Characteristics of very late-onset schizophrenia-like psychosis classified with the biomarkers for Alzheimer’s disease: a retrospective cross-sectional study

ABSTRACT Objectives: We aimed to investigate the association between very late-onset schizophrenia-like psychosis (VLOSLP), a schizophrenia spectrum disorder with an onset of ≥60 years, and Alzheimer’s disease (AD) using biomarkers. Design: Retrospective cross-sectional study. Setting: Neuropsychology clinic of Osaka University Hospital in Japan. Participants: Thirty-three participants were classified into three groups: eight AD biomarker-negative VLOSLP (VLOSLP−AD), nine AD biomarker-positive VLOSLP (VLOSLP+AD), and sixteen amnestic mild cognitive impairment due to AD without psychosis (aMCI−P+AD) participants. Measurements: Phosphorylated tau levels in the cerebrospinal fluid and 18F-Florbetapir positron emission tomography results were used as AD biomarkers. Several scales (e.g. the Mini-Mental State Examination (MMSE), Wechsler Memory Scale-Revised (WMS-R) Logical Memory (LM) I and II, and Neuropsychiatric Inventory (NPI)-plus) were conducted to assess clinical characteristics. Results: Those in both VLOSLP−AD and +AD groups scored higher than those in aMCI−P+AD in WMS-R LM I. On the other hand, VLOSLP+AD participants scored in between the other two groups in the WMS-R LM II, with only VLOSLP−AD participants scoring significantly higher than aMCI−P+AD participants. There were no significant differences in sex distribution and MMSE scores among the three groups or in the subtype of psychotic symptoms between VLOSLP−AD and +AD participants. Four VLOSLP−AD and five VLOSLP+AD participants harbored partition delusions. Delusion of theft was shown in two VLOSLP−AD patients and five VLOSLP+AD patients. Conclusion: Some VLOSLP patients had AD pathology. Clinical characteristics were different between AD biomarker-positive and AD biomarker-negative VLOSLP, which may be helpful for detecting AD pathology in VLOSLP patients.


Introduction
Very late-onset schizophrenia-like psychosis (VLOSLP) is a disease entity for psychotic disorders with an age of onset of over 60 years, proposed by the International Late-Onset Schizophrenia Research Group in 2000 (Howard et al., 2000). It retains concepts from older terms for late-onset psychotic disorders including "late paraphrenia," advocated by Roth circa 1960 (Kay andRoth, 1961;Roth, 1955). VLOSLP features characteristics such as a female preponderance, solitary living, persecutory (especially partition) delusions, and lack of negative symptoms (Howard et al., 1992;Kay and Roth, 1961). Although current operational diagnostic criteria such as the Diagnostic and Statistical Manuals of psychiatric disease (DSM) (American Psychiatric Association, 2022) and International Criteria of Disease (ICD) (World Health Organization, 2019) do not differentiate psychotic disorders, including the schizophrenia spectrum, by age of onset, many researchers have studied late paraphrenia and/or VLOSLP because of their distinct characteristics, unknown etiology, and frequency in psychogeriatric clinical practice.
Although the criteria for VLOSLP states that the psychosis should not be attributed to affective disorders or to focal or structural brain abnormalities, whether psychotic symptoms in VLOSLP are related to neurodegenerative diseases remains uncertain. Several reports showed that many VLOSLP patients were diagnosed with dementia within several years of follow-ups (Brodaty et al., 2003;Kørner et al., 2009). Using nationwide population registry data, Stafford et al. showed that individuals with VLOSLP represented a high-risk group for subsequent dementia (Stafford et al., 2021). Some pathological studies showed the predominance of argyrophilic grain disease, Lewy body disease in late-onset schizophrenia, delusional disorders (Nagao et al., 2014), and primary age-related tauopathy in late-onset schizophrenia (Casanova et al., 2002). In the context of AD, there were few reports about AD pathology in VLOSLP and lateonset psychosis has been rather studied as symptoms of mild behavioral disorder (MBI), pre-dementia late-onset neuropsychiatric symptoms (Creese and Ismail, 2022). Although differentiating late-onset psychotic disorders with concurrent AD pathology from late-onset psychosis attributable to AD with subtle cognitive impairments is considered difficult, clinical characteristics such as details of psychosis may differ (Fischer et al., 2020).
For the present study, we included patients from our neuropsychology clinic's cohort who were diagnosed with VLOSLP and had been tested for AD biomarkers to confirm the presence of VLOSLP patients who were presumed to have a concurrent AD pathology. We then compared the demographic, neuropsychological, and psychiatric characteristics of those with positive and negative results for AD biomarkers. For comparison, we also enrolled patients with amnestic mild cognitive impairment (aMCI) due to AD without psychotic symptoms.
Although we previously reported a case with VLOSLP and AD confirmed with biomarkers and a longitudinal follow-up (Satake et al., 2021), it remains unclear to what extent VLOSLP patients may have Alzheimer's disease (AD) pathology and what characteristics are different between VLOSLP with and without AD pathology. We hypothesized that some VLOSLP patients will be positive for AD biomarkers and that they will have different characteristics from those who are negative for AD biomarkers and those with aMCI due to AD but without psychosis.

