AbstractBackground and ObjectiveThe baseline sleep phenotype of early Alzheimer’s disease (AD) patients entering lecanemab therapy is poorly characterized.
MethodsForty-four consecutive early AD patients (mild cognitive impairment or mild dementia) awaiting lecanemab underwent attended polysomnography, the Epworth Sleepiness Scale, Insomnia Severity Index (ISI), Pittsburgh Sleep Quality Index (PSQI), and Beck Depression Inventory (BDI). Forty-four controls without documented cognitive disorder, referred for polysomnography by the study’s sleep specialist over the same period, were balanced on sex, age and apnea–hypopnea index. Comparisons used independent-samples tests with Benjamini–Hochberg false discovery rate correction within pre-specified families.
ResultsObstructive sleep apnea was prevalent (61.4% moderate-to-severe) despite a body mass index of 22.5 kg/m2; groups were balanced (all p>0.15). Early AD patients showed greater wake after sleep onset (114.0 min vs. 78.7 min; q=0.006), lower sleep efficiency (67.7% vs. 77.2%; q=0.006) and shorter total sleep time (284.0 min vs. 321.1 min; q=0.039) with comparable recording time, and lower REM arousal index (13.9/h vs. 21.3/h; q=0.012). Differences were moderate (|g| 0.49–0.72); sleep-stage distribution and total arousal index did not differ. ISI (3.8 vs. 13.8) and PSQI (4.1 vs. 10.0) were nonetheless lower in AD (both q<0.001; |g| 1.52–1.86); BDI did not differ.
ConclusionsEarly AD patients awaiting lecanemab therapy had a high prevalence of sleep apnea and reduced objective sleep continuity despite reporting fewer subjective sleep complaints than controls, a contrast that may partly reflect the symptomatic sleep-clinic comparator. Objective sleep assessment before initiating lecanemab therapy may therefore be considered.
INTRODUCTIONAlzheimer’s disease (AD) is a progressive neurodegenerative disorder accompanied, beyond cognitive decline, by non-cognitive symptoms that affect daily functioning and quality of life, among which disturbances in sleep–wake regulation are particularly prevalent and clinically relevant [1]. Impaired sleep may promote amyloid-β and tau accumulation [2–5], while AD-related neurodegeneration destabilizes sleep–wake networks [6,7], creating a reciprocal relationship.
The therapeutic landscape of AD has recently shifted with the advent of disease-modifying treatments. Lecanemab, a humanized monoclonal antibody directed against soluble amyloid-β protofibrils, has emerged as a promising intervention in early AD [8]. Despite its growing clinical use, the baseline sleep architecture and subjective sleep perception of patients entering this therapy remain insufficiently characterized. Systematic baseline phenotyping is clinically relevant—both to detect modifiable sleep comorbidities at treatment initiation and to provide a reference against which treatment-related changes can be evaluated.
Polysomnography (PSG) remains the gold standard for objectively characterizing sleep architecture. Yet few studies have systematically examined PSG findings in early AD patients scheduled to receive lecanemab. Moreover, subjective symptoms may diverge from objective measures, as AD patients frequently underreport sleep disturbance owing to impaired interoceptive awareness [9,10]. A further consideration is obstructive sleep apnea (OSA), which is overrepresented in AD [11] and has been independently linked to greater amyloid burden [12,13] and accelerated cognitive decline [14]. We therefore compared early AD patients with non-demented controls, with the groups balanced for age, sex, and apnea–hypopnea index (AHI), to minimize confounding by sleep-disordered breathing in the assessment of non-respiratory sleep parameters.
This study was therefore designed with two aims: 1) to characterize the objective sleep architecture, arousal profile, and OSA burden of early AD patients scheduled to receive lecanemab; and 2) to compare objective PSG findings and subjective sleep-related symptoms with those of non-demented controls balanced on age, sex, and AHI, thereby quantifying the magnitude of objective–subjective sleep discordance.
METHODSStudy Design and ParticipantsThis is the baseline cross-sectional analysis of an ongoing prospective observational cohort study of sleep outcomes in AD patients undergoing lecanemab therapy at Samsung Medical Center, Seoul, Republic of Korea. Between May 1 and August 28, 2025, 44 consecutive patients with mild cognitive impairment (MCI) due to AD or mild AD dementia (hereafter, early AD) were prospectively enrolled before lecanemab initiation (Table 1). The sample size was not determined by a power calculation but comprises all patients consecutively enrolled within this interim window of an ongoing study; the resulting precision is quantified in the Limitations. Eligibility followed established clinical criteria for lecanemab therapy, with biomarker confirmation of amyloid pathology by positron emission tomography [8]. Exclusions comprised other major neurological disorders, unstable psychiatric illness, or severe medical comorbidities.
