Face-Name Associative Recognition Deficits in Subjective Cognitive Decline and Mild Cognitive Impairment

INTRODUCTION

The pathology of Alzheimer’s disease (AD) is manifold and occurs years to decades before the first manifestation of cognitive symptoms [1, 2]. Current research tries to identify and characterize subjects at increased risk for progression to dementia at the preclinical stage of the disease with the aim to target early intervention and to provide an adequate treatment. Preclinical AD is considered as the first stage of AD, initially characterized by an asymptomatic accumulation of amyloid. Further progression of the disease is accompanied by neurodegeneration. Finally, at the transitional stage between preclinical AD and mild cognitive impairment (MCI) due to AD, subtle cognitive impairments on tests with high sensitivity to AD pathology, and self-experienced cognitive decline may arise [2].

Subjective cognitive decline (SCD) may thus represent the first symptomatic manifestation of AD and precede MCI [3]. This assumption is supported by studies which highlighted SCD as a risk factor for future cognitive decline and progression to MCI and AD dementia [4–8].

Several studies have shown that memory clinic patients with SCD, who were unimpaired in traditional neuropsychological tests, exhibit signs of neurodegeneration and dysfunction, attributed to putative pathological processes underlying AD, e.g., gray matter atrophy in the hippocampus and entorhinal cortex [9–12]. Furthermore, a decreased activation of the right hippocampus as well as an increased activation of the right dorsolateral prefrontal cortex during performance on an associative memory test has been reported [13]. Moreover, a study including healthy older individuals with SCD and applying positron emission tomography with Pittsburg Compound B (PiB-PET) showed evidence of brain β-amyloidosis in these individuals compared to healthy controls (HC) without SCD [14]. Subtle decline in cognition at the end of the preclinical stage of AD is, by definition, difficult to detect with standardized neuropsychological tests [3]. Only few studies report slight cognitive impairment on established [15, 14] or novel [16] cognitive tests in patients or individuals with SCD.

Some recently developed tests exhibit increased sensitivity for cognitive deficits in preclinical AD and have been associated with AD pathology [17]. For instance, in healthy individuals an association between a face-name associative memory test and amyloid deposition was found in brain regions underlying memory [18]. Delayed associative face-name recall was found to be impaired in amnestic MCI patients and was the memory test that best predicted conversion from MCI to AD within two years [19].

Face-name association has obvious “face validity”, because it relates to a common everyday cognitive complaint of elderly individuals. However, other associative memory tests may also be useful for early detection of AD. In particular, the CANTAB Paired Associates Learning (CANTAB-PAL) [20] is a widely employed visual-spatial paired associate learning test, which discriminated between non-demented memory clinic patients with or without subsequent progression to AD dementia [21], and in one study was sensitive to progressive memory decline in cognitively normal subjects with memory complaints [22]. It remains to be determined whether different associative tasks yield similar results or whether the type of items to be associated matters. Face-name association requires the association of verbal with biologically meaningful visual face stimuli, which might tap into different brain areas than a visual-spatial task with abstract geometricfigures.

Refinement and further validation of tasks with increased sensitivity to changes occurring early in the course of AD will be a valuable enhancement of the neuropsychological toolbox for clinical studies.

The purpose of the present study was to investigate the sensitivity of a face-name associative memory test in memory clinic patients with SCD and MCI, and to examine associations with self-reported cognitive problems in everyday life. Inspired by the similar Face Name Associative Memory Exam (FNAME) [18], we developed a computerized paired-associate recognition test which also requires pairing of faces and names during learning (FNART; Face-Name Associative Recognition Test). However, the FNART, in contrast to the FNAME and other similar tests [23, 24], does not employ free recall of the name which was learned to go with a face. Rather, the FNART only requires a forced choice about which among several presented names belongs to the face shown. This response format is identical to the format used in previous functional imaging studies of face-name association learning, which have consistently shown stronger hippocampal activation during successful learning of face-name associations, which were subsequently recognized outside the scanner [25, 26]. Interestingly, some of these studies have also found evidence of memory-related activity in entorhinal [27] and parahippocampal [25, 28] cortex, regions which in AD are affected by tau-pathology even earlier than the hippocampus [29]. The associative recognition format has also the practical advantage of being easier, both subjectively and objectively, than recalling a name when shown the face alone. This might avoid floor effects when studying subjects with mild cognitive deficits. Thus, such a test can be administered to both normal and subjects with MCI, which is harder to do with free recall paradigms. In addition, the response format may also reduce feelings of threat and frustration in subjects who are concerned about their memory.