Study design
This was a retrospective observational study with no intervention. The information from all participants was anonymized prior to analysis as unlinked data to prevent the identification of personal information. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. The Osaka University Clinical Research Review Committee (Suita, Japan) approved this study. The requirement of written informed consent was waived by the Review Committee and opt-out was implemented.

Participants
We selected participants from the database of consecutive Japanese patients visiting the neuropsychology clinic in the Department of Psychiatry at Osaka University Hospital for the first time from January 2018 to December 2021. Cases lacking data on Mini-Mental State Examination (MMSE), Clinical Dementia Rating (CDR), and Neuropsychiatric Inventory (NPI)-plus were excluded. We defined VLOSLP according to the criteria adopted by previous studies (Kanemoto et al., 2022). The inclusion criteria were as follows: (1) onset of delusions and/or hallucinations at the age of 60 years or more, and (2) existence of delusions and/or hallucinations at the first visit. The existence of psychosis (delusions and/or hallucinations) was confirmed by the NPIplus sub-items of delusions and hallucinations. The exclusion criteria were as follows: (1) an MMSE score <24; (2) a global CDR score ≥ ; (3) diagnosis of affective disorder; (4) abnormal localized findings on magnetic resonance imaging (MRI) scans, indicating cerebrovascular disease or brain tumor; and (5) a concurrent organic disease that could cause psychosis or physical symptoms suggestive of the disease. We identified the VLOSLP patients who also had data of AD biomarkers (cerebrospinal fluid [CSF] phosphorylated tau [p-tau] or amyloid positron emission tomography [PET]). We then divided VLOSLP patients into those with positive and negative results for AD biomarkers, which we described further in a latter section.
We also selected aMCI patients from the same database. The diagnosis of aMCI was also based on the Petersen's criteria adopted by previous studies (Kazui et al., 2017) which were as follows: (1) a memory complaint documented by the patient or another source; (2) a score of 1.5 standard deviation (SD) below the education-adjusted normal value in the story A recall task in the logical memory (LM) II subtest of the Wechsler Memory Scale-Revised (WMS-R); (3) a score of MMSE ≥ ; 24; (4) a CDR global score = 0.5; and (5) normal basic and instrumental activities of daily living evaluated with Lawton's Physical Self-Maintenance Scale and Instrumental Activities of Daily Living Scale (Lawton and Brody, 1969). For the present analysis, we selected patients with aMCI due to AD without psychosis from all aMCI patients; the exclusion criteria were: (1) existence of psychosis at the first visit; (2) showing negative results for AD biomarkers; (3) onset <60 years; (4) diagnosis of affective disorder; (5) abnormal localized findings on MRI scans, indicating cerebrovascular disease or brain tumor; and (6) coexistence of organic diseases that could cause psychosis or physical symptoms suggestive of them.

