“Cognitive” Criteria in Older Adults With Slow Gait Speed: Implications for Motoric Cognitive Risk Syndrome

Abstract

Background

Motoric cognitive risk syndrome (MCR) is a predementia condition that combines slow gait speed and subjective cognitive concerns (SCC). The SCC criterion is presently unstandardized, possibly limiting risk detection. We sought to (a) characterize SCC practices through MCR literature review; (b) investigate the ability of SCC in slow gait individuals in predicting the likelihood of cognitive impairment in a demographically diverse sample of community-dwelling, nondemented older adults.

Methods

First, we comprehensively reviewed the MCR literature, extracting information regarding SCC measures, items, sources, and cognitive domain. Next, Einstein Aging Study (EAS) participants (N = 278, M_age = 77.22 ± 4.74, %female = 67, M_education = 15 ± 3.61, %non-Hispanic White = 46.3) completed gait, Clinical Dementia Rating Scale (CDR), and SCC assessment at baseline and annual follow-up (M_follow-up = 3.5). Forty-two participants met slow gait criteria at baseline. Generalized linear mixed-effects models examined baseline SCC to predict cognitive impairment on CDR over follow-up.

Results

We reviewed all published MCR studies (N = 106) and documented ambiguity in SCC criteria, with a prevalent approach being use of a single self-reported memory item. In EAS, high SCC endorsement on a comprehensive, validated screen significantly affected the rate of cognitive impairment (CDR; βinteraction = 0.039, p = .018) in slow gait individuals.

Conclusions

An assessment approach that queries across numerous SCC domains was found to predict future decline in clinical dementia status in slow gait older adults. Current SCC practices in MCR, which tend to utilize a single-memory item, may not be the optimal approach. We discuss the implications of SCC criteria validation and standardization to enhance early dementia detection in MCR.

Motoric cognitive risk syndrome (MCR) is a predementia condition initially validated in the early 2010s (1), which combines 2 clinical presentations independently associated with dementia: slow gait speed and self-perceived decline in cognitive functioning. Global prevalence rates of MCR are around 10% (2,3), and the diagnosis has been shown to identify older adults at increased risk for falls (4–6), disability (7–9), cognitive impairment (10), and mortality (11–13), including a number of at-risk older adults who are not identified by other common predementia diagnoses such as mild cognitive impairment (MCI) (2,10). MCR is associated with increased risk for dementia over its individual components of subjective cognitive concerns (SCC) or slow gait alone (10), making it a useful tool to identify individuals at risk for impairment. Since the first publication in 2013 (1), there has been a proliferation of MCR research around the world. The syndrome has international appeal, due to the ease of implementing gait assessment, which is virtually cost free and requires limited training (14–16). Additionally, dementia risk markers such as gait speed are considered more “culture-free” than standard neuropsychological assessment tools, which are often used to detect and track neurocognitive disorders (17,18). As such, MCR has the ability to overcome many barriers to assessing dementia risk in rural, under-resourced, and low-income regions.

The MCR diagnostic criteria were modeled after widely utilized operational standards for MCI, which center on (a) impaired performance on cognitive tests and (b) self-perception of decline in cognitive functioning, in addition to (c) intact daily functioning abilities and (d) absence of dementia (19,20). MCR substitutes gait speed impairment for the objective cognition impairment of MCI, with the standard threshold for “slow” gait being ≥1 standard deviation (SD) below age- and sex-appropriate mean values established within the same cohort (1) (while maintaining the cognitive concerns, intact daily functioning, and absence of dementia criteria from MCI). The slow gait criterion for MCR has been uniformly operationalized within the growing body of MCR studies; however, prior research has highlighted that the SCC criterion of MCR, the second key component of this diagnosis, has been variably implemented (21), possibly resulting in misdiagnosis and weakening the sensitivity of the syndrome to detect risk. Furthermore, the current SCC classification and measurement practices in MCR have not been empirically examined or validated. This lack of attention to SCC in MCR may be a meaningful oversight, as a recent multicenter study found that in older adults with MCR, it is the severity of SCC, rather than the level of gait impairment, that is the strongest predictor of conversion to dementia (even stronger than performance on objective neuropsychological measures) (22).

Beyond MCR, there has been debate within the Alzheimer’s disease and related dementias field pertaining to the operationalization of SCC to classify predementia syndromes, with some arguing that SCC may contribute to misdiagnosis of MCI and should be removed as a part of the classification (23,24,25). Although SCC is frequently the earliest presenting symptom of dementia in the clinical setting (26,27), it is often an unpredictable and deceptive marker of neurodegenerative disease. For example, SCC is ubiquitous in aging populations, with up to 80% of individuals over the age of 70 with normal performance on cognitive testing endorsing some degree of SCC (27). Further, SCC is often attributable to non-neurodegenerative etiologies, including mood, personality, sleep disturbance, chronic pain, and medical conditions (27,28), and, even when SCC is associated with objective cognitive decline and dementia, it becomes increasingly inaccurate with progression of neurodegeneration (29). In essence, the quantification of SCC in aging research, and in MCR specifically, presents a much more complex problem than gait speed measurement, and demands research attention and consensus to optimize detection of at-risk individuals.

The current study focuses on the “cognitive” criteria of MCR. Our objectives were twofold:

• Aim 1. To characterize the current state of SCC classification in the MCR field through a comprehensive analysis of SCC measures, items, cognitive domains, and sources currently in use in all identified published original research studies on MCR.

