A Systematic Review of Neuroimaging Studies Comparing Individuals with Subjective Cognitive Decline to Healthy Controls

Eligibility criteria

English-language, peer-reviewed, human studies, that reported original research with full-text availability were considered eligible for inclusion. Additional inclusion criteria based on the research topic included the necessity of an SCD group (or equivalent condition group). It should be noted that given the variability in nomenclature prior to the recommendations of the International Working Group of the SCD-Initiative [6], the applied definition of SCD included individuals who were identified within an article to have cognitive complaints of any type (e.g., specific to memory, or more general cognitive complaints) in combination with neuropsychological testing sufficient to confirm that participants were functioning within normal limits. Studies were included when participants were aged 55 years and above, to account for the approximate 15-year timeframe to AD conversion that has been proposed [13]. Studies with a lower minimum age were included if they provided age-stratified data with a distinct 55 + age group. Finally, included studies must have included the use of a neuroimaging method, as well as an observational study design (no intervention studies were included). The review protocol for this study was registered with PROSPERO (CRD 42021235940).

Search strategy

PubMed and PsycINFO databases were searched using a combination of relevant keywords agreed upon by the authors (see the Supplementary Material for search terms in full). Following an initial screen to exclude ineligible articles and duplicates, the suitability of each article was independently adjudicated by two of the authors (two of AP, LO, VS, JRG). Where there was disagreement between the initial raters, the third author’s rating determined whether a given study was included.

Data analysis and synthesis

Studies were organized according to the neuroimaging methods employed: MRI (including anatomical MRI, DTI, fMRI, and multi-modal MRI), PET, SPECT, EEG, MEG, and multi-modal neuroimaging combining MRI and PET. For each study, the term/construct of interest (e.g., SCD, subjective memory impairment, subjective cognitive complaints, etc.), number of participants, mean sample age, mean sample education, number of males and females, prevalence of APOE ɛ4, recruitment source, and results relevant to SCD are tabulated.

Study quality ratings

Each study was also rated (by one of AP, LO, VS, CMS, JRG) for quality using the Appraisal Tool for Cross Sectional Studies (AXIS) tool [14], to evaluate risk of bias, study design quality, and quality of reporting. The AXIS tool is comprised of 20 questions and was not designed to yield a total quality score out of 20 as each item cannot be weighted equally [14]. We assessed and interpreted items individually to evaluate each study’s overall quality [14]. Although the AXIS tool [14] is designed to be used with cross-sectional studies, to maintain consistency we also used this scale when rating the quality of longitudinal studies.

Systematic review

The search was conducted on February 14, 2021. Initial search results yielded 619 studies, with 395 articles from PubMed and 224 articles from PsycINFO. Following manual removal of duplicates and evident irrelevant entries (e.g., meta-analyses, intervention studies), 505 abstracts were reviewed for inclusion and exclusion criteria. Following this initial review and manual retrieval of known articles that fit inclusion criteria, 296 full text articles were identified for more in-depth review. After the review of each article by two authors, discrepancies between authors were found for 82 articles. The nature of these 82 discrepancies pertained to correlational studies investigating severity of SCD symptoms in relation to neuroimaging, studies lacking a distinct HC group apart from those with SCD, and studies where the outcomes were focused on creation of novel methods. All discrepancies were successfully reconciled by a third reviewer who was not involved in the initial rating of that article. The exclusions primarily resulted from an undefined SCD group (e.g., lack of discrimination between those with SCD and MCI) or lack of reporting on neuroimaging biomarkers (e.g., reporting of cerebrospinal fluid biomarkers). Results of this process are summarized in Fig. 1 using the PRISMA Flow Diagram [15]. A list of the articles excluded from this review is provided in the Supplementary Material. A final sample of 62 articles published between 2010 and 2021 were included in the final review.

OPEN IN VIEWER

Fig. 1

PRISMA flow diagram for the current systematic review.

Table 1 shows the number of studies included in each neuroimaging modality. Table 2 provides brief descriptions of each neuroimaging technique as well as an overarching summary of the results. Grouped by neuroimaging modality, Tables 3–11 present more detailed information on each study including details on the participant sample and relevant SCD results for each included article. Table 12 shows a count of the significant and non-significant findings separated by neuroimaging modality considering the studies that incorporated multiple neuroimaging modalities.

