Structural Neuroimaging Correlates of Neuropsychiatric Symptoms in Alzheimer’s Disease: A Systematic Literature Review

Alzheimer’s disease (AD), a progressive neurodegenerative disorder characterized by cognitive decline and memory loss, affects millions of people worldwide. Neuropsychiatric symptoms (NPS), for example, apathy and agitation, significantly impact individuals with AD and their caregivers both medically and in terms of quality of life. NPS are reported throughout all stages of AD, and while the frequency of NPS varies, it has been reported that most individuals with AD exhibit one or more NPS at some point [1‒4]. Due to this high prevalence, a more complete understanding of NPS in AD is imperative for better care management and quality of life for people with AD. These symptoms are linked to many negative outcomes including increased disease severity in individuals with AD, risk of institutionalization, caregiver burden, and costs of care [5‒9]. Additionally, NPS are associated with accelerated conversion to AD for individuals with mild cognitive impairment [10‒12], subjective cognitive decline [11], and even cognitively normal individuals [10]. Despite this, there are inconsistencies in the literature regarding the specific brain regions associated with various NPS, and this complicates their interpretation and generalization across studies.

AD progression has been reliably associated with distributed changes in gray matter (GM), some of which are closely related to focal pathologies and specific cognitive deficits [13‒16]. For example, atrophy in the medial temporal lobe, specifically the hippocampus, is associated with the hallmark memory loss of typical AD [13]. Similarly, associations between brain differences and behavioral/psychiatric variables can be measured. Thus, understanding behavioral/psychiatric differences in AD requires evaluating the associations between NPS and regional changes in brain structures. Concretely, AD-associated NPS (AD-NPS) may be related to focal neuroanatomical change in GM. However, it remains unclear which specific brain regions are most strongly associated with individual NPS, creating a gap in the literature that this review aims to address.

Characterizing NPS requires rigorous assessment of a group of highly subjective symptoms, and a number of tools have been developed to address this need. The most commonly reported NPS assessment tool is the Neuropsychiatric Inventory (NPI); it also has a brief format, the Neuropsychiatric Inventory Questionnaire (NPI-Q) [17]. The NPI and NPI-Q are informant-reported measures encompassing 10 observed behavioral changes such that higher scores are associated with increased symptom severity. These behaviors or symptoms include apathy, anxiety, depression, hallucination, delusion, agitation/aggression, disinhibition, irritability/lability, euphoria/elation, and aberrant motor behavior. Other questionnaires assessing NPS either characterize NPS broadly, similar to the NPI, or instead focus on specific symptoms (e.g., apathy when using the Apathy Inventory). These symptoms may be analyzed individually, but some investigators instead use symptom clusters based on a factor analysis. While symptom clusters may be influenced by a myriad of variables (disease stage, informant vs. self-report, etc.), the most commonly used clusters were proposed in 2007 by Aalten and colleagues [18] and include apathy, affective symptoms (anxiety, depression), psychosis (hallucination, delusion), and hyperactivity (agitation/aggression, disinhibition, irritability/lability, euphoria/elation, aberrant motor behavior). Together, these questionnaires and clusters are used to assess dementia-related frequency and severity of behavioral disturbances. However, the prevalence and severity of symptoms can vary significantly depending on the disease stage and some NPS may be mistaken for others. For example, delusions and hallucinations are most often reported in late-stage AD, while apathy is reported across all disease stages [19]. Similarly, apathy is frequently mistaken for depression and this discrepancy may be influenced by symptom perception from either self or informant report [20]. This review aims to synthesize the current literature on AD-NPSs and brain regions associated with specific NPS in AD, with a particular focus on understanding regional GM reduction using magnetic resonance imaging (MRI).

The frequency, severity, and consequences of AD-NPS motivate the necessity of understanding structural brain changes related to NPS. Identifying shared and/or symptom-specific brain regions affected by AD-NPS can offer insights into targeted treatment strategies and improve clinical management of associated symptoms. We aim to resolve discrepancies in the literature and provide a clearer picture of the structural correlates of AD-NPS, potentially guiding future research by identifying critical regions that warrant further investigation in relation to specific NPS. While similar reviews of AD-NPS have discussed gray and white matter changes, as well as amyloid beta deposition as measured with several different imaging tools [21‒23], this review focuses specifically on regional GM reduction in the brains of individuals with clinically diagnosed AD measured with MRI. This review will first summarize the existing literature on AD-NPS, followed by an analysis of regional GM changes and their association with specific symptoms, and conclude with a discussion of the clinical implications of these findings.