• Aim 2. To investigate SCC (key criterion 1 of MCR) in individuals with slow gait speed (key criterion 2 of MCR) and their ability to predict cognitive decline during follow-up in a demographically diverse sample of community-dwelling older adults. We expected that the SCC methodology that best predicted decline would serve as a successful model for operationalizing SCC in older adults with slow gait. This aim may inform SCC criterion considerations in future MCR research.

Method

Aim 1

Subjective cognition criteria in MCR literature review procedures

We undertook a comprehensive review of MCR literature to identify all published original research studies, starting with the first MCR publication in April 2013 (1) through April 2023, utilizing the Google Scholar search engine. Additionally, we reviewed the citations list for each article to ensure we captured all published MCR articles. The variables extracted from each article are presented in Supplementary Table 1.

Aim 2

Participants and data collection

We utilized data from the Einstein Aging Study (EAS), a longitudinal study of community-residing older adults in Bronx, NY, a demographically diverse urban setting. Inclusion criteria are age 70 and above, resident of Bronx, NY, noninstitutionalized, and English speaking. Exclusion criteria include the presence of active psychiatric symptomatology and/or perceptual (eg, visual/auditory) impairments that could interfere with assessments and testing and nonambulatory conditions (30). All protocols were approved by the Einstein Institutional Review Board (Einstein IRB 2017-8022 and 2021-13702) and participants provided written informed consent. In-person assessments were conducted annually and included comprehensive neurological, medical, psychosocial, neuropsychological, and gait examinations. All participants were discussed at a case conference attended by the study neurologist, neuropsychologist, and nurse practitioner to confirm that an individual was free of dementia at baseline and to determine dementia status at follow-up.

For the present study, we included 278 EAS participants who were dementia free and completed gait speed assessment at baseline, with the primary analyses focused on those with slow gait (N = 42). With an initial sample of 322 nondemented participants at baseline (dementia, n = 2), 44 participants were removed due to not completing gait assessment. Participants who were or were not included did not significantly differ on any demographic variable, aside from age (included mean age = 77.2; excluded mean age = 79.8; significant at p < .001; Age was controlled for in all analyses). Analyses were based on baseline gait speed, SCC assessment, and neuropsychological testing, and the longitudinal administration of neuropsychological measures completed during the annual in-person study visit. Among the 278 participants, the average number of assessments, including baseline and yearly follow-up, was 3.3 (median = 4, range 1–7). This particular EAS cohort study was carried out from 2017 through 2022.

Gait speed and motoric cognitive risk syndrome

Investigating SCC within the context of slow gait speed was the focus of this study. As such, we classified individuals as slow gait speed if they met all of the diagnostic criteria for MCR at baseline, setting aside the SCC criterion, which is the independent variable/predictor of key interest. The diagnosis of MCR is based on established criteria (1,10) and includes (1) slow gait, defined as ≥1 SD below age- and sex-appropriate mean values established within the same cohort, including participants with neurological and non-neurological gait abnormalities. To capture gait speed (in centimeters per second), a 4-m walk was measured at usual pace in a quiet, well-lit room; (2) preserved instrumental activities of daily living (IADLs), as measured on the Lawton Brody scale, indicating independence on instrumental activities items (31); (3) absence of dementia diagnosis based on consensus case conference; and (4) the SCC criteria in MCR, that is, the presence of “cognitive concerns,” has been variably applied across studies without clearly defined standardization (see Aim 1 of this study). Given that SCC is the focus of this research, we included any participant who met other MCR criteria (eg, criteria listed above as 1, 2, and 3) and completed SCC assessment. Because these individuals do not meet the full traditional diagnostic criteria of MCR, for the purposes of this study, we call these participants older adults with slow gait speed. All other participants were classified as normal gait speed.

Subjective cognitive concerns assessment

An overview of the SCC assessment is shown in Table 1. We assessed SCC with an expanded 40-item version of the Cognitive Change Index (CCI) (32,33) self-report form, a widely used tool to evaluate perceived decline in various cognitive domains. The CCI-40 includes 40 self-report items, wherein participants are asked to rate their current level of ability (approximately over the past month) compared to 5 years ago on a 5-point Likert scale: 1 = normal ability (no change or better than 5 years ago), 2 = slight/occasional problem (minimal change from 5 years ago), 3 = mild problem (some change from 5 years ago), 4 = moderate problem (clearly noticeable change from 5 years ago), and 5 = severe problem (much worse than 5 years ago). Total scores on the CCI-40 range from 40 to 200, with higher scores indicating more significant self-perceived cognitive difficulties. We included the CCI-40 total score, as well as item-by-item CCI scores, as a continuous predictor in all relevant analyses. We also assessed SCC with a commonly used approach in the MCR field (21), the memory item from the Geriatric Depression Scale (GDS): “Do you feel you have more problems with your memory than most?” (34). Participants respond with a binary yes or no to this item. We included the GDS memory item as a binary predictor in all relevant analyses.

Table 1.

SCC Assessment Cognitive Domains, Number of Items, and Sample Items

Cognitive Domain	Item Count	Sample Items
Cognitive Change Index (CCI-40)*
Memory	21	“Remembering things compared to my age group” “Remembering to pay a bill on time”
Executive functioning	6	“Organizing my daily activities” “Shifting easily from one activity to the next”
Language	4	“Understanding conversations” “Spelling familiar words”
Attention/concentration	3	“Focusing on more than one thing at once” “Concentrating on what I am doing”
Visuospatial/navigation	2	“Navigating to locations” “Finding my way around familiar places”
Mental clarity/efficiency	2	“Thinking quickly” “Thinking clearly”
Calculation	1	“Solving everyday math problems”
Orientation	1	“Recalling the date without looking it up”
Geriatric Depression Scale (GDS)†
Memory	1	“Do you feel that you have more problems with memory than most?