For comparison, 44 controls without a documented cognitive or neurodegenerative disorder were selected from patients who underwent attended overnight PSG ordered by the study’s sleep medicine specialist during the same period, and were chosen to achieve group-level balance in sex, age, and AHI; the balance achieved is reported in Results and Table 2. Controls with a prior diagnosis of dementia, MCI, rapid eye movement (REM) sleep behavior disorder, or other neurodegenerative disease were excluded. Because controls were referred for clinical PSG, they may not represent asymptomatic community-dwelling older adults (see Limitations).
Overnight PSG and Sleep QuestionnairesAll participants underwent a single attended overnight PSG including electroencephalography, electrooculography, chin and leg electromyography, electrocardiography, nasal airflow, respiratory effort, and pulse oximetry. SpO2 records and manually annotated respiratory event data were extracted from Embla RemLogic (Natus Medical Inc.). Each recording was scored by one of seven trained polysomnographic technologists following the American Academy of Sleep Medicine scoring manual, version 3.0 [15], from which the definitions of apnea, hypopnea, and arousal were adopted without modification. The AHI was calculated per hour of total sleep time, and OSA severity was categorized using the conventional cutoffs of 5, 15, and 30/h.
Derived parameters comprised sleep architecture, arousal indices, sleep-disordered breathing measures, and the periodic limb movement index (PLMI) (Table 2).
Participants also completed validated Korean versions of the Epworth Sleepiness Scale (ESS) [16], Insomnia Severity Index (ISI) [17], and Pittsburgh Sleep Quality Index (PSQI) [18], together with the Beck Depression Inventory (BDI). All questionnaires were administered on the same day as PSG to ensure temporal alignment between objective and subjective measures.
Statistical AnalysisStatistical analyses were performed using Python 3.11 (Python Software Foundation; https://www.python.org) with SciPy (https://scipy.org) and statsmodels (https://www.statsmodels.org). Continuous variables were expressed as mean±standard deviation and categorical variables as frequencies and percentages. All analyses used every patient with complete data on the variable under analysis; no imputation was performed. Questionnaire entries were re-verified against source records during revision, and a small number of transcription discrepancies, including one out-of-range ISI value, were corrected. For one participant the ISI, ESS, and BDI were not administered and were treated as missing (n=43); the PSQI was available (n=44). These corrections did not alter the direction or significance of any between-group comparison. All p and q values in Table 2 were recomputed from the source dataset for this revision, and the recomputation was reproduced independently in a second computing environment before adoption.
Because control selection achieved covariate balance rather than exact pair-wise dependency, between-group comparisons used independent-samples tests. For each continuous variable the Shapiro–Wilk test was applied to both groups; Welch’s t-test was used when both were consistent with normality and the Mann–Whitney U test otherwise. Categorical variables were compared using Fisher’s exact test, and Hedges’ g was computed with the pooled standard deviation and the small-sample correction. The Benjamini–Hochberg false discovery rate (FDR) procedure was applied within pre-specified outcome families (sleep continuity, sleep architecture, arousal indices, sleep-disordered breathing, oxygenation, and subjective sleep); sleep latency belonged to the sleep-continuity family, whereas N2 and REM sleep latency and time in bed were reported descriptively and assigned to no family. Spearman rank correlations explored associations between cognitive measures (Mini-Mental State Examination [MMSE], Clinical Dementia Rating [CDR], CDR Sum of Boxes [CDR-SB]) and PSG parameters within the early AD cohort, and a two-tailed p-value <0.05 was considered significant. All outcome families, primary tests, the FDR procedure, and the list of exploratory correlation parameters were pre-specified in the statistical analysis plan before database lock; no post hoc redefinition of outcome families or analytic methods was performed.
In response to peer review, two supplementary analyses were added. Both were specified before the new results were examined and are labelled exploratory. First, item-level ISI and PSQI responses were examined within the early AD cohort, focusing on the three items that directly address sleep maintenance (ISI items 1b and 1c and PSQI item 5b), together with a self-reported habitual sleep efficiency derived from PSQI items 1, 3, and 4. Within-person Spearman correlations between these indices and measured wake after sleep onset (WASO) and sleep efficiency were computed with bias-corrected and accelerated bootstrap 95% confidence intervals (10,000 resamples) and Benjamini–Hochberg correction across the eight pre-specified tests; ISI item 1a (difficulty falling asleep) served as a negative control. Item-level responses were available for the early AD cohort only, because control questionnaire data were retrieved as total scores; a between-group item-level comparison was therefore not possible. Item responses were not imputed, and one participant whose ISI items did not sum to the recorded total was excluded. Second, Hedges’ g with 95% confidence intervals was computed for every continuous comparison, and the minimum detectable standardized effect at the realised sample size was derived; post hoc observed power was not computed.
