Introduction

Cerebral microinfarcts are highly prevalent in older adults [1-5] and are associated with a higher risk of cognitive impairment and dementia [6-8]. Several studies reported a strong association between cerebral microinfarcts and poorer cognitive function and faster progression of dementia [9-12]. For example, a study based on the Religious Orders Study showed increased odds of dementia (OR=1.77; 95%: 1.07, 2.92) in persons with microinfarcts, especially those with multiple microinfarcts [12]. However, that study was conducted in a small older cohort of Catholic clergy (n=425) who have demographic and life-style features unique from other community-based lay populations. Studies are needed to validate the findings in a more representative cohort of the general population with a larger sample size. Dementia is more common in women probably because women live longer than men [13], but less is known about the sex-related variation in the relationship between cerebral microinfarcts and dementia. The purpose of this autopsy-based study is to examine the relationship of cerebral microinfarcts with dementia, overall, and stratified by sex.

Methods

Data and participants

We used data from the ACT study, an ongoing populationbased cohort study of incident dementia, the details of which have been previously described [14,15]. Briefly, a random sample of 2,581 adults 65 years and older free of dementia who were members of Kaiser Permanente Washington (previously Group Health) in the Seattle region were enrolled during 1994-1996 and followed biennially to screen for dementia. A cohort of 811 adults were subsequently enrolled during 2000-2003 with the same methods, and ACT continued enrolling participants from 2004-present to replace attrition from dementia, dropout, and death, ensuring a consistent cohort of ≥2,000 at risk for dementia. Some participants consented to autopsy. The ACT participants underwent biennial follow-up visits for cognitive screening and those who screened “positive” (Cognitive Abilities Screening Instrument score < 86) underwent a full neuropsychological evaluation [15]. Total enrollment over the course of the study is >5,000 with 850 autopsies available at time of this analysis; the current study is based on the 727 autopsied participants who had complete data on presence, count and locations of microinfarcts and dementia (see Figure 1). We excluded 115 people who were “resilient” [16] (no cognitive impairment despite severe neuropathology). The study protocols were approved by Kaiser Permanente Washington Institutional Review Boards.

Identification of cases and controls

Dementia status was categorized into two groups: present and without. The ACT subjects were re-screened every two years using the Cognitive Abilities Screening Instrument (CASI). All subjects showing a significant cognitive deficit on the CASI (< 86, score range 0 - 100) [17] underwent a dementia diagnostic work-up, including neuropsychological testing and physical and neurological examination by a geriatrician and neurologist. The neuropsychological test battery included the CERAD Neuropsychological battery, Wechsler Memory Scale (Paired Associates/Recall and Logical Memory/Recall), Mattis Dementia Rating Scale, Boston Naming Test and selected questions from WAIS-R. Relevant laboratory results and brain CT or MRI results were obtained from Kaiser Permanente records. Clinical dementia diagnoses were made by the ACT Consensus Diagnosis Conference per DSM-IV criteria (procedures are described in detail elsewhere) [18]. Subjects who did not meet criteria for dementia were re-screened at subsequent biennial follow-up visits. Subjects who met criteria for new onset dementia entered annual follow-up for verification of dementia diagnosis and monitoring of cognitive decline. Cases were defined by the presence of dementia, and those without dementia were considered control.

Ascertainment of exposure

Neuropathological examinations were performed by board certified neuropathologists blinded to the clinical diagnosis and status of risk factors. We limited our evaluation of microinfarcts to chronic [19] as acute and subacute infarcts would be expected to be rare shortly before death and would be unlikely to contribute to dementia [8]. Chronic microinfarcts were defined as “a focal lesion attributed to ischemia, found only on microscopic examination, and judged to be temporally remote” [8]. Microinfarcts were measured in cortical gray matter and its underlying white matter, and in deep ganglionic masses that included deep white matter [8]. Based on the judgements of the neuropathologists, microinfarcts were categorized as none, 1, and ≥ 2. Chronic microinfarcts in cortical gray matter regions (frontal, temporal, occipital, and parietal) were classified as cortical, and those with caudate nucleus, putamen, internal capsule, and thalamus were classified as subcortical.

Other variables

AD pathology (CERAD, Braak). Neuritic plaques and neurofibrillary tangles were evaluated by established methods [2025].CERAD staging for neuritic amyloid β plaque was categorized as 0=none, 1=sparse, 2=intermediate, 3=frequent [21]. Braak stage for neurofibrillary tangles was categorized as 0=none (normal), 1=I (very mild), 2=II (mild), 3=III (moderate), 4=IV (moderately severe), 5=V (severe), 6=VI (very severe) [22].

The variables, age at death, sex, APOE e4 genotype, arteriolosclerosis, and cerebral amyloid angiopathy, were included as adjustments in all models. APOE genotyping has been completed using previously established methods [19,20]. APOE e4 status was rated as 0 = absence of the e4 allele, 1 = 1 or 2 e4 alleles. Arteriolosclerosis was evaluated by histologic examination on existing Hematoxylin-Eosin stained sections from each region of interest. Severity of this pathology was graded based on concentric hyaline thickening of vessel walls. Smaller arterioles, less than approximately 50 microns, were used for evaluation. The severity was categorized into four levels: none, mild, moderate, and severe. The presence of cerebral amyloid angiopathy (CAA) was determined by immunohistochemistry to assess degree of compacted amyloid β deposits in the brain leptomeninges and occipital cortex [25], and its severity was ranked on a four-point scale (none, mild, moderate, or severe).

