Changes in LASSI-L performance over time among older adults with amnestic MCI and amyloid positivity: A Preliminary Study.

Abstract

There is a pressing need to develop measures that are sensitive to the earliest subtle cognitive changes of Alzheimer’s disease (AD) to improve early detection and track disease progression. The Loewenstein-Acevedo Scales of Semantic Interference (LASSI-L) has been shown to successfully discriminate between cognitively unimpaired (CU) older adults and those with amnestic mild cognitive impairment (MCI) and to correlate with total and regional brain amyloid load. The present study investigated how the LASSI-L scores change over time among three distinct diagnostic groups. Eighty-six community-dwelling older adults underwent a baseline evaluation including: a clinical interview, a neuropsychological evaluation, Magnetic Resonance Imaging (MRI), and amyloid Positron Emission Tomography (PET). A follow up evaluation was conducted 12 months later. Initial mean values were calculated using one-way ANOVAs and chi-square analyses. Post-hoc comparisons were conducted using Tukey’s Honestly Significant Difference(HSD). A 3 × 2 repeated measures analysis was utilized to examine differences in LASSI-L performance over time. The MCI amyloid positive group demonstrated a significantly greater decline in LASSI-L performance than the MCI amyloid negative and CU groups respectively. The scales that best differentiated the three groups included the Cued A2, which taps into maximum learning capacity, and Cued B2, which assesses the failure to recover from proactive semantic interference. Our findings further support the LASSI-L’s discriminative validity.

1. Introduction

Alzheimer’s Disease (AD) is a chronic degenerative brain disorder characterized by progressive memory loss and cognitive deficits severe enough to interfere with tasks required for independent daily living [National Institute on Aging (NIA), 2019]. It is by far the most prevalent form of dementia, accounting for 60 to 80 percent of all cases. Currently, AD affects more than six million Americans, and this number could grow to 13.8 million by 2060 (Alzheimer’s Association, 2021). The neuropathological changes associated with AD, particularly the accumulation of beta-amyloid (plaques) and twisted strands of tau (tangles), begin decades before the clinical symptoms of the disease manifest (Braak et al., 2011; Jack et al., 2009; Villemagne et al., 2013). Clinically, AD has been conceptualized as progressing along a continuum with three distinct disease phases: preclinical AD, Mild Cognitive Impairment due to AD (MCI), and dementia due to AD (Jack et al., 2018).

Early identification of the disease process, accurate diagnosis, and appropriate management of the clinical manifestations of AD are crucial for clinical care (Budd et al., 2011; Rasmussen & Langerman, 2019; Sun et al., 2018). Timely detection of preclinical AD may help identify a therapeutic window in which emerging disease-modifying agents could more effectively treat the disease before amyloid plaques and neurofibrillary tangles accumulate and result in marked neurodegeneration (Loewenstein et al., 2018b; Wattamwar and Mathuranath, 2010).

Recently, there has been significant progress in the quantification of early brain pathology due to the identification of AD biomarkers that can be detected with structural and molecular neuroimaging, cerebrospinal fluid analysis, and genetic mutation analysis. Despite these advances, AD remains both highly unrecognized and underdiagnosed (Amjad et al., 2018; Barrett et al., 2006; Kotagal et al., 2015). Recently, the NIA and the Alzheimer’s research framework criteria emphasized a biological definition of AD for research purposes (Jack et al., 2018); however, biomarkers such as amyloid and tau PET as well as CSF markers of AD alone are not yet reliable for AD diagnosis, particularly in the preclinical stages and remain costly, invasive, and largely unavailable (Adlard et al., 2014; Jack et al., 2016; Khoury & Ghossoub, 2019; Matsuda et al., 2019). Further, evidence of amyloid and tau pathology alone is insufficient for predicting AD progression, as not all individuals who have pathology in the brain will necessarily exhibit cognitive or biological progression (Nelson et al., 2011; Serrano-Pozo et al., 2014). For the above reasons, there is a pressing need to develop sensitive, valid, and easily accessible cognitive outcome measures that can provide diagnostic value, measure disease progression and their underlying relationship with AD biomarkers (See Loewenstein et al., 2017a; Loewenstein et al., 2017b; Loewenstein et al., 2018a).

