Introduction
Considerable evidence indicates the benefits of sufficient and high-quality sleep for physical and mental health, including cognition. Although the nuances of sleep patterns may change during a person’s lifetime (Ohayon et al., Reference Ohayon, Carskadon, Guilleminault and Vitiello2004), sleep remains critical for well-being at all stages of life. Some of the changes in sleep patterns that can occur in older adults include a reduction in total sleep time, decreased sleep efficiency, decreased quantity of slow wave and REM sleep, increased sleep fragmentation, and increased awakenings after sleep onset (Campbell and Murphy, Reference Campbell and Murphy2007; Dijk et al., Reference Dijk, Duffy and Czeisler2001; Dorffner et al., Reference Dorffner, Vitr and Anderer2015; Ohayon et al., Reference Ohayon, Carskadon, Guilleminault and Vitiello2004; Scullin and Bliwise, Reference Scullin and Bliwise2015; Wolkove et al., Reference Wolkove, Elkholy, Baltzan and Palayew2007). Further, older adults report worse sleep quality compared to younger people (Crowley, Reference Crowley2011).
In addition to changes in sleep patterns, aging is also associated with changes in cognition, which can include a decline in memory, information processing speed, attention and executive function (Davis et al., Reference Davis, Klebe, Guinther, Schroder, Cornwell and James2013; Salthouse, Reference Salthouse2010), and both the quantity and quality of sleep may be a moderator for cognitive aging (Loerbroks et al., Reference Loerbroks, Debling, Amelang and Stürmer2010). Longitudinal data show that adequate and high-quality sleep promotes cognitive health (Dzierzewski et al., Reference Dzierzewski, Dautovich and Ravyts2018; Vance Reference Vance2012), while increased sleep fragmentation and sleep disruption in older adults is associated with a higher rate of cognitive decline over a 3- to 5-year follow-up (Blackwell et al., Reference Blackwell2014; Lim et al., Reference Lim, Kowgier, Yu, Buchman and Bennett2013; McSorley et al., Reference McSorley, Sun Bin and Lauderdale2019).
Other studies that have examined the relationship between sleep problems and cognitive function in older adults have yielded inconsistent results. Some cross-sectional studies have not shown associations between subjective sleep quality and global cognition (Saint Martin et al., Reference Saint Martin, Sforza, Barthélémy, Thomas-Anterion and Roche2012) or any domains of cognition (Blackwell et al., Reference Blackwell2011; Miyake et al., Reference Miyake, Friedman, Emerson, Witzki, Howerter and Wager2000), whereas other cross-sectional investigations have indicated an association between low-quality sleep and worsening of at least one component of cognitive performance, such as decline in processing speed (Bastien et al., Reference Bastien, Fortier-Brochu, Rioux, LeBlanc, Daley and Morin2003), decreased executive function (Nebes et al., Reference Nebes, Buysse, Halligan, Houck and Monk2009; Waters and Bucks, Reference Waters and Bucks2011), and impaired episodic memory (Tsapanou et al., Reference Tsapanou2017). Such inconsistencies may be attributed to differences in controlled variables, such as depressive symptoms, anxiety, or sleep medication use, or the inclusion of heterogeneous populations across studies, such as the inclusion of neurological and psychiatric comorbidities in participants (Bastien et al., Reference Bastien, Fortier-Brochu, Rioux, LeBlanc, Daley and Morin2003; Foley et al., Reference Foley, Masaki, White, Larkin, Monjan and Redline2003; Moore et al., Reference Moore, Bardwell, Ancoli-Israel and Dimsdale2001). Sample heterogeneity is particularly relevant because each domain of cognitive function has an intricate set of multiple sub-processes (Glisky, Reference Glisky and Riddle2007) that may be uniquely linked to sleep quality in different populations. The lack of a uniform set of cognitive measures across studies makes it challenging to compare findings. In addition to the objective measures of cognition, subjective memory complaints, which can occur in the absence of objective cognitive decline, may predict future cognitive impairment and dementia (Choe et al., Reference Choe, Byun, Lee, Sohn, Lee and Kim2018; Jessen, Reference Jessen2010). Only a few studies, however, have examined the link between sleep quality and subjective memory complaints (Gamaldo et al., Reference Gamaldo, Wright, Aiken-Morgan, Allaire, Thorpe and Whitfield2019; Kang et al., Reference Kang2017), and these have indicated a possible connection between poor sleep and memory dysfunction.
