Better Objective Sleep Quality in Older Adults with High Subjective Memory Decline

INTRODUCTION

Subjective memory decline (SMD), also known as subjective memory impairment, subjective memory complaint, and subjective memory concern, refers to a person’s subjective experience of worsening memory function. In older adults, it is one of the diagnostic criteria for the amnestic form of mild cognitive impairment (aMCI), a preclinical stage of Alzheimer’s disease (AD) [1]. Whether SMD alone reflects emerging AD has been controversial, particularly as high levels of SMD are reportedly associated with mood and personality characteristics, especially at a cross-sectional level of analysis in population samples [2–7]. Nevertheless, a growing body of research suggests that older adults who report SMD are at higher risk of cognitive decline and developing dementia over time [8–10]. This observation suggests that SMD may reflect older people perceiving subtle decline in their own memory ability, before objective memory decline is detectable [10, 11]. However, not all research supports this view [12, 13]. Adding to the complexity is that the relationship between subjective and objective memory varies over the course of the neurodegenerative process [14]. A recent longitudinal study found that memory complaints may precede objective memory decline early on, but objective memory decline precedes an improvement in subjective memory, suggesting a role of decreasing awareness for those with emerging memory difficulties [14].

In addition to longitudinal, observational studies, further support for the link between SMD and emerging AD is provided by evidence for an association with AD-related neuropathology. For example, high levels of SMD have been associated with smaller hippocampal, entorhinal cortex, and medial temporal lobe grey matter volumes [15–18]; and more neuritic plaques, diffuse plaques, and neurofibrillary tangles [19]. Moreover, neocortical atrophy was associated with amyloid-β deposition in SMD but not in MCI or AD [20]. Such findings offer support for the amyloid cascade hypothesis whereby amyloid-β deposition reaches a plateau early in the neuropathogenesis of AD, before the development of objective memory deficits, but subsequent downstream pathophysiological processes continue to occur thereafter [21–23]. This highlights the importance of characterizing the earliest stages of an AD process, before further neuropathology confounds the interpretation of possible changes observed in aMCI.

In light of this, it is important to note other characteristics of people with high SMD. During memory tasks, people with SMD have shown greater neural activation than controls on functional magnetic resonance imaging (fMRI) and magnetoencephalography, particularly in the dorsolateral prefrontal cortex [24–26]. Similarly, less impaired MCI participants demonstrated greater hippocampal activation and greater parietal deactivation than controls during a memory task, while more impaired MCI participants demonstrated less hippocampal activation and less parietal deactivation, in a pattern similar to AD patients [27]. Taken together, the results suggest that in the very early stages of AD pathogenesis (i.e., SMD or early MCI), people may compensate for emerging pathology at a functional neural level and thus objective cognitive deficits may be absent (SMD), or less pronounced (early MCI). Although the expression of SMD may be multifactorial, a better understanding of SMD and associated factors provides the opportunity to identify people in the very earliest stages of a neurodegenerative process, when interventions would be most effective. Furthermore, understanding how SMD interacts with other risk factors for memory impairment and AD will also inform future interventions. In view of this, a working group was recently established to facilitate research focused on whether subjective cognitive decline is one of the earliest changes in a neurodegenerative process associated with AD [28].

Meanwhile, there has been a burgeoning interest in how sleep disturbance may relate to memory and cognition in older adults. It has long been known that sleep disturbance is highly prevalent in AD [29, 30], but more recently, research with animal models suggests that sleep deprivation or restriction leads to amyloid-β accumulation, plaque formation, increases in phosphorylated tau, and impairments in new learning and memory [31–33]. Furthermore, additional research indicates that increased amyloid-β levels lead to sleep disruption [34], suggesting a complex bidirectional relationship between the two [35]. Emerging evidence now suggests that poor sleep is already present in people with aMCI; for example, compared to controls, people with aMCI objectively tend to have greater disruption and reduced quality of slow-wave sleep (i.e., deep sleep), greater periods of sleep disruption (wake after sleep onset [WASO]), as well as poorer subjective sleep quality [36–38]. Because of the potential interplay between sleep and amyloid-β, it would be highly informative to know how early in the AD process disrupted sleep occurs, that is, whether sleep disruption occurs later in the process reflecting already developed neuropathology; or whether it occurs early and may contribute to cognitive decline, or be an early marker of subsequent decline. Determining whether sleep disturbance is present in SMD would, therefore, help elucidate the temporal relationship between cognitive changes and sleep disturbance.

