Lewy Bodies, Vascular Risk Factors, and Subcortical Arteriosclerotic Leukoencephalopathy, but not Alzheimer Pathology, are Associated with Development of Psychosis in Alzheimer’s Disease

INTRODUCTION

It is estimated that approximately 50% of patients with Alzheimer’s disease (AD) will develop symptoms of psychosis (delusions and/or hallucinations) at some point in the course of the disease process [1]. Frequencies of these symptoms may vary, with approximately 19% of community dwelling AD patients thought to have delusions and approximately 14% thought to have hallucinations [2]. While delusions are more common than hallucinations, they are often associated with greater insight and therefore are less problematic [3]. Psychotic symptoms in AD are clinically significant, as they have been shown to be associated with increased caregiver burden [4], increased functional decline [4], and more rapid disease progression [5]. The pathobiological mechanisms that underlie psychotic symptoms are unclear, limiting our ability to manage and treat these symptoms.

Previous studies examining clustering of neuropsychiatric symptoms in AD have consistently demonstrated psychosis to be a reliable and valid symptom cluster [6–9]. In fact, Canevelli et al. [10]examined numerous behavioral and psychological sub syndromes using the Neuropsychiatric Inventory Questionnaire (NPI-Q) and found psychosis to be the only reliable and valid symptom cluster. Challenges relating to disentangling these symptoms, both clinically and neurobiologically, are that patients may have either delusions, hallucinations, or both, and it is possible that the different symptoms themselves may have different neurobiological correlates [11]. Most studies to date examining the neurobiological correlates of psychosis in AD have failed to separate delusions from hallucinations, thus limiting their conclusions.

Neuropathological studies investigating psychosis in AD have produced conflicting results, with some studies showing a correlation with severity of Alzheimer pathology [12–14], including neuritic plaques (NPs) and neurofibrillary tangles (NFTs), and others showing no such association [15, 16]. Sweet and colleagues [15] compared the brains of 24 AD patients with psychosis to 25 patients without psychosis, controlling for Lewy bodies (LBs), and found no correlation with NPs or NFTs, thus suggesting that psychotic symptoms in AD may be mediated by an alternate pathway. In contrast, Farber and colleagues [12] followed 109 AD subjects until death and found that the 63% of the sample that developed psychotic symptoms over their disease course had a 2.3 times higher NFT load when compared to patients without psychosis, independent of dementia severity. Murray and colleagues [14] compared AD patients with and without psychosis and found significantly higher rates of phosphorylated tau in the prefrontal cortex of the psychotic group. Meanwhile, Koppel and colleagues [17] found significantly higher levels of phosphorylated tau in the frontal cortex only in females AD patients, while psychosis in males was associated with the presence of α-synuclein pathology, suggesting a possible gender-related effect.

The disparities also appear in relation to cerebrospinal fluid (CSF). Skogseth and colleagues [18] examined CSF of 32 AD patients with neuropsychiatric symptoms and found no correlation between psychosis and tau or amyloid-β levels. However, Koppel and colleagues [13] compared CSF samples of AD patients with and without psychosis and found a correlation with tau levels. In a different disease group, Ballard and colleagues [16] examined 112 autopsy cases of patients with dementia with Lewy bodies (DLB) and found an inverse relationship between NFT burden and psychosis, suggesting that symptoms in this sample were not associated with increased Alzheimer pathology.

Imaging studies examining the correlates of psychosis in AD to date have produced mixed results. Ismail and colleagues [19] in a recent systematic review pooled imaging studies looking at delusions in AD and found an association with right frontal atrophy. The vast majority of imaging studies looking at psychosis in AD cross-sectionally have shown grey matter volume loss most prominent in the frontal lobes [20]. Rosenberg et al. [21] conducted a recent review of brain circuits associated with neuropsychiatric symptoms in AD and concluded that delusions are associated with abnormalities in the frontal cortex, including orbito-frontal, dorsolateral prefrontal, and inferior frontal. Additional brain structures include the right and left insula, and studies have been inconsistent with regard to lateralization. Our group also looked at this question with respect to delusions and found cross-sectionally grey matter volume loss increased in the right-frontal-temporal region [22]. We also examined patients longitudinally and found atrophy more pronounced in regions corresponding to the default mode network and cerebellum [23]. As none of the regions identified to date correspond to areas of atrophy in AD, combined they suggest a different mechanism may underlie the development of delusions in AD.