Anatomical MRI results

16 of 28 studies (eight from Table 3, five from Table 6, and three from Table 8) found significant differences between those with SCD compared to HC using anatomical MRI neuroimaging analyses. Across studies that examined brain atrophy in SCD, reductions in grey matter volume were commonly found in the amygdala [16–18], hippocampus [16–24], entorhinal cortex [18, 21–23], precuneus [16, 25], and various frontal areas ([16, 25–27]. Cortical thickness analyses revealed significantly reduced parenchymal tissue in the left entorhinal cortex [21, 28].

DTI results

In investigations of the microstructural characteristics of white matter, using DTI, 10 of 13 studies (five from Table 4, four from Table 6, and one from Table 8) found widespread differences between SCD and HC groups in various metrics (such as fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), or axial diffusivity (AxD)). Although loss of white matter integrity was widespread, common regions of decreased FA in those with SCD included the corpus callosum [25, 29–31], superior longitudinal fasciculi [29, 31], hippocampus [23, 31], and uncinate fasciculus [28, 29]. Increased MD in those with SCD were also widespread across studies, with common regions including the corpus callosum [25, 29–31], superior longitudinal fasciculus [29, 31]), hippocampus [23, 32], and cingulum [25, 29, 30].

fMRI results

When investigating differences between HC and SCD using fMRI techniques, 11 of 15 studies (six from Table 5, four from Table 6, and one from Table 8) found significant results. Considering resting-state fMRI studies, individuals with SCD exhibited differences in functional connectivity in areas associated with the default mode network (DMN) compared to HC [16, 33]. In studies using task-based fMRI, increased functional activity was found in the precuneus during a divided attention task [34], as well as in the posterior cingulate and thalamus during a working memory task [35]. Interestingly, increased functional connectivity was also seen in the precuneus and thalamus in the studies conducted by Hafkemeijer et al. [16] and Parker et al. [33] using resting-state fMRI.

PET results

Ten of 14 studies (four from Table 7 and six from Table 8) looking at PET yielded significant findings. Six of nine studies that examined Aβ found that individuals with SCD had significantly higher Aβ deposition relative to HC [36–41]. All three studies that investigated tau levels in SCD versus HC found significantly higher tau values in those with SCD [27, 41]. Finally, three of four studies found significant glucose hypometabolism in those with SCD compared to their healthy counterparts [24, 43].

EEG results

There were four studies that used various EEG techniques (e.g., EEG rhythms, amplitudes, and microstate maps) to examine differences between HC and SCD groups (see Table 9). All four studies found significant differences; however, little commonality was found between studies due to the disparate nature of their research questions of interest. See Table 8 for a summary of research findings from EEG studies.

MEG results

Five of the seven studies that used MEG to investigate differences between SCD and HC found significant results. Interestingly, all seven of these studies were comprised of participants from the same recruitment sources, however their sample size increased over time. As each of these studies focused on evaluating unique research questions, the findings from each study yielded unique results which are described in Table 10.

SPECT results

Only one study used SPECT to study differences between HC and those with SCD (see Table 11). This study by Niwa et al. [44] found that individuals with SCD showed hypoperfusion in the caudate, thalamus, and bilateral temporal regions compared to HC.

Study quality ratings

The AXIS tool was used to assess study quality in cross-sectional and longitudinal studies. The AXIS tool does not yield a total numerical score, as a total score would not meaningfully communicate the level of quality across studies as study quality is assessed over 20 components [14]. Table 13 shows a breakdown of the quality ratings for the 62 included studies for each of the 20 items on the AXIS tool. Information on the quality assessment for each individual study can be found in the Supplementary Material.

Can differences be detected between individuals with SCD compared to HC using neuroimaging methods?

A major purpose of this review was to synthesize the neuroimaging literature comparing individuals with SCD to HC. Of the 62 included articles, there were significant findings across each imaging modality. In total, 48 articles identified significant differences between SCD and HC groups. Notably, significant differences were not detected between HC and individuals with SCD in every study. For instance, some studies found significant differences between SCD and other patient groups, such as MCI or AD, while there were no significant differences between the HC and SCD groups. Further, some studies only found differences between HC and SCD when controlling for other variables such as high PiB, or APOE ɛ4, etc. Notably, methodological considerations were very important to the outcomes of each study. Nonetheless, it is clear that significant differences between HC and individuals with SCD have been found using each of the neuroimaging techniques included in the current review. This evidence suggests that there are detectable structural and functional differences in the brains of cognitively normal individuals who report symptoms of SCD compared to those who do not.

