Are Lacunar Strokes Really Different?

Abstract

Background and Purpose

Differences in risk factors between lacunar and nonlacunar infarcts might support a distinct arterial pathological process underlying lacunar infarction.

Methods

We did a systematic review of studies comparing risk factors in patients with lacunar versus nonlacunar infarction. For each risk factor, we calculated study-specific and pooled relative risks (RRs) for lacunar versus nonlacunar infarction.

Results

A total of 16 of 28 studies included risk factors in their ischemic stroke subtype definitions. Hypertension and diabetes appeared commoner among patients with lacunar versus nonlacunar infarction. However, analyses confined to studies using risk factor-free ischemic subtype definitions found only a marginal excess of hypertension with lacunar versus nonlacunar infarction (RR, 1.11; 95% CI, 1.04 to 1.19) and no difference for diabetes (RR, 0.95; 95% CI, 0.83 to 1.09). Atrial fibrillation and carotid stenosis were associated more with nonlacunar than lacunar infarction but less so when only studies using risk factor-free classifications were considered. Otherwise, there was no evidence of differences in risk factor profiles.

Conclusions

Risk factor-free ischemic stroke subtype classification methods should be used for comparing risk factor profiles between lacunar and nonlacunar subtypes.

Approximately one quarter of ischemic strokes are caused by lacunar infarcts: small, deep cerebral infarcts, from 2 to 20 mm in diameter.¹ These are presumed to result from the occlusion of single, small, perforating arteries supplying the deep subcortical areas of the brain. If the occlusive arterial pathology is distinct from the atherothromboembolic processes that occlude larger arteries, causing most other types of ischemic stroke, the best strategies for the investigation and treatment of patients with lacunar infarction might differ from those for patients with other ischemic stroke subtypes.

Current knowledge of the arterial pathology of lacunar infarction is based largely on Fisher’s meticulous clinicopathological studies, in which he serially dissected the vascular supply of a total of 68 lacunar infarcts in 18 postmortem brains.^2-6 He found that most symptomatic lacunar infarcts were associated with occlusion of perforating arteries ≈200 to 800 μm in diameter by atheromatous plaques, with or without complicating thrombus. Most asymptomatic lacunar infarcts were associated with occlusion of perforating arteries ≈40 to 200 μm in diameter by “lipohyalinosis,” a destructive small vessel lesion characterized in the acute phase by fibrinoid necrosis and in the healed phase by loss of normal wall architecture, collagenous sclerosis, and mural foam cells.⁷ However, it is difficult to draw firm conclusions from Fisher’s work because the number of patients included was small, most of the lacunar infarcts were asymptomatic, and infarcts related to stroke symptoms were studied months or even years after the acute event.

Further progress in understanding the arterial pathology of lacunar stroke has been limited. Pathological studies are rare because autopsy rates are declining, lacunar strokes have a low-case fatality rate,¹ and tracing the vascular supply of subcortical lesions is technically difficult and time consuming.⁷ Difficulties in imaging the small perforating intracranial arteries have made informative imaging studies scarce.

An alternative approach has been to compare the risk factor profiles of patients with lacunar infarcts versus those with nonlacunar infarcts because this may reveal differences suggestive of distinct arterial pathologies. However, these studies have tended to have methodological limitations: sample sizes were generally small; risk factors were inconsistently defined; and studies used a variety of different classification methods to define ischemic stroke subtypes (see Appendix A).

Some used a classification based primarily on the clinical features of the stroke syndrome (most commonly the Oxfordshire Community Stroke Project [OCSP] classification⁸), usually refined by the results of brain imaging (ie, if a patient’s computed tomography [CT] or magnetic resonance [MR] brain scan showed an infarct that was relevant to the presenting syndrome but whose site and size suggested a different ischemic stroke subtype classification from the clinical features alone, the patient was reclassified in line with the imaging findings).

Other studies used the clinical and imaging features of the presenting stroke but also included risk factors in their ischemic stroke subtype definitions (most commonly the Trial of Org 10172 in Acute Stroke Treatment [TOAST] classification⁹). This could bias the results of a comparison of the prevalence of such risk factors between subtypes. It might also bias comparisons between subtypes of risk factors not explicitly included in the definitions because many vascular risk factors are associated with each other.

A few studies relied on imaging findings alone to classify ischemic stroke subtypes, regardless of the patients’ symptoms. Some patients in these studies may have been classified on the basis of asymptomatic or remotely symptomatic visible infarcts, whereas those with a definite ischemic stroke but no visible infarct would have been excluded.

In this article, we report the findings of a systematic review of studies comparing the prevalence of a variety of risk factors in patients with lacunar versus those with nonlacunar ischemic stroke.