Notes: The CCI response options are on a 5-point Likert Scale. The GDS response options are yes/no. SCC = subjective cognitive concerns.

Neuropsychological assessment

Participants completed a comprehensive neuropsychological evaluation at baseline and follow-up. Testing included the Clinical Dementia Rating Scale (CDR) (35), which is a global rating scale for the study of individuals across the dementia spectrum. For our analyses, because of the extremely rare cases with CDR ≥ 1 during follow-up, we generated dichotomized CDR scores representing cognitively impaired and noncognitively impaired CDR performances, with CDR scores of 0 being classified as noncognitively impaired and CDR scores ≥0.5 classified as cognitively impaired. A similar approach to CDR score dichotomization has been used elsewhere (eg, see (36)). Comprehensive neuropsychological measures which were administered are shown in Table 2.

Table 2.

Baseline Demographic Characteristics, Neuropsychological Performance, and SCC in the Overall Sample, MCR, and Non-MCR

	Overall Sample		Slow Gait		Normal Gait		p
	N = 278		n = 42		n = 236
	M	SD	M	SD	M	SD
Demographic characteristics
Age	77.217	4.747	78.010	5.576	77.075	4.583	ns
Education years	15.0	3.610	15.210	3.418	14.960	3.648	ns
% female	67.0		65.1		67.3		ns
% White	46.3		41.9		47.0		ns
Depression (adj-GDS)	2.122	1.867	2.762	2.418	2.009	1.733	.016
Gait speed (cm/s)	98.401	27.848	58.289	11.298	105.540	23.519	<.001
SCC
CCI-40 Total	63.360	18.147	66.119	22.755	62.869	17.212	ns
% GDS memory item	13.3		18.6		12.5		ns
Neuropsychological measures
MoCA	23.470	3.538	23.070	3.612	23.540	3.528	ns
DSym	34.600	10.274	30.610	9.415	35.310	10.277	.007
NSF	7.380	2.176	7.360	2.082	7.390	2.197	ns
NSB	5.800	2.178	5.450	1.990	5.860	2.208	ns
TMT A	48.490	21.309	56.480	29.410	47.050	19.230	.008
TMT B	124.340	58.032	141.470	54.227	121.520	58.276	ns
FSL	37.280	12.705	35.860	11.108	37.540	12.973	ns
CAT	27.990	7.403	27.240	7.241	28.120	7.438	ns
MINT	25.610	4.324	26.310	3.659	25.480	4.427	ns
CS 1	14.550	4.157	14.480	3.611	14.570	4.254	ns
CS 2	12.960	4.277	12.620	4.114	13.030	4.311	ns
Benson figure copy	14.830	1.448	14.710	1.582	14.860	1.425	ns
Block design	23.630	9.267	21.260	7.044	24.060	9.559	ns

Notes: adj-GDS = reflects the total Geriatric Depression Scale short form score with the GDS memory item removed; Benson figure copy (25); Block Design (37); CAT = category fluency for animals and vegetables (38); CCI-40 = The Cognitive Change Index 40-Item Version; CS 1 = Craft Story 1; CS 2 = Craft Story 2 (39); Dsym = Digit Symptom Modalities (40); FSL = Phonemic Verbal Fluency for letters F, S, and L (41); MCR = motoric cognitive risk syndrome; MINT = Multilingual Naming Test (42); MoCA = Montreal Cognitive Assessment (43); ns = not significant; non-MCR = non-motoric cognitive risk syndrome; NSB = Number Span Backward (37); NSF = Number Span Forward; SCC = subjective cognitive concerns; TMT A = Trail Making Test A; TMT B = Trail Making Test B (44).

Covariates

Demographic information included self-reported race/ethnicity as defined by the U.S. Census Bureau, years of education, sex, and age. An adjusted version of the Geriatric Depression Scale (GDS, short form) total score was used to screen for depression. The GDS contains dichotomous (yes/no) items and when summed, the total score ranges from 0 to 15, with scores of 5 or above suggestive of clinically significant depression (34). For this study, we excluded the GDS memory item from the total GDS, for an adjusted total score (range 0–14, GDS-adj).

Analyses

For Aim 1, descriptive statistics were employed to characterize the current state of SCC criteria in the MCR field identified through literature review.

For Aim 2, baseline descriptive characteristics were reported in the overall sample and between groups: slow gait speed and normal gait speed. Independent samples t-test and Chi-square tests were employed as appropriate to test for significant group differences at baseline on demographic variables, subjective cognition, and neuropsychological assessment. We used linear mixed-effects models (LME) for continuous outcomes (Number Span Forward, Trail Making Test B, Craft Story 1, and Craft Story 2), and generalized LME for binary longitudinal cognitive outcomes (dichotomized/binary variable: CDR = 0 representing noncognitively impaired, CDR ≥ 0.5 representing cognitively impaired), to examine how CCI total score, CCI item-by-item scores, and GDS memory item at baseline predict to the rate of change in risk of cognitive impairment in slow gait and normal gait individuals over follow-up. All observed data for the longitudinal outcomes were included in the analyses. One advantage of LME and generalized LME models is that they can handle missing data during follow-up under the assumption that data are missing at random, which allows the missing data process to depend on the observed data. CCI total score and/or item/GDS memory item, time, and their interaction terms were included in the model as independent variables, respectively, also adjusting for covariates including age, sex, education, race/ethnicity, and adjusted GDS score. The interaction term, which estimates the effect of CCI scores or GDS memory item on rate of change over follow-up, was reported. Separate models for each outcome of interest were used. An α-value of 0.05 was used for determining statistical significance for all hypothesis-driven analyses. SPSS version 25 software (IBM Corp., Armonk, NY) and SAS 9.4 (SAS Institute Inc., Cary, NC) were utilized.

