Hypertension and Other Determinants of White Matter Lesions in Stroke Patients

Abstract

Hypertension is the main risk factor for both white matter lesions (WMLs) and stroke, but many stroke patients do not have WMLs. To find specific determinants of WMLs, the authors assessed 321 ischemic and hemorrhagic stroke patients who had undergone echocardiography. The patients with WMLs (n=160) were more often hypertensive and had a higher systolic blood pressure than the patients without WMLs. However, in a multivariate analysis, only the following variables remained associated with WMLs: (1) age: odds ratio [OR], 1.08 per year (95% confidence interval [CI], 1.06–1.11); (2) left ventricular relative wall thickness (RWT) ≥0.52: OR, 2.78 (95% CI, 1.59–4.88); (3) lacunar strokes: OR, 4.15 (95% CI, 1.83–9.44); (4) hemorrhagic strokes: OR, 5.36 (95% CI, 1.57–18.39); and (5) female: OR, 1.91 (95% CI, 1.12–3.27). Thus, the main modifiable risk factor for WMLs was RWT, which proved to be an even stronger risk factor than hypertension. This suggests that RWT might be a useful target in the treatment of hypertension to counteract the appearance of WMLs.

White matter lesions (WMLs), also known as leukoaraiosis, are detectable on brain computed tomographic (CT) scans as diffuse hypodensity of the white matter, while on fluid‐attenuated inversion recovery and T2‐weighed magnetic resonance imaging (MRI) they appear hyperdense, and are mainly located around the ventricles, especially close to their horns. White matter damage is caused by a diffuse impairment of the cerebral arterioles of the smallest caliber and, within small vessel disease, is often associated with lacunar lesions,1 which are caused by the occlusion of perforating arterioles of larger caliber.

In addition to unveiling a predisposition to stroke in general (and to lacunar strokes in particular),2, 3, 4 WMLs are frequently found in the presence of cognitive impairment, of which they are an important predictive factor.5, 6, 7

There are two main risk factors for WMLs: advanced age8, 9 and high blood pressure (BP).10, 11, 12 The latter in particular also predisposes patients to all types of stroke, both ischemic and hemorrhagic. In stroke patients, CT scan and MRI reports often describe chronic WMLs in addition to the acute lesion. On the other hand, in many hypertensive patients with stroke, WMLs are not present. Thus, some other risk factors, in addition to hypertension, might be more specifically predictive of cases of WMLs.

With this study, we wanted to search for the specific determinants of stroke‐associated WMLs, considering in particular traditional cardiovascular risk factors and the two main types of stroke: hemorrhagic and ischemic strokes, with the latter subdivided into the subgroups of the Trial of Org 10172 in Acute Stroke Treatment (TOAST) etiological classification.13 Moreover, because the heart, like the brain, is an important target organ of hypertension, we assessed the echocardiographic findings of acute stroke patients, comparing the main echocardiographic parameters between patients with and without WMLs.

Methods

Patients

A total of 321 consecutive patients were examined who had been admitted from April 20, 2013, to March 10, 2015, to the stroke unit of the Sant'Orsola‐Malpighi Hospital in Bologna, Italy. The patients had been admitted for an acute ischemic or hemorrhagic stroke, as defined by the appearance of a neurological deficit persisting for at least 24 hours and the finding of a new lesion of vascular origin on the second CT scan.

All patients had undergone transthoracic echocardiography, as well as the collection of risk factors, routine laboratory investigations, at least two brain CT scans, and carotid ultrasonography.

The patients with ischemic stroke were classified within the five TOAST etiological categories.13 Moreover, according to the second CT result, the patients were subdivided into two groups: (1) patients with WMLs (n=160) and (2) patients without WMLs (n=161). WMLs were graded within four levels of increasing severity, according to van Swieten14.

Because of the retrospective nature of the study, informed written consent could not be obtained. However, the collection of stroke patients' data and their processing for research purposes were approved by our hospital direction.

