Hallucinations and delusions are associated with elevated tau PET signal independent of age, clinical severity, and amyloid burden

Abstract

INTRODUCTION

Psychosis in Alzheimer's disease (AD) is associated with worse outcomes, yet no established biomarkers exist for early diagnosis and intervention. We compared tau positron emission tomography (PET) burden across older individuals with and without psychotic symptoms.

METHODS

[18F]AV1451 tau PET binding was compared between 32 Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects with psychotic symptoms (delusions and/or hallucinations) and 32 ADNI subjects without psychotic symptoms, matched for age, sex, race/ethnicity, and clinical severity. Tau was assessed in a priori regions of interest (ROIs) and in voxelwise analyses, both corrected for amyloid PET burden.

RESULTS

Tau was greater in individuals with psychotic symptoms in the amygdala, hippocampus, frontal cortex, and early, middle, and late Braak stage regions in primary analyses. When considering subgroups, tau binding was greatest in those with concurrent delusions.

DISCUSSION

Greater than expected tau burden for age, clinical severity, and amyloid burden may be relevant for psychotic symptoms in older adults.

Highlights

• Tau positron emission tomography (PET) was elevated in individuals with psychosis

• Elevated tau was independent of Alzheimer's disease (AD) clinical severity and amyloid burden

• There was variability in the regional distribution depending on psychosis type

1. INTRODUCTION

Psychosis is a symptom cluster including false beliefs (e.g., delusions) and/or a false perception of events or objects in the surrounding environment (e.g., hallucinations). ¹ The prevalence of psychotic symptoms within Alzheimer's disease (AD) is between 40% and 60%, typically developing in advanced clinical stages (e.g., Clinical Dementia Rating [CDR] ≥ 2). ² , ³ Neuropsychiatric symptoms, including psychosis, increase caregiver burden considerably ⁴ and these patients often need additional attention leading to long‐term care placement. The specific etiology for psychotic symptoms in AD has yet to be elucidated and is likely complicated by the wide range of reported symptoms. There is a need to identify relevant biomarkers of psychosis to promote earlier diagnosis, more appropriate treatment planning, and better outcomes. ¹

AD dementia is defined clinically as cognitive and functional impairment ⁵ and is confirmed histopathologically as amyloid‐β plaques and neurofibrillary tau tangles in key brain regions at autopsy. ⁶ Tau may be the pathogenic driver underlying the various brain changes associated with psychotic symptoms with progression of AD, as tau is more closely associated with neurodegeneration and clinical severity compared to amyloid pathology in AD. ⁷ , ⁸ , ⁹ Even within AD, different spatial distributions of tau are associated with different cognitive symptoms, suggesting a strong link between the magnitude and regionality of tau with clinical presentation. ⁸ , ⁹

Several studies have investigated psychotic symptoms (specifically, hallucinations, and delusions) and tau in AD. At autopsy, psychotic symptoms were associated with greater hyperphosphorylated tau and neurofibrillary tangle burden in the right frontal cortex. ¹⁰ Using in vivo positron emission tomography (PET) to measure tau burden in the Alzheimer's Disease Neuroimaging Initiative (ADNI), investigators demonstrated that AD patients with psychotic symptoms had greater tau in Braak stage regions compared to AD patients without psychotic symptoms. However, there was no adjustment for clinical severity, and it remains unclear whether tau was a driver of their psychotic symptoms or if it was simply coincident with their more advanced CDR score. ¹¹ In the Translational Biomarkers in Aging and Dementia (TRIAD) cohort, hallucinations and delusions were not associated with tau PET burden in composite Braak stages after adjusting for clinical severity. ¹² However, hallucinations were associated with greater tau in the medial occipital lobe, while delusions were associated with greater tau in the cuneus and occipital lobe, ¹³ suggesting more regionally specific associations with particular psychotic symptoms. Longitudinally, elevated tau PET burden in advanced Braak stages was associated with greater progression of overall neuropsychiatric symptoms. ¹²

In this study, we assessed regional differences in tau PET burden between older adults with psychotic symptoms (+P) and those without psychotic symptoms (−P), focusing on the endorsement of hallucinations and delusions within ADNI. We matched participants for clinical severity to evaluate the regional specificity of the association between tau and psychotic symptoms, above and beyond the tau burden across progressive Braak stages that is characteristic of AD progression. We hypothesized that tau in composite Braak stage regions would not be associated with hallucinations and delusions between CDR matched groups, but that tau in regions processing emotionally salient stimuli (e.g., amygdala, frontal cortex) will have increased tau PET burden.

