Introduction
Subjective cognitive decline (SCD) has been suggested to be an initial manifestation of brain changes related to Alzheimer’s disease (AD) pathology [
1]. SCD has been defined as
a self-experienced persistent decline in cognitive capacity in comparison with a previously normal status and not related to an acute event [
1,
2] and is supposed to precede mild cognitive impairment (MCI), characterized by objective cognitive impairment. Objective cognitive performance in SCD is by definition within normal ranges. The use of standardized tests has been shown to have limited ability to capture differences in performance between groups with and without SCD [
3‐
8]. However, new approaches using more challenging tasks have found subtle deficits in the SCD group [
9‐
12]. Furthermore, the presence of SCD increases the risk of cognitive decline and dementia [
7,
13‐
17] and has been associated with lower volume in the medial temporal lobe, including the hippocampus [
18‐
23], and other AD-related cortical areas [
24‐
26] as compared to cognitively healthy subjects without SCD. Taken together, all of this latter evidence shows that some SCD subjects present structural changes that support the idea of SCD as the first clinical manifestation of AD.
Nevertheless, SCD is a complex syndrome that may be caused by multiple factors besides AD pathology, including other neurological or medical conditions, drug use, or psychological factors (see [
27] for a review). Recently, the SCD-Initiative has proposed a set of specific SCD features, under the name SCDplus, which are associated with an increased likelihood to be an expression of the preclinical stage of AD [
1,
2]. These features are: subjective decline in memory rather than other cognitive domains; onset in the last 5 years; age at onset > 60 years; concerns (worries) associated with SCD; feeling of worse performance than others of the same age group; confirmation of cognitive decline by an informant; and presence of the
APOE ε4 genotype. There is increasing interest in the study of individuals meeting SCDplus features to assess the usefulness of the SCDplus concept for enriching samples at higher risk of cognitive decline in AD secondary prevention trials. However, there is limited knowledge on which of the seven proposed features best relates to objective measures of cognition and/or cerebral changes. Most of the studies on SCD have so far included participants attending memory clinics. Recent evidence points out that the clinical, cognitive, and demographic characteristics [
28,
29] and also the incidence of MCI [
30] are different between clinical and population-based SCD samples. The identification of SCDplus cognitive and brain correlates in a population-based cohort would add to the understanding and usefulness of such a concept beyond memory clinics.
In this context, we hypothesize that some differences may be found in SCD subjects from a population-based cohort depending on the number of SCDplus features that they meet. Therefore, the aim of this study is to describe the cognitive and neuroimaging correlates of SCDplus. For this purpose, we describe the cognitive performance in persons with SCD after having classified them as a function of the number of SCDplus features they meet (≤ 3 and > 3); we investigate whether these features are related to cognitive performance or brain volumes in AD vulnerable areas (hippocampus); and, finally, we explore brain structural patterns related to the presence of SCDplus features and differences amongst non-SCD and SCD groups using an unbiased voxel-based morphometry (VBM) approach.
Discussion
In this study, performed in a large community-based sample of middle-aged healthy subjects, we aimed to determine whether meeting more than three SCDplus features had specific cognitive and neuroimaging correlates, and also to explore which features better predicted cognitive performance and GM volume. Overall, our findings support the use of the concept of SCDplus also in population-based cohorts since performance in subjective decliners was lower than that in non-SCD participants only in the individuals meeting more than three SCDplus features (SCD+ group). This group also showed a decrease of volume in AD-related brain areas as compared to the SCD group meeting fewer features. We found that confirmation of decline by an informant was the best cognitive and GM volume predictor.
In this population-based cohort, 572 out of the 2670 subjects studied reported SCD, representing a prevalence of 21.4%. SCD prevalence is highly variable among studies, depending upon sample recruitment strategies as well as the SCD definition [
40], our findings being in accordance with previous community-dwelling SCD studies [
41,
42]. As compared to non-SCD participants, subjects with SCD reported higher anxiety and depressive symptoms and had higher scores on the SCD-Q questionnaire. Besides the compromise in memory, the most prevalent SCDplus features were a reported decline in the last 2 years and the presence of worries about the decline.
