Incident Cerebral Microbleeds Detected by Susceptibility Weight-Imaging Help to Identify Patients with Mild Cognitive Impairment Progressing to Alzheimer’s Disease

Abstract

Background:

The relationship between cerebral microbleeds (CMB) and Alzheimer’s disease (AD) has not yet been clearly determined, particularly with susceptibility weight-imaging (SWI).

Objective:

To evaluate the SWI sequence using 3T MRI for the detection of CMB, and its ability to differentiate elderly control subjects (CS), stable mild cognitive impairment patients (MCI-s), MCI patients progressing to AD (MCI-p), and AD patients.

Methods:

It was a prospective, monocentric, observational study that took place in Toulouse, France. Participants were 65 years and older, enrolled in three groups: CS, MCI, and AD. Based on the longitudinal analysis of cognitive decline, MCI subjects were retrospectively classified as MCI-s or MCI-p. Each patient had a 4-year follow-up with MRI at baseline (MRI#1) and during the fourth year (MRI#3). CMB were counted on native SWI images juxtaposed to minIP reformatted images.

Results:

150 patients were enrolled: 48 CS, 25 MCI-s, 18 MCI-p, 59 AD. At MRI#1 and at MRI#3, there was no significant difference in the prevalence of CMB between groups (p = 0.75 and p = 0.87). In the MCI-p + AD group, significantly more subjects had≥4 incident CMB compared to the CS + MCI-s group (p = 0.016). In the MCI-p + AD group, the prevalence of patients with >4 CMB was significantly higher at MRI#3 than at MRI#1 (p = 0.008).

Conclusion:

Using SWI, AD and MCI-p patients had developed significantly more new CMB than CS and MCI-s patients during the follow-up. Incident CMB might be suggested as a potential imaging marker of AD progression.

Keywords

Alzheimer’s disease disease progression imaging biomarker incident cerebral microbleeds longitudinal MRI mild cognitive impairment susceptibility weight imaging

INTRODUCTION

The diagnosis of probable Alzheimer’s disease (AD) dementia is based on clinical features. Biomarkers increase the certainty that the basis of the clinical dementia syndrome is the AD pathophysiological process [1, 2]. Biomarkers are even more important for disease identification at stages preceding dementia: mild cognitive impairment (MCI) due to AD or preclinical stage, before the occurrence of the first clinical symptoms [3]. Magnetic resonance imaging (MRI)-related morphologic changes, such as cortical thinning, sulcal enlargement, or hippocampal atrophy, are important but are not sufficiently specific for being used as diagnostic criteria of AD. This lack of specificity even led the International Working Group and the National Institute on Aging and the Alzheimer’s Association to remove these MRI findings from the diagnostic criteria of AD dementia [1, 2]. Thus, it could be relevant to study other potential MRI-biomarkers.

Cerebral microbleeds (CMB) correspond to perivascular collections of hemosiderin deposits [4, 5], which are secondary to the breakdown of hemoglobin. They result from a leakage through damaged cerebral small vessels [5 –7]. They can be associated with vascular risk factors such as older age and chronic hypertension [4 , 9]. In this clinical setting, they tend to be localized in the basal ganglia, brain stem, and cerebellum. Apart from hypertensive vasculopathy, CMB are associated with cerebral amyloid angiopathy (CAA) where they lay in the cortex or at the cortico-subcortical junction with a posterior predilection [4 –7]. In CAA, CMB result from a weakening of blood vessel walls due to amyloid-β (Aβ) deposition [7]. Although CAA is found in the vast majority of AD patients [6–8 , 10–12], the relationship between CMB and AD has not yet been clearly determined.

On MRI, CMB are detected as small, rounded foci of hypointensity on T2*-weighted Gradient Echo (GRE) sequences. This MRI sequence reveals more CMB in AD patients as compared to normal age-matched controls [6 , 13–18].

Susceptibility weight imaging (SWI) is an MRI sequence that was developed several years ago and is becoming increasingly used in daily clinical practice. It maximizes the sensitivity of susceptibility effects, and thus, the detection of CMB [18 –20]. Detection can also be improved by using higher magnetic field strength, such as with 3T MRI scanners. Though SWI improves the detection of CMB, data on the detection of CMB with SWI on 3T MRI in AD patients are scarce.

There is also a lack of data on the clinical relevance of this improved CMB detection in MCI and AD patients. Furthermore, the relationship between CMB, cognitive decline, and patient outcomes is still unclear, as some studies found that AD patients with CMB are more severely cognitively impaired [6 , 22], whereas other studies found no such relationship [10 , 23]. Finally, the association between CMB and progression of MCI patients to AD dementia is not established either, as one study found such association (using SWI) [28], while other studies did not (both using T2*-weighted GRE sequences) [6, 24].

The aim of the study was therefore to evaluate the SWI sequence using 3T MRI for the detection of CMB and its ability to differentiate elderly control subjects (CS), stable MCI patients (MCI-s), MCI patients progressing to dementia (MCI-p) and AD patients. Other imaging parameters, such as lacunes and white matter hyperintensities (WMH) were also assessed to determine if they could help differentiate between patient groups.