Risk Factors

The patients taking antihypertensive treatment, with an average systolic BP during hospitalization ≥140 mm Hg, or with an average diastolic BP ≥90 mm Hg were considered hypertensive. Average BPs (systolic and diastolic) were calculated using the first morning value on the day after admission, on the following day, and on the last day of their stay in the stroke unit. The patients taking antidiabetic treatment, or those with fasting blood glucose ≥126 mg/dL, were considered diabetic. The patients taking treatment with a statin, or with total cholesterol ≥200 mg/dL, were considered hypercholesterolemic. All patients underwent carotid ultrasound for the detection of carotid plaques, and the percentage of stenosis was defined according to the European Carotid Surgery Trial method.15

Brain CT Scan

Brain CT scans were performed by a LightSpeed scanner (GE Medical Systems, Milwaukee, Wisconsin, USA), software revision 07 MW11.10 (GE Healthcare, Waukesha, Wisconsin, USA). All CT scans were reassessed and agreed upon by two expert neuroradiologists.

WML severity was determined as follows.14 Three CT slices were examined: one through the choroid plexus of the posterior horns, one through the cella media, and one through the centrum semiovale. In these three slices, an anterior region and a posterior region were identified, and in each of the two regions, the maximum degree of extension of WMLs was recorded: 0=WMLs absent, 1=only periventricular WMLs, and 2=WMLs extending up to the cortex. By summing up the values of the two regions, an overall value ranging from 0 to 4 was then obtained.

Echocardiographic Parameters

Echocardiographic parameters were obtained by a GE Vivid S5 cardiovascular ultrasound system using the monodimensional, bidimensional, Doppler and color Doppler assessments.

The left atrial diameter was measured in M‐mode from the left parasternal view. The left ventricular mass (LVM) was calculated with the Devereux formula, according to the convention of the American Society of Echo‐cardiography16: LVM=0.8×(1.04×((LVIDD+PWTD+IVSTD)³ – LDIDD³)+0.6 g, where LVIDD is left ventricular internal diameter in diastole, PWTD is posterior wall thickness in diastole, and IVSTD is interventricular septum thickness in diastole.

The LVM index (LVMI) was calculated as the ratio between LVM (in g) and body surface area (BSA; in m²), which, in turn, was calculated with the Dubois formula17 starting from the weight and the height of the patient. The relative wall thickness (RWT) was calculated with the formula (IVSTD+PWTD)/LVIDD. The end‐diastolic and end‐systolic volumes, as well as the ejection fraction of the left ventricle, were calculated from the end‐diastolic and end‐systolic areas according to the formula of Simpson.

Statistical Analysis

The study variables were described with the mean and standard deviation when their distribution was normal; otherwise, they were described with median and interquartile interval. The differences between groups were tested with analysis of variance, Student t test for unpaired data, or Mann‐Whitney test, as appropriate. The differences between percentages were tested with chi‐square. The search for independent associations with WMLs was performed by multiple logistic regression and backward elimination procedure. Logistic coefficients and their standard errors allowed the calculation of odds ratios and 95% confidence intervals. Two‐tailed tests were performed throughout, and P values <.05 were considered significant.

Results

Table 1 shows the main characteristics and risk factors of the two groups of patients. The patients with WMLs were 11 years older than those without WMLs and were more often female. In addition, among patients with WMLs, there was a greater prevalence of hypertension, confirmed by the finding of higher mean values of systolic BP (by nearly 7 mm Hg). Diastolic BP was similar in the two groups, and there was no association between WMLs and other risk factors, such as smoking, diabetes, and hypercholesterolemia. Fina‐lly, among patients with WMLs, there was a higher prevalence of hemorrhagic strokes and carotid stenoses ≥50%.

Table 1.

General Characteristics and Risk Factors of the Two Groups of Patients

	White Matter Lesions Absent (n=161)	White Matter Lesions Present (n=160)	P Value
Age, y	65.2±14.3	76.5±11.1	<.0001
Male	94 (58.4)	64 (40.0)	.001
Hypertension	113 (70.2)	141 (88.1)	.0001
Average SBP, mm Hg	132.1±16.6	139.1±17.3	.0002
Average DBP, mm Hg	76.3±9.3	75.4±10.1	.43
Current smoker	39 (24.2)	30 (18.8)	.23
Ex‐smoker	36 (22.4)	32 (20.0)	.60
Diabetes	25 (15.5)	36 (22.5)	.11
Hypercholesterolemia	17 (10.6)	15 (9.4)	.72
Carotid stenosis ≥50%	62 (38.5)	79 (49.4)	.05
Hemorrhagic stroke	6 (3.7)	14 (8.8)	.06
TOAST classification
Large artery	21 (13.0)	21 (13.1)	.98
Cardioembolism	53 (32.9)	43 (26.9)	.24
Small artery	18 (11.2)	28 (17.4)	.11
Other cause	5 (3.1)	1 (0.6)	.10
Undetermined cause	58 (36.0)	53 (33.1)	.58

Abbreviations: DBP, diastolic blood pressure; SBP, systolic blood pressure; TOAST, Trial of ORG 10172 in Acute Stroke Treatment. Values are expressed as mean±standard deviation or number (percentage).