Assessment of clinical features
We routinely assessed the physical condition, demographic data, and medical history at our neuropsychology clinic and performed standard neuropsychological examinations. In addition, the patients underwent routine laboratory tests and brain neuroimaging. General cognition was assessed using MMSE. Memory was evaluated with the LM I and LM II subtests. Attention was evaluated with weighted raw score sums for the Attention/Concentration (A/C) index composed of Mental Control, Digit Span, and Visual Memory Span subtests in WMS-R. In addition, psychomotor speed, visuospatial cognition, and language were evaluated using the Digit-Symbol-Coding, Block Design, and Information in Wechsler Adult Intelligence Scale-III (WAIS-III). These were conducted by clinical psychologists and neuropsychiatrists specializing in geriatric psychiatry. Neuropsychiatric symptoms were assessed using the NPI-plus, the original NPI-12 with an additional subitem for cognitive fluctuation (Cummings, 1997;Mori, et al., 2012), by neuropsychiatrists specializing in geriatric psychiatry. This study allowed missing data on scales other than MMSE, CDR, and NPI-plus. Detailed information on psychiatric symptoms of VLOSLP participants was collected and summarized through chart review.

Amyloid PET and CSF analysis
We used 18 F-Florbetapir amyloid PET results and CSF p-tau levels as AD biomarkers. We collected the results of amyloid PET and CSF testing performed within three years from the first visit for the comparison of cross-sectional data with other clinical data. When the two biomarkers were not concordant, we followed the result of amyloid PET. 18 F-Florbetapir was injected intravenously as a slow bolus in an antecubital vein at a mean ± SD dose of 270 ± 51 MBq (range, 182-370 MBq). A 20 min list-mode PET scan was obtained after an uptake time of 40 min (range, 40-43 min), following the imaging acquisition guidelines for Amyvid (https://pi.lilly.com/us/amyvid-uspi.pdf). KMa and ES, who are both nuclear medicine specialists, evaluated each PET result as positive or negative for AD.
We obtained CSF samples from patients who were admitted to our university hospital for diagnosis and treatment. All samples were collected via a lumbar puncture between 10:00 and 12:00 while the patient was fasting; the first 1 mL of CSF from each lumbar tap was discarded. All CSF samples were sent to SRL, Inc. (Tokyo, Japan), where both p-tau and total tau (t-tau) levels were measured in duplicate using enzyme-linked immunosorbent assay (Finoscholar™ p-Tau and hTau [NIPRO Corporation, Osaka, Japan]); we adopted the cutoff ≥ 50 pg/ mL of p-tau, which the SRL, Inc. set for the diagnosis of AD.
(VLOSLP−AD), VLOSLP with positive results (VLOSLP+AD), and aMCI due to AD without psychosis (aMCI−P+AD). We compared the three groups using the Kruskal-Wallis test for continuous variables (e.g. some demographic characteristics and neuropsychological test scores). We also performed the Dunn-Bonferroni post hoc test when we found statistically significant differences. The Fisher's exact test was used for the categorical variables (e.g. other demographic characteristics and for NPI-plus sub-items). For delusions and hallucinations on the NPI-plus, the Fisher's exact test was also used to compare responses of the VLOSLP− AD and VLOSLP+AD groups to each sub-question about psychosis. All statistical analyses used were two-tailed with p < 0.05 considered to indicate statistical significance. These analyses were performed using SPSS Statistics for Windows, Version 27.0. (IBM Corp., Armonk, NY, USA).

