|Home | About | Journals | Submit | Contact Us | Français|
The Adult Changes in Thought (ACT) study is a longitudinal population-based prospective cohort study of brain aging and incident dementia in the Seattle metropolitan area. Observational studies using autopsies from ACT indicate that dementia is a convergent syndrome that commonly derives from Alzheimer’s disease (AD), microvascular brain injury (μVBI), and Lewy body disease (LBD), and that these diseases have prevalent subclinical forms that also are commonly co-morbid. The existence of subclinical diseases highlights potential opportunities to intervene before the development of clinically apparent impairments. Our observations suggest that some such interventions already may exist to suppress processes of AD (statin therapy) or μVBI (treatment of hypertension). Reduced burden of LBD is associated with cigarette smoking; although smoking is not recommended as an intervention, these exposure data may provide clues to alternative neuroprotective mechanisms. Self reported anti-oxidant supplementation was without apparent effect in this cohort on indices of AD, μVBI, or LBD. Continued observational studies of brain aging will provide further insight into the convergent complexity of the dementia syndrome and its subclinical forms as well as highlight potential interventions that will require validation in clinical trials.
The Adult Changes in Thought (ACT) study is a long-standing ongoing, population-based prospective cohort study of brain aging and incident dementia (DSM-IV criteria) in men and women aged 65 years and older who are members of the Group Health Cooperative in the Seattle region . ACT is one among a small number of population-based studies of brain aging and dementia with autopsy endpoints in the world. Some features of ACT are unique , it includes community-living men and women from an urban and suburban setting in the USA, and it has a rich pharmacy database. This resource provides the opportunity to conduct observational studies that explore links among cognitive performance, drug exposure, and neuropathologic outcomes. We have termed this undertaking pharmaconeuropathology. Here we review some of the major findings from these studies.
Before doing this, it is worthwhile to note some compromises struck in the design of ACT. First, the ACT cohort consists of approximately 4000 randomly selected volunteers from a total eligible population of about 26,000 people. Of these participants, about 1900 have died, and of these 21% consented to brain autopsy. This percentage of autopsies compares favorably with other population-based studies, such as the Honolulu Asian Aging Study (HAAS) , but is much lower than some studies of more restricted groups that require autopsy consent to enroll such as the Nun Study or the Religious Orders Study [4, 5]. While there are obvious advantages to investigating a more diverse group of people, the tradeoff is the resulting lower autopsy rate that potentially introduces bias. Our group has developed statistical methods to adjust for selection effects thereby preserving the ability to generalize our results . Second, ACT uses the Cognitive Abilities Screening Instrument (CASI) as its neuropsychological screening tool . Some others also use this screening tool, for example HAAS, but it is not an extensive neuropsychologic test battery like ones employed in some other research cohorts. Finally, ACT follows participants every two years from entry until death or diagnosis of dementia by performing medical and cognitive evaluations.
As has been observed by every other population-based study of dementia in the world, ACT participants clearly demonstrate that dementia is a syndrome: a shared clinical outcome that derives from multiple etiologies [2, 3, 5, 8–12]. In our initial work we used nine different measures of brain injury and neurodegeneration: extent of atherosclerosis, brain weight, number of cystic infarcts including lacunar infarcts, neuritic plaque density, Braak stage for neurofibrillary tangles (NFTs), severity of amyloid angiopathy, isocortical and brainstem Lewy bodies (LBs), hippocampal sclerosis, and number of cerebral microinfarcts (CMIs). From these we determined that the clinical diagnosis of dementia was significantly and independently associated with pathologic endpoints of Alzheimer’s disease (AD) as assessed by Braak stage for NFT V or VI, vascular brain injury (VBI) as assessed by > 2 cerebral CMIs, and isocortical Lewy body disease (LBD) as assessed by α-synuclein immunohistochemistry for any LB in the frontal or temporal cortex. The relative risk (RR) for a clinical diagnosis of dementia with this level of AD, VBI, or LBD is 5.9, 4.8, or 5.1, respectively. Since ACT is a population-based study, we could estimate the population-attributable risk for dementia among participants for each of these diseases (Figure 1). As expected, AD was the most common disease contributing to dementia in the ACT cohort. Unexpectedly, despite being a group of men and women with ready access to health care and relatively low prevalence of cigarette smoking, μVBI (as assessed by systematized screening for CMIs) was a close second to AD as a contributor to dementia .