Are the detected differences consistent within each neuroimaging technique?

Taken together, 26 of 41 studies that incorporated structural neuroimaging analyses and 31 of 41 studies that incorporated functional neuroimaging analyses found significant differences between HC and those with SCD (Note: The number of incorporated analyses is greater than the total number of studies included in this review as many studies incorporated more than one neuroimaging technique). Overall, there were some common findings within neuroimaging techniques, particularly for the structural imaging approaches. Functional studies that took similar approaches seemed to have consistent results, although the small number of studies and wide variety of analysis techniques employed led to unique findings across studies. Replication of these findings and applying multimodal approaches for the same participants represent important steps for future research.

Are the regions affected in SCD consistent with those expected in early-stage AD?

Across structural MRI studies, we saw areas of atrophy in regions that are consistent with early AD, such as the hippocampus [45, 46] and entorhinal cortex [47–49]. Although atrophy was detected in brain regions associated with AD, studies also reported structural atrophy beyond the hippocampus and entorhinal cortex, including frontal regions and widespread white matter tracts. It is possible that the additional regions of atrophy identified in these SCD studies relates to the aforementioned variability in etiology of SCD; not all individuals with SCD will convert to AD.

Functional MRI studies of preclinical AD represent a relatively new and emerging area of research. Currently, fMRI biomarkers are not included in the AT(N) biomarker grouping outlined by Jack et al. [12]. Although there are no guidelines for AD specific biomarkers such as altered functional connectivity through analysis of fMRI, previous research has noted the usefulness of resting-state fMRI in its ability to distinguish between groups of HC, MCI, and AD [50–52]. A recent review examining the potential of functional imaging biomarkers for neurodegenerative disorders examined 95 papers focused on functional connectivity in AD and reported that evidence for reduced connectivity in the DMN (including the precuneus, posterior cingulate, and prefrontal cortex) in AD has been well established [53]. In the current systematic review, the findings for individuals with SCD showed decreases in functional connectivity in the DMN consistent with AD. Remarkably, increases in functional connectivity were also detected, mostly in frontal regions, setting the findings for individuals with SCD apart from those with AD. It is possible that at earlier stages of the continuum of cognitive decline, individuals with SCD are able to compensate for cognitive challenges, as demonstrated by increased functional connectivity (prior to observable structural atrophy). These findings are congruent with the Posterior-Anterior-Shift with Aging (PASA) model, which characterizes a shift in functional activity from posterior regions of the brain to frontal regions of the brain, reflective of executive based compensation [54]. However, at later stages such as AD, these compensatory mechanisms may no longer subsist. To investigate these possibilities, more research is needed to capture individuals with SCD, MCI, and AD using the same methods longitudinally.

Similar to fMRI, EEG, and MEG are not included in the current AT(N) biomarker grouping [12]. Recently, Horvath et al. [55] published a critical review of EEG/ERP biomarkers for AD that included more than 300 citations focused on a wide variety of methodologies and types of dementia. They concluded that a majority of studies detected alterations in AD, with variations in latency, amplitude, and source localization. They noted that “outcome variability is a serious issue” that may primarily relate to sample size. The current review included four EEG studies focused on individuals with SCD that had a wide variety of methods and sample sizes; therefore, it is not yet possible to compare findings across EEG/ERP studies on SCD to findings in AD. Similarly, with all the MEG studies on SCD coming from one sample and only one published report using SPECT, the outcomes of these studies cannot yet be thoroughly compared to findings from studies focused on AD. Overall, techniques such as EEG and MEG offer complementary information to MRI methods. In future dementia research, it will be important for techniques such as these to be anchored to other methods that we have a greater knowledge base in (such as MRI or PET).

Within the PET literature, it has been well established that markers of amyloid and tau are measurable hallmarks of AD. The current review revealed mixed findings within the SCD PET literature, with some studies revealing greater tau and amyloid values in SCD relative to HC groups, but other studies finding no significant differences. These mixed findings are almost certainly attributable to the methodological differences between studies, discussed next.

Are there methodological characteristics/differences between the studies that could impact the findings?

In synthesizing neuroimaging studies on SCD versus HC, several methodological differences that could lead to discrepant findings became apparent. Below we discuss several methodological differences including differences in definitions and characterization of SCD (i.e., lack of standardized terms and criteria for SCD), APOE ɛ4 status, participant recruitment location, study design, and inclusion/exclusion criteria which may have led to inconsistent findings and may be exerting a broader impact on the SCD neuroimaging literature.