2. METHODS

2.1. Participants

Data used in the preparation of this article were obtained from the ADNI database (adni.loni.usc.edu). The ADNI was launched in 2003 as a public‐private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early AD. For up‐to‐date information, see www.adni‐info.org.

We included participants who underwent tau PET imaging, amyloid PET imaging, structural T1 MR imaging, and CDR evaluation. As per ADNI inclusion criteria, participants were excluded if they had a long history of ongoing psychotic symptoms associated with other psychiatric disorders. ¹⁴ We identified 32 participants (+P) who endorsed psychotic symptoms (i.e., delusions and/or hallucinations), based on the Neuropsychiatric Inventory (NPI) ¹⁵ or the Neuropsychiatric Inventory Questionnaire (NPI‐Q) ¹⁶ depending on their ADNI visit. We used participant and/or co‐participant responses to NPI/NPI‐Q questions A and B, which ask for a binary endorsement of delusions and hallucinations, respectively. If an endorsement is made, severity is also assessed; however, due to the truncated range and missingness (n = 6) in our sample, severity was not assessed as part of this analysis, but is available in Supplemental Materials. All participants endorsed psychotic symptoms in at least one visit with ADNI. Briefly, the severity for hallucinations ranged from 0 to 2 with a mean of 0.38 and standard deviation of 0.57, while the severity for delusions ranged from 0 to 3 with a mean of 1.04 and standard deviation of 0.87. Multiple episodes of psychosis, particularly those that persist intermittently for 4 or more weeks, are diagnostic criteria for psychosis in neurocognitive disorders. This is only partially captured by the NPI/NPIQ question that asks for endorsements within the last month. This ADNI sample included 15 participants with endorsement of hallucinations (3), delusions (12), or both at more than 1 visit throughout their participation in ADNI. We then matched participants who did not endorse either psychotic symptom (n = 32; −P) algorithmically by creating a demographic profile for each +P participant and matched each with a −P participant based on CDR, age, gender, race, and ethnicity (MatchIt package, R version 4.2.1). The −P group never endorsed hallucinations and delusions throughout their participation in ADNI. For the analysis, we used neuroimaging data and CDR evaluations that were nearest to the NPI/NPIQ data.

2.2. MRI acquisition

Structural T1 MR imaging was conducted under the ADNI protocol (3 Tesla; magnetization prepared rapid gradient echo (MPRAGE) sequence: repetition time (TR) = 2300 ms, echo time (TE) =  minimum full echo, inversion time (TI)  =  900 ms, field of view (FOV) =  208 × 240 × 256 mm³, voxel resolution = 1 × 1 × 1 mm³). Pmod (version 4.2; Pmod Technologies LLC, Switzerland) anatomical segmentation (cortical Hammers‐N30R83‐1MM and cerebellar AAL‐1MM atlas) of the T1 scan was used for tau PET quantification in volume‐weighted bilateral regions of interest (ROIs), including hippocampus, amygdala, frontal lobe, temporal lobe, and parietal lobe. Composite Braak I/II (Hippocampus and parahippocampus), Braak III/IV (fusiform gyrus, lingual gyrus and amygdala), and Braak V/VI (Frontal, Superior Temporal, and Parietal Lobes) ROIs were modified from previously published Freesurfer‐based segmentations to better reflect PMOD‐based segmentations. ¹⁷ , ¹⁸

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature using traditional search engines (e.g., PubMed) and published conference abstracts. While there are many publications on the prevalence of psychotic symptoms and their neurobiological correlates in older adults and neurodegenerative disease, there are not many articles regarding tau positron emission tomography (PET) neuroimaging of psychotic symptoms. Converging evidence from relevant articles are appropriately cited.

2. Interpretation: Tau burden in key regions may be a driver of structural and functional abnormalities that are commonly reported in individuals with psychotic symptoms.