Results

Aim 1: Review of Current SCC Practices in MCR

We identified 106 original research articles published between April 2013 (first MCR publication) and October 2023, which, to our knowledge, encompasses all published manuscripts on MCR. These articles included samples from 26 countries spanning 6 continents (Figure 1). The number of studies published each year since 2013 is shown in Supplementary Figure 1 and demonstrates the increasing research interest and rate of publication in the field. Supplementary Table 1 shows a comprehensive presentation of the SCC criterion and relevant variables from each identified MCR study. Figure 2 provides a brief summary of the SCC practices in terms of (A) measures, (B) items, (C) source, and (D) cognitive domain. (A) “Measures” refers to a psychometrically validated screening instrument of SCC. A measure is comprised of SCC items. (B) “Items” refers to the individual question or prompt related to SCC that may be freestanding or included within an SCC measure. (C) “Source” refers to the individual who is reporting on the participant’s cognitive status. Source may refer to self, a knowledgeable informant (eg, family member or friend), or a clinician’s subjective judgment. (D) “Cognitive domain” refers to the domain of cognition that is queried by an SCC measure or item (eg, memory, attention, language, orientation).

Figure 1.

Country of MCR studies (N = 106). MCR = motoric cognitive risk syndrome. Studies from multiple countries included additional populations from Australia, Belgium, Brazil, Ghana, Ireland, Russia, and South Africa.

Figure 2.

Summary of SCC practices in MCR (studies N = 106) in Terms of (A) Measures, (B) Items, (C) Source, and (D) Cognitive Domain. SCC = subjective cognitive concern. (A) “Measures” refers to a psychometrically validated screening instrument of SCC. A measure is comprised of SCC items. (B) “Items” refers to the individual question or prompt related to SCC that may be freestanding or included within an SCC measure. (C) “Source” refers to the individual who is reporting on the participant’s cognitive status. Source may refer to self, a knowledgeable informant (eg, family member or friend), or a clinician’s judgment. (D) “Cognitive domain” refers to the domain of cognition that is queried by an SCC measure or item (eg, memory, attention, language, orientation).

SCC measures

Almost 40% of the studies did not specify the measure(s) utilized to quantify SCC or no formal measure was utilized (eg, 4 studies classified SCC as a concern about cognition being the reason for study referral, meaning that the entire study sample met the SCC criteria; Figure 2A). Ten of these studies (9.4%) did not appear to use a face valid or psychometrically sound measure/item(s). In terms of number of distinct SCC questionnaires used within a single study to characterize SCC, 39.6% of studies used one questionnaire (n = 42), with 12.3% of studies using 2 (n = 13), 7.5% using 3 (n = 8), and 1.9% using 4 questionnaires (n = 2). Fifteen questionnaires to characterize SCC were identified, including the GDS (n = 30), AD-8 Dementia Screening Interview (n = 12), Health Self-Assessment (n = 6), Short Portable Mental Status Questionnaire (SPMSQ; n = 4), Consortium to Establish a Registry for Alzheimer’s Disease (CERAD; n = 3), CDR (n = 3), Cambridge Mental Disorders of the Elderly Examination (CAMDEX; n = 2), the Subjective Memory Form (n = 2), Mini-Mental State Examination (MMSE; n = 1), The Measurement of Everyday Cognition (E-Cog; n = 1), IADL (n = 1), NeuroTrax (n = 1), the Subjective Cognitive and Motoric Complaint Questionnaire (n = 1), the Functional Autonomy Measurement System (SMAF; n = 1), and the Korean Dementia Screening Questionnaire-P (KDSQ-P; n = 1). More than 26% of studies solely relied on the memory item from the GDS (“Do you feel that you have more problems with memory than most”) to classify SCC, with an additional 22.6% using the memory GDS item in combination with other SCC items or measures (cum% = 49). Notably, the CDR, MMSE, SPMSQ, and NeuroTrax are not measures of SCC; rather, they are measures of objective cognition (MMSE, SPMSQ, and NeuroTrax) or clinical dementia staging (CDR, which includes both subjective and objective cognition components), so it is unclear why and how these measures were used to operationalize SCC.

SCC items

The number of SCC items included in each study ranged from 1 (45 studies) to 15 items (1 study), with the most common approach (43.8%) being to include only 1 SCC item (mode = 1; Figure 2B). For 23.6% of the studies, there was either an unclear or unreported number of SCC items included, a nonpsychometric/nonstandardized approach was utilized (ie, clinician observation of SCC), or SCC was defined as the reason for study referral. Two studies appeared to establish an item cut score to classify SCC (eg, using a combined SCC and “subjective motoric complaints” cut score to classify MCR, see (45) and (46)); instead, the prevailing approach was to include any participant who positively endorsed SCC on any given item.

SCC source

The vast majority of studies utilized self-report to classify SCC (n = 78, 73.6%; Figure 2C). Seven studies utilized different iterations of SCC sources to classify SCC, including report from either (1) self and/or study informant (n = 2), (2) study informant alone (n = 1), (3) self and/or study clinician (n = 2), or self, informant, or study clinician (n = 2). For the remaining studies (19.8%), the source of SCC report was unclear or unspecified.

