Subjective cognitive decline (SCD) is common among older adults and a possible clinical state proposed for research that may represent the earliest detectable sign of cognitive decline in aging in some individuals and may predict the future development of Alzheimer's disease (AD) (1). Individuals with SCD have elevated self-perceived cognitive decline or concerns regarding their cognitive function but perform within the normal range on standard neuropsychological tests. For some individuals with SCD, their symptoms predict a higher risk for future mild cognitive impairment (MCI) and AD diagnosis (2, 3). The primary symptoms of SCD are self-perceived decline or complaints regarding memory. While objective memory impairment represents the classical hallmark of AD, multiple studies have suggested that executive function may decline (4, 5), possibly even prior to memory impairment (6). Investigating neural correlates of executive functioning among individuals with SCD may thus improve our understanding of potential early signs and symptoms of cognitive decline and their underlying neural mechanisms.

Executive function encompasses higher-order cognitive processes that allow for complex mental operations, such as attention, planning, monitoring, inhibition, task switching, and working memory (7). SCD has been reported to be associated with lower executive function performance, resembling deficits seen in AD (8–10). Previous neuroimaging studies with executive function tasks in SCD have yielded somewhat mixed results of hyper- and hypoactivation (10–12). In contrast, AD-related functional and structural pathological changes in executive function and associated brain networks are well-characterized. For example, impaired executive function is associated with reduced metabolism in the cognitive control network (i.e., dorsolateral prefrontal, anterior cingulate, and parietal cortex) (13–16) and decreased structural integrity of this network (17, 18). Individuals with MCI had greater activation during executive function tasks in the cognitive control network in comparison with those without cognitive impairment (19, 20). This heightened activation is often interpreted as a compensatory process to maintain cognitive performance in response to pathological changes in the brain such as cerebral amyloid-beta (Aβ) load, atrophy, and white matter structural damage. Given that SCD may precede MCI (1), mirroring the relationship between MCI and AD, it is possible that the hyperactivation in the cognitive control network observed in SCD may underlie the suboptimal yet maintained performance of executive functioning in an AD pathway, at least for some. However, the literature on neuroimaging studies involving SCD remains limited and inconclusive, displaying both hyper- and hypoactivation patterns. These inconsistencies could be attributed to the inherent complexity of “executive functions,” which comprise various intricate subdomains. To address this gap, we aimed to enhance our understanding of the relationship between executive functioning and SCD symptoms using a different task—the digit-symbol substitution test (DSST). This task may provide additional insights into the underlying mechanisms of executive dysfunction in SCD and its potential ties to AD-related pathways.

In the present study, we employed a computerized version of the DSST [adapted from the Weschler Adult Intelligence Scale-Revised (WAIS-R)] (21). The WAIS-R DSST, a paper-and-pencil test, is a commonly used neuropsychological measure sensitive to age-related cognitive decline (22) and pathology (23) and also shows associations with the functional ability to complete everyday tasks (24). The DSST is “polyfactorial” (24), requiring multiple cognitive operations for successful performance (25), including working memory, cognitive control, attention, and processing speed. Previous fMRI studies have used computerized versions adapted for the fMRI scanner procedures, as in the present study, and they have shown DSST-induced activations in the frontoparietal cognitive control network (26–29), supporting the notion that DSST involves multiple cognitive processes. The fMRI version of the DSST of the present study (28, 29) was designed to induce activations related to working memory and attentional control (see task description below).

Another understudied area in SCD research is the relationship among SCD symptoms, neural function (e.g., hypo- or hyperactivation), and Aß load. A previous study from our group assessed the relationship between cerebral Aß load and DSST-induced activation among a partially overlapping sample (57% of overlap) of cognitively normal older adults (29). This study found that greater Aß deposition was associated with greater activation in the insula, inferior frontal gyrus, precuneus, calcarine, and middle temporal regions, indicating a possible compensatory hyperactivation corresponding to Aß load. Another study from our group in the same sample as the current study (30) found the previously reported positive association between Aß load and SCD symptom severity (31, 32). The relationship between SCD symptom severity and DSST-induced activation with this task has not been tested. Building on these results, we hypothesized that greater SCD symptom severity would be associated with greater cortical DSST-induced activations, since SCD symptoms correlate with Aß, which is correlated with DSST activation. In addition to the main hypothesis, we explored the relationship between DSST-induced activation and task performance (i.e., reaction time of the in-scanner DSST task). We hypothesized that both greater SCD symptom severity and Aß load (i.e., known AD risk factors) would be associated with lower DSST task performance (i.e., slower reaction time).

