Examining Differences in Neuropsychiatric Symptom Factor Trajectories in Empirically Derived Mild Cognitive Impairment Subtypes

Abstract

Objective:

The aim of this study was to examine neuropsychiatric symptom (NPS) factor severity progression over time in empirically derived (ED) mild cognitive impairment (MCI) subtypes.

Methods:

Participants in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study diagnosed with MCI by ADNI protocol using conventional clinical (CC) criteria (n=788) were re-classified using cluster analysis as amnestic, dysnomic, dysexecutive MCI or cluster-derived normal (CC-Normal) using empirical criteria. Cognitively normal(CN) participants (n=207) were also identified. The Neuropsychiatric Inventory-Questionnaire (NPI-Q) was administered from baseline through 48-month follow-up. Exploratory factor analysis (EFA) was completed to determine the NPI-Q factor structure at 6-month follow-up. Multilevel modeling was used to determine NPI-Q symptom severity factor and apathy symptom progression over time by cognitive subtype.

Results:

The EFA revealed that the NPI-Q consisted of two factors: hyperactivity/agitation and mood symptoms. Using clinical and empirical criteria, all MCI groups were identified as having more severe hyperactivity/agitation symptoms than CN participants. However, only the amnestic MCI group identified using empirical criteria showed an increase in symptom severity over time relative to CN participants. Mood factor and apathy symptoms were found to be more severe in dysexecutive and amnestic groups in both models. Similarly, both models identified a significant worsening of mood and apathy symptoms over time for dysexecutive and amnestic groups relative to CN participants.

Conclusions:

This study provides further support that empirical criteria aids in examining the progression of clinical characteristics associated with MCI. Further, it helps to identify which MCI subtypes may be at higher risk for NPS progression.

Introduction

Alzheimer’s Disease (AD) is a growing public health problem, with an estimated 5.5 million people living with the disease in 2017 in the United States. By 2050, these rates are projected to increase to approximately 16 million people in the US¹. There is substantial biological and clinical evidence that neuropathological processes begin years before functional decline is evident². Therefore, the understanding and identification of prodromal elements of AD is crucial for future disease intervention.

The most widely studied prodromal state of progressive dementias such as AD is mild cognitive impairment (MCI), a condition that is conceptualized as the transition between normal cognition and dementia. Patients with MCI have subjective and/or objective neuropsychological evidence of mild memory complaints in the absence of significant functional decline³. In recent years, a series of studies have highlighted the need for empirically derived MCI diagnosis due to increased susceptibility of false positive errors when utilizing conventional MCI criteria^3,4. Conventional criteria utilizes cognitive screening measures (e.g., MMSE), limited neuropsychological assessment (e.g., one memory subtest), and clinical judgement to make a diagnosis⁵ whereas empirical criteria relies heavily on broader neuropsychological assessment (e.g., using several tests to measure across cognitive domains) as well as clinical judgment. Research using empirical criteria has identified several subtypes of MCI: individuals with a primary impairment in memory (amnestic) or a primary impairment in other cognitive domains (non-amnestic) such as a primary impairment in language (dysnomic), a primary impairment in executive function (dysexecutive), as well as a subtype with impairments in more than one domain (multidomain MCI)^6,7.

The identification of these subtypes has shown to improve diagnostic precision of MCI³ and has clarified relationships with prodromal biomarkers of AD as well as identified differing disease trajectories among subtypes^6,8. Further, a study by Thomas and co-workers (2017) identified differing functional status trajectories across empirically derived MCI subtypes, such that dysexecutive individuals experienced steeper rates of functional decline over time than dysnomic or amnestic groups⁹. While many studies have examined biomarkers in empirically derived MCI subtypes, prodromal clinical markers such as neuropsychiatric symptoms (NPS), have not been yet examined in the context of these subtypes.

