Liew Alzheimer's Research & Therapy (2020) 12:35

[Link]

RESEARCH Open Access

Neuropsychiatric symptoms in cognitively

normal older persons, and the association
with Alzheimer’s and non-Alzheimer’s
dementia
Tau Ming Liew1,2,3

Abstract
Background: Neuropsychiatric symptoms (NPS) have been reported to be useful in predicting incident dementia
among cognitively normal older persons. However, the literature has not been conclusive on the differential utilities
of the various NPS in predicting the subtypes of dementia. This study compared the risks of Alzheimer’s and non-
Alzheimer’s dementia associated with the various NPS, among cognitively normal older persons.
Methods: This cohort study included 12,452 participants from the Alzheimer’s Disease Centers across USA, who
were ≥ 60 years and had normal cognition at baseline. Participants completed the Neuropsychiatric Inventory-
Questionnaire at baseline and were followed up almost annually for incident dementia (median follow-up = 4.7
years). Symptom clusters of NPS—as identified from exploratory and confirmatory factor-analyses—were included
in the Cox regression to investigate their associations with incident dementia.
Results: The various NPS showed independent yet differential associations with incident dementia. Although
psychotic symptoms were rarely endorsed by the participants, they predicted much higher risk of dementia (HR 3.6,
95% CI 2.0–6.4) than affective symptoms (HR 1.5, 95% CI 1.2–1.8) or agitation symptoms (HR 1.6, 95% CI 1.3–2.1).
Psychotic symptoms predicted all dementia subtypes, while affective and agitation symptoms differentially
predicted some subtypes. Across dementia subtypes, psychotic symptoms had relatively higher risk estimates than
affective or agitation symptoms, with the risk estimates being particularly high in non-Alzheimer’s dementia.
Conclusions: Among cognitively normal individuals, the presence of NPS may warrant greater clinical vigilance as
precursors to dementia and its subtypes. The findings highlight the need for further research to enrich our
understanding on the neurobiological links between various NPS and dementia subtypes. They may also change
the clinical approach in managing late-life psychotic symptoms, requiring a greater emphasis on dementia
surveillance in the diagnostic criteria of late-life psychotic disorders.
Keywords: Neuropsychiatric symptoms, Psychotic symptoms, Affective symptoms, Agitation, Dementia, Cox
regression

Correspondence: tau_ming_liew@[Link];
[Link]@[Link]; ephltm@[Link]
1
Department of Geriatric Psychiatry, Institute of Mental Health, 10 Buangkok
View, Singapore 539747, Singapore
2
Department of Psychiatry, Singapore General Hospital, Singapore, Singapore
Full list of author information is available at the end of the article

© The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article are included in the article's Creative Commons
licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons
licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain
permission directly from the copyright holder. To view a copy of this licence, visit [Link]
The Creative Commons Public Domain Dedication waiver ([Link] applies to the
data made available in this article, unless otherwise stated in a credit line to the data.
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 2 of 14

Introduction comparative utilities of the various NPS in cognitively

In extant literature, there has been much evidence to normal older persons—as identified by NPI-Q—in pre-
demonstrate the relevance of neuropsychiatric symptoms dicting the subsequent development of Alzheimer’s and
(NPS) — especially among individuals with mild cogni- non-Alzheimer’s dementia. As a secondary aim, in the
tive impairment [1–5] — in predicting incident demen- subset of participants with complete data on the neuro-
tia. For example, in the population-based Cache County psychological tests, this study also examined the com-
study [3], the presence of NPS, even of mild severity, parative utilities of the various NPS in predicting
was associated with conversion from cognitive impair- subsequent changes in the Z-scores of neuropsycho-
ment no dementia to all-cause dementia. In an analysis logical tests.
of the National Alzheimer’s Coordinating Center
(NACC) dataset [4], the presence of at least one NPS in- Methods
creased the risk of conversion from mild cognitive im- Participants and procedures
pairment to dementia (hazard ratio, HR 1.37; 95% CI This cohort study is based on the NACC database [14],
1.1–1.7), while in a separate analysis of the NACC data- involving participants who were recruited from approxi-
set [5], the risk of conversion from mild cognitive im- mately 39 Alzheimer’s Disease Centers across the USA
pairment to dementia was specifically shown to be between September 2005 and May 2018. The study in-
associated with affective symptoms (HR 1.6, 95% CI 1.4– cluded participants who fulfilled the following criteria:
1.9) and psychotic symptoms (HR 1.6, 95% CI 1.2–2.2). (1) age ≥ 60 years, (2) diagnosed as having normal cogni-
Parallel to the literature on NPS in mild cognitive im- tion at baseline (that is, healthy volunteers who have re-
pairment, there is also growing evidence to highlight ceived clinical evaluations at baseline and found not to
similar relevance of NPS among cognitively normal older have mild cognitive impairment or dementia), (3) no
persons. For example, depression and anxiety have been known diagnosis or treatment of schizophrenia, and (4)
shown in recent meta-analyses to be risk factors of Alz- provided baseline information on NPI-Q. At baseline
heimer’s and vascular dementia [6, 7], while early mani- and on an approximately annual basis, the participants
festations of disinhibition and apathy are integral to the would be accompanied by an informant (who knew the
diagnosis of behavioral variant frontotemporal dementia participants well, usually a spouse, child, or friend) and
[8], and hallucinations and delusions (even in the ab- took part in standardized assessments (which included
sence of cognitive symptoms) are among the key diag- clinical history, physical examination, and detailed
nostic criteria for dementia with Lewy Bodies [9]. These neuropsychological testing) to evaluate for incident de-
growing evidences have led to the recent inclusion of mentia. All contributing Alzheimer’s Disease Centers ob-
NPS in the 2018 NIA-AA (National Institute on Aging- tained informed consent from their participants, as well
Alzheimer’s Association) research framework for Alzhei- as received approval by their local institutional review
mer’s disease [10], as being part of the transition phase boards.
that straddles between the equivalence of normal cogni-
tion and mild cognitive impairment. Measures
Yet, several gaps remain in the literature on NPS in The Neuropsychiatric Inventory-Questionnaire (NPI-Q)
cognitively normal older persons. NPS comprise hetero- [11] was measured at the baseline of this study. NPI-Q is
geneous clusters of symptoms, such as those related to an informant-based questionnaire that screens for the
affective regulation, impulse control, and psychosis [2]. presence of NPS in the past month (of note, the focus on
These various NPS have not been directly compared the past month was emphasized with underscore and
among each other, either in their risks of all-cause de- boldface in the questionnaire). To avoid capturing NPS
mentia or across the different subtypes of dementia. As that are unrelated to preclinical neurocognitive disorders
such, the literature has not been certain on the differen- (such as those related to longstanding psychiatric disor-
tial utilities of the various NPS in predicting dementia ders), the NPI-Q in NACC included the following instruc-
or its respective subtypes. Moreover, even though the tions to the NPI-Q interviewers: “For subjects who are
Neuropsychiatric Inventory-Questionnaire (NPI-Q) [11] cognitively normal or whose cognition has not yet been
is the most widely used screening tool for NPS in older evaluated, please report any behaviors or symptoms that
persons [12, 13], its utility has mostly been established were present within the last month, ignoring behaviors
among those with known cognitive impairment (that is, and symptoms that are usual for the subject and have
in mild cognitive impairment and in the various stages been customary throughout his/her life”. NPI-Q has 12
of dementia) [13]. It is not yet conclusive that NPI-Q items that assess NPS in 12 domains, namely depression,
also has similar utility in cognitively normal individuals, anxiety, apathy, sleep, appetite, agitation, irritability, disin-
especially to identify NPS which can be predictive of hibition, elation, motor disturbance, delusions, and hallu-
subsequent dementia. This study sought to examine the cinations. It was administered to informants by trained
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 3 of 14