The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were used to conduct this review [24] (shown in Fig. 1). This review was conducted using the PubMed database to identify studies that utilized MRI and reported GM loss associated with individual or clustered NPS. The search was conducted on March 30th, 2022. Search terms were specified to be in the title and/or the abstract. Three terms were used and combined in two separate searches: first, (Alzheimer’s disease [Title/Abstract]) AND (neuropsychiatric symptoms [Title/Abstract]); and second, (dementia [Title/Abstract]) AND (neuropsychiatric symptoms [Title/Abstract]). No limitations were placed on publication date.

These search terms yielded 3,145 manuscripts. Titles and abstracts were reviewed and, if eligible, the full text was reviewed. Papers were reviewed by one author-reviewer (M.K.R.) unless there was uncertainty about an article. In that case, a second author-reviewer was consulted (D.E.W.). Papers were included if (1) the subject met clinical criteria for probable AD, (2) the primary imaging technique was MRI measuring GM changes, and (3) the NPSs measured were part of the 10 NPS measured in the NPI (see online supplement S1 for inclusion/exclusion criteria; for all online suppl. material, see https://doi.org/10.1159/000543160).

Several questionnaires evaluating NPSs were deemed acceptable with the condition that they measured the NPS of interest. General NPS assessments included the previously discussed NPI (Cumming et al. [17]), as well as the NPI-Q [25] and the Behavioral Pathology in AD Scale (BEHAVE-AD) [26]. Symptom-specific assessments included the Geriatric Depression Scale (GDS) [27], Apathy Evaluation Scale (AES) [28], Modified Apathy Evaluation Scale (MAES) [29], and Apathy Inventory (AI) [30].

Our description of these findings reflects an attempt to harmonize the nature and directionality of the original observations (shown in Fig. 2). For example, “increased cortical thinning” and “reduced cortical thickness” may describe the same phenomenon but in opposite ways, and so we reframe all findings from the perspective of gray matter loss (GML) of several distinct types. These include reductions of density (GML-D), reductions of volume (GML-V), cortical thinning (GML-T), and cortical atrophy (GML-A). GML-D may be understood as the proportion of GM to other tissue types in a specified area while GML-V is the total amount of GM in a specific area [31]. The loss of GM volume, density or cortical thinning is GML-A. GML-T is reduced width of the cortical ribbon [32].

Broadly, multiple NPS were associated with GML in the frontal lobe, cingulate cortex, insular cortex, and temporal cortex, whereas the parietal lobe, occipital lobe, basal ganglia, and thalamus were mentioned less frequently (shown in Fig. 3). Of 29 articles, 15 statistically adjusted for multiple comparisons. Three studies met the inclusion criteria but did not report relevant statistically significant results between NPS and GML-V [33‒35] (see online supplement S2 and Spreadsheet 1 of the online data supplement).

Regarding the articles reviewed under the refined criteria and their characteristics, this review identified 29 articles (Table 1). Eight articles used data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) project [36], while two additional articles used data from the Konkuk dementia registry [37]. Approximately 3,176 individuals with AD were assessed in these studies, not accounting for potential overlap between studies. Two studies reported longitudinal GM changes associated with NPS. Nakaaki and colleagues [38] retrospectively overserved baseline GM volumes associated with developing delusion 2 years later. Additionally, Rafii and colleagues [39] investigated associations between psychotic symptoms at baseline and GM atrophy after 1 year. Of the 29 articles, 27 reported average Mini-Mental State Examination (MMSE) scores. The non-weighted median of the mean MMSE for participants contributing to these articles was 21.16 with a range from 13.60 to 28.07. As a score of 23 or lower indicates dementia [40], the studied samples reflected a range of cognitive ability. All included studies used MRI as their imaging modality, and one study complemented MRI with CT. Structural MRI scans had voxel sizes ranging from 1.0 mm³ to 1.5 mm³ which is consistent with the target voxel size for the ADNI study [41]. Nine articles did not explicitly state voxel size (for more information regarding the included manuscripts, refer to online suppl. Spreadsheet 1).