NPS are psychiatric or behavioral symptoms of dementia that often precede significant cognitive decline¹⁰. NPS are prevalent in both early and later stages of progressive dementias with approximately 50 percent of individuals with MCI experiencing at least one NPS¹¹. The presence of NPS, even of mild severity, is linked with accelerated and higher rates of progression from MCI to dementia¹⁰, increased functional impairment¹², higher need for institutionalized care¹³, poorer quality of life¹⁴, substantial neuropathological burden^15,16, and increased caregiver burden¹⁷.

Investigations into the role that NPS play in MCI as well as risk for conversion to AD have been limited. The majority of studies utilize conventional rather than empirical diagnostic criteria^18,19, examine a single NPS (e.g., anxiety), or examine the overall number or severity of NPS². Previous research suggests that clinically derived subtypes of MCI may have differing NPS profiles, although the literature is mixed regarding these relationships^18,20. Previous work suggests that individuals with amnestic MCI may be more likely to develop NPS, have more severe NPS trajectories, and be more likely to progress to dementia than those with non-amnestic MCI^20,21. Although, other studies found few associations between amnestic MCI and NPS¹⁸. Prodromal differences in NPS profiles may result from early neurodegeneration that is not yet detectable by screening measures.

Recent neuroimaging research provides support for differing NPS profiles by linking certain NPS to brain regions that are commonly affected in specific MCI subtypes. For example, Pa and coworkers (2009) found that individuals with left prefrontal atrophy, a region impaired in those with dysexecutive MCI, were more likely to experience behavioral (e.g., disinhibition) and emotional regulation problems²².

This study proposes to build on previous work by examining the presence and progression in severity of NPS factors in the context of empirically derived subtypes of MCI. Examining NPS symptom factors rather than specific symptoms is likely to lead to increased ease in identifying associations between symptoms and trajectories as well as making intervention more feasible^23,24. Given that NPS often serve as prodromal warning signs of cognitive decline, a better understanding of how these symptoms operate in MCI is likely to facilitate potential for earlier detection and intervention for both cognitive and psychiatric symptoms¹⁰.

Methods

Participants

The present study was approved by the Louisiana State University institutional review board. The study included 788 MCI participants and 207 cognitively normal(CN) participants from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Criteria for ADNI eligibility and diagnostic classifications are described at: http://www.adni-info.org/Scientists/ADNIGrant/ProtocolSummary.aspx. All participants were 55–91 years old, had a modified Hachinski score ≤ 4, and had an informant whom was able to provide an evaluation of functioning. See Table 1 for demographic information. The CN group included individuals who remained cognitively normal throughout their participation in ADNI.

Table 1.

Sample Demographics

	Age, years M(SD)	Gender, % female	Education, years M(SD)	Symptom Severity M(SD)
Conventional Criteria
Cognitively Normal	76.3(5.4)	47.0%	16.0(2.8)	0.62(1.3)
Cluster-Derived Normal	74.4(8.2)	39.1%	16.3(2.6)	1.89(2.6)
Dysexecutive MCI	74.8(7.1)	38.3%	14.5(3.5)	2.52(3.3)
Amnestic MCI	73.2(7.0)	35.6%	15.9(2.8)	2.45(2.9)
Neuropsychological Criteria
Cognitively Normal	76.3 (5.4)	47.0%	16.0(2.8)	0.62(1.3)
Dysexecutive/mixed MCI	75.1(7.3)	38.9%	14.4(3.7)	2.65(3.6)
Dysnomic MCI	74.0(6.0)	36.1%	16.2(2.7)	1.97(2.5)
Amnestic MCI	73.6(7.1)	39.0%	15.9(2.8)	2.37(2.9)

Measures

Neuropsychiatric Inventory-Questionnaire (NPI-Q)

The NPI-Q²⁵ is a brief 12-item, informant-based questionnaire that assesses psychopathology that commonly occurs in progressive cognitive decline. In the ADNI study, the NPI-Q is administered at baseline as well as every 6-months during follow-up visits.

Neuropsychological Testing

Following baseline ADNI diagnosis, participants completed a neuropsychological test battery at baseline and at yearly follow-up. The battery included tests of episodic memory (Rey Auditory Verbal Learning Test [AVLT])²⁶, language (Category Fluency Test-Animals)²⁷, Boston Naming Test (30-items)²⁸, and attention/executive functioning (Trail-Making Test Parts A and B)²⁹.