Methods

Study Identification

We sought studies published in English up to and including June 2003 comparing the prevalence of risk factors among patients with stroke attributable to lacunar versus nonlacunar cerebral infarction. We identified studies by comprehensive textword and MeSH-based electronic searches of Medline and Embase, designed to retrieve articles about ischemic stroke subtypes, especially lacunar strokes (for details see Appendix B, available online only at http://www.strokeaha.org); perusal of reference lists of all relevant articles identified, searching within books on cortical and subcortical stroke, and discussions with colleagues.

We subsequently decided to include only those studies published from 1985 onward because the few earlier studies we identified had very limited access to brain imaging (at that time, CT scanning), and this was often restricted to younger patients. We excluded studies among highly selected groups of patients, for example, randomized controlled trials, and studies in which the definitions used to classify ischemic stroke subtypes were unclear or in which there were data inconsistencies.

Data Extraction

We extracted information from each study on the population studied (community- or hospital-based; inpatients or outpatients; consecutive recruitment or not), ischemic stroke subtype definitions, risk factor definitions, and the numbers of lacunar and nonlacunar patients with each of the following risk factors: hypertension; diabetes; smoking; alcohol consumption; raised cholesterol concentration; previous transient ischemic attack (TIA); atrial fibrillation (AF); and carotid stenosis.

We included in the nonlacunar patient group all the nonlacunar ischemic stroke patients apart from the small proportion of patients with an “unusual” cause of stroke (eg, nonatherosclerotic vasculopathies or hematological disorders), for studies in which these were categorized separately. Both authors independently extracted data from the articles, resolving any disagreements by discussion.

Statistical Analyses

We grouped studies according to whether the classification used to define ischemic stroke subtypes: included the risk factor under study; included various risk factors, but not specifically the risk factor under study; was based on brain imaging alone; or was based on the clinical features of the stroke syndrome, usually refined by brain imaging, but did not include risk factors. For each risk factor, we calculated study-specific and pooled relative risks (RRs) with 95% CIs for lacunar versus nonlacunar infarction using Cochrane RevMan software.¹⁰ We used standard χ² tests to assess statistical heterogeneity between studies. In a post hoc sensitivity analysis, we confined our analyses to community-based studies or studies that had recruited consecutive patients from hospital admissions and outpatient clinics.

Results

Included Studies

We identified 41 potentially relevant studies. From these, we excluded 2 that were published before 1985,^11,12 5 for which it was not clear how ischemic stroke subtypes were defined,^13-17 1 with data inconsistencies,¹⁸ and 5 that included highly selected groups of patients.^19-23 This left 28 studies including 21 980 patients, of whom 5379 had a lacunar infarction.^24-51

The characteristics of the studies included are summarized in Table 1. Sixteen (16 216 patients) used classification methods that included risk factors in the definitions of ischemic stroke subtypes.^{24,27,28,30-33,36,40,42-45,48-50} Ten of these (10 705 patients) used the TOAST classification,^{28,31-33,40,43-45,48,50} which considers the presence of hypertension and diabetes to favor a diagnosis of lacunar infarction, whereas carotid stenosis of >50% and potential sources of cardiac embolism such as AF should be absent for this diagnosis. One study (1262 patients) defined ischemic stroke subtypes on the basis of the site and size of infarction on brain imaging alone.³⁹ Ten studies (4502 patients) defined ischemic subtypes on the basis of the clinical features of the stroke syndrome, generally modified by the site and size of any relevant infarct seen on brain imaging but not including risk factors (Table 1​; Appendix).^{29,34,35,37,38,41,46,47,51,52}

TABLE 1.

