this meta-analysis. High heterogeneity was observed (Q 44.02; 3.5.2. Age
p < 0.01; I2 77.28%). Meta-regression revealed that Lp(a) contrast Subgroup analysis of caseecontrol studies that investigated
level variation was the main source of heterogeneity. Studies that stroke populations with a mean age 55 years revealed a pooled
used categorized Lp(a) contrast levels (i.e. dichotomous, tertile, estimated OR of 1.94 (95% CI 0.86e4.40; Q 8.44, p 0.04;
quartile, or quintile) reported higher RRs, opposed to those studies I2 64.46%). Studies that investigated populations with a mean age
reporting risk per unit or per standard deviation (SD) increase of of >55 years had a pooled estimated OR of 1.37 (95% CI 1.22e1.53,
Lp(a) (QM 22.18, p < 0.01; beta 0.60; se 0.22). The variation in Fig. 4A) with high study heterogeneity (Q 100.05; p < 0.01;
reported effect sizes according to dened Lp(a) contrast levels of I2 92%).
each study are displayed in eFigure 2 in the supplement. Prospective studies [26,50,51,59,60,62] with a mean population
Combining the studies of categorized Lp(a) contrast levels and age 55 years followed a similar trend. Pooled analysis showed a
combining the studies with increase per unit change or per SD higher risk ratio of ischemic stroke with high Lp(a) levels (pooled
increase of Lp(a), revealed a substantial difference in the pooled adjusted RR of 1.36, 95% CI 1.05e1.76; Q 31.08; p < 0.01;
estimated effect sizes with a RR of 1.64 (95% CI 1.37e1.98) and I2 77.48%) compared to studies [13,25,52] with a mean population
0.93 (95% CI 0.84e1.02), respectively (Fig. 3), and dissolved the age >55 years (RR 1.12; 95% CI 0.80e1.59; Q 6.24; p 0.04;
observed heterogeneity in total (I2 of 20.44% and 0.00%, I2 67.93%, Fig. 4B). The proportion of studies reporting linear risk
respectively). calculations did not differ between groups.

3.5. Subgroup analyses 4. Discussion

3.5.1. Sex This meta-analysis of 20 studies that met our stringent inclusion
Three different caseecontrol studies reported data of elevated criteria reinforces the association of Lp(a) with ischemic stroke
Lp(a) serum levels of male ischemic stroke patients [27,47,48]. after adjustment for several risk factors. The pooled adjusted OR
Pooling this data resulted in a signicant positive association be- was 1.41 (95% CI, 1.26e1.57) for all included caseecontrol studies.
tween high Lp(a) levels and ischemic stroke in men (pooled esti- For prospective cohort studies, the pooled adjusted RR was 1.29
mated OR 2.40; 95% CI 1.69e3.41; Q 0.57; p 0.75; (95% CI, 1.06e1.58, Fig. 2). These ndings are in accordance with
I2 0.00%). Pooled analysis of 4 studies [27,47,48,56] that reported previous ndings that have reported that Lp(a) plays a modest role
500 A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503

Fig. 3. Forest plot of prospective studies grouped by contrast levels. Forest plot illustrating number of cases (n) and overall sample size (N), pooled adjusted risk ratio with 95%
condence interval (95% CI), and p value (P) for all prospective studies grouped by comparable reported Lp(a) contrast levels: categorized contrast levels (dichotomous, tertiles,
quartiles, and quintiles, panel A) and more linear relationship (per continuous and per standard deviation increase in Lp(a), respectively, panel B). Data on subgroups were reported
separately. Box sizes represent the individual study weight. aMale patients; bFemale patients.

Fig. 4. Pooled Analysis of Studies (55 vs. >55 years of mean age). Forest plots highlighting age-specic risk differences of casecontrol studies (A) and prospective studies (B)
reporting a mean age of study participants of 55 years or >55 years, respectively. Depicted are number of ischemic stroke patients (n) and total study participants (N), pooled
adjusted risk ratio with 95% condence interval (95% CI), and p value (P). Data on subgroups were reported separately. Box sizes represent the individual study weight. aMale
patients; bFemale patients; cLarge artery atherosclerosis patients; dSmall vessel occlusion patients.