A few imaging studies have investigated the effect of sleep quality on alterations in brain structure in older adults. In adults 55 years and older, self-reported short sleep duration was associated with greater age-related brain atrophy and cognitive decline 2 years later (Lo et al., Reference Lo, Loh, Zheng, Sim and Chee2014). Worse self-reported sleep was also found to be associated with greater hippocampal volume reduction across the adult lifespan (Fjell et al., Reference Fjell2020); however, the effect sizes (ES) were small. Thus, an understanding of how sleep quality may affect brain structure and cognition remains limited.
Given the lack of consistent findings in the literature with respect to sleep quality and neurocognitive performance and the paucity of research with respect to sleep quality and subjective memory complaints, we examined the association between subjective sleep measures and specific aspects of neurocognitive performance and self-reported memory in non-demented middle-aged and older adults with subjective memory complaints. We investigated four domains of objective cognition – simple attention, sustained attention, memory recall, and executive function, and two aspects of subjective memory performance – frequency of forgetting and seriousness of forgetting. We anticipated that sustained attention, memory recall, executive function, and the two self-report memory measures would show significant impairment with poorer sleep quality, while simple attention was less likely to be affected.
Methods
Participants
The present study is a secondary analysis of baseline data collected on 221 middle-aged and older adults, who participated in a randomized, placebo-controlled trial exploring the cognitive effects of daily consumption of pomegranate juice (Identifier: NCT02093130; Siddarth et al., Reference Siddarth2020) approved by the University of California, Los Angeles, Institutional Review Board. Participants were recruited through advertisements, media, and referrals from physicians and families. Our study protocol detailed the methods and procedures, as well as pre-specified inclusion and exclusion criteria. Exclusion criteria included participants with a diagnosis of Alzheimer’s disease or other dementia, a major psychiatric or neurological disorder (e.g. major depressive disorder, schizophrenia, Parkinson’s disease, Huntington’s disease), alcohol or substance abuse, or a history of head trauma or systemic disease affecting brain function (e.g. systemic lupus erythematosus, congestive heart failure). Inclusion criteria were: (1) subjective memory complaints (initially assessed during a phone screen by an affirmative response to the question, “Are you now experiencing or have you experienced memory problems over the past 6 months?” and confirmed by responses to the Memory Functioning Questionnaire [MFQ]); (2) aged 50 or greater; (3) no significant cerebrovascular disease defined as Stroke Risk Factor Prediction < 4 (Wolf et al., Reference Wolf, D’Agostino, Belanger and Kannel1991); (4) normal cognition, defined as Montreal Cognitive Assessment (MOCA) ≥ 25 (Nasreddine et al., Reference Nasreddine2005); and (5) absence of significant depressive symptomology, defined as Beck Depression Inventory-II (BDI-II) < 18 (Beck et al., Reference Beck, Steer and Brown1996). Participants underwent neuropsychological and laboratory testing, including APOE genotyping. Written informed consent was obtained from all subjects in accordance with the University of California, Los Angeles Human Subjects Protection Committee procedures. The trial began in January 2014 and was completed in May 2018, and all data were collected at the Semel Institute for Neuroscience and Human Behavior and Center for Human Nutrition, David Geffen School of Medicine at the University of California, Los Angeles, CA. All data used in the present study were collected at baseline before the intervention was begun.