Characteristics of Studies Included

Study	Year*	Stroke Population Recruited	Mean Age	% Male	% With CT/MR Brain Scan¶	Ischemic Stroke Subtype Classification	Nonlacunar Comparison Group	Risk Factors Studied	Total Ischemic Stroke Patients Included (lacunar infarct patients)†
Ankara40	2002	Consecutive admissions to neurology department with complete evaluation	65	60	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, AF, smoking, HC	264 (66)
Athens48	2000	Consecutive admissions	71	58	100%#	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, TIA, AF, smoking, HC	850 (177)
Barcelona36	1999	Consecutive admissions	66	57	100%	Risk factor-based	All nonlacunar ischemic strokes‡	HBP, DM, TIA, smoking, HC	2720 (399)
Besançon39	2000	Consecutive admissions	69	57	100%	Imaging-based	All nonlacunar ischemic strokes with a single visible infarction on CT/MRI	HBP, DM, TIA, AF, smoking, alcohol, HC	1262 (243)
Boston45	2000	Consecutive admissions	69	54	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes	HBP, DM, AF smoking, alcohol, HC	410 (109)
Buenos Aires44	2001	Consecutive admissions	64	64	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, AF, smoking, alcohol, HC	234 (105)
Cincinnati49	1999	Admissions or autopsied cases (among blacks)	NR	NR	97%#	Risk factor-based (NINDS)	All nonlacunar ischemic strokes‡	HBP, DM, smoking	167 (39)
Edinburgh38	1999	Consecutive admissions and neurovascular clinic outpatients	NR	NR	NR	Risk factor-free	Partial anterior circulation strokes (TACI and POCI excluded)	Ipsilateral CS	608 (215)
Erlangen31	2001	Community-based (13% of original cohort excluded because of incomplete investigation)	73	44	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, smoking	522 (120)
Germany28	2001	Admissions (multicenter)	68	58	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, TIA, smoking, alcohol, HC	4842 (1028)
Izmir32	1997	Consecutive admissions to stroke care unit	63	56	100%	Risk factor-based (TOAST)	All nonlacunar ischemic strokes	HBP, DM, TIA, smoking, HC, ipsilateral CS	1529 (198)
L’Aquila46	1998	Community-based (unknown No. of original cohort excluded because of lack of brain imaging)	74	53	100%	Risk factor-free	Nonlacunar ischemic strokes, excluding CE strokes or strokes of uncertain or mixed etiology	HBP, DM, smoking, HC, ipsilateral CS	393 (242)
London29	2001	Community-based	72	48	NR	Risk factor-free	All nonlacunar ischemic strokes	HBP, DM, AF, smoking, alcohol	862 (282)
Lund (a)41	1989	Consecutive admissions of minor stroke patients to neurology department who were examined using CT, brain, and catheter angiography	58	75	100%	Risk factor-free	Nonlacunar ischemic strokes excluding potentially CE strokes	HBP, DM, TIA, smoking, alcohol, HC, ipsilateral CS, contralateral CS	122 (61)
Lund (b)34	1994	Consecutive admissions with brain imaging, carotid ultrasound, and echocardiography	73	55	100%	Risk factor-free	All nonlacunar ischemic strokes	AF, ipsilateral CS	166 (49)
Maastricht (a)25	1991	Admissions, excluding infratentorial infarcts	70	NR	100%	Risk factor-free	Atherothrombotic infarcts (excluding CE strokes for all comparisons except AF)‡	HBP, DM, AF, ipsilateral CS, contralateral CS	247 (103)
Maastricht (b)26	1996	NR	NR	NR	NR	Risk factor-free	Nonlacunar ischemic strokes	HBP, DM	869 (287)
Manchester37	1998	Acute stroke admissions	NR	NR	56%§	Risk factor-free	All nonlacunar ischemic strokes (excluding POCI for CS comparisons)	AF, ipsilateral CS, contralateral CS	305 (80)
Oxford35	1990	Community-based	73	49	85%§	Risk factor-free	Nonlacunar carotid territory ischemic strokes	HBP, DM, TIA, AF, smoking	304 (102)
Riyadh24	1999	Consecutive admissions who had CT brain	59	68	100%	Risk factor-based	All nonlacunar ischemic strokes	HBP, DM, AF, smoking	756 (248)
Rochester42	1999	Community-based	76	41	92%#	Risk factor-based (NINDS)	All nonlacunar ischemic strokes‡	HBP, DM, TIA, smoking	442 (72)
Rome51	1995	Consecutive stroke patients hospitalized within 12 hours of onset of event	67	58	100%	Risk factor-free	All nonlacunar carotid territory ischemic strokes	HBP, DM, TIA, AF, smoking	517 (170)
San Diego43	1993	Consecutive admissions to and outpatients seen at hospital	61	66	98%#	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, TIA, smoking	448 (133)
Seoul (a)30	1999	Admissions who had a CT or MRI and catheter angiography	61	76	100%	Risk factor-based	Large vessel infarcts, excluding CE strokes‡	HBP, DM, smoking, alcohol	153 (49)
Seoul (b)33	2001	Admissions within 7 days of onset of event	62	63	100%#	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, TIA, smoking, HC	969 (215)
Taiwan50	1997	Consecutive admissions	66	57	Most	Risk factor-based (TOAST)	All nonlacunar ischemic strokes‡	HBP, DM, AF smoking, alcohol, HC, ipsilateral CS	637 (195)
Texas47	1991	Inpatients and outpatients seen in neurology department	61	75	100%	Risk factor-free	Nonlacunar carotid territory infarcts, excluding CE and unclassified infarcts	HBP, DM, TIA, smoking, ipsilateral CS, contralateral CS	109 (55)
USA27	1987	Admissions (multicenter)	68	47	97%#	Risk factor-based (NINDS)	All nonlacunar ischemic strokes	HBP, DM, TIA, AF	1273 (337)

^*Year of publication.

^†Total of 28 studies with 21 980 ischemic stroke patients included in the lacunar vs nonlacunar comparisons, and 5379 patients with lacunar ischemic stroke.