The association of psychosis with cerebrovascular risk factors is a recent topic of interest. Steinberg et al. [24] examined the correlation between cardiovascular risk factors and neuropsychiatric symptoms in AD patients, finding a correlation only with antihypertensive medications. There have been numerous studies suggesting white matter hyperintensities may play a role in depression and mild cognitive impairment as well as early AD [25]. However, few studies have examined postmortem markers of cerebrovascular disease in AD patients with psychosis.

Research to date has yet to reach a conclusion regarding the neuropathological correlates of psychosis in AD, although it is possible based on prior studies that Alzheimer pathology, cerebrovascular disease and LBs may all play a role. Limitations of previous studies include the fact that very few studies distinguished delusions from hallucinations; the sample sizes were small in many cases; most studies excluded analyses of vascular pathology; and correlations were made to clinical diagnosis without neuropathological confirmation. We plan to explore the role of Alzheimer pathology, LBs and cerebrovascular disease in the development of psychotic symptoms in AD using data from the National Alzheimer’s Coordinating Center (NACC) database, and to compare our findings in clinically diagnosed AD and neuropathologically confirmed patients as well as sub-stratify patients into delusions versus hallucinations. We predict that our findings will differ across the subgroups, reflecting different disease mechanisms for each symptom.

METHODS

RESULTS

We identified 1,073 subjects in the NACC database who fulfilled the criteria for analysis with available neuropathology data, including 890 cAD and 728 npAD subjects (Fig. 1). The cAD and npAD groups are overlapping groups as not all the cAD subjects had a neuropathological diagnosis of AD. Of the npAD patients who were not clinically diagnosed with AD (false negatives, n = 116), 42 (36.2%) were diagnosed with DLB, 11 (9.5%) were diagnosed with Parkinson’s disease dementia, 12 (10.3%) were diagnosed with vascular dementia, and 51 (44.0%) were diagnosed with possible AD. The average MMSE score of the npAD group was 13.6±8.0, so the group appeared to meet the threshold for dementia.

DISCUSSION

Our samples of cAD and npAD patients with and without psychosis were matched demographically with regards to age and education. Concordant clinical findings observed in both cAD and npAD samples included the fact that AD+P patients were not more functionally impaired at their last clinical visit compared to the AD-P group, contrary to existing literature that suggests psychosis represents a more severe form of AD [1, 5, 6, 39]. However, in breaking down symptoms into delusions and hallucinations, we found a clear separation in terms of impact on disease severity. AD patients with hallucinations, both with and without delusions, had more advanced disease (higher CDR scores) and were more cognitively and functionally impaired (lower MMSE scores and higher FAQs) at the last study visit prior to death, although in the latter there was no association observed between hallucinations and cognition in the npAD group. Contrary to previous studies [4], AD+D subjects were not more functionally impaired and, in fact, had significantly less clinical disease severity at the visit prior to death when compared to AD–P subjects as measured by the MMSE and global CDR. One potential explanation for this finding is that patients with delusions had a longer interval between the last clinical assessments to time of death, so it is possible that the disease progression was less advanced at the last assessment. Another explanation for this finding is that a certain level of cognitive ability is required to form and express a delusion, and that these symptoms remit as cognition declines.

In terms of discordant clinical findings between the cAD and npAD samples, patients in the cAD sample with psychosis had a longer duration of illness. This could indicate slower progression and thus less disease severity in patients with AD and psychosis. This would be contrary to some existing literature, which suggests that patients with psychosis undergo a more rapid disease progression [5]. One potential explanation is that patients with psychosis, due to the nature of their symptoms, may seek medical attention earlier, thus creating an artificially longer disease duration. An alternative explanation is that perhaps the rate of decline varies at different stages of the AD continuum, with more rapid progression in earlier stages. Currently, the literature is equivocal in regards to associations of AD+P and age of onset and illness duration [1].

In terms of the neuropathological analyses, concordant findings between the cAD and npAD sample included the observation that LBs were correlated with psychosis. The cAD sample showed a stronger correlation with delusions while the npAD sample showed a stronger association with hallucinations. The finding of association of psychosis with LB is consistent with Jacobson et al. [40], who found a strong correlation between visual hallucinations and LBs, both in patients with DLB and clinical AD, and a rather poor correlation with other markers of AD neuropathology (NFTs and NPs). These findings are also consistent with the work of Ballard et al. [16] who concluded that hallucinations in patients with DLB are likely mediated by an alternate disease mechanism from AD pathology, as he observed them to be inversely related to both NPs and NFTs.

