Assessment of Instrumental Activities of Daily Living in Older Adults with Subjective Cognitive Decline Using the Virtual Reality Functional Capacity Assessment Tool (VRFCAT)

Abstract

Background

Continuing advances in the understanding of Alzheimer's disease progression have inspired development of disease-modifying therapeutics intended for use in preclinical populations. However, identification of clinically meaningful cognitive and functional outcomes for individuals who are, by definition, asymptomatic remains a significant challenge. Clinical trials for prevention and early intervention require measures with increased sensitivity to subtle deficits in instrumental activities of daily living (IADL) that comprise the first functional declines in prodromal disease. Validation of potential endpoints is required to ensure measure sensitivity and reliability in the populations of interest.

Objectives

The present research validates use of the Virtual Reality Functional Capacity Assessment Tool (VRFCAT) for performance-based assessment of IADL functioning in older adults (age 55+) with subjective cognitive decline.

Design

Cross-sectional validation study.

Setting

All participants were evaluated on-site at NeuroCog Trials, Durham, NC, USA.

Participants

Participants included 245 healthy younger adults ages 20–54 (131 female), 247 healthy older adults ages 55–91 (151 female) and 61 older adults with subjective cognitive decline (SCD) ages 56–97 (45 female).

Measures

Virtual Reality Functional Capacity Assessment Tool; Brief Assessment of Cognition App; Alzheimer's Disease Cooperative Study Prevention Instrument Project–Mail-In Cognitive Function Screening Instrument; Alzheimer's Disease Cooperative Study Instrumental Activities of Daily Living–Prevention Instrument, University of California, San Diego Performance-Based Skills Assessment–Validation of Intermediate Measures; Montreal Cognitive Assessment; Trail Making Test- Part B.

Results

Participants with SCD performed significantly worse than age-matched normative controls on all VRFCAT endpoints, including total completion time, errors and forced progressions (p≤0001 for all, after Bonferonni correction). Consistent with prior findings, both groups performed significantly worse than healthy younger adults (age 20-54). Participants with SCD also performed significantly worse than controls on objective cognitive measures. VRFCAT performance was strongly correlated with cognitive performance. In the SCD group, VRFCAT performance was strongly correlated with cognitive performance across nearly all tests with significant correlation coefficients ranging from 0.3 to 0.7; VRFCAT summary measures all had correlations greater than r=0.5 with MoCA performance and BAC App Verbal Memory (p<0.01 for all).

Conclusions

Findings suggest the VRFCAT provides a sensitive tool for evaluation of IADL functioning in individuals with subjective cognitive decline. Strong correlations with cognition across groups suggest the VRFCAT may be uniquely suited for clinical trials in preclinical AD, as well as longitudinal investigations of the relationship between cognition and function.

Continuing advances in the understanding of Alzheimer's disease (AD) progression have ignited widespread interest in the development of disease-modifying therapeutics intended for use in preclinical, largely asymptomatic populations. With at least eight clinical trials of such drugs currently underway (1), industry investment in early intervention and prevention of AD is growing. Nevertheless, reliable detection of clinically meaningful cognitive and functional change in individuals who are, by definition, asymptomatic, remains a significant challenge. The recently revised FDA draft guidance for development of early–stage AD drugs underscores the current lack of clear, acceptable endpoints for preclinical AD, particularly with respect to functioning (2). Since functional impairment remains a key differential diagnostic for dementia, preservation of functioning in preclinical and prodromal individuals is generally assumed. The recent FDA draft guidance adopts this assumption by specifying that the onset of subtle functional decline marks the progression from Stage 2 to Stage 3 prodromal disease (2).

Unfortunately, identifying the onset of functional impairment is non–trivial, and often depends on the type of functioning assessed. Although basic activities of daily living (ADLs) are largely preserved in prodromal AD, there is ample evidence for decline in more complex aspects of functioning in prodromal and even preclinical disease (1, 3, 4, 5, 6). Deficits in instrumental ADLs (IADLs) such as shopping, navigating public transportation, and cooking are well–documented in mild cognitive impairment (MCI) (1, 3, 6). Declines in IADLs have also been demonstrated in cognitively unimpaired individuals who later develop MCI or AD (3, 4, 6), suggesting that declines in complex functioning may sometimes precede objective cognitive decline. Longitudinal data from the Sydney Memory and Ageing Study supports this notion, demonstrating that subtle but measurable impairment in IADLs requiring high cognitive demand at baseline both predates and predicts the subsequent diagnosis of both MCI and AD dementia (7). These findings highlight that reliable evaluation of both cognitive performance and functioning across the entire AD continuum will be critical to assessment of treatment response at each stage. While several candidate measures have been proposed to increase the sensitivity of cognitive evaluations in preclinical AD (8, 9), reliable functional assessment has remained far more elusive (1).