It is essential to realize that these statistical estimates cannot be used to draw mechanistic conclusions. For example, while neuritic plaque density was not significantly independently-associated with the clinical diagnosis of dementia when put in competition with Braak stage for NFTs, this does not mean that the processes that underlie neuritic plaque formation are unimportant in the pathogenesis of AD. The same can be said for territorial infarcts vs. microinfarcts. Moreover, there are additional diseases that can cause dementia, (e.g., Frontotemporal Lobar Degenerations or Creutzfeldt-Jakob disease) but ACT is insufficiently powered to detect the contribution of these less common diseases in our 65 years of age and older population-based cohort. Finally, it is important to note that approximately 12% of the population-attributable risk for dementia is unexplained by these point estimates. There are several possible reasons for this outcome that are not exclusive. One is that less severe forms of these three common diseases can produce dementia in some people, e.g., Braak stage IV AD, 2 CMIs, or subcortical LBD. Another is that there may be other moderately common contributors to dementia in the population that are not captured by neuropathologic examination. Alternatively, since the sum of the error for each estimate exceeds 12%, it is possible that there are no other common contributors to dementia in the ACT cohort. We will discuss this more in the next section.
While the population-attributable risk estimates are very helpful in considering the burden of diseases that can cause dementia in the ACT cohort, they do not reflect the complex mixtures of diseases that may exist in individuals. To demonstrate this complexity visually, we devised a “summary neuropathology score” that is the sum of: (i) Braak stage for NFTs expressed as a number (0 to 6) divided by 2 to give values of 0, 0.5, 1, 1.5, 2, 2.5, and 3; (ii) number of CMIs with > 3 expressed as 3 to give a values of 0, 1, 2, or 3; and (iii) LBD coded as 0 (none), 1 (brainstem), 2 (limbic), or 3 (isocortical). Thus the minimum score is 0 and the maximum is 9. The same numerical score could represent different disease contributors to cognitive impairment or dementia. For example, a cumulative neuropathology score of 1.5 could translate into a Braak stage for NFTs of III if only AD was present, or a Braak stage for NFTs of II and 1 CMI. Although the RR for causing dementia is similar for each disease, they are not identical and thus the score is not strictly weighted. Nevertheless, the RRs are similar and so scaling each disease from scores of 0 to 3 and expressing the sum provides a simple visual representation of the co-morbid burden of disease in a given individual.
As a first analysis we divided the ACT autopsy cohort of individuals with last clinical evaluation within two years of death into three functionally-defined subsets: Dementia, Not Dementia (ND) with low cognitive function, and ND with high cognitive function. Dementia or ND was diagnosed using DSM-IV criteria. The ND group was divided into low and high cognitive function by CASI score . CASI is a 100-point scale, and among the entire ACT cohort, scores < 90 are in the lowest quintile. The ND low cognitive function group had last CASI < 90. The ND high cognitive function group had CASI > 90.
Figure 2 plots the cumulative relative frequency (CRF) of the overall summary neuropathology score for these three groups. As expected, the overall burden of diseases progresses from lowest in the ND & CASI > 90 group to highest in the Dementia group. One summary estimate of disease burden is 50% of total CRF (CRF50). The CRF50 for summary neuropathology score progressively increased from approximately 1.25 in the highest functioning group (ND & CASI > 90), to 2.0 for the intermediate functioning group (ND & CASI < 90), to 3.5 for the lowest functioning group (Dementia). Non-parametric ANOVA (Kruskal-Wallis test) of the summary neuropathology score among these three functionally-defined groups had P < 0.0001 and Dunn’s corrected repeat paired comparison had P < 0.01 for each of the three pairs. These cross sectional observations from groups defined by cognitive performance within two years of death indicate that increasing overall burden of diseases that can contribute to the dementia syndrome is significantly associated with decreasing cognitive function.