healthcare professionals, with each of its items rated on a between numbers and letters in an ascending order.
4-point Likert scale: 0 = not present, 1 = mild (noticeable, Both Trail Making Tests (Part A and Part B) were mea-
but not a significant change), 2 = moderate (significant, sured by the correct lines connected divided by the time
but not a dramatic change), and 3 = severe (very marked to completion, as this computation was previously
or prominent, a dramatic change). To ensure the NPI-Q shown to provide more accurate scores [27]. The Z-
was administered correctly, all interviewers at Alzheimer’s scores for the cognitive tests were computed using the
Research Centers had to undergo a compulsory online age-, sex-, and education-adjusted normative calculator
certification course ([Link] that was published for the NACC participants [27], while
npiq/[Link]) before they were allowed to administer the global Z-score was computed by averaging the Z-
NPI-Q. In the current dataset, the NPI-Q had demon- scores of the cognitive tests within the neuropsycho-
strated adequate internal-consistency reliability, with a logical battery.
Cronbach’s alpha of 0.74 (95% CI 0.72–0.76).
The diagnoses of mild cognitive impairment and de-
mentia were made based on all available information Statistical analyses
from standardized assessments [14], with 69.9% made The factor structure of NPI-Q was first examined to
via consensus conference and the remainder made by understand how the different NPS cluster with each
single clinicians. Mild cognitive impairment was diag- other in the population of cognitively normal older per-
nosed using the modified Petersen criteria [15]. Demen- sons. The clustering of the symptoms (instead of the
tia was diagnosed using either the McKhann (1984) focus on individual symptoms) is useful because most of
criteria [16], DSM-IV (Diagnostic and Statistical Manual the NPS tend to co-occur in clinical practice, and hence,
of Mental Disorders–Fourth Edition) criteria [17], or this approach can provide more meaningful interpreta-
McKhann (2011) criteria [18], with further classification tions of the NPS in clinical practice [12, 13] and has
into the subtypes of Alzheimer’s dementia [16, 18], vas- been the suggested approach by some of the researchers
cular dementia [19], dementia with Lewy Bodies [9, 20, in the field [12, 32]. The clustering of symptoms also
21], frontotemporal lobar degeneration [8, 20, 22–26], helps to avoid the issue of multicollinearity among cor-
and other subtypes. related NPS (highly correlated covariates tend to render
In NACC database, the detailed neuropsychological the results erratic in regression analyses), while at the
battery included cognitive tests [27, 28] which covered same time increases the power of NPS in predicting spe-
the domains of immediate memory (Craft Story 21 Im- cific subtypes of dementia (rarer symptoms, when evalu-
mediate Recall), visuospatial abilities (Benson Complex ated in isolation, tend to have limited power in
Figure Copy), delayed memory (Craft Story 21 Delayed predicting the less common subtypes of dementia; in
Recall and Benson Complex Figure Recall), language contrast, the clustering of NPS increases the number of
(Multilingual Naming Test, and Verbal Fluency–Animal participants who endorse each symptom cluster and
and L-words), attention (Number Span Test Forward hence improves the overall power of NPS).
and Backward), processing speed (Trail Making Test To identify the symptom clusters of NPS, the study
Part A), and executive function (Trail Making Test Part sample was randomly split into two (based on an 80:20
B). These cognitive tests are briefly described here, with split), with 80% of the sample used for exploratory fac-
the full details available in a recent publication [27]. tor analysis (EFA) to identify the factor structure of
Craft Story 21 assesses the ability to provide verbatim re- NPI-Q, while the remaining 20% of the sample reserved
call of a short story immediately after hearing it, and 20 for confirmatory factor analysis (CFA) to validate the
min later [29]. Benson Complex Figure Test assesses the identified factor structure from EFA. This combined
ability to copy a simplified form of the Rey-Osterrieth approach of EFA and CFA was used because the factor
figure and to draw from memory the same figure after structure of NPI-Q has not been known among the
20 min [30]. Multilingual Naming Test assesses the abil- older persons with normal cognition. Although the fac-
ity to name 32 objects which are shown in picture forms tor structure of NPI-Q has been much studied in the
[31]. Verbal Fluency Test measures the number of ani- literature [13], the prior factor structures of NPI-Q
mals and L-words that a participant can name in 1 min. have been derived from patients with known cognitive
Number Span Test assesses the ability to repeat a series impairment (that is, mild cognitive impairment and de-
of numbers (ranging from 2 to 9 digits), both in the for- mentia) and may not be directly applied to the current
ward and backward sequence. Trail Making Test (Part population of older persons with normal cognition—this
A) requires the participants to connect the circles in as- is especially pertinent given the recent suggestion that
cending numerical order (from 1 to 25) as quickly as the factor structure of NPI-Q may vary across different
possible, while Trail Making Test (Part B) requires the cognitive status (such as normal cognition, mild cogni-
participants to connect the circles while alternating tive impairment, or dementia) [13].
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 4 of 14

EFA was conducted with maximum likelihood estima- significance of non-proportionality), the variables that vio-
tion methods and oblique rotation (oblimin), with the lated the proportional-hazard assumption were identified
number of factors in EFA identified using Horn’s parallel using the scaled Schoenfeld residuals and included in the
analysis [33]. An item in NPI-Q is deemed to belong to a Cox regression as stratified variables [40].
particular factor if it has a factor loading of ≥ 0.40 in that Three sensitivity analyses were conducted to evaluate
factor. CFA was conducted in structural equation model- the consistency of the results when some parts of the Cox
ing using maximum likelihood estimation with robust sta- regression were modified. They include the following:
tistics. In CFA, alternative factor structures were also
compared to identify the best-fitting model (the other al- (1) Using the severity scores of each symptom cluster
ternative factor structures are further described in the (instead of the binary variables which were based
footnote of Additional file 3). The model that fulfilled the on the presence or absence of each symptom
criteria of excellent fit (that is, fulfilling all of the following cluster). As described earlier, the items in NPI-Q
four criteria: Root Mean Square Error of Approximation ≤ were scored on 4-point Likert scales. The scores of
0.05, Standardized Root Mean Square Residual ≤ 0.05, the respective items in each symptom cluster could
Comparative Fit Index ≥ 0.95, and Tucker-Lewis Index ≥ be summed to produce the severity score of each
0.95) [34] was used to constitute the symptom clusters of symptom cluster.
NPS in the subsequent analyses. (2) Using inverse probability weighting (IPW) [41] in
The symptom clusters of NPS (at baseline) were then in- the Cox regression to account for participants who
cluded in the Cox proportional-hazard regression—based did not have follow-up data (beyond the baseline
on the full sample—to evaluate their associations with inci- visits). IPW is a well-accepted strategy to minimize
dent dementia. Time to event was defined as the duration potential bias in the results related to differential
from the baseline visit to the clinical diagnosis of demen- risks between those with and without follow-up
tia, as well as to the various subtypes of dementia (namely, data. The probabilities of being “complete cases”
Alzheimer’s dementia, vascular dementia, dementia with (those with follow-up data) were generated from lo-
Lewy Bodies, frontotemporal lobar degeneration, and gistic regression. The inverse of the probabilities
other subtypes of dementia). All the symptom clusters were then used as weights in the Cox regression, so
were concurrently included in the Cox regression (that is, that the results bear more semblance to those who
each symptom cluster was included as a separate variable) dropped out and are less biased towards partici-
to evaluate the independent risks that were attributable to pants who provided follow-up data [41, 42]. Further
each of the symptom clusters. They were included as bin- details on IPW are available in Additional file 1.
ary variables based on whether the participants endorsed (3) Redefining the symptom clusters by items with factor
the presence of each symptom cluster at baseline (that is, a loadings of ≥ 0.20 in the EFA (instead of ≥ 0.40).
symptom cluster would be coded as “yes” when the partici-
pants endorsed at least one of the symptoms within the In addition, in the subset of participants with complete
cluster). The Cox regression adjusted for demographic in- data on neuropsychological tests, a secondary analysis
formation (age, sex, ethnicity, and years of education) as was conducted to examine the comparative utilities of
well as baseline confounders that may potentially predict the symptom clusters of NPS in predicting changes in
both the exposure of interest (NPS) and the outcome of the Z-scores of neuropsychological tests over time. The
interest (dementia) [35], namely, APOE e4 status and use mixed-effect linear regression was used to account for
of antidepressants. APOE e4 status was included as a po- intra-individual correlations in the repeated measure-
tential confounder because it is a known predictor of de- ments of neuropsychological tests. It was conducted as-
mentia (outcome of interest) [36], while at the same time, suming unstructured covariance, the random effects of
there has also been reports of its association with NPS (ex- intercept and time (to allow each participant to vary in
posure of interest) [37]. Similarly, antidepressant has been his neuropsychological scores over time), and the fixed
reported to be associated with dementia (outcome of inter- effects of the other covariates in the model. All the stat-
est) [38, 39], and its use may potentially also modify the istical analyses were conducted in Stata (version 14).
manifestation of NPS (exposure of interest). The primary
analysis was based on participants with follow-up data be- Results
yond the baseline visits (also known as complete case ana- The total sample size was 12,452, with a median age of 72
lysis). The proportional-hazard assumption of Cox (interquartile range (IQR) 67–79; full range 60–104) and a
regression was tested statistically based on whether the median education of 16 years (IQR 14–18; full range 0–
Schoenfeld residuals were associated with time—in the 29). Figure 1 presents the flow diagram related to partici-
event there was significant violation of the proportional- pant selection, while Table 1 shows the participant charac-
hazard assumption (p ≤ 0.05 in the global test on statistical teristics, as well as the comparison between participants
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 5 of 14

