Relative risk analysis

To determine the relative risk of each of the cerebrovascular outcome measures for the three cohorts, pathological outcomes were dichotomized to indicate the presence/absence of each cerebrovascular event. The risk of a positive finding was then calculated for each cohort, separately. Ninety-five percent confidence intervals were based on the score test for binomial outcomes [2]. Two AD cases were missing measures of atherosclerosis but other measures were available, thus they were excluded from the analysis when comparing atherosclerosis across groups.

Regression analysis

To test the hypothesis that the DS cohort was less likely to have atherosclerosis or arteriolosclerosis but more likely to have CAA than sporadic AD and control cases, we used multinomial logistic regression. Atherosclerosis, arteriolosclerosis and CAA were graded on an ordinal scale: none, mild, moderate to severe. Cumulative logistic regression was used to model the probability of the severity of CAA findings based on the autopsy case cohort [2]. The cumulative-logit models thus represent the full range of severity and allow for tests of whether each of the cohorts was associated with progressive degrees of severity for each of the cerebrovascular outcomes.

All computations were executed in the graphical and programming environment R [48].

Atherosclerosis, arteriolosclerosis and age in DS, AD and controls

As shown in Fig. 2, the DS cohort is younger than the AD and control cohorts. The relative youth of the DS cohort may explain the absence of arteriolosclerosis, while control (16/37 = 43%) and AD cases (36/80 = 45%) (Table 2) showed mild to severe arteriolosclerosis (Fig. 3). Similarly the relative youth may explain the low frequency of atherosclerosis in DS cohort (7/31 = 22.6% -Table 2) and these lesions were in the mild to moderate range (Fig. 4), whereas the frequencies in AD and control cohorts were 43/77 (55.8%) and 28/37 (75.7%), respectively (Table 2). However, in cases that were under 70 years of age, only one control case showed mild arteriolosclerosis and 6 sporadic AD cases showed mild to moderate arteriolosclerosis. Atherosclerotic lesions were observed in a subset of DS cases (25.3%) and these lesions were in the mild to moderate range (Fig. 4). To determine whether younger sporadic AD cases and controls show less frequent and severe atherosclerosis comparable to our DS cohort, we separated the bottom 10% of the AD and control cohorts such that they overlapped with the overall age distribution of the DS cohort. This cutoff left 4 control cases and 8 AD cases. The younger controls showed no CAA, no atherosclerosis and only one case had mild arteriolosclerosis. In contrast, one AD case showed mild atherosclerosis, one showed moderate arteriolosclerosis and almost half showed CAA.

CAA severity increases with age in DS cases

Participants in the DS cohort were on average 20 to 25 years younger at autopsy than participants in the AD or Control cohorts (Fig. 2). The strong separation between the groups, that is DS versus AD plus Control, meant that age-at-autopsy and cohort were effectively confounded in any attempt to account for their relative contributions to the observed distribution of CAA. We found it productive, however, to address the question by analysis of the association with age-at-autopsy in each cohort, separately. Our results showed in particular that age-at-autopsy was strongly correlated with CAA severity among DS participants, but not among participants in the AD or the Control cohorts (Fig. 5). Thus, using a cumulative logistic regression model with CAA severity as the dependent variable, we estimated the probability of the severity of CAA based on age at autopsy for each cohort. In each case the null model, excluding age as predictor, was compared to the alternative model, including age as predictor, using a Chi-squared likelihood ratio statistic (LRS) based on 1 degree of freedom (df). The results showed that a trend with age-at-autopsy was strongly associated with CAA finding in the DS cohort at the 5% level of significance (LRS = 6, df = 1, P-value = 0.014). Again at the 5% level of significance, the results also showed that age-at-autopsy was not significantly associated with CAA findings in the AD cohort (LRS = 0.57, df = 1, P-value = 0.45) nor in the Control cohort (LRS = 2.43, df = 1, P-value = 0.119).