The echocardiographic parameters obtained in the two groups are reported in Table 2. The main factor characterizing the patients with WMLs was greater left ventricular wall thickness, especially RWT. The LVMI of the patients with WMLs was greater as well, although it was at a lower level of significance. In addition, WMLs were associated with a higher prevalence of aortic regurgitation and aortic stenosis.

Table 2.

Main Echocardiographic Parameters in the Two Groups of Patients

	White Matter Lesions Absent (n=161)	White Matter Lesions Present (n=160)	P Value
Left ventricular end‐diastolic volume, mLa	92 [75–110]	91 [74–112]	.70
Left ventricular end‐systolic volume, mLa	30 [23–40]	30 [23–39]	.40
Left ventricular ejection fraction, %a	65 [60–71]	67 [61–71]	.21
Left ventricular mean wall thickness, cma	1.11±0.20	1.18±0.19	.001
Relative wall thicknessa	0.46±0.10	0.51±0.10	<.0001
Left ventricular mass index, g/m2 a	107 [86–134]	119 [98–143]	.02
Left atrium, cma	4.1±0.7	4.3±0.8	.08
Mitral regurgitation	70 (43.5)	86 (53.8)	.07
Tricuspid regurgitation	56 (34.8)	64 (40.0)	.33
Aortic regurgitation	34 (21.1)	58 (36.3)	.003
Aortic stenosis	7 (4.3)	18 (11.3)	.02

Values are expressed as median [interquartile interval], mean±standard deviation, or number (percentage).

A few values were missing (on average 5 per group).

To establish which risk factors and echocardiographic parameters were independently associated with WMLs, a series of multiple logistic regressions was performed, with backward elimination of the nonsignificant variables. The initial model included the 16 variables that in Tables 1 and 2 were associated with WMLs and P values <.20 (age, sex, hypertension, average systolic BP, diabetes, carotid stenosis ≥50%, hemorrhagic stroke, small artery, other cause, mean left ventricular wall thickness, RWT, LVMI, left atrium, mitral regurgitation, aortic regurgitation, and aortic stenosis). All patients without missing data for the above variables contributed to the model (n=309, 155 without WMLs and 154 with WMLs). Table 3 lists the variables that remained significantly associated with WMLs at the end of the elimination procedure: age, small artery, RWT, hemorrhagic stroke, and female sex (men had an inverse relationship). The other variables, including hypertension and average systolic BP, were eliminated. This model explained 25% of WML variability (R ²=0.25, P<.0001).

Table 3.

Variables Independently Associated With White Matter Lesions

Variable	Coefficient β±SE	χ2	P Value
Age, y	0.079±0.013	39.5	<.0001
Small artery	1.467±0.417	12.4	.0004
Relative wall thickness	3.893±1.392	7.8	.005
Hemorrhagic stroke	1.673±0.623	7.2	.007
Male	−0.664±0.271	6.0	.01
Intercept	−7.504±1.135	43.7	<.0001

Results of a series of multiple logistic regressions, with backward elimination procedure of the variables that were not significantly associated with white matter lesions. The initial model (n=309, 155 without white matter lesions and 154 with white matter lesions) included the 16 variables that were associated with white matter lesions with P values <.20 in Tables 1 and 2.

To obtain clinically useful relative risk values, women were included in the model and RWT was dichotomized at the cutoff of 0.52. This value was chosen, after comparison with other possible cutoffs, because it was associated with the maximum χ² value: 75 of 109 patients had WMLs (69%) with RWT ≥0.52 vs 79 of 200 (40%) with RWT <0.52 (χ²=24.2; P<.0001). The logistic regression was repeated after these changes, and odds ratios with 95% confidence intervals were obtained from the beta coefficients and their standard errors (Table 4). Table 4 shows that the patients with WMLs had a fourfold risk of having a small artery (lacunar) stroke and a fivefold risk of having a hemorrhagic stroke, compared with patients without WMLs. In addition, Table 4 shows that each year of age (within the age range of our sample) could increase the risk of WMLs by 8%, and that women had an almost double risk of WMLs compared with men. Finally, the only potentially modifiable risk factor was an echocardiographic parameter, the RWT, which was associated with an almost threefold risk of WMLs for values ≥0.52.

