Episodic and Semantic Autobiographical Memory in Mild Cognitive Impairment (MCI): A Systematic Review

Introduction: Mild cognitive impairment (MCI) is a syndrome defined as a decline in cognitive performance greater than expected for an individual according to age and education level, not interfering notably with daily life activities. Many studies have focused on the memory domain in the analysis of MCI and more severe cases of dementia. One specific memory system is represented by autobiographical memory (AM), which has been largely studied in Alzheimer’s disease and its effect on AM; however, the impairment of AM in moderate forms of decline, such as MCI, is still controversial. Objective: The main aim of this systematic review is to analyze the functioning of autobiographical memory in patients with MCI, considering both the semantic and the episodic components. Materials: The review process was conducted according to the PRISMA statement. The search was conducted until 20 February 2023 in the following bibliographical databases: PubMed, Web of Science, Scopus, and PsycInfo, and twenty-one articles were included. Results: The results highlight controversial findings concerning the semantic component of AM since only seven studies have found a worse semantic AM performance in patients with MCI compared to the HC group. The results of impaired episodic AM in individuals with MCI are more consistent than those concerning semantic AM. Conclusions: Starting from the evidence of this systematic review, further studies should detect and investigate the cognitive and emotional mechanisms that undermine AM performance, allowing the development of specific interventions targeting these mechanisms.


Introduction
The term mild cognitive impairment (MCI) indicates a syndrome conceptualized as a decline in cognition more severe than expected for an individual according to age and schooling. This impairment does not notably interfere with daily activities [1]. Since its first proposal [2], the MCI construct has been evolving: it was originally used to describe a pattern of impairments involving the memory domain, while afterward, it broadened to embrace many other domains [3]. Petersen [4] hypothesized a categorization into multiple subtypes. This hypothesis distinguishes MCI into (a) amnestic MCI single domain, in which there is an impairment in the memory domain only (aMCIsd); (b) amnestic MCI multiple domains, characterized by impairments in memory and other cognitive domains (aMCImd); (c) non-amnestic MCI single domain, marked by an impairment in one domain that is not memory (naMCIsd); and, finally, (d) non-amnestic MCI multiple domains, characterized by at least two impairments in two cognitive domains other than memory (naMCImd). These subtypes have different development pathways [5], and individuals with aMCI convert more frequently into Alzheimer's disease (AD) [5]. Therefore, many studies have focused on the memory domain since it remains relevant in the analysis and outcome of MCI, as in more severe cases of dementia. One specific memory system is represented by autobiographical memory (AM), which is important for social functioning [6] since it gives people the sense of a subjective timeline. Thanks to AM, individuals can mentally travel back, acquiring a sense of "self" that can exist in the subjective time [7]. AM goes beyond the mere recall of past events: it creates a sense of extended self through time in order to reflect and evaluate events related to the self [8]. Therefore, AM represents a special form of memory that regards the individual's life experiences, and it comprises multiple forms of long-term memory [9], including episodic and semantic components [10]. Autobiographical episodic memory refers to the recall of specific episodes from one's past, such as an unexpected accident. On the other hand, autobiographical semantic memory refers to the general knowledge about the self, such as the name of the street in which one lived as a child. Investigating the functioning of autobiographical memory and its different phenomenological aspects in both physiological and pathological decline occurring with aging is a challenge of the current study [11], relevant for its role in integrating, interpreting, and evaluating past events and self [12].
In healthy aging, it has been demonstrated that AM declines with aging and that the episodic component is more affected than the semantic one [11]. Piolino et al. [11] found that episodic AM decline becomes apparent after 60 years. In pathological aging, while the impact of Alzheimer's disease on AM is largely studied [13,14], it is still controversial in moderate forms of decline, such as MCI. Recent studies have evidenced episodic AM impairments in patients with MCI [15,16]. While recalling past events, these studies found that patients with aMCI had a worse performance in episodic memory but recalled more semantic details. Patients with aMCI typically show a malfunctioning of the hippocampus [17], which could cause an isolated impairment of episodic memory [15]. However, there is still a lack of clear evidence about the effect of MCI on AM and its components.
Therefore, the main objective of this systematic review is to analyze the functioning of autobiographical memory in patients with MCI, considering both the semantic and the episodic components.