RESULTSBaseline Characteristics of the Early AD CohortForty-four patients with early AD scheduled for lecanemab therapy were enrolled (Table 1): mean age 70.1±8.8 years, predominantly female (27/44, 61.4%), non-obese (body mass index [BMI] 22.5±2.5 kg/m2), and with early-stage disease (MMSE 24.14±2.97; CDR 0.56±0.16). Apolipoprotein E (APOE) genotyping was available for all 44 participants: 25/44 (56.8%) carried at least one ɛ4 allele (ɛ3/ɛ4: 19/44, 43.2%; ɛ4/ɛ4: 6/44, 13.6%), an ɛ4 allele frequency of 31/88 (35.2%), consistent with biomarker-confirmed AD cohorts. Most participants (39/44, 88.6%) were receiving an acetylcholinesterase inhibitor, and cardiovascular-metabolic comorbidities were common (Table 1).
OSA prevalence was high: 5 patients (11.4%) had a normal AHI, whereas 27 (61.4%) met criteria for moderate-to-severe OSA (AHI ≥15/h)—severe in 13 (29.5%), moderate in 14 (31.8%), and mild in 12 (27.3%)—despite the non-obese body habitus of the cohort.
Demographic and Polysomnographic Comparison with ControlsCases and controls were balanced on the pre-specified selection variables, with no significant differences in age (70.1±8.8 years vs. 68.9±8.2 years, p=0.480), sex (61.4% female in both groups, p>0.999), height, weight, or BMI (all p>0.15) (Table 2). Combined AHI was comparable (21.76±14.27/h vs. 22.21±15.39/h, p>0.999; standardized mean difference 0.03) and OSA severity distributions did not differ (all p>0.63); the interpretive consequences of AHI balancing are addressed in the Discussion.
Sleep continuityDespite comparable sleep-disordered breathing, early AD patients showed reduced sleep continuity. WASO was longer (113.99±59.41 min vs. 78.69±35.39 min, q=0.006, Hedges’ g=0.72), sleep efficiency was correspondingly lower (67.65%± 16.29% vs. 77.19%±11.69%, q=0.006), and total sleep time was reduced (283.95±83.03 min vs. 321.12±65.21 min, q=0.039); WASO expressed as a percentage differed in the same direction (q=0.006). Recording time did not differ (time in bed 414.62± 65.85 min vs. 414.01±55.29 min, p=0.183). The 37.2-minute reduction in total sleep time was largely attributable to the 35.3-minute increase in WASO, with smaller contributions from sleep latency (2.5 minutes) and time in bed (0.6 minutes); these closely related measures primarily reflect a common sleep-continuity domain and should not be interpreted as independent abnormalities. Sleep latency was also longer in early AD (median 12.5 [interquartile range, IQR 3.9–19.5] min vs. 5.5 [2.9–13.9] min; q=0.049), but the difference was small (Hodges–Lehmann shift +2.5 minutes, both groups within the normal range) and sensitive to individual observations, significance being lost in 47 of 88 leave-one-out replicates whereas total sleep time remained significant in 88 of 88; it is therefore interpreted cautiously.
Sleep stage distributionThe proportional distribution of non-REM (NREM) stages was broadly similar, although N2% tended to be lower in early AD (54.50%±13.00% vs. 59.21%±8.95%, p=0.051). N3 sleep was low in both groups (3.75%±6.67% vs. 3.05%±5.21%, p=0.355), and REM sleep percentage did not differ.
Arousal indicesThe total arousal index did not differ between groups (21.68± 11.24/h vs. 22.60±9.87/h, p=0.534), nor did respiratory, spontaneous, snore, movement-related, or NREM arousal indices. The REM arousal index was lower in early AD (13.94±13.26/h vs. 21.32±14.36/h, q=0.012), addressed in the Discussion.
Sleep-disordered breathing and oxygenationThe maximum hypopnea duration was nominally shorter in early AD (57.42±19.73 s vs. 72.34±32.23 s, p=0.008) but did not survive FDR correction (q=0.068). Oxygen saturation indices did not differ (all p>0.35), nor did periodic limb movement indices. Exploratory stratification by APOE ɛ4 carrier status (25 carriers vs. 19 non-carriers) showed no significant PSG differences (all p>0.34), although power was limited.