We found a significant negative impact of SCD in cognition, mainly driven by the SCD+ group (individuals meeting > 3 SCDplus features). A significant lower performance in this group was observed as compared to both non-SCD subjects and subjects in the SCD– group, while no differences between these latter groups were found. These findings could not be attributable to confounders, since adjustments for demographics and mood scores were implemented in the analysis. Differences were focused in the episodic memory domain, which mainly relies on the integrity of medial temporal lobe regions [
43]. There is a well-described spread of neuropathology from these regions to wide cortical areas in the course of symptomatic AD, symptoms being related to the topographical distribution of tau pathology. Recently, evidence of increased tau deposition has been reported in the medial temporal lobe of healthy elders with SCD [
44], and our findings on the predominance of memory subclinical deficits in subjects with SCD may be related to such changes.
Our results for cognitive outcomes partially mirror the findings reported by Fernández-Blázquez et al. in subjects who fulfill SCDplus features [
45]. They found lower scores in SCDplus as compared to non-SCD subjects only for verbal episodic memory measured with the Free and Cued Selective Reminding Test (FCSRT). Interestingly, in their study they found differences between SCDplus and SCD restricted to the delayed free recall scores. In our study we used the MBT, which is a test that has similar advantages to the FCSRT (i.e., controlled learning and distinction between storage and retrieval) but overcomes its limitations, such as the ceiling scores achieved by subjects with any or subtle memory deficits [
46]. By using a larger sample and the MBT, we were able to demonstrate global memory differences, in both free and cued measures, between the SCD+ and SCD– groups.
We also found differences in the Coding subtest of the WAIS-IV. This kind of task, which measures attention and psychomotor and processing speed, is one of the most sensitive tools to detect cognitive impact due to any brain insult but lacks disease-related specificity [
47]. Noteworthy, this test was included in the longitudinal data-derived preclinical Alzheimer cognitive composite (PACC) and, together with episodic memory and orientation measures, has shown good performance in detecting decline in cognitively healthy subjects with evidence of AD pathology [
48].
With regard to the ability of specific SCDplus features to predict cognitive performance, in our study, confirmation of the decline by an informant was the feature that best related to objective cognitive performance in memory and executive function tasks. In addition, this feature was also the only one that predicted hippocampal volume and related to lower GM volume in AD-relevant regions. This is in line with several reports highlighting the value of informant ratings in SCD and preclinical AD in which informant complaints predicted progression to AD dementia in longitudinal studies [
49,
50]. Recently, Valech et al. [
51] demonstrated the superiority of informant-related ratings over self-reported ones when it comes to discriminating between controls and biomarker-defined preclinical AD subjects. In addition, they found that informant measures also correlated better with CSF AD biomarkers than self-derived measures. The rationale behind this phenomenon is proposed to be two-fold: first, some subjects in the AD preclinical stage may present some initial form of anosognosia (unawareness of the own difficulties); and second, a proportion of the cognitive decliners report subjective changes related to other conditions, such as psychoaffective symptoms, or personality traits, such as neuroticism. In this regard, in our study, subjects with SCD reported higher anxiety and depressive symptoms as compared to non-SCD subjects, as previously reported [
28,
52]. This finding is in agreement with the attentional bias toward negative information that depressed individuals usually present [
53], which would make them more sensitive to their cognitive failures [
54] and points toward an overlap between the subjective perception of decline due to preclinical AD and due to mood-related causes. In both scenarios, anosognosia and mood-driven complaints, external reports appear as less biased indicators of the presence of actual subclinical decline. Investigators in the INSIGHT-preAD study have further explored this by developing an “awareness indicator”, operationalized as the difference between informant-reported and self-reported scores. This type of measure seems promising since, by classifying subjects in “low
versus high awareness” groups, significant differences that were not captured by isolated SCD scores appear in amyloidosis and reduced cortical metabolism [
4].