Research Questions
This systematic review aims to answer the following research questions:

1.
Is autobiographical memory impaired, and to what degree in patients with MCI? 2.
Which components of AM are impaired and how are they impaired in MCI?

Search Strategy
The present systematic review adhered to the guidelines developed by the PRISMA-Statement [18,19] and was registered on the Open Science Framework (https://osf.io/dn2 5x, accessed on 10 November 2022). The search was conducted until 20 February 2023 in the following bibliographical databases: PubMed, Web of Science, Scopus, and PsycInfo. The review is based only on English, Italian, French, and Spanish articles. The search syntax can be found in Table 1.

Inclusion/Exclusion Criteria
According to the objectives of this review, the following inclusion criteria have been adopted: (I) randomized cross-sectional or longitudinal studies that aim at evaluating autobiographical memory in patients with MCI; (II) articles that specify the instruments for evaluating autobiographical memory; (III) studies that include a healthy control group; (IV) studies that clearly specify which criteria have been used for diagnosing MCI; (V) studies that analyzed differences in autobiographical memory performance between the different subtypes of MCI.
Specific exclusion criteria were: (I) studies that included clinical patients (people with neurological conditions, with metabolic or autoimmune diseases, with cardiovascular or oncological problems, or with diagnoses of dementia); (II) studies that included other diagnoses of cognitive impairment such as CIND, AAMI, or AACD; (III) studies that measured autobiographical memory only with the use of neurophysiological measures; (IV) gray literature (pre-print papers that have not undergone the peer-review process, Ph.D. dissertations).
Two researchers independently performed a screening of the selected articles. By screening titles and abstracts, non-relevant articles were excluded, which allowed including 70 studies. Afterward, the reading of the full texts resulted in 21 retained articles. This process is described in Figure 1.

Studies' Risk of Bias
The present systematic review adopted the Cochrane Collaboration's tool for assessing the risk of bias [20]. For this systematic review, we considered the following risk of biases: (i) Attrition bias (which can be defined as the bias consequent of the presence of incomplete outcome data); (ii) Reporting bias (i.e., the bias resulting from the reporting of selective outcomes or the absence of reporting relevant results; reporting bias was considered low risk if all prespecified outcomes were reported, as suggested by Higgins et al. [20]).
As a complement to these types of biases, two other biases were included: (i) Sample bias (i.e., bias resulting in samples that do not represent the general

Studies' Risk of Bias
The present systematic review adopted the Cochrane Collaboration's tool for assessing the risk of bias [20]. For this systematic review, we considered the following risk of biases: (i) Attrition bias (which can be defined as the bias consequent of the presence of incomplete outcome data); (ii) Reporting bias (i.e., the bias resulting from the reporting of selective outcomes or the absence of reporting relevant results; reporting bias was considered low risk if all prespecified outcomes were reported, as suggested by Higgins et al. [20]).
As a complement to these types of biases, two other biases were included: (i) Sample bias (i.e., bias resulting in samples that do not represent the general population, undermining the generalization of results, or lack of demographic information, such as female ratio, mean age, or mean schooling years); (ii) Measurement bias (i.e., bias due to using non-validated tasks to measure autobiographical memory).
A summary of the risk of bias in reviewed studies can be found in Table 2. Nine studies [16,[21][22][23][24][25][26][27][28] did not meet any of the considered bias. Attrition and reporting biases posed low risks in all of the included studies. Instead, sample bias risk was high in six articles [15,[29][30][31][32][33], due to the lack of demographic information (e.g., female ratio, mean age, mean years of schooling). Moreover, measurement bias risk was high in eight studies  due to the use of non-validated or ad hoc tasks to measure autobiographical memory.  [28] − − − − "+" high risk of bias; "−" low risk of bias.