METHODS

Population

The study population was recruited from the ROSAS study, which was described in a previous article [25]. Briefly, it was a prospective, monocentric, observational study that took place in the memory clinic setting of Toulouse University Hospital, France. This study was designed to evaluate the biomarkers in AD, MCI patients, and elderly controls. There was a 3- to 4-year follow-up for each patient, with annual cognitive examinations, biological tests, and longitudinal MRI.

Participants were men and women, aged 65 years and older, enrolled in three groups as follows:

Patients with AD met revised Alzheimer’s dementia criteria of Diagnostic and Statistical Manual version IV (DSM-IV-TR), had a Mini-Mental State Examination ((12≤MMSE <26) score), and a global Clinical Dementia Rating Scale (CDR) score of 0.5 or higher (mild to moderate stage);

MCI patients did not meet diagnostic criteria of dementia due to AD of DSM-IV-RT, had a memory disorder detected by the Rey Auditory Verbal Learning test (RAVLT) (<1 standard deviation (SD) of the age-adjusted mean), an MMSE score of 24 or higher, and a CDR global score of 0.5. MCI subjects were retrospectively classified as MCI-s if they remained MCI until the end of the study, or MCI-p if they progressed to AD dementia according to the above-mentioned criteria during the study;

Control subjects (CS) were those without memory complaint at the time of the interview and had no memory impairment detected by the RAVLT (value within±1 SD of the age-adjusted mean), had an MMSE score of 26 or higher, and a global CDR score of 0.

Participants unable to speak or write French, under legal protection, subjects with brain tumor, stroke or other neurologic diseases that may explain their cognitive deficit (e.g., Parkinson’s disease, multiple sclerosis, epilepsy), with a diagnosis of vascular dementia according to the NINDS-AIREN criteria [26] or other types of dementia, with serious illness, or participating in a clinical trial were excluded.

Participants and their formal caregiver took part in the study on a voluntary basis and gave their written informed consent at selection. The Ethics Committee of Toulouse University Hospital approved the study protocol and all its amendments.

At baseline and at every follow-up visit, face-to-face interviews were held for neuropsychological evaluation of study participants by a trained neurologist. Clinical examinations of study participants were done twice a year for patients with AD dementia or MCI, and once a year for CS.

Demographics (level of education, age, gender, and living arrangements) and medical characteristics (risk factors, cardiac, vascular and psychiatric diseases, pharmacological treatments) were recorded for all participants. Clinical assessments of cognition consisted of MMSE, global CDR score, ADAS-Cog 11, RAVLT, and Trail Making Test (TMT) A and B. Physical impairment evaluation was based on Alzheimer’s Disease Cooperative Study-Activities of Daily Living (ADCS-MCI-ADL) scale administration. Neuropsychiatric symptoms (NPS) were measured by total and individual items of the Neuropsychiatric Inventory (NPI-12) scale.

Genetic data

ApoE ɛ4 genotyping was performed for all participants and carrier status was defined as the presence of one or both ApoE ɛ4 alleles (“carrier” status) or the absence of the ApoE ɛ4 allele (“non-carrier” status).

MRI data collection

According to the ROSAS study protocol, MRI was not compulsory for all patients. The sub-population of participants who accepted to undergo MRI scans is described in the present study (see Table 1).

Table 1

Population’s characteristics

		CS	MCI-s	MCI-p	AD	p-value
Number		48	25	18	59
Age
	Mean±SD	72.875±5.931	77.840±4.571	76.889±4.922	78.407±6.240	<0.0001
Gender
Female	n (%)	24 (50.00)	11 (44.00)	13 (72.22)	42 (71.19)	0.03
Education Level	Mean±SD
		10.000±4015	8.480±3819	8.889±4599	7.983±3692	0.06
ApoE4 Carrier
Yes	n (%)	14 (29.17)	7 (28.00)	5 (27.78)	34 (57.63)	0.006
Hypertension
Yes	n (%)	21 (43.75)	12 (48.00)	9 (50.00)	31 (52.54)	0.84
Cholesterol Total	Mean±SD (mmol/l)
		5.408±879	5.320±978	5.800±1363	5.532±974	0.62
Diabetes
Yes	n (%)	4 (8.33)	2 (8.00)	1 (5.56)	9 (15.25)	0.52
Smoking Habit
No	n (%)	29 (60.42)	16 (64.00)	13 (72.22)	48 (81.36)	0.35
Has stopped	n (%)	17 (35.42)	8 (32.00)	4 (22.22)	10 (16.95)	.
Yes	n (%)	2 (4.17)	1 (4.00)	1 (5.56)	1 (1.69)	.

CS, control subjects; MCI-s, stable mild cognitive impairment patients; MCI-p, progressor mild cognitive impairment patients; AD, Alzheimer’s disease patients.

MRI scans were acquired up to 3 times during the study: at Baseline (MRI #1), during the second year (MRI #2), and at the last visit during the fourth year (MRI #3). Data were sent to an Imaging Core Lab (Bioclinica) for quality control and centralized analysis. They were acquired at a single site using a Philips Achieva 3T MRI scanner (Philips Medical Systems, Best, The Netherlands). The MRI protocol was composed of the following sequences: sagittal 3D T1-weighted, sagittal 3D T2-weighted, axial FLAIR, axial SWI and axial diffusion weighted, and diffusion tensor imaging (DWI and DTI).