Table 4.

Relative Risk of the Variables Independently Associated With White Matter Lesions

Variable	OR (95% CI)	χ2	P Value
Age, y	1.08 (1.06–1.11)	38.9	<.0001
RWT ≥0.52	2.78 (1.59–4.88)	12.8	.0004
Small artery	4.15 (1.83–9.44)	11.6	.0007
Hemorrhagic stroke	5.36 (1.57–18.39)	7.1	.008
Female	1.91 (1.12–3.27)	5.7	.02

Abbreviations: CI, confidence interval; OR, odds ratio. The variables shown are the same listed in Table 3, after dichotomization of relative wall thickness (RWT) at its best cutoff (0.52) and after conversion of sex from male to female.

Table 5 shows that there was no quantitative relationship between the three risk factors (age, female sex, and RWT) and the severity of WMLs according to van Swieten's scale. These variables differed significantly between patients without WMLs and even mild WMLs, but they did not differ between patients with mild WMLs and with moderate to severe WMLs.

Table 5.

Absence of Quantitative Relationship Between Risk Factors and Severity of White Matter Lesions According to van Swieten

Risk Factor	White Matter Lesions Absent (n=155)	White Matter Lesions 1 or 2 (n=82)	White Matter Lesions 3 or 4 (n=72)	P Value
Age, y	64.9±14.4	75.3±11.8a	77.9±10.1b	<.0001
Relative wall thickness	0.46±0.10	0.50±0.10b	0.51±0.10b	.0001
Female	66 (42.6)	50 (61.0)a	44 (61.1)b	.005

Values are expressed as mean±standard deviation or number (percentage). P values were obtained by analysis of variance or χ² test.

P<.0001 compared with absence of white matter lesions.

P=.002 compared with absence of white matter lesions.

^a P=.007 compared with absence of white matter lesions.

^b P=not significant compared with white matter lesions 1 or 2.

Discussion

This study has confirmed the important unfavorable prognostic significance of WMLs, which was found to be associated with lacunar and hemorrhagic strokes. As far as risk factors for the presence of WMLs are concerned, in addition to advanced age and female sex, which are not modifiable, the only independent factor was an echocardiographic parameter, ie, RWT, which was found to be a stronger risk factor than hypertension itself in multivariate analysis.

Relative Wall Thickness

It is widely known that hypertension is associated with WMLs10, 11, 12 and with all types of stroke, ie, ischemic and hemorrhagic. It would thus be useful, for both the understanding of the phenomenon and possible preventive applications, to try to understand what elements characterize the specific type of hypertension that is associated with WMLs in stroke patients. From this study, it is shown that WMLs are favored by hypertension associated with a specific target organ damage, ie, left ventricular hypertrophy with increased wall thickness (concentric hypertrophy). In fact, mean wall thickness, LVMI, and especially RWT (ie, the thickness related to the internal end‐diastolic diameter of the left ventricle) were strongly associated with WML. These associations cancelled out the other associations of WMLs with hypertension and average systolic BP in multivariate analysis.

Previously, in patients without stroke, Japanese authors found an association with left ventricular diastolic dysfunction, which was mainly caused by myocardial hypertrophy.18, 19 Moreover, we had already found high LVMI values in patients with small vessel disease (often without visible lesions on CT scan).20 In addition, in hypertensive patients, an association between silent WMLs and RWT had also been reported.21 However, the relevance of these echocardiographic parameters as risk factors for WMLs had never been demonstrated in stroke patients.

It is possible that the impairment of very small cerebral vessels underlying WMLs is the consequence of a continuous and prolonged, even if not very marked, hypertensive status capable of leading to both WMLs and concentric left ventricular hypertrophy over time.

