MRI evidence of past cerebral microbleeds in a healthy elderly population

Subjects and methods.

Briefly, the ASPS is a prospective population-based single-center study on the effects of cerebrovascular risk factors on brain parenchyma and function.8 It has been approved by the local ethical committee. This cross-sectional study comprises individuals randomly selected from the official community register after stratification by sex and 5-year age groups. All subjects underwent a structured clinical interview, physical and neurologic examination, three blood pressure readings, EKG, echocardiography, and laboratory testing, including blood cell count and a complete blood chemistry panel. The inclusion criteria for the study were no history of neuropsychiatric diseases, including previous cerebrovascular attacks and dementia, and normal results of neurologic examination. We then invited every fourth individual of the baseline cohort of 1998 volunteers to participate in phase 2 of the study, which included MRI of the brain, Doppler sonography, SPECT, and neuropsychological testing. The overall rate of positive response into phase II was 89%, and MRI of the brain was obtained in 458 participants. After 3 years, we invited all of them to undergo a second MRI. Following our observation of incidental past microbleeds in patients with primary ICH,4 we extended the MRI protocol for the 3-year visit by a gradient-echo T2*-weighted sequence to increase the sensitivity for detecting hemosiderin deposits. This made MRI of 280 volunteers available for our study. The ages of these individuals (149 men, 131 women) ranged from 44 to 79 years (mean age 60 years).

Risk factor diagnosis was based on historical information and findings at examination, as described previously.9 In short, arterial hypertension was considered present if the subject had a history of arterial hypertension with repeated pressure readings >160/95 mm Hg, was treated for arterial hypertension, or had two readings at examination that exceeded this limit. A diagnosis of diabetes mellitus was based on current treatment or fasting blood glucose levels >140 mg/dL. Physical examination, the Rose questionnaire, and EKG findings served to establish cardiac disease comprising sources of cardiac embolism, coronary heart disease, and left ventricular hypertrophy.9

MRI was performed on a 1.5-Tesla superconducting magnet (Gyroscan S15 or ACS, Philips, Best, the Netherlands). The imaging protocol consisted of conventional spin-echo mixed intensity and T2-weighted (repetition time [TR]/echo time [TE] 2300–2600/20 and 90 msec) and gradient-echo T2*-weighted (TR/TE 600–800/16–20 msec; flip angle 20°) sequences in the transverse orientation. Slice thickness was uniformly 5 mm, and the interslice gap was 10%. All scans were read independently by three experienced investigators (F.F., R.S., P.K.). Gradient-echo T2*-weighted images served to outline focal areas with signal loss. They consisted of homogeneous rounded lesions with a diameter of 2 to 5 mm (figure 1, A through C). Areas of symmetric hypointensity of the globus pallidus, likely to represent calcification, were disregarded. White matter hyperintensities (WMH) were specified and graded according to our scheme into absent, punctate, early confluent, and confluent abnormalities.10 We disregarded caps and “pencil-thin” periventricular lining, because they represent normal anatomic variants.11 Areas of ischemic parenchymal destruction—that is, lesions exhibiting signal isointensity with CSF in the center—were diagnosed as lacunes (<10 mm in diameter) or infarcts. These interpretations were done on the conventional spin-echo images.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. (A) Gradient-echo T2*-weighted axial MRI (repetition time [TR]/echo time [TE] 620/16 msec; flip angle 20°) shows small foci of signal loss in the thalamus and putamen (arrows). (B) The thalamic lesion is barely visible on a conventional spin-echo sequence (TR/TE 2500/90 msec), and the putaminal lesion is not seen. Multiple small lacunes or enlarged Virchow-Robin spaces are present in the basal ganglia bilaterally. (C) Subcortical area of signal loss in the frontal region (arrow) of another participant in the Austrian Stroke Prevention Study.

We used the Statistical Package of Social Sciences for data analysis. Interrater agreement in interpretation of the various MRI abnormalities was determined by calculating the kappa coefficient for each pair of raters.12 Final lesion definitions relied on majority judgment or on a further consensus reading of all raters in case of complete disagreement. We used the chi-square test to compare the frequency distribution of categorical variables between subgroups of individuals with and without focal hypointensities. Continuous variables were compared with Student’s t-test. Because of the inequality of subgroup sizes, we did not perform more complex statistical analyses.

Results.