Clinical Diagnostic Classification

MCI was diagnosed by ADNI using the following conventional clinical (CC) criteria: 1) self-, informant-, and clinician-reported memory problems; 2) performing below an education-adjusted normative cut-off score on delayed recall of Story A of the WMS-R Logical Memory Test 3), global clinical dementia rating (CDR) score of 0.5; and 4) adequate general cognitive and functional performance such that a diagnosis of dementia could not be made by the site physician at baseline ³⁰.

Neuropsychological Diagnostic Classification

Empirically derived (ED) diagnostic classification was determined by the procedure outlined by Bondi and colleagues (2014)⁶. All ADNI CN and MCI participants were reclassified using the Jak/Bondi neuropsychological-based method applied to their baseline data. Five neuropsychological subtests from ADNI were chosen because of their well-established use in the assessment of early cognitive decline and consistent administration across all four ADNI study periods⁶. These measures were not used in determining ADNI diagnosis at initial screening as only the WMS-IV Logical Memory subtest was used per ADNI study procedure. Each measure was converted to an age-corrected z-score using widely-accepted normative data as outlined by Bondi and colleagues⁶: Mayo Older Americans Normative Study³¹ for the Rey AVLT and National Alzheimer’s Coordinating Center normative^27,32 for the remaining neuropsychological measures. Participants were considered to have MCI if any one of the following criteria were met: 1) they performed >1 SD below the age-corrected normative mean, on both scores within at least one cognitive domain (i.e., memory, language, or attention/executive function); 2) performed >1 SD below the age-corrected normative mean, in each of the three cognitive domains.

Statistical Analyses

Descriptive Statistics

Differences in baseline demographic characteristics between groups were evaluated using chi-square tests for categorical variables. Analysis of variance (ANOVA) was used to examine differences between groups for continuous demographic and neuropsychological variables.

Cluster Analysis

Cluster analytic methodology outlined in Bondi et al. (2014) was utilized to determine MCI subtype⁶. In the first hierarchical cluster analysis, individuals diagnosed as MCI using CC criteria was completed. In previous studies as well as the current study, the first cluster analysis aids in identifying individuals who have been diagnosed with MCI, but are cognitively normal by neuropsychological standards⁶. A second hierarchical analysis was conducted using ED criteria, which has been used in previous studies to identify an additional ED-MCI cluster⁶. Ward’s method was used in both analyses to calculate the distance between each cluster (squared Euclidean distance) and merge clusters together that produced the smallest increase in overall distances within clusters.

Factor Analysis

The factor structure of the NPI-Q was evaluated using exploratory factor analysis (EFA) with an oblique rotation, as significant correlation among the factors was expected (Mplus, 7.2)³³. A robust estimator appropriate for ordinal item responses was used (WLSMV) as there are only four response categories on the NPI-Q severity scale³⁴. Two questions (delusions and hallucinations) were excluded from analyses due to low base rate in the sample. Based on the findings of the factor analysis, composites were calculated by summing the NPI-Q severity scores. For the current study, the baseline assessment was excluded from analysis due to a lack of variability in symptom severity³⁵ and because fewer participants were given the NPI-Q at baseline compared to at their 6-month follow-up visit. Therefore, by beginning analysis with the 6-month visit rather than at baseline, this allowed for a larger sample size.

Multilevel Modeling(MLM)

MLM was used to determine whether there were differences in NPI-Q symptom severity composite scores over time by group. The visit variable included 8 time points that spanned over 4 years. The first analysis using CC criteria included four groups: a cognitively normal group(CN), a cluster-derived normal group (CC-Normal), a dysexecutive MCI group (CC-EF), and an amnestic MCI (CC-Amnestic) group. The second analysis using ED criteria, included four groups: a cognitively normal group (CN), a dysexecutive MCI group (ED-EF), a dysnomic MCI group (ED-Lang), and an amnestic MCI group(ED-Amnestic). In both analyses, CN participants were used as the reference group. Age and education were included as covariates as they were found to differ between groups. The intercept was included as a random effect in the model.