A frequently cited measure from the NPI (or NPI-Q) is the total score, or the sum of all the individual domain scores which ranges from 0 to 144 with higher scores representing increased symptom severity [17]. While the total score may not offer insight into a patient’s specific symptoms, it may provide a useful measure of a patient’s (or caregiver’s perceived) overall NPS “load.” This review identified five articles that reported structural brain abnormalities associated with total NPI or NPI-Q score. One out of five employed whole-brain analysis, while four examined a priori regions of interest.

Regarding associations of GML with NPI total score, there was some evidence of GML lateralization, specifically the left temporal gyrus [44, 62]. Regions with bilateral associations included the temporal [49] and frontal lobes [9, 62], left insula [44], amygdala [49], and left ventral nucleus [50]. Online supplement S3 of the online data supplement summarizes the relevant findings.

Apathy is one of the most frequently reported AD-NPS and its prevalence increases with disease severity. Even in mild AD, half of all cases exhibit this symptom [19, 65, 66]. Apathy is of course characterized by a reduction or loss of motivation [28] affecting emotion, cognition, and/or social interactions [67]. In this review, apathy was most often found to be measured using the NPI/NPI-Q and occasionally using the AES, MAES, and AI.

Broadly, apathy was found to be associated with GML throughout the brain in the basal ganglia [43, 52], the frontal [42, 43, 48, 51, 54, 58, 60], temporal [48, 51, 52, 58, 64], occipital [48, 64], parietal [48], insular [51, 52, 58], and cingulate cortices [42, 43, 48, 51, 58, 60]. Generally, these findings were left lateralized, and they were identified using both whole-brain and region of interest analyses. The most consistent regions with evidence of GML were the orbitofrontal cortex, inferior frontal cortex, left insula, posterior cingulate, and putamen. Online supplement S4 of the online data supplement summarizes the relevant findings.

Affective symptoms, including anxiety and depression, are frequently observed in AD and have been proposed to comprise an affective syndrome in prior work [18, 68]. In the general population, anxiety and depression are often comorbid, and having both is associated with higher symptom severity [69, 70]. Hayata and colleagues [46], who operationalized comorbid anxiety and depression in individuals with AD as an affective syndrome, reported GML-A using whole-brain analysis. Reported regions included the frontal cortex (including the lateral orbitofrontal region, pars triangularis, and precentral gyrus), the insular cortex, and regions of the temporal lobe (including the temporal pole, inferior and superior temporal cortex, and the fusiform gyrus) [46]. Online supplement S5 of the online data supplement summarizes the relevant findings.

Anxiety is a symptom characterized by feelings of fear or worry that may be accompanied by physical symptoms (e.g., excessive sweating, increased heart rate) that interfere with a person’s activities of daily living [17]. In AD, anxiety is estimated at a prevalence of 39% [19, 71]. Regarding the reviewed articles, anxiety was assessed using the NPI-Q or the BEHAVE-AD assessments. Two articles reported significant associations with anxiety.

Relative to healthy comparison participants, Mohamed Nour and colleagues [52] found significant GML-V in the left parahippocampal cortex, posterior cingulate gyrus, left insula, and bilateral putamen. Meanwhile, Tagai and colleagues [59] found that anxiety was correlated with GML-V of the right precuneus and inferior parietal lobule. Both studies utilized whole-brain analysis and reported significant cingulate GML-V. Online supplement S6 of the online data supplement summarizes the relevant findings.

Depression is a common AD-NPS with approximately 40% of individuals with AD developing depression [19, 72]. In a substantial number of patients with AD, depressive symptoms become clinically significant. Based on a recent meta-analysis, major depressive disorder was observed in 14.8% of individuals with AD [73]. In this review, articles used the NPI-Q to assess depression.

Two articles reported significant depression-associated structural brain differences in AD of the frontal, temporal, and insular cortices. There was some evidence of lateralization of GM changes to the right for the middle and medial temporal region. Mohamed Nour and colleagues [52] found evidence between higher depression scores and GML-V of the middle frontal gyrus, right middle temporal gyrus, temporal pole, right insula, right parahippocampal cortex, and hippocampus. Additionally, Hu et al. [47] found that higher depression scores were associated with GML-A of the superior and left middle frontal gyri. Both studies utilized whole-brain analysis. Online supplement S7 of the online data supplement summarizes the relevant findings.