Statistical analyses

The analyses were conducted in 727 participants with both microinfarcts and dementia data available among 850 autopsy participants. The process of creating the cohort is shown in Figure 1. We excluded 8 participants without known microinfarct status and 115 participants who were resilient to Alzheimer’s disease (81 with BRAAK stage of V or VI but no dementia and 34 with severe CERAD but no dementia). We compared participant characteristics in the overall cohort and sub-cohorts stratified by sex, respectively. Due to the sample size of the study, we reported absolute standardized difference (ASD) instead of p-value to estimate the magnitude of difference between groups. It is commonly accepted that ASD >10% indicates imbalanced characteristics between comparison groups [23].

Then we fit five multivariable logistic regression models to estimate adjusted odds ratio (aOR) and 95% confidence interval (CI) for dementia associated with each of five microinfarct measures, respectively, adjusting for the other covariates. Microinfarcts were used in four different ways: categorical variable indicating presence (present vs. not), continuous variable measuring number of microinfarcts, categorical variable categorizing number of microinfarcts (one vs. none, two or more vs. none), and categorical variable by location of microinfarcts: cortical (present vs. not) and subcortical microinfarcts (present vs. not). The regression models were adjusted for age, sex, APOE e4 carrier, CAA, and arteriolosclerosis. Because the objective of the study was to examine whether the association between microinfarct and dementia varied by sex, we ran the regression models separately in males and females. Finally, we checked for interaction between microinfarct and sex in the overall cohort. To understand the association between dementia and other covariates, we also report ORs with 95% CI in the model with the main exposure of presence of microinfarcts. Similarly, we also estimated the interaction between sex and each of the covariates and conducted the subgroup analysis stratified by sex. All analyses were conducted using SAS 9.4.

Results

Participant characteristics

Among 727 participants with available microinfarct data in the autopsy cohort, 381 (52.4%) had dementia. Among 308 males and 419 females dementia was present in 150 (48.7%) and 231 (55.1%) individuals, respectively. Compared to participants with no dementia, a higher proportion of those with dementia were older, were APOE e4 carriers, and had higher severity of Braak (V-VI), CERAD (moderate and severe), CAA (moderate and severe) and arteriolosclerosis (moderate and severe) (Table 1). Distribution of these characteristics was generally similar among males and females (Table 1).

Table 1. Characteristics of the Participants

Microinfarcts and dementia

Microinfarcts were present in 356 (49%) participants; 144 (40.4%) had one microinfarct; 211 (59.3%), two or more; 257 (72.2%), cortical; and 216 (60.7%), subcortical microinfarcts. Compared with participants without dementia, more of those with dementia had two or more microinfarcts, although there were no differences in the distribution of a single microinfarct (Table 1). The aOR (95% CI) for dementia associated with the presence of any microinfarct was 1.36 (0.97-1.91; Table 2). aORs (95% CIs) for males and females were 1.24 (0.75-2.06) and 1.45 (0.91-2.30), respectively. There was no interaction by sex in this association (p for interaction, 0.34; Table 2).

When microinfarcts were used as a continuous variable, the aOR estimate indicated that increasing number of microinfarcts by one, the odds of developing dementia would increase 7% (Table 2); however, the association was not statistically significant. The association was significant in males only with having one more microinfarct increasing the odds of dementia by 16%, although the interaction with sex was not significant with p-value at 0.32 (Table 2), when microinfarcts were used as a categorical variable, aORs (95% CIs) for dementia associated with one microinfarct and two or more microinfarcts were 1.24 (0.80-1.92) and 1.46 (0.98-2.18), respectively (Table 2). A similar numeric dose-response or graded association was also observed in males but not in females (Table 2). When microinfarcts were used as a continuous variable, aORs (95% CIs) for dementia associated with increase in the number of microinfarcts by one in the overall cohort and among males and females were 1.07 (0.98-1.16), 1.16 (1.01-1.33), and 1.01 (0.911.13), with no evidence of sex interaction in the relationship (p for sex interaction, 0.32; Table 2).

In the overall cohort, aOR (95% CI) for dementia associated with cortical microinfarcts was 1.19 (0.83-1.69), which was higher in males (OR, 1.29; 95% CI, 0.76-2.19) than in females (aOR, 1.11; 95% CI, 0.69-1.80), but the between-sex difference was not statistically significant (p for sex interaction, 0.96; Table 2). In contrast, subcortical microinfarcts had a significant association with dementia in the overall cohort (OR, 1.65; 95% CI, 1.14-2.38), which also had no interaction by sex (p for sex interaction, 0.55; Table 2).

Table 2. Association of microinfarcts presence, number, location with dementia adjusting for age, sex, APOE e4, arteriolosclerosis, cerebral amyloid angiopathy 

Note: * p <0.05

Other predictors of dementia

When compared with participants younger than 80 years, aOR (95% CI) for dementia associated with age 80-89 years was 3.66 (1.95-6.86), which was significantly higher in females (aOR, 8.38; 95% CI, 2.74-25.62) than in males (aOR, 2.18; 95% CI, 0.96-4.92; p for interaction, 0.03; Table 3). Respective aOR (95% CI) associated with ≥90 years was 6.67 (3.55-12.55), which was also significantly stronger for females (p for interaction, 0.03; Table 3). APOE e4 carrier (vs. non-carrier) state was associated with higher odds of dementia (aOR, 1.76; 95% CI, 1.19-2.61), without any evidence of interaction by sex (p for interaction, 0.53; Table 3). Moderate to severe CAA was associated with dementia (aOR, 2.62; 95% CI, 1.63-4.22), which was similar for both males and females (p for interaction, 0.7; Table 3). Moderate and severe arteriolosclerosis were also significantly associated with dementia (aOR, 2.62; 95% CI, 1.63-4.22), which was stronger in females (p for interaction, 0.06 and 0.11 for moderate and severe arteriolosclerosis; Table 3).

Note: The estimates of covariates were derived from the model with the main exposure of presence of microinfarcts.