Results
Demographic characteristics, AD biomarkers, and APOE genotyping in the selected participants A enrollment flowchart is shown in Figure 1. Of the 805 patients who visited our neuropsychological clinic from January 2018 to December 2021, 679 had NPI-plus and CDR data. Of these, 36 met the criteria for VLOSLP, and 100 met the criteria for aMCI without psychosis. Finally, 8 patients in VLOSLP−AD, 9 patients in VLOSLP+AD, and 16 patients in aMCI−P+AD were enrolled as participants in this study. The numbers of testing for amyloid PET and CSF p-tau were not significantly different between the groups (Table 1). Five participants had CSF p-tau levels <50 pg/mL in the VLOSLP−AD group; all in the aMCI−P+AD and VLOSLP+AD groups had >50 pg/mL. In one participant in the VLOSLP−AD, amyloid PET and CSF p-tau showed conflicting results (PET showed a negative result but p-tau was positive at 61 pg/mL); we adopted the result of amyloid PET in classifying this patient. Table 1 summarizes other demographic characteristics, CSF tau, and Aβ concentrations, and the results of APOE genotyping in the three groups. No significant differences in education years, MMSE scores, CDR scores, CDR sum of boxes, sex distribution, percentage living alone, cholinesterase inhibitor use, and antipsychotic use were seen. Age and onset age showed significant differences (p = 0.017 and 0.024, respectively). The post hoc analysis showed participants in VLOSLP+AD were older than those in aMCI−P+AD at the first The proportion of APOE ϵ4 carriers was also not significantly different. Table 2 summarizes the results of the neuropsychological tests. LM I, II, and weighted raw score sums for A/C of WMS-R showed significant differences (p = 0.001, 0.001, and 0.017, respectively). Dunn's post hoc analysis revealed the following: in LM I, participants in VLOSLP−AD and VLOSLP+AD showed higher scores than those in aMCI−P+AD Results of the Digit-Symbol-Coding, Block Design, and Information subtests in WAIS-III were not significantly different in the three groups. Table 3 summarizes the NPI-plus results. The VLOSLP−AD and VLOSLP+AD groups did not differ significantly in delusions and hallucinations. However, there were significant differences in disinhibition (38% in VLOSLP−AD, 22% in VLOSLP +AD, and 0% in aMCI−P+AD), irritability (50% in VLOSLP−AD, 0% in VLOSLP+AD, and 6% in aMCI−P+AD), and nighttime behaviors (75% in VLOSLP−AD, 22% in VLOSLP+AD, and 25% in aMCI−P+AD). There were no significant differences among the three groups in the other psychiatric symptoms. Table 4 describes the psychotic content elicited from VLOSLP patients on their initial visit. Four in eight and five in nine VLOSLP−AD and VLOSLP+AD patients, respectively, harbored partition delusions. Delusion of theft was shown in two in eight VLOSLP−AD patients and five in nine VLOSLP+AD patients. That someone intruded into their house and stole something was the most frequent delusion.