SCC domain

The majority of studies (n = 63, 57.4%) utilized SCC items that related to memory alone (Figure 2D). An additional 17 studies (16%) assessed SCC across multiple cognitive (eg, memory, processing speed, executive functioning) or adaptive functioning domains (eg, cognitive problems resulting in difficulty in daily activities). For a large proportion of MCR studies (n = 26, 23.6%), the cognitive domain assessed was unclear, unspecified, or unreported.

Aim 2

Overview

Participants’ (N = 278) age at baseline ranged from 70.4 to 93.7 (mean = 77.217, SD = 4.74) years, the sample was 67% female, and educational achievement averaged 15 ± 3.61 years. The sample was 46.3% non-Hispanic White, 39.81% non-Hispanic Black, with the remaining participants combined into an “other” ethnic/racial group (13.89%) due to small sample sizes in these populations, including 31 Hispanic White (9.6%), 9 Hispanic Black (2.8%), 4 Asian (1.2%), and 1 (0.3%) “more than 1 race” participant. At baseline, a total of 42 participants (15.1%) met criteria for slow gait speed, with 236 classified as normal gait.

Baseline characteristics of the sample, including age, education, race/ethnicity, gait speed, depression, scores on objective cognitive measures, and SCC (total CCI and GDS memory item) are presented in Table 2 in the overall sample, and stratified by gait speed status. Slow and normal gait participants did not differ by age, education, sex, or race/ethnicity. Individuals with slow gait showed more symptoms of depression (t(276) = 2.431, p = .016) than normal gait participants. Objective cognitive performance showed worse scores in the slow gait group on a measures of processing speed and attention (DSym, t(269) = 2.730, p = .007; TMT A, t(273) = 2.669, p = .008). Slow and normal gait participants did not differ significantly on cognitive performances on a cognitive screen (MoCA), measures of working memory (DSB), executive functioning (TMT B, FSL), language (CAT, MINT), and memory (CS).

Baseline subjective cognition in overall sample, and by gait speed status

The slow and normal gait groups did not differ on overall subjective concern reporting (CCI) or by endorsement of memory problems on the GDS item (Table 2). In the overall sample (n = 278), only 5 participants (1.54%) did not endorse any SCC item on the CCI. Only 43 individuals in the sample (13.3%) positively endorsed the GDS memory item. An item-by-item analysis (Supplementary Table 2) revealed 4 individual CCI items that significantly distinguished between slow and normal gait status, including items 2, 12, 37, and 39.

SCC and decline in clinical dementia status and objective cognition

Among the 278 participants, the average numbers of assessments at baseline and over follow-up was 3.3 (median = 4, range 1–7). Table 3 shows the CCI total score and CCI items which are significantly associated with decline on CDR and neuropsychological outcomes in individuals with slow gait speed (n = 42). We present data for NSF, TMT B, CS, and CDR, which showed statistically significant association between baseline CCI and rate of decline over follow-up (Table 3). In individuals with slow gait, total SCC endorsement significantly affected rate of change in log odds on a global dementia scale CDR (βinteraction = 0.039, p = .018). Total SCC endorsement did not affect rate of change in log odds on a global dementia scale in the normal gait group. Total SCC endorsement was not associated with rate of decline on any cognitive measures in the slow gait group.

Table 3.

Baseline Reporting of Overall Subjective Cognitive Concerns (CCI-40 total score), Individual Subjective Cognitive Concerns (CCI items), and GDS Memory Item and Associations with Rate of Decline on Neuropsychological Outcomes and Cognitive Status in Participants With Slow Gait (n = 42)