Traditional neuropsychological tests commonly used to capture cognitive deficiencies associated with AD include the use of list-learning tasks [e.g., the Hopkins Verbal Learning Test-Revised (HVLT-R; Brandt & Benedict, 2001)]; immediate and delayed recall for story passages [e.g., Logical Memory of the Wechsler Memory Scale – Fourth edition (WMS-IV; Wechsler, 2009)]; Verbal Paired Associate learning tasks [e.g., The Verbal Paired Associates Subtest, of the WMS-IV (Wechsler, 2009)]; and immediate and delayed recall of geometric designs [e.g., Brief Visual Memory Test-Revised (BVMT; Benedict et al., 1996)]. While research suggests sensitivity values ranging from 80% to 90% and specificity values ranging from 82% to 90% for traditional neuropsychological assessments in differentiating individuals with AD dementia from non-demented controls (Bloudek et al., 2011), these instruments have also proven to lack sensitivity to the earliest subtle cognitive changes that are present in the preclinical stages of AD (Brooks et al., 2010; Loewenstein et al., 2016; Loewenstein et al., 2018a; Loewenstein et al., 2018b; Thomas et al., 2018a). In addition, traditional neuropsychological assessments also fail to successfully discriminate AD from other neurodegenerative conditions that might present with impaired cognition during the very early stages of disease (Hornberger & Piguet, 2012). For example, rate of forgetting and episodic memory deficits, which have been traditionally identified as early deficiencies observed in AD (Grober et al., 2008; Salmon, 2011; Saxton et al., 2004), are also commonly present during the early stages of other neurodegenerative conditions, such as, fronto-temporal dementias and other neurological conditions that affect medial temporal lobe structures (Hornberger et al., 2010; Hornberger & Piguet, 2012). Importantly, most traditional neuropsychological assessments do not employ controlled learning paradigms which might result in the use of individualized strategies for learning leading to greater intra-individual variance in the initial acquisition stage.

There is increasing evidence that persons with Preclinical AD may be uniquely susceptible to semantic interference (Loewenstein et al., 2003; Loewenstein et al., 2004) and while some traditional neuropsychological measures may include several semantically related items on competing to-be-remembered lists, these have an insufficient number of shared to-be-remembered targets belonging to the same semantic category which would be required to adequately elicit sufficient semantic interference. Particularly, individuals with AD are vulnerable to two types of semantic interference: Proactive Semantic Interference (PSI), which has been defined as old learning interfering with the new learning of new semantically related targets; and the Failure To Recover From Proactive Semantic Interference (fPSI), which occurs when the new list-to be remembered is presented again and the individual is still unable to recover from the interference effect (Loewenstein et al., 2016; Loewenstein et al., 2018b; Matias-Guiu et al., 2017; Sanchez et al., 2017).

The Loewenstein-Acevedo Scales for Semantic Interference & Learning (LASSI-L) is a cognitive stress test paradigm with robust psychometric properties (Crocco et al., 2014; Curiel et al., 2013), that has demonstrated high sensitivity to the subtle cognitive changes present in the preclinical and prodromal stages of AD (Crocco et al., 2018; Curiel et al., 2018; Loewenstein et al., 2016; Loewenstein et al., 2018b). The LASSI-L is considered a “cognitive stress test” because by employing controlled learning and cued recall over two trials it maximizes the storage of to-be-remembered semantic information, reducing the individualized use of learning strategies. More importantly, the LASSI-L employs semantically similar list of targets, measuring susceptibility to PSI, frPSI, and the total amount of semantic intrusion errors (IE) produced during these trials. Studies have shown that performance on subscales tapping into PSI and frPSI on the LASSI-L effectively discriminated between community-dwelling older adults with amnestic mild cognitive impairment (aMCI) and older adults without cognitive impairment (Crocco et al., 2014; Curiel et al., 2013; Matias-Guiu et al., 2017). Further, frPSI has been related to brain volumetric loss on MRI in AD prone areas among older adults with MCI (Loewenstein et al., 2017a; Loewenstein et al., 2017b). Intrusion errors on scales tapping into PSI and frPSI were able to differentiate individuals with MCI who had positive brain amyloid (Amy+) visualized on PET/CT from those with MCI who were amyloid negative (Amy−) and presumptive other non-AD conditions (Loewenstein et al., 2018b). Matias-Guiu and colleagues (2018) found that the LASSI-L outperformed the Free and Selective Reminding Test in Europe in the diagnosis of early AD with significantly greater area explained under the receiving operating characteristic curve (ROC). Intrusion errors and increase susceptibility to semantic interference on other measures have been identified by the recent literature as sensitive markers of early AD-related changes (Libon et al., 2011; Salmon & Bondi, 2009; Thomas et al., 2018b).

