Refining memory assessment of elderly people with cognitive impairment: Insights from the short-term memory binding test

Highlights

• Short-term memory binding (STMB) can inform about the predementia stages of AD.

• Tailoring the test difficulty to the disease severity helps trace the AD continuum.

• STMB and hippocampal volume identify MCI patients who may be on the AD trajectory.

Abstract

Alzheimer's disease (AD) affects temporary memory for bound features more remarkably than for individual features. Such selective impairments manifest from presymptomatic through dementia stages via titration procedures. A recent study suggested that without titration and with high memory load the binding selectivity may disappear in people at risk of AD such as those with Mild Cognitive Impairment (MCI). We compared data from two studies on temporary binding which assessed people with MCI and controls using different memory loads (2 or 3 items). Selective binding impairments were found in MCI, but relative to controls, such selectivity was contingent upon memory load (i.e., present with 2 items). Further analysis with MCI people who tested positive to neuroimaging biomarkers (i.e., hippocampal atrophy) confirmed that this specific binding impairments are a feature of prodromal AD. The temporary binding task has been recently suggested by consensus papers as a potential screening tool for AD. The results presented here inform on task properties that can maximize the reliability of this new assessment tool for the detection of memory impairments in prodromal cases of AD.

Introduction

Memory assessment in individuals at risk of Alzheimer's disease (AD), such as those with Mild Cognitive Impairment (MCI), has long focused on episodic memory functions (Fields, Ferman, Boeve, & Smith, 2011; Parra, Abrahams, Logie, & Della Sala, 2010; Parra et al., 2011). Examples are Paired Associates Learning (PAL) tasks (Sahakian et al., 1988), the Face Name Associative Memory Exam (FNAME) (Amariglio et al., 2012, Rentz et al., 2013), the Free and Cued Selective Reminding test (FCSRT) (Barbeau et al., 2004; Buschke, 2014; Grober, Buschke, Crystal, Bang, & Dresner, 1988; Sarazin et al., 2007), and other episodic memory tests (Ivanoiu et al., 2005). These tests are known to assess functions of the hippocampus which are essential to episodic memory formation i.e., associative memory (Tulving, 2002). Tests assessing associative memory functions of the hippocampus are considered markers for AD (Auriacombe et al., 2010, Barbeau et al., 2008, Dubois et al., 2010, Rentz et al., 2013, Sarazin et al., 2007). To uphold the claim that the associative function is that selectively impaired in AD, it is necessary to demonstrate that such impairments are greater than those found when patients remember the individual items. For instance, memory for faces (Sperling et al., 2003), lists of words (Gallo, Sullivan, Daffner, Schacter, & Budson, 2004), or locations (Stehli, Chubb, & Jacob, 2003), are functions affected by AD. This makes it difficult to ascertain that holding associations between these items in memory (e.g., faces and locations, faces and names) is the hallmark of AD. This is important because item memory and associative memory dissociate (Chalfonte, Verfaellie, Johnson, & Reiss, 1996; Old & Naveh-Benjamin, 2008) and the form of representation claimed to be specifically affected by the hippocampal amnesia of AD is the latter. This caveat i.e., limited underlying constructs, has been recently highlighted by a recent consensus paper (Costa et al., 2017).

The Visual Short-Term Memory Binding Test (VSTMBT) was developed to investigate if the function responsible for binding features within object representations is affected by AD above and beyond that supporting single feature processing (Parra, Abrahams, Logie, Mendez, et al., 2010). The test assesses participants’ memory for single features such as shapes and for combination of features such as shape-color bindings. When memory load is controlled for (i.e., via titration to keep patients’ and controls’ memory performance for individual features at the same level), patients with AD show memory binding deficits which are far greater than those found when memory for single features is assessed (Della Sala, Parra, Fabi, Luzzi, & Abrahams, 2012; Parra et al., 2009; Parra, Abrahams, Logie, & Della Sala, 2010; Parra, Abrahams, Logie, Mendez, et al., 2010; Parra et al., 2011). Such specific increase of the cost of binding has been observed since the preclinical stages of AD. This fits well current trends in the assessment of AD which have shifted toward a new lexicon (Costa et al., 2017, Dubois et al., 2010, Dubois et al., 2016) that encourages the detection of subtle cognitive impairments in stages prior to dementia. The VSTMBT detects such early impairments, even when other novel and traditional tests have failed (Parra, Abrahams, Logie, Mendez, et al., 2010; Parra et al., 2011).

The results form a recent study (Koppara et al., 2015) suggest that memory load may be a factor precluding the specificity of the VSTMBT (i.e., greater cost of binding in patients than in controls). Previous studies have manipulated memory load by presenting patients and controls with a different number of to-be-remembered items (Della Sala, Kozlova, Stamate, & Parra, 2018; Della Sala et al., 2012; Parra et al., 2009; Parra, Abrahams, Logie, & Della Sala, 2010; Parra, Abrahams, Logie, Mendez, et al., 2010; Parra et al., 2011). Such manipulation rested on the assumption that VSTM stores a limited number of items (Luck & Vogel, 1997; Vogel, Woodman, & Luck, 2001) and that increasing the number of items above such a limit (i.e., 4) would overload memory, rendering the task more challenging and performance poorer. Titration aimed at reducing differences at baseline (i.e., memory for single features). This led to suggesting that patients with AD present with a selective deficit of VSTMB. (Koppara et al., 2015) showed that without titration (i.e., patients and controls tested with the same visual arrays of 3 items), the selectivity of the VSTMBT holds for people with Subjective Cognitive Deficits (SCD) but not for people with MCI. Considering that memory binding is maintained to be selectively impaired in AD and that MCI is an uncertain clinical category which holds limited value to predict future risk of dementia, it is important to demonstrate the precise testing conditions with which selective impairments of VSTMB can be found. Is the specific impaired ability to binding features in VSTM that has been considered a hallmark of AD. Hence, identifying such a hallmark in MCI people might provide more reliable evidence of AD pathology as the likely underlying mechanism. To address these outstanding issues, in the present paper we present data from groups of healthy older adults and people with MCI who were assessed with the VSTMBT using arrays of 2 and 3 items and without titration. If the above-mentioned selectivity is contingent upon memory load, it would be observed only under the low memory load condition (i.e., 2 items).