	Number Span Forward			Trail Making Test B			Craft Story 1			Craft Story 2			Clinical Dementia Rating
	est	stderr	p	est	stderr	p	est	stderr	p	est	stderr	p	est	stderr	p
Overall reporting of subjective cognitive concerns (SCC)
CCI-40	0.003	0.004	0.361	0.106	0.102	0.304	0.001	0.007	0.930	0.009	0.007	0.220	0.038	0.016	0.018*
Individual SCC items
CCI1	−0.189	0.171	0.276	−3.776	4.130	0.366	−0.399	0.309	0.204	−0.128	0.320	0.690	0.617	0.271	0.025*
CCI2	0.214	0.157	0.179	−1.247	3.948	0.754	0.047	0.289	0.870	0.283	0.293	0.340	0.724	0.298	0.017*
CCI5	0.053	0.099	0.594	0.800	2.385	0.739	0.070	0.178	0.694	0.232	0.178	0.199	0.972	0.461	0.038*
CCI12	0.214	0.157	0.179	−1.247	3.948	0.754	0.047	0.289	0.870	0.283	0.293	0.340	0.724	0.298	0.017*
CCI13	−0.067	0.300	0.822	−2.503	7.284	0.733	0.002	0.536	0.995	0.620	0.538	0.256	1.370	0.599	0.025*
CCI15	0.096	0.105	0.363	1.542	2.532	0.546	−0.070	0.190	0.712	0.204	0.191	0.291	0.486	0.241	0.047*
CCI16	0.029	0.100	0.771	1.570	2.438	0.523	−0.091	0.179	0.611	0.189	0.180	0.300	0.587	0.249	0.021*
CCI21	0.102	0.146	0.485	4.232	3.472	0.231	0.129	0.261	0.622	0.534	0.256	0.043	0.818	0.397	0.043*
CCI22	0.074	0.114	0.521	5.968	2.702	0.033*	−0.066	0.207	0.750	0.127	0.211	0.549	0.368	0.235	0.123
CCI23	0.173	0.451	0.703	23.152	10.213	0.029*	0.442	0.806	0.586	1.443	0.798	0.078	.	.	.
CCI25	−0.083	0.162	0.608	12.303	3.487	0.001*	−0.040	0.292	0.891	0.206	0.298	0.492	0.018	0.276	0.946
CCI27	0.031	0.123	0.797	6.322	2.780	0.029*	0.041	0.221	0.855	0.321	0.220	0.153	0.885	0.538	0.104
CCI28	0.161	0.108	0.144	5.171	2.504	0.046*	0.050	0.199	0.800	0.259	0.200	0.202	−0.036	0.238	0.879
CCI29	0.188	0.099	0.065	3.979	2.364	0.101	0.230	0.181	0.210	0.394	0.178	0.033	0.636	0.240	0.009*
CCI32	−1.323	0.639	0.045*	48.388	33.363	0.155	−2.860	1.062	0.010*	−2.315	1.131	0.047*	.	.	.
CCI33	0.131	0.110	0.243	5.099	2.526	0.051	0.044	0.201	0.825	0.257	0.202	0.211	0.810	0.341	0.020*
CCI36	−0.146	0.154	0.349	−4.948	4.474	0.276	−0.398	0.273	0.152	−0.190	0.282	0.503	0.872	0.335	0.011*
CCI37	−0.074	0.116	0.529	4.658	2.741	0.098	−0.047	0.207	0.819	0.172	0.212	0.422	1.390	0.518	0.009*
CCI-40	0.199	0.108	0.073	5.608	2.482	0.030*	0.205	0.199	0.310	0.375	0.198	0.066	0.300	0.275	0.279
The Geriatric Depression Screen Memory Item
GDS	−0.020	0.270	0.939	10.600	6.853	0.130	0.312	0.479	0.519	0.568	0.483	0.247	0.711	0.442	0.111

Notes: *Indicates a statistically significant result. CCI = Cognitive Change Index; GDS = Geriatric Depression Scale (memory items); SCC = subjective cognitive concerns. This table shows the longitudinal association between baseline SCC and rate of decline in slow gait participants (n = 42). Empty cells are due to nonconvergent models.

In slow gait participants, high endorsement on 19 individual items showed significant higher rate of decline in clinical dementia status and cognitive functioning. Supplementary Table 3 shows the text and cognitive domain of each CCI item associated with rate of decline in slow gait participants. Items 1 (βinteraction = 0.617, p = .026), 2 (βinteraction = 0.725, p = .018), 5 (βinteraction = 0.972, p = .039), 12 (βinteraction = 0.724, p = .018), 13 (βinteraction = 1.371, p = .025), 15 (βinteraction = 0.486, p = .049), 16 (βinteraction = 0.587, p = .021), 21 (βinteraction = 0.818, p = .043), 29 (βinteraction = 0.636, p = .009), 33 (βinteraction = 0.812, p = .020), 36 (βinteraction = 0.872, p = .011), and 37 (βinteraction = 1.391, p = .009) had significant impact on rate of change in log odds on a global dementia scale. Seven items showed significant effect in rate of decline on measures of executive functioning (TMT B) and attention (DSF). For TMT B, these items included item 22 (βinteraction = 5.968, p = .034), 23 (βinteraction = 23.152, p = .029), 25 (βinteraction = 12.304, p = .001), 27 (βinteraction = 6.322, p = .029), 28 (βinteraction = 5.171, p = .046), and 40 (βinteraction = 5.608, p = .030). For DSF, the item 32 (βinteraction = -1.32, p = .045) had a significant impact in rate of decline over follow-up. Item 32 was significantly associated with rate of decline in learning (CS1; βinteraction = -2.860, p = .010) and memory (CS2; βinteraction = -2.315, p = .047). The GDS memory item was not significantly associated with rate of decline in log odds on the CDR nor rate of decline on any neuropsychological test.

Although not the focus of the present study, results for participants with normal gait are presented in Supplementary Table 4 and described briefly here. As stated above, total SCC endorsement on the CCI-40 did not affect rate of change in log odds on a global dementia scale in the normal gait group. The GDS memory item was also not associated with rate of decline. Twelve individual CCI items were found to predict decline on objective tests of attention, set-shifting, and learning/memory. These SCC items were within the cognitive domains of memory (Items 3, 4, 7, 21, 25, and 32), language (Item 22), attention (Item 39), mental efficiency (items 37,38), executive functioning (Item 13), and mental calculation (Item 36). Seven of these sensitive items overlapped with items predictive of decline in the MCR group, including Items 13, 21, 22, 25, 32, 36, and 37.

Discussion

There has been a proliferation of research on the predementia condition MCR over the last decade, with over 100 published MCR studies to date. First characterized in 2013 by Verghese and colleagues (1) in the EAS cohort in New York City, original research on MCR has emanated from more than 26 different countries spanning 6 continents. Across the growing MCR literature, there is clear operationalization of the slow gait criterion of MCR. However, there is significant variability, lack of clarity, and conflicting application of the SCC criterion. As the MCR field grows, it is essential to carefully examine the SCC criteria, as the lack of operationalization of one of the 2 key MCR criteria may limit the ability to uniformly assign diagnosis and draw conclusions across MCR studies. More broadly, the confusion surrounding SCC likely limits risk detection of the MCR construct. In response to these concerns, we characterized the current SCC practices in MCR within the entire field to clearly highlight the need for attention to this criterion (Aim 1); and investigated, within a single aging study, an SCC assessment approach that was optimized for predicting future decline in slow gait older adults (Aim 2). Overall, our findings demonstrated that a comprehensive assessment across numerous SCC domains was effective at identifying risk for cognitive impairment in older adults with slow gait speed.