A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503 501

in stroke risk [16,17]. gave rise to the assumption that Lp(a) may play an underlying role.
The association becomes particularly evident in (1) studies Therefore, we decided to analyze the results of studies reporting
including younger patients and (2) studies that categorized Lp(a) data on younger stroke populations (55 years). In general, the
rather than having analyzed it on a continuous scale (Fig. 3). Both studies included within our meta-analyses were heterogeneous
ndings are not unexpected. First, moderate risk factors may with regards to age. Further studies should evaluate the ischemic
exhibit higher RRs in the young due to the low absolute risk in this stroke risk of young adults with high Lp(a) and address the risk of
group. Also, because environmental risk factors accumulate with recurrent stroke in young adult stroke survivors with high Lp(a)
age, the effect of a weaker, genetic risk factor like Lp(a) will also be serum levels.
more pronounced in the young in terms of relative risk. Although Meta-regression analysis of caseecontrol studies revealed that
the observed difference could be due to chance, these two phe- stroke subtype was a source of heterogeneity. More specically,
nomena possibly explain the different RRs in our age-specic large artery atherosclerosis and strokes of undetermined source
subgroup analysis (Fig. 4). Second, studies that use a categoriza- were signicantly associated with higher ORs. In contrast to case-
tion of Lp(a) levels are less bound to the constraints of the under- econtrol studies, most prospective studies did not report data
lying regression model. In contrast, studies assessing risk per SD or regarding stroke subtypes. To elucidate critical patient subgroups
per unit increase in Lp(a) tend to yield lower RR and force a linear susceptible for ischemic stroke at higher Lp(a) serum levels, future
association on the scale of the regression model which most likely prospective studies should provide more information on ischemic
does not correspond to the true doseeresponse relationship stroke subtypes and respective subgroups [68e70]. In particular,
(eFigure 2). We recommend future studies use standardized cate- prospective cohort studies seldom report patient ethnicities
gorized Lp(a) contrast levels. Comparing results between several whereas caseecontrol studies have increasingly focused on the
studies using various contrasts may also be a limitation of our racial differences of stroke risk in relation to high Lp(a) levels.
analysis. Converting eligible prospective studies (i.e. studies that However, our subgroup analysis of ethnic-specic risk differences
reported risk ratios from tertiles, quartiles, quintiles, or risk per SD of caseecontrol studies did not reveal that a particular ethnicity
increase) to a common unit of exposure (tertiles), as previously was at an increased risk (see eFigure 4).
introduced by Danesh et al. [63], revealed, as expected, less het- Interventional studies addressing Lp(a) and ischemic stroke are
erogeneity and a less pronounced risk ratio of 1.32 (95% CI scarce and our systematic literature search only yielded one hit
1.05e1.66, I2 37.66%), compared to pooled data of categorized [53]. The authors of the study reported that daily aspirin (150 mg
contrasts displayed in Fig. 3. However, not all prospective studies for 4 weeks) decreased Lp(a) levels by >45%. However, the study
were eligible for this analysis, which has to be considered when lacked clinical or control data. Hand searching revealed
interpreting our results. another study (by Leebmann et al. [10]), which mainly focused on
Sex-specic risk differences were inconsistent between case- Lp(a) and coronary events as the primary endpoint, but also
econtrol and prospective studies and show contradictory results reported the number of cerebrovascular events as a secondary
(eFigure 3). Previous large, prospective studies have shown a sex- outcome parameter. In a cohort of 170 patients with Lp(a)-
specic risk difference of high Lp(a) for overall stroke, with an hyperlipoproteinemia and the maximum tolerable lipid lowering
increased risk in men [64,65]. This meta-analysis, however, can medication dose, the authors found that lipid apheresis effectively
only conrm the presumably higher risk of ischemic stroke in men reduced the frequency of cardiovascular and cerebrovascular
for caseecontrol studies. Analysis of prospective cohort studies did events over a follow-up period of 2 years. New pharmacological
not equally reproduce ndings of the caseecontrol studies, which substances, e.g. the anti-sense oligonucleotide of apolipoprotein-B
are more susceptible to reverse causation and overestimation of the mRNA such as mipomersen or inhibitors of the proprotein con-
true relative risk. In fact, the meta-analysis of prospective studies vertase subtilisin kexin type 9 (PSCK9) might develop as reasonable
surprisingly indicated the opposite. High Lp(a) clearly seemed to be candidates for future treatment strategies [71]. If these evolving
a risk factor for ischemic stroke in women (eFigure 3B). This nding therapeutic options prove to be effective treatment strategies for
might be inuenced by the higher proportion of studies reporting high Lp(a) levels, the results of this study should encourage clini-
linear risk calculations that was present among males and the fact cians to determine Lp(a) serum levels in stroke patients more
that the Women's Health study compared women with very high frequently.