Measures
Sleep quality
Sleep data were obtained using a self-report questionnaire, the Pittsburgh Sleep Quality Index (PSQI). The PSQI assesses sleep quality exploring seven components of sleep. Each item has a range of 0 to –3, and the seven items are summed for a Total PSQI score (lower scores indicating better sleep quality). With a cut-off score of 5, a PSQI total has a sensitivity of 89.6% and specificity of 86.5% for identifying disordered sleep (Buysse et al., Reference Buysse, Reynolds, Monk, Berman and Kupfer1989). The Total PSQI score was used as a continuous variable in analyses. In addition, we also examined the seven component scores. The distribution of these item scores did not allow their use as continuous variables in analyses, and hence they were dichotomized as below, depending on their distribution, such that there were sufficient participants in each group (see Table 1):
-
1. Subjective sleep quality: Good (score 0–1) versus bad (score 2–3);
-
2. Sleep latency: Less than or equal to 15 minutes (score 0) versus more than 15 minutes (score 1–3);
-
3. Sleep duration: More than 7 hours (score 0) versus 7 hours or less (score 1–3);
-
4. Habitual sleep efficiency: more than 85% (score 0) versus 85% or less (score 1–3);
-
5. Sleep disturbance (i.e. sleep fragmentation or sleep disruption): Less disturbed (score 0–1) versus more disturbed (score 2–3);
-
6. Use of sleeping medication: No use (score 0) versuss use (score 1–3);
-
7. Daytime dysfunction: No problem (score 0) versus problem with daytime dysfunction (score 1–3).
a reported as n (%).
b reported as median score (range).
Obstructive sleep apnea (OSA)
In addition to the PSQI, the STOP-BANG questionnaire (Chung et al., Reference Chung, Abdullah and Liao2016) was used to screen for OSA. This instrument assesses four subjective items and four clinical items: S – “Do you snore loudly?”; T – “Do you often feel tired, fatigued, or sleepy during daytime?”; O – “Has anyone observed you stop breathing during your sleep?’; and P – “Do you have or are you being treated for high blood pressure?” The BANG part of the questionnaire consists of four demographic questions associated with risk for OSA: B – increased body mass index (over 35 kg/m2); A – aged over 50 years; N – neck circumference over 40 cm; and G – male gender. For each question, answering “yes” scores 1, while a “no” response scores 0, resulting in a range of 0–8. A standard cut-off score of 3 is usually taken to indicate high risk for OSA.
Objective measures of cognitive functioning
Within 1 month of the PSQI assessment, participants underwent a full neuropsychological assessment that focused on four domains: (1) Simple attention: measured using the five T-scores of the Continuous Performance Test (CPT) (Conners, Reference Conners2000): detectability, hit reaction time, standard error, reaction time by interstimulus interval, and standard error by interstimulus interval; (2) Sustained attention (ability to maintain focus on a specific task for an extended period of time) (Fortenbaugh et al., Reference Fortenbaugh, DeGutis and Esterman2017) measured by hit reaction time by block and standard error by block (Conners, Reference Conners2000); (3) Executive function: this domain included two major components of executive function, namely cognitive set-shifting and the ability to inhibit responses. Set-shifting ability (adjusted for processing speed) was measured by subtracting the Trail Making Test A time score from Trail Making Test B time score (Arbuthnott and Frank, Reference Arbuthnott and Frank2000). Inhibition was measured using T-scores of commission errors, and hit reaction time of CPT; high T-scores for both commission errors and hit reaction times are indicators of impulsivity (Conners, Reference Conners2000). and (4) Recall memory: measured using the delayed recall score of Buschke–Fuld Selective Reminding Test (SRT) (Buschke, Reference Buschke1973) and the delayed recall score of the Brief Visual Memory Test – Revised (BVMT-R) (Benedict et al., Reference Benedict, Schretlen, Groninger, Dobraski and Shpritz1996). T-scores and raw scores were converted to z-scores using the overall sample mean and standard deviation. For variables in which better performance was represented by lower values (e.g. T-scores of CPT, timed scores of the Trail Making Test), z-scores were reversed so that a higher score indicated better performance. A domain z-score was obtained by averaging z-scores.
Subjective assessment of memory functioning
Participants completed MFQ (Gilewski et al., Reference Gilewski, Zelinski and Schaie1990); two factor scores were used in the analysis from this 64-item questionnaire: frequency of forgetting (MFQ1) and seriousness of forgetting (MFQ2). Higher scores reflect higher levels of self-perceived memory functioning. Factor structure is stable across age groups and internal consistency is high, with Cronbach’s alpha values for its factor scores ranging from 0.83 to 0.94 (Zelinski et al., Reference Zelinski, Gilewski and Anthony-Bergstone1990). The MFQ shows moderate concurrent validity with objective memory performance (Zelinski et al., Reference Zelinski, Gilewski and Anthony-Bergstone1990).