Figure 6 graphs the estimated probability curves for the DS cohort. The estimated probability of a severe finding is shown as an increasing function of age and becomes the most likely finding at about age 54 years. A finding of ‘none’ is shown as the most likely finding before age 43 years, beyond which the most likely finding was estimated to be ‘mild’, at least until the age of about 54 years. Figure 5 graphs the estimated probabilities of each finding at selected ages, including a 95% confidence interval for the population value. At younger ages, say 46 to 50 years in Fig. 5, the confidence intervals overlap appreciably among the various findings. But at the older ages, say 56 years and above, the intervals separate, and the likelihood of a severe finding tends to dominate. Note that above age 59 years in DS, severe CAA was more common than mild or moderate CAA. In contrast to these results, the estimated probability curves (not shown) were basically flat in the AD and Control cohorts, reflecting the non-significant trends with age-at-autopsy in these groups. Nevertheless, the observed trends in the DS cohort appeared to be of independent interest.

CAA in controls, AD and DS

The current study observed CAA in 87.1% of DS cases, 18.9% of controls and 72.2% of AD cases. This is significantly higher than a previous report using neuroimaging outcomes to detect microbleeds in a larger cohort [14]. In a separate study, the frequency of CAA in a prospectively followed cohort of nondemented individuals without DS (n = 59 cases with an average age of 83.9 years at autopsy) has been reported as high as 62% [17]. In contrast, in a NACC study of 140 nondemented control cases without AD neuropathology with an average age at death of 83.5 years, 7.5% exhibited CAA [9]. The NACC study was consistent with the Medical Research Council of Cognitive Function and Ageing Study in England and Wales where 7% of 109 nondemented control cases (70–103 years) exhibit CAA [44]. In contrast, the Adult Changes in Thought (ACT) study noted that 15.7% of control cases (n = 89, 70 years to over 90 years) had CAA [54]. In studies of AD cases, consistently more frequent CAA is observed that range between 37%–83% of autopsy cases [22, 44, 49, 54]. Thus, the range of cases affected by CAA in our AD and control cohorts is not inconsistent with previous reports.

The presence of CAA in 87.1% of DS cases is high both relative to the AD cohort examined here as well as previous reports in nondemented controls and AD cases, suggesting that CAA may be a critical vascular pathology associated with aging in DS. A recent neuroimaging study using T2*-based MR outcome measures observed that CAA was present in 31% of cognitively impaired people with DS [14].

More frequent CAA is observed in autosomal dominant AD subjects, who also exhibit Aβ overproduction and develop early onset AD. In a study by Ringman and colleagues, autosomal dominant AD cases who were of similar age to the current cohort of DS autopsy cases (average age 53.5 years) showed more moderate to severe CAA (63.3%) as compared to older sporadic AD cases (39.2%, mean age 79.3 years) [49]. In contrast, in rare autopsy cases of DS with partial trisomy 21, where APP is not overexpressed, AD neuropathology and CAA was absent even at 72 or 78 years of age at autopsy [19, 46]. Thus, higher levels of CAA in DS are most likely due to APP overexpression and consequent increased Aβ accumulation. CAA in people with DS is more consistent with the CAA prevalence in early onset autosomal dominant AD.

Arteriolosclerosis and Atherosclerosis in AD, controls and DS

Atherosclerosis and arteriolosclerosis were uncommon in the DS cohort, which may be due to their younger ages (50–59 years). In a NACC study of older nondemented autopsy cases lacking AD pathology, 77.6% of cases showed arteriolosclerosis and 82.9% showed atherosclerosis (n = 140, mean age 83.5 years) [9]. In studies of nondemented autopsy cases under 70 years of age (50–70 years), 68.6% (n = 1008) were positive for arteriolosclerosis [29]. Further, Ighodaro et al. showed using NACC autopsy data from cases at an average age of 50 years that only 20% exhibited arteriolosclerosis.

