Blood Markers of Coagulation, Fibrinolysis, Endothelial Dysfunction and Inflammation in Lacunar Stroke versus Non-Lacunar Stroke and Non-Stroke: Systematic Review and Meta-Analysis

Lacunar stroke is an important stroke subtype, accounting for one quarter of ischaemic strokes. Its aetiology probably differs from other stroke subtypes, usually being the symptomatic manifestation of small vessel disease rather than large vessel atheroma or cardioembolism.

A previous review [1] of symptomatic lacunar stroke versus mainly non-stroke controls suggests a pathogenic role for endothelial dysfunction but this could simply reflect having an ischaemic stroke in general [2]. Another review [3] found C-reactive protein (CRP), P-selectin and homocysteine to differ significantly between ischaemic stroke (in general) and non-stroke controls, but did not assess levels of blood markers between ischaemic stroke subtypes.

We sought to clarify if differences exist in blood markers between lacunar stroke and other ischaemic stroke subtypes by reviewing the literature for studies measuring coagulation, fibrinolysis, endothelial dysfunction and inflammation. We sought to disentangle the acute phase response by splitting the analysis on timing of the blood draw in relation to the stroke.

This review has been prepared in accordance with The PRISMA statement [4]. We extracted data and conducted the meta-analysis in accordance with MOOSE [5], which was modified for our needs using 3 reporting standards [6,7,8].

We used OVID to search MEDLINE (from 1966) and Embase (from 1980) on July 15th, 2012 using a strategy (see Appendix) developed with advice from the Cochrane Stroke Group (http://stroke.cochrane.org/).

We included English language studies published in full measuring blood markers in plasma or serum as follows: tissue plasminogen activator (tPA) as an indicator of fibrinolytic and thrombotic state; plasminogen activator inhibitor (PAI) as an inhibitor of tPA; fibrinogen as a measure of coagulation; D-dimer as a measure of fibrinolysis; homocysteine as a marker of endothelial toxicity; von Willebrand factor (vWF) as a marker of endothelial damage; E-selectin, P-selectin, intercellular adhesion molecule-1 (ICAM) and vascular cellular adhesion molecule-1 (VCAM) as markers of endothelial activation and CRP, interleukin-6 (IL-6) and tumour necrosis factor-alpha (TNF-α) as markers of inflammation. We excluded studies that did not provide information on blood markers in lacunar stroke and a control group.

We extracted data on study population (sample size, age, sex, co-morbidities, current medications); study design; control group (including details on matching for age, sex and covariates where given); blinding of stroke assessor to blood markers and assay investigator to stroke diagnosis; blood markers assessed; time to all blood draws; whether the authors gave detailed methods on how samples were obtained, stored and processed, and results of blood marker levels (both measures of average and spread where given). We recorded method of stroke diagnosis, criteria/system used for stroke subtyping, details of imaging, definition of lacunar stroke and the grade of clinician involved in the stroke diagnosis. We contacted one author [9] to clarify time of blood draw. Some studies drew blood at multiple time points. We recorded data at multiple time points where a relevant comparator was also available. We avoided duplicate publications.

Our gold standard definition of lacunar stroke was all of the following: reference to one of the classical clinical lacunar syndromes; no evidence of cortical dysfunction; imaging (including noting a normal scan does not exclude a lacunar diagnosis), and mention of expertise of person who subtyped the stroke. We relied on the clinical stroke definitions used in the primary papers but tried to harmonize these to a clinical stroke syndrome with support from imaging where available [10].

We used the Review Manager 5 software (the Cochrane Collaboration) to calculate standardized mean differences (SMD, where data were in a suitable format) using the inverse variance method and a fixed effects model with 95% confidence intervals (CI). Our forest plots compare lacunar to non-stroke or to non-lacunar stroke (atherothrombotic stroke and/or cardioembolic stroke, as appropriate). Studies [11,12,13] reporting a geometric mean with 95% CI were converted to standard deviations (SD) using methods described in the Cochrane Handbook [14].