Data analysis
Prior to analyses, all data were screened for outliers and normality assumptions. Baseline clinical characteristics were described using total number (percent) for categorical data and mean (standard deviation) for continuous data. A separate general linear model was estimated to examine if sleep quality measured by Total PSQI was significantly associated with objective cognitive performance in each of the four domains, controlling for age, gender, education level, and baseline MOCA score. We also examined whether age, gender, OSA risk, and APOE-4 status moderated the relationship between cognitive functioning and sleep, by including the interactions in turn (age × PSQI; gender × PSQI; OSA risk × PSQI, ApoE4 × PSQI) in the models. The moderating effects of age, gender, OSA risk, and ApoE4 were tested since all these variables are known to affect sleep and/or cognitive performance (Gosselin et al., Reference Gosselin, Baril, Osorio, Kaminska and Carrier2019; Krishnan and Collop, Reference Krishnan and Collop2006; Lim et al., Reference Lim, Yu, Kowgier, Schneider, Buchman and Bennett2013; Scullin and Bliwise, Reference Scullin and Bliwise2015). Similar general linear models were estimated to examine if sleep quality was associated with MFQ1 and MFQ2, the two subjective memory scores. Finally, for those cognitive domains and subjective memory scores that yielded significant associations with Total PSQI, we analyzed each PSQI component in a separate model to ascertain which specific sleep component, if any, was associated with cognitive performance and subjective memory complaints. Statistical analyses were performed using IBM SPSS Statistics v21. Using Bonferroni correction for multiple comparisons, findings were considered significant at 0.0125 (0.05/4; two-tailed) for the four objective cognitive domain scores and 0.025 (0.05/2; two-tailed) for the two MFQ scores.
Results
Of 221 participants, 15 were excluded from the analyses because they did not complete neuropsychological testing (n = 7) or lacked PSQI scores (n = 8); 3 participants with outlying values on TMT-B−TMT-A (z-scores of ≥ 4) were also excluded. As a result, the final sample size was 203, with a mean age of 60.4 (SD 6.5). The sample was predominantly Caucasian (69.5%) and included a higher proportion of women (69.0%), with a mean educational level of 16.4 (2.1) years and mean Total PSQI of 4.9 (2.9) (Table 1). Cognitive test scores are presented in Table 2.
Abbreviations: TMT, Trail Making Test; BVMT-R, The Brief Visual Memory Test – Revised.
For the scores in simple attention, sustained attention, and executive function, lower score indicates better performance.
a Executive function uses the average of three z-scores, time on TMT part B−A, T-score for commission errors, and T-score for hit reaction time.
PSQI and cognitive domains
Poorer sleep quality, as measured by Total PSQI score, was significantly associated with worse scores on sustained attention [F(1,197) = 10.03, p = 0.002, partial r = −0.22 (Figure 1)]. The Total PSQI score was not significantly associated with simple attention, executive function, or recall memory (Table 3). ApoE4, OSA risk, gender, and age did not moderate the relationship between PSQI and cognitive domain scores (all interactions p > 0.2).
a Adjusted for age, gender, education, and MOCA.
b Unstandardized estimates.
c n = 203; F-statistics are F(1,197).
d Significant result, p < 0.0125 (0.05/4).
e n = 201; F-statistics are F(1,195).
f Significant result, p < 0.025 (0.05/2).
Post hoc analyses examining the relationships between specific sleep components and sustained attention revealed significant correlations between sustained attention and subjective sleep quality as well as sleep disturbance. The group reporting good sleep quality (n = 174) showed significantly better performance in sustained attention compared with the group with bad sleep quality (n = 29) [F(1,197) = 12.61, p < 0.0001, Cohen’s d ES = 0.72] (Supplement 1). The group with less sleep disturbance (n = 149) performed better than the group with more sleep disturbance (n = 54) [F(1,197) = 6.83, p < 0.009, ES = 0.42]. No significant associations between sustained attention and the other PSQI components were found.