Overview
The twenty-one articles included in this systematic review involved 1046 participants. Among these, 462 were included in the MCI group, while 584 were included in the healthy control group. The female percentage ranged between 23% and 79% in the MCI group and between 44% and 86% in the HC group. The mean age ranged between 63.13 years (SD = 5.78) and 81.8 (SD = 7.8) years in the MCI group and between 62.94 (SD = 5.73) years and 78.35 (SD = 5.75) years in the healthy controls. The mean years of schooling varied from 7.6 (SD = 2.6) to 16.68 (SD = 3.96) in the MCI group and from 7.9 (SD = 2.5) to 16.06 (SD = 2.80). Five articles did not report years of education [29][30][31][32][33]. Moreover, all the studies have been conducted in Europe, except for six articles that have been placed either in Canada [16,21,27,36] or in Australia [29,30]. These characteristics are summarized in Table 3.

MCI Diagnosis
In this systematic review, 20 studies out of 21 based their diagnosis of MCI on Petersen's criteria. Only the study of Irish et al. [23] referred to Winblad's criteria. These criteria are more thoroughly described in Table 4.
Twenty studies out of twenty-one focused on one specific subtype of MCI, that is, the amnestic subtype. Since the studies included in this systematic review intended to focus on autobiographical memory impairments, most articles have not considered the nonamnestic MCI subtype. Only the study by Davidson et al. [36] has not specified whether they included patients with MCI in general or if they focused only on the amnestic subtype.
In Table 5, it is possible to observe the assessed cognitive domains and the neuropsychological tests that the authors have utilized in their studies.

Episodic and Semantic Autobiographical Memory
All the studies included in this systematic review report an impairment in the episodic component of AM in patients with MCI compared to healthy controls.
The situation is more controversial for what concerns the semantic component. Fourteen studies out of twenty-one have studied this aspect, while seven articles [26,32,[34][35][36][37][38] focused only on the episodic component. Among the 14 articles that analyzed autobiographical semantic memory, 4 of them [21,24,29,33] have not found any significant differences in semantic memory scores between patients with MCI and healthy controls. On the other hand, seven studies [22,23,25,28,30,31,39] have found a worse semantic performance in patients with MCI when compared to the HC group. The other three studies [15,16,27] found more semantic details in patients' recollections than controls.

Internal and External Details
In five out of twenty-one studies [15,16,21,27,34], autobiographical memory was assessed following the protocol that was developed and standardized by Levine et al. [40]. According to this procedure, the autobiographical memories were transcribed and segmented in order to distinguish external and internal details. Internal details corresponded to episodic memory since they reflected information regarding the main event. On the other hand, external details were not specific to the main episode and were scored as semantic memory (concerning general knowledge of facts or events related to the self).
Three [15,16,27] out of the five articles that used this method found that controls produced more internal details (episodic memories) than patients with aMCI, whereas patients with aMCI produced more external details (semantic memories) than controls. Therefore, patients with aMCI recall fewer episodic, event-specific details and more semantic details. The remaining two studies [21,34] observed the same trend for internal details: patients with MCI recalled fewer internal details than controls. However, they did not detect a significant difference in the amount of recalled external details.

Temporal Gradient
Regarding the recall of autobiographical memories, some authors [41] found evidence of a significant Ribot-like temporal gradient in patients' performance, with better preservation of remote memories than recent ones.