Not all studies could be meta-analysed due to heterogeneity of reporting, e.g. studies reporting medians with either interquartile ranges or minimum and maximum values, as we could not assume normality of distributions. Consequently, tPA, PAI, E-selectin, P-selectin, ICAM, VCAM, CRP and TNF-α were reviewed solely by summary of individual study data. Where studies provided data at more than one acute time point, we meta-analysed only the first time point as this most often corresponded with data for non-stroke comparators. In order to see if an acute phase response affected the results we split our analysis into ‘acute' and ‘chronic' (bloods drawn up to and after 21 days of stroke, respectively).

We identified 1,468 full papers. In all, 1,389 titles were excluded following a survey of titles and abstracts, leaving 79 for reading. Of these 10 were excluded (unable to translate) and 32 did not meet inclusion criteria (duplicates, asymptomatic subjects, non-relevant blood markers and no control group). Hand-searching identified a further 5 papers. Therefore, 42 papers were eligible, including 4,816 ischaemic strokes, of which 2,196 were lacunar and 2,500 non-stroke controls (see online suppl. fig. 1 and suppl. table 1; for all online suppl. material, see www.karger.com/doi/10.1159/000356789). In a further 4 papers [15,16,17,18] blood was collected prior to stroke. These studies are reported separately (see online suppl. material).

Over 50% of studies (22/42) used TOAST [19] to subtype ischaemic stroke. Just 4 studies [11,20,21,22] (fewer than 10%) met our gold standard definition of lacunar stroke. More than half of the studies (57%) reported a ‘minimal' definition of lacunar stroke; 1 study [23] failed to define lacunar stroke. Most studies recruited patients consecutively (31/42); 1 study recruited non-consecutively [20] and 10/42 did not report on recruitment. Two thirds of studies reported excluding cases based on co-existing disease such as concurrent infection, cancer, inflammatory disease and renal failure. One third of studies did not report on exclusion criteria. Most studies reported matching controls by age and sex to cases (24/42). Some studies matched age only (6/42) and 1 study matched sex only; 3 studies did not report on matching. Matching for co-morbidities varied from study to study. Less than 20% of studies (8/42) reported blinding of stroke assessor to blood marker values. Fewer still reported blinding of laboratory staff to stroke data (see online suppl. tables 3 and 4 for critical appraisal of included studies).

Tissue Plasminogen Activator. Here, we included 5 studies [24,25,26,27,28] (251 lacunar strokes) of which 1 study [26] contributed 50% of the data. There were insufficient data in an appropriate format for meta-analysis. Individual studies suggest tPA was significantly higher in lacunar stroke versus non-stroke controls, both acutely and chronically. Meanwhile, tPA does not appear to differ between lacunar stroke and other stroke subtypes, either acutely or chronically.

Plasminogen Activator Inhibitor. A total of 7 studies were included [24,25,26,27,28,29,30] (336 lacunar strokes) but available data did not permit meta-analysis. Jood et al. [26] report levels of PAI significantly higher in lacunar stroke versus non-stroke controls acutely and chronically. They also report lower PAI in lacunar stroke versus atherothrombotic and cardioembolic stroke acutely (significance not given) but not chronically [26]; 4 other studies report no difference between lacunar and non-lacunar stroke acutely [24,27,28,29].

Fibrinogen. We included 9 studies [20,23,25,31,32,33,34,35,36] (622 lacunar strokes) of which 6 permitted a meta-analysis (fig. 1). Fibrinogen showed no difference in lacunar stroke versus non-stroke controls acutely but was significantly higher in lacunar stroke chronically. Fibrinogen was significantly lower in lacunar stroke versus other ischaemic stroke subtypes, both acutely [atherothrombotic SMD -0.37 (95% CI -0.51, -0.22); cardioembolic SMD -0.83 (-1.15, -0.51)] and chronically (fig. 1), although the chronic data comprised only 1 study [32]. In 2 [31,34] of 3 [31,33,34] studies that could not be meta-analysed no difference was reported in levels of fibrinogen between lacunar and non-lacunar stroke in the acute phase; the third study [33] did not report on whether its findings were significant.

