Cerebral microbleeds (CMBs) are becoming increasingly of interest as a factor related to the pathophysiology of stroke and cerebral amyloid angiopathy. CMBs can be detected on gradient echo MRI, also known as T2* weighted MRI (T2* MRI), through the superparamagnetic effect of hemosiderin deposits.1 2 Their aetiology is not fully understood but histopathologically verified cases have shown either an association with amyloid microangiopathy, atherosclerosis with fibrohyalinosis or the coexistence of both.3

CMBs are associated with increased risk of stroke, especially in patients with non-traumatic intracerebral haemorrhage, where they may be present in 33–80% of cases.4^–10 They are also common in ischaemic stroke but the prevalence depends on patient age, stroke history and comorbidity in the brain.6 9 11^–13 In subgroups within the stroke populations, CMBs may be the consequence of familial amyloid angiopathy,14 CADASIL15 or Moyamoya disease.16 The few extant studies on patients with clinically diagnosed Alzheimer’s disease report a prevalence ranging from 18% to 27%.17^–19 Because CMBs can reflect amyloid deposition, their association with apolipoprotein (Apo) E allele has been investigated, but the findings are not consistent. Some studies suggest that the 4 is more prevalent, others suggest that the 2 is more prevalent and still others have shown that both 4 and 2 may be more prevalent in those with CMBs.20

Despite the emerging interest in CMBs in other than the mainly clinical stroke populations, there is scarce information on CMBs in community dwelling older populations. For example, little is known about the prevalence by age and sex, primary cerebral locations or their association with genetic markers. In the three prevalence studies of community dwelling subjects the reported prevalences were 3.1%, 4.7% and 6.4%.21^–23 Two of these studies22 23 suggest that the prevalence increases with age but mean age in the samples was only 64 years. A higher prevalence (9.8%) was reported by Tsushima et al in a large unselected clinic based population with a mean age of 56.6 years.8 There are no available data from community dwelling older cohorts, which are likely to be at greatest risk for developing CMBs.

This study presents the age and sex prevalence of CMBs in a large population based cohort of older men and women, with a mean age of 76 years, who participated in the Age Gene/Environment Susceptibility (AGES)-Reykjavik Study. We also investigated whether known risk factors for CMBs account for the sex and age related prevalence of CMBs.

MATERIALS AND METHODS

The study population

The AGES-Reykjavik Study is a single centre prospective study based on the Reykjavik Study. The Reykjavik Study was initiated in 1967 by the Icelandic Heart Association to study cardiovascular disease and risk factors. The cohort included men and women born between 1907 and 1935 who lived in Reykjavik at the 1967 baseline examination. Re-examination of surviving members of the cohort was initiated in 2002 as part of the AGES-Reykjavik Study. The AGES-Reykjavik Study is designed to investigate aging using a multifaceted comprehensive approach that includes detailed measures of brain function and structure. The study design has been described previously.24 Briefly, as part of a comprehensive examination, all participants answered a questionnaire, underwent a clinical examination and had blood drawn. All consenting participants without contraindications were offered a brain MRI on a dedicated machine in the study centre. Here we present data on the first 2300 participants examined from September 2002 to February 2004; this sample is generally representative of the original Reykjavik Study cohort.24 The study was approved by the Institutional Review Board (IRB, VSN 00-063) of Iceland, the Icelandic Data Protection Committee and the IRB serving the National Institute on Aging. All subjects signed an informed consent prior to participating in the study.

Of the first 2300 participants, 2020 had an MRI; the sequences needed to detect CMBs were available for 1962 participants. The reasons for missing MRI data for 14.7% (n = 338) of the sample were: standard MRI contraindications (implants/metallic foreign bodies (n = 99), claustrophobia (n = 88), could not lie supine in the scanner (n = 12), too ill (n = 9) and too large for the scanner gantry (n = 6)). Other reasons for missing MRI data included: scheduling because of MRI equipment maintenance (n = 60), only participated in home visits (n = 40), poor image quality for CMB reading (n = 6) and refusal (n = 18).

MRI acquisition protocol

MR images were acquired on a 1.5 T Signa Twinspeed system (General Electric Medical Systems, Waukesha, Wisconsin, USA). The image protocol consisted of the following pulse sequences: a T2* weighted gradient echo type echo planar (GRE-EPI) sequence optimised to maximise the susceptibility effect of hemosiderin (time to echo (TE) 50 ms; repetition time (TR) 3050 ms; flip angle (FA) 90°; field of view (FOV) 220 mm; matrix 256×256); a proton density/T2 weighted fast spin echo (FSE) sequence (TE1 22 ms; TE2 90 ms; TR 3220 ms; echo train length 8; FA 90°; FOV 220 mm; matrix 256×256); and a fluid attenuated inversion recovery (FLAIR) sequence (TE 100 ms; TR 8000 ms; inversion time 2000 ms; FA 90°; FOV 220 mm; matrix 256×256). These sequences were acquired with 3 mm thick interleaved slices. Additionally, images were acquired with a T1 weighted three-dimensional spoiled gradient echo sequence (TE 8 ms; TR 21 ms; FA 30; FOV 240 mm; matrix 256×256; slice thickness 1.5 mm). All images were acquired to give full brain coverage and slices were angled parallel to the anterior commissure–posterior commissure line.

