The Effects of Sleep Disturbances on Fatigue and Cognition in Multiple Sclerosis
The Effects of Sleep Disturbances on Fatigue and Cognition in Multiple Sclerosis
Fatigue is a prevalent symptom in patients with multiple sclerosis (MS) and is associated with considerable impairment in quality of life as well as loss of occupational capacity. Sleep disturbances are regarded as a critical factor in the development of fatigue and are frequently observed in individuals with MS. However, they often remain underrecognized, undiagnosed, and consequently untreated.
Polysomnography, the gold standard for assessing sleep architecture and quality, has rarely been applied in the investigation of sleep disorders in MS. Accordingly, uncertainties remain regarding the prevalence and extent to which sleep disturbances contribute to fatigue in this population. Moreover, emerging evidence suggests an association between sleep disorders and cognitive dysfunction in MS. Yet, it is unclear whether cognitive impairment arises from the sleep disorder itself, from the resulting fatigue, or from other independent factors.
Pharmacological treatments for MS-related fatigue remain limited, given heterogeneous and frequently non-replicable effects. Non-pharmacological interventions such as physical activity, cognitive behavioral therapy, and psychoeducation have shown promise but yield variable outcomes. The development of novel and effective therapeutic strategies requires a more comprehensive understanding of the etiology of fatigue. To date, the role of sleep disturbances and their relationship to cognitive performance in MS have not been adequately investigated.
The objective of this project is to determine the prevalence and characteristics of sleep disorders in MS patients with fatigue using polysomnography and to examine their relationship with cognitive impairment. In addition, the study will compare sleep quality parameters and the prevalence of sleep disorders across different MS subtypes (relapsing-remitting, primary progressive, and secondary progressive). Furthermore, within a sub-study, it will be investigated whether the type of immunotherapy has an influence on the aforementioned aspects.
Finally, the project seeks to integrate artificial intelligence (AI) into polysomnography analysis to streamline data evaluation and facilitate the future assessment of therapeutic interventions.
The study will be conducted as a non-invasive, non-interventional, longitudinal observational trial including MS patients with fatigue and a control group of patients with subjective sleep complaints but without MS. Recruitment will take place over 36 months at two centers: the Department of Neurology at the University Hospital Düsseldorf and the Maria Hilf Clinics in Mönchengladbach. Additional recruitment will be supported by community-based neurologists in the Mönchengladbach region to broaden the study cohort and ensure representativeness of the study population.
Approximately 382 MS patients are expected to be enrolled. The number of control participants will be determined by the proportion of MS patients presenting with sleep disorders and will be recruited consecutively from the neurological sleep laboratory of the Maria Hilf Clinics. For AI training, retrospective polysomnography data from the past five years (N ≥ 10,000 patients) at the Maria Hilf Clinics will be utilized.
The study protocol includes overnight polysomnography to assess sleep quality, along with comprehensive clinical evaluation, neuropsychological testing, and validated questionnaires addressing fatigue, subjective sleep quality, daytime sleepiness, depression, and anxiety.
Based on manually scored polysomnography, AI models will be trained to identify key parameters of sleep quality. The findings of this study will advance the understanding of the role of sleep disturbances in MS-related fatigue and will facilitate the integration of AI into sleep research, thereby streamlining the evaluation of future therapeutic approaches.
Project Objectives
1.1. In-depth characterization of sleep disorders in MS patients with fatigue.
Exploratory analysis:
Which specific types of sleep disorders are most frequently observed in MS patients with fatigue?
1.2. Comparison of subjective fatigue severity and objectively measured fatigability between patients with and without abnormalities in sleep-related parameters.
Hypothesis:
Fatigue severity and fatigability differ significantly between MS patients with diagnosed sleep disorders and those without, with greater fatigue observed in patients with sleep disorders.
Exploratory analysis:
Which sleep-related parameters show the strongest correlation with fatigue severity and fatigability in MS patients?