Subjective Sleep and Objective–Subjective DiscordanceThe principal adverse objective pattern described above was a moderate reduction in sleep continuity, with a lower REM arousal index as a secondary finding; early AD patients nonetheless reported fewer subjective sleep complaints than controls (Fig. 1). ESS did not differ (6.5±4.1 vs. 7.9±4.9, p=0.153) (Fig. 1B). In contrast, ISI was substantially lower in early AD (3.81±4.67 vs. 13.84±5.92, p<0.001, Hedges’ g=−1.86) (Fig. 1C), as was PSQI (4.11±2.73 vs. 9.95±4.64, p<0.001, g=−1.52) (Fig. 1D). BDI scores were comparable (11.4±7.8 vs. 10.7±7.6, p=0.622), arguing against differential depressive burden as a full explanation. Symptom burden differed accordingly: 22 of 44 controls (50.0%) scored 15 or higher on the ISI compared with 1 of 43 early AD patients (2.3%), and a PSQI above 5 was present in 34 of 44 controls (77.3%) versus 12 of 44 patients (27.3%).
The joint distribution of ESS and ISI (Fig. 1A) showed tight clustering of early AD patients in the low-symptom quadrant across both AHI severity strata (<15/h vs. ≥15/h).
Exploratory Associations between Cognitive Severity and Sleep ArchitectureSpearman rank correlations between cognitive measures (MMSE, CDR, CDR-SB) and PSG parameters were examined within the early AD cohort (Table 3 and Fig. 2A). A suggestive pattern emerged for slow-wave sleep: MMSE correlated positively with N3% (ρ=+0.315, p=0.037) (Fig. 2B) and CDR-SB inversely (ρ= −0.394, p=0.008) (Fig. 2C), indicating that better cognition and lower dementia severity were associated with preserved slow-wave sleep.
A second pattern was observed for nocturnal arousal and movement measures, uniformly in the direction of lower values with greater dementia severity: higher CDR and CDR-SB were associated with lower total, REM and movement arousal indices and with lower PLMI (ρ=−0.33 to −0.43, all p<0.03) (Table 3). An inverse association between CDR and WASO (ρ=−0.341, p=0.023) is noted separately, as WASO reflects wake duration rather than arousal frequency. None of these correlations survived FDR correction within the correlation family and should be regarded as hypothesis-generating.
Multiple-Comparison CorrectionWithin each pre-specified outcome family (Table 2), reduced sleep continuity, a lower REM arousal index, and lower ISI and PSQI scores remained significant after Benjamini–Hochberg correction (all q<0.05); sleep latency also reached the threshold and is interpreted with the limits described above. The shorter maximum hypopnea duration (q=0.068) did not survive correction and is reported as exploratory.
Item-Level Subjective Responses and Effect Sizes (Exploratory)Within the early AD cohort, responses to the ISI and PSQI items that directly address sleep maintenance were largely at floor: 75.0% scored 0 on ISI item 1b (difficulty staying asleep) and only 7.5% scored 2 or more, and 60.0% scored 0 on ISI item 1c (waking too early), in a cohort whose mean WASO was 114.0 minutes and mean sleep efficiency 67.7%. PSQI item 5b was the most responsive item (mean 1.24 of 3; 43.9% scoring 2 or more), indicating that the instruments were not silent on nocturnal awakening (Supplementary Table 1 in the online-only Data Supplement). None of these items showed a detectable association with the corresponding objective measure within the same patient: Spearman correlations with measured WASO and sleep efficiency ranged from ρ=+0.001 to +0.156, none nominally significant and all q=0.996 after Benjamini–Hochberg correction; total scores behaved similarly (Supplementary Tables 2 and 3 in the online-only Data Supplement). Self-reported habitual sleep efficiency derived from the PSQI exceeded measured sleep efficiency (median 93.3% vs. 69.8%; Wilcoxon p=5.2×10−9), a mean within-person difference of 23.5 percentage points, and the two were uncorrelated (ρ=−0.06, p=0.71) (Supplementary Table 4 in the online-only Data Supplement). Effect sizes for all continuous comparisons are given in Supplementary Table 5 (in the online-only Data Supplement). These analyses are exploratory and are constrained by restricted item variability, the recall period of the questionnaires, and limited power (see Limitations).
DISCUSSIONPrincipal FindingsThree principal findings emerged. Early AD patients entering lecanemab evaluation carried a high burden of OSA despite a non-obese habitus. After balancing for age, sex, and AHI they showed moderately poorer objective sleep continuity than controls (|g| 0.49 to 0.72); because recording time was identical, the greater WASO, lower sleep efficiency and shorter total sleep time represent a single sleep-continuity domain rather than independent abnormalities, whereas sleep-stage distribution and total arousal index did not differ. Subjective complaints were nevertheless lower by a standardized margin two to three times larger, although that margin may partly reflect the symptomatic sleep-clinic comparator; within the early AD cohort itself, item-level responses addressing sleep maintenance showed no detectable association with the corresponding objective measures.