SWI protocol used 1-mm thick slices reconstructed every 0.5 mm, a 0.81×1.00 mm in-plane resolution and TR/TE settings of 21.5/15 ms. In addition to those native images, Minimum Intensity Projection (minIP) reconstructed slices were centrally generated (6 mm thick, with 3 mm overlap).

FLAIR protocol used 4 mm slices with no gap, a 0.78×0.64 mm in-plane resolution and TR/TE/TI settings of 11000/125/2800 ms.

3D T1 data consisted of a sagittal 3D TFE sequence with 1-mm thick slices, a 1.46×1.00 in-plane resolution and TR/TE settings of 9.9/4.6 ms.

Only subjects with good quality scans (no significant imaging artifacts) were considered for analysis.

Image analysis

SWI data were reviewed by one reader (HB), blinded to the clinical status and time of acquisition. The reader looked for CMB on native SWI images juxtaposed to minIP reformatted images.

CMB were defined on SWI as a round area of signal loss, less than or equal to 10 mm in diameter. Hypointense signal areas due to calcification in the globus pallidus and flow void artifacts in the cerebral pial vessels were carefully excluded [27]. CMB were further categorized by location as follows: Deep: deep grey matter (basal ganglia and thalamus), internal, external, and extreme capsules; Lobar: cortical grey matter and subcortical white matter; and Infratentorial: brain stem and cerebellum.

FLAIR images were also reviewed by the same reader, still blinded to clinical status and time of acquisition. Supratentorial and infratentorial lacunes were counted.

White matter hyperintensities (WMH) were semi-automatically delineated on FLAIR data and assessed in order to determine their volume. This volume was then normalized by the intracranial volume, automatically measured on the 3DT1 sequence.

MRI findings for each group were subsequently compared at baseline (MRI #1) and at study end (MRI #3) to look for potential changes over the 3-year follow-up period.

Statistical analysis

Descriptive statistics are presented as mean±standard deviation (SD) or median (Q1; Q3) for continuous and count variables, and as numbers and percentages for categorical variables. For demographic data, groups were compared using a Kruskal-Wallis test for continuous variables and a χ² test for categorical variables.

The prevalence of CMB and lacunes was compared among groups using a Logistic Regression adjusted for confounding factors (age, ApoE ɛ4, sex, and education level). p-values correspond to Type 3 tests.

The comparison of the number of CMB among groups was performed using a Poisson Regression adjusted for confounding factors.

The incidence of CMB between MRI #1 and MRI #3 was compared among groups using Exact-Tests from a Logistic Regression, non-adjusted for confounding factors.

The comparison of the percentage of patients with strictly more than 4 CMB between MRI #1 and MRI #3 was performed with a McNemar’s test and the comparison among groups at a given visit using a Logistic Regression adjusted for confounding factors.

The volumes of WMH were compared among groups using a non-parametric approach (Wilcoxon test, unadjusted for confounding factors) and a parametric approach (ANCOVA, adjusted for confounding factors after log-transformation of data).

p-values correspond to overall comparison among groups and no multiplicity adjustment was applied in case multiple group comparisons were performed on a given variable.

RESULTS

Demographic data

The characteristics of the population are listed in Table 1.

As described earlier, only a subset of participants of the ROSAS study participated in the MRI sub-study [25].

A total of 150 patients were enrolled in the study: 48 CS, 25 MCI-s, 18 MCI-p, and 59 patients with dementia due to AD.

The four groups significantly differed in terms of age (p < 0.0001). The AD and MCI patients were older than CS., however the difference between MCI-p and MCI-s was not so clear.

We also observed differences in gender (p = 0.03) with more women enrolled in the AD and MCI-p groups.

There were more ApoE ɛ4 carriers among patients with dementia due to AD than in other groups (p = 0.006 for the difference among the 4 groups).

There was no difference in cardiovascular risk factors between the groups.

Of the enrolled patients, 145 underwent MRI #1 (47 CS, 24 MCI-s, 18 MCI-p, 56 AD), 115 underwent MRI #2 (43 CS, 17 MCI-s, 16 MCI-p, 39 AD), and 84 underwent MRI #3 (35 CS, 13 MCI-s, 13 MCI-p, 23 AD).

Prevalence of CMB

Number and percentage of patients with at least 1 CMB is shown in Table 2.

Table 2

Number, prevalence and median number of CMB by subject at MRI #1 and #3

	CS	MCI-s	MCI-p	AD	p
Number and prevalence	29 (62%)	30 (71%)		38 (68%)	p = 0.55
of CMB at MRI #1	29 (62%)	17 (71%)	13 (72%)	38 (68%)	p = 0.75
Number and prevalence	28 (80%)	21 (81%)		19 (86%)	p = 0.80
of CMB at MRI #3	28 (80%)	11 (85%)	10 (77%)	19 (86%)	p = 0.87
Median number of CMB	1 (0; 2)	1 (0; 4)		1 (0; 2)	p = 0.11
by subject at MRI #1	1 (0; 2)	2 (0; 5.25)	1 (0.25; 3)	1 (0; 2)	p = 0.14
(Q1;Q3)	1 (0; 2.5)		1 (0; 2)		p = 0.83
Median number of CMB	2 (1; 2)	2 (1; 7.75)		3 (1; 5)	p < 0.001
by subject at MRI #3	2 (1; 2)	2 (1; 4)	2 (1; 14)	3 (1; 5)	p < 0.001
(Q1;Q3)	2 (1; 2)		3 (1; 7.5)		p = 0.009

CMB, cerebral microbleeds; CS, control subjects; MCI-s, stable mild cognitive impairment patients; MCI-p, progressor mild cognitive impairment patients; AD, Alzheimer’s disease patients.