Results

Cluster Analysis

Cluster analysis of 788 participants diagnosed with MCI using CC criteria resulted in three subtypes. 326 participants who performed within the normal range on all neuropsychological measures were considered to be cluster derived normal (CC-Normal). 336 individuals who were primarily impaired on memory measures were considered to be amnestic MCI (CC-Amnestic). 126 participants who performed most poorly on attention/executive function measures and had mildly impaired performance on memory and language were considered to have dysexecutive/mixed MCI (CC-EF). At baseline, significant differences were found between groups on age, education, and overall NPI-Q symptom severity, but not gender. See Table 1 for demographic characteristics of the sample.

The second hierarchical cluster analysis using the 462 individuals diagnosed with MCI using the ED criteria resulted in three subtypes. 300 individuals who were primarily impaired on memory measures were considered to be amnestic MCI (ED-Amnestic). 100 participants who performed most poorly on attention/executive function measures and had mildly impaired performance on memory and language were considered to have dysexecutive/mixed MCI (ED-EF). 62 participants who had primary impairment on language measures and mild impairments in other domains were classified as dysnomic (ED-Lang). At baseline, significant differences were found between groups on age, education, and overall NPI symptom severity, but not gender. See Table 1 for demographic descriptive statistics.

Factor Analysis

The results of the EFA are shown in Table 2. The EFA revealed a two-factor solution: the first factor consisted of symptoms that are related to “agitation/hyperactivity” and the second factor consisted of symptoms related to “mood problems”. Apathy was not included in either factor due to high cross-loading on both factors, however, a follow-up analysis was conducted to examine differing trajectories of apathy severity amongst MCI subtypes.

Table 2.

Factor Loadings of NPI-Q Scores at the 6-Month Visit

Symptom	Agitation/Hyperactivity	Mood Problems
Agitation/Aggression	.67	.24
Dysphoria/Depression	−.01	.76
Anxiety	.22	.53
Euphoria/Elation	.71	−.03
Apathy/Indifference	.40	.34
Disinhibition	.80	.00
Irritability/Lability	.59	.25
Aberrant Motor	.56	−.03
Nighttime Behavior	.07	.55
Appetite/Eating	.00	.52

Note: Bold indicates factor loading of .4 or greater

Multilevel Modeling(MLM)

In both the CC and ED models that examined agitation/hyperactivity symptoms, there were main effects for group, such that the CC-Normal, CC-EF, ED-EF, CC-Amnestic, and ED-Amnestic groups exhibited more symptoms than CN participants. There were no main effects of visit, age, or education. In the model using CC criteria, no group by visit interaction was significant. However, in the ED criteria model a group by visit interaction was observed for the ED-Amnestic group, which suggests that ED-Amnestic individuals experienced more severe agitation/hyperactivity over time. See Table 3 for a summary of agitation/hyperactivity model results.

Table 3.

Estimates and Effect Sizes of Variables in the Agitation/Hyperactivity Model

	Estimate	S.E.	t
Conventional Criteria
Intercept	.49	.46	1.06
Demographic Variables
Age	.00	.01	−.51
Education	.00	.01	.32
Visit	.01	.02	.69
Group main effect
Cluster-derived Normal**	.38	.13	3.02
Dysexecutive MCI***	.80	.17	4.87
Amnestic MCI***	.72	.13	5.55
Group × visit
Cluster-derived Normal × visit	.03	.24	1.08
Dysexecutive MCI × visit	.02	.03	.49
Amnestic MCI × visit	.04	.03	1.43
Neuropsychological Criteria
Intercept	.57	.46	1.22
Demographic Variables
Age	.00	.00	−.60
Education	.00	.01	.16
Visit	.01	.02	.70
Group main effect
Dysexecutive MCI***	.80	.17	4.58
Dysnomic MCI***	.47	.13	3.65
Amnestic MCI***	.63	.13	4.99
Group × visit
Dysexecutive MCI × visit	.02	.03	.50
Dysnomic MCI × visit	.01	.02	.31
Amnestic MCI × visit*	.05	.02	2.01