Importantly, although baseline performance on the LASSI-L (subscales susceptible to PSI and frPSI) predicted aMCI individuals who later progressed to a dementia stage or PreMCI individuals who later progressed to a stringent diagnosis of MCI (Crocco et al., 2021), there have been no studies conducted up to date regarding changes in performance on these LASSI-L subscales. The aim of the current investigation was to determine which elements of the LASSI-L were most prone to change over a follow-up period among three distinct diagnostic groups of older adults: aMCI (Amy+), aMCI (Amy−), or cognitively unimpaired older adults (CU: amyloid negative). To our knowledge, this represents the first investigation into longitudinal changes in LASSI-L performance among older adults that are at higher versus lower risk of developing AD.

2. Methods

In the present study we recruited 86 community-dwelling older adults from the 1Florida Alzheimer’s Disease Research Center (ADRC), Clinical Core site, (Dr. Duara, Principal Investigator) at Mount Sinai Medical Center, Miami Beach, Florida. The investigation was carried out in accordance with the latest version of the declaration of Helsinki. The study design was reviewed by an appropriate ethical committee. Informed consent of the participants was obtained after the nature of the procedures were fully explained. The participants ages ranged from 54 to 98 years. From the total sample, 47% of individuals identified themselves as Hispanics, 51% as Caucasian, and 2% as African Americans. Among the Hispanic group, 87% were tested in Spanish and 13% were tested in English. All measurements utilized were available in the Spanish language and administered by a fully bilingual research associate. Specific demographic variables are described in detail in Table 1. All participants underwent an extensive baseline evaluation that included a clinical interview, a neuropsychological evaluation, Magnetic Resonance Imaging (MRI) of the brain, and an amyloid Positron Emission Tomography (PET) to quantify amyloid load in the brain. The clinical interview included the Clinical Dementia Rating scale (CDR; Morris, 1997), which was administered by an experienced clinician. The neuropsychological evaluation included the total recall and delayed recall subtests from the HVLT-R (Brandt, 1991); Delayed recall from the Logical Memory subtest from the National Alzheimer’s Coordinating Center Uniform dataset, version 3.0 (NACC UDS; Beekly et al., 2007); the Category Fluency task (animals, vegetables, and fruits) (Binetti, 1995), the Block Design subtest of the WAIS-IV (Wechsler, 2008), and Parts A & B of the Trail Making Test (Reitan, 1958).

Table 1.

Demographic Characteristics by Diagnostic Group

	CU (Amy−)	MCI (Amy−)	MCI (Amy+)	F-Value or X 2	p-value
Age (54–98)	69.72	71.78	72.68	1.05	.354
	(SD=6.5)	(SD=8.1)	(SD=7.4)
Education (6–22)	15.64	14.50	15.59	1.16	.317
	(SD=3.1)	(SD=3.4)	(SD=3.6)
MMSE (23–30)	29.27a	28.18c	26.3900b	22.07	<.001
	(SD=.8)	(SD=1.7)	(SD=1.9)
Sex (% female)	45.5%	50.0%	50.0%	.14	.93
Language (% Spanish)	54.5%	50.0%	38.2%	1.68	.43
Time to re-evaluation	14.06	14.96	15.60	1.19	.310
(11.07 – 27.76)	(SD=2.1)	(SD=3.7)	(SD=3.9)
Centiloid Value	−.79b	2.46b	73.02a	194.28	<.001
(−21.71 – 125.48)	(SD=9.4)	(SD=11.1)	(SD=23.6)

Note. Bolded values are statistically significant Means with different alphabetic superscripts are statistically different using the Tukey’s Honestly Significant Difference Test (Tukey’s HSD)

Abbreviations: CU=Cognitively unimpaired; MCI=Mild Cognitive Impairment; Amy=amyloid; MMSE=Mini-Mental state examination

All individuals were administered the LASSI-L during the baseline evaluation and on the annual follow up visit. The LASSI-L was not used for diagnostic purposes with regards to group classification. Based on their performance on the clinical and diagnostic evaluation depicted above, participants were classified into three distinct diagnostic groups using the following criteria:

2.1. Cognitively Unimpaired group - Amyloid Negative (CU (Amy−) n=25)

Participants were classified as CU (Amy−) if: a) there were no subjective cognitive complaints made by the participant and/or a collateral informant; b) the Global CDR score was 0, confirming no evidence of memory or other cognitive domain decline after an extensive clinical interview with the participant and informant; c) performance on all memory (e.g., HVLT-R or delayed paragraph recall from the NACC UDS 3.0) and non-memory measures (e.g., Category Fluency, Trails A and B, WAIS-IV Block Design subtest) were less than 1.0 SD below normal limits for age, education, and language of testing; d) the amyloid PET scan was read as an amyloid negative scan by an experienced rater.