PSQI and subjective memory complaints
Total PSQI was significantly associated with both general frequency of forgetting (MFQ1) [F(1,195) = 12.11, p = 0.001, partial r = −0.24 (Figure 2A)] and seriousness of forgetting (MFQ2), [F(1,195) = 9.95, p = 0.002, partial r = −0.22 (Figure 2B)] (Table 3). None of the interaction terms of PSQI total with age, gender, ApoE4, and OSA risk reached significance.
Sleep latency and daytime sleepiness were significantly associated with MFQ1 and MFQ2. Participants with sleep latency ≤ 15 minutes (n = 87) reported significantly less frequency of forgetting [F(1,195) = 8.7, p = 0.004, ES = 0.42] and less seriousness of forgetting [F(1,195) = 12.95, p < 0.001, ES = 0.52] compared to participants with sleep latency > 15 minutes (n = 116). Compared to the group who reported daytime sleepiness (n = 106), the group with no daytime sleepiness (N = 97) had less frequency of forgetting [F(1,195) = 12.69, p < 0.001, ES = 0.51] and less seriousness of forgetting [F(1,195) = 7.87, p = 0.006, ES = 0.40]. Lastly, the group with less sleep disturbance (n = 149) reported less seriousness of forgetting [F(1,195) = 6.53, p = 0.011, ES = 0.41] (Supplement 2).
Discussion
In this cross-sectional study of a large sample of non-demented older adults, we found that poorer sleep quality, specifically fragmented or disrupted sleep, was significantly associated with worse performance in sustained attention and increased memory complaints, and not related to simple attention and memory recall. While there was an association of sleep concerns with poorer executive function, this association did not survive the statistical correction for multiple comparisons. The ES of the significant associations ranged from 0.4 to 0.7, indicating that they were medium to large effects and thus clinically meaningful. The sample of older adults recruited for this study had mild memory complaints, putting them at increased risk for future cognitive decline; therefore, this finding of an association between sleep and cognition in an at-risk sample is important and may serve to inform future preventive interventions that can target sleep to see if they improve cognitive functioning as well in at-risk older adults.
This association of poorer sleep and a reduction in sustained attention is consistent with a previous study indicating that as little as one night of sleep fragmentation is correlated with impaired sustained attention (Martin et al., Reference Martin, Engleman, Deary and Douglas1996). In particular, we found that performance in sustained attention measures was strongly associated with self-reported sleep quality (with a large ES of 0.7). Maintaining sustained attention involves a complex interplay of several brain regions and poor sleep quality has been associated with an increased rate of atrophy across frontal, temporal, and parietal regions (Sexton et al., Reference Sexton, Storsve and Walhovd2014). Thus while the current findings are insufficient to postulate an exact mechanism of why poorer sleep and sustained attention may be related, it is plausible that the widespread cortical atrophy associated with sleep quality may be a cause of worse performance in sustained attention.
By contrast, self-reported sleep quality was not significantly associated with simple attention and memory. While there was an association between poorer sleep quality and worse executive function, this did not survive correction. In the present study, we measured executive functioning using two components: set-shifting and inhibition (Miyake et al., Reference Miyake, Friedman, Emerson, Witzki, Howerter and Wager2000). Previous studies have shown significant associations between set-shifting and sleep quality (Blackwell et al., Reference Blackwell2006; Reference Blackwell2011; Yaffe et al., Reference Yaffe2007). Moreover, in a study of 157 healthy older adults, reports of poorer sleep quality were significantly associated with worse performance on tasks of working memory and set-shifting compared to reports of better sleep quality (Nebes et al., Reference Nebes, Buysse, Halligan, Houck and Monk2009). The discrepancy in findings among studies could be related to the younger age of the participants included in our study, who were on average more than 10 years younger than those in the other studies. In addition, while Blackwell et al. (Reference Blackwell2011) and Nebes et al. (Reference Nebes, Buysse, Halligan, Houck and Monk2009) used the PSQI as a subjective sleep quality measure, Blackwell et al. (Reference Blackwell2006) used objective sleep measurements derived from actigraphy.