D-Dimer. Here, 9 studies were included [25,30,32,34,37,38,39,40,41] (364 lacunar strokes) of which 4 could be meta-analysed (fig. 2). D-dimer was significantly higher in lacunar stroke versus non-stroke controls, both acutely [SMD 1.42 (1.14, 1.69)] and chronically [SMD 3.22 (2.65, 3.78)]. D-dimer was significantly lower in lacunar versus non-lacunar stroke, both atherothrombotic [acute SMD -3.59 (-4.06, -3.12)] and cardioembolic [acute SMD -5.73 (-6.38, -5.09)], acutely and chronically.

The largest (n = 128 lacunar strokes) of the 5 studies [39] that could not be meta-analysed found no difference in D-dimer between lacunar and atherothrombotic stroke acutely (which disagrees with the meta-analysis) and found D-dimer significantly lower in lacunar versus cardioembolic stroke acutely (in agreement with the meta-analysis).

Homocysteine. Overall, we included 9 studies [9,11,12,29,35,42,43,44,45] (747 lacunar strokes) of which 8 could be meta-analysed (fig. 3). Homocysteine was significantly higher in lacunar stroke versus non-stroke controls, both acutely [SMD 0.55 (0.42, 0.69)] and chronically. Studies comparing lacunar to non-lacunar stroke drew blood acutely only and found no difference [atherothrombotic SMD -0.02 (-0.22, 0.18); cardioembolic SMD 0.16 (-0.02, 0.35)]. In 1 study [29] not included in the meta-analysis no difference was found in homocysteine between lacunar and non-lacunar stroke acutely.

von Willebrand Factor. We included 6 studies [13,27,31,35,46,47] (293 lacunar strokes) of which 2 [13,35] were meta-analysable (fig. 4). vWF was significantly higher in lacunar stroke versus non-stroke controls acutely. vWF was significantly lower in lacunar versus non-lacunar stroke [atherothrombotic SMD -0.34 (-0.61, -0.08); cardioembolic SMD -0.38 (-0.62, -0.14)] acutely. The 4 non-meta-analysable [27,31,46,47] studies show conflicting results (online suppl. table 1).

E-Selectin. Here, 4 studies were included [27,28,35,46] (130 lacunar strokes) but available data did not permit meta-analysis; 1 study [46] found E-selectin significantly higher in lacunar stroke versus non-stroke controls acutely but not at 1 month. Individual study reports suggest no difference between lacunar and non-lacunar stroke.

P-Selectin. A total of 7 studies were included [27,28,30,31,46,48,49] (227 lacunar strokes) but available data did not permit meta-analysis. P-selectin was significantly higher in lacunar versus non-stroke acutely in 2 studies [46,49] but 1 study [30] found no difference. Another study [48] found P-selectin significantly lower in lacunar versus atherothrombotic stroke acutely but all other studies found no difference or did not report significance levels. Tsai et al. [48] continued to show levels to be significantly lower at 1 month but this difference disappeared at 3 months. Only 1 study [31] found P-selectin significantly lower in lacunar versus cardioembolic stroke acutely; all other studies found no difference or did not report significance levels.

Intercellular Adhesion Molecule-1. We included 5 studies [21,27,28,50,51] (344 lacunar strokes) but available data did not permit meta-analysis. In 1 study [21] significantly higher ICAM was found in lacunar stroke versus non-stroke controls acutely; another study [50] found ICAM significantly higher in lacunar versus non-stroke controls chronically. There was no difference between lacunar and other stroke subtypes acutely and no data for lacunar versus non-lacunar stroke chronically.

Vascular Cellular Adhesion Molecule-1. Here, 3 studies were included [27,28,51] (121 lacunar strokes) but available data did not permit meta-analysis. Results suggested no difference between lacunar stroke and non-stroke acutely and no difference between lacunar and non-lacunar stroke acutely. No studies measured VCAM chronically.

C-Reactive Protein. We included 8 studies [29,34,35,39,47,49,52,53] (490 lacunar strokes) but available data did not permit meta-analysis; 2 studies [39,52] provided just over 50% of the data. CRP was higher in lacunar stroke versus non-stroke acutely in 3 studies [34,49,52], significantly so in 2, and the other [34] did not report if the higher value was significant. CRP was significantly lower in lacunar versus atherothrombotic stroke in 1 study [34] acutely; all other studies reported no difference (or did not state significance) between lacunar and non-lacunar stroke, acutely and chronically.