Given the increasing prevalence of clinical trials for early intervention in preclinical AD, there is a growing need for measures of functioning with improved sensitivity to functional deficits in healthier, non–demented individuals (see (10, 6)). Although performance–based tools such as the Virtual Reality Functional Capacity Assessment Tool (VRFCAT) (11) and the University of California, San Diego Performance– Based Skills Assessment (12) have begun to show promise (see (10) for review) (13, 14), the most commonly utilized functional measures continue to be informantreported questionnaires, such as the Alzheimer's Disease Cooperative Study – Activities of Daily Living (ADCSADL), the Functional Activities Questionnaire (FAQ), and the Disability Assessment for Dementia (DAD), among others. These measures, though helpful in laterstage disease, often lack sensitivity to subtle functional deficits in non–demented preclinical and prodromal populations. Furthermore, a 2009 systematic review of all available informant–based IADL scales reported that none of the 12 questionnaires evaluated could be considered an adequate measurement of function due to deficient psychometric properties or lack of sufficient psychometric information (15). An updated 2016 review identified eight additional scales, but again failed to make a recommendation for the use of one or more specific instruments (16).

The Virtual Reality Functional Capacity Assessment Tool (VRFCAT) is a direct performance–based assessment of IADL functioning that uses a computer interface and is appropriate for multiple populations. Utilizing a realistic virtual environment, the VRFCAT assesses a subject's ability to complete instrumental activities associated with a shopping trip, including searching the pantry at home, making a shopping list, taking the correct bus to the grocery store, shopping in the store, paying for groceries, and returning home. In previous studies, the VRFCAT has demonstrated strong psychometric characteristics including high test–retest reliability, lack of practice effects, and strong correlations with cognition (17–18). The VRFCAT has shown sensitivity to functional impairment in schizophrenia, as well as sensitivity to age differences in the accuracy and efficiency of IADL performance (18).

The primary goal of the present work is to assess the VRFCAT's sensitivity to potential differences in IADL functioning between older adults (aged 55 and over) with and without subjective cognitive decline (SCD). SCD – defined as the subjective experience of cognitive decline from a previous normal state that is independent of objectively measurable decline in cognitive performance (19) – is gaining increasing attention as a potential preclinical marker for MCI or dementia due to AD (20, 21). Multiple epidemiological studies have shown that subjective cognitive decline is a notable risk factor for progression to dementia (22), with one meta–analysis reporting that 25% of unimpaired older adults with SCD developed MCI leading to AD in the next four years (23). As a result, SCD presents a preclinical population uniquely suited for the validation of assessment tools proposed for use in preclinical AD. In order to elucidate the relationship between SCD, objective cognitive performance, and IADL functioning, the current study also examines correlations between cognitive and functional measures in older adults with SCD.

Method

Participants

Participants included 245 healthy younger adults (YAs) ages 20–54 (131 female), 247 healthy older adults ages 55–91 (151 female) and 61 older adults with subjective cognitive decline (SCD) ages 56–97 (45 female). All subjects were recruited as a part of an ongoing normative study. Younger participants were recruited through a variety of online and paper advertisements, including Craigslist, paper flyers, and local newspaper ads. Older adults and individuals with SCD were recruited primarily though the Duke University Bryan Alzheimer's Disease Center Prevention Registry (ADPR), which includes approximately 3,600 individuals ages 55 to 95 who joined the registry to volunteer as participants in research related to cognitive aging. Initial screening was completed by phone. Although individuals with diagnoses of MCI were not specifically recruited for this study, this diagnosis was not exclusionary for individuals in the SCD group. Thus, those reporting a current diagnosis of MCI were admitted to the SCD group if they met all other inclusion criteria for enrollment. Of the 61 participants enrolled in the SCD group, six reported a current MCI diagnosis.

All participants received compensation at a rate of $50 per visit. Participants with SCD were asked to provide an informant to supply collateral information. Informants were compensated $20 for participation. Participants who failed the screening at the first visit were compensated a minimum of $10.