Lp(a) levels (>44 mg/dl) to women with low levels resulting in
strong effect sizes [62]. Hence, the published data on sex- 4.1. Limitations
dependent risk differences remain unclear. This is particularly
true as other factors in the comparison between men and women Our systematic search considered studies from 2006 and 2009
can obscure the true effect: In contrast to our pooled prospective to December 2014 and should therefore be considered an update of
results for females, Berger et al. prospectively analyzed post- previous reviews [16,17]. As we only addressed ischemic stroke, we
menopausal women and did not nd an association between followed stricter inclusion criteria and therefore not all studies
elevated Lp(a) and ischemic stroke. However, approximately 40% of included in previous meta-analyses were eligible for inclusion into
women had received hormonal therapy. This might have contrib- our re-assessment. Still, the results are comparable thereby rein-
uted to the negative nding which is in line with interventional forcing the pathological role of Lp(a) in ischemic stroke. As present
studies that have shown a signicant decrease in Lp(a) by hormonal in the previous analysis by Smolders et al. [17], high levels of het-
replacement therapy in postmenopausal women [66]. Further, erogeneity are also present in our analysis. However, meta-
post-menopausal women in the study had a mean age of ~70 years. regression identied varying Lp(a) contrast levels and ischemic
Children with high Lp(a) serum levels have more than a fourfold stroke subtypes as the primary sources of heterogeneity. If studies
risk of ischemic stroke [11], and more than a tenfold increased risk reported multiple Lp(a) contrasts, the contrast that revealed the
of recurrent ischemic stroke (children with Lp(a) levels >90th highest risk ratio was used for further analyses. Meta-analyzing
percentile) [12]. Caseecontrol studies have shown that high Lp(a) data with high vs. low contrast irrespective of the categorization
levels also seem to be a relevant risk factor in young adults will lead to misclassication and result in an underestimation of
[27,58,67,68]. The sifap study [2], which is, so far, the largest study the true effect that can be expected in a quartile or quintile analysis.
on stroke in the young (18e55 years), emphasized the importance Individual patient characteristics differed substantially between
of atherosclerotic pathology in this cohort of stroke survivors. This studies, making direct comparisons of study results problematic
502 A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503

[14,72]. A major limitation is the lack of access to individual patient References

data, as it would have allowed for more standardized analyses in
terms of dened Lp(a) contrast levels and adjustments, as well as [1] A.S. Go, D. Mozaffarian, V.L. Roger, E.J. Benjamin, J.D. Berry, M.J. Blaha, et al.,
Executive summary: heart disease and stroke statisticse2013 update: a report
for the investigation of within-study heterogeneity. The absence of from the American heart association, Circulation 127 (1) (2013) 143e152.
standardized detection methods for Lp(a) limit the comparability of [2] A. Rolfs, F. Fazekas, U. Grittner, M. Dichgans, P. Martus, M. Holzhausen, et al.,
studies because (1) not all studies have used isoform insensitive Acute cerebrovascular disease in the young: the stroke in young Fabry pa-
tients study, Stroke 44 (2) (2013) 340e349.
detection assays and (2) reported storage times of frozen samples [3] J. Putaala, T. Tatlisumak, Prime time for dissecting the entity of cryptogenic
varied substantially across studies (see eTable 5). Moreover, the lack stroke, Stroke 45 (4) (2014) 950e952.
of Mendelian randomization studies that have investigated a ge- [4] L. Berglund, R. Ramakrishnan, Lipoprotein(a): an elusive cardiovascular risk
factor, Arterioscler. Thromb. Vasc. Biol. 24 (12) (2004) 2219e2226.
netic relationship between Lp(a) and stroke did not allow us to [5] G. Utermann, The mysteries of lipoprotein(a), Science 246 (4932) (1989)
assess a causative role of Lp(a) in ischemic stroke. 904e910.