^‡Ischemic strokes with unusual causes excluded (generally<5% of ischemic stroke patients studied).

^§Siriraj score or Guy’s Hospital Stroke Diagnostic Scale was used to distinguish between ischemic and hemorrhagic stroke in the absence of brain imaging.

^¶% of the total No. of ischemic stroke patients included (shown in final column), except where otherwise stated.

^#% of total cohort studied (all ischemic strokes±hemorrhagic strokes).

NINDS indicates National Institute of Neurological Disorders and Stroke; TACI, total anterior circulation infarct, POCI, posterior circulation infarct; HBP, hypertension; DM, diabetes mellitus; HC, hypercholesterolemia; CS, carotid stenosis; NR, not reported; CE, cardioembolic.

Five studies (2522 patients) were community based, whereas 23 (19 458 patients) were hospital based, mostly recruiting hospital admissions (Table 1). The mean age of the patients was higher in the community than in the hospital-based studies (73 versus 66 years), and the proportion of males was higher in the hospital-based studies. Almost all patients studied had CT or MR brain imaging. Three studies excluded presumed cardioembolic stroke,^25,30,41 2 excluded infratentorial infarcts,^41,51 and 1 excluded total anterior circulation and posterior circulation strokes (Table 1).³⁸​ Table 2 shows the definitions given for the various risk factors studied.

TABLE 2.

Definitions of Risk Factors in Studies Included^*

Study	Hypertension		Diabetes		Smoking	Alcohol†	Raised Cholesterol (mmol/L)	Severe Carotid Stenosis
	Pre/post Stroke	Cutoff (mm Hg)	Pre/post Stroke	Cutoff (mmol/L)				Cutoff	Method of Measurement
Ankara40	Pre or post	160/90	Pre or post	Fasting >7.8 or random >11.1	Current smoking or within 5 years	ND	>5.7	ND	ND
Athens48	Pre	160/90	Pre	Fasting >6.0	Current daily smoking or within previous year	ND	>6.5	ND	ND
Barcelona36	Pre or post	160/90	Pre or post	Fasting >6.1	Current smoking or within 5 years	ND	>6.5 or triglycerides >1.71	ND	ND
Besançon39	NR	NR	NR	NR	NR	NR	NR	ND	ND
Boston45	Pre or post	140/90	NR	NR	>10 cigarettes/day	>2 units/day	NR	ND	ND
Buenos Aires44	Pre or post	140/90	Pre or post	Random >11	>10 cigarettes/day	>2 units/day	>5.7	ND	ND
Cincinnati49	Pre	NR	Pre	NR	Current smoking	ND	ND	ND	ND
Edinburgh38	ND	ND	ND	ND	ND	ND	ND	>70%	Ultrasound
Erlangen31	Pre or post	160/95	Pre or post	Fasting >6.7	Current smoking	ND	ND	ND	ND
Germany28	Pre	160/90	Pre or post	NR	Current smoking	Daily alcohol consumed	>5.7 before stroke or lipid-lowering medication	ND	ND
Izmir32	Pre	160/90	Pre	Fasting >6.0	Regular smoking	ND	>6.5	>50 %	Ultrasound, MRA or catheter angiography
L’Aquila46	Pre or post	150/90	Pre	Fasting >7.8	Daily smoking within at least last 2 months	≥2 units/day for women and ≥3 units/day for men	>5.7	>50%	Ultrasound
London29	Pre	160/95	Pre	Fasting >7.8	Ever smoked	ND	ND	ND	ND
Lund (a)41	Pre or post	160/90	NR	NR	NR	Alcohol abuse	NR	≥50%	Catheter angiography
Lund (b)34	ND	ND	ND	ND	ND	ND	ND	≥50%	Ultrasound
Maastricht (a)25	Pre or post	160/90	Pre or post	NR	ND	ND	ND	≥50%	Ultrasound
Maastricht (b)26	NR	NR	NR	NR	ND	ND	ND	ND	ND
Manchester37	ND	ND	ND	ND	ND	ND	ND	>70%	Ultrasound
Oxford35	Pre	160/90	NR	NR	Ever smoked	ND	ND	ND	ND
Riyadh24	Pre or post	160/90	Pre or post	Fasting >7.7	NR	ND	ND	ND	ND
Rochester42	NR	NR	NR	NR	NR	ND	ND	ND	ND
Rome51	NR	NR	NR	NR	NR	ND	ND	ND	ND
San Diego43	NR	NR	NR	NR	≥5 cigarettes/day at time of stroke	ND	ND	ND	ND
Seoul (a)30	Pre	NR	Pre	NR	Current smoking	>7.5 units/day	ND	ND	ND
Seoul (b)33	Pre or post	160/90	Pre or post	Fasting >6.1	Current smoking	ND	>5.7	ND	ND
Taiwan50	NR	NR	NR	NR	NR	NR	>6.2	≥50%	Ultrasound, MRA or catheter angiography
Texas47	NR	NR	NR	NR	NR	ND	ND	≥50%	Ultrasound
USA27	Pre	NR	Pre	NR	ND	ND	ND	ND	ND

^*Definitions for previous TIA and AF not shown.

