Introduction
Insomnia is a sleep-wake disorder characterized by difficulty initiating or maintaining sleep, along with an impairment of daytime functioning [
1,
2] whose prevalence in the general population oscillates from 4 to 20%, according to different series [
3‐
5].
Longitudinal epidemiological studies have linked poor sleep quality with a higher risk of late-life dementia [
6,
7], and sleep fragmentation has been associated with a higher incidence of Alzheimer’s disease (AD) [
8]. Understanding how insomnia and other causes of sleep disruption generate a higher vulnerability for AD constitutes a focus of major interest, given the potential of sleep quality as a therapeutic target for dementia prevention.
Increasing evidence suggests that sleep deprivation promotes the accumulation of β-amyloid and tau in the brain, which may be an important mechanism linking sleep disturbances and cognitive impairment [
9,
10]. However, other mechanisms may drive this association. For instance, brain structural differences in individuals with poor sleep quality may contribute to lower the threshold for cognitive impairment [
11‐
13]. In support of this hypothesis, previous neuroimaging studies have described lower gray matter volume involving well-known AD-vulnerable regions, such as precuneus, hippocampus, and cingulate gyrus in patients with insomnia [
12,
14‐
18]. In addition, two independent studies have found, respectively, that poor sleep quality is associated to a higher rate of cortical atrophy [
19] and reduced volume in brain regions usually affected in mild cognitive impairment and AD [
13] in cognitively unimpaired adults. However, these studies have not evaluated potential interactions between sleep quality and
APOE genotype, although previous evidence suggests that sleep quality interacts with
APOE genotype in determining the risk of AD and the burden of β-amyloid and tau pathology in the brain [
20,
21].
On the other hand, diffusion tensor imaging studies have shown decreased fractional anisotropy (which denotes microstructural integrity loss) in several white matter tracts in patients with insomnia and community-dwelling individuals with self-reported poor sleep quality [
11,
22,
23].
In the present study, we aimed to characterize the pattern of cognitive performance, gray matter morphometry, and white matter microstructure associated with the presence of insomnia in a cohort of middle/late-middle-aged cognitively unimpaired individuals from the ALFA (ALzheimer and FAmilies) study [
24]. Notably, the sample used in the present study has been enriched with AD risk factors, therefore potentiating possible associations between sleep quality and AD-related brain changes. We hypothesize that individuals with insomnia will display poorer performance in neuropsychological tests, lower brain volume involving areas usually involved in AD and altered white matter microstructure compared with non-insomniacs, with a more deleterious effect of insomnia being expected among
APOE-ε4 carriers.
Discussion
In the present study involving cognitively unimpaired individuals at increased risk for AD, we found that insomnia was associated to poorer performance in some executive functions and to a distinctive brain macro- and microstructural pattern, characterized by cortical and subcortical GMv differences and decreased white matter diffusivity. In addition, we found that the association between insomnia and gray matter volume is modulated by the APOE-ε4 status, so that APOE-ε4 carriers tend to display lower gray matter volumes in the presence of insomnia, but higher volumes when insomnia is not present.
The demographic and clinical profile of participants with insomnia from this study is similar to that reported in individuals with poor sleep quality from population-based studies [
47,
48]. However, by using a study sample enriched for AD risk factors, we have been able to detect neuroimaging findings that differ from those reported in previous studies, particularly those showing decreased mean and axial diffusivity in white matter tracts, which may be relevant for understanding the association between poor sleep quality and AD in individuals at higher risk for this disease.
