Age-related episodic memory decline and the role of amyloid-β: a systematic review

ABSTRACT Aging has been associated with the functional decline of episodic memory (EM). Unanswered questions are whether the decline of EM occurs even during healthy aging and whether this decline is related to amyloid-β (Aβ) deposition in the hippocampus. Objective: The main purpose of this study was to investigate data on the relationship between the age-related EM decline and Aβ deposition. Methods: We searched the Cochrane, MEDLINE, Scopus, and Web of Science databases and reference lists of retrieved articles that were published in the past 10 years. The initial literature search identified 517 studies. After screening the title, abstract, key words, and reference lists, 56 studies met the inclusion criteria. Results: The overall results revealed that increases in Aβ are related to lower hippocampal volume and worse performance on EM tests. The results of this systematic review revealed that high levels of Aβ may be related to EM deficits and the progression to Alzheimer’s disease. Conclusions: We discussed the strengths and pitfalls of various tests and techniques used for investigating EM and Aβ deposition, methodological issues, and potential directions for future research.


INTRODUCTION
A ging has been associated with a functional decline in episodic memory (EM). An assessment of the memor y function in healthy aging can represent a meaningful alternative since Alzheimer's disease (AD) tends to be diagnosed at more advanced stages, especially when memory impairments appear. Approximately, 30% of elderly individuals with normal aging, above 60 years old, can have high levels of amyloid-β (Aβ) deposition in the hippocampus. 1 Hence, they tend to be more likely to develop AD. 2 Age-related memory decline, especially in those related to EM, is common in aging, and Aβ deposition has been associated with a progressive conversion to AD. 3 EM can be understood as the integration of the "what," "when," and "where" (or "which") components, which is a substantial factor in the social life of an individual. 4 However, it is unclear whether hippocampal volume atrophy and age-related EM decline in healthy aging are entirely related to the deposition of Aβ. 5 One of the reasons is because there are links between the neocortex and the hippocampus. In this way, it is essential to understand the relationship of the hippocampus with Aβ deposition to further unravel the other areas related to EM. 3 The hypothetical model of biomarkers in AD indicated that the presence of Aβ is related to the functioning of the vascular systems and growth. 6,7 This model describes the integration of AD biomarkers, which may reflect an underlying pathophysiological sequence of the following events: (1) the presence of amyloid-β42 in the cerebrospinal fluid (CSF) is first detected (the same is true for tau blood biomarkers); 8 (2) levels of tau protein in CSF are significantly increased; (3) hypometabolism of fluorodeoxyglucose occurs; (4) brain atrophy occurs; and (5) cognitive decline is noted.
The model of biomarkers in AD suggests that the relationship between Aβ and cognitive impairment is not immediately successive, and therefore it should be less evident than the relationship between neurodegenerative biomarkers and cognitive impairment. 8 Notwithstanding, the deposition of Aβ before the clinical diagnosis of AD remains a challenge, since less, if any, relationship between a decline in cognition in healthy aging and the deposition of Aβ in the hippocampus can be observed. In this manner, it is comprehensible that cortical thickness exerts a substantial influence on Aβ deposition and the consequent decline in EM. 9 Age-related dementia typically begins slowly and gradually with brain atrophy before the onset of clinical symptoms. 7,8,10 Some authors have suggested that the cognitive decline in aging is related to undetected diseases and, hence, is not a characteristic of normal aging. 11 An important implication of detectable brain atrophy and dementia is that some cases with undetected diseases can help create inferences about normal brain aging. 12 This is of particular concern, since the proportion of elderly people with undetected neurodegenerative disease is expected to increase, potentially leading to baseless conclusions of accelerated age-related decline in the cortical areas vulnerable to disease pathology, especially in the entorhinal cortex and the hippocampus. 13 Therefore, the contribution of latent pathology to age-related decline in healthy aging remains an open question. However, differences in AD biomarkers, such as brain atrophy and Aβ deposition, may clarify this issue. This review is based on the assumption that the presence of Aβ in people with healthy aging can serve as a prodromal state of AD. Thus, brain atrophy and Aβ deposition are biomarkers of the disorder. Here, our main purpose was to investigate the data on the relationship between EM and Aβ deposition in healthy aging.