Sleep-Disordered Breathing Burden in Lecanemab CandidatesThe high prevalence of moderate-to-severe OSA in this lean Korean early AD cohort (61.4%) substantially exceeds rates reported in unselected community-dwelling older adults and aligns with prior evidence that OSA is overrepresented in AD [11] and that its severity is linked to greater amyloid burden [12,13]. Intermittent hypoxia and sleep fragmentation have been implicated in amyloid-β and tau pathology and may accelerate cognitive decline [12,13], potentially modulating the effects of anti-amyloid therapy, although direct evidence for such an interaction is lacking. Given the magnitude of unrecognized OSA observed here, baseline polysomnographic screening in lecanemab candidates warrants prospective evaluation as a potentially modifiable comorbidity.
AHI Balancing as a Design Choice: Interpretive ConsiderationsA distinctive methodological feature of this analysis is the inclusion of AHI as a balancing variable in addition to age and sex, used to reduce confounding by sleep-disordered breathing [12,14]. Consequently, AHI and AHI-derived respiratory indices were comparable by design, and their non-significant between-group differences should not be interpreted as evidence of biological equivalence. This balancing approach improves comparability for non-respiratory sleep parameters but limits inferences regarding the contribution of OSA itself to sleep disturbance in AD.
Objective–Subjective Sleep DiscordanceA clinically relevant observation is the contrast between reduced objective sleep continuity and minimal subjective complaints, although its magnitude depends on the comparator. One alternative explanation is that the ISI and PSQI contain few items addressing sleep maintenance, so that their total scores would not be expected to track an isolated increase in WASO. Our exploratory item-level analysis provides supportive but not definitive evidence against this as a complete account: the items that do address sleep maintenance showed no detectable association with measured WASO or sleep efficiency within the same patient, and self-reported habitual sleep efficiency was higher than measured sleep efficiency; the constraints on this inference are set out in the Limitations. The pattern is consistent with prior PSG studies reporting impaired insight into sleep quality in AD [9,10] and with impaired awareness of sleep disturbance, including anosognosia [1], although this design cannot establish the mechanism. Comparable BDI scores (p=0.622) argue against differential mood reporting. The between-group difference in ISI and PSQI may nonetheless reflect, in part, the referral context, since the control group was itself symptomatic; the size of the discordance cannot be read as a measure of impaired sleep perception in early AD, and comparison with community-dwelling controls would be required to establish its true extent. The within-cohort analyses are unaffected, as they involve no comparator. The clinical implication is more modest: in some patients with early AD, questionnaire scores were low despite substantially reduced measured sleep continuity; questionnaires alone may therefore provide an incomplete baseline characterization, and objective assessment may complement them, with the modality chosen according to the clinical question.
Lower REM Arousal Index: A Secondary ObservationThe reduction in REM arousal index in early AD (13.94/h vs. 21.32/h; q=0.012) warrants cautious interpretation, given the absence of differences in total, NREM, and respiratory-driven arousal measures and the susceptibility of REM arousal physiology to medication effects and to REM sleep amount. Two non-mutually-exclusive explanations may contribute. First, 88.6% of the cohort were taking acetylcholinesterase inhibitors, which have been reported to modulate REM sleep in AD [19,20], and greater REM consolidation may translate into a lower arousal index per hour of REM. Second, AHI balancing removed the dominant respiratory-driven source of REM arousals, so residual differences may reflect intrinsic disease biology, consistent with early degeneration of the brainstem networks mediating cortical arousal during REM. We regard this as a signal rather than a core conclusion.
Cognitive Severity, Arousability, and Slow-Wave SleepWithin the early AD cohort, greater dementia severity was associated with lower total, REM and movement arousal indices and with lower PLMI in exploratory analyses (Table 3); a separate inverse association between CDR and WASO is interpreted cautiously, as WASO reflects wake duration rather than arousal frequency. None survived FDR correction and all are hypothesis-generating. The direction—lower arousal and movement indices with more advanced disease—is biologically plausible in the context of degeneration of wake-promoting neuronal networks in AD [6,7] and converges with the between-group finding of reduced REM arousal index. Paralleling this, MMSE correlated positively and CDR-SB inversely with N3%, consistent with prior evidence linking preserved slow-wave sleep to better cognition in older adults at risk for AD [6]. Because sleep is implicated in glymphatic amyloid-β clearance and amyloid dynamics [2–4], while slow-wave sleep supports hippocampus-dependent memory consolidation [6], the consistency of this association across two independent cognitive scales provides rationale for investigating slow-wave sleep as a candidate longitudinal biomarker in lecanemab-treated patients.