Tumour Necrosis Factor-alpha. Overall, 5 studies were included [21,22,27,47,53] (252 lacunar strokes) but available data did not permit meta-analysis; 1 study [21] provides 45% of the data. In 2 studies [21,22] levels of TNF-α were found to be significantly higher in lacunar stroke versus non-stroke controls acutely. TNF-α was significantly lower in 2 [27,47] and no different in 2 [22,53] studies reporting on lacunar versus non-lacunar stroke acutely. No studies measured TNF-α chronically.

Interleukin-6. A total of 9 studies were included [21,22,23,27,47,53,54,55,56] (340 lacunar strokes) of which 5 could be meta-analysed (fig. 5). IL-6 was significantly higher in lacunar stroke versus non-stroke acutely [SMD 1.3 (0.9, 1.7)], and significantly lower in lacunar versus non-lacunar stroke [atherothrombotic SMD -0.37 (-0.63, -0.10); cardioembolic SMD -0.52 (-0.82, -0.22)] acutely.

In 1 non-meta-analysable study [21] IL-6 was found to be significantly higher in lacunar stroke versus non-stroke acutely, in agreement with the meta-analysed studies; 2 non-meta-analysable studies [27,47] found IL-6 significantly lower in lacunar versus non-lacunar stroke acutely, also in agreement with the meta-analysed studies.

The only long-term study [55] found no difference between lacunar stroke and non-stroke controls from a wholly female cohort measured 3.5 years after stroke.

This review assessed blood markers of coagulation, fibrinolysis, endothelial dysfunction and inflammation in lacunar stroke versus non-stroke controls and other ischaemic stroke subtypes. While many markers were higher in lacunar stroke than in non-stroke controls, they were mostly lower in lacunar versus non-lacunar stroke. A brief summary follows.

tPA/PAI were significantly higher in lacunar stroke versus non-stroke, acutely and chronically, and did not differ between lacunar and non-lacunar stroke, acutely or chronically.

Fibrinogen did not differ between lacunar stroke and non-stroke acutely, although we only used a single time point from Kataoka et al. [32] (bloods drawn at <24 h). If we consider their second sample (at 7 days) as a further acute measurement, lacunar stroke tends towards being significantly higher than non-stroke. Fibrinogen was significantly higher in lacunar stroke versus non-stroke chronically. Fibrinogen was significantly lower in lacunar versus non-lacunar stroke, acutely and chronically. However, studies excluded from the meta-analysis tended to show no overall difference between lacunar and non-lacunar stroke and chronic data came from a single study.

D-dimer was significantly higher in lacunar stroke versus non-stroke, acutely and chronically, and significantly lower in lacunar versus non-lacunar stroke, acutely and chronically.

Homocysteine was significantly higher in lacunar stroke versus non-stroke, acutely and chronically, but did not differ between lacunar and non-lacunar stroke acutely (but there were no chronic phase studies).

vWF was significantly higher in lacunar stroke versus non-stroke, acutely, with conflicting evidence chronically. vWF was significantly lower in lacunar than non-lacunar stroke acutely (2 studies), with conflicting but non-meta-analysable evidence in other studies both acutely and chronically.

E-selectin was significantly higher in lacunar stroke versus non-stroke acutely (only 1 study) but not chronically and did not differ between lacunar and non-lacunar stroke, either acutely or chronically (only 1 study).

P-selectin was significantly higher in lacunar stroke versus non-stroke acutely in some but not all studies, and in the only study that reported a chronic measurement. P-selectin did not differ between lacunar and non-lacunar stroke, either acutely or chronically (only 1 study).

ICAM was significantly higher in lacunar stroke versus non-stroke, acutely and chronically (only 1 study), and did not differ between lacunar and non-lacunar stroke acutely (with no studies chronically).

VCAM did not differ between lacunar stroke and non-stroke nor between lacunar and non-lacunar stroke acutely. There were no studies chronically.