Search strategy
The PRISMA guidelines were used. 14,15 Exhaustive electronic searches were conducted on the studies that were published from 2010 to 2021 in the following databases: COCHRANE, Medline, Scopus, and Web of Science (PROSPERO: CRD42020190981). The following search strategy was used: Aging OR Senescence OR Aged OR Elderly OR "Healthy Aging" OR "Aging Well" OR "Healthy Ageing" AND "Amyloid beta-Peptides" OR "Amyloid beta Peptides" OR "Alzheimer beta-Protein" OR "Alzheimer's ABP" OR "beta-Amyloid Protein" OR "Amyloid beta Protein" OR "Amyloid Protein A4" OR "beta Amyloid" OR "Amyloid AD-AP" OR "Amyloid" OR "Amyloid Substance" OR "Amyloid Fibril" OR β-amyloid OR amyloid-β OR "amyloid β-peptide" OR "amyloid-beta" AND "Histology" OR "Histocytochemistry" OR "Immunohistochemistry" OR "Immunolabeling Technique" OR "Immunolabeling Technic" OR "Immunogold Technique" OR "Immunohistocytochemistry" OR Positron-Emission Tomography" OR "Positron Emission Tomography" OR "Positron Emission Tomography Imaging" OR "PET Scan" OR "PET Imaging" AND "Episodic Memory" OR "Autobiographical Memory" OR "Prospective Memory." The key words were chosen even in the absence of a specific term (according to the MeSH) to prioritize sensitivity over the specified theme. In addition, we examined the reference lists in the retrieved studies.

Selection criteria
We included studies that investigated Aβ deposition in healthy aging and its relationship with decline in EM. [16][17][18][19][20] We adopted the following inclusion criteria: (a) we investigated Aβ in the hippocampus in individuals with healthy aging and (b) we used the tasks to assess EM. We excluded studies that (a) used animal models, (b) did not assess the hippocampus, (c) evaluated another type of memory instead of episodic, and (d) were literature reviews.

Data extraction
For each study, data were extracted independently by two authors (GS and JS) using a structured form. The discrepancies were resolved by consulting a third author (NA) if needed. If there was insufficient information in the studies, the respective author was contacted. The following variables were extracted: (1) demographic and clinical characteristics (e.g., number of patients); (2) study design; (3) characteristics of the techniques; (4) task for assessing EM; and (5) main findings.

Quality assessment
We performed individual and comprehensive quality assessments for each study. The studies were also evaluated based on internal validity (i.e., selection bias or attrition bias) and construct validity (i.e., adequacy of the operational criteria used). In general, the quality and evidence of the studies were assessed based on three main measures, namely, (a) limitations (e.g., poorly designed strategies), (b) consistency of the results, and (c) accuracy (i.e., ability to generalize findings and provide sufficient data). A quality assessment was conducted using the PEDro scale and the Appraisal Tool for Cross-Sectional Studies (AXIS).

RESULTS
The initial search of the databases identified 517 studies. After screening the title, abstracts, key words, and article references, a total of 56 studies were in compliance with the inclusion criteria. Figure 1 presents the diagram flow and the details used to identify studies in our review. General characteristics of the studies Table 1 presents the selected studies, which were published between 2010 and 2021. Thirty studies (53.6%) were published between 2010 and 2016, and the other 26 studies (46.4%) were published between 2017 and 2021. Regarding the sample size, the studies showed a variance between 45 and 2,908 participants, the majority (76.8%) being more than 100 participants.
A longitudinal study design was used in almost all of the selected studies (71.4%), whereas eight studies (28.6%) employed a cross-sectional design. Follow-up in these studies ranged between 6 months and 23 years. Diagnostic criteria also varied between studies, that is, 24 studies used the Mini-Mental State Examination

Techniques for investigating Aβ
We observed four types of techniques for investigating Aβ. The main technique used was positron emission tomography (PET), performed in 53 studies (94.6%). Three types of radioligands were used for the PET technique. Pittsburgh compound B was used in most of the studies (75%), while florbetapir F 18 and flutemetamol F 18 were used by a limited number of studies, i.e., nine (16%) and two (3.6%) studies, respectively. In addition, electrochemiluminescence techniques (in one study, 1.8%), CSF collection (in one study, 1.8%), and selected reaction monitoring (in one study, 1.8%) were also used.
Tasks used for the assessment of episodic memory Most studies (67.9%) utilized between two and four different neuropsychological tests to assess EM. Approximately, 25 of these studies (44.6%) concentrated their assessments based on two types of tests. Predominantly, most of them used neuropsychological measures, such as California Verbal Learning Test (CVLT; 35.7%), Rey Complex Figure Test (RCFT; 19.6%), Logical Memory Recall (17.8%), Rey Auditory Verbal Learning Test (16%), and the Wechsler Memory Scale-Revised (12.5%). Five studies did not specify the tests used in the neuropsychological assessment.

Main findings
The main findings are presented in Table 1. The scope of the selected studies included heterogeneous objectives. 21,22 Nevertheless, all the studies evaluated common aspects of Aβ deposition in cortical structures, especially in the hippocampus, and its relationship with EM. Hence, the results have specificities. In general, most studies (83.9%) indicated that the increase in Aβ was related to worse performance in EM tests. One of these studies even indicated that Aβ deposition in women makes them more vulnerable to declines in EM. 23 In contrast, some studies (16.1%) indicated that the increase in Aβ did not correspond to differences in EM.

