Analysis of positron emission tomography hypometabolic patterns and neuropsychiatric symptoms in patients with dementia syndromes

Abstract Aims To estimate the proportions of specific hypometabolic patterns and their association with neuropsychiatric symptoms (NPS) in patients with cognitive impairment (CI). Methods This multicenter study with 1037 consecutive patients was conducted from December 2012 to December 2019. 18F‐FDG PET and clinical/demographic information, NPS assessments were recorded and analyzed to explore the associations between hypometabolic patterns and clinical features by correlation analysis and multivariable logistic regression models. Results Patients with clinical Alzheimer's disease (AD, 81.6%, 605/741) and dementia with Lewy bodies (67.9%, 19/28) mostly had AD‐pattern hypometabolism, and 76/137 (55.5%) of patients with frontotemporal lobar degeneration showed frontal and anterior temporal pattern (FT‐P) hypometabolism. Besides corticobasal degeneration, patients with behavioral variant frontotemporal dementia (36/58), semantic dementia (7/10), progressive non‐fluent aphasia (6/9), frontotemporal lobar degeneration and amyotrophic lateral sclerosis (3/5), and progressive supranuclear palsy (21/37) also mostly showed FT‐P hypometabolism. The proportion of FT‐P hypometabolism was associated with the presence of hallucinations (R = 0.171, p = 0.04), anxiety (R = 0.182, p = 0.03), and appetite and eating abnormalities (R = 0.200, p = 0.01) in AD. Conclusion Specific hypometabolic patterns in FDG‐PET are associated with NPS and beneficial for the early identification and management of NPS in patients with CI.


| INTRODUC TI ON
Cognitive impairment (CI), including mild cognitive impairment (MCI) and dementia, has been a major global health problem because of the aging of the world's population. Dementia is broadly characterized by cognitive and psychological dysfunction that significantly impairs daily and social functioning. Overall, approximately more than half of dementia cases are attributable to Alzheimer's disease (AD), followed by vascular dementia (VaD), dementia with Lewy bodies (DLB), and frontotemporal lobar degeneration (FTLD). Currently, early identification, differential diagnosis, and improved therapy have become an important part of CI management.
[18F]-Fluorodeoxyglucose positron emission tomography (18F-FDG-PET) is a biomarker for neuronal degeneration with good diagnostic accuracy, thus it is widely used in current clinical research and has now been recommended as a reliable tool for diagnosis in patients with CI. 1 AD shows hypometabolism predominantly in a posterior pattern, including in the posterior temporoparietal association cortex and posterior cingulate cortex. Patients with DLB share the characteristic of a posterior pattern of hypometabolism that is seen in patients with AD and also demonstrate hypometabolism in the occipital lobe. 2 This pattern is characterized by the "cingulate island sign," which demonstrates a parieto-occipital pattern of hypometabolism and relatively preserved posterior cingulate metabolism. 3,4 The metabolic abnormality in FTLD is predominant in the frontal and anterior temporal lobes, cingulate gyri, uncus, insula, and the subcortical areas, including basal ganglia and medial thalamic regions. 5-7 AD-pattern hypometabolism represents decreased clinical stability, with accelerated progression from amnestic MCI to AD. 8 Moreover, longitudinal FDG-PET data revealed that (i) greater medial temporal and posterior cingulate hypometabolism were related to memory decline; (ii) asymmetrical lateral temporal hypometabolism was related to language dysfunction; (iii) hypometabolism in the lateral parietal lobe and precuneus hypometabolism were related to visuospatial dysfunction and hallucinations; and (iv) progressive frontal lobe hypometabolism was related to executive dysregulation and depression. [9][10][11] Until now, there has been a lack of studies on hypometabolic patterns in large samples of FDG-PET in patients with CI, especially in patients with FTLD subtypes. In addition, there have been few studies that evaluate the correlation between hypometabolic patterns in neuroimaging and neuropsychiatric symptoms (NPS) in patients with dementia.
Therefore, we performed a multicenter study to estimate the proportion of the primary hypometabolic pattern in a large sample of patients with a variety of MCI and dementia syndromes, evaluate the associations between AD-pattern (AD-P) and frontotemporal lobe-pattern (FT-P) hypometabolism and NPS, and examine the apolipoprotein E (APOE) ε4 allele, amyloidβ (Aβ), or Tau deposition on PET scans in this cohort of patients.

| Clinical assessment
The clinical assessment was performed by neurologists specialized in dementia care and included a detailed history taken from the primary caregivers of the patient, a physical examination, cognitive assessments, laboratory tests (including thyroid/liver/kidney function tests, vitamin B12 level, folate level, syphilis serology, and APOE genotype), and neuroimaging (brain MRI/CT and PET-CT). were assessed with the 12-item NPI using the information provided by their caregivers. The composite score of each subscale ranges between 0 (no NPS) and 12 and the total composite score between 0 (no NPS) and 144.