3. Future directions: There may be variability in the magnitude and spatial pattern of Alzheimer's disease (AD) ‐type tau accumulation beyond the well‐established Braak staging in individuals with psychotic symptoms. Key considerations for future work include individual brain regions (as opposed to typical composites used in AD) and heterogeneity in psychotic symptom type, onset, severity, and medication use.

2.3. PET acquisition

[18F]AV1451 tau PET imaging was conducted under the ADNI protocol (370 MBq (10 mCi); 30‐min dynamic scan, comprised of six 5‐min frames acquired 75 min postinjection). Using Pmod (version 4.2, Pmod Technologies LLC, Switzerland), we calculated standardized uptake value ratio (SUVR; 75–115 min postinjection; inferior cerebellar gray matter reference region) on the voxel‐level.

[18F]Florbetaben and [18F]Florbetapir ([18F]AV45) amyloid PET imaging was conducted under the ADNI protocol ([18F]Florbetaben: 300 MBq (8.1 mCi); four 5‐min frames acquired 90 min postinjection; [18F]Florbetapir: 370 MBq (10.0 mCi); four 5‐min frames acquired 50 min postinjection). Amyloid PET SUVR was calculated within a global composite ROI including Thal phase regions and was harmonized across amyloid PET tracers into the Centiloid scale as previously described. ¹⁹

2.4. Statistical analysis

We assessed our hypothesis that tau burden would be elevated in the +P group compared to CDR‐matched –P group primarily through paired t‐tests on the ROI level, adjusting for global amyloid burden. Adjusting for amyloid burden further removed variance in tau burden due to AD in a similar manner as CDR matching +P and –P groups, but avoids limiting the sample size via exact matching and unwanted variance from inexact matching. We then performed a series of exploratory analyses to further understand (1) the potential regionality and laterality beyond our a prior ROIs, and (2) the potential bias from grouping delusions and hallucinations together regardless of whether they were endorsed before, after, or concurrently within 1 year of the tau PET scan. First, we performed a paired t‐test on the voxel‐level (unadjusted p‐value < 0.001), adjusting for global amyloid burden. Second, there were five subgroups in the exploratory analyses—no psychotic symptoms (−P), psychotic symptoms endorsed after tau PET (n = 10), psychotic symptoms endorsed prior to tau PET (n = 7), concurrent delusions (n = 12), and concurrent hallucinations (n = 3). One participant endorsed concurrent delusions and concurrent hallucinations. They were included in the concurrent delusions group because that was the more common symptom overall, but sensitivity analyses including them in the concurrent hallucinations group did not change results. For a full breakdown of psychotic symptoms and timing relative to tau PET, see Supplemental tables 1 and 2. To explore subgroups, we report estimated marginal means for ROI‐level tau burden, adjusting for age, sex, race, ethnicity, education, CDR, and global amyloid burden. We further descriptively present participant level scans as sample sizes in exploratory groups were too small for voxel‐wise analyses. All ROI‐level statistics were performed with the lme4, emmeans, and p.adjust packages in R version 4.2.1, correcting for multiple comparisons using the Benjamini‐Hochberg method, and all exploratory voxelwise statistics were performed in SPM12 (MATLAB), uncorrected for multiple comparisons using a significance threshold of 0.001.

3. RESULTS

Sixty‐four participants from ADNI were evaluated in this study, with 32 reporting psychotic symptoms (+P) and 32 reporting no psychotic symptoms (−P), matched for CDR and demographics (Table 1). In this analytic sample, the average age was 77 years old, 52% were women, the mean education was 16 years, and the majority self‐identified as non‐Hispanic White. Demographic variables were not different between groups by design. Similarly, global CDR was not different between groups, but ranged from CDR 0 to CDR 2. The majority of participants were CDR 0.5 (53%) and CDR 1 (31%). Global amyloid burden was nonsignificantly greater in the +P group compared to the –P group (Table 1).

TABLE 1.

Demographic characteristics for the entire analytic sample and by individuals with no psychotic symptoms and individuals with psychotic symptoms.