[6] T. Cho, R. Romagnuolo, C. Scipione, M.B. Boffa, M.L. Koschinsky, Apolipopro-
tein(a) stimulates nuclear translocation of b-catenin: a novel pathogenic
4.2. Conclusion mechanism for lipoprotein(a), Mol. Biol. Cell 24 (3) (2013) 210e221.
[7] R. Clarke, J.F. Peden, J.C. Hopewell, T. Kyriakou, A. Goel, S.C. Heath, et al., Ge-
netic variants associated with Lp(a) lipoprotein level and coronary disease,
In conclusion, our meta-analysis emphasizes the importance of N. Engl. J. Med. 361 (26) (2009) 2518e2528.
high Lp(a) as an independent but modest risk factor in ischemic [8] P.R. Kamstrup, A. Tybjaerg-Hansen, R. Steffensen, B.G. Nordestgaard, Geneti-
stroke. This becomes especially evident in younger stroke pop- cally elevated lipoprotein(a) and increased risk of myocardial infarction, JAMA
301 (22) (2009) 2331e2339.
ulations. Conicting sex-specic results warrant further studies, [9] P.R. Kamstrup, A. Tybjrg-Hansen, B.G. Nordestgaard, Extreme lipoprotein(a)
preferably using standardized categorized Lp(a) contrast levels. levels and improved cardiovascular risk prediction, J. Am. Coll. Cardiol. 61 (11)
(2013) 1146e1156.
[10] J. Leebmann, E. Roeseler, U. Julius, F. Heigl, R. Spitthoever, D. Heutling, et al.,
Author contribution Lipoprotein apheresis in patients with maximally tolerated lipid-lowering
therapy, lipoprotein(a)-hyperlipoproteinemia, and progressive cardiovascu-
lar disease: prospective observational multicenter study, Circulation 128 (24)
Study concept: Nave AH, Lange KS, Ebinger M. (2013) 2567e2576.
Acquisition, analysis, or interpretation of data: All authors. [11] S.M. Sultan, N. Schupf, M.M. Dowling, G.A. Deveber, A. Kirton, M.S.V. Elkind,
Review of lipid and lipoprotein(a) abnormalities in childhood arterial
Drafting of the manuscript: Nave AH, Lange KS. ischemic stroke, Int. J. Stroke 9 (1) (2014) 79e87.
Critical revision of the manuscript for important intellectual [12] N.A. Goldenberg, T.J. Bernard, J. Hillhouse, J. Armstrong-Wells, J. Galinkin,
content: All authors. R. Knapp-Clevenger, et al., Elevated lipoprotein (a), small apolipoprotein (a),
and the risk of arterial ischemic stroke in North American children, Haema-
Statistical analysis: Nave AH, Leonards CO.
tologica 98 (5) (2013) 802e807.
Obtained funding: Ebinger M, Endres M. [13] D. Gurdasani, B. Sjouke, S. Tsimikas, G.K. Hovingh, R.N. Luben,
Administrative, technical, or material support: Nave AH, Leo- N.W.J. Wainwright, et al., Lipoprotein(a) and risk of coronary, cerebrovascular,
and peripheral artery disease: the EPIC-Norfolk prospective population study,
nards CO, Ebinger M.
Arterioscler. Thromb. Vasc. Biol. 32 (12) (2012) 3058e3065.
Study supervision: Ebinger M. [14] H.J. Milionis, Lipoprotein (a) and stroke, J. Clin. Pathol. 53 (7) (2000) 487e496.
[15] L. Christogiannis, H.J. Milionis, M. Elisaf, Lipoprotein (a) and stroke : an
overview, Open Clin. Chem. J. (2010) 38e43.
Sources of funding [16] S. Erqou, S. Kaptoge, P.L. Perry, E. Di Angelantonio, A. Thompson, I.R. White, et
al., Lipoprotein(a) concentration and the risk of coronary heart disease, stroke,
and nonvascular mortality, JAMA 302 (4) (2009) 412e423.