^†Values converted from grams into United Kingdom units where necessary.

ND, no data, indicates that the risk factor concerned was either not studied or extractable data were not given in the publication(s); NR, not reported, even though the risk factor concerned was studied; MRA, magnetic resonance angiography.

Hypertension

A total of 25 studies (20 850 patients, 5034 with lacunar infarction) presented data on hypertension.^{24-33,35,36,39-51} Most defined hypertension on the basis of raised blood pressure before or after the stroke (Table 2). Overall, hypertension was commoner among patients with lacunar infarcts (pooled RR lacunar versus nonlacunar infarction, 1.22; 95% CI, 1.20 to 1.25; Figure 1). However, there was substantial statistical heterogeneity between results of the individual studies (χ²_24df=105.70; P<0.00001), partly arising from the different methods used to define ischemic stroke subtypes (Figure 1). The apparent excess of hypertension in lacunar infarction was confined to studies in which the presence of hypertension favored a diagnosis of lacunar infarction (pooled RR, 1.25; 95% CI, 1.21 to 1.28) and studies including risk factors other than hypertension in the definitions of ischemic stroke subtypes (pooled RR, 1.28; 95% CI, 1.23 to 1.34; Figure 1). Among studies defining ischemic stroke subtypes using a risk factor-free classification, the increased prevalence of hypertension among those with lacunar infarction was marginal (pooled RR, 1.11; 95% CI, 1.04 to 1.19), with no statistical heterogeneity between studies (Figure 1).

Figure 1.

RRs (lacunar vs nonlacunar) for hypertension, grouped by ischemic stroke subtype classification method. The RR for each study is shown as a square, with size denoting statistical weight of the study. Horizontal lines represent 95% CIs. Diamonds represent pooled RRs, with 95% CIs represented by the width of the diamonds. N indicates total number of lacunar or non-lacunar patients; n, number of lacunar or non-lacunar patients with hypertension; RR, relative risk; CI, confidence interval. Heterogeneity between four groups: χ²_3df=18.22; P=0.0004.

Diabetes Mellitus

A total of 25 studies (20 851 patients, 5035 with lacunar infarction) presented data on diabetes mellitus.^{24-33,35,36,39-51} Only 13 gave a clear definition of diabetes, generally comprising a history of diabetes before the stroke or raised blood glucose on admission (Table 2). There was a significant excess of diabetes in lacunar versus nonlacunar infarction among studies using a classification in which diabetes favors a diagnosis of lacunar infarction (pooled RR, 1.25; 95% CI, 1.17 to 1.34) and among studies using a classification including risk factors other than diabetes (pooled RR, 1.20; 95% CI, 1.09 to 1.32). However, among studies with a risk factor-free classification, there was no difference in the prevalence of diabetes in lacunar versus nonlacunar infarction (pooled RR, 0.95; 95% CI, 0.83 to 1.09; Figure 2).

Figure 2.

RRs (lacunar vs nonlacunar) for diabetes, grouped by ischemic stroke subtype classification method. Figure format as in Figure 1. N indicates total number of lacunar or non-lacunar patients; n, number of lacunar or non-lacunar patients with diabetes; RR, relative risk; CI, confidence interval. Heterogeneity between four groups: χ²_3df=18.22; P=0.0004.

Atrial Fibrillation

Fourteen studies (8087 patients, 2266 with lacunar infarction) presented data on AF.^{24,25,27,29,34,35,37,39,40,44,45,48,50,51} Overall, there was a stronger association between AF and nonlacunar than lacunar infarction (pooled RR for all 14 studies, 0.35; 95% CI, 0.30 to 0.40; Table 3). There was substantial heterogeneity between the individual study results (χ²_13df=85.08; P<0.00001), largely explained by different ischemic subtype classifications (Table 3). The association of AF with nonlacunar infarction was particularly pronounced among studies in which the presence of AF favored a diagnosis of nonlacunar infarction (pooled RR, 0.13; 95% CI, 0.09 to 0.19) and was less extreme for studies using a risk factor-free classification (pooled RR, 0.51; 95% CI, 0.42 to 0.62; Table 2).

TABLE 3.