Measures

The Virtual Reality Functional Capacity Assessment Tool (VRFCAT)

The VRFCAT (Figure 1) presents participants with multiple instrumental activities of daily living including: navigating a kitchen, catching a bus to a grocery store, finding/purchasing food in a grocery store, and returning home on a bus. Participants complete all objectives in order. If a given objective cannot be completed (defined as five errors or more than five minutes spent on a given objective), the participant is forced to progress and assigned the worst possible score; the program then offers assistance and continues along to the next objective. The VRFCAT takes approximately 20 minutes to complete for cognitively healthy adults, and up to 35 minutes for those with SCD or MCI. Primary endpoints include Total Adjusted Time (time to complete all objectives, adjusted for instructions and error messages), Total Errors, and Total Forced Progressions. The VRFCAT consists of a tutorial and six alternate versions to allow for repeated assessment and evaluation of change over time.

Figure 1.

During the VRFCAT, participants complete 12 objectives associated with multiple instrumental activities of daily living including: navigating a kitchen, catching a bus to a grocery store, finding/purchasing food in a grocery store, and returning home on a bus

Alzheimer's Disease Cooperative Study Prevention Instrument Project – Mail– In Cognitive Function Screening Instrument (ADCS–MCFSI)

Assessment of subjective cognitive decline was completed using the ADCS–MCFSI (24). The ADCSMCFSI is a 14–item self–administered questionnaire assessing recent changes (over the last year) in cognition and functional activities that are commonly affected in the development of MCI. Scores on the ADCS–MCFSI range from 0 (no decline over the past year), to 14 (definite decline in all areas assessed), with higher scores indicating greater subjective impairment. For the present study, a ADCS–MCFSI total score equal to or greater than 4 was required for individuals assigned to the SCD group, indicating subjective decline in a minimum of four areas.

Brief Assessment of Cognition (BAC App)

The BAC App has been described in detail elsewhere (25). Briefly, the BAC App provides tablet–based administration and scoring of the original pen–and–paper Brief Assessment of Cognition (BACS) (26, 27), which includes assessment of the following domains: Verbal Memory (List Learning Task), Working Memory (Digit Sequencing Task), Verbal Fluency (Letter Fluency Task and Semantic Fluency Task), Processing Speed (Symbol Coding Task), Motor Function (Token Motor Task), and Executive Functioning (Tower of London Task). The BAC App has been clinically validated and has demonstrated equivalency to the original measure (25). In the present study, the BAC App was augmented with the addition of the following tasks to assess delayed episodic verbal memory: Delayed Free Recall, Delayed Recognition, and Forced Choice Recognition.

University of California, San Diego Performance–Based Skills Assessment – Validation of Intermediate Measures (UPSA–2–VIM)

The UPSA–2–VIM is a standard rater–administered performance–based measure of functional capacity utilizing physical props (28). The UPSA–2–VIM takes about 40 minutes to administer and measures performance in several domains of everyday living, including counting money, planning an outing, and reading a bus schedule. Raw scores from each subtest are transformed to yield scores (ranging from 0 to 20) for each and a summary score ranging from 0 to 100. Higher scores reflect better performance.

Other Measures

The Montreal Cognitive Assessment (MoCA) and the Trail Making Test – B (TMT–B) are commonly used standardized assessments, which have been described in detail elsewhere. Briefly, the MoCA (29) is a short (10–minute) cognitive status assessment scored from 0 to 30, with higher scores indicating better performance. The MoCA is commonly used for screening and follow–up evaluation of healthy older adults and those with MCI. The TMT–B (30) is a brief, timed assessment of executive functioning and processing speed. TMT–B scores are reported in seconds (up to 300 seconds), with lower scores indicating better performance.

The Alzheimer's Disease Cooperative Study Instrumental Activities of Daily Living – Prevention Instrument (ADCS–ADL–PI) (31) is a 15–item informantreported measure of functioning designed for use in a primary prevention context. The ADCS–ADL–PI was completed by informants of participants with SCD, either in person or via phone interview.

Procedure

Participants provided written informed consent prior to engagement in any study–related activities. Participants completed completed two study visits. At Visit 1, all participants completed a screening assessment that included administration of the MoCA as well as a general health and demographic questionnaire; participants of ages 55 and older were given the MCFSI. Participants with MoCA scores of 22 and greater were eligible for enrollment in the YA and OA groups. This cut–off was based on recent recommendations for suitable cut–off scores that are inclusive of the normative range across diverse demographics (e.g., (32)). Participants of ages 55 and older were eligible for the SCD group with MoCA scores of 16 and higher and ADCS–MCFSI scores of 4 and greater, indicating subjective decline in four or more distinct cognitive areas over the past year.