Based on our results that showed many similarities between MCI-p and AD groups, and on the clinical hypothesis that MCI-p and AD patients represent the same pathological process at earlier (MCI) or later (AD) stages, these groups of patients were unified for analysis. This new group was called MCI-p + AD, and was compared to the other combined group called CS + MCI-s.

At MRI #1 and at MRI #3, there was no significant difference in the prevalence of CMB between groups (p = 0.75 and p = 0.87).

At MRI #1, the median number of CMB was not different between the groups (p = 0.11).

At MRI #3, the median number of CMB was significantly higher in the MCI-p + AD group as compared to the CS + MCI-s group (p = 0.009).

Incidence of CMB

Number and percentages of patients with at least one new CMB detected between MRI #1 and MRI #3 is provided in Table 3.

Table 3

Number and percentage of patients with incident CMB between MRI #1 and MRI #3

	CS	MCI-s	MCI-p	AD	p
≥1 incident CMB	17 (50%)	14 (53.8%)		12 (54.5%)	p = 0.90
	17 (50%)	7 (53.8%)	7 (53.8%)	12 (54.5%)	p = 0.99
≥2 incident CMB	9 (26.5%)	11 (42.3%)		9 (40.9%)	p = 0.40
	9 (26.5%)	5 (38.5%)	6 (46.2%)	9 (40.9%)	p = 0.60
≥3 incident CMB	3 (8.8%)	8 (30.8%)		7 (31.8%)	p = 0.06
	3 (8.8%)	2 (15.4%)	6 (46.2%)	7 (31.8%)	p = 0.02
≥4 incident CMB	1 (2.9%)	7 (26.9 %)		5 (22.7%)	p = 0.02
	1 (2.9%)	2 (15.4%)	5 (38.5%)	5 (22.7%)	p = 0.02
3 (6.4%)		10 (28.6%)		p = 0.016

CMB, cerebral microbleeds; CS, control subjects; MCI-s, stable mild cognitive impairment patients; MCI-p, progressor mild cognitive impairment patients; AD, Alzheimer’s disease patients.

An example of an incident CMB is also illustrated in Fig. 1.

Fig.1

Incident cerebral microbleeds (CMB) in the periventricular white matter. Brain MRI with Axial SWI sequence, using Minimum Intensity Projection (minIP) reconstructed slices. This figure shows an image of MRI #1 (left) and MRI #3 (right), in the same patient, at the same level. These images illustrate the formation of a periventricular CMB (arrow) between MRI #1 and MRI #3.

There was no significant difference between groups in the incidence of 1 or more CMB between MRI #1 and MRI #3 (p = 0.99). But interestingly, in the MCI-p + AD group, significantly more subjects had 4 or more incident CMB than the CS + MCI-s group (p = 0.016).

With this threshold of 4 or more incident CMB, sensitivity and specificity for detecting MCI-p or AD patients (compared to CS and MCI-s patients) were 28.6% and 93.8% respectively, with a positive predictive value at 76.9% and a negative predictive value at 64.3%.

Furthermore, sensitivity and specificity for detecting MCI-p patients compared to MCI-s patients were 38.5% and 84.6%, respectively, with a positive predictive value at 71.4% and a negative predictive value at 57.9%.

Table 4 indicates the number and percentages of patients with more than 4 CMB at MRI #1 and at MRI #3.

Table 4

Number and prevalence of patients with more than 4 CMB

	CS + MCI-s	MCI-p + AD	p¹
Patients >4 CMB at MRI #1	8 (11.3 %)	10 (13.5 %)	p = 0.92
Patients >4 CMB at MRI #3	7 (14.6 %)	13 (37 %)	p = 0.20
p²	p = 0.063	p = 0.008

p¹: Comparison of the percentage of patients with >4 CMB between CS + MCI-s group and MCI-p+: AD group at MRI 1 and at MRI 3. p²: Comparison of the percentage of patients with >4 CMB between MRI 1 and MRI 3 in each groups. CMB, cerebral microbleeds; CS, control subjects; MCI-s, stable mild cognitive impairment patients; MCI-p, progressor mild cognitive impairment patients; AD, Alzheimer’s disease patients.

There was no significant difference in the prevalence of patients with >4 CMB between the groups at MRI #1 and at MRI #3 (p > 0.05).

In the MCI-p + AD group, the prevalence of patients with >4 CMB was significantly higher at MRI #3 than at MRI #1 (p = 0.008).

Lacunes and volume of white matter hyperintensities

The number and prevalence of patients with at least one lacune and the volume of WMH are listed in the Table 5.