LimitationsSeveral limitations merit acknowledgment. First, although AHI was comparable between groups, the control group was drawn from PSG referred by a sleep medicine specialist and was itself symptomatic. This referral context enriches the comparator for sleep-related symptoms and is the most important design limitation of the study: the between-group difference in subjective measures may therefore reflect, in part, the symptomatic composition of the comparator and should not be attributed to impaired sleep perception in early AD alone; comparison with community-dwelling controls would be required to establish the true extent of the discordance. Second, the sample size was fixed by consecutive enrolment rather than by a power calculation. For a single two-sided comparison at α=0.05, 44 participants per group provided 80% power to detect Cohen’s d=0.60 and 64% power to detect d=0.50; power after FDR correction is lower. The null findings for sleep-stage distribution and for most arousal indices should therefore be read as inconclusive rather than as evidence of equivalence, since their confidence intervals exclude a large but not a small or moderate difference (Supplementary Table 5 in the online-only Data Supplement). The shorter maximum hypopnea duration (q=0.068) and the exploratory cognition–sleep correlations did not survive correction; in leave-one-out analysis the MMSE–N3 association lost nominal significance in 11 of 44 replicates, whereas the CDR-SB–N3 association remained significant in all 44. Third, the cross-sectional design precludes causal inference; longitudinal follow-up is ongoing and will be reported separately. Fourth, single-center recruitment from a tertiary academic memory clinic in Korea, the predominance of women, and the exclusion of advanced disease may limit generalizability. Fifth, acetylcholinesterase inhibitor use in 88.6% of AD patients introduces a pharmacological confound that cannot be disentangled in observational data. Sixth, item-level responses were available only for the early AD cohort, so those analyses are within-group and cannot establish that the pattern is specific to AD; restricted item variability attenuates correlation independently of sleep perception, at n=40 to 41 the minimum detectable |ρ| with 80% power is 0.43 and the confidence intervals extend to +0.44, so a weak or moderate association cannot be excluded, and the PSQI refers to habitual sleep over one month whereas PSG measured a single laboratory night subject to a first-night effect. Finally, the absence of fluid biomarkers precludes direct correlation with amyloid burden, an important direction for the planned follow-up.
Clinical Implications and Future DirectionsThese baseline findings indicate that early AD patients entering lecanemab therapy may carry a substantial and largely unrecognized burden of sleep-disordered breathing, together with a reduction in sleep continuity that was not accompanied by proportionate subjective complaints. Baseline sleep evaluation, including objective assessment for OSA, may therefore be considered before lecanemab initiation to identify potentially modifiable sleep comorbidity and to establish a reference for longitudinal follow-up. Longitudinal studies should determine whether objective measures add value to questionnaires and whether lecanemab modifies sleep architecture in early AD.
Supplementary MaterialsThe online-only Data Supplement is available with this article at https://doi.org/10.17241/smr.2026.03685.
Supplementary Table 1.Item-level responses to sleep-maintenance items of the ISI and PSQI (early AD cohort)
Supplementary Table 2.Within-person association between subjective sleep-maintenance items and measured sleep continuity (early AD cohort)
Supplementary Table 3.Context: association between ISI and PSQI total scores and measured sleep continuity (early AD cohort)
Supplementary Table 4.Self-reported habitual sleep efficiency versus measured sleep efficiency (early AD cohort, n=42)
Supplementary Table 5.Effect sizes for every continuous between-group comparison in Table 2
Availability of Data and MaterialThe datasets generated or analyzed during the study are not publicly available due to institutional review board restrictions on the sharing of individual-level clinical and polysomnographic data but are available from the corresponding author on reasonable request.
NOTESEthics Statement
The study protocol was approved by the Institutional Review Board of Samsung Medical Center (No. 2024-12-050-015). All AD participants provided written informed consent, with additional consent from a legally authorized representative where appropriate. For control participants, whose PSG data were obtained from the institutional sleep database, the consent requirement was waived and all data were fully anonymized. The study was conducted in accordance with the Declaration of Helsinki.
Author Contributions
Conceptualization: Yi-Seul Choo, Heekyoung Kang, Hee Jin Kim, Eun Yeon Joo. Data curation: Yi-Seul Choo, Heekyoung Kang, Seok-Yeol Yang, Jun Pyo Kim. Formal analysis: Yi-Seul Choo, Heekyoung Kang. Funding acquisition: Hee Jin Kim, Eun Yeon Joo. Investigation: Yi-Seul Choo, Heekyoung Kang, Seok-Yeol Yang, Hea Ree Park, Jun Pyo Kim, Sang Won Seo. Methodology: Yi-Seul Choo, Heekyoung Kang, Hea Ree Park, Hee Jin Kim, Eun Yeon Joo. Project administration: Hee Jin Kim, Eun Yeon Joo. Resources: Sang Won Seo, Hee Jin Kim, Eun Yeon Joo. Software: Yi-Seul Choo, Heekyoung Kang. Supervision: Hee Jin Kim, Eun Yeon Joo. Validation: Yi-Seul Choo, Heekyoung Kang, Seok-Yeol Yang, Hea Ree Park. Visualization: Yi-Seul Choo, Heekyoung Kang. Writing—original draft: Yi-Seul Choo, Heekyoung Kang. Writing—review & editing: all authors.