Several studies have analyzed the cognitive correlate of insomnia or poor sleep quality, leading in some cases to inconsistent results [
47‐
50]. Our findings are in line with a meta-analysis of 24 studies showing worse performance in executive functions among individuals with insomnia [
50]. Subsequent studies have also reported altered executive functions in patients with insomnia [
51,
52] and community-dwelling individuals with self-reported poor sleep quality [
53]. We did not find differences in episodic memory performance, despite previous evidence of poorer memory performance among individuals with insomnia [
50]. This could be explained by a selection bias towards patients with more severe insomnia (possibly linked to worse cognitive performance) in studies conducted at Sleep Units. Also, we cannot exclude an effect of insomnia on other cognitive domains, such as language and visuoperceptual or visuospatial abilities, as they were not evaluated in our study. Regarding the relationship between cognitive and neuroimaging findings, our group previously described a positive correlation between speed processing and thalamic, as well as superior longitudinal fasciculus (SLF) volume in cognitively unimpaired adults, which is consistent with our results showing a trend towards lower processing speed, as well as lower thalamic volume and altered diffusivity in the SLF in insomniacs [
54].
Regarding the interaction analyses in cognitive performance, although our findings did not survive correction for multiple comparisons, the potential interaction between
APOE and insomnia in delayed episodic memory performance deserves further study, as it is consistent with our finding of lower hippocampal volume in
APOE-ɛ4 carriers with insomnia (as opposed to non-insomniac
APOE-ɛ4 carriers), considering the pivotal role of the hippocampus in episodic memory formation [
55], and it is also in line with previous evidence of a negative interaction between
APOE-ɛ4 and sleep disturbance on memory performance [
56].
Our findings of lower gray matter volume in orbitofrontal and parietal cortex, as well as middle cingulate gyrus, recapitulate some of the main brain volume differences previously reported in patients with insomnia [
14,
15,
17,
18,
57,
58], which supports the existence of a brain structural signature associated to this condition. Regarding possible mechanistic links between these alterations and poor sleep quality, it has been hypothesized that orbitofrontal cortex abnormalities may predispose to insomnia due to altered sensing of the optimal temperature for sleep [
14,
59,
60]. As far as we are concerned, lower thalamic volume has not been previously reported in patients with insomnia, although it has been associated with increased sleep fragmentation variability in cognitively unimpaired elderly subjects [
61]. Thalamic involvement in sleep disturbances is biologically plausible, since regulation of wakefulness and sleep cycles largely relies on a neural network involving neurons in the brainstem, hypothalamus and basal forebrain that provides excitatory input to the thalami and cortical regions [
62]. Also, degeneration of this nucleus in familial and sporadic fatal insomnia, a rare subtype of prion diseases, leads to prominent sleep disturbances [
63].
Our finding of lower gray matter volume in precuneus and posterior cingulate cortex in individuals with insomnia could be linked to the higher vulnerability for cognitive impairment that has been observed in association with poor sleep quality, as these regions are early involved in AD [
64]. Considering that poor sleep quality has been associated with higher levels of β-amyloid deposition in the brain [
65‐
67], changes in these regions (i.e., precuneus and posterior cingulate cortex) could be related to a higher prevalence of individuals with preclinical AD in the insomnia group. An alternative hypothesis is that structural differences observed in individuals with insomnia may represent pre-existing morphological traits that could confer a higher vulnerability to both insomnia and cognitive impairment.
Unexpectedly, we found a greater volume in the left caudate in individuals with insomnia. The interpretation of this finding should be considered with caution, as higher caudate volume has not been previously reported in individuals with insomnia. Even so, with these limitations in mind, one potential explanation would be the existence of a higher prevalence of individuals with preclinical AD among those with insomnia, based on previous evidence of increased caudate size in presymptomatic
PSEN1 mutation carriers, which is a genetic cause of AD [
68], and previous findings suggest a transient size increase in some brain structures during early stages of AD [
69,
70].
Our interaction analyses point to
APOE as a potential modulator in the association between sleep and brain structure, with a higher detrimental effect of insomnia on brain structure being observed among
APOE-ɛ4 carriers. This is in line with previous evidence suggesting that
APOE-ɛ4 carriers may be more vulnerable to different environmental factors, such as lifestyle and vascular risk factors [
71], and also with a previous study showing that better sleep quality attenuates the effect of
APOE-ɛ4 on AD incidence and neurofibrillary tangle burden [
20].