Fig. 1 Participant enrolment and exclusion details. MCI, mild cognitive impairment; NACC, National Alzheimer’s Coordinating Center; NC, normal
cognition; NPS, neuropsychiatric symptoms

who did and did not develop dementia. About a quarter of In EFA (based on 80% of the randomly split sample,
the participants (23.0%) only had baseline data and did n = 9962), NPI-Q demonstrated three factors among its
not contribute to the follow-up data, while the rest of the scale items. The scree plot is shown in Additional file 3,
participants had a median follow-up of 4.7 years (IQR while the results on the three factors are presented in
2.4–7.6 years; full range 0.6–12.4 years)—the comparison Table 2. Factor 1 comprises depression, anxiety, and
of characteristics between those with and without follow- apathy and likely represents those with affective
up data is further presented in Additional file 2. At base- symptoms. Factor 2 includes disinhibition, agitation, and
line, 31.6% of the participants reported at least one NPS, irritability and possibly describes NPS related to agita-
of which the most common symptoms were related to de- tion symptoms. Factor 3 includes hallucinations and de-
pression (13.1%), irritability (11.3%), and sleep (10.5%). lusions and represents those with psychotic symptoms.
During follow-up, 724 participants converted to dementia Based on pre-specified factor loading of ≥ 0.40, four
(of which 76.7% were Alzheimer’s dementia, 7.9% vascular items from the original NPI-Q (sleep, appetite, elation,
dementia, 5.4% dementia with Lewy Bodies, 1.9% fronto- and motor disturbance) did not load in any of the three
temporal lobar degeneration, and 8.2% dementia due factors; however, sleep and appetite marginally loaded in
other or unknown etiologies). factor 1 while elation and motor disturbance marginally
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 6 of 14

Table 1 Demographic information of the study participants at baseline (n = 12,452), and comparison between those did and did
not develop dementia during the follow-up period
Variable Overall sample Participants who did Participants who p valuea
(n = 12,452) not develop dementia developed dementia
(n = 11,728) (n = 724)
Age, median (IQR; full range) 72 (67–79; 60–104) 72 (67–78; 60–104) 79 (74–84; 60–99) < 0.001
Male sex, n (%) 4514 (36.3) 4241 (36.2) 273 (37.7) 0.400
Ethnicity, n (%) < 0.001
White 9902 (79.5) 9279 (79.1) 623 (86.0)
African American 1773 (14.2) 1696 (14.5) 77 (10.6)
Others/unknown 777 (6.2) 753 (6.4) 24 (3.3)
Years of education, median (IQR; full range) 16 (14–18; 0–29) 16 (14–18; 0–29) 16 (12–18; 3–29) < 0.001
APOE e4 genotype, n (%) < 0.001
Two copies of e4 allele 242 (1.9) 212 (1.8) 30 (4.1)
One copy of e4 allele 2528 (20.3) 2309 (19.7) 219 (30.2)
No e4 allele/unknown 9682 (77.8) 9207 (78.5) 475 (65.6)
History of psychiatric diagnosis, n (%) 3513 (28.2) 3308 (28.2) 205 (28.3) 0.950
Use of antidepressants, n (%) 2278 (18.3) 2133 (18.2) 145 (20.0) 0.210
Use of anxiolytics, n (%) 1436 (11.5) 1362 (11.6) 74 (10.2) 0.260
Use of antipsychotics, n (%) 86 (0.7) 83 (0.7) 3 (0.4) 0.350
Presence of NPS, n (%)
Any NPS 3941 (31.6) 3635 (31.0) 306 (42.3) < 0.001
Depression 1629 (13.1) 1488 (12.7) 141 (19.5) < 0.001
Anxiety 1107 (8.9) 1027 (8.8) 80 (11.0) 0.035
Apathy 574 (4.6) 506 (4.3) 68 (9.4) < 0.001
Sleep 1306 (10.5) 1212 (10.3) 94 (13.0) 0.024
Appetite 686 (5.5) 601 (5.1) 85 (11.7) < 0.001
Agitation 733 (5.9) 658 (5.6) 75 (10.4) < 0.001
Irritability 1411 (11.3) 1301 (11.1) 110 (15.2) < 0.001
Disinhibition 327 (2.6) 293 (2.5) 34 (4.7) < 0.001
Elation 114 (0.9) 99 (0.8) 15 (2.1) < 0.001
Motor disturbance 160 (1.3) 143 (1.2) 17 (2.3) 0.009
Delusions 97 (0.8) 78 (0.7) 19 (2.6) < 0.001
Hallucinations 41 (0.3) 35 (0.3) 6 (0.8) 0.016
IQR interquartile range, MMSE Mini-Mental State Examination, NPS neuropsychiatric symptoms
a
Test of difference between participants who did and did not developed dementia during the follow-up period: chi-square test for categorical variables, and the
Mann-Whitney U test for continuous variables. Boldfaced p values are ≤ 0.05

loaded in factor 2 (each with factor loading between 0.20 dementia, compared to 5.3% among those without
and 0.40). In CFA (based on 20% of the randomly split affective symptoms; 8.6% among participants with agita-
sample, n = 2490), the three-factor model above (based tion symptoms compared to 5.4% among those without;
on factor loading of ≥ 0.40 in EFA) was the only one that and 18.7% among participants with psychotic symptoms
fulfilled the criteria of excellent fit (Additional file 4). In compared to 5.7% among those without. The risk associ-
contrast, the other models that included the four add- ated with the three symptom clusters is further quanti-
itional items (sleep, appetite, elation, and motor disturb- fied in the Cox regression, with the results presented in
ance) demonstrated inadequate fit. Table 4. The three symptom clusters of NPS were sig-
Participants who endorsed the three symptom clusters nificantly associated with the risk of all-cause dementia.
tended to have a higher proportion of incident dementia However, the risk associated with psychotic symptoms
during the follow-up period. As shown in Table 3, 8.1% (HR 3.6) was comparatively higher than those of
of the participants with affective symptoms developed affective (HR 1.5) or agitation symptoms (HR 1.6). This
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 7 of 14

Table 2 Exploratory factor analysis of the Neuropsychiatric as shown in Table 4. Affective symptoms were associ-
Inventory-Questionnaire and factor loadings of the scale items, ated with the risk of Alzheimer’s dementia, vascular de-
based on 80% of the randomly split sample (n = 9962). Factor mentia, and dementia with Lewy Bodies (HR 1.4–2.8);
loadings of ≥ 0.40 are highlighted in bold, indicating items agitation symptoms were associated with Alzheimer’s
which belong to the respective factor. For the purpose of dementia, frontotemporal lobar degeneration, and other
clarity, only factor loadings of ≥ 0.20 are shown in the table
subtypes of dementia (HR 1.7–4.4), while psychotic
Items in the Neuropsychiatric Factors
Inventory-Questionnaire
symptoms were associated with all the subtypes of de-
1 2 3 mentia (HR 2.2–13.9). Notably, the risk estimates of
1. Depression—Does the patient seem 0.739 psychotic symptoms were relatively higher compared to
sad or say that he/she is depressed?
those of the other two symptom clusters, as well as
2. Anxiety—Does the patient become upset 0.515 higher in non-Alzheimer’s dementia (HR 4.5–13.9) com-
when separated from you? Does he/she
have any other signs of nervousness such pared to Alzheimer’s dementia (HR 2.2). The findings
as shortness of breath, sighing, being remained robust in the three sensitivity analyses, with
unable to relax, or feeling excessively tense? the results presented in Additional files 5, 6, and 7. In
3. Apathy—Does the patient seem less 0.427 particular, the findings remained similar even when the
interested in his/her usual activities or in three symptom clusters were defined by items with fac-
the activities and plans of others?
tor loadings of ≥ 0.20 in the initial EFA (Additional file 7),
4. Sleep—Does the patient awaken you 0.388
during the night, rise too early in the
which included the additional items of sleep and appetite
morning, or take excessive naps during in affective symptoms, as well as the additional items of
the day? elation and motor disturbance in agitation symptoms.
5. Appetite—Has the patient loss or gained 0.237 In the secondary analysis, the symptom clusters of
weight, or had a change in the type of NPS were further compared in their utilities in predict-
food he/she likes?
ing changes in the Z-scores of neuropsychological tests
6. Disinhibition—Does the patient seem to 0.654 over time (Table 5). Overall, the 3 symptom clusters
act impulsively, for example, talking to
strangers as if he/she knows them, or saying were all associated with declines in global Z-scores of
things that may hurt people’s feelings? neuropsychological tests, although psychotic symptoms
7. Agitation—Is the patient resistive to help 0.652 predicted a larger magnitude of decline in the global Z-
from others at times, or hard to handle? score (regression coefficient − 0.25) than affective or agi-
8. Irritability—Is the patient impatient and 0.571 tation symptoms (regression coefficient − 0.08, respect-
cranky? Does he/she have difficulty coping ively). Affective symptoms affected the visuospatial,
with delays or waiting for planned activities?
memory, processing speed, and executive function do-
9. Elation—Does the patient appear to feel 0.368 mains, and largely spared the language and attention do-
too good or act excessively happy?
mains. Agitation symptoms affected the language,
10. Motor disturbance—Does the patient 0.268
engage in repetitive activities such as
attention, and executive function domains, and largely
pacing around the house, handling spared the visuospatial, memory, and processing speed
buttons, wrapping string, or doing domains. Psychotic symptoms mainly affected the pro-
other things repeatedly?
cessing speed and executive function domains, and par-
11. Hallucinations—Does the patient have 0.562 tially affected the memory and language domains.
hallucinations such as false visions or
voices? Does he or she seem to hear
or see things that are not present? Discussion
12. Delusions—Does the patient have false 0.490 Summary of findings
beliefs, such as thinking that others are This study utilized a large sample and a longitudinal
stealing from him/her or planning to study design to examine the comparative utilities of vari-
harm him/her in some way?
ous NPS in predicting dementia and its subtypes. Three
symptom clusters of NPS were identified among cogni-
differential risks across the NPS are also visible in the tively normal older persons, namely psychotic, affective,
Kaplan-Meier curves in Fig. 2. Among the participants and agitation symptoms. Although psychotic symptoms
without NPS, a quarter of them would have developed were rarely endorsed by the participants, those who did
dementia by 12.0 years. This duration shortened to 10.1 report psychotic symptoms had a relatively higher risk of
years in the presence of affective symptoms, 9.1 years in all-cause dementia compared to those with affective or
the presence of agitation symptoms, and 4.1 years in the agitation symptoms. Psychotic symptoms also predicted
presence of psychotic symptoms. all the subtypes of dementia, while affective symptoms
The three symptom clusters were also examined in predicted Alzheimer’s dementia, vascular dementia, and
their associations with the various subtypes of dementia, dementia with Lewy Bodies, and agitation symptoms
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 8 of 14