Parameter	No history of psychosis (N = 32)	Psychotic symptoms (N = 32)	Overall (N = 64)
Age at tau scan
Mean (SD)	76.6 (6.07)	77.0 (6.37)	76.8 (6.18)
Median [Min, Max]	77.0 [65.0, 94.0]	77.0 [65.0, 89.0]	77.0 [65.0, 94.0]
Gender
Female	14	14	28
Male	18	18	36
Education
Mean (SD)	15.5 (2.31)	15.7 (2.16)	15.6
Median [min, max]	16.0 [12.0, 20.0]	16.0 [12.0, 20,0]	16 [12.0, 20.0]
MMSE
Mean (SD)	25.0 (4.70)	23.8 (4.80)	24.4 (4.75)
Median [min, max]	25.5 [13, 30]	24.0 [9, 30]	24.5 [9, 30]
Race
White	32 (100%)	30 (93.8%)	62 (96.9%)
Asian	0 (0%)	1 (3.1%)	1 (1.6%)
Black or African American	0 (0%)	1 (3.1%)	1 (1.6%)
Ethnicity
Hispanic or Latino	2 (6.25%)	1 (3.13%)	3 (4.92%)
Not Hispanic or Latino	30 (93.75%)	31 (96.88%)	61 (95.31%)
Amyloid Centiloids
Mean (SD)	73.6 (40.2)	72.5 (43.6)	73.0 (41.6)
Median [min, max]	77.3 [−4.27, 140]	71.2 [−3.31, 193]	73.6 [−4.27, 193]
CDR global
0	2 (6.3%)	2 (6.3%)	4 (6.3%)
0.5	17 (53.1%)	17 (53.1%)	34 (53.1)
1	10 (31.3%)	10 (31.3%)	10 (31.3%)
2	3 (9.4%)	3 (9.4%)	3 (9.4%)
Days between tau PET and psychotic symptoms
Mean (SD)		244 (1,238)

After removing CDR and demographic effects through matching and amyloid effects through covariate adjustment, tau burden was significantly greater in the +P group compared to the −P group in the amygdala (0.21 SUVR [0.05, 0.37] p_adj = 0.043), hippocampus (0.14 SUVR [0.02, 0.25], p_adj = 0.043), frontal lobe (0.14 SUVR [0.02, 0.25], p_adj = 0.043), and early (0.17 SUVR [0.03, 0.30], p_adj = 0.043), middle (0.21 SUVR [0.02, 0.41], p_adj = 0.045), and late (0.14 SUVR [0.02, 0.25], p_adj = 0.043) Braak stage regions, but not the temporal (0.19 SUVR [‐0.009, 0.38], p_adj = 0.061) or parietal lobes (0.14 SUVR [0.002, 0.27], p_adj = 0.055; Figure 1, Table 2). The standardized effect size was similar across regions, demonstrating 0.44–0.55 standard deviation increases in tau burden for the +P group compared to −P group (Table 2). In exploratory, but complementary voxelwise analyses (Figure 2), clusters of elevated tau burden in the +P group compared to the −P group were evident bilaterally in the amygdala, hippocampus, parahippocampus, thalamus, basal ganglia (putamen, caudate, pallidum, nucleus accumbens), cingulate gyrus, temporal cortex (entorhinal cortex, fusiform, lingual, inferior temporal, temporal pole), orbitofrontal cortex, lateral occipital cortex and unilaterally in the right inferior temporal cortex, right inferior parietal cortex, and right frontal cortex (rostral middle, superior, precentral). We also observed small clusters in the corpus callosum and superior cerebellum. Further breaking down the +P group by endorsed symptoms in relation to tau PET timing (Supplemental Tables 1 and 2), tau burden in amygdala, frontal cortex, temporal cortex, parietal cortex, and early, middle, and late Braak stage regions was significantly greater in the concurrent delusions group compared to the −P group (Table 3). Participant level tau SUVR images (Figure 3) demonstrate variability in regionality, including laterality, and magnitude that differs from typical Braak staging.

FIGURE 1.

Distribution of tau burden (SUVR) between individuals with no psychotic symptoms (−P) and individuals with psychotic symptoms (+P), adjusted for amyloid burden. SUVR, standardized uptake value ratio.

TABLE 2.

Tau burden (SUVR) group means and 95% confidence interval for individuals with no psychotic symptoms and individuals with psychotic symptoms, adjusted for amyloid burden, within different ROIs.