Pooled RRs for Risk Factors Other Than Hypertension and Diabetes Studied, Grouped by Stroke Subtype Classification Method

Risk Factor	Stroke Subtype Classification	No. of Studies/Ischemic Strokes/Lacunar Infarcts RR (95% CI) Heterogeneity Statistics
AF	AF included in definitions	6/3668/989 0.13 (0.09 to 0.19) χ25df=7.03; P=0.22
	Other risk factors included in definitions	1/756/248 0.12 (0.05 to 0.29)
	Imaging-based	1/1262/243 0.59 (0.47 to 0.74)
	Risk factor-free	6/2401/786 0.51 (0.42 to 0.62) χ25df=11.96; P=0.04
	All groups combined*	14/8087/2266 0.35 (0.30 to 0.40) χ213df=85.08; P=<0.00001
Ipsilateral carotid stenosis	Stenosis included in definitions	2/2166/393 0.08 (0.03 to 0.25) χ21df=0.83; P=0.36
	Risk factor-free	7/1684/681 0.35 (0.28 to 0.44) χ26df=17.72; P=0.007
	All groups combined†	9/3850/1074 0.30 (0.24 to 0.37) χ28df=27.98; P=0.0005
Contralateral carotid stenosis	Risk factor-free	4/661/279 0.21 (0.11 to 0.41) χ23df=4.56; P=0.21
Smoking	Other risk factors included in definitions	15/14943/3153 1.06 (0.99 to 1.12) χ214df=48.97; P=<0.0001
	Imaging-based	1/1262/243 1.22 (1.02 to 1.46)
	Risk factor-free	6/2305/911 1.00 (0.92 to 1.08) χ25df=11.79; P=0.04
	All groups combined‡	22/18510/4307 1.05 (1.00 to 1.10) χ222df=65.16; P=<0.00001
Alcohol	Other risk factors included in definitions	5/6276/1486 1.11 (0.96 to 1.29) χ24df=4.42; P=0.35
	Imaging-based	1/1262/243 1.41 (0.98 to 2.04)
	Risk factor-free	2/984/343 0.95 (0.68 to 1.34) χ21df=0.27; P=0.60
	All groups combined¶	8/8522/2072 1.12 (0.98 to 1.27) χ27df=7.10; P=0.42
Previous TIA	Previous TIA included in definitions	8/12964/2546 0.88 (0.78 to 1.00) χ27df=7.03; P=0.43
	Imaging-based	1/1262/243 1.11 (0.83 to 1.48)
	Risk factor-free	4/1052/388 1.26 (0.97 to 1.64) χ23df=8.37; P=0.04
	All groups combined§	13/15278/3177 0.95 (0.86 to 1.05) χ212df=21.71; P=0.04
Raised cholesterol	Other risk factors included in definitions	9/12455/2492 1.26 (1.18 to 1.35) χ28df=27.98; P=0.0005
	Imaging-based	1/1262/243 1.20 (0.92 to 1.57)
	Risk factor-free	2/515/303 0.75 (0.58 to 0.97) χ21df=0.14; P=0.71
	All groups combined#	12/14232/3038 1.22 (1.15 to 1.30) χ211df=42.79; P=<0.0001

^*Heterogeneity between 4 groups: χ²_3df=66.1; P<0.00001

^†Heterogeneity between 3 groups: χ²_1df=9.4; P=0.002

^‡Heterogeneity between 3 groups: χ²_3df=4.4; P=0.1

^¶Heterogeneity between 3 groups: χ²_2df=2.4; P=0.3

^§Heterogeneity between 3 groups: χ²_2df=6.3; P=0.04

^#Heterogeneity between 2 groups: χ²_2df=14.7; P=0.0007.

Carotid Stenosis

Nine studies (3850 patients, 1074 with lacunar infarction) presented data on ipsilateral stenosis.^{25,32,34,37,38,41,46,47,50} Definitions of severe stenosis varied, but >50% or >70% stenosis on ultrasound were the commonest (Table 2). Four of the studies using risk factor-free classifications also gave data on contralateral stenosis.^25,37,41,47 Overall, there was an excess of ipsilateral carotid stenosis among patients with nonlacunar infarction. The association was more pronounced among studies in which severe carotid stenosis favored a diagnosis of nonlacunar infarction (RR, 0.08; 95% CI, 0.03 to 0.25) and was less extreme among studies using risk factor-free ischemic subtype definitions (RR, 0.35; 95% CI, 0.28 to 0.44). A similar result was observed for contralateral stenosis (RR, 0.21;95% CI, 0.11 to 0.41; Table 3).

Other Risk Factors

There was no clear association between smoking, excess alcohol consumption, or history of previous TIA and lacunar versus nonlacunar infarction, irrespective of the method used to define ischemic stroke subtypes (Table 3). Where given, definitions of smoking or alcohol excess varied considerably between studies (Table 2). The definition of raised cholesterol in most studies was based on the blood cholesterol concentration after stroke (Table 2). Although the overall pooled RR of 1.22 (95% CI, 1.15 to 1.30), suggested that raised cholesterol predisposes more to lacunar than to nonlacunar infarction, studies using risk factor-free ischemic subtype definitions found no definite association of raised cholesterol with lacunar versus nonlacunar infarction (Table 3).