Discussion
We classified patients as AD biomarker-positive or AD biomarker-negative VLOSLP groups using amyloid PET results and CSF p-tau levels. We further characterized both groups by comparing them with patients with aMCI due to AD without psychosis. The results revealed that despite similar general cognitive abilities and sex distribution among the three groups, they had significant demographic, neuropsychological, and phenomenological differences along several characteristics.
In our cohort, several VLOSLP patients were positive for AD biomarkers. Previous pathological studies on late-onset schizophrenia and delusional disorders did not suggest the presence of an AD   pathology (Casanova et al., 2002;Nagao et al., 2014). This difference may be due to heterogeneity in inclusion criteria [e.g. Nagao et al. excluded patients with delusions of theft (Nagao et al., 2014), while 7 of 17 participants with VLOSLP showed delusions of theft in our study]. Moreover, the participants also differed in age: the median age of participants in VLOSLP+AD group was 84.0 years, while the average age at death was reported as 63.3 years by Nagao et al. (Nagao et al., 2014) and 77.3 years by Casanova et al. (Casanova et al., 2002). Such differences may have led to the absence of AD pathology in their reports. Other longitudinal reports have suggested that some VLOSLP patients may transition to AD dementia after a few years (Brodaty et al., 2003;Stafford et al., 2021). To our knowledge, this study is the first to show that a considerable number of VLOSLP patients may have AD pathology using AD biomarkers. As many as nine of 17 (52.9%) VLOSLP patients in this study had AD biomarkers, perhaps because of advanced age. In the Biomarkers For Identifying Neurodegenerative Disorders Early and Reliably cohort, 24.2% of participants had subjective cognitive decline (SCD) at a mean age of 69.7 years. In the Alzheimer's Disease Neuroimaging Initiative cohort, 25.5% of MCI cases at the mean age of 72.1 years were amyloid PET-positive (Hansson et al., 2018). Additionally, in European memory clinic cohorts, 46.3% and 40.4% of SCD cases with a mean age of 66.5 years were positive for Aβ 42 and p-tau, respectively, in the CSF (Wolfsgruber et al., 2019). Furthermore, the number of people with AD pathology increases with age up to mid-90s (Nelson et al., 2011). The median age of VLOSLP patients in our cohort was over 80 years; thus, it is plausible that more than 50% of them had AD pathology. However, despite the lack of significant differences in age at examination, the participants in the VLOSLP+AD group had a higher age of onset than those in the VLOSLP−AD group. This difference is difficult to attribute solely to age-dependent positivity for AD biomarkers. We have also previously reported a case in which a VLOSLP patient was positive for AD biomarkers and progressed to AD dementia over a two-year follow-up (Satake et al., 2021). To conclude that AD pathology in VLOSLP is incidental and unrelated to psychotic symptoms would be hasty.
In the psychiatric symptoms assessed by NPIplus, the presence of disinhibition, irritability, and nighttime behaviors was the most frequent in VLOSLP−AD among the three groups. Although there is insufficient literature about psychiatric symptoms other than psychosis and negative symptoms in VLOSLP, emotional dysregulation, impulsivity, and sleep disturbances have often been emphasized in the schizophrenia spectrum disorders (Cohrs, 2008;Freeman et al., 2009;Hoptman, 2015;Joseph and Siddiqui, 2022). Irritability, disinhibition, and sleep disturbances are also frequent in MCI and neurodegenerative diseases (Apostolova et al., 2014;Guarnieri et al., 2012). Among those diseases, irritability was reported to be a frequent initial symptom in argyrophilic grain disease (Togo et al., 2005), and sleep disturbances were more common in MCI with Lewy bodies than in MCI due to AD (Donaghy et al., 2020). The higher prevalence of these symptoms in VLOSLP−AD may be explained as psychiatric symptoms of psychotic disorders and potential comorbidities of neurodegenerative diseases other than AD.
Regarding misidentification delusions, one patient with VLOSLP−AD and three patients with VLOSLP+AD were evaluated as having phantom border symptom in the NPI-plus. Fisher et al. reported those delusions to be more dementiarelated delusions than persecutory delusions (Fischer et al., 2020). However, all of the phantom border delusions scored in NPI in the present study were associated with partition delusion that "someone intruded into the house," and these delusions were considered to be mainly persecutory delusions rather than misidentification delusions, and were not necessarily of a quality that would lead to suspicion of dementia. The proportions of the participants harboring persecutory delusions and partition delusions were 87.5% and 50% in VLOSLP−AD, and 88.9% and 55.6% in VLOSLP+AD, respectively. The overall similarity in the content of psychotic symptoms between VLOSLP−AD and +AD made it difficult to predict the AD biomarker results on their own. However, given the small sample size and the lack of uniformity in the assessment of psychosis, it is not possible to determine from these results alone whether AD biomarker-positive and AD biomarker-negative VLOSLP psychosis are truly similar. Future evaluation of semi-structured psychosis with a larger sample may reveal features that may be useful in differentiating between the two.
The present study proposes that AD pathology may be directly involved in the emergence of psychosis in VLOSLP with AD pathology. However, we think it is challenging to attribute the psychosis solely to AD pathology. Recent pathological studies have shown that in people older than 80 years, individuals with AD pathology often have other neurodegenerative changes and vascular diseases (Beach and Malek-Ahmadi, 2021;Yu et al., 2020); even in older VLOSLP patients with AD pathology, the influence of these age-related copathologies may not be ignored. We previously reported that there is a certain number of VLOSLP patients with positive results for indicative biomarkers of DLB (Kanemoto et al., 2022). Therefore, VLOSLP+AD patients in this study also might have Lewy body disease pathology. However, the involvement of a concurrent AD pathology in the pathogenesis of VLOSLP should be considered. With regard to pharmacotherapy, the usefulness of antipsychotic drugs for VLOSLP was already demonstrated (Howard et al., 2018;Scott et al., 2011). However, such uses of antipsychotics could be tailored to the specific neuropathology the patient has. In the future, new drugs, especially diseasemodifying ones, may be candidates for the treatment of VLOSLP with specific neuropathologies. Further clarification of the link between neuropathology and VLOSLP would be helpful for the treatment of VLOSLP.