At present, the extent of the relationship between indices of sleep quality and SMD in older adults is yet to be fully detailed. There have, however, been some indications of a relationship with subjective indices of sleep. One study demonstrated that more subjective complaints about sleep were reported in older adults with SMD compared to those without SMD (54.6% versus 26.3%) [39]. Another study showed that people with SMD reported higher levels of sleep disturbance on a depression rating scale [10]. In mixed age population-based samples, there is evidence of an association between SMD and subjective sleep indices, including indicators of possible sleep disorders, such as insomnia or sleep disordered breathing [40, 41]. However, these studies are all focused on subjective sleep indices, which are often uncorrelated with objective sleep indices [42]. Subjective sleep is, however, often correlated with mood [43–45]; while mood itself is commonly associated with SMD [2, 5–7]. Thus, clarification of previous findings using objective sleep measures, and further clarifying the role of mood are both necessary in any consideration of the relationship between sleep and SMD. In terms of objective measures, one study has demonstrated that the waking electroencephalography (EEG) of people with SMD lies on a continuum somewhere between healthy aging and aMCI, with greater frontal delta amplitude and lower parietal, occipital, and central alpha amplitude than controls [46]. Although that study was conducted during restful waking rather than sleep, the results indicate potential alterations to the sleep-wake cycle, likely due to functional neuronal changes. Given that it is unclear if EEG changes persist into sleep or whether they would be sufficient to be observed as more significant sleep-wake disturbance, further investigation is required at both the macro (gross sleep disturbance and movement) and micro (sleep stage changes and EEG) levels. Taken together, these studies warrant further investigation of sleep patterns in older individuals with SMD. Lastly, because sleep quality has been shown to differ in older women compared to older men even in healthy ‘successful agers’ [47], and some research indicates gender differences in rates of SMD [12, 49], the potential interaction with gender also needs to be taken into account.

Therefore, the aim of the present study was to examine whether indices of both objective and subjective habitual sleeping patterns would predict SMD status in an older age sample without self-reported sleep disorders. As SMD is proposed to increase the future risk of MCI and dementia, it was expected that indices of poor sleep quality would be associated with increased likelihood of high SMD.

MATERIALS AND METHODS

RESULTS

There were no significant differences in age, gender, education, overall cognitive status, retrospective memory performance, mood, self-reported number of prescribed medications, average daily alcohol use, or number of vascular conditions between the high and low SMD groups (see Table 1). There were, however, significant differences in nocturnal actigraphic sleep variables, whereby, contrary to expectation, WASO, Number of Nocturnal Awakenings, Sleep Fragmentation, and Time in Bed were lower in the high SMD group, while percent sleep efficiency was greater in the high SMD group, and Bed-time was later in the high SMD group (see Table 2). For example, low SMD participants had an average of nine minutes longer of wakefulness after falling asleep than high SMD participants. There were no group differences in actigraphic SOL or TST, or any of the subjective sleep variables.