With regard to other SCDplus features, we found that being older than 60 years, even after age adjustment, predicted lower performance in Coding. This finding suggests higher executive/processing speed impact of SCD than that expected by age. In contrast, we found a positive relationship between having worries about the decline and Matrix Reasoning performance. An association between higher insight and good reasoning abilities could underlie this unexpected association.
As expected, the SCD+ group showed lower GM volume than the SCD– one in areas known to be affected by AD. On the other hand, we found that the non-SCD group showed no significant differences with respect to the SCD+ group, and also displayed lower GM volumes than the SCD– group. This effect overlapped with those regions with decreased GM volume in the SCD+ group as compared to the SCD– group. Indeed, in an independent population composed of subjects in the AD continuum from healthy controls to preclinical and symptomatic AD, we previously reported a nonlinear association pattern between brain volumes and CSF biomarker levels in some brain regions [
55]. In this previous study, the parahippocampus and some parietotemporal regions presented an initial increase in volume in the low to intermediate biomarker abnormality level that later reverted. In this context, we hypothesize that the observed increase in GM volumes found in the SCD– group presented here might be capturing this initial effect of AD pathology that may be present in some subjects. Within the SCD group, on the contrary, the SCDplus classification criteria showed the capacity to detect individuals who, on average, displayed brain morphological differences associated with preclinical AD stages. In this regard, note that our sample was recruited from the general population rather than from memory clinics as in most SCD studies found in the literature. Therefore, the particular recruitment strategy in our study may underlie this unexpected finding which deserves further investigation. It is known that the recruitment strategy could highly affect SCD characteristics. Perrotin et al. [
28] found differences between SCD subjects recruited from a memory clinic and those from the community. While both groups showed increased β-amyloid deposition and anxiety, subclinical depression and brain atrophy was found only in the SCD subjects who sought medical advice. Using VBM they reported brain atrophy in similar areas to those we found in our study in this latter group, suggesting that the presence of worry of enough intensity to induce help-seeking behavior would be a relevant feature of increased likelihood for AD. Similarly, Abdelnour et al. [
29] reported worse neuropsychological performance in individuals with SCD attending a memory clinic than those recruited as study participants from the community, which may reflect a higher enrichment of AD in this group. Help-seeking has been recently suggested to occur concurrent with NIA–AA preclinical stage 2 (i.e., evidence of positive amyloid and tau markers) and to be strongly related to future cognitive decline [
17,
30]. In the whole picture, our findings in the ALFA population-based cohort may reflect changes at an earlier stage to help-seeking behavior that can be captured at the group level by neuroimaging and challenging cognitive assessments.
This study is not free of limitations. The most important is the absence of biomarker data to be used as proxies of AD pathology. Ongoing studies in the follow-up of a subsample will solve this and will enable a refined analysis to further assess the usefulness of using the number of SCDplus features as a proxy of a continuum within SCD. In addition, as already highlighted throughout this paper, the value of each individual SCDplus feature seems to have a different weight in the prediction of cognitive performance. Ideally, homogeneous subgroups of subjects fulfilling each of the possible feature combinations would aid to disentangle interactions and reveal combinations of features of special interest. However, the sample size and the frequency of SCDplus features in the current study prevented us from using this approach.
Acknowledgements
This publication is part of the ALFA study (ALzheimer and FAmilies). The authors would like to express their most sincere gratitude to the ALFA project participants, without whom this research would have not been possible.
Collaborators of the ALFA study are: Jordi Camí, Grégory Operto, Stavros Skouras, Karine Fauria, Anna Brugulat-Serrat, Gemma Salvadó, Albina Polo, Laia Tenas, Paula Marne, Xavi Gotsens, Tania Menchón, Anna Soteras, Laura Hernandez, Ruth Dominguez, Sandra Pradas, Maria Pascual, Paula Marne, Maria León, Gema Huesa, Marc Vilanova, Sabrina Segundo, Jordi Huguet, and Aleix Sala-Vila.