Microbleed evaluation

CMBs were defined as a focal area of signal void within the brain parenchyma that: (1) is visible on T2* weighted GRE-EPI images and is smaller or invisible on T2 weighted FSE images (“blooming effect”); (2) is not abutting a parenchymal defect; and (3) does not show any other structure in the area of signal void. Using these criteria, CMBs can be differentiated from areas of signal void based on vascular flow voids (which do not show the blooming effect) (fig 1), past larger haematomas associated with parenchymal defects, and cavernomas, which are generally associated with areas of increased signal within the flow void. Areas of symmetric hypointensities of the globus pallidus likely to represent calcification or non-haemorrhagic iron deposits in the globus pallidus were excluded. These criteria are similar to those used in previous studies,2 4 23 except there was no size criterion for CMBs.

Images were viewed on a Dicom work station. The presence and number of CMBs, as well as the slice number of each CMB, was assessed and recorded by the neuroradiologist. This information was accessed by three radiographers trained to record the anatomical location and size of each CMB, up to 30 CMBs. In the case of very high numbers of CMBs, it was difficult to distinguish individual microbleeds because of coalescing, so the size and location of additional CMBs over 30 were not recorded (fig 2). Scored anatomical locations of CMBs included the cerebral lobes (frontal, parietal, temporal, occipital), external capsule, internal capsule, basal ganglia (putamen, globus pallidus, thalamus) and infratentorium (cerebellum, medulla oblongata, pons, mesencephalon). The size of each CMB was assessed on the T2* weighted GRE-EPI images by measuring the largest diameter of the lesion.

Reliability of assessing the presence or absence of microbleeds was 0.71 and 0.73 for the neuroradiologists compared with a neuroradiologist who, for quality control purposes, read the same scans in Leiden University Medical Centre. Intra-rater reliability (kappa) based on two ratings by one observer (neuroradiologist) who scored the majority of CMBs was 1.0 based on 19 brains read within 1 week.25

Apolipoprotein E genotyping

Apolipoprotein E genotype was determined by using standard DNA amplification and restriction isotyping.26 Subjects were classified by genotype and divided into four groups for the analysis: 22 and 23; 33; 34; and 44. The ApoE 24 genotype was excluded from the analysis.

Measure of risk factors

Putative risk factors for CMBs2 were investigated in relation to the age and sex prevalence of CMBs. Diabetes was defined when the participant gave a history, or was on glucose modifying mediation, or when fasting blood glucose was >7 mmol/l. Hypertension was defined as measured systolic blood pressure ⩾140 mm Hg, diastolic pressure ⩾90 mm Hg, self-reported doctor’s diagnosis of hypertension or using antihypertensive medications, assessed by medication vial brought to the study centre by the participant. A history of transient ischaemic attack (TIA)/stroke and smoking were assessed by questionnaire.

Statistical analysis

Logistic regression models were used to estimate the association of the presence of CMBs, as a binary variable, with age, sex and Apo E genotype. A general linear mixed model was used to determine if mean microbleed size was associated with age, sex or location (lobar vs other) among those with microbleeds. A random effect for subject was used to account for subjects with multiple microbleeds. Microbleed size was analysed on the square root scale to normalise the distribution of the variable. To investigate the extent to which the associations of age and sex to CMBs reflected factors co-varying these two variables, we added to the model hypertension, diabetes, smoking history and self-reported history of stroke or TIA. The level of significance was set at 0.05. SAS V.9.1 was used for the analysis.27

RESULTS

Prevalence

Evidence of CMBs was found in 11.1% (95% CI 9.8% to 12.6%; n = 218) of all subjects. Those with CMBs had a mean age of 77 (SD 6, range 67–92) years compared with 76 (SD 6; range 66–93) years for those without CMBs (table 1). The prevalence of CMBs significantly increased with age (p = 0.0001), both in men (p = 0.006) and women (p = 0.007) (fig 3). The prevalence of CMBs in men was significantly higher than in women (14.4% (95% CI 12.2% to 17.0%) vs 8.8% (95% CI 7.3% to 10.5%); age adjusted p = 0.0002). The median size of the CMBs was 6 mm; 9% were ⩾10 mm in diameter (fig 4). Microbleed size did not correlate with sex (p = 0.69), age (p = 0.12) or location (p = 0.17). These results did not change when microbleeds >10 mm were removed from the analysis. The size of the CMBs was positively associated with hypertension (p = 0.01) but no longer significantly when microbleeds >10 mm were removed (p = 0.11).

In this cohort, having a CMB was significantly associated with the homozygote ApoE 4 genotype (p = 0.01) (table 1). The prevalence of hypertension, diabetes, past and current smoking in the cohort is shown in table 1. After adding age and sex into the model, there was no association of CMBs with current or prior history of smoking. The associations of CMB presence with diabetes (p = 0.06) and hypertension (p = 0.07) were borderline significant and with stroke and TIA significant (p = 0.02). Adjusting for the above variables did not appreciably change estimated associations of CMBs with sex or age. Microbleed size was positively associated with hypertension (p = 0.01), but not with other risk factors.