| Imaging acquisition
Acquisition procedures for FDG-PET and Aβ-PET have been fully described in a previous study. 12 Briefly, a T1-weighted sequence was acquired on a 3.0-T GE Healthcare scanner or a 3.0-T Siemens  and then smoothed. The regional SUVR z score was defined as: (single patient's SUVR-mean SUVR observed in healthy controls)/SD of SUVR value observed in healthy controls. A regional z score ≥ 2 was considered to define positive findings for semiquantitative interpretation at the regional level. 13 In this study, considering the

| APOE genotyping
Genomic DNA was extracted from peripheral blood stored at −80 °C, and the APOE gene was amplified by polymerase chain reaction (PCR), with details in Appendix S1. We determined all genotypes without knowledge of the patient's status.

| Statistical analysis
The Skewness-Kurtosis test was used to check the normal distribution of the data. Since the age, education years, course of disease, the scores of MMSE and MoCA did not satisfy the normal distribution, the data were described as the medians (interquartile range, IQR). The qualitative variables were expressed as frequency, and the chi-squared test was used to compare the two independent groups ( Figure 1B-D) for qualitative variables. Linear regressions were run to analyze the associations between accumulated frequencies of hypometabolic patterns and age at PET-CT performed ( Figure 1E) and the course of disease ( Figure 1F). The correlation between clinical features and metabolism pattern was evaluated by chi-squared tests and described by Pearson contingency coefficient for two qualitative variables (such as the correlations between AD-P/FT-P and Aβ deposition/Tau aggregation/APOE ε4 carrier/12 items of NPI in Table 3), or Spearman's correlation between qualitative variables and qualitative variables (such as the correlations between AD-P/ FT-P and scores of MMSE, MoCA, and NPI in Table 3). Multivariable logistic regression was used to produce individual predicted probability using the cross-validated method of the leave-one-out principle, which drops the data of one subject and re-estimates the parameters. The cross-validated predicted probabilities were used to assess discriminatory performance of hypometabolism of AD-P or FT-P, amyloid deposition, and Tau Table S1.   Table S2 show that patients with Aβ deposition, Tau aggregation, or APOE ε4 allele had higher probabilities of AD-P hypometabolism. In addition, the proportion of AD-P hypometabolism in all CI patients was significantly positively related to age (R 2 = 0.262, p = 0.000) but not the course of the disease.  Table 3).

| RE SULTS
The ROC curves in Figure 2

| DISCUSS ION
In this multicenter neuroimaging cohort study among patients with CI, the majority of AD, DLB, and CBD patients demonstrated AD-P hypometabolism, while FTLD and its main subtypes, including bvFTD, SD, PNFA, FTLD/ALS, and PSP, showed predominantly FT-P hypometabolism, whereas no specific hypometabolic patterns were found in MCI. The combination of metabolic pattern and biomarkers can effectively distinguish AD/FTLD from other types of dementia.
In addition, the higher proportion of AD-P hypometabolism was associated with old age, Aβ deposition, Tau aggregation, and worse cognition, although not with APOE ε4 allele or the course of disease. Furthermore, a higher proportion of FT-P hypometabolism was related to the presence of hallucinations, anxiety, and appetite and eating abnormalities in AD.

| Interpretation of results
In the present study, over half of patients with clinical AD and DLB showed AD-P hypometabolism, which was consistent with previous neuroimaging research. 22,23 In addition, hypometabolism in the lateral occipital cortex and the "cingulate island sign" specifi-   42 Other neuroimaging studies found appetite and eating abnormalities were associated with hypometabolism in the orbitofrontal cortex, 43,44 or the atrophy of the medial temporal cortex. 45 However, we cannot demonstrate the significant associations between FT-P hypometabolism with other NPS like previous literature [9][10][11]36,46 since the differences in target population or sample-size.

AD-P FT-P AD-P FT-P AD-P FT-P
These findings have potential clinical and programmatic relevance for standardizing early diagnosis of dementia, and improving the accuracy of differential diagnosis. The significant differences in hypometabolic patterns across CI types makes FDG-PET a helpful parameter in the differential diagnosis of CI, while the independent application of AD-P and FT-P hypometabolism plays a minor role in the diagnosis of MCI and DLB.

| Limitations
We present a multicenter cohort study with comprehensive neurological assessment and molecular imaging biomarkers in a large group of CI patients. First, a little part of the target participants had "clinical diagnosis" due to the absence of blood and CSF biomarkers, it might weaken the strength of our finding. Moreover, the main limitation in this study is the qualitative interpretation of the FDG-PET images with AD-P or FT-P and the lack of amyloid and Tau deposition pattern. The lack of detailed descriptions of brain regions limited us from doing more analysis, likely a detailed evaluation of the relationship among NPS, hypometabolic areas, Aβ, and Tau deposition patterns.

| CON CLUS ION
These results suggest that brain hypometabolic patterns are (i) significantly different across CI types, (ii) associated with old age, Aβ deposition, Tau aggregation, and worse cognition, and (iii) related to the presence of hallucination, anxiety, and appetite and eating abnormalities in AD. In addition, the combined application of AD-P or FT-P hypometabolism and biomarkers can effectively distinguish AD/FTLD from other types of dementia. Furthermore, specific hypometabolic patterns are associated with NPS and beneficial for the early identification and management of NPS in patients with CI.

AUTH O R CO NTR I B UTI O N S
Dr. YJ had full access to all of the data in the study and take re-

ACK N OWLED G M ENTS
The present study was supported by the National Natural Science

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.