On the other hand, we also found that
APOE-ɛ4 carriers without insomnia tend to display higher gray matter volumes compared with non-carriers, which was not expected. Assuming the hypothesis that sleep quality is gradually deteriorated as AD pathology accumulates in the brain [
72], a potential explanation for this finding would be that the presence of insomnia among
APOE-ɛ4 carriers (which are more likely to harbor AD neuropathological change than non-carriers) may be associated with a more advanced stage in the preclinical phase, whereas
APOE-ɛ4 carriers without insomnia may include a higher proportion of individuals in an earlier AD preclinical stage, where neuroinflammation may still overcome neurodegeneration [
70], resulting in overall greater gray matter volumes in this group.
We found lower diffusivity values involving widespread white matter tracts, exclusively in the right hemisphere. Previous studies have shown a right predominant loss of white matter integrity in patients with insomnia [
11,
22]. Studies in human healthy volunteers have reported an asymmetry in brain hemisphere activity during wakefulness (with left-hemisphere predominance) that is reversed during sleep [
73,
74]. Whether differences in the laterality pattern across the sleep-wake cycle may be related to a higher vulnerability of the right hemisphere white matter tracts to insomnia-related disruption deserves further investigation. On the other hand, a critical difference between our findings and those reported in previous studies is that we found decreased, rather than increased diffusivity associated with insomnia [
11,
23]. Acute ischemic lesions, tumoral lesions, and inflammation are among the main established causes of MD reduction in the brain tissue [
75]. These three scenarios have in common a reduction in water molecules diffusivity due to their confinement to the intracellular compartment, either due to cellular swelling or cellular proliferation. Thus, one potential explanation for our findings is the existence of insomnia-related neuroinflammation involving white matter. In support of this hypothesis, a recent meta-analysis reported an association between insomnia and elevated systemic inflammatory markers [
76], and murine model studies have shown that circadian clock disruption induces astrogliosis [
77] and sleep disturbance is associated with higher expression of pro-inflammatory interleukins and microglial activation in mouse brains [
78,
79]. An alternative explanation is that our results could have been driven by a hypothetical higher prevalence of preclinical AD among individuals with insomnia, as decreased MD in white matter has been previously associated with early β-amyloid deposition [
80]. This could also explain the difference between our findings and those from previous studies, as our sample has been enriched for AD risk factors, therefore facilitating the detection of AD-related changes.
The main strengths of our study are the large size and characteristics of the study sample and its multimodal approach. Although using a sample enriched for AD risk factors may preclude the generalizability of our results, it is better suited to detect brain structural differences that might be driven by AD pathology. On the other hand, an important limitation of the present study is that we have used a subjective measure that queries for the core criteria of insomnia but does not provide more detailed information on sleep quality. In this sense, having used more specific subjective or objective sleep measures may have resulted in more robust statistical associations between sleep quality and neuroimaging and cognitive outcomes. The fact that gray matter volume differences did not survive correction for multiple comparisons (which is inherent to the small effect size of poor sleep quality on gray matter volume) and the lack of AD biomarkers are other relevant limitations in order to interpret our findings. However, we plan to address some of these questions in a further study including data from a longitudinal cohort nested in the ALFA study that incorporates CSF and PET biomarkers.
Acknowledgements
This publication is part of the ALFAstudy (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. The authors would like to express recognition and heartfelt gratitude to Mrs. Blanca Brillas for her outstanding and continued support to the Pasqual Maragall Foundation to make possible a Future without Alzheimer’s.
Collaborators of the ALFA study are: Jordi Camí, Marta Crous-Bou, Carme Deulofeu, Ruth Dominguez, Xavi Gotsens, Laura Hernández, Gema Huesa, Jose María González de Echavarri, Jordi Huguet, María León, Paula Marne, Eider Martínez de Arenaza, Tania Menchón, Marta Milà, Maria Pascual, Albina Polo, Sandra Pradas, Aleix Sala-Vila, Sabrina Segundo, Mahnaz Shekari, Anna Soteras, Laia Tenas, Marc Vilanova, Natàlia Vilor-Tejedor.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.