Because recruitment took place within the context of a larger normative study, participants with a history of documented neurologic or psychiatric disorders and/ or medical conditions interfering with daily function or cognition were excluded, with two exceptions. These exceptions were: 1) Participants with a history of depression and/or anxiety were included if they were stable (treated or non–treated) for a period of three months or more prior to participation; 2) As noted above, participants with a diagnosis of MCI or currently treated with cholinesterase inhibitors were admitted to the SCD group if they met all other inclusion criteria.

Eligible participants went on to complete the TMTB, VRFCAT, and BAC App assessments. The TMT–B assessment was added to the study after several YAs had already completed Visit 1, and is therefore only available for a portion of YAs assessed. Participants also completed the WMS IV Logical Memory subtest (33), and the Test of Premorbid Functioning (34) (both to be discussed in a separate publication). Participants of ages 55 and older completed the MCFSI at this first visit. Participants who endorsed four of more items on the MCFSI questionnaire were categorized as having SCD and were asked to provide an informant to complete the ADCS–ADL–PI.

At Visit 2 (7 to 14 days later), participants completed the UPSA–2–VIM as well as alternate versions of the VRFCAT and BAC App (the latter of the two will be presented in a separate publication regarding test–retest reliability).

Analysis

Data were analyzed with IBM® SPSS® Statistics 23.0 (2015) (35), using a 5% significance level and a 95% confidence interval. Demographic and rating scale data are reported as means and standard deviations (SDs). Continuous variables for all rating scales were compared by one–way analysis of variance with the Scheffe post hoc test, which corrects alpha for simple and complex mean comparisons. Raw scores for individual BAC App and VRFCAT measures are reported as means and SDs.

All measures were normally distributed (Kolmogorov– Smirnov normality test) so we chose to employ parametric tests to evaluate group differences. Cohen's d was used to estimate the effect size for key differences (e.g., differences in cognitive and functional measures) and computed as SCD mean – YA mean or OA mean divided by the pooled standard deviation (Cohen's d = xCC – xYA or OA / SD pooled) (36, 37, 38, 39, 40). Cohen (1988) criteria were employed for interpreting effect size d = 0.20 – 0.49 as small, d = 0.50 – 0.79 as medium, and d ≥ 0.80 as large; effect sizes above 1.00 are considered very large, indicating that the difference between the two means is larger than 1 SD.

Bonferroni corrections for multiple comparisons (p–value/number of tests) were made when several dependent or independent statistical tests were performed simultaneously (as specified in Results). To assess the relationships between the BAC App, VRFCAT, and other cognitive and functional scales, a series of Pearson r correlations were computed. Given the large number of computed correlations, Type I error rate was minimized by setting alpha to 0.01.

Results

Demographic data for participants in the younger adult (YA), older adult (OA), and subjective cognitive decline (SCD) groups are displayed in Table 1. Statistically significant differences were not assessed among the three groups as subject recruitment was stratified with respect to age, sex, race, and education. Mean ADCS–ADL–PI score for participants in the SCD group was 38.93 (SD=5.81). Mean ADCS–MCFSI score for participants in the SCD group was 5.62 (SD=1.79), with a range of 4–11.5. Mean ADCS–MCFSI score for participants in the OA group was 1.23 (SD=1.18), with a range of 0–3.

Table 1.

Demographic Characteristics

	Young Adults		Older Adults		Subjective Cognitive Decline
n	245		247		61
	n	%	n	%	n	%
Gender
Male	114	46.53	96	38.87	16	26.23
Female	131	53.47	151	61.13	45	73.77
Racea
African American	75	30.61	50	20.24	21	34.23
Asian	10	4.08	0	0.00	0	0.00
American Indian	1	0.39	0	0.00	0	0.00
Caucasian	150	61.22	197	79.76	40	65.57
Other	8	3.27	0	0.00	0	0.00
Ethnicity
Hispanic (Latino)	33	13.47	3	1.21	4	6.56
Not Hispanic (Latino)	212	86.53	244	98.79	57	93.44
	Mean	SD	Mean	SD	Mean	SD
Education (years)	14.41	2.23	15.21	2.71	14.13	2.75
Age (years)	39.00	10.27	68.46	7.98	71.38	9.21

Young Adults: Age 20 – 54 years; Older Adults: Age > 54 years; SD: Standard Deviation; Data for Race was not reported for one subject (0.41%) in the Young Adult group.