Table 1 summarizes the demographic characteristics of the sample, as well as SCD and neuropsychological scores and in-scanner task performance (detailed sample descriptions are in Supplementary Table S1). The mean age of participants was 73.6 (SD = 7.2), 63% were females, and 90% were Caucasians. The mean years of education of 15.7 (SD = 3.1) indicate a college level of education. In terms of genetic risk, 22% of participants had at least one ε4 APOE allele (with a breakdown of 23% for community volunteers and 21% for memory clinic participants, accounting for three and one instance of missing data, respectively). There was no significant difference in SCD symptom severity observed between participants carrying at least one ε4 APOE allele and those without [t_(22.8) = 0.13, p = 0.90, Welch's t-test]. Participants had high (average 94.0%) task accuracy. Slower in-scanner DSST reaction time was associated with greater SCD symptom severity (β = 0.34, p = 0.01) but not Aβ load [β = −0.09, p = 0.50; R² = 0.11, F_{(2, 55)} = 3.46, p = 0.04; Figure 1: GREEN section]. The association between reaction time and SCD symptom severity remained significant when adjusting for demographic factors (age, sex, race, and education; Table 2).

We found greater activation during the experimental condition (compared to the control condition, Table 3) in the primary visual processing area (calcarine), executive control network (dorsolateral prefrontal cortex, inferior and superior parietal lobes, and precuneus), salience network (insula and dorsal cingulate cortex/supplemental motor area), and subcortical regions (hippocampus and dorsomedial thalamus; Supplementary Figure S2).

Greater SCD symptom severity was associated with lower DSST-induced activation in the bilateral dorsomedial thalamus (cluster-wise FWE p < 0.05, Figure 1: BLUE section). This association was robust to covariates (age, sex, race, education, recruitment source, global Aβ load, neuroticism, GDS, in-scanner DSST reaction time, and framewise displacement motion; Table 4). To evaluate a possible effect driven by outliers, we repeated this analysis using the non-parametric Spearman's rank correlation. The result confirmed the significant association between SCD symptoms and DSST-induced activation in bilateral dorsomedial thalamus [r_s(63) = −0.38, p = 0.002].

We found no significant associations between DSST-induced activation and in-scanner DSST reaction time. Aβ load was not associated with either DSST activation or task performance.

We investigated brain activation during an executive function task among older adults with varying levels of SCD symptom severity by using a computerized DSST task that taps working memory and attentional control. Contrary to our hypothesis, we found a negative association between SCD symptom severity and DSST-induced activation in the subcortical region. Greater SCD symptom severity was associated with lower dorsomedial thalamus activation during DSST. The analyses of in-scanner DSST task performance showed no association between DSST-induced activation and reaction time. However, worse task performance (i.e., slower reaction time for correctly answered trials) was associated with greater SCD symptom severity. Task performance was not associated with Aß load. We have previously reported that greater SCD symptom severity was associated with greater Aβ load in the same participants (30). These collective results suggest that neural function and executive task performance are more directly associated with SCD symptom severity than Aβ load.

Self-perceived cognitive decline and concerns likely represent metacognitive awareness (i.e., reflection on and knowledge about one's own cognitive process). Subjective cognitive concerns may arise due to subtle changes in the usage of neural resources and underlying neurobiological changes in the aging brain. The cognitive control network is the typical neurofunctional structure that mediates executive function and associated processes (16). The present study observed the expected task-induced activations in the regions that belong to this network. However, we found the dorsomedial thalamus to be the neural correlate of SCD symptom severity during the executive function task. Recent studies have proposed the contributions of the dorsomedial thalamus to complex cognitive operations by emphasizing its anatomical connection to the frontal cortex (58, 59). The dorsomedial thalamus has reciprocal connections with the prefrontal cortex, including the dorsal anterior cingulate [e.g., studies in both animal (60) and human studies (61)]. This “trans-thalamic” pathway connecting the medial temporal lobe and prefrontal cortex may provide the function of orchestrating the temporal fidelity to execute cognitive operations (59, 62). The DSST fMRI paradigm requires coordinating a sequence of temporally discrete information steps: encoding a cue, holding it in working memory with a short delay, identifying the matched symbol while eliminating distractors, and coordinating visuomotor information to choose the correct key. Therefore, lower dorsomedial thalamus activation may indicate that suboptimal trans-thalamic cognitive processes, which are not manifest as an overt impairment (i.e., low accuracy), underlie SCD symptoms.