ROI	Unstandardized estimate	Standardized estimate	T‐statistic, p‐value
Amygdala	0.21 [0.05, 0.37]	0.52 [0.12, 0.92]	T = 2.6, p = 0.043
Hippocampus	0.14 [0.02, 0.25]	0.53 [0.06, 0.99]	T = 2.3, p = 0.043
Frontal lobe	0.14 [0.02, 0.25]	0.54 [0.10, 0.98]	T = 2.5, p = 0.043
Temporal lobe	0.19 [−0.009, 0.38]	0.46 [−0.02, 0.93]	T = 1.9, p = 0.061
Parietal lobe	0.14 [0.002, 0.27]	0.44 [0.005, 0.87]	T = 2.1, p = 0.055
Braak I/II	0.17 [0.03, 0.30]	0.55 [0.11, 1.0]	T = 2.5, p = 0.043
Braak III/IV	0.21 [0.02, 0.41]	0.50 [0.04, 0.96]	T = 2.2, p = 0.045
Braak V/VI	0.14 [0.02, 0.25]	0.52 [0.08, 0.95]	T = 2.4, p = 0.043

Note: Unstandardized (units: SUVR) and standardized (units: standard deviations) difference in tau burden with their 95% confidence intervals, adjusted for amyloid burden. Reference was the CDR and demographics‐matched group without psychotic symptoms. Separate models were run for each ROI and p‐values were adjusted using the Benjamini‐Hochberg method.

Abbreviations: CDR, Clinical Dementia Rating; ROI, region of interest; SUVR, standardized uptake value ratio.

FIGURE 2.

Voxelwise paired t‐test between CDR and demographic matched individuals with psychotic symptoms (+P) and individuals without psychotic symptoms (−P), adjusted for amyloid burden. Voxels in green show areas of greater tau burden in the +P group compared to the −P group. No significant voxels found in the other direction (i.e., greater tau burden in the −P group compared to +P group). Significance was determined at p < 0.01 without multiple comparisons correction. CDR, Clinical Dementia Rating.

TABLE 3.

Tau burden (SUVR) group means and their 95% confidence interval for individuals with no psychotic symptoms, post‐tau psychotic symptoms, pre‐tau psychotic symptoms, concurrent delusions, and concurrent hallucinations, adjusted for amyloid burden, within different ROIs.

ROI	Individuals with no psychotic symptoms	Individuals with post‐tau psychotic symptoms	Individuals with pre‐tau psychotic symptoms	Individuals with concurrent delusions	Individuals with concurrent hallucinations	F‐statistic, p‐value
Amygdala mean 95% CI	1.39 [0.96 –1.83]	1.63 [1.11–2.15]	1.48 [0.98–1.97]	1.80 [1.36–2.23]	1.38 [0.69–2.08]	F = 2.7 p = 0.047
Hippocampus mean 95% CI	1.24 [0.95–1.54]	1.44 [1.09–1.79]	1.30 [0.96–1.64]	1.51 [1.21–1.81]	1.18 [0.71–1.66]	F = 2.4 p = 0.059
Frontal lobe mean 95% CI	1.09 [0.82–1.35]	1.19 [0.87–1.51]	1.12 [0.81–1.42]	1.36 [1.09–1.63]	1.05 [0.63–1.48]	F = 4.0 p = 0.025
Temporal lobe mean 95% CI	1.37 [0.94–1.81]	1.44 [0.92–1.96]	1.34 [0.84–1.84]	1.76 [1.32–2.21]	1.29 [0.59–1.99]	F = 3.4 p = 0.025
Parietal lobe mean 95% CI	1.29 [0.97–1.61]	1.37 [0.99–1.75]	1.34 [0.97–1.70]	1.58 [1.26–1.91]	1.19 [0.68–1.70]	F = 3.4 p = 0.025
Braak I/II mean 95% CI	1.28 [0.97–1.59]	1.44 [1.07–1.81]	1.33 [0.97–1.69]	1.58 [1.27–1.90]	1.21 [0.72–1.71]	F = 3.3 p = 0.025
Braak III/IV mean 95% CI	1.54 [1.09–1.99]	1.65 [1.11–2.20]	1.56 [1.04–2.08]	1.99 [1.53–2.45]	1.52 [0.79–2.24]	F = 3.5 p = 0.025
Braak V/VI mean 95% CI	1.16 [0.89–1.43]	1.25 [0.93–1.58]	1.19 [0.88–1.50]	1.44 [1.17–1.72]	1.10 [0.67–1.54]	F = 4.1 p = 0.025

Note: Tau burden (SUVR) group estimated marginal means and their 95% confidence interval for individuals with no psychotic symptoms, post‐tau psychotic symptoms, pre‐tau psychotic symptoms, concurrent delusions, and concurrent hallucinations, adjusted for age, sex, CDR, race/ethnicity, and amyloid burden. Separate models were run for each ROI and p‐values were adjusted using the Benjamini‐Hochberg method.