Mean performance for YA, OA, and SCD groups on each VRFCAT summary measure is displayed in Figure 2. Statistically significant differences were demonstrated among the three groups for all three summary measures, including VRFCAT Total Adjusted Time, Total Errors and Total Forced Progressions (p≤0.001 for all, after Bonferonni correction). Scheffe's post hoc tests showed a significant difference between the YA group, the OA group, and the SCD group, with the OA group performing significantly worse than the YA group and the SCD group performing significantly worse than the OA group on all three measures (p≤0.001 for all comparisons).

Figure 2.

Mean VRFCAT Total Adjusted Time (Panel A; Cohen's d: YA–OA = 0.99, SCD–YA = 1.45, SCD–OA = 0.65), Total Errors (Panel B; Cohen's d: YA–OA = 0.48, SCD–YA = 1.08, SCD–OA = 0.55) and Total Forced Progressions (Panel C; Cohen's d: YA–OA = 0.66, SCD–YA = 1.12, SCDOA = 0.55) for participants in each group

Mean Adjusted Time and Errors for each VRFCAT objective are displayed in Figures 3A and 3B, respectively. For Adjusted Time, Bonferonni corrected (p≤0.002) omnibus ANOVA tests were significant for differences among the groups on all VRFCAT objectives with the exception of Objective 4: Pick up the Billfold on the Counter (p=0.014), and Objective 11: Wait for the Correct Bus to Your Apartment (p=0.013). Scheffe's post hoc tests performed for objectives with significant omnibus tests showed significant differences between the SCD and OA groups for eight of the remaining objectives; post–hoc tests were not significant for Objective 5: Exit the Apartment and Head to the Bus, and Objective 6: Wait for the Correct Bus to the Grocery Store.

Figure 3.

Mean Adjusted Time (Panel A) and Errors (Panel B) for VRFCAT objectives 1–12. Error bars depict +/– one standard error

For VRFCAT Errors (Figure 3B), Bonferonni corrected (p≤0.002) omnibus ANOVA tests were significant for differences among the groups on Objective 3: Cross Off Ingredients in Kitchen, Objective 9: Shop for Grocery Items on List, Objective 10: Pay for Groceries with Exact Change, Objective 11: Wait for and Board Correct Bus Home, and Objective 12: Pay for Bus with Exact Change. Scheffe's post hoc tests performed for objectives with significant omnibus tests showed significant differences between the SCD and OA groups on all but Objective 9: Shop for Grocery Items on List (p>0.1).

Group means and standard deviations for cognitive performance measures are displayed in Table 2, along with effect size (Cohen's d) calculations for comparisons between groups. The SCD group performed significantly worse than the OA group on objective cognitive tests. On the MoCA, the SCD group performed 0.92 SDs (2.67 points) lower than the OA group. On the TMT–B, the SCD group took an average of 37.62 seconds (0.69 SDs) longer than the OA group to complete the task.

Table 2.