Table 5

Number and prevalence of lacunes at MRI #1 and MRI #3

	CS	MCI-s	MCI-p	AD	p
Number and prevalence of lacunes at MRI #1	6 (13%)	11 (29%)		13 (25%)	p = 0.36
	6 (13%)	5 (23%)	6 (38%)	13 (25%)	p = 0.34
Number and prevalence of lacunes at MRI #3	7 (20%)	10 (38%)		7 (30%)	p = 0.43
	7 (20%)	5 (38%)	5 (38%)	7 (30%)	p = 0.56
Median volume of WMH at MRI #1	0.11	0.47		0.66	p < 0.01
	0.11	0.3	0.57	0.66	p < 0.01
Mean volume of WMH at MRI #1	0.25±0.34	0.83±0.94		1.06±1.18	p = 0.045
	0.25±0.34	0.73±0.93	0.96±0.97	1.06±1.18	p = 0.047
Median volume of WMH at MRI #3	0.2	0.31		0.68	p = 0.026
	0.2	0.25	0.9	0.68	p = 0.032
Mean volume of WMH at MRI #3	0.35±0.42	0.84±1.03		1.37±1.49	p = 0.86
	0.35±0.42	0.51±0.73	1.2±1.22	1.37±1.49	p = 0.65
Increase in median volume WMH	0.05		0.2		p = 0.002
between MRI #1 and MRI #3
Increase in mean volume WMH	0.09±0.11		0.48±0.62		p = 0.45
between MRI #1 and MRI #3

Mean and median volume (in square centimeter) of WMH at MRI #1 and at MRI #3, and increase in mean and median volume of WMH between MRI #1 and MRI #3. WMH, white matter hyperintensities; CS, control subjects; MCI-s, stable mild cognitive impairment patients; MCI-p, progressor mild cognitive impairment patients; AD, Alzheimer’s disease patients.

At MRI #1 and MRI #3, there was no significant difference in the prevalence of lacunes among the groups (p = 0.34 and p = 0.56).

At MRI #1, the median volume of WMH was significantly different among groups, with a higher lesion load in AD and MCI-p (p < 0.01). After adjustment for confounding factors, this difference remained significant (p = 0.047).

At MRI #3, there was a significant difference in median volume of WMH between groups, with greater volumes found in AD and MCI-p (p = 0.032), though this finding disappeared following adjustment for confounding factors (p = 0.65).

After adjustment, mean increase in volume of WMH did not differ between groups either (p = 0.45).

DISCUSSION

The main finding of the study is the higher incidence of CMB in AD and MCI-p patients, as compared to CS and MCI-s patients.

Using SWI, we have demonstrated that it is common to see one or two incident CMB during a 4-year follow-up period. We also showed that 4 or more incident CMB are correlated to AD pathology or clinical progression of MCI patients to AD dementia. Thus, it could be relevant to monitor the number of CMB on longitudinal MRI: a rapid increase in the number of CMB could be a potential AD biomarker or predict progression to AD. These results are supported by studies that found a higher incidence of CMB in AD patients [28] and MCI-p patients [29] and by studies that found a correlation between the presence of CMB and the severity of cognitive impairment [6 , 22].

The other results support this hypothesis. At MRI #1, the median number of CMB was the same between the different groups, but it was higher at MRI #3 in MCI-p and AD patients, confirming that CMB appear faster for those patients.

In this study, we did not find any difference in the prevalence of CMB in the different groups. Others studies, using T2*-weighted GRE sequences, found that the prevalence of CMB was higher in AD [6 , 13–18] and in MCI patients [8 , 18] than in CS. It could be the consequence of the greater sensitivity of SWI [18 –20] that makes the detection of one CMB a normal occurrence in people over 75 years of age, with no clinical implication. It is still to be discussed as some studies found a greater prevalence of CMB, with SWI, in MCI and AD patients than in CS [28, 29].

We also showed that the differences between groups were more significant when CS were grouped together with MCI-s and when MCI-p were grouped together with AD. This suggests that MCI-p subjects already share many characteristics with AD patients, especially on incident CMB, median number of CMB, and volume of WMH. Thus, in MCI patients, incident CMB, detected with a longitudinal MRI follow-up, could help differentiate patients who will progress to AD dementia, from patients who might remain cognitively stable.

In our study, the prevalence of CMB ranged from 62% to 86% depending on the groups. This was much higher than the prevalence seen in the literature, which is usually between 20% and 40% [4 , 28]. It could be related to the high sensitivity of SWI and the use of 3T MRI. Groos [19] and Sepehry [18] showed that in the same patients, CMB had a prevalence of 25% and 18%, respectively, using T2*-weighted GRE sequences, and a prevalence of 40% using SWI. Nandigam et al. [20] showed that T2*-weighted GRE sequences identified only 33% of CMB seen with SWI.

Our results can also be explained by the joint use of minIP reconstructions that helped to increase the sensitivity of CMB detection. To our knowledge, no study has investigated the impact of minIP reconstruction on CMB detection. It could be a major factor explaining the high prevalence of CMB in our study, and therefore, a study comparing the prevalence of CMB with and without minIP could be interesting to perform.

Compared to other studies, our subjects are older, with a mean age of 76.3-year-old. Vernooij et al., showed on 1.5T MRI [9] that the overall prevalence of CMB increases with age, from 17.8% in patients between 60 and 69 years of age, to 38.3% in patients older than 80. The advanced age of our patients may also possibly explain the higher prevalence of CMB in our study.