Number and location

Among those with CMBs, 61% had one CMB, 18% had two CMBs and 9% had three CMBs. There were 12% with more than three but less than 22 CMBs; 1.4% had more than 30 CMBs. In 87% of subjects with ⩽30 CMBs, 61% had CMBs in the cortical or subcortical cerebral hemispheres, 6% had CMBs in the basal ganglia regions and 19% had lesions in the infratentorium. In 10% of subjects the CMBs were located in two of three regions, and in 3% they were spread across all three regions.

An analysis based on individual lesions (n = 447) showed that CMBs were most frequently located in the cerebral lobe. Over one-third of lesions were located in the posterior region (parietal and occipital lobes), and 24% were in the frontal lobe (table 2). The cerebellum was the second most common location of the CMBs (15.7%), followed by the basal ganglia region (10.4%). Out of 447 CMBs, 10.3% (n = 46) were either located in the cortical grey matter or on the border of the grey and white matter.

DISCUSSION

Based on a large community based cohort, we estimated the prevalence and location of CMBs and their relationship with age and sex, with and without adjusting for potential covariates, including hypertension, diabetes, history of smoking, self-reported history of TIA or stroke. CMBs were found to be common, significantly associated with age and more frequent in men than women. Homozygosity for Apo E 4 was also associated with the presence of CMBs.

The sex by age distribution is of some note: men 65–69 years of age had a higher prevalence than similarly aged women, but by age 85 years or greater, the prevalence was similar in men and women. Because this was a cross sectional study, it is not known whether this trend reflects selective mortality of men with CMB or an increase in the rate of CMB formation in women. The prevalence pattern is similar to sex differences in the prevalence of vascular disease.28 We did not find that adjustment for cardiovascular risk factors substantially altered age and sex associations, suggesting there are other risk factors for CMBs. More specific measures of cardiovascular risk factors are needed to investigate how much of the age and sex associations they explain.

Few CMB prevalence studies have been carried out on community based cohorts, and those that have were based on younger cohorts. In the ASPS22 (mean age ∼60 years), the prevalence of CMBs was 6.4%; there were no differences in prevalence between men and women.22 In the Framingham Study, the prevalence in participants (mean age 64 years) was reported to be 4.7%; in those older than 75 years the prevalence was 12.6%.23 In that study, men had proportionately more CMBs than women, as we found in our study. The difference in the distributions of the age of participants in these two cohorts compared with ours may explain some of the differences in prevalence. Other factors can also explain the differences across community based studies. The level of relevant comorbidity in the sample, such as hypertension, will influence prevalence rates. This is observed when comparing the prevalence of CMBs in community studies to the prevalence in clinical studies.

In addition to different samples, MRI hardware and software differences need to be considered when comparing rates across studies.2 The susceptibility induced signal loss on MR images, which reflects a CMB, depends on several MRI hardware and acquisition parameters,2 29 30 including variation in imaging parameters such as flip angle, slice thickness and slice gap, MRI magnet strength and echo (TE) times. Furthermore, the susceptibility artefacts from intracranial tissues will differ depending on airborne, air–brain and bone–brain interfaces.29 31

Differences across studies may also reflect differences in the size of the counted CMBs. For instance, the Framingham Study is based on lesions <10 mm23 and the ASPS is based on lesions 2–5 mm in size.22 In our study, which did not use size criteria, 65% of the CMBs were 6 mm or less (sample prevalence 7.2%) and 9% were >10 mm. It should be noted, the size of the lesion on the MR sequence does not reflect actual lesion size but rather the distortion of the magnetic field by the hemosiderin deposit; therefore, the amount of blood leaked from a vessel does not necessarily correlate with the severity of the pathology underlying the leakage. In our cohort, CMB size did not correlate with sex, age or location. A positive association between CMB size and hypertension was no longer found when lesions >10 mm were removed from the analysis, which may imply that hypertension could lead to larger CMBs. This will be further examined in the prospective study data.

It has been suggested that location in the cerebral lobes indicates an underlying amyloid angiopathy (CAA) while locations in the basal ganglia or the thalamus reflect hypertensive or atherosclerotic disease.5 In our study, CMBs were observed most frequently in the cerebral lobes, predominantly with a posterior hemispheric location with a relative over-representation in the parietal lobe. This localisation, plus the finding that Apo E 4 homozygotes are at increased risk for CMBs could indicate that the underlying pathology of these lesions is CAA. CAA is associated with the Apo E 4 and the posterior cerebral hemisphere has been shown to be the preferential site of cerebral amyloid deposition in a population based autopsy study32 and in Alzheimer’s disease, as shown by FDDNP-positron emission tomography studies.33 There is also recent evidence that amyloid plaques in Alzheimer’s disease might represent the sites of initial microbleeds.34 CMBs may therefore be implicated in CAA and Alzheimer’s disease. Future studies of risk factors for CMBs and their prognostic significance in the general population are needed to determine their aetiology and clinical implications.