Cognitive Performance

Cognitive Test	Younger Adults (YA)			Older Adults (OA)			OAs with Subjective Cognitive Decline (SCD)
	n	Mean	SD	n	Mean	SD	n	Mean	SD	ANOVA	Cohen's d
MoCA	241	26.51	2.25	237	26.78	1.84	61	24.11	3.66	F(2,538)=33.898 p < 0.001	SCD–YA = 0.79 SCD–OA = 0.92 YA–OA = 0.13
TMT–B (seconds)	185	76.87	37.11	232	91.18	39.33	60	128.80	66.79	F(2,476)= 33.214 p < 0.001	SCD–YA = 0.96 SCD–OA = 0.69 YA–OA = 0.37
BAC–APP Measures
Verbal Memory	243	44.88	8.69	245	41.26	8.33	60	34.37	9.54	F(2, 547) = 37.911 p < 0.001	SCD –YA = 1.15 SCD –OA = 0.77 YA–OA = 0.43
Digit Sequencing	243	20.32	4.17	245	20.38	3.78	60	18.22	3.93	F(2, 547) = 7.721 p = 0.001	SCD –YA = 0.52 SCD –OA = 0.56 YA–OA = 0.02
Delayed Free Recall	166	8.90	3.29	246	8.02	3.32	59	5.63	3.92	F(2, 470) = 20.402 p < 0.001	SCD –YA = 0.90 SCD –OA = 0.66 YA–OA = 0.27
Recognition	166	28.93	1.60	245	28.64	1.60	60	27.02	3.40	F(2, 470) = 22.686 p < 0.001	SCD –YA = 0.72 SCD –OA = 0.61 YA–OA = 0.18
Forced Choice Recognition	166	14.93	0.28	243	14.88	0.38	60	14.45	1.79	F(2, 468) = 10.597 p < 0.001	SCD –YA = 0.37 SCD –OA = 0.33 YA–OA = 0.15
Token Motor	243	89.16	29.91	243	59.52	23.44	59	52.61	21.57	F(2, 545) = 94.846 p < 0.001	SCD –YA = 1.40 SCD –OA = 0.31 YA–OA = 1.10
Semantic Fluency	243	23.44	6.20	244	21.25	5.54	61	18.33	5.57	F(2, 547) = 21.349 p < 0.001	SCD –YA = 0.87 SCD –OA = 0.53 YA–OA = 0.37
Letter Fluency	243	33.34	9.36	238	32.66	9.49	60	27.93	8.33	F(2, 531) = 8.181 p < 0.001	SCD –YA = 0.61 SCD –OA = 0.53 YA–OA = 0.07
Symbol Coding	235	51.24	12.71	237	36.73	10.11	60	29.85	11.27	F(2, 531) = 135.104 p < 0.001	SCD –YA = 1.79 SCD –OA = 0.64 YA–OA = 1.26
Tower of London	233	17.28	2.70	237	16.61	2.95	58	15.07	3.59	F(2, 527) = 13.745 p < 0.001	SCD –YA = 0.69 SCD –OA = 0.47 YA–OA = 0.24

BAC–APP: Brief Assessment of Cognition – Application; Bonferroni adjusted p value = 0.005 (0.05 p value/12 cognitive tests); SD: Standard Deviation

On the BAC App, statistically significant differences were demonstrated among the three groups for all ten individual BAC App tests (p≤0.001 for all, after Bonferonni correction). Scheffe's post hoc tests showed significant differences between the SCD and OA groups (p values range from 0.04 to p≤0.001) for all BAC App tests except Token Motor (p=0.196). Consistency of findings across all cognitive measures indicated the sensitivity of these cognitive assessments in detecting objective cognitive differences between the OA and SCD groups, which were defined only through subjective reporting of cognitive decline on the ADCS–MCFSI.

Correlational analyses were conducted to examine the relationship between the VRFCAT and the UPSA–2– VIM, as well as the relationship between cognitive and functional measures in the SCD group. UPSA–2–VIM total score was strongly correlated with all VRFCAT Total Adjusted Time (r=–0.56, p<.01), Total Errors (r=–0.47,p<.01), and Forced Progressions (r=–0.56, p<0.01).

In all cases, decreased performance on the VRFCAT measure was associated with a corresponding decrease in UPSA–2–VIM performance.

VRFCAT measures were also strongly correlated with cognitive performance. Table 2 displays correlations between summary VRFCAT measures, MoCA, TMTB, and BAC App tests for the YA, OA, and SCD groups. Across all three groups, total completion time of the VRFCAT was associated with processing speed (Symbol Coding r >0.500). In the SCD group, VRFCAT performance was strongly correlated with cognitive performance across nearly all tests with significant correlation coefficients ranging from 0.3 to 0.7. Only Forced Choice Recognition was not significant due in part to ceiling performance across all groups. In the SCD group, VRFCAT summary measures all had correlations greater than r=0.5 with MoCA performance and BAC App Verbal Memory (p<0.01 for all). VRFCAT Total Time was also strongly correlated with measures of processing speed (r=–0.714, p<0.01 for Symbol Coding) and executive functioning (r=–0.439 for Tower of London and r=0.545 for TMT–B; p<0.01 for both).

Discussion

The past two decades have brought a series of failed clinical trials for new treatments of Alzheimer's disease (37–38). The failure of symptomatic treatments, coupled with increased understanding of disease progression, has catalyzed interest in disease–modifying treatments intended for largely asymptomatic, preclinical AD populations. However, selection of appropriate functional endpoints in preclinical AD poses a considerable challenge due to the preservation of those functional activities assessed by traditional informant–based measures.