Sensitivity Analysis

When we repeated our analyses including only community-based studies or studies that had recruited consecutive patients from hospital admissions and outpatient clinics, we found very similar results for all risk factors (data available from authors on request).

Discussion

Many studies have explored the association of various risk factors with different subtypes of ischemic stroke. Their pooled results must be interpreted carefully, with consideration given to various potential sources of heterogeneity.

Ischemic Stroke Subtype Classification Bias

The most important and striking difference was in the classification systems used to define ischemic stroke subtypes. Many studies included the risk factors being studied in their definitions of ischemic stroke subtypes, which may lead to bias (referred to hereafter as “classification bias”) when assessing differences in risk factor profiles between lacunar and nonlacunar ischemic strokes. The most appropriate classification system for investigating possible differences in risk factors between ischemic stroke subtypes should, ideally, be free of etiological assumptions about risk factors and so based solely on the clinical features of the stroke syndrome along with the appearances on brain imaging (ie, the site and size of the relevant lesion).⁸

Classification bias was of particular importance in the results for hypertension and diabetes. The apparent excess of hypertension and diabetes among lacunar versus nonlacunar infarction patients disappeared when only studies using risk factor-free classifications were considered.

Classification bias also affected the results for AF and carotid stenosis. The excess of AF and carotid stenosis among nonlacunar infarction patients was (unsurprisingly) more extreme among studies in which the presence of AF or carotid stenosis mitigated against a diagnosis of lacunar infarction. Emboli from the heart can occasionally occlude small, perforating cerebral vessels, and so it may be difficult to ascertain whether AF is causal or simply a manifestation of generalized vascular disease. Similarly, although carotid stenosis is more prevalent among nonlacunar infarcts, it does occur in association with some lacunar infarcts. The similarity of RRs for ipsilateral and contralateral carotid stenosis supports the concept that carotid stenosis is unlikely to cause most lacunar infarctions. However, the benefit from carotid endarterectomy among patients with lacunar infarction and severe carotid stenosis suggests that artery-to-artery emboli and low flow resulting from carotid stenosis may play an etiological role in some lacunar infarcts.⁵³

Variation Between Studies in Stroke Patient Population Studied

A further potential source of variability is in the population of patients studied. Most studies we identified were hospital based. The ideal study population would include all patients with an incident stroke in a particular geographical area (ie, a community-based stroke study), regardless of whether or not they attended hospital. A recent meta-analysis of community-based studies comparing the risk factor profiles of different ischemic stroke subtypes found only 4 such studies (and all but 1 small study including 102 patients that had not been published previously were included in our systematic review).⁵⁴ Community-based studies should avoid spurious differences in risk factor profiles between ischemic stroke subtypes arising because of hospital admission selection bias. But diagnosis of ischemic stroke subtype will first require appropriately timed brain imaging to exclude primary intracerebral hemorrhage, and patients managed entirely outside hospital are unlikely to access such imaging. Thus, for our purposes here, a series of patients recruited consecutively from outpatients as well as hospital admissions is unlikely to be any more biased than a community-based stroke register. Reassuringly, our results were essentially unchanged by a sensitivity analysis including only those studies based in the community or studying inpatient admissions and outpatients.

Potential selection bias could be reduced by using multivariate analysis methods to adjust the results for each risk factor for the confounding effects of other risk factors (eg, age, gender, etc). We could calculate only univariate associations because we did not have the individual patient data necessary for multivariate analyses. However, the few studies that performed both types of analyses found little difference between the results of univariate and multivariate analyses.^{25,26,29,35,46}

Variable Misclassification of Ischemic Stroke Subtypes

Another source of variation between studies using risk factor-based classification systems such as TOAST is the reliance on a number of investigations (such as carotid ultrasound, transcranial Doppler, echocardiography, etc) apart from brain imaging to allow assignment of an ischemic stroke subtype because access to these investigations is bound to vary between centers and according to patient characteristics such as age. Furthermore, the TOAST classification does not allow assignment of an ischemic stroke subtype when there is >1 potential cause of stroke, which occurred in 7% of ischemic strokes in a large, hospital-based stroke register.⁵⁵ In this case, or in the case of incomplete investigation, patients are placed in the “undetermined etiology” category. In the studies that used the TOAST classification, the proportion of patients in this category varied widely from 8%⁴⁵ to 41%.³³ This must be partly the result of variable access to diagnostic investigations but could also reflect inconsistent application of the TOAST criteria.⁵⁶ The large and variable proportion of patients in the “undetermined” subtype category (some of which will be lacunar, and others nonlacunar, in unknown proportions) introduces heterogeneity between the studies. A classification that can assign a stroke subtype to all (or at least almost all) ischemic stroke patients in the study population will be less prone to such heterogeneity, favoring classifications based mainly on clinical features of the stroke syndrome.⁸