Table 3 A breakdown of the eventual diagnosis among participants who reported each of the three symptom clusters of
neuropsychiatric symptoms at baseline
Dementia subtype Presence of affective symptomsa Presence of agitation symptomsa Presence of psychotic symptomsa
Yes (n = 2372) No (n = 10,080) Yes (n = 1773) No (n = 10,679) Yes (n = 123) No (n = 12,329)
All-cause dementia 193 (8.1%) 531 (5.3%) 153 (8.6%) 571 (5.4%) 23 (18.7%) 701 (5.7%)
Alzheimer’s dementia 136 (5.7%) 419 (4.2%) 111 (6.3%) 444 (4.2%) 12 (9.8%) 543 (4.4%)
Vascular dementia 15 (0.6%) 42 (0.4%) 9 (0.5%) 48 (0.5%) 2 (1.6%) 55 (0.5%)
Dementia with Lewy Bodies 16 (0.7%) 23 (0.2%) 8 (0.5%) 31 (0.3%) 4 (3.3%) 35 (0.3%)
Frontotemporal lobar degeneration 8 (0.3%) 6 (0.1%) 8 (0.5%) 6 (0.1%) 2 (1.6%) 12 (0.1%)
Other or unknown subtype of dementia 18 (0.8%) 41 (0.4%) 17 (1.0%) 42 (0.4%) 3 (2.4%) 56 (0.5%)
a
Affective symptoms included depression, anxiety, and apathy. Agitation symptoms included disinhibition, agitation, and irritability. Psychotic symptoms included
delusions and hallucinations

predicted Alzheimer’s dementia, frontotemporal lobar clinical features for this subtype) [9]. At the same time,
degeneration, and other subtypes of dementia. Across this study also added new insight to some of the existing
the subtypes of dementia, the risk estimates of psychotic literature. For example, agitation symptoms (such as dis-
symptoms were relatively higher compared to those of inhibition and irritability) were shown to predict Alzhei-
the other two symptom clusters, as well as higher in mer’s dementia in this study, which is a less established
non-Alzheimer’s dementia compared to Alzheimer’s de- finding compared to their association with behavioral
mentia. Compared to affective and agitation symptoms, variant frontotemporal dementia [8]. Similarly, psychotic
psychotic symptoms also predicted a larger magnitude symptoms were shown to predict all the subtypes of de-
of decline in the Z-scores of neuropsychological tests mentia in this study, which is also a less known finding
over time, although the 3 symptom clusters demon- compared to its association with dementia with Lewy
strated different profile of decline in the cognitive do- Bodies [9].
mains—psychotic symptoms mainly affected the The finding on the significant association of psychotic
processing speed and executive function, while affective symptoms, across all the subtypes of dementia, is prob-
symptoms affected the visuospatial, memory, processing ably one of the less expected. Apart from in the context
speed, and executive function, and agitation affected lan- of dementia with Lewy Bodies [9], the association be-
guage, attention, and executive function. tween psychotic symptoms and incident dementia has
not been conclusively demonstrated among cognitively
Interpretation of findings normal older persons [43–45]. This is possibly related to
Some of the findings from this study are consistent with the relatively infrequent occurrence of psychotic symp-
what have been known in the literature. For example, toms in cognitively normal older persons and, conse-
affective symptoms (such as depression and anxiety) quently, the difficulties in capturing them in research
have been shown to predict Alzheimer’s dementia and settings. Prior studies that reported the lack of associ-
vascular dementia in recent meta-analyses [6, 7] and ation [46, 47] often had relatively small number of par-
have also been included as part of the diagnostic criteria ticipants with psychotic symptoms, which may limit
for dementia with Lewy Bodies (as the supportive meaningful interpretation of the results. For example, in

Table 4 Associations between the neuropsychiatric symptoms (affective, agitation, and psychotic symptoms) and incident dementia
Dementia subtype Presence of affective symptomsa Presence of agitation symptomsa Presence of psychotic symptomsa
HR (95% CI) b
p value HR (95% CI) b
p value HR (95% CI)b p value
All-cause dementia 1.5 (1.2–1.8) < 0.001 1.6 (1.3–2.1) < 0.001 3.6 (2.0–6.4) < 0.001
Alzheimer’s dementia 1.4 (1.1–1.7) 0.018 1.7 (1.3–2.2) < 0.001 2.2 (1.1–4.6) 0.027
Vascular dementia 1.9 (1.0–3.5) 0.042 1.2 (0.6–2.5) 0.649 5.7 (1.3–25.6) 0.023
Dementia with Lewy Bodies 2.8 (1.5–5.3) 0.002 0.7 (0.3–1.7) 0.399 13.9 (3.8–50.7) < 0.001
Frontotemporal lobar degeneration 2.7 (0.8–9.4) 0.114 4.4 (1.3–15.0) 0.019 8.7 (2.0–38.7) 0.004
Other or unknown subtypes of dementia 1.3 (0.7–2.6) 0.433 2.2 (1.1–4.3) 0.024 4.5 (1.4–14.4) 0.011
HR hazard ratio, CI confidence interval
a
Affective symptoms included depression, anxiety, and apathy. Agitation symptoms included disinhibition, agitation, and irritability. Psychotic symptoms included
delusions and hallucinations
b
Model adjusted for baseline variables of age, sex, ethnicity, years of education, APOE e4 status, and use of antidepressants. Significant risk estimates (with p ≤
0.05) are highlighted in bold
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 9 of 14

Fig. 2 The Kaplan-Meier curves reflecting the risk of dementia in the presence of a affective symptoms, b agitation symptoms, and c
psychotic symptoms
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 10 of 14