In order to address this challenge, the present study examined the sensitivity of the VRFCAT – a performancebased measure of IADL functioning – to differences between healthy younger adults, healthy older adults, and healthy older adults with SCD.

Older adults with SCD performed significantly worse than OAs without SCD on nearly all cognitive measures assessed (Forced Choice Recognition was the only exception), suggesting that SCD was associated with objective cognitive decline in the present sample (Table 2). In addition, the VRFCAT was sensitive to differences between all groups, with YAs significantly outperforming OAs and OAs significantly outperforming those with SCD. Importantly, performance on all three summary measures (Total Adjusted Time, Total Errors, and Total Forced Progressions) distinguished between the groups (Figure 2). Findings comparing YAs and OAs replicate earlier work (18), while findings in SCD suggest that the VRFCAT may be sensitive to the earliest detectable declines in IADL functioning in preclinical AD. Strong correlations between the VRFCAT and the UPSA–2–VIM, a rater–administered performance–based measure of functioning, confirm previous reports (17, 18) and offer additional evidence in support of the construct validity of the VRFCAT.

Results also demonstrated strong relationships between VRFCAT outcomes and performance on a diverse set of cognitive measures (Table 3). In the SCD group, performance on tests of verbal memory and executive functioning, both of which demonstrated objective decline, showed remarkably strong correlations (r>0.5) with VRFCAT performance. This finding is particularly striking given that these domains are often among the first to decline and exhibit the most robust changes observed in preclinical AD populations compared to healthy controls who remain stable over time (39–40).

Table 3.

Correlations between VRFCAT Measures and Cognitive Performance

	YA			OA			SCD
	VRFCAT TOTAL ADJUSTED TIME	VRFCAT ERRORS	VRFCAT FORCED PROGRESSIONS	VRFCAT TOTAL ADJUSTED TIME	VRFCAT ERRORS	VRFCAT FORCED PROGRESSIONS	VRFCAT TOTAL ADJUSTED TIME	VRFCAT ERRORS	VRFCAT FORCED PROGRESSIONS
MoCA	–.364**	–.154*	–.096	–.350**	.172**	.258**	–.538**	–.503**	–.537**
TMT–B	.336**	.132	.063	.533**	.350**	.453**	.545**	.400**	.465**
BAC App:
Verbal Memory	–.422**	–.298**	–.221**	–.257**	–.136*	–.238**	–.570**	–.522**	–.554**
Digit Sequencing	–.377**	–.228**	–.103	–.362**	–.268**	–.248**	–.408**	–.433**	–.292*
Delayed Free Recall	–.333**	–.194*	–.163*	–.266**	–.170**	–.279**	–.497**	–.433**	–.507**
Recognition	–.199*	–.136	–.089	–.185**	–.096	–.191**	–.457**	–.164	–.467**
Forced Choice Recognition	–.162*	–.077	–.151	–.148*	–.180**	–.261**	–.12	–.113	–.086
Token Motor	–.357**	–.299**	–.254*	–.259**	–.184*	–.292**	–.406*	–.475**	–.367*
Forced Choice Recognition	–.376**	–.249**	–.159*	–.331**	–.168**	–.282**	–.394**	–.315*	–.343*
Semantic Fluency	–.321**	–.212**	–.136*	–.345**	–.219**	–.277**	–.429**	–.327*	–.317*
Letter Fluency	–.298**	–.147*	–.113	–.324**	–.150*	–.234**	–.226	–.308*	–.099
Symbol Coding	–.526**	–.306**	–.267**	–.512**	–.268**	–.440**	–.714**	–.477**	–.571**
Tower of London	–.387**	–.274**	–.255**	–.189**	–.159*	–.249**	–.439**	–.544**	–.357*
VRFCAT:
VRFCAT
TOTAL ADJUSTED TIME	1	.637**	.662**	1	.626**	.812**	1	.665**	.857**
VRFCAT ERRORS	.637**	1	.522**	.626**	1	.676**	.665**	1	.618**
VRFCAT FORCED PROGRESSIONS	.662**	.522**	1	.812**	.676**	1	.857**	.618**	1

** Correlation is significant at the 0.01 level (2–tailed). * Correlation is significant at the 0.05 level (2–tailed).

The present study also evaluated the use of the BAC App to assess cognition in individuals with SCD. Like the VRFCAT, the BAC App leverages technology to reduce rater burden by providing automated stimulus presentation and scoring (see (25)). In the present study, the BAC App demonstrated sensitivity to objective cognitive differences between YAs, OAs, and individuals with SCD (Table 2). These findings provide support for the BAC App as a method of cognitive assessment in trials for preclinical AD.