Definitions and characterization of SCD

Across the studies in this systematic review, there was a lack of consistency in the terminology and criteria characterizing individuals with subjective reports of cognitive decline. Across the 62 studies, eight different terms were used to describe those with subjective cognitive decline or SCD-equivalent classifications. Overall, the term SCD was used in just over half of the included studies (see Table 14 for the frequency counts of each of these eight terms). Almost all of the studies that used the term SCD were published after 2014, which may suggest these studies are following the guidance of Jessen and colleagues of the SCD-Initiative, an international working group focused on the advancement of knowledge in the realm of SCD. However, some of the studies that used other terms to refer to SCD such as SMI or SCI were also published after 2014, and do not follow the recommendations set out by the SCD-Initiative [6]. As a result, some studies may have only included individuals with subjective complaints in the domain of memory, while others may have included individuals with subjective decline in other domains, such as executive function (which SCD is intended to capture).

Studies often lacked detailed descriptions of SCD reports such as severity of cognitive concerns, which cognitive domains are affected, onset and length of SCD, the persistence of SCD symptoms, and corroboration of decline from an informant. For example, in documenting and determining inclusion criteria for SCD, some studies only asked one or two questions about whether the person experiences SCD. Meanwhile other studies included completion of questionnaires with specific cut-off scores (e.g., Cognitive Change Index and SCD-Questionnaire), however the rationale for the specified cut-off scores is not typically provided. Furthermore, some studies used comprehensive assessments with several cognitive measures to ensure objective cognitive performance of the participants fell within the normal range, while others only administered cognitive screeners or select subtests from different cognitive measures.

APOE ɛ4 status

In the literature, it is widely accepted that the presence of the APOE ɛ4 allele is associated with an increased risk of developing late-onset AD [56–59]. However, in this review, there was variability across studies in reporting APOE ɛ4 status in their groups. Out of the 62 included articles, only 20 articles included information on APOE ɛ4 status. These were seen across anatomical MRI, DTI, fMRI, multi-modal MRI, PET, and MRI & PET studies. No articles reported APOE ɛ4 status that utilized EEG, MEG, or SPECT.

Participant recruitment location

Most of the articles included in the current review recruited their samples from memory clinics. Of the 62 articles included in this systematic review, only 11 articles investigated differences between HC and individuals with SCD in community samples, rather than memory clinics. In many studies, it appeared that recruitment from a memory clinic was part of the inclusion criteria for their SCD group. The results of the 11 studies that utilized community recruitment (or both clinic and community recruitment) for their SCD groups were mixed. Eight of these 11 studies found significant differences between HC and individuals with SCD [21, 60–63]. Community recruitment was not utilized in stand-alone PET, MEG, or SPECT studies. Studies that recruited individuals with SCD from memory clinics may yield participant samples with greater levels of concern about their SCD compared to studies that recruited individuals with SCD from the community who were not actively seeking medical care [7]. Furthermore, as most SCD participants were recruited from memory clinics, most were located within larger cities and these studies tended to include participants with higher levels of education (e.g., a mean of 16 years of education was not uncommon). It is unclear if this represents a sampling bias or whether individuals with higher levels of education are more likely to report and seek care for experiences of SCD. It is also possible that individuals with SCD and lower levels of education may yield different findings.

Study design

Nearly all the studies included in this review were cross-sectional. It is possible that different results would be revealed if within-subject longitudinal study designs were utilized. Investigating SCD longitudinally will be important in understanding who is more likely to progress from SCD to AD-related or non-AD related dementia in the future.

There was also a wide range in sample size observed in the literature. In the current systematic review, sample sizes for the SCD groups ranged from 5 to 210, which may also have impacted the statistical power to detect differences from HC groups.