This study was funded by the Federal Ministry of Education and [17] B. Smolders, R. Lemmens, V. Thijs, Lipoprotein (a) and stroke: a meta-analysis
Research through the grant G.2.15 of the Center for Stroke Research of observational studies, Stroke 38 (6) (2007) 1959e1966.
Berlin (CSB). [18] D. Moher, A. Liberati, J. Tetzlaff, D.G. Altman, Preferred reporting items for
systematic reviews and meta-analyses: the PRISMA statement, Ann. Intern
Med. 151 (4) (2009) 264e269. W64.
[19] D.F. Stroup, J.A. Berlin, S.C. Morton, I. Olkin, G.D. Williamson, D. Rennie, et al.,
Disclosures Meta-analysis of observational studies in epidemiology: a proposal for
reporting. Meta-analysis of observational studies in epidemiology (MOOSE)
AHN, KSL, COL, BS, WD, MEn and MEb, have nothing to disclose group, JAMA 283 (15) (2000) 2008e2012.
[20] M. Egger, G.D. Smith, A.N. Phillips, Meta-analysis: principles and procedures,
and no conicts of interest. EST has received honoraria for pre-
BMJ 315 (7121) (1997) 1533e1537.
sentations, advisory board activities, or data monitoring committee [21] R.D. Riley, J.P.T. Higgins, J.J. Deeks, Interpretation of random effects meta-
activities from Abbott, Aegerion, Amgen, Astra-Zeneca, Boehringer- analyses, BMJ 342 (2011) (Feb10 2):d549ed549.
[22] M. Egger, G. Davey Smith, M. Schneider, C. Minder, Bias in meta-analysis
Ingelheim, Bristol-Myers Squibb, Fresenius, Genzyme, a Sano
detected by a simple, graphical test, BMJ 315 (7109) (1997) 629e634.
company, Kaneka, Merck Sharp & Dohme, Novartis, Roche. [23] L. Bax, L.-M. Yu, N. Ikeda, H. Tsuruta, K.G.M. Moons, Development and vali-
dation of MIX: comprehensive free software for meta-analysis of causal
research data, BMC Med. Res. Methodol. 6 (2006) 50.
Acknowledgments [24] W. Viechtbauer, Conducting meta-analyses in R with the meta for package,
J. Stat. Softw. 36 (3) (2010) 1e48.
[25] J.S. Berger, A.P. McGinn, B.V. Howard, L. Kuller, J.E. Manson, J. Otvos, et al.,
AHN had full access to all the data in the study and takes re- Lipid and lipoprotein biomarkers and the risk of ischemic stroke in post-
sponsibility for the integrity of the data and the accuracy of the data menopausal women, Stroke 43 (4) (2012) 958e966.
analysis. For general support and excellent statistical advice in [26] S. Uchiyama, N. Nakaya, K. Mizuno, Y. Ohashi, N. Tajima, T. Kushiro, et al., Risk
factors for stroke and lipid-lowering effect of pravastatin on the risk of stroke
conducting the systematic review and meta-analysis we thank Dr. in Japanese patients with hypercholesterolemia: analysis of data from the
Ulrike Grittner from the Center for Stroke Research Berlin and the MEGA study, a large randomized controlled trial, J. Neurol. Sci. 284 (1e2)
Department of Biostatistics and Clinical Epidemiology of the (2009) 72e76.
[27] M. Rigal, J.B. Ruidavets, A. Viguier, R. Petit, B. Perret, J. Ferrieres, et al., Lipo-
Charite protein (a) and risk of ischemic stroke in young adults, J. Neurol. Sci. 252 (1)
(2007) 39e44.
[28] B.S. Kim, H.S. Jung, O.Y. Bang, C.S. Chung, K.H. Lee, G.M. Kim, Elevated serum
Appendix A. Supplementary data lipoprotein(a) as a potential predictor for combined intracranial and extra-
cranial artery stenosis in patients with ischemic stroke, Atherosclerosis 212
(2) (2010) 682e688.