2.2. MCI Suspected Alzheimer’s Disease, Amyloid Positive (MCI (Amy+) n=28)

Participants in the MCI (Amy+) group: a) fulfilled Petersen’s criteria for MCI (Petersen et al., 2014), b) had subjective cognitive complaints expressed by the participant and/or informant; c) Global CDR scale score of 0.5; d) had impaired performance on memory measures (i.e., 1.5 SD or greater below the mean, accounting for age, education, and language of testing) on either the HVLT-R or delayed paragraph recall from the NACC UDS 3.0, and in addition to the memory impairment might had impairment on 1 non-memory measure (e.g., Category Fluency, Trails A and B, WAIS-IV Block Design subtest); e) the Mini Mental State Examination (MMSE; Folstein, et al., 1975) score > 23; f) had the PET scans read as amyloid positive by an experienced rater.

2.3. MCI Suspected Non-Alzheimer’s Disease, Amyloid Negative (MCI (Amy−) n=33)

Participants in the MCI (Amy−) group fulfilled all criteria outlined for the MCI (Amy+) group, with the exception that the amyloid PET scan was read as amyloid negative by an experienced rater.

2.4. Experimental Measure (LASSI-L)

The LASSI-L was not used for diagnostic determination in this study to avoid potential issues of circularity. This novel cognitive stress employs a paradigm that includes both controlled learning and cued recall to maximize the learning and storage of a list of to be remembered semantically related targets. The presented words belong to one of three distinct semantic categories (i.e., fruits, articles of clothing, and musical instruments) (Curiel et al., 2013).

Participants were tested in their preferred language, English or Spanish. Previous research has shown the LASSI-L to be culturally fair and valid in either language (Curiel et al., 2019; Matias-Guiu et al., 2017). During the administration of the LASSI-L, the examinee is instructed to remember a list of 15 common words. The cued recall after the second presentation is a measure of maximum storage capacity (Trial A2). A unique aspect of the LASSI-L paradigm is the presentation of a second competing list of to-be-remembered words that is presented in the same manner as the first list. That is, List B is presented immediately following the second cued recall trial of List A. As with List A, there are two consecutive presentations of List B. The second list introduces different words but shares the same previously presented semantic categories in order to elicit a considerable amount of PSI (Trial B1). Unlike other traditional memory paradigms, the presentation of List B for a second time and the subsequent recall of this second list of words provides an opportunity for the individual to recover from the effects of PSI. In the current study we focused on the following LASSI-L subscales: Cued A2 recall (measure of maximum storage capacity), Cued B1 recall (measure of PSI), and Cued B2 recall (measure of frPSI) as these scales have shown to successfully discriminate aMCI from CU adults (Crocco et al., 2014; Curiel et al., 2013; Matias-Guiu et al., 2017). We also focused on the amount of intrusion errors produced on Cued B1 and Cued B2 subscales as these have shown to differentiate individuals with MCI who are amyloid positive versus amyloid negative (Loewenstein et al., 2018b). These intrusion errors typically consist of words from the first list of semantically similar target items or other non-target words. Lastly, we calculated a percentage of intrusion errors by comparing total intrusion errors to total responses produced by each diagnostic group. The percentage of intrusion errors (PIE) was calculated by computing the total intrusion errors / (total intrusion errors plus total correct responses) for Cued B1 and Cued B2 subscales on the LASSI-L.

Each participant received the LASSI-L at baseline and during their first annual follow-up visit, which occurred approximately 12 months after their initial evaluation. This longitudinal assessment allowed us to measure cognitive deterioration over time for each of the distinct groups with different amyloid status.