The computerized presentation of the FNART also allowed assurance of a minimal exposure time of the face-name associations in order to control depth of encoding, and enables the exploration of decision times during recognition, which may yield additional diagnostic information.

The main objective of the current study was to examine whether the FNART would discriminate not only between MCI and HC, but also between SCD and HC.

We also wanted to examine whether specific complaints about cognitive functions, at least in the SCD group, might be related with, and possibly reflect, objective memory function. We assessed complaints about everyday cognitive functioning by the self-rating version of the Everyday Cognition (ECog) questionnaire.

Furthermore, we explore whether performance on FNART is related to performance on CANTAB-PAL, which requires the recall of previously learned object-position associations.

We hypothesized that the FNART and the CANTAB-PAL would be moderately correlated, as both are paired associate tasks which, however, differ in item content. The number of subjects with data on CANTAB-PAL performance is smaller because this test became available to us only after the study of the FNART had started.

MATERIALS AND METHODS

In this cross-sectional study, patients were recruited within the outpatient clinic at the Clinical Research Center for Neurodegenerative Diseases (KBFZ) at the University Hospital Bonn. HC were acquired using a press release. This sample does not overlap with previous cohorts from our research group [9, 16].

The study was approved by the ethics committee of the University of Bonn (Medical Faculty). Participants were informed about the study procedure and gave their written informed consent before being enrolled.

RESULTS

Descriptive statistics of the three groups are listed in Table 1. The groups did not differ significantly with regard to age (F(2,86) = 0.056, p = 0.946), years of education (F(2,86) = 0.181, p = 0.835) or distribution of gender (χ²(2, N = 89) = 1.632, p = 0.442). However, the GDS score did differ significantly between the groups (F(2,82) = 10.705, p < 0.001), such that SCD and MCI patients had higher scoresthan HC.

As expected by group definition, comparisons including the CERAD total score indicated a significantly lower overall performance in the MCI group (M = 73.79, SD = 10.61) compared to HC (M = 89.57, SD = 5.72, p < 0.000) and SCD. However, HC and SCD (M = 87.06, SD = 6.18) did not differ significantly (p = 0.651).

DISCUSSION

The aim of the present study was to investigate whether associative recognition memory is a sensitive measure for detection of subtle memory impairment. We found that associative recognition on the demanding FNART is not only impaired in MCI but also in individuals with SCD who performed normal on tests of an established neuropsychological battery.

Our finding of a slowing in memory RT in MCI cases is consistent with other findings previously reported for AD cases [49, 50] and could be indicative for a general cognitive slowing associated with the disease, memory search and the absence of confidence.

While RT for correct choices was faster in HC (in line with previous findings [51]) and SCD, RT for correct and incorrect decisions did not differ in MCI cases. The difference in RT was negatively related to recognition accuracy on FNART. This suggests that the ease of successful retrieval is reflected in both, better and faster recognition. In subjects with MCI, accuracy (58%) was only slightly better than chance (50%), so that their decisions were largely based on guessing rather than on associative retrieval. Whether or not the RT differences in forced-choice memory paradigms will yield additional diagnostic information beyond accuracy is an interesting issue in need of further research.

In a reduced sample we observed similar results regarding FNART performance across the three groups. In addition, performance on the short visual memory test CANTAB-PAL, which requires the recognition of previously learned association between an abstract (nonverbal) object and a spatial location, was impaired in MCI— as previously shown [52, 53]— but not in SCD. Impairment in associative memory performance on CANTAB-PAL in aMCI cases was shown to be determined by a decrease of short-term memory capacity, difficulties with memory consolidation and less effective executive functioning as, e.g., reduced responsemonitoring [53]. Thus, preserved executive functioning in SCD cases may account for the normal recognition performance on CANTAB-PAL. Because of the small sample sizes and the limited power to discriminate SCD from HC with the CANTAB-PAL, the dissociation between the two paired associate tasks regarding their sensitivity to SCD might be more apparent than real. However, if substantiated with larger samples, such a dissociation could result from face-name association being more reliant on extrahippocampal brain regions [25] where tau pathology occurs early in the course of AD.

Although the association of both tests were substantial (r = –0.547) as expected given both are associative learning tests, the above mentioned specifics as, e.g., abstract visual-spatial versus ecologically valid face-verbal associations and differential involvement of executive functions may account for the fact that both still share less than 30% variance.