Another hypothesis ascribes a prominent role to the stiffness of the aortic root that, in addition to hypertension, would favor upstream left ventricular hypertrophy and downstream WMLs, due to a more intense propagation of the pulse wave up to the smallest cerebral vessels.22, 23

The increase in myocardial wall thickness is not the only WMLs‐associated target organ damage. Retinopathy24 and chronic kidney disease25 have been found to be associated with WMLs. These associations were independent from hypertension, but they were not found to be stronger than the association between WMLs and hypertension.

Predisposition to Hemorrhagic Strokes

An increased risk of intracerebral hemorrhage in patients with WMLs had already been reported.26 In fact, most intracerebral hemorrhages, ie, those known as “typical,” are due to the rupture of a deeply located, small perforating artery. Typical hemorrhages are therefore the expression, like WMLs, of damage to small cerebral vessels. Thus, it seems reasonable that the two conditions may coexist.

In our study, the risk of hemorrhagic stroke was fivefold in patients with WMLs compared with patients without WMLs. With such an elevated hemorrhagic risk, some prudence is advisable when treating with thrombolytic therapy patients with ischemic stroke and WMLs. Indeed, two studies have reported an increased risk of post‐thrombolysis hemorrhagic transformation in these patients.27, 28 In addition, a less recent study reported an increased risk of intracerebral hemorrhage in patients with ischemic stroke and WMLs who underwent anticoagulation with warfarin.29

Age and Women

Advanced age is the most important risk factor for WMLs.8, 9 As previously stated, many years of hypertension must be present for concentric left ventricular hypertrophy, aortic root stiffness, and WML‐associated, small cerebral vessel damage to develop.

In epidemiological studies, women are frequently older than men; thus, many associations with women are not confirmed when age is included in a multivariate analysis. This was not the case in the present study. In a multivariate analysis, female sex was independently associated with WMLs, with an almost double relative risk compared with male sex. Thus, women would actually be more predisposed to WMLs then men, as already reported by others.30

Study Limitations

This study, like all retrospective studies, may have selection biases. For example, we selected patients who had undergone echocardiography. This investigation was performed in most of our patients, but not in all, so that cardioembolic and undetermined strokes were slightly overrepresented in our sample. However, this should not have influenced the distribution of the various types of stroke between patients with and without WMLs.

Age was the strongest factor associated with WMLs, and the age difference between patients with and without WMLs was considerable (11 years). Indeed, this large age spread suggests that other age‐associated factors, that were not considered in this study, might play a relevant role in the development of WMLs.

Another limitation comes from a study result. It is generally thought that a quantitative relationship should exist between causes and effects. Instead, we found no quantitative relationship between the three independent risk factors (age, female sex, and RWT) and the severity of WMLs according to van Swieten's definition. A more accurate method of WMLs quantification, such as MRI assessment of WML volume, might have shown this quantitative relationship.

Finally, this study showed that a specific target organ damage caused by hypertension (RWT) is more strongly associated with another target organ damage (WMLs) than hypertension itself. This apparent paradox is explainable considering that RWT may provide a better quantitative estimate of the hypertensive burden during a prolonged period compared with the simple diagnosis of hypertension or the average of three BP measurements during the stay in a stroke unit. In this regard, multiple follow‐up BP measurements, ambulatory BP monitoring, or the measurement of biochemical variables reflecting both hypertensive status and left ventricular hypertrophy (such as angiotensin and noradrenaline) might have led to different results.

Conclusions

WMLs are frequent in stroke patients, demonstrated in half of the patients in our study sample. In particular, this study confirmed the strong association between WMLs and lacunar strokes and highlighted that WMLs have an even stronger association with intracerebral hemorrhages. In addition, considering the predisposition to cognitive impairment characterizing patients with WMLs,5, 6, 7 the search for factors capable of predicting WMLs development is of considerable interest. M‐mode echocardiography provides a simple parameter, ie, RWT, which is associated with an almost triple relative risk of WMLs. This risk seems to be greater than the risk associated with hypertension. Controlling BP and checking that RWT remains within normal or low values could prove a useful strategy to avoid the development of WMLs. However, only a longitudinal study in patients without WMLs at baseline will be able to demonstrate, beyond the simple association, the ability of this parameter to predict WMLs and their potential fatal consequences.

Disclosures

The study was self‐financed: no sources of funding have to be acknowledged. The authors declare that they have no conflicts of interest.

J Clin Hypertens (Greenwich). 2016;18:907–912. DOI: 10.1111/jch.12788. © 2016 Wiley Periodicals, Inc.