2.5. Amyloid Imaging

PET/CT imaging was obtained using a 3D Hoffmann brain phantom to establish a standardized acquisition and reconstruction method. Participants were infused with [18-F] florbetaben (Neuraceq; Life Molecular Imaging) 300 MBQ over a 3-minute period. Scanning commenced 70 to 90 minutes after the infusion for a duration of 20 minutes. We scanned all participants on a Siemens Biograph 16 PET/CT scanner operating in 3D mode (55 slices/frame, 3mm slice thickness 128 × 128 matrix). The PET data was reconstructed into a 128 × 128 × 63 (axial) matrix with voxel dimensions of 0.21 × 0.21 × 0.24 cm.

A small number of participants had Florbetapir as their amyloid tracer (22 % of subjects). Reconstruction was performed using the manufacturer- supplied software and included corrections for attenuation, scatter, random coincidences and dead time. Images for regional analyses were processed using Fourier analysis followed by direct Fourier reconstruction. Images were smoothed with a 3 mm Hann filter. Following reconstruction, image sets were inspected and, if necessary, corrected for inter-frame motion. Images were obtained from the top of the head to the top of the neck and CT data was employed for initial attenuation correction and image reconstruction in the sagittal, axial, and coronal planes.

The PET/CT scans, including the outline of the skull, co-registered linearly (i.e., trilinear interpolation) with 12 degrees of freedom, onto the volumetric MRI scan using a T1-weighted (MP-RAGE) (Lizarraga et al., 2018; Smith et al., 2004). Region-of-interest (ROI) boundaries were defined manually using the structural MRI for anatomical reference, and criteria that have been proven to provide highly reproducible outcomes (Desikan et al., 2006). This registration process ensured that the florbetaben PET/CT image had the same accurate segmentation and parcellation as in the MRI scan. Average activity was calculated in the ROIs corresponding to cerebellar gray matter and cerebral cortical regions. While quantitative values called centiloids were calculated, these values were used only to provide mean values for the three groups, as outlined in Table 1. A Centiloid value (CL) represent a measure of total amyloid burden that can be applied to multiple amyloid traces. In our laboratory a CL value of 40 or above denotes a positive amyloid scan with 100% concordance with expert visual ratings, while a CL value of <17 denotes an amyloid negative scan with over 95% concordance with expert visual reading.

2.6. Visual Ratings of Amyloid PET Scans

All amyloid PET scans were interpreted by an experienced reader (RD) who was blind to the cognitive and clinical diagnosis, using a methodology similar to that described by Seibyl and colleagues (2016). Images were displayed using a reader-adjustable gray scale to provide optimal discrimination of the cerebellar gray matter from white matter. Subsequently, all the amyloid PET scan slices were viewed using this gray scale adjustment. Tracer uptake was assessed in six cortical regions (orbitofrontal, frontal, parietal, lateral temporal, occipital and precuneus/posterior cingulate cortex, combining values from the left and right hemispheres) using the regional cortical tracer uptake (RCTU) system (Bullich et al., 2017). A final dichotomous (A+ versus A−) diagnosis was rendered. Loewenstein and colleagues (2018a) found extremely high agreement between the experienced reader and an independent rater in reading these scans.

2.7. Statistical Analyses

The SPSS Statistics software was used to analyze the data. All initial mean values were calculated using a series of one-way ANOVAs and chi- square analyses. Specific post-hoc comparisons involving 2 or more means were conducted using the Tukey’s Honestly Significant Difference (HSD) procedure for mean comparisons. A 3 × 2 repeated measures analysis with Diagnostic Group as the between subjects’ factor and Time as the within subjects factor was utilized to examine differences in changes on the LASSI-L variables over time. Our primary focus was on the Group × Time interaction term for which we have provided measures of effect size η²

Initial Cued Recall A2 was entered into the model as a covariate to test if the Group × Time interaction for Cued B2 over time remained statistically significant.

3. Results

As illustrated on Table 1 none of the three diagnostic groups differed significantly regarding demographic variables such as age, years of educational attainment, gender, or language of testing. The CU group had the highest MMSE scores, followed by the MCI (Amy−) group and the MCI (Amy+) groups respectively. All means were statistically different from each other using the Tukey’s Honestly Significant Difference Test (HSD). Adding the MMSE in subsequent analyses as a covariate did not alter the obtained results, so unadjusted scores are reported below. Also, as expected, the CU group had the lowest centiloid scores followed by the MCI (Amy−) and MCI (Amy+) groups, respectively. As indicated in Table 2, there are overall Group Effects on all LASSI-L subscales. However, in repeated measures analyses, it is the interaction term rather than main effect that is most important in evaluating changes in LASSI-L scores over time. Specifically, the Group × Time interaction term depicts the differences between the distinct diagnostic groups on the LASSI-L subscales comparing baseline and follow up evaluation. As depicted in Table 2, there were statistically significant differences in Group × Time interaction effects for LASSI-L Cued A2 (maximum learning) [F(2,83)=7.24; p<.001] and LASSI-L Cued B2 (frPSI) [F(2,83)=4.55; p=.013]. The effect sizes η² for these statistically significant interaction terms were eta squares of 14.8% and 9.9%, which represent large and medium to large effect sizes (See Cohen, 1988). For non-significant Group × Time Interactions for percentage of intrusion errors on Cued B1, η²=051 indicated a moderate effect size which may have reached statistical significance with an increased n. The effect sizes for Cued B1 recall and Cued B2 intrusions were η²=.004 and η²=<.0001 respectively, which are considered extremely small and trivial effect sizes with no practical importance.