Working memory is a key component of cognitive operations during DSST performance (27), and our computerized version of DSST is particularly designed to induce working memory activities. To recognize the symbol–number pair that is held as a cue in working memory, there are two possible processing routes: recollection and familiarity (63). Although these two processing routes are mainly investigated in the context of episodic (long-term) memory, the same dual processes are also speculated to be present in working memory and short-term memory (64–66). Recollection refers to the retrieved details of previous experience associated with a given item, whereas familiarity is a mere sense of the previous exposure to a given item and is thought to be mediated in the perirhinal cortex through the dorsomedial thalamus pathway (67–69). The DSST (our version and others) does not require recollection of the elaborate details that were associated with the cue (e.g., when/where the cue was acquired), so familiarity-based memory may be a more likely processing route. Earlier studies have suggested that the process of aging is associated with impaired recollection but not familiarity (70). However, a growing number of studies support the notion that familiarity-based memory is also associated with healthy aging and may be more closely associated with AD-related processes (71, 72). The decline of familiarity-based memory may be a possible means to identify those at risk for developing Alzheimer's disease (71). We suggest that the association between lower thalamic activation and greater SCD symptoms may indicate deteriorating familiarity-based working memory function at an early stage of cognitive decline.

While in-scanner DSST performance showed no direct association with neural activation, worse performance was associated with greater SCD symptom severity (Figure 1: GREEN section). Furthermore, there was no association between task performance and Aβ load. However, we previously reported a positive association between Aß load and SCD symptom severity among these same participants (30) (Figure 1: ORANGE section). With our core finding of the association between brain activation and SCD symptom severity (Figure 1: BLUE section), we suggest that SCD symptoms act as a unifying factor connecting all the variables we tested: Aß load, task performance (reaction time), and brain activation (Figure 1). In the visual summary of results related to this cohort, accumulating Aß load was related to elevated SCD symptom severity (Figure 1: an orange arrow); however, Aß load itself was not related to DSST task performance or brain activation. Instead, we posit that the contribution of Aß load to executive function at the behavioral and neural levels is via SCD symptom severity. It is pertinent to point out that this indirect contribution through SCD symptoms could be specific to our task and cohort, warranting further investigation to better understand the relationship between Aß and SCD.

Collectively, our results underscore the significance of SCD symptoms as an early sign of AD-associated cognitive decline as they not only appeared to reflect Aß load but also deteriorating neural function. In addition to Aß, the relationship of other AD neuropathologies with SCD symptoms needs to be further investigated. In one SCD study, CSF biomarkers of total tau pathology but not Aß were associated with cognitive decline (73). As seen in studies in individuals with MCI, tau pathology may have a direct association with cognitive function in SCD. Although the relationship between Aβ load and SCD is somewhat inconsistent (74, 75), SCD symptoms may be at least partially accounted for by Aβ-associated neural alterations. A more recent study (76) reported that Aβ load was not associated with the general severity of SCD but rather with elevated SCD-related worry and awareness of memory deficit. These suggest that SCD symptoms may encompass broad aspects of metacognitive awareness of not only Aβ-associated cognitive decline but also associated psychological factors, including depressive symptoms, worry, and neuroticism (77).

In this cross-sectional fMRI study, it is unclear how to interpret “low” activation in the thalamus. We lack data on neural function prior to the onset of SCD symptoms, which limits interpretation. The present study employed the approach of investigating SCD symptoms as a continuous variable (i.e., SCD symptom severity) rather than a categorical/diagnostic group in combined clinical and community-dwelling samples. First, the continuous variable approach has been used in SCD research (2, 32, 78), and this approach avoids an arbitrary operationalization of SCD as a diagnostic classification. Second, our approach to investigate combined clinical and community-dwelling samples allowed us to capture neurobiological features that were represented in a wider population (i.e., higher generalizability, not limited to clinical/help-seeking SCD samples). This approach has been previously used in an fMRI study in SCD (79), and it is concordant with recent rising efforts to understand the SCD symptoms among the understudied community-dwelling sample (80–83). Despite various advantages of treating SCD as a continuous variable, there may be qualitative differences between SCD symptoms in memory clinic patients and “questionnaire-discovered” individuals who have SCD symptoms in community samples. We attempted to address this point by analytically controlling for recruitment sources in the present findings. We also have a limited sample size; however, our study has one of the largest samples of those with SCD and both fMRI and amyloid PET (10–12). Considering the known concerns regarding the heterogeneity of individuals with SCD, we need longitudinal studies to interpret inferences about changes accurately. Future studies should also include younger participants (e.g., 50s) to further understand the neural basis of earlier cognitive decline.

We found that lower dorsomedial thalamus activation during an executive function task was associated with subjective decline/concerns in older adults. The self-perceived decline in memory and other cognitive function may be clinically informative because these symptoms may reflect an individual's own longitudinal trajectory of cognitive capabilities. Future studies should not be limited to the investigation of the memory domain but consider the subtle decline of executive functions as a neural feature of SCD.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the University of Pittsburgh Institutional Review Board. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Study conception and design: AM, HK, HA, and BS. Data acquisition: HK, BL, AC, CM, WK, HA, and BS. Analysis and interpretation of data: AM, HK, ML, BL, HA, and BS. Drafting and critical revision: AM, HK, ML, HA, and BS. All authors contributed to the article and approved the submitted version.