Limitation
Apart from the small number of participants, there are several limitations. First, selection bias in this study cannot be overlooked. Although we attempted to generalize the results of this study to general VLOSLP, two substantial biases existed: the implementation of testing for AD biomarkers and the fact that participants visited our neuropsychology clinic. AD biomarker testing was performed for research and clinical necessity. AD biomarkers are of greater value for assessing neuropathology in younger patients (Ossenkoppele et al., 2015). Therefore, we may have tended to perform AD biomarker testing in relatively young MCI patients without psychosis, with a view to drug trials and future clinical use of disease-modifying drugs. In contrast, our use of AD biomarker testing for VLOSLP patients was not subject to age of onset bias. The differences in the selection criteria for testing might have led having older participants in the VLOSLP groups than in the aMCI−P+AD group. With regard to patient visit bias, patients who visited our neuropsychology clinic likely presented with more cognitive decline than the general VLOSLP population, as VLOSLP patients often do not voluntarily visit memory or psychiatric clinics owing to preserved function and lack of insight (Lam et al., 2016). However, many VLOSLP patients in this study were referred from community psychiatric clinics and outreach teams organized by psychiatrists for the differential diagnosis of mild dementia AD biomarker-positive and -negative VLOSLP 11 with psychosis and VLOSLP. Second, the AD biomarkers we used may not reflect pathology perfectly. However, 18 F-Florbetapir PET is a reliable tracer for AD and is predictive even of early-stage disease (Clark et al., 2011;Palmqvist et al., 2015); CSF p-tau is also reliable in detecting early-stage AD (Mattsson et al., 2017). Furthermore, the CSF ttau, Aβ1-42 and Aβ1-42/1-40 scores of participants in VLOSLP and aMCI groups were compatible with the presumed pathologies (Blennow and Zetterberg, 2018;Hansson et al., 2019;Mattsson et al., 2009). Third, we had several statistical limitations. Although many statistical comparisons were conducted, we did not conduct statistical correction for multiple comparisons. We focused on avoiding the risk of a β-error rather than α-error in this small sample size exploratory study. Note that when the Benjamini-Hochberg procedure was performed to evaluate the false discovery rate in each table, only WMS-R LM1, 2 and the weighted raw score sums of A/C were considered significant, and these differences were considered relatively robust. In addition, due to the small sample size, it was not possible to perform a multivariate analysis. Fourth, as mentioned previously in the discussion, we did not control confounders for psychosis (e.g. physical disorders, financial backgrounds, recent impactful events, sensory impairments, and medications). They were difficult to regulate due to lack of data. Finally, some VLOSLP patients had cognitive impairment that could be considered MCI. Our classification for VLOSLP could have included a group with MCI and psychosis. However, VLOSLP patients performed better on the episodic memory test than aMCI−P+AD patients, while they performed worse on the attention test. This suggests that while a small percentage of the VLOSLP patients in this study met the criteria for MCI, they remained distinct from the MCI patients. Future longitudinal studies with a larger sample size and more detailed assessment of psychotic symptoms that address many of the limitations of this study are needed.

Conclusions
Our findings suggest that VLOSLP is a prodromal feature of dementia or that VLOSLP independently confers risk for later dementia via other mechanisms (Stafford et al., 2021). A number of VLOSLP patients had AD pathology in our retrospective cohort. The VLOSLP patients who tested positive for AD biomarkers were older at onset than those who tested negative. Episodic memory impairment in AD biomarker-positive VLOSLP was intermediate between aMCI due to AD without psychosis and AD biomarker-negative VLOSLP. The prevalence of psychiatric symptoms such as disinhibition, irritability, and nighttime behaviors differed between VLOSLP with positive and negative AD biomarkers and aMCI due to AD without psychosis. These clinical differences may be helpful for estimating the pathology of VLOSLP.

Conflict of interest
None.

Description of authors' roles
YS and HK designed the study. YS wrote the initial draft of the manuscript. HK, KMa, ES, KMo, TM, KY, and MI contributed to the interpretation of data and assisted in the preparation of the manuscript. DT, TS, FK, SS, SG, and TW contributed to data collection and interpretation and reviewed the article. All authors approved the article and agreed to be accountable for all aspects of the work.