Table 2.

LASSI-L Scores by Diagnostic Group and Evaluation Time Period

	CU Amy− (N=25)		MCI Amy+ (N=28)		MCI Amy− (N=33)		Group	Time	Group × Time F and η2
	Initial	Follow-up	Initial	Follow-up	Initial	Follow-up
LASSI-L	13.36	13.60	10.61	8.75	11.85	11.58	24.88	7.47	7.24
Cued A2 Recall	(SD=2.0)	(SD=1.3)	(SD=2.4)	(SD=2.9)	(SD=1.9)	(SD=2.5)	p<.001	p=.001	p=.001
(Maximum Learning; 0–15)									η2=.148
LASSI -L	8.20	8.48	4.61	5.36	5.52	6.12	20.66	2.92	.18
Cued B1 Recall	(SD=2.9)	(SD=2.3)	(SD=2.1)	(SD=2.6)	(SD=2.3)	(SD=1.6)	p<.001	p=.09	p=.84
(PSI; 0–15)									η2=.004
LASSI-L	11.56	12.08	8.82	7.11	9.55	9.21	21.09	2.94	4.55
Cued B2 Recall	(SD=2.4)	(SD=1.7)	(SD=2.6)	(SD=3.1)	(SD=2.5)	(SD=2.7)	p<.001	p=.09	p<.013
(frPSI; 0–15)									η2=.099
LASSI-L	.16	.10	.45	.33	.33	.18	22.38	46.31	2.45
Cued B1	(SD=.15)	(SD=.11)	(SD=.21)	(SD=.17)	(SD=.21)	(SD=.13)	p<.001	p<.001	p=.09
Percent Intrusion Errors (PIE)									η2=.051
LASSI-L	.10	.10	.31	.34	.18	.19	17.86	.50	.25
Cued B2 Percent Intrusion	(SD=.11)	(SD=.10)	(SD=.19)	(SD=.24)	(SD=.14)	(SD=.16)	p<.001	p=.48	p=.78
Errors (PIE)									η2=<.001

Note. Bolded values are statistically significant

Abbreviations: LASSI-L= The Loewenstein-Acevedo Scales for Semantic Interference & Learning; CU Amy− =Cognitively Unimpaired Amyloid negative; MCI Amy+ =Mild Cognitive Impairment Amyloid positive; MCI Amy−=Mild Cognitive Impairment Amyloid negative

Using Tukey HSD post-hoc mean comparisons, the MCI (Amy+) group evidenced a significant reduction in scores from baseline evaluation to follow up visit on Cued A2 and Cued B2 relative to the MCI (Amy−) or the CU comparison groups. Tables 3a and 3b provide 95th percentile confidence intervals for all pre-test and post-test means for all diagnostic groups.

Table 3a.

LASSI-L Cued A2 Recall Across Multiple Time Points

	Time	Mean (SE)	95% CI
CU Amy−	Baseline	13.36 (0.42)	12.52 – 14.20
	Follow-up	13.60 (0.47)	12.67 – 14.53
MCI Amy+	Baseline	10.61 (0.40)	9.81 – 11.40
	Follow-up	8.75 (0.44)	7.87 – 9.63
MCI Amy−	Baseline	11.85 (0.37)	11.12 – 12.58
	Follow-up	11.58 (0.41)	10.76 – 12.39

Abbreviations: LASSI-L=The Loewenstein-Acevedo Scales for Semantic Interference & Learning; CU Amy− =Cognitively Unimpaired Amyloid negative; MCI Amy+ =Mild Cognitive Impairment Amyloid positive; MCI Amy−=Mild Cognitive Impairment Amyloid negative)

Table 3b.