Within the MCI sample we observed a floor effect in the FNART performance indicated by near chance level (50%). This effect is due to the relatively high task difficulty of the FNART. Therefore, the use of this test could be limited regarding follow-up examinations within MCI and AD samples and the discrimination between MCI and more impaired AD patients. On the other hand, it is advantageous for the detection of subtle cognitive impairment in individuals with SCD and preclinical AD, i.e., our primary target population.

There is accumulating evidence showing that memory clinic patients with self-experienced cognitive decline and normal test performance have increased risk for abnormal levels of AD biomarkers of amyloidosis and tau-mediated neurodegeneration and dysfunction [9, 13]. Furthermore, biomarker evidence is relevant for clinical decision making and prediction of future cognitive decline and is of particular importance for the characterization of preclinical AD [3]. In this study, information on CSF biomarker status was available only in a minor number of patients. Biomarker positivity versus negativity might account for the observed variability in test performance especially in SCD cases, beside the extent of cognitive reserve and coexisting pathologies or diseases. Further studies incorporating data on AD biomarkers will clarify whether the FNART can discriminate between SCD/MCI cases due to AD and SCD/MCI due to other causes.

Self-reported cognitive function in daily life is associated cross-sectionally [54, 55] and longitudinally [56] with episodic memory and executive functioning. In our study, we assessed function in daily life with the self-rating instrument “Everyday Cognition” (ECog-39). In the current study, patients with MCI and SCD complained particularly about difficulties in everyday memory. All patients were recruited in a memory clinic, and memory complaints are the most prominent complaint in these settings, and part of the definition of MCI and most SCD criteria. Thus it is not a surprise to us that patients differ from controls on the memory subscale, but that this is a rather specific complaint not evident for the other functions assessed with the ECog. Interestingly, self-reported memory difficulties were significantly related to FNART performance in the SCD group, but not in MCI. In the subgroup analysis, self-reported memory difficulties tended toward significance with CANTAB-PAL performance for SCD but not MCI. This suggests that some subjects with SCD correctly sense and report minor memory deficits, which can be unmasked only with challenging tests like the FNART or the CANTAB-PAL. No such relationships were found in the MCI sample. This might be due to a reduced awareness of cognitive deficits and their impact on daily activities/everyday function in some MCI cases, consistent with previous studies [57–59]. Underestimation of functional deficits in MCI has been associated with a CSF biomarker profile indicative of AD and progression to AD [57, 58].

As also reported in previous studies [57], functional deficits in our MCI and SCD cases were linked to depressive symptoms.

In sum, our results provide further evidence that associative memory assessment is not only sensitive for age-related cognitive decline in episodic memory function in healthy elderly (Associative Deficit Hypothesis [60]) but also for detecting subtle cognitive decline that might be attributed to AD pathology,

Thus, associative memory assessment seems to be suitable for the investigation of subtle cognitive deficits in preclinical AD. Most assessment batteries in this field rely on non-associative item memory, the inclusion of a short associative memory test might be a useful complement. Several studies found associative memory to be more impaired than item-memory in healthy elderly [61] and MCI [62] and our findings would be consistent with an early sensitivity of associative memory. However, Didic et al. hypothesized that item memory but not associative memory could be affected already in the “clinically silent” transentorhinal stage of AD [63]. This interesting conjecture will need to be tested with psychometrically matched item- and associative memory tasks in biomarker-based studies of the initial stages of preclinical AD.

A number of novel test developments have recently been reviewed by Rentz et al. [17]. These included, for example, a face-name association tests (FNAME [18]), but also a short-term memory binding test [64], and variants of word list learning tasks(Memory Capacity Test from H. Buschke). Furthermore, composites of traditional, challenging tests (e.g., Preclinical Alzheimer Cognitive Composite [65]) with proposed sensitivity to early AD related changes have been developed. At present, there are yet not enough comparative data in the ability of these tests to identify biomarker-defined AD in the preclinical or prodromal stages of the disease.

In summary, findings of this study suggest that associative recognition is a sensitive neuropsychological measure for detecting subtle cognitive impairment in memory clinic subjects with SCD, a group known to be at increased risk for having preclinical AD. Thus, the FNART, together with other recently developed tests [17, 16], could be a useful supplement to established test batteries for observational and intervention studies in memory clinic patients.

Further longitudinal and biomarker studies should examine the predictive validity of the FNART and aim to identify factors that influence the variability in associative memory accuracy in MCI and especially in SCD cases.