It is increasingly clear that the public health burden of Alzheimer’s disease (AD) is not solely due to cognitive decline but also to neuropsychiatric symptoms (NPS). NPS such as agitation, depression, anxiety, and apathy are highly prevalent in AD (Tschanz et al., Reference Tschanz2011) and increase caregiver burden, health care costs, and risk of institutionalization (Storti et al., Reference Storti, Quintino, Silva, Kusumota and Marques2016). Nonpharmacologic interventions are considered first-line treatment for NPS in AD but are not consistently used nor always effective and there are no FDA-approved medications for NPS in AD (although risperidone is approved for short-term treatment of agitation in Europe). Several drugs are currently being studied for repurposing for NPS in AD including methylphenidate for apathy (Rosenberg et al., Reference Rosenberg2013) and selective serotonin reuptake inhibitors (Porsteinsson et al., Reference Porsteinsson2014), cannabinoids (Outen et al., Reference Outen2021), and prazosin for agitation in AD among others. But treatments might be more effective if we could target biological mechanisms.

While our understanding of mechanisms of NPS in AD is at a much earlier stage than our understanding of core AD mechanisms of cognition, there are increasing clues from neuroimaging studies. Depression in AD has been associated with decreased volume or cortical thickness of the anterior cingulate cortex (ACC) in several studies (Liu et al., Reference Liu2017; Zahodne et al., Reference Zahodne, Gongvatana, Cohen, Ott and Tremont2013) as well as decreased hippocampal volume (Elcombe et al., Reference Elcombe2015; Xie et al., Reference Xie2013). Apathy has been associated with changes to the frontal-subcortical reward circuitry including the inferior temporal cortex, posterior cingulate cortex (PCC), and frontal lobe structures including the orbitofrontal cortex as well as the ACC (Rosenberg et al., Reference Rosenberg, Nowrangi and Lyketsos2015). Agitation in AD is innately difficult to study due to the challenges of imaging agitated patients, but the limited data to date suggest associations with atrophy in the left inferior frontal/insula and bilateral retrosplenial cortices (Rafii et al., Reference Rafii2014) and with frontolimbic atrophy (Trzepacz et al., Reference Trzepacz2013), as well as decreased white matter integrity in the anterior cingulum bundle (Tighe et al., Reference Tighe2012). Psychosis in AD shows prominent associations with atrophy of the temporal and parietal lobes, and less so with frontal and occipital lobe structures and subcortical structures (Fischer et al., Reference Fischer, Ting, Millikin, Ismail and Schweizer2016). It is important to note that most prior studies report associations between NPS and decreased regional brain volumes and cortical thickness and white matter integrity.

This is the context to consider “Neuropsychiatric symptoms and brain morphology in patients with mild cognitive impairment and Alzheimer’s disease dementia” by Siafarikas et al. (Reference Siafarikas2021) in this issue of International Psychogeriatrics. They examined the cross-sectional associations of NPS with MRI features in 235 subjects with mild cognitive impairment (MCI) or AD. NPS was measured with the Neuropsychiatric Inventory-Informant version (NPI-Q), a widely used instrument for assessing NPS in AD in 12 neuropsychiatric domains. To address the issue of multiple comparisons given the 12 NPI-Q domains, they used principal components analysis to identify four subfactors of NPS including elation, psychosis, depression, and motor behavior. Additionally, they examined associations of MRI features with the NPI domains. They measured 170 MRI features including regional brain volumes and cortical thickness, using a voxel-based (unbiased) analysis. The strengths of the study include a large sample size and sophisticated MRI methods.

The authors report a mixture of novel and replicative findings. Elation was associated with greater thickness of right caudal ACC. Previously reported associations of apathy, anxiety, and sleep- and night-time behavior to the left temporal lobe were replicated. There was an interaction with diagnosis/disease stage: the associations of psychosis with left postcentral cortical thickness and depression with left superior sulcus cortical thickness were positive in MCI and negative in AD.