Table 5 Associations between the neuropsychiatric symptoms (affective, agitation, and psychotic symptoms) and the longitudinal
changes in the Z-scores of the neuropsychological tests
Neuropsychological Cognitive Sample Presence of affective symptomsa Presence of agitation symptomsa Presence of psychotic symptomsa
tests (Z-score) domains size
Regression coefficient p value Regression coefficient p value Regression coefficient p value
(95% CI)b (95% CI)b (95% CI)b
Global Z-score Global 5380 − 0.08 (− 0.12, − 0.04) < 0.001 − 0.08 (− 0.12, − 0.03) 0.001 − 0.25 (− 0.44, − 0.06) 0.010
Benson Complex Visuospatial 6304 − 0.08 (− 0.15, – 0.01) 0.030 − 0.07 (− 0.15, 0.01) 0.069 − 0.19 (− 0.67, 0.29) 0.431
Figure Copy
Craft Story 21 Immediate 5520 − 0.14 (− 0.22, − 0.06) < 0.001 − 0.02 (− 0.11, 0.06) 0.608 − 0.33 (− 0.75, 0.09) 0.121
Immediate Recallc memory
Craft Story 21 Delayed 5517 − 0.18 (− 0.26, − 0.10) < 0.001 − 0.05 (− 0.14, 0.04) 0.284 − 0.52 (− 0.86, − 0.18) 0.003
Delayed Recallc memory
Benson Complex Delayed 6289 − 0.10 (− 0.17, − 0.03) 0.006 − 0.10 (− 0.19, − 0.02) 0.017 − 0.27 (− 0.58, 0.04) 0.092
Figure Recall memory
Multilingual Language 5501 − 0.07 (− 0.16, 0.02) 0.106 − 0.10 (− 0.20, − 0.00) 0.041 − 0.73 (− 1.17, − 0.28) 0.001
Naming Test
Verbal Fluency– Language 12,418 − 0.09 (− 0.13, − 0.05) < 0.001 − 0.06 (− 0.11, − 0.02) 0.007 − 0.10 (− 0.28, 0.08) 0.264
Animal
Verbal Fluency– Language 6298 0.00 (− 0.06, 0.07) 0.938 − 0.11 (− 0.18, − 0.03) 0.004 − 0.13 (− 0.42, 0.15) 0.368
L-words
Number Span Attention 5532 − 0.04 (− 0.11, 0.03) 0.253 − 0.10 (− 0.18, − 0.02) 0.014 − 0.20 (− 0.52, 0.12) 0.217
Test Forward
Number Span Attention 5531 − 0.04 (− 0.11, 0.03) 0.270 − 0.13 (− 0.21, − 0.05) 0.001 − 0.16 (− 0.50, 0.18) 0.350
Test Backward
Trail Making Processing 11,327 − 0.11 (− 0.15, – 0.07) < 0.001 − 0.04 (− 0.09, 0.00) 0.061 − 0.30 (− 0.46, − 0.14) < 0.001
Test Part A speed
Trail Making Executive 11,249 − 0.11 (− 0.15, − 0.07) < 0.001 − 0.07 (− 0.12, − 0.03) 0.001 − 0.29 (− 0.45, − 0.14) < 0.001
Test Part B function
CI confidence interval
a
Affective symptoms included depression, anxiety, and apathy. Agitation symptoms included disinhibition, agitation, and irritability. Psychotic symptoms included
delusions and hallucinations
b
Model adjusted for baseline variables of age, sex, ethnicity, years of education, APOE e4 status, and use of antidepressants. Significant estimates (with p ≤ 0.05)
are highlighted in bold
c
For this outcome measure, the “exchangeable covariance” was used in the mixed linear regression because the model could not converge when the
“unstructured covariance” was used

one of the studies of 1408 participants [46], there were cognitively normal older persons. The finding is also not
only 5 participants with delusions and 5 participants inconsistent with a recent study of NPS among individ-
with hallucinations, which resulted in a rather wide con- uals with mild cognitive impairment [32], which simi-
fidence interval of the reported HR (ranging from 0.08 larly highlighted the higher risk estimates of psychotic
up to 6.37). On the other hand, prior studies that re- symptoms, compared to the other NPS, in predicting in-
ported the significant association either did not adjust cident dementia.
for key confounders relevant to dementia [48–51] (such
as educational attainment and APOE e4 status, and Potential implications of the findings
hence may not provide definitive conclusion on the asso- As shown in this study, the various NPS in cognitively
ciation), or recruited community-dwelling individuals normal older persons can be useful in predicting demen-
who possibly have had undiagnosed mild cognitive im- tia and its subtypes, as well as predicting cognitive de-
pairment [52–55] (and hence may not be representative cline over time. They may aid in identifying high-risk
of cognitively normal individuals with psychotic symp- populations, which may then prompt more intensive in-
toms). In contrast to the prior studies, the current study terventions to prevent cognitive decline (such as those
addressed some of the aforementioned limitations in the related to risk factor modification, physical exercise, and
literature (by capturing a slightly larger number of par- cognitive training) [56, 57] as well as the consideration
ticipants with psychotic symptoms, adjusting for key of enrolment into preventive trials for dementia. The
confounders related to dementia, and including only findings also highlight the need for further research to
cognitively normal individuals) and hence may possibly clarify on the neurobiological links of NPS with cogni-
afford a clearer answer on the association between tive impairment, as well as with the different subtypes of
psychotic symptoms and incident dementia among dementia. Prior studies have implicated some of the
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 11 of 14

neurobiological underpinnings of NPS, including the which are more relevant to pre-dementia states. MBI-C
frontal-subcortical circuits and the monoaminergic sys- also focuses on NPS over a 6-month duration (instead of
tem in the brain stem [12]. However, the literature re- the 1-month duration in NPI-Q), with an intent to ex-
mains unclear on whether the various NPS share a clude transient NPS such as those related to bereave-
common neurobiological pathway, or involve distinct ment, changes in living situations, or transient poor
pathways, in the pathogenesis of dementia. The findings sleep [59]. Future research should investigate whether
from this study provided suggestive evidence that the the differences across these two scales may have any im-
various NPS potentially involve distinct brain regions (as pact on their utilities in dementia prognostication, and
shown by the differential profile of cognitive decline), whether one may be better than the other in capturing
which eventually manifest with the different subtypes of NPS among cognitively normal individuals.
dementia. Further research in this area of neurobio- The findings on psychotic symptoms may have impli-
logical links may potentially enrich our understanding cations to the clinical approach in managing psychotic
on the pathogenesis of dementia as well as identify novel symptoms among cognitively normal older persons. To
drug targets to inform the development of disease- date, late-life psychotic disorders remain diagnoses of
modifying drugs for dementia [12]. exclusion, often requiring the exclusion of underlying
It is noteworthy that the NPS can remain useful even neurodegenerative processes [60, 61]. However, the
when they were measured using a screening tool such as current diagnostic criteria of late-life psychotic disorders
the NPI-Q. To date, NPI-Q remains the most commonly [60, 61] do not dictate the extent of clinical investigations
used screening tool for NPS in older persons [12, 13] that are needed to exclude neurodegenerative processes.
and is one of the few NPS tools that most clinicians in This has translated into the prevailing practice where cli-
the field are well-versed in. However, NPI-Q was devel- nicians would not uncommonly exclude dementia based
oped primarily to identify NPS in patients with existing on a cross-sectional assessment (during the initial evalu-
cognitive impairment [13] and included the instruction ations) and using cognitive tests which may be insensi-
to specifically capture NPS that have occurred “since a tive to more subtle cognitive deficits (such as the
patient first began to experience cognitive problems.” Its MMSE). In the presence of apparently normal cognition
utility, as well as sensitivity, in capturing NPS among during the initial evaluations, the clinical suspicion
cognitively normal individuals is not yet certain. As would often shift to primary psychiatric disorders [60],
shown in this study, even a screening tool such as NPI- with the subsequent care focused mostly on relieving
Q can be sufficiently useful to capture clinically relevant psychotic symptoms and less on monitoring for incident
NPS among cognitively normal individuals. It provides a dementia [61]. The findings from this study suggest that
convenient approach to operationalize the routine profil- psychotic symptoms in cognitively normal older persons,
ing of NPS among cognitively normal individuals, which albeit being less common than other NPS, may indicate
is especially pertinent given the recent inclusion of NPS a very high risk of dementia. The apparently normal cog-
as being part of the 2018 NIA-AA research framework nition during the initial evaluations may not be suffi-
for Alzheimer’s disease [10]. Notwithstanding the cient to exclude underlying neurodegenerative processes,
current findings, clinicians who intend to use NPI-Q and there remains a need for continued and close sur-
among cognitively normal individuals may need to adapt veillance to detect the onset of cognitive decline in this
the questionnaire, similar to what was done in NACC, group of individuals. The findings may have implications
to include the following instructions to the NPI-Q inter- to the future diagnostic criteria of late-life psychotic dis-
viewer: “For subjects who are cognitively normal or orders, suggesting the need for greater clarity on the ex-
whose cognition has not yet been evaluated, please re- tent of clinical investigations in the diagnostic criteria—
port any behaviors or symptoms that were present such as specifying the need for repeated administration
within the last month, ignoring behaviors and symptoms of a neuropsychological battery (or in its absence, an-
that are usual for the subject and have been customary other briefer but sensitive cognitive test such as the
throughout his/her life.” Future research should also Montreal Cognitive Assessment) [27, 28, 62, 63]—before
consider further comparative studies between NPI-Q a diagnosis of late-life psychotic disorder can be accur-
and a separate tool, the Mild Behavioral Impairment– ately made. Potentially, the future diagnostic criteria of
Checklist (MBI-C) [58], which was specifically developed late-life psychotic disorders may also incorporate the
to capture NPS in cognitively normal individuals. NPI-Q newer biomarkers of dementia (such as those related to
and MBI-C have several key differences. Compared to amyloid protein, tau protein, and neuronal injury) [10]
MBI-C, NPI-Q has the strength of being much briefer in to definitively exclude underlying neurodegenerative
length (34 items versus 12 items, respectively). However, processes, especially when these biomarkers become
MBI-C may be better tailored to individuals without more accessible to general clinicians in the foreseeable
cognitive impairment, with its descriptions on NPS future. Such approach is not inconsistent with the
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 12 of 14