REFERENCES1. Bliwise DL. Sleep disorders in Alzheimer’s disease and other dementias. Clin Cornerstone 2004;6(Suppl 1A):S16-28.
2. Ju YE, Lucey BP, Holtzman DM. Sleep and Alzheimer disease pathology—a bidirectional relationship. Nat Rev Neurol 2014;10:115-9.
3. Xie L, Kang H, Xu Q, Chen MJ, Liao Y, Thiyagarajan M, et al. Sleep drives metabolite clearance from the adult brain. Science 2013;342:373-7.
4. Kang JE, Lim MM, Bateman RJ, Lee JJ, Smyth LP, Cirrito JR, et al. Amyloid-β dynamics are regulated by orexin and the sleep-wake cycle. Science 2009;326:1005-7.
5. Spira AP, Gamaldo AA, An Y, Wu MN, Simonsick EM, Bilgel M, et al. Self-reported sleep and β-amyloid deposition in community-dwelling older adults. JAMA Neurol 2013;70:1537-43.
6. Mander BA, Winer JR, Jagust WJ, Walker MP. Sleep: a novel mechanistic pathway, biomarker, and treatment target in the pathology of Alzheimer’s disease? Trends Neurosci 2016;39:552-66.
7. Oh J, Eser RA, Ehrenberg AJ, Morales D, Petersen C, Kudlacek J, et al. Profound degeneration of wake-promoting neurons in Alzheimer’s disease. Alzheimers Dement 2019;15:1253-63.
8. van Dyck CH, Swanson CJ, Aisen P, Bateman RJ, Chen C, Gee M, et al. Lecanemab in early Alzheimer’s disease. N Engl J Med 2023;388:9-21.
9. Most EI, Aboudan S, Scheltens P, Van Someren EJ. Discrepancy between subjective and objective sleep disturbances in early- and moderate-stage Alzheimer disease. Am J Geriatr Psychiatry 2012;20:460-7.
10. Hita-Yañez E, Atienza M, Cantero JL. Polysomnographic and subjective sleep markers of mild cognitive impairment. Sleep 2013;36:1327-34.
11. Emamian F, Khazaie H, Tahmasian M, Leschziner GD, Morrell MJ, Hsiung GY, et al. The association between obstructive sleep apnea and Alzheimer’s disease: a meta-analysis perspective. Front Aging Neurosci 2016;8:78.
12. Sharma RA, Varga AW, Bubu OM, Pirraglia E, Kam K, Parekh A, et al. Obstructive sleep apnea severity affects amyloid burden in cognitively normal elderly. A longitudinal study. Am J Respir Crit Care Med 2018;197:933-43.
13. Liguori C, Mercuri NB, Izzi F, Romigi A, Cordella A, Sancesario G, et al. Obstructive sleep apnea is associated with early but possibly modifiable Alzheimer’s disease biomarkers changes. Sleep 2017;40:p. zsx011.
14. Yaffe K, Laffan AM, Harrison SL, Redline S, Spira AP, Ensrud KE, et al. Sleep-disordered breathing, hypoxia, and risk of mild cognitive impairment and dementia in older women. JAMA 2011;306:613-9.
15. Troester MM, Quan SF, Berry RB; American Academy of Sleep Medicine. The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specifications. Version 3. Darien, IL: American Academy of Sleep Medicine 2023.
16. Cho YW, Lee JH, Son HK, Lee SH, Shin C, Johns MW. The reliability and validity of the Korean version of the Epworth sleepiness scale. Sleep Breath 2011;15:377-84.
17. Cho YW, Song ML, Morin CM. Validation of a Korean version of the insomnia severity index. J Clin Neurol 2014;10:210-5.
18. Sohn SI, Kim DH, Lee MY, Cho YW. The reliability and validity of the Korean version of the Pittsburgh sleep quality index. Sleep Breath 2012;16:803-12.