There are a few limitations to the current study. First, because prior work on SCD has applied heterogeneous methods of evaluating subjective decline, there is no guarantee that the current findings would hold if SCD were operationalized in a different manner. A cut–off score of 4 on the ADCS–MCFSI was chosen to ensure moderate subjective impairment, and this may have increased the probability of objective decline in our SCD group relative to other studies. Inclusion of six subjects with a current diagnosis of MCI may also have contributed. However, since the aim of the current study was to assess the use of the VRFCAT in the early portion of the AD continuum, we decided that eliminating those with an MCI diagnosis would not be useful, particularly given the likelihood that more individuals in the SCD group may meet diagnostic criteria for MCI if a full clinical evaluation were available.

The findings from the present study suggest that the VRFCAT provides a sensitive tool for evaluation of IADL functioning in SCD with strong correlations to cognitive performance. Taken together with prior findings regarding the strong psychometric properties of the VRFCAT, including strong test–retest reliability and lack of practice effects, these findings suggest that the VRFCAT may be uniquely suited for use in clinical trials in preclinical AD, as well as in longitudinal investigations of the relationship between cognition and function throughout the AD continuum.

Funding: This research was supported National Institutes of Health under NIMH 2R44 MH084240 (ASA) and NIA 1R44AG03191 (RSE). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health

Disclosures: AS Atkins is a full–time employee of NeuroCog Trials, Durham, NC, USA, and has received support from National Institute of Mental Health and National Institute on Aging. A Khan is a full–time employee of NeuroCog Trials, Durham, NC, USA, and has received support from National Institute of Mental Health, Janssen, Celgene, Teva Pharmaceuticals, and Stanley Medical Research Foundation. D Ulshen, A Vaughn and D Balentin are full–time employees of NeuroCog Trials, Durham, NC, USA, H. Dickerson as a NeuroCog employee. LE Liharska is a consultant to NeuroCog Trials. B Plassman currently or in the past 3 years has received funding from the National Institute on Aging and the National Institute of Environmental Health Sciences, and has served as a consultant for Takeda. K Welsh–Bohmer is a consultant to NeuroCog Trials. She currently or in the past 3 years has received funding from the National Institute on Aging and has received honoraria, served as a consultant, or advisory board member for Takeda, Biogen, Roche, T3D Therapeutics, Diffusion Pharmaceutical company, and Merck. RSE Keefe currently or in the past 3 years has received investigator–initiated research funding support from the Department of Veteran's Affair, Feinstein Institute for Medical Research, GlaxoSmithKline, National Institute of Mental Health, National Institute on Aging, Novartis, Psychogenics, Research Foundation for Mental Hygiene, Inc., and the Singapore National Medical Research Council. Dr. Richard Keefe currently or in the past 3 years has received honoraria, served as a consultant, speaker, or advisory board member for Abbvie, Acadia, Aeglea, Akebia, Akili, Alkermes, ArmaGen, Astellas, Avanir, AviNeuro/ChemRar, Axovant, Biogen, Boehringer–Ingelehim, Cerecor, CoMentis, Critical Path Institute, FORUM, Global Medical Education (GME), GW Pharmaceuticals, Intracellular Therapeutics, Janssen, Lundbeck, Lysogene, MedScape, Mentis Cura, Merck, Minerva Neurosciences Inc., Mitsubishi, Monteris, Moscow Research Institute of Psychiatry, Neuralstem, Neuronix, Novartis, NY State Office of Mental Health, Otsuka, Pfizer, Regenix Bio, Reviva, Roche, Sangamo, Sanofi, Sunovion, Takeda, Targacept, University of Moscow, University of Texas Southwest Medical Center, and WebMD. Dr. Keefe receives royalties from versions of the BAC testing battery, the MATRICS Battery (BACS Symbol Coding), and the Virtual Reality Functional Capacity Assessment Tool (VRFCAT). He is also a shareholder in NeuroCog Trials, Inc. and Sengenix.

Ethical standard: The study protocol was approved by a central Institutional Review Board. All participants provided written informed consent prior to participation.