direction taken by the recently proposed consensus cri- with incident dementia. Psychotic symptoms predicted
teria for psychosis in Alzheimer’s disease [64], where all the subtypes of dementia, while affective and agita-
there is an increasing emphasis that late-life psychotic tion symptoms differentially predicting some subtypes.
disorders may potentially be manifestations of preclinical Psychotic symptoms also had higher risk estimates than
and prodromal Alzheimer’s disease. affective or agitation symptoms, with its risk estimates
being particularly high in non-Alzheimer’s dementia.
Limitations The findings demonstrate that the various NPS in cogni-
Several limitations should be considered. First, the partici- tively normal older persons—even when measured using
pants in the study involved those who volunteered at the a screening tool such as NPI-Q—can be useful in pre-
Alzheimer’s Disease Centers, and may not necessarily rep- dicting dementia and its subtypes, and may potentially
resent those in the community. Additional file 8 provides aid in identifying high-risk populations for preventive
a comparison of the prevalence estimates of NPS between trials and interventions. They highlight the need for fur-
the current sample and a recently published community ther research to clarify the neurobiological links between
sample [46]—although the prevalence estimates of NPS in various NPS and dementia subtypes, which may poten-
the current study may not be dissimilar from those of the tially enrich our understanding on the pathogenesis of
community sample, they are still marginally higher than neurocognitive disorders. They may also change the clin-
those from the community sample and are not inconsist- ical approach in managing late-life psychotic symptoms,
ent with our understanding that clinical samples may have requiring a greater emphasis on dementia surveillance in
higher risk of dementia than community samples (and the diagnostic criteria of late-life psychotic disorders.
hence also have higher prevalence of NPS than commu-
nity samples, given that the presence of NPS can be Abbreviations
CFA: Confirmatory factor analysis; CI: Confidence interval; EFA: Exploratory
viewed as a marker of higher risk). Notwithstanding the factor analysis; HR: Hazard ratio; IQR: Interquartile range; IPW: Inverse
limitation, the findings from this study can be especially probability weighting; NACC: National Alzheimer’s Coordinating Center; NIA-
relevant to patients in current healthcare services, where AA: National Institute on Aging-Alzheimer’s Association; NPI-
Q: Neuropsychiatric Inventory-Questionnaire; NPS: Neuropsychiatric
patients often present voluntarily to health services (not symptoms; USA: United States of America
unlike those who volunteered in this study). Second, this
is the first study in the literature to identify the factor Acknowledgements
structure of NPI-Q among cognitively normal older per- The NACC database is funded by NIA/NIH Grant U01 AG016976. NACC data
sons. Although the findings provide clear evidence of the are contributed by the NIA-funded ADCs: P30 AG019610 (PI Eric Reiman,
MD), P30 AG013846 (PI Neil Kowall, MD), P50 AG008702 (PI Scott Small, MD),
three symptom clusters (no cross-loading between the fac- P50 AG025688 (PI Allan Levey, MD, PhD), P50 AG047266 (PI Todd Golde, MD,
tors in EFA; consistent evidence in CFA; as well as face PhD), P30 AG010133 (PI Andrew Saykin, PsyD), P50 AG005146 (PI Marilyn Al-
validity of the symptom clusters), the factor structure of bert, PhD), P50 AG005134 (PI Bradley Hyman, MD, PhD), P50 AG016574 (PI
Ronald Petersen, MD, PhD), P50 AG005138 (PI Mary Sano, PhD), P30
NPI-Q will still benefit from further validation in other AG008051 (PI Thomas Wisniewski, MD), P30 AG013854 (PI M. Marsel Mesu-
populations of cognitively normal older persons. Third, lam, MD), P30 AG008017 (PI Jeffrey Kaye, MD), P30 AG010161 (PI David Ben-
30.1% of the diagnoses (normal cognition, mild cognitive nett, MD), P50 AG047366 (PI Victor Henderson, MD, MS), P30 AG010129 (PI
Charles DeCarli, MD), P50 AG016573 (PI Frank LaFerla, PhD), P50 AG005131
impairment, or dementia) were made by single clinicians. (PI James Brewer, MD, PhD), P50 AG023501 (PI Bruce Miller, MD), P30
They may not necessarily be as accurate as those made via AG035982 (PI Russell Swerdlow, MD), P30 AG028383 (PI Linda Van Eldik,
consensus conference. Fourth, although this study ex- PhD), P30 AG053760 (PI Henry Paulson, MD, PhD), P30 AG010124 (PI John
Trojanowski, MD, PhD), P50 AG005133 (PI Oscar Lopez, MD), P50 AG005142
cluded individuals who reported prior diagnosis or treat- (PI Helena Chui, MD), P30 AG012300 (PI Roger Rosenberg, MD), P30
ment of schizophrenia, there were still a small number of AG049638 (PI Suzanne Craft, PhD), P50 AG005136 (PI Thomas Grabowski,
participants (0.7%) who had current use of antipsychotics. MD), P50 AG033514 (PI Sanjay Asthana, MD, FRCP), P50 AG005681 (PI John
Morris, MD), and P50 AG047270 (PI Stephen Strittmatter, MD, PhD).
It is plausible, though less likely, that this group of partici-
pants may have under-reported prior history of schizo-
Authors’ contributions
phrenia and may still be using antipsychotics for the The author read and approved the final manuscript.
treatment of schizophrenia. However, there is a more
plausible explanation to the current use of antipsychotics, Funding
where this class of medications is not uncommonly pre- TML was supported by research grants under the National Medical Research
Council of Singapore (grant number NMRC/Fellowship/0030/2016 and
scribed for adjunct or off-label indications, such as for de- NMRC/CSSSP/0014/2017). The funding sources had no involvement in any
pressive symptoms, agitation, or even insomnia. part of the project.

Conclusion Availability of data and materials

This study identified three symptom clusters of NPS The data were obtained from the National Alzheimer’s Coordinating Center
(NACC). For further information on access to the database, please contact
among cognitively normal older persons and demon- NACC (contact details can be found at [Link]
strated their independent yet differential associations researcher_home.html).
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 13 of 14