Inclusion/exclusion criteria

Many of the studies included in this review listed exclusion criteria regarding their sample such as history of a mental disorder, depression, prescription of psychotropic or psychoactive medications, history of stroke, history of addiction, history of traumatic brain injury that could cause cognitive impairment, or structural brain abnormalities such as white matter hyperintensities or vascular problems. Careful consideration of inclusion/exclusion criteria is imperative, given that some of the above mentioned criteria (e.g., depression, vascular problems) may contribute the presentation of SCD and can occur independent of, or be considered integral to, the presentation of AD. Molinuevo and colleagues [7] recommended three areas of information that should be included in research on SCD including information on the measurement approach, defined cut-off scores for subjective report measures of SCD, and defined cut-off scores for measures evaluating objective cognition. In the current review, a large majority of studies adequately described their measurement approach such that they included information on the research environment and recruitment source, the names of the measures used, who completed the measures, how the measures were administered, the comparison group (i.e., HC), and the domains of cognition that were assessed. Although a large majority of the studies reviewed provided detailed information on their measurement approach, several of these studies did not provide defined cut-off scores the subjective and objective report measures. While some studies provided defined cut-off scores for the subjective report measures and/or objective cognitive measures., many simply stated that performance by the SCD and HC groups were within normal limits.

Limitations

The search strategy in the current review was limited to published studies written in English available on the PubMed and PsycINFO databases. Systematic searches were not conducted outside of these databases nor were unpublished studies solicited. It is possible that studies included in this review are skewed towards those with significant findings due to the “file drawer problem” [64]. Although the current systematic review provides a comprehensive synthesis of neuroimaging findings in individuals with SCD, there were several types of studies that were not included. Specifically, this review sought to detail group differences between individuals with SCD compared to HC. This specific criterion led to the exclusion of studies that did not include a separate SCD group from HC (i.e., control groups comprised of healthy individuals with or without reports of SCD). This criterion also excluded studies without a distinct control group (i.e., healthy individuals without SCD) but investigated differences between individuals with SCD relative to a cognitively impaired group, such as MCI or AD. Similarly, studies that only investigated relationships or associations between the severity of SCD symptoms and neuroimaging metrics were also excluded if they did not report group differences in neuroimaging metrics between individuals with SCD and HC.

Future directions

There are several specific recommendations for future research that can be made based on the existing literature. In particular, studies investigating SCD should engage in standardized research practices outlined by the SCD-Initiative, using both the proposed standardized terminology and group classification of individuals with SCD. Further, SCD-plus criteria should be applied to specifically examine individuals with SCD that have an increased risk of future cognitive decline. Notably, the criteria for SCD-plus has been recently updated by Jessen et al. [8] and recommendations for operationalizing SCD have been well described by Molinuevo et al [7]. Relatedly, the inclusion of more details on the SCD groups being studied will be useful in contextualizing the findings of each study. Specifically, including data on the nature of the self-experienced cognitive decline such as persistence of symptoms, length of time experiencing SCD, and the cognitive domain(s) affected, will help contextualize results based on participant characteristics. It will also be important for future studies to include information on the diversity of the samples as well as consider whether there are cultural contributions to the presentation and/or identification of SCD. Collecting information on APOE ɛ4 from participants should be continued in order to investigate the relationship between APOE, SCD, and AD. Additionally, longitudinal study designs should be used to provide valuable information regarding conversion rates and distinguishing between individuals with SCD who progress from subjective to objective cognitive decline based on neuropsychological measures and those who remain cognitively stable. Longitudinal designs would also help determine whether SCD is indicative of early AD or if it is related to other forms of dementia. Furthermore, multiple measurements could help to model the progression of change over time (e.g., Ashford & Schmitt [65]). Finally, along with consistency, multimodal methods and replications with large sample sizes will be crucial for translating research findings on individuals with SCD into practice.

Conclusions

Biomarker research has focused on the identification of pre-clinical indicators of AD. We have completed the most comprehensive overview of published neuroimaging studies comparing individuals with SCD to HC to date. The current review found that differences can be detected between individuals with SCD and HC across all included neuroimaging modalities, anatomical MRI, DTI, fMRI, PET, EEG, MEG, SPECT, and multi-modal neuroimaging. These findings suggest that neuroimaging is capable of detecting differences between individuals experiencing both normal and non-normal aging, even during SCD where no objective impairments in cognition have yet taken place. Although significant differences were found between individuals with SCD and HC across the neuroimaging modalities included in this review, results were inconsistent both within and across neuroimaging modalities likely due to methodological variations. Due to the inconsistencies in methodology, the level of generalizability of the findings across studies is unknown. Since SCD may arise from other etiologies apart from preclinical AD, it is important that future research in this area is designed and conducted purposefully, recognizing the impact of methodology on the study outcome. Implementation of the recommendations for future neuroimaging studies on SCD are imperative to understanding the earliest changes in the brain that relate to cognitive decline and potential future development of AD.