Hierarchical logistic regression was used to determine which variables predicted SMD group status (see Tables 3 and 4). Age, gender, and education did not significantly predict SMD group in the first step, Nagelkerke R² = 0.02, χ ²  = 2.98 (3), p = 0.395; nor did depression, anxiety, and stress ratings significantly add to the prediction of SMD in the second step, ΔNagelkerke R² = 0.02, χ ²  = 2.37 (3), p = 0.500. In the third step, using objective indices of sleep, WASO_Acti, TST_Acti, and SOL_Acti significantly predicted SMD, ΔNagelkerke R² = 0.07, χ ²  = 9.80 (3), p = 0.020, with WASO_Acti contributing significant unique variance to the model Wald = 7.15, p = 0.008, OR = 0.97 (CI 95%  = 0.96–0.99). This indicated that longer WASO was associated with a reduced risk of having high SMD. In other words a one minute increase in WASO would be equivalent to a 2.7% reduced risk of having high SMD (see Table 3). To further shed light on this interpretation, we computed WASO in hours and re-ran the regression, which resulted in a Wald = 7.15, p = 0.008, OR = 0.21 (CI 95%  = 0.06–0.66), indicating that a one hour increase in time spent awake after falling asleep would result in a 79% reduced risk of having high SMD, or conversely, every hour of WASO is associated with 4.87 times greater likelihood of having low SMD. In the fourth step, the interaction terms between WASO_Acti, TST_Acti, and SOL_Acti with gender did not significantly improve the prediction of SMD beyond that predicted by the previous steps, ΔNagelkerke R² = 0.01, χ ²  = 0.98 (3), p = 0.806.

Using subjective rather than objective sleep variables in the third step, Sleep Disturbance_PSQI, SOL_PSQI, TST_PSQI, and Daytime Dysfunction_PSQI did not significantly add to the prediction of SMD, Δ Nagelkerke R² = 0.01, χ ²  = 0.83 (4), p = 0.935 (see Table 4). Finally, the interaction terms between Sleep Disturbance_PSQI, SOL_PSQI, TST_PSQI, and Daytime Dysfunction_PSQI with gender did not significantly add to the prediction of SMD, Δ Nagelkerke R² = 0.02, χ ²  = 2.92 (4), p = 0.572. Due to differences in missing data between the subjective sleep variables and actigraphy, regression coefficients varied slightly between the objective and subjective models, in the first two steps (demographics and mood), and thus we have reported the coefficients resulting from the first steps of the analyses in separatetables.

DISCUSSION

The main findings of the present study indicate that in a large sample of community-dwelling older adults who were screened for self-reported sleep disorders and significant emotional distress, those who demonstrated less objective sleep disruption were, unexpectedly, more likely to report high levels of SMD. In particular, less wakefulness after sleep onset (actigraphic WASO) was predictive of high SMD. This association was not accounted for by a potential interaction with gender, or current emotional status; and, subjective indices of sleep did not predict SMD status.

Given the view that high levels of SMD may represent a very early preclinical stage of AD, the current findings are not convergent with previous reported associations between disrupted sleep in AD [29, 30], and the proximal preclinical stage –aMCI [36, 37]. Also, in a further two studies, although group differences were not statistically significant, sleep disruption was greater in aMCI participants as compared to controls [38, 67]. The current study finding of less sleep disruption associated with high levels of SMD is, therefore, initially perplexing.

A possible explanation is that compensatory sleep behavior is instigated in response to the requirement by people with high SMD to exert greater cognitive effort to maintain memory function in day-to-day life; thereby leading to cognitive exhaustion and a heightened need for restorative sleep. Partial support for this interpretation is provided by a recent large cross-sectional population study whereby better self-reported sleep quality was associated with poorer objective memory performance in community-based older adults [68]. In the current study, we found a similar effect but in this instance with objective measures of sleep. In further considering this interpretation, it should be noted that other forms of exertion or exhaustion have been reported as promoting less disrupted sleep patterns in older age; for example, following sleep deprivation, older adults have been observed to display longer sleep duration, fewer nocturnal awakenings, and deeper sleep compared to baseline [69]. In addition, cognitive exertion has also been reported to impact sleep quality in older people [70, 71]. More specifically, an intervention for older adults with insomnia involved eight weeks of increasingly complex, computer-based cognitive training for 20–30 minutes, three times per week. Compared to controls, who performed less cognitively demanding computer-based tasks, the cognitive training group demonstrated reduced sleep onset latency, and increased sleep efficiency on actigraphy after the intervention [70]. Another study demonstrated that engaging in one and a half hours of motor learning tasks prior to bedtime resulted in changes to sleep architecture, as indicated by increased sleep spindles [71], which are known to be associated with memory consolidation [72]. Explanations proposed for these prior findings included the possibility that cognitive training or new learning may improve sleep by increasing cognitive fatigue and decreasing cognitive arousal pre-sleep [70]. This model of sleep-enhancing factors provides a basis for why less disrupted sleep is associated with high levels of SMD.