It is important to note that everyone in the highest functioning group (ND & CASI > 90) had at least some evidence of one of these diseases at autopsy, suggesting widely prevalent clinically silent disease in this group. Moreover, the burden of disease even among these high functioning individuals varied continuously from very low to high with approximately 15% having summary neuropathology scores that exceeded the CRF50 for individuals in the Dementia group. The same relationships held with the ND & CASI < 90 group but shifted to the right with approximately 25% of this group having cumulative disease burden that exceeded the CRF50 for individuals in the Dementia group.
In summary, the three diseases that are significantly associated with a clinical diagnosis of dementia in ACT participants— AD, μVBI, and LBD—are highly prevalent among individuals with high cognitive function within two years of death. Furthermore, about 15% of this group carries levels of disease that might reasonably be expected to produce dementia. While these observations are not new [3, 5, 18, 19], our summary neuropathology score provides a useful tool to express these relationships among co-morbid AD, μVBI, and LBD.
Figure 3 presents the separate components of our summary neuropathology score for each ACT participant in either the Dementia or ND & CASI > 90 groups arranged from lowest to highest overall score. There is a highly varying mixture of AD, μVBI, and LBD among patients in the Dementia group (Figure 3A). This is a very important point to consider when contemplating organization of clinical trials for disease- modifying therapies or the management of patients once such therapies are developed, since there are not as yet validated intra vitam neuroimaging or biomarkers for each of these diseases. As mentioned above, it is unusual to have no evidence of any of these three diseases in a patient who expressed the dementia syndrome. The more common observation is that a participant with dementia has disease burden that is below the cutoffs used in our estimates of population-attributable risk. This seems to favor the interpretation that there is a spectrum of susceptibility to developing dementia. The basis for varying vulnerability among individuals is not clear but likely includes both genetic and environmental factors, some of which are likely modifiable.
Figure 3B presents the same data from the ND & CASI > 90 group. While the burden of overall disease is lower in this group than the Dementia group, there again is a highly varying mixture of disease type and amount among high-functioning participants. These data show that clinically silent disease also is convergent with roughly equivalent distribution of disease types as observed in the Dementia group. This is a critical point because it suggests that there is not a common pattern to the development of these diseases but rather heterogeneity among individuals. Perhaps even more importantly, these data indicate that many—even most—older individuals without dementia have initiated neurodegenerative diseases(s) and thereby present an opportunity of tremendous public health potential to intervene before these diseases manifest clinically. Two key elements to achieving this goal will be effective disease-modifying interventions and the ability to diagnose clinically silent disease during life.
In summary, diseases significantly associated with the clinical diagnosis of dementia variably converge among individuals in both clinically overt and clinically silent states.
Despite hundreds of experimental animal and clinical studies, only a handful have investigated modification of human neuropathologic endpoints by common exposures, an area of investigation we describe as pharmaconeuropathology. Although autopsy studies have limitations, they also bring a unique perspective to understanding mechanisms of disease that is not reflected in animal models or captured in clinical trials. Two major limitations to pharmaconeuropathology are that most autopsy series are comprised of non-representative cohorts and that most have limited exposure data. ACT meets both of these challenges. Here we review our results for statins, antioxidant supplements, treatment for DM or hypertension, and cigarette smoking.