CRP was significantly higher in lacunar stroke versus non-stroke, acutely and chronically (only 1 study) and did not differ between lacunar and non-lacunar stroke acutely or chronically (only 1 study).

TNF-α was significantly higher in lacunar stroke versus non-stroke acutely with no studies chronically. There was conflicting evidence on levels of TNF-α in lacunar versus non-lacunar stroke acutely with no studies chronically.

IL-6 was significantly higher in lacunar stroke versus non-stroke acutely but did not differ chronically. IL-6 was significantly lower in lacunar versus non-lacunar stroke acutely, but there were no chronic phase studies.

This suggests that plasma marker elevation in lacunar stroke is likely to reflect the process of having a stroke rather than that systemic inflammation or endothelial dysfunction is specific to lacunar stroke. The available data were limited and do not exclude the possibility that peripheral inflammatory or endothelial dysfunction processes are associated with lacunar stroke specifically.

There were limitations to the studies. Most were small, with varying methods and an inconsistent definition of ‘lacunar stroke', as highlighted previously [57]. Papers reviewed used the term lacunar stroke to reflect a clinical entity, i.e. clinical presentation with a stroke, but definitions varied and we refer readers to the new neuroimaging standards [10]. We were not able to differentiate different mechanisms of lacunar stroke. Most lacunar strokes are due to recent small subcortical infarcts, and most of these relate to intrinsic small vessel disease. However, they also arise from atherothromboembolism (large artery) or cardioembolism in a small proportion of patients and it was not possible to differentiate these cases.

There was heterogeneity across several aspects of the methods. Many used TOAST [19] but as this uses risk factors to categorize patients it potentially introduces classification bias. A patient with an unclear diagnosis of lacunar stroke but concurrent hypertension or diabetes might (rightly or wrongly) be classified as ‘lacunar' using this system, although hypertension and diabetes were equally prevalent risk factors between ischaemic stroke subtypes in 21,980 stroke patients when subtypes were classified without risk factors [58]. Several did not report on whether their findings achieved statistical significance; in the absence of an explicit statement, we report this as ‘not stated'. In some studies blood was drawn after overnight fasting, whereas in others non-fasting blood was collected. Studies used different units of measurement and assay methods. None reported on whether the patients had recent infection or neutrophilia, or if these patients were excluded. Timing of blood draw in relation to stroke varied but is important to account for each marker's individual ‘response curve' which changes over time. Fassbender et al. [59] found levels of IL-6 to rise rapidly following onset of ischaemic stroke, reaching a plateau at 10 h until 3 days before returning to normal by day 7. They did not subtype stroke and hence their study was not included in this review. Between-study heterogeneity on time to blood draw complicates subsequent analysis, although meta-analyses use within-study data and so will have minimized any effect of between-study variation.

Our review had limitations. We did not study markers in cerebrospinal fluid. We did not review the association of marker levels with lesion size or clinical outcome as data were sparse. Ahmad et al. [60] found markers of neuronal damage correlated with infarct size, which might explain why marker levels in non-lacunar stroke were frequently higher than in lacunar stroke in the acute phase. We were not able to analyse differences between groups reported as top versus bottom quantiles.

Our review had strengths, including assessment of differences between stroke subtypes, quality assessment of included studies, meticulous extraction of data and meta-analysis thereof, wherever suitable data were available. Previous reviews compared lacunar stroke to non-stroke controls only and therefore did not distinguish lacunar stroke specifically from stroke in general.

To determine if there is a difference in coagulation, fibrinolysis, endothelial dysfunction and inflammation in lacunar versus other stroke subtypes requires a large prospective study of blood markers in accurately phenotyped patients with lacunar versus non-lacunar stroke classified using non-risk-factor-based definitions. Future studies should clearly define and diagnose lacunar stroke, avoid subtyping stroke using risk factor-based classifications, explicitly report negative findings and significance levels, and obtain blood several weeks post-stroke to avoid confounding from an acute phase response.

Search Strategy.

The authors acknowledge Dr. Stephen Makin for assistance in collecting some of the data. Stewart Wiseman is supported by a Principal's Career Development PhD Scholarship from the University of Edinburgh.

None.