We found that subjective perceptions of memory functioning, both general frequency and seriousness of forgetting, were significantly linked to the PSQI total score, sleep latency, and daytime sleepiness. Such associations have been previously reported in older African American and Korean adults, where lengthened sleep latency was associated with severe subjective memory complaints (Gamaldo et al., Reference Gamaldo, Wright, Aiken-Morgan, Allaire, Thorpe and Whitfield2019; Kang et al., Reference Kang2017). Thus, adults with frequent memory complaints tend to report longer periods to fall asleep, or decreased sleep efficiency. Subjective memory problems also have been shown to be an early indicator of subsequent cognitive decline in older individuals (Gifford et al., Reference Gifford2014; Reisberg et al., Reference Reisberg, Shulman, Torossian, Leng and Zhu2010). Thus, our finding that sleep disturbances, even in participants with only a mild degree of self-reported sleep problems, were associated with a greater degree of subjective memory complaints is noteworthy. It is possible that those reporting poor sleep are at increased risk for future cognitive decline, and vice versa.
A strength of our study is that we assessed a broad set of neurocognitive measures in a relatively large sample, but some potential limitations need consideration when interpreting the results. Sleep quality was not evaluated using objective measures such as polysomnography or actigraphy. Examining brain activity through different sleep stages may provide more insight in terms of neurophysiological mechanisms of how sleep affects the different domains of cognitive function. Sleep quality was measured using self-reports, a common and convenient method, and previous research indicates that PSQI self-report measures correlate with objective measures of sleep (e.g. Bliwise et al., Reference Bliwise, Holm-Larsen and Goble2014). However, not all investigations show significant correlations between the PSQI self-report measure and objective sleep measures in both younger and older adults (Grandner et al., Reference Grandner, Kripke, Yoon and Youngstedt2006). Thus, the PSQI may not be the best sleep quality measure for older adults, due to its reliance on participant’s recall capacity (Landry et al., Reference Landry, Best and Liu-Ambrose2015). In addition, PSQI questions were designed to track sleep quality within a 1-month period (Backhaus et al., Reference Backhaus, Junghanns, Broocks, Riemann and Hohagen2002) and can measure various patterns of sleep problems; however, insomnia or other sleep disorders that can affect cognition cannot be diagnosed using this assessment tool. Also, we analyzed components of sleep using categories, and may have thus not been able to detect some nuances in associations with cognition and memory self-reports; for example, not only short sleep, but long sleep (e.g. >9 hours) can also be associated with worse cognitive functions and the ability to detect such nonlinear relationships may have been lost by dichotomizing sleep duration. Due to the lack of available data, we were also unable to examine sleep duration as a continuous variable or examine the factor scores of PSQI in addition to the PSQI component scores.
Conflict of interest
The University of California, Los Angeles, owns a U.S. patent (6,274,119) entitled “Methods for Labeling ß-Amyloid Plaques and Neurofibrillary Tangles;” which has been licensed to Ceremark Pharma, LLC. Dr. Small is among the inventors and has a financial interest in Ceremark Pharma, LLC. Dr. Small also reports having served as an advisor to and/or having received lecture fees from AARP, Acadia, Allergan, Avanir, Axovant, the Gerontological Association of America, Handok, Herbalife, Lundbeck, Lilly, Genentech, Medscape, Otsuka, Recruitment Partners, Roche, and Theravalues. The other authors have no conflicts of interest to declare.
Funding
This work was supported by the Fran and Ray Stark Foundation Fund for Alzheimer’s Disease Research, The Wonderful Company, LLC, John and Barbara McLoughlin Gift Fund (DM), the Parlow-Solomon Professorship (GWS), and the National Institute on Aging of the National Institutes of Health (K23AG049955 to JMD). The original study protocol was registered in ClinicalTrials.gov (Identifier: NCT02093130).
Description of authors’ roles
P. Siddarth and G.W. Small designed the study. L. Ercoli and K. Miller supervised the data collection and collected the data. P. Siddarth was responsible for the statistical design of the study. P. Siddarth and K. Thana-udom carried out the statistical analysis. P. Siddarth wrote the paper and R. Ojha, D. Merrill, J.M. Dzierzewski, G.W. Small, and L. Ercoli assisted with writing.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1041610220003324