However, even if all ischemic strokes are assigned a subtype, there will still inevitably be some misclassification. In particular, some lacunar infarcts will be misclassified as small cortical infarcts and vice versa⁵⁷ because the clinical features of the stroke syndrome alone are of limited accuracy in distinguishing these subtypes, and frequently, the relevant infarct is not visible on CT or MR brain scan. The extent of misclassification will depend partly on the proportion of patients with brain imaging and the type and timing of imaging used, which varied between the studies included. Small recent infarcts are more likely to be seen with diffusion-weighted MRI, but none of the studies we identified used this technique. The effects of misclassification of subtypes on risk factor associations are difficult to predict and will only be clarified by further studies using modern imaging techniques (including diffusion-weighted MR whenever possible) in large numbers of patients.

Variable Definitions of Risk Factors

Variability in the definitions used for the risk factors studied (Table 2) could also account for some of the heterogeneity between the results of different studies.

Summary

Our results suggest that the controversial assertion that hypertension and diabetes are particularly associated with lacunar infarction may arise almost entirely from classification bias. Hypertension and diabetes are risk factors for ischemic stroke in general, but their presence does not help to distinguish the ischemic stroke subtype. In addition, although AF and carotid stenosis are associated more with nonlacunar than lacunar infarction, this association is not as extreme as risk factor-based classification systems might suggest. Finally, there is no clear evidence of any association between smoking, previous TIA, excess alcohol consumption, or raised cholesterol and lacunar versus nonlacunar ischemic stroke subtypes.

Appendix A

Method		Description
Based on risk factors as well as clinical and brain imaging features	TOAST	Ischemic stroke classified into one of 5 categories. Large-artery atherosclerosis: Clinical findings include cortical, cerebellar, or brain stem dysfunction and on brain imaging cortical, cerebellar, brain stem or subcortical lesions >1.5 cm are considered to be of potential large-artery atherosclerotic origin. Diagnosis requires supportive evidence by duplex imaging or arteriography of >50% stenosis of an appropriate intracranial or extracranial artery. Potential sources of cardiogenic embolism, such as AF should be excluded, and a history of TIAs in the same vascular territory supports the clinical diagnosis. Cardioembolism: Clinical and brain imaging findings are similar to those described for large artery atherosclerosis. At least 1 cardiac source of embolism, such as AF, must be identified. Previous TIAs in >1 vascular territory supports the diagnosis. Potential large-artery atherosclerotic sources of thrombosis or embolism should be absent. Lacunar: Clinical findings of one of the lacunar syndromes should be present. Brain imaging should be normal or show a relevant brain stem or subcortical hemispheric lesion of diameter <1.5cm. A history of diabetes mellitus or hypertension supports the diagnosis. Potential cardiac sources of embolism, such as AF, should be absent, and the large extracranial arteries should not demonstrate >50% stenosis. Undetermined etiology: Includes patients with ≥2 potential causes of stroke (eg, AF and >50% stenosis of extracranial arteries). Other determined etiology: Includes patients with rare causes of stroke (eg, nonatherosclerotic vasculopathies and hematologic disorders.
	NINDS	Large-artery atherosclerosis: Clinical and brain imaging findings as described in TOAST. A history of TIAs is considered more common than among those with other types of stroke. Clinical diagnosis rests on finding evidence of arterial stenosis or occlusion at ≥1 sites. Cardioembolism: Clinical and brain imaging findings as described in TOAST. The basis for the clinical diagnosis is the demonstration of a cardiac-transcardiac source of embolism (such as AF) and no evidence of other causes of stroke. Lacunar: Clinical findings of a lacunar syndrome with normal brain imaging or relevant lesion. No mention of risk factors that specifically support a lacunar diagnosis. Undetermined etiology: Cerebral infarction in the absence of stenosis or occlusion of extracranial or intracranial arteries, cardiac source of embolism, or other demonstrable mechanism. Unusual cause: Includes patients with rare causes of stroke (eg, nonatherosclerotic vasculopathies and hematologic disorders.
Based on brain imaging only		Site and size of visible infarction on CT or MRI scan used to classify stroke subtypes irrespective of patient’s symptoms. Patients with a definite ischemic stroke but no visible infarct excluded.
Based on clinical features and brain imaging only	eg, OCSP	Clinical stroke syndrome (eg, TACI, PACI, LACI, and POCI) used to assign stroke subtype, then revised in light of site and size of any relevant infarct seen on CT or MRI scan.

NINDS indicates National Institute of Neurological Disorders and Stroke: Classification of Cerebrovascular Diseases III; TACI, total anterior circulation infarction; PACI, partial anterior circulation infarction; LACI, lacunar infarction; POCI, posterior circulation infarction.