Further support for this explanatory model is provided by the observation that people reporting high levels of concern about their memory appear to be compensating more at both a behavioral and a neuronal level. People with SMD report higher engagement in everyday memory compensation strategies, particularly if experiencing high levels of stress [73]. Furthermore, older adults with SMD have shown neuronal compensation during an episodic memory task, as indicated by greater activation on fMRI in the dorsolateral prefrontal cortex, with concurrently reduced activation in the hippocampus, despite showing no differences in memory performance compared to controls [25]. Thus, people with concerns about their memory, yet who do not present with clearly identified objective memory deficits, may be maintaining normal levels of memory performance through greater cognitive effort or energy expenditure. This may also suggest a non-linear trajectory between sleep and emerging neurodegenerative disease, whereby sleep is less disrupted in the earliest stages of neurodegeneration (SMD), but deteriorates in more developed stages (aMCI/AD). This is similar to research showing a non-linear trajectory between SMD and objective memory performance [14], and between amyloid-β deposition and neocortical atrophy [20].

With regard to the found relationship between SMD and objective but not subjective measures of sleep, we propose that objectively determined alterations in the sleep-wake cycle are associated with SMD, as distinct from subjective perceptions of sleep, which previous studies have frequently associated with mood or sleep disorders (e.g., [10, 39–41]). However, in the current study, potential participants were excluded if these confounding factors were present.

The explanatory model proposed for the unexpected findings is speculative, and further research is required to specifically test this explanation. In interpreting our data it should be noted that by using actigraphy but not polysomnography to index sleep behavior, we do not have indices of sleep depth or stages, and this will be important to include in further evaluation of SMD and sleep patterns. Such sleep EEG data would be of particular interest given that it may reflect functional neuronal changes observed in SMD in previous research [46]. It will also be interesting to further explore why people with high SMD go to bed later and spend less time in bed. A further issue is that while SMD is argued to be associated with increased risk of later emergence of AD, it is recognized that there are also multiple modifiable risk factors that will contribute to large individual variations in likelihood of cognitive decline over time [74–76], and not all SMD is a reflection of the earliest changes of AD. For example, SMD is often associated with mood. However, this was not the case in the current study, possibly due to the exclusion of people if they reported neurological or psychiatric conditions, which are often associated with co-morbid mood disorders. The present results regarding SMD and sleep would probably need to account for these confounding factors in less healthy community samples. Furthermore, it remains possible that other factors not examined in this study underlie the relationship between sleep and SMD. Clearly, further longitudinal analyses of large cohorts will be required to fully evaluate the associations between sleep and SMD. It will be particularly informative to understand the association with sleep in those who have SMD and do indeed experience cognitive decline, in comparison to those who do not decline. Despite these limitations, these preliminary findings about sleep quality and high SMD in a community sample of older people are of significance at a time when sleep disturbance is increasingly being reported as showing a relationship to cognitive decline and models of AD. With the advent of widely available wearable technologies that index sleep behavior, an exciting possibility is that algorithms could be developed whereby an unexplained change in sleep patterns in an older person with SMD could warrant further clinical investigation and monitoring. Moreover, as there is a push to identify people at the earliest stages of an AD process, the findings could prove clinically useful for early detection of memory decline in aging populations. Subsequent intervention could then be provided early, such as cognitive-behavioral interventions to promote compensatory behaviors for memory challenges in everyday activities, which have been found to be effective in aging populations (e.g., [77]).