Finally, the origin of CMB in AD is not yet completely elucidated. CMB have recently been shown to be associated with increased serum level of vascular endothelial growth factor (VEGF) [30]. Since this parameter had not been analyzed in our study, we cannot conclude whether it had influenced the high prevalence of CMB in our population. However, considering that a large biobank of plasma samples has been created during the study, it may be of interest for future investigations to measure VEGF in samples of patients whose MRI has been analyzed.

In this study, we also assessed the presence of lacunes and the volume of WMH. There was no difference in the prevalence of lacunes among groups, neither was there a difference in the cardiovascular risk factors. Those results suggest that patients with vascular dementia were correctly excluded from this study. It also means that CMB detected in our subjects are more likely to be related to CAA, rather than hypertensive vasculopathy.

The volume of WMH was higher in the MCI-p and AD groups. This is consistent with the literature, as many studies showed the independent relationship between AD and WMH [13 , 31–34]. The higher volume of WMH in MCI-p than in MCI-s is also consistent with the literature as studies found more WMH in CS who will develop AD dementia [31 –33] and in MCI-p patients [35].

Because of their lack of specificity, MRI findings are no longer included in diagnostic criteria of AD dementia [1 , 36]. The diagnosis of typical AD can be made in the presence of an amnestic syndrome of the hippocampal type that can be associated with various cognitive or behavioral changes, and at least one of the following changes indicative of in vivo AD pathology: a cerebrospinal fluid profile consisting of decreased Aβ_1 - 42 levels together with increased T-tau or P-tau concentrations, an increased retention of amyloid positron emission tomography tracer or mutations in genes of amyloid precursor protein, presenilin 1, and presenilin 2 [1, 2]. MRI is usually performed only once, at the onset of the disease, in order to exclude differential diagnoses. This study confirmed that a single MRI has low diagnostic impact in AD dementia, as we did not find any significant difference at MRI #1 between NC, MCI-s, MCI-p, and AD patients (except for the volume of WMH).

On the other hand, this study showed that longitudinal MRI follow-up could be interesting, not only to track disease progression using structural MRI, but also to look at vascular endpoints as there were more incident CMB in MCI-p and AD patients. Thus, this longitudinal MRI follow-up could improve the diagnosis of AD dementia and could help predict which MCI patients will convert to AD dementia.

Our study has some limitations. The mean age of our groups was different, with CS patients being 5 years younger than MCI and AD patients. This was due to the matching of subjects that was done on the population of the ROSAS study [25] and was not retained in the sub-population participating in the MRI study. However, this bias has been attenuated by the adjustment used in the statistical tests. Furthermore, the groups were comparable at the beginning of the study, with no difference among those groups in the prevalence of CMB, the median number of CMB, the ratio of patients with 4 or more CMB, and the prevalence of lacunes. The only difference at MRI #1 was an increased volume of WMH in MCI-p and AD patients, which is consistent with the literature. This lack of difference at baseline supports the meaning of the differences found on the evolution between MRI #1 and MRI #3, especially for CMB.

Another limitation of this study was the high rate of patients lost to follow-up. On the 150 patients enrolled in the study, only 84 completed all 3 MRIs. This was mainly explained by comorbidities in these elderly patients leading to withdrawal from the study, and the technical difficulties to perform multiple MRI in patients with severe AD.

The last limitation is that the study was monocentric, with a relatively small population. However, in the literature we found very few prospective studies with more patients, reporting on the use of SWI on AD. The relatively high number of MCI-p patients in our cohort also compensates the small size sample of the study. Besides, consistency in imaging data was inherently higher than in a multicentric setting.

Conclusion

Our study demonstrated that using SWI, we could observe that AD and MCI-p patients developed significantly more new CMB than CS and MCI-s patients during a 4-year MRI follow-up period.

This suggests incident CMB could be an interesting biomarker in AD, especially in identifying MCI-p patients.

This study confirmed that a single MRI had a low impact on the diagnosis of AD dementia, but that longitudinal MRI follow-up could be valuable for monitoring disease progression by identifying new CMB.

DISCLOSURE STATEMENT

Authors’ disclosures available online (http://j-alz.com/manuscript-disclosures/17-0470r1).

References

Dubois

, Feldman

, Jacova

, Hampel

, Molinuevo

, Blennow

, DeKosky

, Gauthier

, Selkoe

, Bateman

, Cappa

, Crutch

, Engelborghs

, Frisoni

, Fox

, Galasko

, Habert

, Jicha

, Nordberg

, Pasquier

, Rabinovici

, Robert

, Rowe

, Salloway

, Sarazin

, Epelbaum

, de Souza

, Vellas

, Visser

, Schneider

, Stern

, Scheltens

, Cummings

(2014) Advancing research diagnostic criteria for Alzheimer’s disease: The IWG-2 criteria. Lancet Neurol 13, 614–629.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

Sperling

(2011) Potential of functional MRI as a biomarker in early Alzheimer’s disease. Neurobiol Aging 32, S37–S43.

Greenberg

, Vernooij

, Cordonnier

, Viswanathan

, Al-Shahi Salman

, Warach

, Launer

, Van Buchem

, Breteler

; Microbleed Study Group (2009) Cerebral microbleeds: A guide to their detection and interpretation. Lancet Neurol 8, 165–174.