Ethics approval and consent to participate as risk factors for progression from CIND to dementia: the Cache County
All contributing Alzheimer’s Disease Centers obtained informed consent Study. Am J Geriatr Psychiatry. 2013;21(11):1116–24.
from their participants, as well as received approval by their local institutional 4. Rosenberg PB, Mielke MM, Appleby BS, Oh ES, Geda YE, Lyketsos CG. The
review boards. Research using the NACC database was also approved by the association of neuropsychiatric symptoms in MCI with incident dementia
University of Washington Institutional Review Board. and Alzheimer disease. Am J Geriatric Psychiatry. 2013;21(7):685–95.
5. Liew TM. Symptom clusters of neuropsychiatric symptoms in mild cognitive
Consent for publication impairment and their comparative risks of dementia: a cohort study of 8530
Not applicable. older persons. J Am Med Director Assoc. 2019;20(8):1054.e1051–9.
6. Diniz BS, Butters MA, Albert SM, Dew MA, Reynolds CF 3rd. Late-life
depression and risk of vascular dementia and Alzheimer’s disease:
Competing interests systematic review and meta-analysis of community-based cohort studies. Br
The author declares no competing interests. J Psychiatry. 2013;202(5):329–35.
7. Becker E, Orellana Rios CL, Lahmann C, Rucker G, Bauer J, Boeker M. Anxiety
Author details as a risk factor of Alzheimer’s disease and vascular dementia. Br J Psychiatry.
1
Department of Geriatric Psychiatry, Institute of Mental Health, 10 Buangkok 2018;213(5):654–60.
View, Singapore 539747, Singapore. 2Department of Psychiatry, Singapore 8. Rascovsky K, Hodges JR, Knopman D, Mendez MF, Kramer JH, Neuhaus J,
General Hospital, Singapore, Singapore. 3Saw Swee Hock School of Public van Swieten JC, Seelaar H, Dopper EGP, Onyike CU, et al. Sensitivity of
Health, National University of Singapore, Singapore, Singapore. revised diagnostic criteria for the behavioural variant of frontotemporal
dementia. Brain. 2011;134(Pt 9):2456–77.
Received: 17 December 2019 Accepted: 20 March 2020 9. McKeith IG, Boeve BF, Dickson DW, Halliday G, Taylor JP, Weintraub D,
Aarsland D, Galvin J, Attems J, Ballard CG, et al. Diagnosis and management
of dementia with Lewy bodies: fourth consensus report of the DLB
Consortium. Neurology. 2017;89(1):88–100.
Supplementary information 10. Jack CR Jr, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB,
Supplementary information accompanies this paper at [Link] Holtzman DM, Jagust W, Jessen F, Karlawish J, et al. NIA-AA Research
1186/s13195-020-00604-7. Framework: toward a biological definition of Alzheimer’s disease. Alzheimers
Dement. 2018;14(4):535–62.
Additional file 1. Details on the conduct of inverse probability 11. Kaufer DI, Cummings JL, Ketchel P, Smith V, MacMillan A, Shelley T, Lopez OL,
weighting to account for those who dropped out of the study after the DeKosky ST. Validation of the NPI-Q, a brief clinical form of the
first visit. Neuropsychiatric Inventory. J Neuropsychiatry Clin Neurosci. 2000;12(2):233–9.
Additional file 2. Comparison of demographic information between 12. Geda YE, Schneider LS, Gitlin LN, Miller DS, Smith GS, Bell J, Evans J, Lee M,
participants with and without longitudinal follow-up data. Porsteinsson A, Lanctot KL, et al. Neuropsychiatric symptoms in Alzheimer’s
disease: past progress and anticipation of the future. Alzheimers Dement.
Additional file 3 Scree plot from exploratory factor analysis, based on 2013;9(5):602–8.
80% of the randomly-split sample (n = 9962). 13. Canevelli M, Adali N, Voisin T, Soto ME, Bruno G, Cesari M, Vellas B.
Additional file 4 Fit indices of various models in confirmatory factor Behavioral and psychological subsyndromes in Alzheimer’s disease using
analysis, based on 20% of the randomly-split samples (n = 2490). the Neuropsychiatric Inventory. Int J Geriatr Psychiatry. 2013;28(8):795–803.
Additional file 5. The first sensitivity analysis using the severity scores of 14. Beekly DL, Ramos EM, van Belle G, Deitrich W, Clark AD, Jacka ME, Kukull
the symptom-clusters. WA. The National Alzheimer’s Coordinating Center (NACC) Database: an
Alzheimer disease database. Alzheimer Dis Assoc Disord. 2004;18(4):270–7.
Additional file 6. The second sensitivity analysis to include participants 15. Petersen RC, Morris JC. Mild cognitive impairment as a clinical entity and
without follow-up data using inverse probability weighting. treatment target. Arch Neurol. 2005;62(7):1160–3 discussion 1167.
Additional file 7. The third sensitivity analysis which redefined the 16. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM:
symptom-clusters by NPI-Q items with factor loadings of ≥0.20 in the ex- Clinical diagnosis of Alzheimer’s disease. Report of the NINCDS-ADRDA
ploratory factor analysis. Work Group* under the auspices of Department of Health and Human
Additional file 8. Comparison of the prevalence-estimates of neuro- Services Task Force on Alzheimer’s Disease 1984, 34(7):939–939.
psychiatric symptoms between the current sample and those of a 17. American Psychiatric Association. Diagnostic and Statistical Manual of Mental
previously-published community sample. Disorders: DSM-IV-TR. Washington: American Psychiatric Association; 2000.
18. McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas
CH, Klunk WE, Koroshetz WJ, Manly JJ, Mayeux R, et al. The diagnosis of
Author details dementia due to Alzheimer’s disease: recommendations from the
Supplementary information accompanies this paper at [Link] National Institute on Aging-Alzheimer’s Association workgroups on
1186/s13195-020-00604-7.1Department of Geriatric Psychiatry, Institute of diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;
Mental Health, 10 Buangkok View, Singapore 539747, Singapore. 7(3):263–9.
2
Department of Psychiatry, Singapore General Hospital, Singapore, Singapore. 19. Román GC, Tatemichi TK, Erkinjuntti T, Cummings JL, Masdeu JC, Garcia JH,
3
Saw Swee Hock School of Public Health, National University of Singapore, Amaducci L, Orgogozo J-M, Brun A, Hofman A, et al. Vascular dementia.
Singapore, Singapore. Diagn Criteria Res Stud. 1993;43(2):250.
20. Litvan I, Bhatia KP, Burn DJ, Goetz CG, Lang AE, McKeith I, Quinn N, Sethi
Received: 17 December 2019 Accepted: 20 March 2020 KD, Shults C, Wenning GK. Movement Disorders Society Scientific Issues
Committee report: SIC Task Force appraisal of clinical diagnostic criteria for
Parkinsonian disorders. Mov Disord. 2003;18(5):467–86.
References 21. McKeith IG, Dickson DW, Lowe J, Emre M, O'Brien JT, Feldman H, Cummings
1. Monastero R, Mangialasche F, Camarda C, Ercolani S, Camarda R. A J, Duda JE, Lippa C, Perry EK, et al. Diagnosis and management of dementia
systematic review of neuropsychiatric symptoms in mild cognitive with Lewy bodies: third report of the DLB consortium. Neurology. 2005;
impairment. J Alzheimers. 2009;18(1):11–30. 65(12):1863–72.
2. Ismail Z, Smith EE, Geda Y, Sultzer D, Brodaty H, Smith G, Aguera-Ortiz L, 22. Bensimon G, Ludolph A, Agid Y, Vidailhet M, Payan C, Leigh PN. Riluzole
Sweet R, Miller D, Lyketsos CG, et al. Neuropsychiatric symptoms as early treatment, survival and diagnostic criteria in Parkinson plus disorders: the
manifestations of emergent dementia: provisional diagnostic criteria for NNIPPS study. Brain. 2009;132(Pt 1):156–71.
mild behavioral impairment. Alzheimers Dement. 2016;12(2):195–202. 23. Armstrong MJ, Litvan I, Lang AE, Bak TH, Bhatia KP, Borroni B, Boxer AL,
3. Peters ME, Rosenberg PB, Steinberg M, Norton MC, Welsh-Bohmer KA, Dickson DW, Grossman M, Hallett M, et al. Criteria for the diagnosis of
Hayden KM, Breitner J, Tschanz JT, Lyketsos CG. Neuropsychiatric symptoms corticobasal degeneration. Neurology. 2013;80(5):496–503.
Liew Alzheimer's Research & Therapy (2020) 12:35 Page 14 of 14