Fig. 1Subjective sleep measures in early AD patients evaluated for lecanemab and in controls. A: Joint distribution of ESS and ISI scores. Circles denote participants with an AHI below 15/h and stars those with an AHI of 15/h or above; bold crosses denote group means. Dashed lines mark the conventional thresholds for excessive daytime sleepiness (ESS>10) and for at least moderate insomnia (ISI≥15). B–D: ESS (B), ISI (C) and PSQI (D) scores by group. Boxes show the median and interquartile range with whiskers extending to 1.5 times the interquartile range, open diamonds show the group mean, and individual data points are jittered horizontally. ESS was compared using Welch’s t-test, and ISI and PSQI using the Mann–Whitney U test; q values are Benjamini–Hochberg false discovery rate–corrected within the subjective sleep outcome family. The ESS and ISI were not administered to one early AD participant (n=43); the PSQI was available for all participants (n=44). AD, Alzheimer’s disease; ESS, Epworth Sleepiness Scale; ISI, Insomnia Severity Index; AHI, apnea–hypopnea index; PSQI, Pittsburgh Sleep Quality Index. Fig. 2Associations between cognitive severity and sleep architecture in the early AD cohort (n=44). A: Spearman rank correlation coefficients (ρ) between three cognitive measures (MMSE, CDR, CDR-SB) and 26 polysomnographic parameters. The coefficient is printed in each cell so that the panel can be read without colour; cell shading indicates the magnitude and direction of the correlation, and coefficients in bold are nominally significant (p<0.05). None of the 78 associations was significant after Benjamini–Hochberg false discovery rate correction within the correlation family. B: MMSE score against slow-wave (N3) sleep as a percentage of total sleep time (ρ=+0.315, p=0.037). C: CDR-SB against N3 percentage (ρ=−0.394, p=0.008). In B and C the y axis is broken between 17% and 34% to accommodate a single observation with N3=36.4%; markers are semi-transparent so that overlapping observations with identical values remain distinguishable. Dashed lines show a linear trend for visualization only and do not represent the Spearman correlation. In leave-one-out analysis the association in B lost nominal significance in 11 of 44 replicates, including on removal of the single observation with N3=36.4%, whereas the association in C remained significant in all 44 replicates. All associations are exploratory and hypothesis-generating. AHI, apnea–hypopnea index; CDR, Clinical Dementia Rating; CDR-SB, CDR Sum of Boxes; MAI, movement arousal index; MMSE, Mini-Mental State Examination; NREM, non-rapid eye movement; ODI, oxygen desaturation index; PLMI, periodic limb movement index; REM, rapid eye movement; RERA, respiratory effort-related arousal; SpO2, peripheral oxygen saturation; TST, total sleep time; WASO, wake after sleep onset. Table 1Demographics and clinical characteristics of the early AD cohort (n=44) Values are mean±standard deviation for continuous variables and n (%) for categorical variables. OSA severity was determined from AHI in all 44 patients. APOE ɛ4 carrier status was defined as the presence of at least one ɛ4 allele (ɛ3/ɛ4 or ɛ4/ɛ4); the 19 ɛ3/ɛ4 heterozygotes shown here are not the same group as the 19 ɛ4 non-carriers referred to in the exploratory stratification in Results. Table 2Comparison of demographics, polysomnographic findings, and subjective sleep measures between the early AD and control groups
Values are mean±standard deviation or n (%). Cell-level n is shown where it differs from 44. The between-group test is shown for each variable. q values are Benjamini–Hochberg FDR–corrected within the outcome families shown; demographic variables, the two descriptive sleep-latency variables and the OSA severity distribution were not corrected (-). † Sleep latency met the FDR threshold (q=0.049), but the effect was small and sensitive to leave-one-out analysis; it is therefore interpreted cautiously in the Results. AHI, apnea–hypopnea index; A+H+RERA, apnea+hypopnea+respiratory effort-related arousal; BDI, Beck Depression Inventory; BMI, body mass index; ESS, Epworth Sleepiness Scale; FDR, false discovery rate; ISI, Insomnia Severity Index; NREM, non-rapid eye movement; ODI, oxygen desaturation index; OSA, obstructive sleep apnea; PSQI, Pittsburgh Sleep Quality Index; REM, rapid eye movement; RERA, respiratory effort-related arousal; SpO2, peripheral oxygen saturation; TST, total sleep time; WASO, wake after sleep onset. Table 3Spearman rank correlations between cognitive measures and polysomnographic sleep parameters in the early AD cohort (n=44) Spearman rank correlation coefficients (ρ) and corresponding two-tailed p-values. None of the reported associations survived Benjamini–Hochberg false discovery rate correction within the correlation family and should therefore be regarded as hypothesis-generating. AHI, apnea–hypopnea index; CDR, Clinical Dementia Rating; CDR-SB, CDR Sum of Boxes; MAI, movement arousal index; MMSE, Mini-Mental State Examination; NREM, non-rapid eye movement; ODI, oxygen desaturation index; PLMI, periodic limb movement index; REM, rapid eye movement; RERA, respiratory effort-related arousal; SpO2, peripheral oxygen saturation; TST, total sleep time; WASO, wake after sleep onset. |
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||