The psychosis cluster is composed of the delusion and hallucination NPS [18]. Prevalence of psychosis in AD has been estimated at 41% [74]. In this review, two publications reported GML in brain structures associated with psychosis, measuring regions of interest. Cotta Ramusino and colleagues [54] saw cortical GML-A in the frontal lobe (considered as a whole) to be associated with scores on the psychosis cluster. Additionally, Siafarikas and colleagues [57] found that higher psychosis scores were associated with GML-V of the left postcentral gyrus. Online supplement S8 of the online data supplement summarizes the relevant findings.

Hallucinations are a sensory perception of something not there [17] such as seeing or hearing a person who is not present. Based on recent reviews, approximately 15% of individuals with AD experience hallucinations [19, 71]. Despite occasional manifestation in early disease stages, it is generally understood that hallucinations appear in later stages of AD [74, 75].

Broadly, hallucinations were found to be associated with bilateral GML in the frontal and cingulate cortex. In a study assessing MCI and mild AD, Rafii and colleagues [39] found that greater hallucination scores at baseline were correlated with GML-A 12 months later in regions of the anterior and posterior cingulate cortices, lateral frontal cortex, and medial orbitofrontal cortex. Cotta Ramusino and colleagues [54] reported the frontal lobe as having GML-A associated with higher hallucination scores for individuals with AD compared to non-AD dementia/MCI. Both studies assessed regions of interest. Online supplement S9 of the online data supplement summarizes the relevant findings.

Delusions are firmly held beliefs for something that is not true [17]. Recent reviews reported that 23–31% of individuals with AD have delusions [19, 71]. Six articles reported significant GML associated with delusions, as measured using the NPI.

Broadly, GML was reported in regions of the frontal [38, 39, 43, 54, 63], temporal [38, 39, 54, 63], and parietal lobe [43], as well as the insular [38] and cingulate cortices [38, 39]. Two studies utilized whole-brain analysis and the remaining four measured regions of interest. There were four overlapping regions mentioned by at least two articles associated with increased delusion scores. These included the orbitofrontal cortex, left insula specifically the claustrum, and the anterior and posterior cingulate. There was evidence of left lateralization of insular GML. Online supplement S10 of the online data supplement summarizes the relevant findings.

Hyperactivity is a cluster of symptoms including agitation/aggression, disinhibition, irritability/lability, aberrant motor behavior, and euphoria/elation [18]. The latter symptom was not included in this review due to no significant reported brain structural GML correlates of euphoria/elation. This is likely due to the low frequency at which euphoria/elation is reported in AD [19]. Siafarikas and colleagues [57] reported a cluster including euphoria and disinhibition that they called the elation cluster. These authors found elevated elation scores were associated with GML-T of the right anterior cingulate cortex in a sample of people with MCI and AD. Online supplement S11 of the online data supplement summarizes the relevant findings.

In the NPI, agitation and aggression are terms used to describe the subsection of the questionnaire focusing on uncooperative and resistant behavior [17] and will be referred to as simply aggression. Aggression can include verbal symptoms like screaming and behaviors like throwing objects and physically assaulting others. A recent meta-analysis concluded that the prevalence of aggression in patients with AD was 27.8% [76].

Six articles reported significant GML associated with increased aggression scores as measured with the NPI and NPI-Q. Implicated regions included the frontal lobe [47, 54, 61], insula [43, 61], cingulate [43, 48], fusiform gyrus [48], hippocampus and amygdala [61]. Two articles reported left lateralization of the insula. Online supplement S12 of the online data supplement summarizes the relevant findings.

Disinhibition is characterized as impulsiveness and inappropriate social behavior such as making crude remarks or sharing very personal information publicly [17]. Based on recent meta-analyses, this symptom has a prevalence of 5–17% in individuals with AD [19, 71]. Two articles reported significant GML associated with high disinhibitions scores. Both articles measured GML in regions of interest that did not overlap.