As well, in terms of vascular pathology, psychosis was not correlated with specific lesions such as stroke and lacunar infarcts, but both cohorts showed a correlation between AD+P/AD+DH and SAL. Hypertension (in AD+P), diabetes (in AD+DH) and age at quitting smoking (AD+H) were also positively correlated with psychotic symptoms. Overall, all the vascular risk factors appeared to be positively associated with psychosis and its subgroups, although many of them did not reach significance. This provides the first compelling evidence that psychotic symptoms in AD patients may be mediated by a vascular mechanism. This is somewhat consistent with the findings of Steinberg et al. [24] who correlated NPI-Q scores, specifically the affective cluster, with antihypertensive use but found no other associations. Bidzan et al. [41] similarly found that vascular factors, as measured by the Hachinski score, were correlated with depression and anxiety, but not with hallucinations or delusions. Unfortunately we did not look at anxiety and depressive symptoms in our study so it is hard to comment on how strong this correlation is in relation to affective symptoms. Previous research found that vascular burden may disrupt local connections between the frontal lobe and subcortical areas through disruptions of cholinergic transmissions, which may intensify neuropsychiatric symptoms [42]. Our findings suggest that targeting the modifiable aspects, such as managing vascular risk factors (e.g., controlling diabetes and hypertension), may reduce the risk of psychosis.

We found a number of significant discordant neuropathological findings when comparing the cAD and npAD groups. In cAD patients, psychosis was associated with more advanced Alzheimer pathology, including Braak stage, NIA Reagan scores, CERAD scores, and NPs. Specifically, advanced Braak stage was associated with delusions with or without hallucinations; NIA Reagan scores correlated with all psychotic groups; CERAD scores correlated with both delusions and hallucinations; and NPs correlated with delusions only. These findings are consistent with prior studies [12, 14] and support the concept that AD+P may represent a more aggressive form of the disease, associated with more advanced neuropathology [5]. Importantly, however, there were no significant correlations between Alzheimer pathology and psychotic symptoms in patients with npAD, suggesting that the observed findings in cAD may be driven by misdiagnosis, specifically false positive diagnosis of AD. While it would be of interest to know what the ultimate diagnosis was in the case of patients who were false positive, unfortunately this data was not available. Our paper is the first to highlight the differences between cAD and npAD with regard to psychosis and Alzheimer pathology and may well explain the discrepancy in the field to date. The majority of studies reporting a positive [12–14] and negative [15, 18] correlation with Alzheimer pathology relied on clinical diagnosis as opposed to neuropathologically confirmed diagnosis. However, clinical and neuropathological expression of AD is not always correlated. Monsell et al. found that some patients who have AD neuropathology are asymptomatic despite neuritic plaque burden, and suggested that neurofibrillary tangles has a larger role in determining the clinical expression of AD [43]. Another potential explanation for the discrepancy between cAD and npAD may reflect differences in disease severity. In order to meet criteria for npAD, patients by definition must have severe disease while patients with cAD may have had a broader range of disease severity, thus making it easier for an association to be detected.

There are some limitations of the current study. We relied on data obtained from different centers across the United States, so it is possible that there was some variation in the collection of data. We also used many univariate tests in our statistical analysis, which increased the probability of a Type I error. Moreover, we used clinical data obtained from the last study visit, whether or not the patients had active psychotic symptoms at that time, as opposed to looking at the course of symptoms over time. There was also variation in the time between last visit and time of death, which may have affected the results. Additional, prospective studies will be needed to better define different trajectories of disease in patients with and without psychosis. It is possible that some patients with active psychotic symptoms in fact had delirium, given it may be hard to differentiate the two conditions in advanced dementia. The NPI-Q captures symptoms that have occurred in the previous month, so it is possible that some informant reports may reflect a delirium that had resolved by the time of the study visit. Also, it is possible that clinical differences between patients were not detected secondary to the floor effects of the instruments used (MMSE, CDR, FAQ) given that most patients had advanced dementia.

In summary, AD patients with psychotic symptoms do not appear to differ demographically or clinically from AD patients without psychosis, but further breakdown of psychotic symptoms showed that patients with hallucinations appear more cognitively and functionally impaired at last visit prior to death, while there is some evidence to suggest that delusions are associated with better cognitive and functional status at last visit prior to death. We demonstrated that in the clinically diagnosed AD population, including false positive diagnosis of AD, psychosis is associated with AD pathology. However, in the pathologically diagnosed AD population, psychotic symptoms do not vary with the severity of AD pathology but instead are associated with LBs, subcortical ischemic vasculopathy, and cerebrovascular risk factors including hypertension, smoking, and diabetes. Combined, these findings suggest that LBs and cerebrovascular disease are important risk modifiers of psychosisin AD.