Supplementary data related to this article can be found at http://
[29] R.K. Dhamija, P. Gaba, S. Arora, a Kaintura, M. Kumar, J. Bhattacharjee,
A.H. Nave et al. / Atherosclerosis 242 (2015) 496e503 503

Homocysteine and lipoprotein (a) correlation in ischemic stroke patients, [51] F. Canou-Poitrine, G. Luc, J.-M. Bard, J. Ferrieres, J. Yarnell, D. Arveiler, et al.,
J. Neurol. Sci. 281 (1e2) (2009) 64e68. Relative contribution of lipids and apolipoproteins to incident coronary heart
[30] H.J. Milionis, T.D. Filippatos, T. Loukas, E.T. Bairaktari, A.D. Tselepis, M.S. Elisaf, disease and ischemic stroke: the PRIME Study, Cerebrovasc. Dis. 30 (3) (2010)
Serum lipoprotein(a) levels and apolipoprotein(a) isoform size and risk for 252e259.
rst-ever acute ischaemic nonembolic stroke in elderly individuals, Athero- [52] K.K. Teo, L.B. Goldstein, B.R. Chaitman, S. Grant, W.S. Weintraub,
sclerosis 187 (1) (2006) 170e176. D.C. Anderson, et al., Extended-release niacin therapy and risk of ischemic
[31] K. Taseva, S. Fischer, J. Passauer, N. Weiss, S.R. Bornstein, U. Julius, Factors stroke in patients with cardiovascular disease: the atherothrombosis inter-
inducing cardiovascular events in patients treated by lipoprotein apheresis, vention in metabolic syndrome with low HDL/high triglycerides: impact on
Atheroscler. Suppl. 14 (1) (2013) 45e50. global health outcome (AIM-HIGH) trial, Stroke 44 (10) (2013) 2688e2693.
[32] B. Wiberg, J. Sundstro m, J. Arnlov, A. Tere
nt, B. Vessby, B. Zethelius, et al., [53] G.S. Ranga, O.P. Kalra, H. Tandon, J.K. Gambhir, G. Mehrotra, Effect of aspirin
Metabolic risk factors for stroke and transient ischemic attacks in middle- on lipoprotein(a) in patients with ischemic stroke, J. Stroke Cerebrovasc. Dis.
aged men: a community-based study with long-term follow-up, Stroke 37 16 (5) (2007) 220e224.
(12) (2006) 2898e2903. [54] E. Emanuele, E. Peros, C. Tomaino, E. Feudatari, L. Bernardi, G. Binetti, et al.,
[33] C.J. Glueck, J. Munjal, D. Aregawi, M. Agloria, M. Winiarska, Q. Khalil, et al., Relation of apolipoprotein(a) size to Alzheimer's disease and vascular de-
Thrombophiliaehypobrinolysis and atherothrombotic cardiovascular dis- mentia, Dement. Geriatr. Cogn. Disord. 18 (2004) 189e196.
ease  age 45 years, Transl. Res. 150 (2) (2007) 93e100. [55] L. Sun, Z. Li, H. Zhang, A. Ma, Y. Liao, D. Wang, et al., Pentanucleotide TTTTA
[34] J. Zhou, X. Cui, X. Jin, J. Zhou, M. Fu, C. Zhong, et al., Association between li- repeat polymorphism of apolipoprotein(a) gene and plasma lipoprotein(a) are
poprotein (a) level on admission and the incidence of subsequent cardio- associated with ischemic and hemorrhagic stroke in Chinese: a multicenter
vascular events in patients with acute coronary syndrome, Int. J. Cardiol. 158 case-control study in China, Stroke 34 (2003) 1617e1622.
(3) (2012) 464e466. [56] R.J. Wityk, S.J. Kittner, J.L. Jenner, J.R. Hebel, A. Epstein, M.A. Wozniak, et al.,
[35] M.A. Ban ~ os-Gonz alez, M.A. Pen ~ a-Duque, E. Angles-Cano, M.A. Martinez-Rios, Lipoprotein (a) and the risk of ischemic stroke in young women, Athero-
A. Bahena, B. Valente-Acosta, et al., Apo(a) phenotyping and long-term sclerosis 150 (2) (2000) 389e396.
prognosis for coronary artery disease, Clin. Biochem. 43 (7e8) (2010) [57] D.Q. Peng, S.P. Zhao, J.L. Wang, Lipoprotein (a) and apolipoprotein E epsilon 4
640e644. as independent risk factors for ischemic stroke, J. Cardiovasc Risk 6 (1999)
[36] L.A. Simons, J. Simons, Y. Friedlander, J. McCallum, A comparison of risk factors 1e6.