GML in the frontal, temporal, and cingulate cortices was reported in association with disinhibition scores. Finger and colleagues [45] observed GML-T of the right frontal pole. Serra and colleagues [56] found the right middle frontal gyrus and right precentral gyrus to have GML-V. The temporal and cingulate cortex were both found to be associated with disinhibition where higher disinhibition was found to be related to GML-T of the left middle temporal gyrus [45] and GML-V of the bilateral cingulate cortex [56]. Online supplement S13 of the online data supplement summarizes the relevant findings.

Irritability is often characterized by being abnormally sensitive where one is quick to anger or be annoyed [17]. Based on recent meta-analyses, irritability has been estimated in 36–41% of individuals with AD [19].

Two articles reported GML associated with increased irritability scores as measured using the NPI, but the regions of interest evaluated in the two articles did not overlap. However, findings from both articles were right lateralized, and in total, three regions were identified in association with increased irritability. GML-V of the right posterior cingulate gyrus and right superior parietal [48] as well as GML-A of the right posterior insular cortex was found to be associated with higher irritability scores [51]. Online supplement S14 of the online data supplement summarizes the relevant findings.

Aberrant motor behavior is characterized by repetitive, seemingly purposeless activities [77]. This symptom is more commonly reported in men than women [78]. Aberrant motor behavior has been reported in 18–32% of individuals with AD [19, 71]. Three articles reported significant GML associated with increased aberrant motor behavior scores. Results from these articles reported significant GML associated with increased aberrant motor behavior scores using the NPI and NPI-Q.

One article utilized whole-brain analysis, and two measured regions of interest. There were no overlapping regions across the three studies. However, several frontal lobe regions were reported by Hu and colleagues [47] to have GML-V including the right inferior and right middle frontal, right olfactory, and right medial orbitofrontal gyri as well as the bilateral gyrus rectus. Rosen and colleagues [55] found the dorsal anterior cingulate cortex extending into the left premotor cortex to have GML-V [55] associated with higher aberrant motor behavior scores. Additional regions found to have evidence of GML associated with aberrant motor behavior were GML-A of the amygdala [53], GML-A of the pallidum [47], and GML-V of the dorsal anterior cingulate [55]. Online supplement S15 of the online data supplement summarizes the relevant findings.

AD and aging are most often associated with neurodegeneration manifested as atrophy, thinning, or volumetric reductions, and in this context, our review focused on GML. However, in the course of our review, we noted certain articles reporting GM increase associated with specific NPS. While this outcome challenges normal expectations regarding age- and disease-related neurodegenerative trajectories, we describe these findings here for completeness.

To explore this phenomenon further, we conducted a post hoc review for articles identified in the original PubMed search that mentioned GM increases associated with AD-NPSs. Five articles were identified [45, 48, 58, 61, 79], three of which also reported AD-NPS-GML. None of the five articles measured the same NPS. However, Finger et al. [45], and Trzepacz et al. [61], both reported left cortical parietal GM increases associated with disinhibition and aggression, respectively. While the specific region of the parietal cortex did not overlap (inferior vs. superior regions), the matched lateralization may warrant further investigation. These findings suggest that certain behavioral symptoms may be associated with localized GM increases. Aggression was associated with the largest number of regions showing GM increases including right rostral anterior cingulate, bilateral pallidum, right fusiform, right hippocampus, and left superior parietal [61]. Other associations of GM increases with specific NPS included the amygdala with depression [79], left inferior parietal with disinhibition [45], left pars orbitalis with irritability [48], and right lingual gyrus and left cuneus with apathy [58].

Increases in GM are unexpected in a neurodegenerative disease such as AD, where progressive atrophy is more commonly reported. However, these changes may reflect underlying behavioral/psychiatric symptoms rather than AD-NPS as such. For instance, in bipolar disorder [80], schizophrenia [81], and autism spectrum disorder [82], there are significant increases in putamen volume. Additionally, Okada and colleagues [83] found that individuals with bipolar disorder had larger bilateral caudate and left pallidum volumes. It is known that AD is often associated with comorbidities which may influence disease presentation and projection. These comorbidities may impact brain volume independently or synergistically with AD pathology, potentially leading to paradoxical focal GM increases. Regarding a mechanistic account of GM increases associated with NPS, we can only speculate. It is possible that GM increases may reflect compensatory or neuroinflammatory responses related to comorbidities, though further research would be necessary to evaluate this account. Notable limitations of these findings include small sample sizes and methodological differences across studies. Future research should aim to replicate these findings in larger cohorts, with attention to specific NPS and neuroanatomical regions associated with GM increases.