The association of the elation factor with increased ACC cortical thickness is, to our knowledge, new to the field. The NPI “elation” domain consists of clinical features that are similar to mania or hypomania of bipolar disorder, but in this study the elation subfactor is weighted to NPI items “elation” and “disinhibition” which reflects not only increased energy and activity but impulsivity and poor judgment and therefore more directly relevant to common presentations of AD rather than psychiatric presentations. A similar subfactor was identified by Vilalta-Franch et al. (Reference Vilalta-Franch2010) who described a “hypomanic” group in which symptoms of euphoria and disinhibition clustered together.

The ACC has been thought of as an executive region connecting prefrontal cognitive skills with limbic regions involved with affective regulation (Nowrangi et al., Reference Nowrangi2021) and has been previously linked to euphoria. An important question is how to interpret the meaning of increased cortical thickness there. Cortical thickness has been largely interpreted as a reflection of underlying volume, and thus many brain regions show cortical thinning in AD, presumably reflecting neuronal loss due to neurodegeneration and resulting in increased clinical dysfunction. The meaning of increased cortical thickness as revealed by MRI is not clear. Do brain regions and the constituent neuronal cells become hypertrophic, reorganize (arborize), or proliferate as a compensation or as a sign of injury to underlying physiological changes? Are there subtle signs of vascularization, edema, or inflammation not detected by MRI (e.g. at high enough field strength) that may account for increased volume? Or is increased thickness due to changes in nonneuronal cell types such as astrocytes and microglia? What role does the temporal course of the disease and symptom presentation play on brain structure? These questions are largely unanswered in the study of NPS of AD. It would be helpful to have pathological correlations to better understand these mechanisms, but findings of increased MRI volumes in other diseases may provide additional context. For example, increase in prefrontal cortical volume was found in first-episode, treatment-naïve MDD patients (Qiu et al., Reference Qiu2014). Parkinson’s patients early in the disease show an increase in volume in regions comprising the dorsal and ventral attention networks (Sawczak et al., Reference Sawczak, Barnett and Cohn2019).

It seems then that the interaction with diagnosis (MCI vs. AD) is particularly important and may point to the significance of disease stage, suggesting that earlier in the disease psychosis and depression are associated with increased cortical thickness, and later with cortical thinning. As with elation, the meaning of increased cortical thickness is not straightforward to interpret, but the assertion that the brain’s initial response is to compensate for deficits early in disease by increasing cortical thickness (through various mechanisms) but fails later in the course of disease resulting in cortical thinning is interesting. It is a hypothesis that has been suggested before as an inverted U shape pattern of cortical thickness alterations. The interaction has been seen in other AD studies related to presenilin-1 mutation carriers who follow a nonlinear trajectory with regional increases during a presymptomatic period followed by decreases over time (Fortea et al., Reference Fortea2011; Sala-Llonch et al., Reference Sala-Llonch2015).

The psychosis subfactor in this study was weighted toward hallucinations and delusions which is to be expected. In this study, psychosis was associated with decreased volume of left postcentral gyrus which is in line with other research indicating associations with regions in the parietal lobe (Bruen et al., Reference Bruen, McGeown, Shanks and Venneri2008; Rafii et al., Reference Rafii2014) including the PCC (Fischer et al., Reference Fischer, Ting, Millikin, Ismail and Schweizer2016). The postcentral gyrus includes primary somatosensory cortex, which has been implicated in processing somatic sensations such as touch, proprioception, and pain, which may be associated with hallucinations, but prior reports associating delusions with right frontal atrophy were not replicated. Depression was associated with cortical thickness in the left superior temporal gyrus, roughly replicative of prior reports of decreased temporal lobe volume (Hayata et al., Reference Hayata2015; Lebedev et al., Reference Lebedev, Beyer, Fritze, Westman, Ballard and Aarsland2014; Lee et al., Reference Lee2012). It should be noted that their depression factor was similarly weighted toward depression and apathy and thus may reflect mechanisms underlying both, which may account for differing results from those in the literature.