for coronary heart disease and ischaemic stroke: the Dubbo study of [58] J. Peynet, J.L. Beaudeux, F. Woimant, F. Flourie , V. Giraudeaux, E. Vicaut, et al.,
Australian elderly, Hear Lung Circ. 18 (5) (2009) 330e333. Apolipoprotein(a) size polymorphism in young adults with ischemic stroke,
[37] S. Tsimikas, G.W. Duff, P.B. Berger, J. Rogus, K. Huttner, P. Clopton, et al., Pro- Atherosclerosis 142 (1) (1999) 233e239.
inammatory interleukin-1 genotypes potentiate the risk of coronary artery [59] C.A. Glader, B. Stegmayr, J. Boman, H. Stenlund, L. Weinehall, G. Hallmans, et
disease and cardiovascular events mediated by oxidized phospholipids and al., Chlamydia pneumoniae antibodies and high Lipoprotein(a) levels do not
lipoprotein (a), J. Am. Coll. Cardiol. 2014 (2014). predict ischemic cerebral infarctions. Results from a nested case-control study
[38] S. Ishikawa, K. Kotani, K. Kario, K. Kayaba, T. Gotoh, Y. Nakamura, et al., Inverse in Northern Sweden, Stroke 30 (1999) 2013e2018.
association between serum lipoprotein(a) and cerebral hemorrhage in the [60] G. Alfthan, J. Pekkanen, M. Jauhiainen, J. Pitk aniemi, M. Karvonen,
Japanese population, Thromb. Res. 131 (2) (2013) e54ee58. J. Tuomilehto, et al., Relation of serum homocysteine and lipoprotein(a)
[39] S.-B. Jeon, S. Chun, S. Choi-Kwon, H.-S. Chi, H.-W. Nah, S.U. Kwon, et al., concentrations to atherosclerotic disease in a prospective Finnish population
Biomarkers and location of atherosclerosis: matrix metalloproteinase-2 may based study, Atherosclerosis 106 (1994) 9e19.
be related to intracranial atherosclerosis, Atherosclerosis 223 (2) (2012) [61] J. Suk Danik, N. Rifai, J.E. Buring, P.M. Ridker, Lipoprotein(a), hormone
442e447. replacement therapy, and risk of future cardiovascular events, J. Am. Coll.
[40] H. Yokokawa, A. Goto, K. Terui, Y. Funami, K. Watanabe, S. Yasumura, Preva- Cardiol. 52 (2) (2008) 124e131.
lence of metabolic syndrome and serum marker levels in patients with four [62] J. Suk Danik, N. Rifai, J.E. Buring, P.M. Ridker, Lipoprotein(a), measured with an
subtypes of cerebral infarction in Japan, J. Clin. Neurosci. 15 (7) (2008) assay independent of apolipoprotein (a) isoform size, and risk of future car-
769e773. diovascular events among initially healthy women, JAMA 296 (11) (2006)
[41] B. Chakraborty, G. Vishnoi, B. Goswami, S.H. Gowda, D. Chowdhury, 1363e1370.
S. Agarwal, Lipoprotein(a), ferritin, and albumin in acute phase reaction [63] J. Danesh, R. Collins, P. Appleby, R. Peto, Association of brinogen, C-reactive
predicts severity and mortality of acute ischemic stroke in North Indian pa- protein, albumin, or leukocyte count with coronary heart disease: meta-
tients, J. Stroke Cerebrovasc. Dis. 22 (7) (2013) e159ee167. analyses of prospective studies, JAMA 279 (18) (1998) 1477e1482.
[42] G.T. Jones, A.M. van Rij, J. Cole, M.J.A. Williams, E.H. Bateman, S.M. Marcovina, [64] A.A. Ariyo, C. Thach, R. Tracy, Lp(a) lipoprotein, vascular disease, and mortality
et al., Plasma lipoprotein(a) indicates risk for 4 distinct forms of vascular in the elderly, N. Engl. J. Med. 349 (22) (2003) 2108e2115.
disease, Clin. Chem. 53 (4) (2007) 679e685. [65] T.T. Nguyen, R.D. Ellefson, D.O. Hodge, K.R. Bailey, T.E. Kottke, H.S. Abu-Leb-
[43] T. Ohira, P.J. Schreiner, J.D. Morrisett, L.E. Chambless, W.D. Rosamond, deh, Predictive value of electrophoretically detected lipoprotein(a) for coro-
A.R. Folsom, Lipoprotein(a) and incident ischemic stroke: the atherosclerosis nary heart disease and cerebrovascular disease in a community-based cohort
risk in communities (ARIC) study, Stroke 37 (6) (2006) 1407e1412. of 9936 men and women, Circulation 96 (5) (1997) 1390e1397.