Abbreviations: CDR, Clinical Dementia Rating; ROI, region of interest; SUVR, standardized uptake value ratio.

FIGURE 3.

Representative slice from participant level tau SUVR scans in standardized space for exploratory psychotic symptoms subgroups. Age range is provided to prevent identification in addition to CDR and amyloid burden (CL). The yellow box indicates one participant endorsed both concurrent delusions and concurrent hallucinations, but was included in the concurrent delusions group for analyses. CDR, Clinical Dementia Rating; CL, Centiloid; SUVR, standardized uptake value ratio.

4. DISCUSSION

Neurofibrillary tau tangles, measured with AV‐1451 tau PET, were elevated in the amygdala, hippocampus, frontal lobe, and early, middle, and late Braak stage regions, above and beyond the level expected for age, amyloid burden, and clinical severity in older adults. Elevated tau burden may contribute to the structural, functional, and neuroinflammatory changes commonly reported in these same regions for individuals with psychotic symptoms. ²⁰ , ²¹ , ²² , ²³ , ²⁴ As tau PET may become increasingly incorporated into the clinical workflow for the early differential diagnosis of dementia patients, it will be critical to understand how tau pathology may contribute to symptom development and progression to reduce patient and caregiver burden. Tau PET burden, greater than expected for age, amyloid, and clinical severity, could be a useful biomarker of psychosis, which may necessitate a more proactive approach to symptom management through medication and caregiver education. Including psychosis as a surrogate endpoint for anti‐tau clinical trials will provide further insight into this potential mechanistic link, particularly as it relates to tau clearance in the amygdala and hippocampus.

Our results add to the growing body of work that aims to elucidate the neurobiological correlates of psychotic symptoms. Previous work has identified elevated tau PET burden across Braak stage regions without correction for disease severity in ADNI ¹¹ and no association with Braak stage regions for hallucinations and delusions specifically in TRIAD. ¹² , ²⁵ After adjusting for age, clinical severity, and amyloid burden to understand psychosis‐specific tau burden, we found evidence of elevated tau in the amygdala, hippocampus, frontal lobe, as well as early, middle, and late Braak stage regions. While approximately equal standardized effect sizes across regions indicate that the tau elevation was not strongest in regions involved in emotional processing compared to composite Braak stage regions, our results still identify tau elevation in limbic regions as one key feature of individuals with psychosis. The amygdala is often implicated in neuroimaging studies of psychosis because of its role in moderating the stress response to an emotionally salient environmental cue (e.g., delusions, hallucinations). ²⁶ , ²⁷ , ²⁸ Limbic regions have neuronal projections through the thalamus and basal ganglia to cortical regions, including the frontotemporal lobes that are also involved in the salience network. ²⁹ Exploratory clusters of elevated tau in those with psychotic symptoms compared to those without were observed in specific subregions of the frontal and temporal cortex as well as the thalamus, basal ganglia, cingulate gyrus, and occipital cortex. Additionally, right‐sided laterality was observed in subregions of the frontal, temporal, and parietal cortices. Smaller clusters in corpus callosum and superior cerebellum may be noise‐related at this significance level and sample size, but other studies have demonstrated brain changes in the cerebellum ³⁰ and white matter ³¹ , ³² in relation to psychosis in non‐AD patients.