Discussion
This systematic review aimed to analyze AM performance in patients with MCI, specifically focusing on the main features of AM alteration. Generally, the large number of studies included in the first screening can confirm the interest in this topic. AM has a critical role in forming one's identity, and impairments in AM can have devastating consequences for patients and their families [51]. These aspects have led us to focus on this specific memory component to study the level of impairment that affects patients with MCI. Moreover, we aimed to understand further whether the episodic (i.e., the recall of specific episodes from one's past) or the semantic (i.e., the general knowledge about the self) components of AM are impaired in patients with MCI. This systematic review highlights the presence of impairments in episodic autobiographical memory. However, results are more controversial when assessing the semantic component of AM.
Firstly, it is important to underline that these results are only generalizable to some MCI subjects since 20 out of 21 studies focused on 1 specific subtype of MCI, that is, the amnestic subtype. It is known that individuals affected by aMCI are more likely to develop Alzheimer's disease (AD) [5], it is also likely that impairment in AM is more common in aMCI rather than the non-amnestic subtype. This aspect can certainly influence the generalizability of the results of this systematic review, but it gives us an overlook of autobiographical memory performance in patients with aMCI.
All the studies included in this systematic review report an impairment in the episodic component of AM in patients with aMCI compared to healthy controls. This finding can have different effects on MCI patients. Indeed, research has demonstrated that people rely on their autobiographical memories to achieve various social, practical, and psychological goals [52]. Indeed, AM has important functions in three main domains: social, directive, and self. The social function involves using AM to connect with others. The directive function refers to using AM for solving problems and planning future behavior, while the self function refers to the fact that people can use AM to develop, maintain, and express an enduring self-concept [52].
Our results demonstrate a more controversial situation concerning the semantic component of AM.
Past research investigating AM in healthy aging has shown that age-related decline is particularly marked in episodic autobiographical memory rather than the semantic component [40]. For what concerns pathological aging, specifically MCI, it has been shown that patients with aMCI typically show a malfunctioning of the hippocampus [17], and this could cause an isolated impairment of episodic memory in patients with MCI [15]. On the other hand, it has been theorized that semantic memory is less dependent on the hippocampus and, therefore, could be less affected by the neuropathology associated with MCI [16].
Our systematic review confirms these hypotheses. Regardless of the utilized test, the episodic component is impaired in all the included studies. Instead, only seven studies [22,23,25,28,30,31,39] have found a worse semantic performance in patients with MCI when compared to the HC group. In addition, three studies [15,16,27] found the opposite effect, i.e., more semantic details in patients' recollections compared to controls. This result could be explained by the fact that the latter assessed AM following the standardized scoring procedure developed by Levine et al. [40]. This procedure allows researchers to derive both episodic and semantic information using the same test by segmenting a single transcribed autobiographical narrative into internal event-specific and external semantic details. On the other hand, most authors use separate tests to assess the different components of AM. For example, the most widely used measure is the autobiographical memory interview (AMI) [42]. The AMI is a semi-structured interview consisting of two parts; each independently assesses the two AM components, i.e., the episodic and semantic ones [42].
Therefore, the differences in the semantic memory scores could be attributed to these different assessment procedures. Another possible interpretation could be that, in Levine's procedure, confabulations (which are inaccurate or false narratives produced to give information about the world or the "self") [53] could be scored as semantic (external) details; therefore, resulting in a better semantic performance for patients with MCI.
Another important aspect to consider is the temporal gradient of autobiographical memories. Indeed, past research has shown evidence of better preservation of AM for older memories than for more recent ones, and this effect is known as Ribot's law [54].
For what concerns episodic memory, seven studies [24][25][26]28,31,37,39] confirmed Ribot's law for patients with aMCI. However, three studies [16,34,35] found the opposite effect. The methodological assessment of AM could explain these results. Two out of the three studies that found the opposite effect used Levine's protocol. It could be possible that by considering the internal/external details, the temporal gradient could be inverted.
On the other hand, for what concerns semantic memory, only three studies [24,25,31] confirmed Ribot's law for semantic memory, while four articles [16,21,28,39] found the opposite effect. This aspect could mean that semantic memory is more easily accessible for recent experiences than for older ones.