Note: S.E. = Standard Error; MCI = Mild Cognitive Impairment

^*p<.05,

^**p < .01,

^***p < .001

With respect to mood symptoms (See Table 4), there were no differences in the models utilizing CC or ED criteria, which demonstrated group main effects for CC-Normal, CC-EF, ED-EF, CC-Amnestic, and ED-Amnestic groups. These results indicate that these individuals experience more severe mood symptoms than CN participants. There were no main effects of visit, age, or education. Group by visit interactions were observed for CC-EF, ED-EF, CC-Amnestic and ED-Amnestic groups, but not for ED-Lang participants. This indicates that regardless of criteria, dysexecutive and amnestic groups, but not ED-Lang individuals, experienced more severe mood problems over time.

Table 4.

Estimates and Effect Sizes of Variables in the Mood Model

	Estimate	S.E.	t
Conventional Criteria
Intercept	.50	.44	1.14
Demographic Variables
Age	.00	.01	−.23
Education	.00	.01	.18
Visit	.02	.02	.79
Group main effect
Cluster-derived Normal**	.40	.13	3.31
Dysexecutive MCI***	.63	.16	3.93
Amnestic MCI***	.57	.13	4.52
Group × visit
Cluster-derived Normal × visit	.02	.03	.57
Dysexecutive MCI × visit*	.08	.03	2.07
Amnestic MCI × visit**	.09	.03	3.36
Neuropsychological Criteria
Intercept	.61	.44	1.40
Demographic Variables
Age	.00	.01	−.44
Education	.00	.01	.01
Visit	.02	.02	.82
Group main effect
Dysexecutive MCI***	.66	.17	3.86
Dysnomic MCI***	.44	.13	3.51
Amnestic MCI***	.54	.13	4.29
Group × visit
Dysexecutive MCI × visit*	.07	.04	2.00
Dysnomic MCI × visit	.02	.03	.74
Amnestic MCI × visit**	.08	.03	2.88

Note: S.E. = Standard Error; MCI = Mild Cognitive Impairment

^*p<.05,

^**p < .01,

^***p < .001

Follow-up analyses examining the trajectory of apathy symptoms among differing MCI subtypes using CC criteria revealed that CC-Normal, CC-Amnestic, and CC-EF participants experienced more apathy symptoms at baseline when compared to CN participants. However, group by visit interactions were only observed for CC-Amnestic and CC-EF groups, indicating that these groups experience more apathy symptoms over time. The model utilizing ED criteria demonstrated that all ED-MCI groups experienced more severe apathy symptoms at baseline compared to CN participants. Similar to CC findings, group by visit interactions were revealed for ED-Amnestic and ED-EF groups, indicating that these groups experience more apathy symptoms over time. The group by visit interaction was not observed in the ED-Lang group. See Table 5 for results.

Table 5.

Estimates and Effect Sizes of Variables in the Apathy Symptom Model

	Estimate	S.E.	t
Conventional Criteria
Intercept	.00	.17	.01
Demographic Variables
Age	.00	.00	.09
Education
Visit	.01	.01	.64
Group main effect
Cluster-derived Normal*	.12	.53	2.36
Dysnomic MCI**	.14	.07	2.16
Amnestic MCI***	.23	.05	4.22
Group × visit
Cluster-derived Normal × visit	.01	.01	.53
Dysexecutive MCI × visit***	.06	.01	3.97
Amnestic MCI × visit**	.03	.01	2.39
Neuropsychological Criteria
Intercept	.33	.23	.14
Demographic Variables
Age	−.00	.00	−1.40
Education	−.00	.01	−35
Visit	.01	.01	.42
Group main effect
Dysexecutive MCI*	.14	.07	1.93
Dysnomic MCI***	.29	.09	3.24
Amnestic MCI***	.21	.06	3.59
Group × visit
Dysexecutive MCI × visit***	.06	.02	3.76
Dysnomic MCI × visit	−.00	.02	−.32
Amnestic MCI × visit**	.03	.01	2.75