McAleese

, Alafuzoff

, Charidimou

, De Reuck

, Grinberg

, Hainsworth

, Hortobagyi

, Ince

, Jellinger

, Gao

, Kalaria

, Kovacs

, Kövari

, Love

, Popovic

, Skrobot

, Taipa

, Thal

, Werring

, Wharton

, Attems

(2016) Post-mortem assessment in vascular dementia: Advances and aspirations. BMC Med 14, 129.

Van der Flier

(2012) Clinical aspects of microbleeds in Alzheimer’s disease. J Neurol Sci 322, 56–58.

Schrag

, McAuley

, Pomakian

, Jiffry

, Tung

, Mueller

, Vinters

, Haacke

, Holshouser

, Kido

, Kirsch

(2010) Correlation of hypointensities in susceptibility-weighted images to tissue histology in dementia patients with cerebral amyloid angiopathy: A postmortem MRI study. Acta Neuropathol 119, 291–302.

Cordonnier

, Van der Flier

(2011) Brain microbleeds and Alzheimer’s disease: Innocent observation or key player? Brain 134, 335–344.

Vernooij

, van der Lugt

, Ikram

, Wielopolski

, Niessen

, Hofman

, Krestin

, Breteler

(2008) Prevalence and risk factors of cerebral microbleeds: The Rotterdam Scan Study. Neurology 70, 1208–1214.

10.

Hanyu

, Tanaka

, Shimizu

, Takasaki

, Abe

(2003) Cerebral microbleeds in Alzheimer’s disease. J Neurol 250, 1496–1497.

11.

Nakata-Kudo

, Mizuno

, Yamada

, Shiga

, Yoshikawa

, Mori

, Nishimura

, Nakajima

, Nakagawa

(2006) Microbleeds in Alzheimer disease are more related to cerebral amyloid angiopathy than cerebrovascular disease. Dement Geriatr Cogn Disord 22, 8–14.

12.

Kalaria

, Ballard

(1999) Overlap between pathology of Alzheimer disease and vascular dementia. Alzheimer Dis Assoc Disord 13, S115–S123.

13.

Kester

, Goos

, Teunissen

, Benedictus

, Bouwman

, Wattjes

, Barkhof

, Scheltens

, van der Flier

(2014) Associations between cerebral small-vessel disease and Alzheimer disease pathology as measured by cerebrospinal fluid biomarkers. JAMA Neurol 71, 855–862.

14.

Shams

, Martola

, Granberg

, Li

, Shams

, Fereshtehnejad

, Cavallin

, Aspelin

, Kristoffersen-Wiberg

, Wahlund

(2015) Cerebral microbleeds: Different prevalence, topography, and risk factors depending on dementia diagnosis— The Karolinska Imaging Dementia Study. AJNR Am J Neuroradiol 36, 661–666.

15.

Pettersen

, Sathiyamoorthy

, Gao

, Szilagyi

, Nadkarni

, St George-Hyslop

, Rogaeva

, Black

(2008) Microbleed topography, leukoaraiosis, and cognition in probable Alzheimer disease from the Sunnybrook Dementia Study. Arch Neurol 65, 790–795.

16.

Olazarán

, Ramos

, Boyano

, Alfayate

, Valentí

, Rábano

, Álvarez-Linera

(2014) Pattern of and riskfactors for brain microbleeds in neurodegenerative dementia. Am J Alzheimers Dis Other Demen 29, 263–269.

17.

Cordonnier

, Van der Flier

, Sluimer

, Leys

, Barkhof

, Scheltens

(2006) Prevalence and severity of microbleeds in a memory clinic setting. Neurology 66, 1356–1360.

18.

Sepehry

, Lang

, Hsiung

, Rauscher

(2016) Prevalence ofbrain microbleeds in Alzheimer disease: A systematic review andmeta-analysis on the influence of neuroimaging techniques. AJNR Am J Neuroradiol 37, 215–222.

19.

Goos

, van der Flier

, Knol

, Pouwels

PJW

, Scheltens

, Barkhof

, Wattjes

(2011) Clinical relevance of improved microbleed detection by susceptibility-weighted magnetic resonance imaging. Stroke 42, 1894–1900.

20.

Nandigam

, Viswanathan

, Delgado

, Skehan

, Smith

, Rosand

, Greenberg

, Dickerson

(2008) MR imaging detection of cerebral microbleeds: Effect of susceptibility-weighted imaging, section thickness, and field strength. AJNR Am J Neuroradiol 30, 338–343.

21.

Goos

, Kester

, Barkhof

, Klein

, Blankenstein

, Scheltens

, Van der Flier

(2009) Patients with Alzheimer disease with multiple microbleeds: Relation with cerebrospinal fluid biomarkers and cognition. Stroke 40, 3455–3460.

22.

Nagasawa

, Kiyozaka

, Ikeda

(2014) Prevalence and clinicoradiological analyses of patients with Alzheimer disease coexisting multiple microbleeds. J Stroke Cerebrovasc Dis 23, 2444–2449.

23.

Van der Vlies

, Goos

, Barkhof

, Scheltens

, van der Flier

(2012) Microbleeds do not affect rate of cognitive decline in Alzheimer disease. Neurology 79, 763–769.

24.