Gradient-echo T2*-weighted MRI of the brain showed focal areas of signal loss consistent with microbleeds in 18 of 280 (6.4%) volunteers. Their numbers ranged from one to five (mean 2.5) lesions, and most would have gone undetected on the conventional spin-echo sequence (see figure 1, A and B). The kappa coefficients of interrater agreement for presence of microbleeds ranged from 0.4 to 0.65. According to Fleiss,12 such values reflect fair to good agreement. Hypointense foci were seen in cortico-subcortical locations in 10 individuals (see figure 1C), in the basal ganglia and thalami in 6, and infratentorially in 3. In one, foci of signal loss were noted both cortico-subcortically and in the basal ganglia.

Other morphologic abnormalities of the brain consisted of WMH in 188 (67%) volunteers, lacunes in 23 (8%), and infarcts in 6. More extensive early confluent to confluent WMH and lacunes were significantly more frequent in patients with microbleeds (table 1). The kappa coefficients for WMH rating were between 0.65 and 0.68 and ranged from 0.38 to 0.6 for interrater agreement regarding the presence of lacunes. Table 2 shows the comparison of demographic variables and risk factors between individuals with and without microbleeds. Individuals with focal MRI hypointensities were significantly older and more often hypertensive, with significantly higher mean systolic and diastolic blood pressures. All seven volunteers with hypointensities in the basal ganglia and infratentorially, but only 5 of 10 individuals with foci of signal loss limited to cortico-subcortical regions, had arterial hypertension (p = 0.04). The mean ages of these subgroups were not significantly different.

Discussion.

We noted focal areas of signal loss on gradient-echo T2*-weighted MRI in 18 of 280 (6.4%) participants in a cross-sectional study of neurologically asymptomatic elderly volunteers. These lesions were seen in cortico-subcortical regions of the brain, in the basal ganglia, and in infratentorial locations. According to histopathologic correlation in subjects with ICH, these abnormalities are likely to represent hemosiderin deposits following earlier microbleeds.4,5,7 Individuals with MRI evidence of microbleeds were significantly older, had a significantly higher frequency of hypertension, and had a higher rate of early confluent to confluent WMH and lacunes. These associations support the role of microangiopathy in the pathogenesis of focal MRI hypointensities. Hypertension is a well-established cause of small vessel disease, and white matter damage and lacunes have been commonly reported to represent microangiopathy-related tissue damage.13-15 Foci of signal loss in cortico-subcortical regions were significantly less often associated with hypertension than microbleeds in the basal ganglia or infratentorial regions. Cortical and lobar ICH, predominantly of elderly patients, has been frequently related to amyloid angiopathy,16,17 and concomitant MRI evidence of past bleedings has been reported in this setting.5 The cortico-subcortical microbleeds in some of our elderly without hypertension may reflect the early stage of this disorder.

Potential causes of the observed foci of signal loss other than microangiopathy-related microbleeds are rare and likely to be excluded. Foci of dense calcification may have an MRI appearance similar to that of old microbleeds. Calcifications, however—although much more readily detectable in CT—have never been reported in CT studies of normal subjects except in the globus pallidum. Occult vascular malformations should also be considered in the differential diagnosis. Although a center of high signal intensity on both T1- and T2-weighted images is common for such lesions,18 very small malformations seen preferentially on gradient-echo T2*-weighted MRI may appear hypointense only. Foci of hemosiderin deposition can also stem from the rupture of minute vessels caused by shearing injury in the wake of severe head trauma,19 but individuals with such a history were not included in the ASPS.

So far, we can only speculate on the clinical significance of our observation. Besides an overall interest in better prediction of a person’s risk for intracerebral bleeding, some estimate on the likelihood of small vessel rupture would be most desirable for individuals who are taking drugs that affect blood clotting. More recently, a clinical trial of anticoagulation for secondary prevention in patients with transient ischemic attacks had to be terminated because of an unacceptably high rate of intracerebral bleeds.20 Clinically, hypertension is considered the prime risk factor for ICH, both spontaneously and following anticoagulation, but >50% of patients with ICH may be nonhypertensive.21 Evidence of leukoaraiosis on CT has also been associated with a higher risk of bleeding.20,22,23 In comparison, signs of past microbleeds on MRI appear to convey such information even more directly. Anecdotal support for this assumption comes from observation of a 59-year-old participant in the ASPS. Initially, the baseline MRI of this hypertensive woman had been considered abnormal only for a few punctate WMH. Five months later, she experienced an acute thalamic hemorrhage. In retrospect, an area of signal loss, probably consistent with a previous microbleed, was noted on the first MRI examination at the site of subsequent bleeding (figure 2, A and B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. (A) Baseline examination of a 59-year-old woman in the Austrian Stroke Prevention Study. Minute foci of signal loss in the right thalamus and left putamen (arrows) on conventional T2-weighted scans (repetition time/echo time 2500/90 msec). Five months later, this patient experienced an acute hemorrhage into the right thalamus, where the area of signal loss had been located.