An all-inclusive review of the MCR literature (Aim 1) revealed striking lack of clarity and variability in how the SCC criteria has been operationalized (Supplementary Table 1). For example, almost half of all studies did not clearly report the name of the SCC questionnaire utilized, a quarter did not clarify how many SCC items were considered in the criterion or used a nonpsychometric approach (eg, SCC was reason for study inclusion, clinician observation of SCC), and another quarter did not indicate the cognitive domain (eg, memory, executive functioning) of the SCC being assessed. In some studies, SCC was the reason for study referral (hence, all participants in a cohort met the SCC criteria), virtually making the SCC criteria irrelevant. In other studies, objective cognitive performance, not subjective cognition, was assessed, which is a clear deviation from the intended SCC criteria and the spirit of the MCR diagnosis, which intentionally does not include objective cognition. We speculate that the disparate approaches to SCC assessment result at least in part from the retrospective nature of the young MCR literature, which draws upon existent databases and/or established longitudinal cohort study protocols, thus limiting the SCC items available to assign MCR diagnosis. Furthermore, multicenter MCR studies, which integrate data across distinct aging studies, likely suffer from restriction of available SCC items that are common across sites. Unfortunately, the tremendous variability of the SCC criteria may ultimately undermine the MCR diagnosis, as it suggests that the intended construct of SCC is likely not being captured uniformly across studies. For this reason, it will be essential for future MCR study design and implementation to pay special attention to the SCC criteria moving forward. More broadly, it may be time for the MCR field to examine if the SCC are indeed meaningfully contributing to the diagnosis, or whether SCC should possibly be excluded to optimize risk detection (as has begun to occur in the MCI field, from which the MCR criteria were initially derived; see (23,24,25)).

Despite the confusing and largely undefined landscape of SCC in MCR, a few trends and common practices emerged. First, most MCR studies assessed memory SCC, which is highly consistent which the broader literature on subjective cognitive decline beyond MCR, wherein 60% of items probe episodic memory concerns (47). Second, most studies assessed self-reported SCC, rather than informant or clinician reported concerns. Third, the most frequent method was to use just one item/question to classify SCC, with the vast majority of studies using 4 or fewer items. Finally, the most common approach to SCC assessment was to utilize the GDS memory item, “Do you feel you have more problems with your memory than most?” (34), with approximately half of all MCR studies using this item either alone or in combination with other SCC methods. This was the item used in the first landmark, multicenter MCR study (10), which likely led to the proliferation of this item in subsequent research.

Unfortunately, the common practices in SCC criteria largely conflict with the primary findings of the current study (Aim 2), which identified that a comprehensive, multi-item, multicognitive domain SCC assessment is best optimized to predict cognitive and clinical dementia status decline in slow gait older adults. It is important to note that slow and normal gait participants at baseline did not differ on any demographic variable or on most neuropsychological measures. Slow gait participants were more depressed (although not reaching a clinically elevated threshold) and performed worse on tests with strong psychomotor speed demands, possibly related to their significantly slowed gait speed. Slow and normal gait participants did not differ in overall endorsement of SCC at baseline. However, high SCC on the CCI-40 total score at baseline in older adults with slow gait was associated with increased risk for decline in clinical dementia status over longitudinal follow-up. This relationship was not observed in normal gait individuals, suggesting that SCC plays a unique role in predicting dementia risk in slow gait populations specifically, with key implications for SCC classification in the MCR field at large. Unfortunately, as noted above, the most common approach in the field is to include any participants who endorse SCC on any one given SCC item. In our study, only 5 participants in the overall sample did not endorse at least 1 SCC item, calling into question how MCR studies are using the SCC criterion to clearly delineate the diagnosis above and beyond the gait speed criterion. Additionally, no study has clearly established level of SCC endorsement (eg, low, medium, high) to classify SCC in MCR. In light of our findings, it is likely that most MCR studies to date are not capturing SCC across a broad enough range such that high SCC endorsement can be quantified. Furthermore, including any older adult who expresses SCC on any given cognitive item is likely not a sufficient threshold for a diagnosis, given the near ubiquitous endorsement of some extent of SCC in aging populations (27).

Because our results support that high levels of SCC predict risk for decline in slow gait participants, it follows that future work in larger samples should investigate developing a complementary approach to the SCC criteria to what is used to classify slow gait in MCR. Overall, our key findings support that high levels of SCC at baseline predicts cognitive decline among individuals with slow gait, laying the groundwork for how future studies in larger, multicenter samples may approach the SCC criterion of MCR. For example, the slow gait speed threshold is classified broadly as 1 SD below the mean. As such, perhaps SCC endorsement, quantified as 1 SD above the mean within a given cohort, may represent a meaningful level of SCC to capture those at risk for future decline. A comprehensive SCC screen (eg, the CCI-40) requires only 5–8 minutes to administer. When combined with a 2–3 minute gait speed assessment, this approach will yield a more sensitive risk detection screen in MCR than what is the current practice in the field, with relatively minimal time investment.