Implications of CAA in DS

The contribution of CAA pathology to the development of AD neuropathology and/or dementia is increasingly recognized as being critical [33, 55]. Further, cerebrovascular lesions may be one of the underlying causes of variable (and modest) clinical trial outcomes as they may be more or less engaged in individual patients and possibly blunting responses to Aβ interventions. In DS, the higher frequency of CAA is likely to lead to several possible outcomes: First, CAA can lead to vascular dysfunction including impaired constriction and dilation, which is consistent with reports of reduced FDG-PET with age in DS and also associated with cognitive decline [51]. Second, CAA can lead to blood brain barrier disruptions and microhemorrhages [50, 58], which also may contribute (in addition to APP overexpression) to the earlier age of onset of dementia in people with DS. Interestingly, a recent review by Buss and colleagues suggests that despite the presence of CAA in DS, there appears to be less intracerebral hemorrhage suggesting possible protective mechanisms; this will be an exciting avenue of research for future studies [13].

It is important to consider some caveats to the current study, which include the use of observational data, the relatively small sample size of our DS cohort and the younger age of this cohort. Autopsy studies of DS are a challenge as the number of tissue donations are relatively small, thus our samples size is smaller relative to sporadic AD cohorts. Further, people with DS who volunteered for this research study and consented to autopsy may also represent a biased sample. Age is virtually impossible to control for in the comparison of these three groups as people with DS have a significantly younger age of onset of AD neuropathology and development of dementia.

The higher frequency of CAA in DS compared to sporadic AD is interesting in that cerebrovascular pathology in DS appears to have a unique signature, i.e. significant CAA and low or absent atherosclerosis and arteriolosclerosis, which could impact cognition and age of onset of dementia in DS. Weller et al. have suggested that CAA may impact therapeutic outcomes and may predispose people to vasogenic edema and hemorrhagic complications due to reduced drainage of Aβ associated with affected vessels [59]. Experimental evidence by the same group shows that periarterial lymphatic drainage is impaired with age and CAA. With an increase in production of Aβ in DS it seems likely that CAA and the presence of Aβ plaques would form at a younger age and with less severity of arteriolosclerosis in DS than in sporadic AD. The present study supports the hypothesis that a life-long increase in production of Aβ in DS predisposes to an earlier onset of age-related impairment of periarterial elimination of Aβ from the brain and thus may accelerate the onset of the pathological features of AD [4] .

While the hypothesis of reduced periarterial elimination of Aβ from the brain is compelling, in DS it does not readily account for findings from other groups of Aβ’s role in cerebrovascular disease and plasma levels of Aβ. For example, Gomis and colleagues show that plasma Aβ1–40 levels are associated with cerebrovascular small vessel disease in acute lacunar stroke and suggest that vascular Aβ is primarily Aβ1–40, which alters endothelial functions [24]. In DS, plasma data show that the risk of dementia in DS is increased as levels of Aβ1–42 decline but Aβ1–40 levels increase [53]. As reviewed by Biffi and Greenberg, a high Aβ 40:42 ratio appears to be an important index of vascular amyloid formation [10]. However, Carmona-Iragui and colleagues did not find differences in CSF Aβ40 that corresponded to microbleeds by neuroimaging in DS [14]. Therefore, investigation of plasma markers and CAA presence in DS as related to clinical dementia may become an important avenue of investigation.

If we extend these findings in the context of considering possible interventions for people with DS to prevent AD, it will be important to target CAA pathology specifically. For example, there is a clinical trial of ponezumab (PF-04360365) that is specifically targeting CAA as a treatment for AD that may be very relevant for the DS population [5, 34]. Thus, CAA pathology in DS may be a significant factor to consider in the design of future clinical trials.

In future, it will be important to link neuropathological vascular findings with clinical outcomes (e.g. severity of dementia in DS) and to further explore the additional downstream pathologies associated with CAA. For example, the role of microhemorrhages in DS and their link to the extent of CAA in DS is as yet unknown. Further it will be interesting in future studies to distinguish CAA that is associated with capillary Aβ and associated CAA findings such as microinfarcts, superficial siderosis, etc. Neuroimaging may also provide novel insights for vascular pathology in DS as has been reported previously [14]. Using T2* magnetic resonance imaging, Carmona-Iragui and colleagues observed CAA in 31% and 38%, respectively in people with DS who were nondemented compared with those with dementia. Linking neuroimaging outcomes to subsequent autopsy studies will be a challenge for studies of DS but should evolve over time.