The overarching goal of this review was to identify brain regions impacted by NPS to elucidate symptom-specific patterns of GML in AD. We examined 29 articles and found that the prefrontal, temporal, and cingulate cortices were frequently associated with AD-NPS, suggesting that GML in these regions may contribute to these behavioral symptoms. Also of note, each individual NPS was associated with GML in more than one region of the brain, a finding consistent with theories positing distributed brain system contributing to NPS-associated GML in AD [84, 85].

The cingulate [86, 87], prefrontal [88, 89], and temporal cortices [90] have been associated with psychiatric disorders beyond AD, which aligns with their roles in high-level cognitive, affective, and social processes. These three brain regions contribute to a shared set of cognitive processes including but not limited to executive functions, emotion regulation, and memory [91‒93]. Collectively, the functional associations of these brain regions may well underlie the associations we observed with AD-NPS. Future research should explore the involvement of cingulate, frontal, and temporal cortices in AD-NPS, considering potential therapeutic targets.

While this review highlights frequent associations between AD-NPS and the frontal, temporal, and cingulate cortices, some NPS were not associated with GML in these regions, including anxiety, irritability, hallucination, and depression. For instance, anxiety and irritability showed no association with frontal lobe GML, despite evidenced involvement of the prefrontal cortex-amygdala circuit in generalized anxiety disorder [94]. Additionally, while the irritability literature is sparse in non-demented adults, research in children has shown evidence of the association between irritability and increased functional activation in the dorsolateral prefrontal cortex and the inferior frontal gyrus [95]. Additionally, irritability and hallucination are associated with temporal lobe abnormalities in both schizophrenia [86, 96] and epilepsy [97], but these NPS-GML associations were not reported in this review. These findings may indicate a difference between the brain systems involved in these symptoms between AD and non-AD populations.

We found no association between depression and cingulate GML. This was somewhat surprising as the subcallosal cingulate is a deep brain stimulation site for treatment-resistant depression [98]. Additionally, depression in older adults has been associated with volumetric changes in the cingulate [99]. One possible reason for a lack of evidence supporting cingulate GML and depression in this review may be the exclusion criteria in the reviewed studies. Both studies observing depression and AD-associated GML utilized data from the ADNI study, and ADNI excluded participants with major depressive disorder. This exclusion criterion would necessarily have truncated the range of depressive symptomology represented in ADNI, and this may in turn have limited the capacity for secondary analysis of ADNI-originated data to observe outcomes associated with clinically meaningful levels of depression. However, with the exceptions of depression, anxiety, and irritability, the frequency of cingulate GML with AD-NPSs is compelling evidence of possible causal associations.

This review suggests a unique association between apathy in AD and GML in the basal ganglia. The basal ganglia play a critical role in the cortico-basal ganglia-thalamo-cortical loop. In this loop, the basal ganglia are responsible for resolving competing sensory and motor inputs that are involved in behavior control [100]. The basal ganglia’s role in resolving sensory and motor inputs may underlie deficits in purposeful behavior observed in apathy [101, 102]. Apathy has also been associated with basal ganglia dysfunction in non-AD neurodegenerative diseases and focal basal ganglia lesions [90]. While basal ganglia GML may not be unique to apathy in AD, there is an established connection between symptoms of apathy and the basal ganglia. Given the overlap between apathy and depression, future studies should use expert ratings to distinguish these symptoms for improved rigor.

This review highlights GML associations with AD-NPS, yet it is limited to structural MRI measures favoring region of interest analyses over whole-brain approaches. Though informative, GML cannot capture all known brain changes associated with AD-NPS (white matter, vasculature, functional connectivity, etc.). Nonetheless, many structural changes observed here align with PET and SPECT studies showing disordered metabolism and perfusion in AD [103‒110]. Further, functional brain network alterations would also provide more insight into NPS-associated brain changes. Examining intrinsic functional networks could deepen insights, as disruptions in networks like the default mode network are linked to several psychiatric disorders and are among the earliest changes in AD [111, 112].