ADNI was not primarily designed to anchor tau PET imaging and neuropsychiatric symptoms, yet still represents a rich dataset to generate mechanistic hypotheses from exploratory analyses. The complex spatial heterogeneity of elevated tau in subgroups with psychosis is presented in Figure 3. Across subgroups, individuals with concurrent delusions demonstrated the highest tau burden; however, we are unable to disentangle this effect from larger sample size, greater severity (delusions: 1.04 ± 0.87 vs hallucinations: 0.38 ± 0.57), or overall neuropsychiatric profile (Supplemental Materials). The timing of concurrent delusions (59 ± 93 days) compared to concurrent hallucinations (23 ± 20 days) likely did not influence this finding. Visual inspection of Tau PET reveals substantial heterogeneity in tau burden (severity and pattern) among +P participants with concurrent delusions, including a small subset with markedly elevated uptake. Notably, these individuals were themselves heterogeneous with respect to age, amyloid burden, and neuropsychiatric symptom profiles, suggesting that no single clinical or biomarker feature accounts for the observed variability. If psychosis is a consequence of tau, tau PET would be elevated even in the post‐tau psychosis group, which we observed after adjusting for age, CDR, and amyloid burden among other covariates; alternatively, if psychosis is a driver of tau, through aberrant network connectivity, this would not be the case. Because psychosis in AD could reflect a more severe or accelerated biological, cognitive, and/or functional trajectory, elucidating the temporal relationship between disease course and psychosis will require longitudinal studies incorporating repeated tau PET imaging, cognitive performance, and functional measures before, concurrently, and after psychosis. It will be critical to collect information regarding the age of first psychosis, duration and severity of psychosis endorsements, and type of delusions/hallucinations (e.g., persecution, jealousy, auditory, visual) to understand the heterogeneity. While we do not have information on the reason for dropout, we speculate that the burden on participants may be too high as the onset of neuropsychiatric symptoms often coincides with transition to long‐term care, particularly for more severe cases. This highlights the need to develop timely and effect treatments. In our sample, 15 participants endorsed at two or more visits. A sensitivity analysis demonstrated similar, but nonsignificant findings such that the effect was largest in the amygdala (0.405) followed by the hippocampus (0.172). This could be due to the smaller sample size, individuals with milder forms of psychosis being able to stay in the study for more than two timepoints, or both. Regardless, a larger elevation in tau with delusions compared to hallucinations is an important observation to study further. Previous studies found that delusions were among the most strongly associated NPI symptoms with elevated tau burden in the cuneus and occipital lobe, ²⁵ and an increasing dependence and faster cognitive decline in individuals with concurrent delusions compared to hallucinations. ³³

While neuropsychiatric symptoms encompass a broad and clinically important spectrum, this study focuses specifically on hallucinations and delusions, as these psychotic features represent some of the most severe manifestations and lead to the greatest patient burden. We cannot rule out that co‐occurring NPS contributed to the elevated tau PET signal in key regions, but co‐occurring NPS were also endorsed in the group without psychosis, albeit at a lower frequency (e.g., apathy and irritability were the two most common endorsements in the psychosis group (88% each), but these were also present in the matched groups (47% and 56%, respectively; p < 0.05)).

Psychotic symptom severity was associated with tau, but not amyloid or neurodegeneration. ¹³ Tau pathology of the AD type (i.e., 3R/4R isoform, paired helical filaments, neurofibrillary tangles) is found in older individuals without cognitive impairment, ³⁴ , ³⁵ , ³⁶ individuals with primary age related tauopathy, ³⁷ individuals with traumatic brain injury, ³⁸ specific mutations of frontotemporal dementia, ³⁹ , ⁴⁰ and dementia with Lewy bodies (DLB). ⁴¹ , ⁴² The magnitude and spatial distribution of AD‐type tau pathology is relevant for different symptom profiles, ⁸ , ⁹ , ⁴³ even outside of the AD continuum, which motivated the inclusion of individuals with a CDR of 0 and non‐Braak regions. However, we further investigated these two sources of potential bias. Two +P participants who were CDR 0 at baseline remained CDR 0 throughout their participation in ADNI (as did their matched ‐P pairs). It is unlikely that clinical progression in these two +P individuals, but not their −P matched pairs, biased our results. Similarly, a sensitivity analysis restricted to amyloid‐positive individuals on the Alzheimer's continuum (Centiloid > 20; n = 27 +P and −P pairs) demonstrated a similar pattern of results as in the whole sample with slightly larger effect sizes.