Epidemiological studies suggest a protective effect of 3hydroxy- 3-methylglutaryl-coenzyme-A reductase inhibitors (“statins”) against development of cognitive impairment , all-cause dementia [21, 22], and AD . However, more recent large prospective cohort studies [24, 25] have not supported these earlier findings. Furthermore, two large randomized clinical trials of statins for prevention of coronary heart disease failed to provide evidence of a protective effect against cognitive decline [26, 27]. Although we did not find an overall significant protective effect of statins against dementia in an initial analysis of data from ACT , preliminary findings from a larger subject sample suggested a protective effect for AD in the ACT cohort, particularly in subjects younger than age 80 years of age at enrollment .
Based on these studies, we sought evidence of reduced risk for dementia by examining associations between statin use and pathologic endpoints in ACT participants . After controlling for age at death, gender, CASI score at entry, brain weight, and number of CMI, the odds ratio (OR) for each one-unit increase in Braak stage for NFTs was lower in statin users compared to non-users (OR 0.44; 95% CI: 0.20, 0.95). The OR for each one-unit increase in neuritic plaque score did not differ between groups (OR 0.69; 95% CI: 0.32, 1.52). The risk for combined AD pathologic changes derived from both Braak stage and neuritic plaque score  was lower in statin users (OR 0.20; 95% CI: 0.05, 0.86). These findings suggest that statin use may provide some limited but significant protection against the development of AD pathologic changes.
Evidence from transgenic mouse models, autopsy studies, and epidemiologic investigations support a pathogenic role for increased free radical damage to diseased regions of brain in AD (reviewed in ); however, results from clinical trials of vitamin E (α-tocopherol) supplements are only mildly encouraging [33, 34] with a trend for partial protection of a composite verbal memory measure with α-tocopherol treatment in one study . In a clinical observational study conducted by our colleagues, dietary supplementation with α-tocopherol, vitamin C, or both by ACT participants was not associated with reduced risk of developing dementia over 5.5 years of follow-up .
In an effort to limit confounding by co-morbid diseases, we separated ACT autopsies based on neuropathologic features into those with high level AD, high level μVBI, high level AD and μVBI, and no/low level AD or μVBI . Neither average age nor average number of years of education was significantly different among groups. We used three quantitative in vivo markers of free radical injury: F2-Isoprostanes (IsoPs) that reflect injury to all tissue elements, F4-Neuroprostanes (NeuroPs) that reflect free radical injury to neuronal membranes, and F2 Adrenoprostanes (AdrenoPs) that reflect free radical injury to synaptic vesicle membranes in gray matter . We observed significantly increased F4-NeuroPs and F2-AdrenoPs in cerebral cortex of patients with AD and increased levels in all three markers of free radical injury in individuals with μVBI as well as those with co-morbid AD and μVBI. These data suggest that AD produces free radical injury in neurons while μVBI produces free radical injury that affects all tissue elements in cerebral cortical gray matter. Greatest free radical damage was observed in patients with co-morbid AD and μVBI, raising the possibility that AD and μVBI may have additive effects on oxidative damage. AD and large vessel disease are known to cause free radical-mediated injury to brain . Our data are the first to show that μVBI is associated with increased free radical injury to cerebral cortex. We determined that typical antioxidant supplement use was not associated with suppressed levels of free radical damage to brain in controls, AD, μVBI, or AD plus μVBI, consistent with clinical observations from the full ACT cohort .
DM appears to increase risk for dementia in the elderly; however the underlying mechanisms, intersection with AD and μVBI, and effects of therapy remain unclear. We tested the hypothesis that DM promotes specific neuropathologic processes contributing to dementia and that these processes may be suppressed by anti-diabetic therapy . We analyzed pathologic endpoints and biochemical data from ACT participants who came to autopsy (n=259) and for whom there was information on DM status (n=196). Biochemical analysis was conducted on a subset of these cases who had rapidly frozen brain tissue (n=57). DM was diagnosed based on information obtained from ACT pharmacy and clinical records.