Note: S.E. = Standard Error; MCI = Mild Cognitive Impairment

^*p<.05,

^**p < .01,

^***p < .001

Discussion

The current study provides further support that NPS differ in their reported severity level over time in different MCI subtypes. The current study utilized both the CC ADNI criteria as well as ED criteria to due to findings of increased false positive error rates when utilizing CC criteria^4,36. Cluster analyses using an updated database with additional participants identified similar groups as in prior analyses of ADNI data⁶.

To identify NPS factors, an EFA was conducted, in which items loaded on two factors: an “agitation/hyperactivity” factor and a “mood problems” factor. These findings are similar to results found in other studies^24,37. Other studies have identified a “psychosis” factor²⁴, but due to a low base rate in the current sample, this study did not identify a “psychosis” factor. Apathy demonstrated high cross-loadings on both factors. This is consistent with prior work in patients with dementia showing that apathy is not consistently related to one set of NPS³⁸. Due to research suggesting its importance as a feature of dementia, apathy was analyzed separately³⁹. In the current study, all CC- and ED-MCI participants displayed increased symptoms of apathy compared to CN participants. However, only the CC-Amnestic, CC-EF, ED-Amnestic, and ED-EF groups experienced these increased symptoms over time. These results are in line with previous research identifying increased levels of apathy in those at risk for dementia ^40,41.

Further, all MCI participants displayed more agitation/hyperactivity and mood symptoms at baseline compared to CN participants. This finding was consistent regardless of whether CC criteria or ED criteria were used. CC criteria and ED criteria also both identified the dysexecutive and amnestic groups as experiencing more severe agitation/hyperactivity symptoms than CN participants. However, only the ED criteria identified that the amnestic group experienced more severe agitation/hyperactivity symptoms over time, whereas using CC criteria did not. This finding suggests that initial classification with the CC criteria may not be sensitive enough to identify changes in these symptoms over time. NPS (e.g., agitation) are clinically important for several reasons including their association with increased mortality and faster rate of decline in patients with AD⁴². Although symptom severity did not differentially increase over time compared to other groups, baseline symptom severity of agitation and aggression was highest in the ED-EF group. This is consistent with research showing atrophy of frontal regions of the brain is associated with increased symptoms of agitation³⁷ and individuals diagnosed with dysexecutive MCI demonstrate atrophy in frontal regions of the brain²².

Regarding mood symptoms, there were no differences in model prediction as both the CC and ED criteria models demonstrated that amnestic and dysexecutive individuals experienced more severe symptoms over time. This finding is consistent with research demonstrating that individuals of different subtypes of MCI⁴³ frequently endorse depressive symptomatology.

In the current study, the CC-Normal group exhibited increased agitation/hyperactivity symptoms and mood symptoms compared to CN participants. However, this group did not show significant increases in either agitation/hyperactivity or mood symptoms over time. It is possible that the CC-Normal group may be classified in other studies as individuals with subjective cognitive decline (SCD). Individuals experiencing NPS, such as anxiety or depression are more likely to report SCD, even when not displaying objective cognitive impairments⁴⁴. Despite not demonstrating impairments during initial testing, SCD has been shown to be associated with future cognitive decline⁴⁵. Therefore, individuals reporting SCD warrant follow-up testing to monitor for decline in objective cognitive performance over time. Previous research utilizing the ADNI database has found that individuals reporting SCD have been shown to have to have fewer CSF biomarkers for AD and were less likely to convert to AD than the ED MCI subtypes in a prior study³.

Alternatively, the elevated symptoms in this group at baseline may represent one reason for potential misdiagnosis of MCI given their association with objective cognitive performance and subjective cognitive complaints. NPS symptoms, such as anxiety or depression, are associated with a decline in memory performance in healthy functional older adults^15,46 as well as reduced processing speed^47,48. One CC criteria for MCI is subjective cognitive complaints and subjective cognitive complaints may be related to depressive symptoms⁴⁴.