LASSI-L Cued B2 Recall Across Multiple Time Points

	Time	Mean (SE)	95% CI
CU Amy−	Baseline	11.56 (0.50)	10.56 – 12.56
	Follow-up	12.08 (0.52)	11.04 – 13.12
MCI Amy+	Baseline	8.82 (0.48)	7.88 – 9.77
	Follow-up	7.11(0.50)	6.12 – 8.09
MCI Amy−	Baseline	9.55 (0.44)	8.67 – 10.42
	Follow-up	9.21 (0.46)	8.31 – 10.12

Abbreviations: LASSI-L=The Loewenstein-Acevedo Scales for Semantic Interference & Learning; CU Amy− = Cognitively Unimpaired Amyloid negative; MCI Amy+ = Mild Cognitive Impairment Amyloid positive; MCI Amy− = Mild Cognitive Impairment Amyloid negative

Since these are both maximum learning measures for two different lists, follow-up analyses were conducted to determine if maximum learning on list A influenced learning on list B. When initial cued recall A2 (maximum learning at baseline) was entered into the model as a covariate in the Group × Time interaction for Cued B2 over time, the result remained statistically significant [F(2,82)=6.55; p<.002]. This indicates that the lack of learning on Cued B2 (frPSI) was not related to the initial strength of maximal learning for the original Target A list.

4. Discussion

The LASSI-L, particularly measures sensitive to frPSI have consistently been shown to successfully discriminate between cognitively unimpaired older adults and those with amnestic MCI in a wide array of investigations headed by different research groups in the United States, Europe and South America, has been shown to be more highly related to biomarkers of AD such as amyloid load and atrophy in AD prone regions than traditional memory measures as well as non-traditional memory measures such as the Free and Cued Selective Reminding test (Matias-Guiu et al., 2017; 2018; Sanchez et al, 2016; Loewenstein et al, 2017a,b; Loewenstein et al, 2016). More recently, Crocco et al (2021) found that baseline measures of the LASSI-L susceptible to PSI and frPSI identified those with PreMCI that that progressed to aMCI or dementia or returned to cognitively normal status over a 2–3 year period.

The present study represents the first attempt to evaluate performance of the LASSI-L itself over a 12-month period among three distinct operationally defined diagnostic groups: CU, MCI (Amy+), and MCI (Amy−). The study also evaluated if repeated evaluation would differ with regards to these different diagnostic groups over time.

Being able to track changes in cognition over time to estimate disease progression is crucial in both the clinical and the research settings. Accordingly, the current preliminary study is highly relevant to determining whether the LASSI-L could potentially be administered across multiple time points in order to track disease progression at a group level. We agree with the notion, that reliable change indices (RCI; See Duff, 2012) should be considered in evaluating clinically meaningful changes over time. However, in this preliminary study with a modest number of participants, this would require robust standard deviations that could be examined relative to mean difference that reflected ethnicity, sex and education to calculate meaningful reliable change scores. Thus, our current findings reflect statistically significant Diagnostic × Time effects, but meaningful change on the individual level awaits further study.

Results of this investigation provided further evidence that the LASSI-L is a valid instrument that could likely discriminate MCI individuals without AD pathology from those who are amyloid positive. There is increasing evidence that examining intrusions and other qualitative errors may be associated with increased underlying AD pathology over time (Thomas et al., 2018a). Other measures such as the Memory Binding Test (MBT) which is sensitive to the ability to bind semantically related material has shown potential relative to traditional neuropsychological measures (See Loewenstein et al., 2018b). The current investigation clearly showed that the MCI (Amy+) group performed significantly worse in relation to the MCI (Amy−) group and the CU group respectively, with regards to a reduction in LASSI-L performance over approximately a 12-month follow up period and obtained effect sizes were relatively large. The decline in LASSI-L subscales that best differentiated these groups included the Cued A2, which measures maximum learning capacity, and Cued B2 that taps into the fPSI. Performance on the LASSI-L for the MCI (Amy−) and CU groups remained relatively stable over a 12-month period. While practice effects have been routinely found using list-learning tasks when these are repeated among cognitively unimpaired individuals (Bartels et al., 2010; Goldberg et al., 2015; Woods et al., 2006), the LASSI-L is distinct than traditional memory tests due to inclusion of a second “to be remembered” list, as well as a large number of semantically related targets presented over two trial periods. Thus, the lack of obtained practice effect is likely related to a higher level of complexity in the task and there does not appear to be issues with regression to the mean or other artifacts. As the CU (Amy−) and the aMCI (Amy−) groups remained stable over time, this increases our confidence that reduction in scores over time in the MCI (Amy+) group is likely related to disease progression and increased pathology, though additional research with a more robust sample size and longer follow-up periods to study this finding in greater depth.