Staekenborg

, Koedam

, Henneman

, Stokman

, Barkhof

, Scheltens

, van der Flier

(2009) Progression of mild cognitive impairment to dementia: Contribution of cerebrovascular disease compared with medial temporal lobe atrophy. Stroke 40, 1269–1274.

25.

de Mauléon

, Soto

, Kiyasova

, Delrieu

, Guignot

, Galtier

, Lilamand

, Cantet

, Lala

, Sastre

, Andrieu

, Pueyo

, Ousset

, Vellas

(2017) The ROSAS cohort: Aprospective, longitudinal study of biomarkers for Alzheimer’sdisease. Strategy, methods and initial results. J PrevAlzheimers Dis. doi: 10.14283/jpad.2017.8

26.

Román

, Tatemichi

, Erkinjuntti

, Cummings

, Masdeu

, Garcia

, Amaducci

, Orgogozo

, Brun

, Hofman

, Moody

, O’Brien

, Yamaguchi

, Grafman

, Drayer

, Bennett

, Fisher

, Ogata

, Kokman

, Bermejo

, Wolf

, Gorelick

, Bick

, Pajeau

, Bell

, DeCarli

, Culebras

, Korczyn

, Bogousslavsky

, Hartmann

, Scheinberg

(1993) Vasculardementia: Diagnostic criteria for research studies. Report of theNINDS-AIREN International Workshop. Neurology 43, 250–260.

27.

Cordonnier

, Potter

, Jackson

, Doubal

, Keir

, Sudlow

, Wardlaw

, Al-Shahi Salman

(2009) Improving interrater agreement about brain microbleeds: Development of the Brain Observer MicroBleed Scale (BOMBS). Stroke 40, 94–99.

28.

Yates

, Desmond

, Phal

, Steward

, Szoeke

, Salvado

, Ellis

, Martins

, Masters

, Ames

, Villemagne

, Rowe

; AINL Research Group (2014) Incidence of cerebral microbleeds in preclinical Alzheimer disease. Neurology 82, 1266–1273.

29.

Kirsch

, McAuley

, Holshouser

, Petersen

, Ayaz

, Vinters

, Dickson

, Haacke

, Britt

3rd , Larsen

, Kim

, Mueller

, Schrag

, Kido

(2009) Serial susceptibility weighted MRI measures brain iron and microbleeds in dementia. J Alzheimers Dis 17, 599–609.

30.

Zhang

, Li

, Zhang

, Li

, Sun

, Zhang

, Wang

(2016) Association of serum vascularendothelial growth factor levels andcerebral microbleeds in patients with Alzheimer’s disease. Eur J Neurol 23, 1337–1342.

31.

Prins

, van Dijk

, den Heijer

, Vermeer

, Koudstaal

, Oudkerk

, Hofman

, Breteler

(2004) Cerebral white matterlesions and the risk of dementia. Arch Neurol 61, 1531–1534.

32.

Lee

, Viqar

, Zimmerman

, Narkhede

, Tosto

, Benzinger

, Marcus

, Fagan

, Goate

, Fox

, Cairns

, Holtzman

, Buckles

, Ghetti

, McDade

, Martins

, Saykin

, Masters

, Ringman

, Ryan

, Förster

, Laske

, Schofield

, Sperling

, Salloway

, Correia

, Jack

Jr , Weiner

, Bateman

, Morris

, Mayeux

, Brickman

; Dominantly Inherited Alzheimer Network (2016) White matter hyperintensities are a core feature of Alzheimer’s disease: Evidence from the dominantly inherited Alzheimer network. Ann Neurol 79, 929–939.

33.

Mortamais

, Artero

, Ritchie

(2014) White matter hyperintensities as early and independent predictors of Alzheimer’s disease risk. J Alzheimers Dis 42, S393–S400.

34.

Grimmer

, Faust

, Auer

, Alexopoulos

, Förstl

, Henriksen

, Perneczky

, Sorg

, Yousefi

, Drzezga

, Kurz

(2012) White matter hyperintensities predict amyloid increase in Alzheimer’s disease. Neurobiol Aging 33, 2766–2773.

35.

Defrancesco

, Marksteiner

, Deisenhammer

, Kemmler

, Djurdjevic

, Schocke

(2013) Impact of white matter lesions and cognitive deficits on conversion from mild cognitive impairment to Alzheimer’s disease. J Alzheimers Dis 34, 665–672.

36.

Dubois

, Hampel

, Feldman

, Scheltens

, Aisen

, Andrieu

, Bakardjian

, Benali

, Bertram

, Blennow

, Broich

, Cavedo

, Crutch

, Dartigues

, Duyckaerts

, Epelbaum

, Frisoni

, Gauthier

, Genthon

, Gouw

, Habert

, Holtzman

, Kivipelto

, Lista

, Molinuevo

, O’Bryant

, Rabinovici

, Rowe

, Salloway

, Schneider

, Sperling

, Teichmann

, Carrillo

, Cummings

, Jack

Jr (2016) Proceedings of the Meeting of the International Working Group (IWG) and the American Alzheimer’s Association on “The Preclinical State of AD”; July 23, 2015; Washington DC, USA, Preclinical Alzheimer’s disease: Definition, natural history, and diagnostic criteria. Alzheimers Dement 12, 292–323.