Limits, Implications, and Suggestions for the Future
This review highlighted several limitations in examining autobiographical memory performance in patients with MCI. The main limitation is that MCI is a varied phenomenon. In a recent systematic review [55], authors highlighted the difficulty of diagnosing MCI. They found that MCI prevalence rates range from 1.2 to 87%. This aspect can be attributed to the lack of a comprehensive standardized neuropsychological evaluation to delineate the aging profile associated with MCI. Therefore, studying autobiographical memory in these patients can be affected by the heterogeneity of this nosological category. Moreover, 20 out of 21 studies focused only on 1 MCI subtype, i.e., aMCI. This aspect is another limitation since it prevents further generalizability of the results. Furthermore, considering only individuals classified as aMCI is difficult to understand. In fact, the classification of a person as having aMCI is generally based on poor performance in tests that evaluate verbal short-or long-term memory, using tests such as memory span, Rey's words, etc., which are all tests that do not assess semantic or episodic aspects of memory. Furthermore, it is well known that people with MCI have dysfunctions in simple [56] and higher-order executive functions [57]. Can it be ruled out that people with executive dysfunction do not have problems with autobiographical memory? Therefore, it cannot be excluded that patients classified as naMCI may present a decline in the episodic and/or semantic components of autobiographical memory. Therefore, the selective choice of subjects classified as aMCI in research on autobiographical memory is based on an unjustified assumption, namely the equivalence of the various forms of memory.
Another limitation could be represented by the partial recovery of articles, including studies published only in some languages that could have excluded relevant information, thus undermining the generalizability of our results.
Regardless of these limitations, at the end of this review emerges an actual need to investigate this specific type of memory in pathological aging. Considering the results of this review, it would be desirable for future research to investigate this aspect to find some gravity indicators of the MCI pathological profile. Moreover, it could be useful to focus on the controversial results related to autobiographical semantic memory, developing a new task more sensitive to detecting changes in this specific component. Finally, it would be interesting to study whether it is possible to draw a relationship between the native language and performance in autobiographical semantic memory.

Conclusions
Overall, this systematic review highlights the presence of impairments in the autobiographical memory performance in patients with MCI. Specifically, all the included studies confirmed that episodic AM is impaired in patients with MCI, while the situation is more controversial when assessing the semantic component.
Thanks to AM, individuals can mentally travel back, acquiring a sense of "self" that can exist in the subjective time [7]. This special form of memory is crucial for developing and maintaining a sense of identity [12]. Moreover, AM has been shown to influence social interactions and abilities such as problem solving and planning [52]. Therefore, starting from the evidence of this systematic work, further studies should detect and deepen the cognitive and emotional mechanisms that undermine AM performance, allowing the development of specific interventions targeting these mechanisms [58]. Indeed, it has been shown that impairments in this memory system are frequent in older adults with depression [59]. For this reason, interventions that alleviate stress and improve mood can enhance AM. Examples are reminiscence therapy and life review, related but distinct interventions that enhance mood and cognitive functions [58]. However, standardized methods of AM stimulation for rehabilitating patients with Alzheimer's disease or MCI are underdeveloped [60]. An exception is represented by the REMau program (réminiscence autobiographique) [60], whose goal is to improve both the episodic and semantic components of AM, focusing on bettering the orientation in time and the chronology of personal events and teaching strategies for accessing memories of the targeted event. This program showed benefits not only for the AM performance but also for the patients' moods. Therefore, it is extremely important to develop more standardized methods targeting AM in order to improve the cognitive and psychological conditions of the elderly.
In conclusion, past autobiographical memories are extremely significant to older people, as well as to their caregivers, since they are associated with one's identity and emotional state, as well as with mood, social functioning, and abilities such as problem solving. This systematic review has highlighted the presence of AM impairment in patients with MCI. These impairments could worsen MCI patients' performances in many domains, not only related to cognitive functioning, and can be affected by many impaired aspects in elderly people, such as sleep quality and mental health [61,62]. Therefore, it is extremely important to investigate these aspects further, also considering some relevant aspects, such as cognitive reserve [63], in order to develop psychological interventions to improve memory of the past.