Overall, accurate diagnosis and identification of NPS, and how they change over time, has the potential for numerous short- and long-term health benefits. The current study has demonstrated that use of ED criteria predicted the onset of more severe agitation/hyperactivity symptoms over time. Through the use of the ED criteria, diagnoses will become more accurate, and new treatment interventions can be established. Due to the different NPS experienced by individuals in the different MCI subtypes, different treatments may be more beneficial for one subtype of MCI as compared to the others. For example, nonpharmacological interventions tailored for individuals with dementia demonstrated effectiveness in reducing symptoms such as agitation and depression. Therefore, identifying NPS profiles within ED subtypes of MCI has implications for diagnostic clarity and treatment effectiveness.

Limitations

The ADNI database provides an excellent opportunity to study longitudinal relationships among variables related to cognitive decline in potential neurodegenerative processes. The benefits of utilizing this database include a large sample size consisting of clinical and non-clinical populations. However, there are some limitations to the current study, including those associated with analyzing pre-collected data such as sample collected, study methodology, alterations to protocols and being limited to specific assessments^49,50.

Previous research suggests that NPS prevalence, severity, and trajectories may differ between types of dementia (e.g., AD type vs. frontotemporal dementia). Given that ADNI aims to recruit participants that are on the trajectory for or have AD-type dementia, this limits the generalizability of these findings to individuals with AD dementia⁵¹. Further, ADNI protocol requires exclusion of participants with more severe NPS symptoms (e.g., psychosis) at baseline. Therefore, this may help to explain why a psychosis factor was not identified in the current study but has been identified in previous research.

Due to ADNI’s use of CC criteria which relies heavily on the presence of memory impairment to diagnose MCI, individuals who have presentations consistent with non-amnestic MCI may not have been included in the initial sample. Further, individuals re-classified with ED criteria (e.g., ED-Lang) show primary impairment in one domain (e.g., language), but may also demonstrate memory impairments. Therefore, these individuals may represent a MCI subtype described in other studies as “MCI, Amnestic, Multidomain”⁵². Although individuals with non-amnestic MCI presentation have been described, the prevalence rates of these MCI subtypes are not clear.

Regarding limitations related to methodology, given that fewer participants were administered the NPI-Q at baseline than at their 6-month follow up, analysis of the NPI-Q items began at the 6-month follow up. Therefore, symptoms prior to the onset of cognitive testing are not known. Additionally, although other research has utilized the NPI-Q to measure NPS in individuals with MCI, it was developed to assess NPS in individuals with dementia⁵³. Therefore, the utilization of the NPI-Q with a population that may not have symptoms severe enough to warrant a diagnosis of dementia may be inappropriate. Future studies should examine relationships between ED-MCI subtypes and NPS severity and progression using measures that may be more appropriate for the population such as the Mild Behavior Checklist (MBI-C)⁵⁴. Lastly, due to its cross-loading on both symptom factors, apathy was analyzed separately, rather than being included in the symptoms factors.

Conclusion

The current study examined the associations among ED-MCI subtypes and NPS factors. The current study extends prior work on MCI subtypes derived in ADNI based on both CC and ED criteria. Utilizing ED criteria has previously identified differences in MCI subtypes in their association with biomarkers of pathological aging and associations with functional impairment over time. In the current study, differences were observed both in baseline NPS and change over time across subtype. This highlights the utility of examining ED-MCI subtypes when studying NPS in older adults with cognitive decline. These differences may partially explain variability across NPS prevalence rates in studies which define MCI in varying ways and often include participants with different patterns of cognitive deficits.

Key Points:

• Empirically-derived mild cognitive impairment (MCI) subtypes have different neuropsychiatric symptom factor trajectories.

• Amnestic MCI participants experienced more severe agitation/hyperactivity symptoms compared to other MCI subtypes.

• Amnestic and dysexecutive MCI participants experienced more severe apathy symptoms over time compared to dysnomic MCI subtypes.