In terms of specific SCC items, we analyzed baseline slow and normal gait group differences as well as the predictive utility of each CCI item and the GDS memory item. Although slow and normal gait participants did not differ in overall reporting of SCC, 4 individual items (Supplementary Table 2) did significantly distinguish between gait status. An item-by-item analysis of the CCI-40 showed that high endorsement of SCC on 19 specific items were uniquely associated with rate of decline in key objective cognitive domains and clinical dementia status in slow gait participants. These 19 items ranged from memory, executive functioning, visuospatial/navigation, language, and calculation related SCC, suggesting that assessing SCC in slow gait speed groups across comprehensive cognitive domains, not just memory, may be most sensitive to risk. Interestingly, 3 out of the 4 items (2, 12, and 37) that successfully distinguished gait status were associated with significant cognitive decline over follow-up. Given the limitation in sample size in our study, future research in larger, multicenter, and multicultural samples should explore if these specific 3–4 items may constitute a possible brief SCC screen to include to enhance diagnostic discriminability and risk prediction.

We also investigated the utility of baseline endorsement of the GDS memory item (“Do you feel that you have more problems with memory than most”) in slow gait individuals, and did not find that this item was significantly associated with decline on any objective cognitive outcome or clinical dementia status, potentially undermining the current widespread use of this item in MCR studies. The use of a depression screen in many MCR studies, rather than a validated SCC measure, raises a question as to which construct, depression or SCC, is being assessed. Depression is an independent risk factor for dementia, an early marker or prodrome of dementia, as well as an accelerating factor of cognitive decline in the context of dementia (48) and a strong association between depression and SCC is well established (27,28). This association is possibly confounded in the context of MCR, given that in a pooled analysis of 22 cohorts from 17 countries, older adults with MCR were significantly more likely to self-report symptoms of depression, and depression was strongly associated with MCR in all cohorts examined (2,10). Participants with slow gait in the present study expressed significantly higher symptoms of depression than those with normal gait. Moving forward, attention to the possible overlap between SCC and depression, and cessation of depression questionnaire use to classify SCC, will be essential to effectively operationalize SCC in MCR.

Another challenge for SCC assessment in the MCR field is the international, cross-cultural use of the construct. The perception and experience of SCC, and the willingness to endorse SCC on a formal measure, or to report SCC to a medical professional have been shown to vary by demographic variables (27). Key factors such as sex (49), educational attainment (50,51), household income (50), ethnic/racial backgrounds (52) and/or lesbian, gay, bisexual, or transgender (LGBT) (53) status have all be associated with SCC prevalence and endorsement. Unfortunately, we are not aware of any studies that explore such complexities as they relate to SCC in the MCR population. Clearly, there is great need for future research which provides cross-cultural validation of the SCC criteria in MCR to facilitate accurate and timely diagnosis of dementia for underserved, rural, low-income regions around the globe.

The present study has several notable strengths, including being the first to comprehensively examine SCC classification in MCR literature. Furthermore, we clearly demonstrated the link between high baseline endorsement of SCC to cognitive and clinical dementia status decline in a demographically diverse cohort of older adults with slow gait. Our study, however, is not without limitations. First, this study was conducted in a single cohort, yielding a small population of older adults meeting criteria for slow gait (n = 42). Our sample was North American, English speaking, and urban, and thus may not be generalizable to other regions of the world where the MCR construct is utilized. Informant report of SCC, which may enhance the accuracy of subjective cognitive assessment as neurodegeneration progresses, was not available for analysis in this sample. Future research should explore self- versus informant report of SCC in MCR. The longitudinal analyses were limited by a relatively short period of annual follow-up visits available for inclusion. We were not able to explore incident dementia due to the limited follow-up period as well. Overall, future work should explore SCC in MCR within larger samples, derived from multisite, multicultural settings over a longer follow-up period to further explore the link between baseline SCC and decline in MCR.

In conclusion, the “motoric” criteria in MCR have been well validated, whereas the “cognitive” criteria have not been fully operationalized, possibly limiting the capacity of the MCR construct to accurately detect dementia risk. Our comprehensive review of the MCR literature showed striking variability and lack of attention to SCC classification. We further showed that high endorsement on a multi-item, multidomain-validated SCC questionnaire in older adults with slow gait predicts future progression from normal cognitive status to cognitive impairment. Our findings suggest that the current SCC practices in MCR may not be optimized to detect risk and should thus be reexamined. We propose that a thoughtful validation and standardization of the SCC criteria in MCR in future research will be necessary to ensure construct validity of the SCC classification, facilitate data harmonization across the MCR field, and advance accurate diagnosis of early dementia in at risk populations.

Funding

This work was supported by the National Institutes of Health, National Institute of Neurological Disorders and Stroke [F31NS127520 to C.O.N.]; National Institute on Aging [3P01AG003949 to R.B.L.]; and The CUNY Graduate Center Dissertation Fellowship to C.O.N. Support was also provided by the Leonard and Sylvia Marx Foundation, the Czap Foundation, and the Hollander Family Foundation.

Conflict of Interest

None.

Author Contributions

C.N. took the lead on study conceptualization, manuscript writing, revisions/edits, and funding, and conducted supplementary analyses. Q.G. and C.W. led the statistical analyses and provided support on manuscript revision/edits. M.J.K led study implementation, data collection, and provided support for revisions/edits. R.B.L. led study conceptualization, funding, and supplementary revisions/edits. J.V. provided support in study conceptualization and revisions/edits. L.A.R. co-led study conceptualization and provided supplementary manuscript writing and revisions/edits support.