Quality assessment
Based on the PEDro scale, which comprised 11 items, 56 studies obtained an overall average of 5.7 points, approaching the score considered by the moderate-to high-quality instrument (≥6.0 points -moderate to high quality). Overall, 53.6% were considered of moderate quality (4.0-5.0 points) and the other 46.4% were of moderate to high quality (6.0-9.0 points). Based on the AXIS, the overall average was 16.2 points on a scale ranging from 0 to 20 points. Overall, 28.6% (16 studies) scored from 0 to 15 points and the other 71.4% (40 studies) reached ≥15.0 points, showing a lower risk of bias and higher quality of the studies. Both the assessment tools showed good results with regard to the quality of the selected studies.
The discrepancies between the evaluations within these two scales may be related to the objective for which each one was designed. While AXIS provides an evaluation of more general items, 24 PEDro provides an assessment for more specific aspects in clinical trials, 25 so the lowest PEDro score, especially in items related to blinding, can signal possible biases, like exaggerating or reporting fewer symptoms. These biases can induce different rates of, for example, co-intervention, friction, and placebo effect. 26,27 Furthermore, our results suggest that future studies should exercise caution in the definition and methodological execution, so that the studies are in accordance with the guidelines of the designs to which they are proposed.

DISCUSSION
There are significant increases in research aimed at identifying diagnostic and prognostic biomarkers of AD. The updated AD research structure proposed by the National Institute on Aging and Alzheimer's Association working group defines AD as a biological construct, and the research focused on the diagnosis of AD in those who are alive using biomarkers that cover the presymptomatic and symptomatic stages of the disease. Biomarkers are grouped into those of Aβ deposition, pathological tau, and neurodegeneration. Although it is possible that Aβ plaques and neurofibrillary tau deposits are not the cause of AD pathogenesis, these abnormal protein deposits define AD as a unique neurodegenerative disease among various diseases that can lead to dementia.   The preclinical imaging methods used by diagnostic of amyloid accumulation and neurodegeneration (i.e., imaging PET) and biofluids (CSF and blood plasma) are very expensive and difficult to use in research. Therefore, the attention of researchers is directed to the search for predictors of the disease, which indirectly reflect the functional activity of the structures involved in the pathological process.
With regard to the prognostic value, the most relevant studies sought to identify biomarkers at early stages of AD pathogenesis, particularly the studies involving groups of healthy elderly individuals, characterized by the accumulation of cerebral amyloid plaques in the absence of clinical symptoms of mild cognitive impairment (MCI) or dementia. The hippocampus is a key brain region that processes EM and is a primary structure that is susceptible to the accumulation of amyloid plaques. However, some structures such as the entorhinal cortex and the cingulate gyrus are also relevant to the matter. Currently, some studies aimed at studying the relationship of minimal cognitive impairment, including the decline of EM, in different groups of subjects such as those with AD symptoms and/or MCI and the elderly with preservation of cognitive functions. [28][29][30] The drawback is that the methods and procedures used to investigate EM vary and do not always accurately reflect EM. It is also important to mention that visual processing 31,32 is one of the main biomarkers for cognitive decline. 33 An episode can be classified as "what" happens "where," with contextual information (temporal "when," or circumstantial "which") that fosters contextual and behavioral criteria. 34 As follows, the hippocampal formation is necessary for learning and memory, [35][36][37] particularly for spatial components. 38 The trigger between pyramidal neurons in the CA1 and CA3 regions is clearly correlated with the location of an individual. 38,39 The ablation of the hippocampal formation, in turn, impairs spatial navigation ability.
The tests such as CVLT and RCFT have considerable clinical predictive value and reliability and are widely used for the clinical assessment of EM. 16,17 However, classic neuropsychological tests are limited to evaluating the recall of focal elements, such as lists of images or words, freely or with the use of cues. 18,19 Thus, they are more associated with semantic or verbal processes than episodic processes, and they do not consider contextual elements. 19 Most studies that were discussed in this review attempted to circumvent this hindrance by employing at least two types of classic neuropsychological tests to assess EM. However, the set of tests that are used by these studies may not provide an assessment that fully integrates "what," "where," and "when" components that are essential for predicting real memory performance in everyday life. 19,20 . Additionally, using a battery of tests can be tiring for respondents, especially for the elderly people. The Treasure Hunt task 20 is proposed to investigate three components (i.e., what, where, and when) and be useful for further studies. This task was developed by Cheke 20 and is proposed to assess the memory of object information ("what"), location information ("where"), and temporal order information ("when") within the same testing paradigm. This is important because it integrates all three features into a classic what-where-when framework. This task can also identify the extent and pattern of EM deficits that might be present in several diseases. Although no normative data have been published for the Treasure Hunt task, its application can still be useful. It is difficult to identify whether there is a specific deficit of EM, and the Treasure Hunt task can be an important alternative or complement to the existing assessment tools.