In nearly 5500 participants from the National Alzheimer's Coordinating Center, DLB pathology was associated with psychosis as the first presenting NPS symptom. ⁴⁴ The majority of our ADNI sample was enrolled prior to the systematic collection of DLB related data and information on DLB status is largely unavailable. Using the NPI/NPIQ as a proxy, motor symptoms were endorsed by 44% in the group with psychosis and 22% in the group without psychosis (χ ²(1) = 2.6, p = 0.11). We cannot rule out DLB as the driver of psychotic symptoms in these individuals, but the ADNI recruitment criteria may minimize this possibility by excluding any significant neurological disease other than AD (e.g., Parkinson's disease, multi‐infarct dementia, Huntington's disease, normal pressure hydrocephalus, brain tumor, progressive supranuclear palsy, seizure disorder, subdural hematoma, multiple sclerosis, history of significant head trauma followed by persistent neurological deficits, known structural brain abnormalities) as well as any significant systemic illness or unstable medical condition. Even if DLB were present in some of these individuals, tau PET may still be a meaningful biomarker for psychosis through its interaction with Lewy bodies, ⁴⁵ particularly in the amygdala. ⁴⁶ Future studies should investigate multiple pathologies (e.g., tau and DLB in the amyloid negative context) to understand if specific pathologies or combinations of pathologies are associated with earlier onset of psychosis, faster progression of psychosis, and worse clinical outcomes.

Limitations of the study include the complexity of neuropsychiatric symptom profiles, the representativeness of ADNI, the cross‐sectional analyses, and the small sample size. The overall neuropsychiatric profiles of individuals with hallucinations and/or delusions and their timing of endorsements relative to tau PET contribute to within‐group variability, which we were only able to explore given the small sample size and highlight in Figure 3. Additionally, the ADNI selection criteria may be too stringent to capture the full range of neuropsychiatric symptoms that are present across the AD continuum, excluding those with severe symptoms and higher healthcare needs. These factors may bias our results towards the null. Despite the small cross‐sectional sample, our results converge with other studies that identify neurobiological changes in the amygdala ⁴⁷ and hippocampus, ²⁸ thalamus, ⁴⁸ putamen, ⁴⁹ frontotemporal lobes, ²⁰ and late Braak stage regions, ¹² , ²⁵ as well as increasing dependence and faster cognitive decline in individuals with concurrent delusions compared to hallucinations. ³³

Longer longitudinal studies are needed to observe consecutive endorsements of psychosis, or none at all, to refine group definitions for those with psychosis and those without psychosis. Larger longitudinal studies with symptom type, onset, duration, and severity, concomitant medication use, tau, and other comorbid pathologies are needed to validate these cross‐sectional findings and identify symptom‐specific regionality. Future studies should also investigate the link between 3R/4R neurofibrillary tangles measured by AV1451 in non‐AD tauopathies to further our understanding of hallucinations and delusions in different dementia syndromes. Further, 3R or 4R tau specific radiotracers ⁵⁰ as well as measures of soluble phosphotau epitopes and other posttranslational modifications, ⁴³ co‐pathologies (e.g., DLB), and inflammation ²⁴ would add information about potential mechanisms by which elevated neurofibrillary tangles may be associated with hallucinations and delusions.

Overall, a pathological biomarker for psychotic symptoms in AD could aid in early diagnosis, inform therapeutic agents in clinical trials, and complement existing neurobiological biomarkers (e.g., neuroinflammation, neuronal metabolism, white matter connectivity, and gray matter atrophy). Elevated tau burden, above and beyond what is expected with age, amyloid burden, and clinical AD severity, in key brain regions provides insight into the various neurodegenerative and neuroinflammatory profiles reported in individuals with psychotic symptoms. There may not be one single cause of psychosis, but many contributing factors, and neurofibrillary tau tangles in key regions may represent one pathway to psychotic symptoms.

CONFLICTS OF INTEREST STATEMENT

Dr. William C. Kreisl is currently employed by Eisai, Inc.; however, his work on this project does not necessarily reflect the position or views of Eisai. No other authors have anything to disclose. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Informed consent was obtained from all research participants prior to enrollment into the ADNI study.

Supporting information

Supporting information

Supporting information

Johnson AS, Houlihan H, Ziaggi G, et al. Hallucinations and delusions are associated with elevated tau PET signal independent of age, clinical severity, and amyloid burden. Alzheimer's Dement. 2026;18:e70264. 10.1002/dad2.70264