Methodological differences (whole-brain or region of interest analysis) across studies may influence analysis and reporting. Further, many studies did not account for (i.e., adjust for) the impact of age and/or dementia severity on GML in their analyses. While certain patterns of GML are often associated with dementia, GML is also a well-characterized consequence of healthy aging [113‒115]. Controlling for age could help isolate disease-related degeneration, while leveraging established regions of interest implicated in both dementia and non-dementia behavioral symptoms may provide deeper insights into the interplay between brain changes and behavioral manifestations across AD and non-AD psychiatric disorders.

This review also relied on a clinical rather than a pathological diagnosis of AD which typically relies on an in vivo or postmortem confirmation of amyloid beta and tauopathies [116]. Without biomarkers, clinical diagnoses yield a “probable AD” diagnosis. Few studies provided biomarker confirmation of AD, which is an important context for interpreting these findings. Additionally, while symptoms such as psychosis and hallucination can be experienced in AD, these symptoms are also common in Lewy body dementia [117]. This overlap introduces a natural skepticism regarding the assurance that all participants in these studies had AD and not another form of dementia; comorbidity of AD and Lewy body pathologies (among others) are also common and could influence the specificity of our findings. Considering both pathological and clinical aspects of AD is therefore critical for understanding disease. Despite the advent of in vivo biomarkers for AD neuropathology, clinical insight remains essential for diagnosis, management, and treatment. Lastly, several studies in this review focused on early disease stages or combined multiple disease stages which may limit insight into symptoms prevalent in later stages. Further research on symptom progression and brain changes in advances stages of AD would address this gap.

While the NPS clusters reported by Aalten and colleagues [18] are commonly used to describe behavioral AD subsyndromes, it is important to acknowledge other noteworthy studies that report alternative clustering schemes. Cheng and colleagues [118] reported many of the same factors as Aalten et al. [18], while Canevelli and colleagues [119] found 34 different clusters across 15 studies. Such variation suggests the need for more descriptive clustering techniques. Many factors could influence clustering such as disease stage, cultural norms and differences, and sample size. Large, diverse samples may better define AD-NPS clusters.

Finally, terminology for NPS differs between disciplines. Some literature uses, “behavioral and psychological symptoms of dementia” or “neuropsychiatric features,” for example. Differences in NPS descriptors may have limited the articles identified by this review. Additionally, some articles may have used select NPS in their analysis but were not identified in the literature search due to limiting search terms to be in either the abstract or the title. These limitations affect all systematic reviews, but they remain noteworthy.

Our review of structural brain differences associated with AD-NPS noted especially frequent associations with frontal, temporal, and cingulate cortices across many different NPS as well as more distinct regional differences for some specific NPS. This suggests complex, multiregional involvement of GML in association with AD-NPS, although robust symptom-specific patterns were not observed. While this outcome was unexpected, these findings highlight key brain structures implicated in AD-NPS and may serve as a reference for studies aiming to elucidate symptom-specific GML patterns. The heterogeneity of AD-NPS-GML associations presents challenges for translation to clinical applications, and focused longitudinal studies of AD-NPS with structural brain imaging are needed. Nonetheless, our central observation is that regional neurodegeneration, particularly in the frontal, temporal, and cingulate regions, may be associated with NPS in AD. This finding represents a necessary first step towards an overall understanding of AD-NPS and patient-specific variability in NPS among patients with AD.

A statement of ethics is not applicable because this study is based exclusively on published literature.

The authors have no conflicts of interest to declare.

Research reported in this publication was supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under Award No. F99NS139537.

Conceptualization: M.K.R., C.J.P., D.L.M., J.B., V.S.P., and D.E.W.; methodology: M.K.R., D.L.M., J.N.B., V.S.P., and D.E.W.; validation: M.K.R. and D.E.W.; data curation: M.K.R.; writing – original draft preparation: M.K.R.; writing – review and editing: M.K.R, C.J.P., D.L.M, J.N.B., V.S.P., and D.E.W.; visualization: M.K.R., C.J.P., and D.E.W.; and supervision: D.E.W. All authors have read and agreed to the published version of the manuscript.

All data generated or analyzed during this study are included in this article and its online supplementary material. Further inquiries can be directed to the corresponding author.