The divergence between age and cognitive functioning may be grounded in the medial temporal lobe (MTL). The MTL, comprising interconnected structures of the hippocampus, dentate gyrus, peri-and entorhinal cortices, and parahippocampal gyrus, undergoes a prolonged period of postnatal development in humans, nonhuman primates, and rodents, with different maturation timelines. The MTL also impacts learning and memory functions differently in time. 40 These areas have functions in general mnemonic processes, specifically EM in humans and rats. It has been suggested that the function of the hippocampus is to integrate details of events that have been experienced, including spatial locations, together to gather episodic memories. 41 EM is a multisensory neurocognitive process of linking many elements. The unification of many elements in an episode occurs because of the long-term potentiation in the MTL and the activity of hippocampal neurons in the theta phase. 42,43 Studies have been demonstrated that theta cycles determined the process packaging of principal cell spiking into functional ensembles via the provision of discrete windows in which incoming streams of information from different systems are processed. 44,45 In addition, there is ample evidence that the N-methyl-d-aspartate glutamatergic receptor has a fundamental function in inducing synaptic plasticity and memory formation for various tasks involving aversive conditioning, 47 training of spatial memory, 48 nonspatial, and nonaversive tasks 49 in rodents 46 . These studies provided evidence that the deposition of neocortical and hippocampal Aβ in elderly people with normal cognitive aging is associated with functional changes in EM. 50 . In general, aging is associated with a reduced ability to modulate MTL activity, that is, with aging, the hippocampus shows a decrease in activation, and the entorhinal cortex decreases inhibition during an EM task. In addition, among elderly individuals, Aβ deposition was associated with a reduction in the entorhinal cortex regions associated with standard network functioning. 1,5,9 It is strongly suggested that in elderly individuals with a high concentration of Aβ, the preclinical processes of AD have begun despite normal cognitive functioning, even in the absence of changes in clinical findings. It is important to mention that this can be detected over a short period of time through the use of neuropsychological measures. 51 Age and Aβ deposition contribute to a collapse of the network between the hippocampus and regions of the neocortex, suggesting the declines in EM.
Aging-related dementia usually begins gradually, with hippocampal atrophy manifesting several years after the onset of clinical symptoms. Aβ deposition is a part of the pathophysiology of AD, and therefore Aβ is more focused as a biomarker of AD. It is known that the role of Aβ in neurodegeneration culminates in a cascade of harmful events such as dementia and AD. However, this review noted that even in the presence of cognitive aging, Aβ can be detected in some individuals. The factors or issues that make individuals to be at risk of developing AD are still unclear. 51,52 One key to improving our understanding of the relationship between normal aging and initial stages of AD is related to neuroplasticity and cognitive decline that results from a lack of a compensatory response to the accumulation of Aβ. The leading genetic risk factor for AD, namely, apolipoprotein ε4, is related to neural plasticity. The high levels of cognitive reserves are associated with the level of education and socioeconomic status of an individual and delay in the diagnosis of AD 53 . Individuals with greater cognitive reserves can maintain cognitive function in the face of the higher levels of cerebral Aβ deposition 54 . Weak correlations between the levels of Aβ deposition and cognitive function suggest that other mechanisms, such as functional compensation, influence cognitive ability. The views that cognitive function can be maintained during aging by compensatory cognitive processes and that this decline is seen when a person is no longer able to compensate for a decrease in the function of primary brain structures and circuits are well supported in the contemporary research. 55,56 High-performing elderly adults are quite interesting in this regard. High-performing elderly individuals exhibit global preservation of the cerebral cortex, especially the anterior cingulate gyrus, and the volume of the hippocampus is higher than in people of normal age. The histological analyses of this group also revealed lower amyloid burden and lower neurofibrillary tangles compared with cognitively normal elderly controls. Thus, further studies of high-performing older adults are likely to provide additional insights in the role of amyloid deposition in the hippocampus during the aging process. 57 Despite the strengths of this review, it has some limitations. First, this review included studies that involved age ranges, which created challenges in interpreting these results (e.g., age is directly associated with amyloid deposition). Second, several studies used tests that assessed only specific parts of EM (e.g., delayed recall), thus raising the need to use extensive neuropsychological assessments or tasks, such as the Treasure Hunt task, that encompass all three main aspects of EM. Third, the analysis of various brain regions raises the need to define possible neural networks that are responsible for EM processing.
Finally, we noted the need for further research in this area. It is necessary to investigate this concept in order to understand the prodromal symptoms of AD and the emergence of new practices and techniques that can identify and map Aβ deposition during healthy aging.