As previously noted, when initial Cued Recall A2 was entered into the model as a covariate, the Group × Time interaction for Cued B2 over time still remained statistically significant [F (2,82) = 6.55; p<.002], thereby indicating that the decline in Cued B2 recall was not related to a general learning ability. This is consistent with previous work by Curiel and colleagues (2018) in a cross-sectional study, which showed that a decline in learning on a list task susceptible to frPSI was not influenced by initial learning ability. In previous studies we have shown that Cued B2 recall on the LASSI-L is a sensitive marker of early AD, and that frPSI might constitute an early and unique cognitive deficiency of AD (Crocco et al., 2014; Loewenstein et al., 2018a; Loewenstein et al., 2018b). More specifically, this deficiency is posited to represent deficits in inhibitory systems and selective impairments in source memory, particularly given the strong relationship to amyloid load and volumetric loss in AD prone regions, such as, the hippocampi, entorhinal cortex, rostral frontal regions, inferior lateral ventricle, precuneus, inferior temporal lobules, temporal poles, superior parietal lobules, and posterior cingulate (Loewenstein et al., 2017a; Loewenstein et al., 2017b; Loewenstein et al., 2018b). The results of this study further support the notion that performance on scales tapping into frPSI may decline over time among individuals who have measurable amyloid brain pathology. It is important to note that Brooks and Loewenstein (2010), and Sanchez & colleagues (2017) first introduced the notion that early amyloidosis in AD likely does not have a direct effect on cognitive performance, but that downstream effects such synaptic disconnection between medial temporal, limbic and frontal lobe neocortical functions and frontal circuitry ubique may be causing LASSI-L deficits to AD and that these do not represent the same disconnection effects seen in strategic infarctions or conditions, such as, Parkinson’s Disease. Clearly future resting and task-based fMRI studies could provide clearer answers to these questions.

Strengths of this study include the use of carefully operationalized groups, as well as the inclusion of a CU group for comparison purposes. In this study amyloid positivity was established using expert visual reads, which is still considered the gold standard in the field (Haller et al., 2020). With regards to potential limitations, while the current results are promising and highly relevant, the study included a relatively modest sample that was followed over a relatively brief follow-up period. It is our intention to attempt to replicate these findings utilizing a larger sample size, a longer follow-up period, and to incorporate MRI data at baseline and follow-up periods, so as to correlate changes in performance over time with volumetric loss in AD prone regions. While the current study did not indicate differences among Spanish-speaking versus English-speaking participants, increasing the sample size could also provide a more through analyses of the effects of ethnic, linguistic, and cultural diversity. A curious finding was that correct responses rather than intrusion errors appeared to show decline in aMCI (Amy+) participants. As noted by Brooks and Loewenstein (2010), tests that are sensitive to AD may not be the best measures to estimate decline over time. Indeed, relative insensitive tests, such as, the MMSE (Folstein et al.,1975), Montreal Cognitive Assessment (MOCA;Nasreddine et al., 2005), CDR (Morris, 1997) sum of boxes and the Alzheimer’s Disease Cooperative Study ADL scale adapted for patients with MCI (ADCS-MCI-ADL) (Galasko et al., 1997) are much less sensitive to early AD deficits but are often used to track change over time. Correct responses generally have more restrictive ranges than intrusion errors which may in part have contributed to the obtained findings.

Finally, although these are preliminary findings, our data suggest that the LASSI-L may have the potential to be utilized at different time points to detect longitudinal cognitive change with minimal practice effects for older adults without significant brain amyloid deposition. More importantly, our findings also suggest that changes in performance across time relate to amyloid positivity in the brain, supporting the LASSI-L’s utility as a diagnostic tool. Previous work has shown that baseline LASSI-L performance on tests sensitive to frPSI could predict progression of preMCI, an intermediate cognitive state between normal cognition and MCI, to MCI or dementia over a 2 to 3-year period (Loewenstein et al., 2012; Crocco et al, 2021). Therefore, longitudinal LASSI-L performance in a preMCI cohort may be worthy of continued research.