A discussion of how increased or decreased brain structural volume is related to clinical presentation would not be complete without considering their function as components of coherent circuits or networks. Born out of Norman Geschwind’s pioneering work in disconnection syndromes, a “disconnectionist” framework provides this field with the understanding that higher functional deficits result from lesions of white matter between association cortices that act as relay stations between primary motor, sensory, and limbic areas (Catani and ffytche, Reference Catani and ffytche2005). This framework has taken shape within the Alzheimer’s Association Neuropsychiatric Syndromes of AD Professional Interest Group which put forth a tripartite hypothetical model based on proposed networks and circuits (Geda et al., Reference Geda2013). Taking the ACC again as an example, it has been implicated in both depression and apathy where there are overlapping features to both: initiation, motivation, and goal-directed behaviors. The rich connectivity between the ACC and other neural structures such as those within the limbic lobe (e.g. hippocampus, amygdala, anterior thalamus) suggests its role as a “behavioral hub” instrumental for affective regulation. Further exploration of these potential connections may reveal features further differentiating depression from apathy. As such, future research would be well-served to consider NPS of AD within a network/circuit framework to more fully capture the breadth of these complex behavioral presentations; however, a similar discussion of what increased or decreased connectivity (instead of volume) mean is likely to continue ensue.

The authors chose a data-driven approach to reducing the number of NPS variables using PCA and report a four-factor solution. This approach gives advantages in the statistical analyses by reducing the number of comparisons, but each factor is a new phenotype which merits interpretation. Their “psychosis” factor is quite readily interpretable as being weighted toward delusions and hallucinations. Their “elation” and “depression” factors are closely related to the corresponding NPI domains but with a different mix of symptoms that include other common NPS in dementia; “elation” has significant weight from disinhibition and “depression” from apathy. It is not clear whether these factors are a more or less accurate way to describe NPS phenotypes in dementia or whether these results are replicable. Moreover, a systematic review of PCA analyses of NPI domains reported low concordance between different studies, the only highly replicable domain being psychosis (Canevelli et al., Reference Canevelli2013).

There are two ways in which these results may impact this rapidly changing field. First, there is increasing evidence that NPS in cognitively unimpaired adults may be risk factors for future dementia, and NPS frequently are evident prior to any cognitive complaints or symptoms (Wise et al., Reference Wise, Rosenberg, Lyketsos and Leoutsakos2019). This has led to the concept of mild behavioral impairment (MBI), a parallel concept to MCI in which NPS in five domains are predictors of dementia, including drive and motivation, mood and affect, impulse control, social appropriateness, thought, and perception. There is a validated checklist for assessing MBI severity (MBI-C; Ismail et al., Reference Ismail2017) and MBI has been reported to be associated with early AD cognitive changes, biomarkers, and pathology. The NPI-Q factors in this study partly map onto MBI domains and may reflect MBI mechanisms.

Second, this study and others like it offer the possibility of developing mechanistically based treatments for NPS in MCI and AD, with the further possibility that we might find preventive interventions for older adults with MBI. Newer therapeutic modalities such as repetitive transcranial magnetic stimulation and transcranial direct current stimulation, for example, are in their developmental infancy with respect to cognitive and NPS symptoms but are appealing because of their minimal invasiveness, low side-effect profile, and ability to target functional brain networks. It is clear that the study of NPS in AD is early, but the need for other novel mechanistically based therapeutics is needed. This is the excitement and challenge of following up on these novel findings.

Understanding the biological mechanisms underlying NPS in AD is related to the theme of this issue of International Psychogeriatrics. NPS is one of the major contributors to caregiver distress and management of NPS falls squarely on caregiver’s shoulders. As outlined above, the hope is that improved understanding of mechanism will lead to improved treatments for NPS, and especially for “person-centered” personalized treatment strategies which may directly improve the lives of AD patients and their caregivers alike.