We observed two patterns of injury in patients with dementia depending on DM status. Dementia without DM had greater neuritic plaque score and increased F2-isoPs in cerebral cortex, while dementia with DM had greater number of CMIs and increased cortical IL-6 concentration. CMI burden was greater in deep cerebral structures in treated diabetics with dementia, perhaps a reflection of more severe DM in treated patients. Neuritic plaque score tended to be greater in untreated than treated diabetics with dementia.
These novel characterizations of two different patterns of cerebral injury in patients with dementia depending on DM status may have etiologic and therapeutic implications. Indeed, one interpretation of our data, especially when combined with results of others, is that treatment of DM might lower neuritic plaque burden. However, we cannot formally exclude the possibility that treatment of DM might also carry negative consequences as reflected in increased CMIs .
In this study, we assessed common pathologic changes in the aging brain to determine what type of injury is most closely associated with elevated blood pressure (BP) [13, 41, 42]. We also examined the effect of anti-hypertensive treatment. At each ACT study visit, BP is measured by trained research assistants and recorded twice in a sitting position at least five minutes apart. We used the average systolic and diastolic BP values from the last ACT study visit, and we excluded those with more than two years between last ACT study visit and death. The presence of more than 2 CMIs, but not any other pathologic score, was independently associated with systolic BP (SBP) in younger participants (age 65–80, n=137), but not in older participants (age >80, n=91). The RR for >2 CMI with each 10 mmHg increase in SBP was 1.15 (95% CI: 1.00, 1.33) in the younger participants, adjusted for age-at-entry, gender, and time to death. This RR was particularly strong in younger participants not taking antihypertensive medications (RR: 1.48; 95% CI: 1.21, 1.81). Significant associations were not observed in participants treated for hypertension. Findings for diastolic BP were negative.
Several case-control or cohort studies from across the globe have reported that cigarette smoking is inversely associated with Parkinson disease (PD) [43–45]. These observations have inspired experimentation and speculation on the role of nicotine as a modulator of dopaminergic neurodegeneration . In contrast to PD, outcomes from a similar number of observational studies of smoking and the risk of clinically diagnosed AD or all-cause dementia have been inconsistent (reviewed in ). Despite these clinical and experimental studies, the neuropathologic correlates of smoking remained unknown. We addressed this gap in knowledge by determining the association between lifetime cigarette use and pathologic changes of AD or LBD in ACT participants .
We observed that heavy lifetime cigarette smoking (> 50 pack-years) was associated with significantly reduced RR for LBD (0.43; 95%CI 0.15, 0.90) compared to moderate smokers, but not AD-type pathologic changes, after correcting for selection bias, with significantly reduced frequency of LBD in the substantia nigra (P<0.05 by Fisher’s exact test). These findings are the first to associate reduced LBD with any exposure, and substantiate observational study results that have associated cigarette smoking with reduced risk of PD.
Our observational studies from ACT autopsies indicate that dementia is a convergent syndrome that is significantly associated with AD, μVBI, and LBD, and that these diseases have prevalent subclinical forms that also are commonly comorbid. The existence of subclinical forms of disease is very important because it highlights potential opportunities to intervene before the development of clinically apparent impairments. Our findings suggest that some such interventions may already exist for AD (statin therapy) and μVBI (treatment of hypertension and perhaps treatment of DM). While cigarette smoking obviously is not recommended as an intervention, it may provide clues to alternative mechanisms to suppress LBD. Anti-oxidant supplementation was without apparent effect on measures of AD, μVBI, LBD, or free radical injury. Continued observational studies of brain aging will provide further insight into the convergent complexity of the dementia syndrome, and highlight potential interventions that will require validation in clinical trials, ideally designed to test interventions that prevent or delay onset of these common diseases that contribute to late life dementia.
U01 AG06781, R01 AG23801, P50 AG05136, P50 NS62684, and the Nancy and Buster Alvord Endowment.
CONFLICT OF INTEREST None declared.