[44] W. Zhang, X.-A. Zhang, Prognostic value of serum lipoprotein(a) levels in [66] M.A. Espeland, S.M. Marcovina, V. Miller, P.D. Wood, C. Wasilauskas,
patients with acute ischemic stroke, Neuroreport 25 (4) (2014) 262e266. R. Sherwin, et al., Effect of postmenopausal hormone therapy on lip-
[45] N.H. Petersen, A.B. Schmied, J.A. Zeller, H. Plendl, G. Deuschl, P. Zunker, Lp(a) oprotein(a) concentration, Circulation 97 (1998) 979e986.
lipoprotein and plasminogen activity in patients with different etiology of [67] H.W. Hahmann, D. Scha tzer-Klotz, T. Bunte, D. Becker, H.J. Schieffer, The sig-
ischemic stroke, Cerebrovasc. Dis. 23 (2e3) (2007) 188e193. nicance of high levels of lipoprotein (a) compared with established risk
[46] L. Ma, J. Wu, D. Niu, R. Yu, J. Song, C. Zhang, et al., Serum lipoprotein(a) factors in premature coronary artery disease: differences between men and
complexes with beta2-glycoprotein I levels in patients with ischemic stroke, women, Atherosclerosis 144 (1) (1999) 221e228.
Clin. Chim. Acta 429 (0) (2014) 163e167. [68] M. Nagayama, Y. Shinohara, T. Nagayama, Lipoprotein(a) and ischemic cere-
[47] B. Boden-Albala, D.E. Kargman, I.-F. Lin, M.C. Paik, R.L. Sacco, L. Berglund, brovascular disease in young adults, Stroke 25 (1) (1994) 74e78.
Increased stroke risk and lipoprotein(a) in a multiethnic community: the [69] S. Shintani, S. Kikuchi, H. Hamaguchi, T. Shiigai, High serum lipoprotein(a)
Northern Manhattan stroke study, Cerebrovasc. Dis. 30 (3) (2010) 237e243. levels are an independent risk factor for cerebral infarction, Stroke 24 (7)
[48] S.-Y. Li, Y. Gao, W.-N. Ma, H.-C. Wang, G. Zhou, W.-C. Guo, et al., The rela- (1993) 965e969.
tionship between serum lipoprotein (a) levels and ischemic stroke risk: a [70] T. Hiraga, M. Shimada, M. Okubo, K. Nakanishi, T. Kobayashi, T. Murase, Lip-
cohort study in the Chinese population, Inammation 37 (3) (2014) 686e693. oprotein(a) is an independent risk factor for multiple cerebral infarctions,
[49] G.T. Jones, M. Deng, G.D. Hammond-Tooke, S.P.A. McCormick, A.M. Van Rij, Atherosclerosis 122 (1) (1996) 29e32.
Increased plasma lipoprotein(a) found in large-artery atherosclerotic, but not [71] F. Kronenberg, Genetic determination of lipoprotein(a) and its association
small-artery occlusive, stroke, Clin. Chem. 55 (2009) 1888e1890. with cardiovascular disease: convenient does not always mean better,
[50] S.S. Virani, A. Brautbar, B.C. Davis, V. Nambi, R.C. Hoogeveen, A.R. Sharrett, et J. Intern Med. 276 (3) (2014) 243e247.
al., Associations between lipoprotein(a) levels and cardiovascular outcomes in [72] H.H. Hobbs, A.L. White, Lipoprotein(a): intrigues and insights, Curr. Opin.
black and white subjects: the atherosclerosis risk in communities (ARIC) Lipidol. 10 (3) (1999) 225e236.
study, Circulation 125 (2) (2012) 241e249.