24. Brooks BR, Miller RG, Swash M, Munsat TL. El Escorial revisited: revised 47. Donovan NJ, Amariglio RE, Zoller AS, Rudel RK, Gomez-Isla T, Blacker D,
criteria for the diagnosis of amyotrophic lateral sclerosis. Amyotroph Lateral Hyman BT, Locascio JJ, Johnson KA, Sperling RA, et al. Subjective
Scler Other Motor Neuron Disord. 2000;1(5):293–9. cognitive concerns and neuropsychiatric predictors of progression to
25. Neary D, Snowden JS, Gustafson L, Passant U, Stuss D, Black S, Freedman M, the early clinical stages of Alzheimer disease. Am J Geriatr Psychiatry.
Kertesz A, Robert PH, Albert M, et al. Frontotemporal lobar degeneration: a 2014;22(12):1642–51.
consensus on clinical diagnostic criteria. Neurology. 1998;51(6):1546–54. 48. Korner A, Lopez AG, Lauritzen L, Andersen PK, Kessing LV. Delusional
26. Litvan I, Agid Y, Calne D, Campbell G, Dubois B, Duvoisin RC, Goetz CG, disorder in old age and the risk of developing dementia: a nationwide
Golbe LI, Grafman J, Growdon JH, et al. Clinical research criteria for the register-based study. Aging Ment Health. 2008;12(5):625–9.
diagnosis of progressive supranuclear palsy (Steele-Richardson-Olszewski 49. Korner A, Lopez AG, Lauritzen L, Andersen PK, Kessing LV. Acute and
syndrome): report of the NINDS-SPSP international workshop. Neurology. transient psychosis in old age and the subsequent risk of dementia: a
1996;47(1):1–9. nationwide register-based study. Geriatr Gerontol Int. 2009;9(1):62–8.
27. Weintraub S, Besser L, Dodge HH, Teylan M, Ferris S, Goldstein FC, Giordani 50. Korner A, Lopez AG, Lauritzen L, Andersen PK, Kessing LV. Late and very-late
B, Kramer J, Loewenstein D, Marson D, et al. Version 3 of the Alzheimer first-contact schizophrenia and the risk of dementia--a nationwide register
Disease Centers’ neuropsychological test battery in the Uniform Data Set based study. Int J Geriatr Psychiatry. 2009;24(1):61–7.
(UDS). Alzheimer Dis Assoc Disord. 2018;32(1):10–7. 51. Brodaty H, Sachdev P, Koschera A, Monk D, Cullen B. Long-term outcome of
28. Liew TM. Developing a brief neuropsychological battery for early diagnosis late-onset schizophrenia: 5-year follow-up study. Br J Psychiatry. 2003;183:
of cognitive impairment. J Am Med Direct Assoc. 2019;20(8):1054.e1011–20. 213–9.
29. Craft S, Newcomer J, Kanne S, Dagogo-Jack S, Cryer P, Sheline Y, Luby J, 52. Acosta I, Borges G, Aguirre-Hernandez R, Sosa AL, Prince M.
Dagogo-Jack A, Alderson A. Memory improvement following induced Neuropsychiatric symptoms as risk factors of dementia in a Mexican
hyperinsulinemia in Alzheimer’s disease. Neurobiol Aging. 1996;17(1):123–30. population: a 10/66 Dementia Research Group study. Alzheimers Dement.
30. Possin KL, Laluz VR, Alcantar OZ, Miller BL, Kramer JH. Distinct 2018;14(3):271–9.
neuroanatomical substrates and cognitive mechanisms of figure copy 53. Kohler S, Allardyce J, Verhey FR, McKeith IG, Matthews F, Brayne C, Savva
performance in Alzheimer’s disease and behavioral variant frontotemporal GM. Cognitive decline and dementia risk in older adults with psychotic
dementia. Neuropsychologia. 2011;49(1):43–8. symptoms: a prospective cohort study. Am J Geriatr Psychiatry. 2013;21(2):
31. Ivanova I, Salmon DP, Gollan TH. The multilingual naming test in 119–28.
Alzheimer’s disease: clues to the origin of naming impairments. J Int 54. Ostling S, Palsson SP, Skoog I. The incidence of first-onset psychotic
Neuropsychological Soc. 2013;19(3):272–83. symptoms and paranoid ideation in a representative population sample
32. Dietlin S, Soto M, Kiyasova V, Pueyo M, de Mauleon A, Delrieu J, Ousset PJ, followed from age 70-90 years. Relation to mortality and later development
Vellas B. Neuropsychiatric symptoms and risk of progression to Alzheimer’s of dementia. Int J Geriatr Psychiatry. 2007;22(6):520–8.
disease among mild cognitive impairment subjects. J Alzheimers Dis. 2019; 55. Ostling S, Skoog I. Psychotic symptoms and paranoid ideation in a
70(1):25–34. nondemented population-based sample of the very old. Arch Gen
33. Horn JL. A rationale and test for the number of factors in factor analysis. Psychiatry. 2002;59(1):53–9.
Psychometrika. 1965;30:179–85. 56. Livingston G, Sommerlad A, Orgeta V, Costafreda SG, Huntley J, Ames D,
34. Marsh HW, Hau KT, Grayson D: Goodness of fit in structural equation Ballard C, Banerjee S, Burns A, Cohen-Mansfield J, et al. Dementia
models. In: Contemporary Psychometrics: A Festschrift for Roderick P prevention, intervention, and care. Lancet. 2017;390(10113):2673–734.
McDonald. Edited by McDonald RP, Maydeu-Olivares A, McArdle JJ. New 57. Ngandu T, Lehtisalo J, Solomon A, Levalahti E, Ahtiluoto S, Antikainen R,
Jersey: Lawrence Erlbaum Associates; 2005. Backman L, Hanninen T, Jula A, Laatikainen T, et al. A 2 year multidomain
35. Rothman KJ. Modern epidemiology, 3rd ed. edn. Philadelphia: Wolters intervention of diet, exercise, cognitive training, and vascular risk monitoring
Kluwer Health/Lippincott Williams & Wilkins; 2008. versus control to prevent cognitive decline in at-risk elderly people (FINGER): a
36. Baumgart M, Snyder HM, Carrillo MC, Fazio S, Kim H, Johns H. Summary of randomised controlled trial. Lancet. 2015;385(9984):2255–63.
the evidence on modifiable risk factors for cognitive decline and dementia: 58. Ismail Z, Aguera-Ortiz L, Brodaty H, Cieslak A, Cummings J, Fischer CE,
a population-based perspective. Alzheimers Dement. 2015;11(6):718–26. Gauthier S, Geda YE, Herrmann N, Kanji J, et al. The Mild Behavioral
37. Panza F, Seripa D, D'Onofrio G, Frisardi V, Solfrizzi V, Mecocci P, Pilotto A. Impairment Checklist (MBI-C): a rating scale for neuropsychiatric symptoms
Neuropsychiatric symptoms, endophenotypes, and syndromes in late-onset in pre-dementia populations. J Alzheimers Dis. 2017;56(3):929–38.
Alzheimer’s disease: focus on APOE gene. Int J Alzheimers Dis. 2011;2011: 59. Liew TM, Yu J, Mahendran R, Ng TP, Kua EH, Feng L. Neuropsychiatric and
721457. cognitive subtypes among community-dwelling older persons and the
association with DSM-5 mild neurocognitive disorder: latent class analysis. J
38. Chan JYC, Yiu KKL, Kwok TCY, Wong SYS, Tsoi KKF. Depression and
Alzheimers Dis. 2018;62(2):675–686. [Link]
antidepressants as potential risk factors in dementia: a systematic review
60. American Psychiatric Association. Diagnostic and statistical manual of
and meta-analysis of 18 longitudinal studies. J Am Med Direct Assoc. 2019;
mental disorders: DSM-5. Washington, DC: Amer Psychiatric Pub
20(3):279–286.e271.
Incorporated; 2013.
39. Kodesh A, Sandin S, Reichenberg A, Rotstein A, Pedersen NL, Ericsson M,
61. Reinhardt MM, Cohen CI. Late-life psychosis: diagnosis and treatment. Curr
Karlsson IK, Davidson M, Levine SZ. Exposure to antidepressant medication and
Psychiatry Rep. 2015;17(2):1.
the risk of incident dementia. Am J Geriatr Psychiatry. 2019;27(11):1177–88.
62. Liew TM. The Optimal short version of Montreal cognitive assessment in
40. StataCorp. Stata survival analysis reference manual release 14. Texas: Stata
diagnosing mild cognitive impairment and dementia. J Am Med Dir Assoc.
Press; 2015.
2019;20(8):1055.e1–1055.e8. [Link]
41. Seaman SR, White IR. Review of inverse probability weighting for dealing
63. Liew TM. A 4-Item case-finding tool to detect dementia in older persons. J
with missing data. Stat Methods Med Res. 2013;22(3):278–95.
Am Med Dir Assoc. 2019;20(12):1529–34.e6. [Link]
42. Liew TM. Depression, subjective cognitive decline, and the risk of
2019.06.015.
neurocognitive disorders. Alzheimers Res Ther. 2019;11(1):70.
64. Fischer CE, Ismail Z, Youakim JM, Creese B, Kumar S, Nuñez N, Ryan Darby R,
43. Kwak Y, Yang Y, Koo M-S. Late-onset psychosis; is it real? Dement
Di Vita A, D'Antonio F, de Lena C, McGeown WJ, Ramit R, Rasmussen J, Bell
Neurocogn Disord. 2015;14(1):1–11.
J, Wang H, Bruneau MA, Panegyres PK, Lanctôt KL, Agüera-Ortiz L, Lyketsos
44. Lagodka A, Robert P. Is late-onset schizophrenia related to C, Cummings J, Jeste DV, Sano M, Devanand DP, Sweet RA, Ballard C.
neurodegenerative processes? A review of literature. L'Encephale. 2009;35(4): Revisiting criteria for psychosis in Alzheimer's disease and related
386–93. dementias: toward better phenotypic classification and biomarker research.
45. Cohen CI. Very late-onset schizophrenia-like psychosis: positive findings but J Alzheimers Dis. 2020;73(3):1143–56. [Link]
questions remain unanswered. Lancet Psychiatry. 2018;5(7):528–9.
46. Geda YE, Roberts RO, Mielke MM, Knopman DS, Christianson TJ,
Pankratz VS, Boeve BF, Sochor O, Tangalos EG, Petersen RC, et al. Publisher’s Note
Baseline neuropsychiatric symptoms and the risk of incident mild Springer Nature remains neutral with regard to jurisdictional claims in
cognitive impairment: a population-based study. Am J Psychiatry. 2014; published maps and institutional affiliations.
171(5):572–81.