1.3. Comparison of self-reported fatigue severity (FSMC) with fatigability as an objectively measurable, short-term phenomenon.
Hypothesis:
There is a significant association between self-reported fatigue severity and objectively measured fatigability, with higher subjective fatigue scores corresponding to greater motor and cognitive fatigability.
Exploratory analysis:
To what extent do discrepancies exist between subjective fatigue and objectively measured fatigability? Which moderating factors explain possible divergences between subjective perception and objective performance decline?
1.4. Comparison of fatigability between patients with and without abnormalities in sleep-related parameters.
Hypothesis:
Patients with sleep-related abnormalities demonstrate significantly greater objectively measured fatigability compared to patients without such abnormalities.
Exploratory analysis:
Which specific sleep-related parameters are associated with increased motor and/or cognitive fatigability? Do distinct patterns emerge for cognitive versus motor fatigability depending on the type of sleep disorder?
1.5. Comparison of cognitive performance between MS patients with and without abnormalities in sleep-related parameters.
Hypothesis:
Information processing speed as well as learning and memory performance are significantly reduced in MS patients with sleep disorders compared to those without.
Exploratory analysis:
Which specific sleep-related parameters are most strongly associated with cognitive impairments in MS patients?
1.6. Comparison of cognitive performance, subjective fatigue, and fatigability in MS patients with sleep disorders versus control patients without MS but with the same sleep disorders.
Hypothesis:
Information processing speed, learning, and memory performance are significantly lower in MS patients with sleep disorders compared to control individuals without MS who present with the same types of sleep disorders.
Hypothesis:
Fatigue severity and fatigability are significantly greater in MS patients with sleep disorders compared to control individuals without MS with the same types of sleep disorders.
Exploratory analysis:
What differences in the impact of sleep disorders on cognitive performance, fatigue, and fatigability can be identified between MS patients and control individuals without MS?
1.7. Comparison of sleep medicine parameters and prevalence of sleep disorders in patients with fatigue across different MS subtypes.
Exploratory analysis:
Are there significant differences in sleep-related parameters among MS patients with different disease courses?
Exploratory analysis:
Are there significant differences in the prevalence of sleep disorders among MS patients with different disease courses?
1.8. Investigation of longitudinal changes in objectively measured sleep parameters, fatigue, fatigability, and cognition over a 12-month period in patients with MS.
Exploratory analysis:
How do polysomnography-derived sleep parameters change in patients with MS over time?
Exploratory analysis:
To what extent does subjectively perceived fatigue, as assessed by the Fatigue Scale for Motor and Cognitive Functions (FSMC), change over a 12-month period?
Exploratory analysis:
To what extent does objectively measurable fatigability change over a 12-month period?
Exploratory analysis:
To what extent do cognitive functions, as measured by the Symbol Digit Modalities Test (SDMT), Verbal Learning and Memory Test (VLMT), and Brief Visuospatial Memory Test-Revised (BVMT-R), change over a 12-month period?
Hypothesis:
Patients with MS who were diagnosed with a sleep disorder at baseline and subsequently received treatment will report lower levels of subjective fatigue after 12 months compared with patients with MS in whom no sleep disorder was treated following the baseline assessment.
Hypothesis:
Patients with MS who were diagnosed with a sleep disorder at baseline and subsequently received treatment will exhibit lower fatigability after 12 months compared with patients with MS in whom no sleep disorder was treated following the baseline assessment.
Hypothesis:
Patients with MS who were diagnosed with a sleep disorder at baseline and subsequently received treatment will demonstrate better performance on the SDMT, VLMT, and BVMT-R at the 12-month follow-up compared with baseline.
1.9. Investigation of the associations between changes in sleep parameters and changes in subjective motor and cognitive fatigue, fatigability, and objective cognitive performance over a 12-month period.
Exploratory analysis:
Are changes in sleep parameters associated with changes in subjective motor and cognitive fatigue, fatigability, and objective cognitive performance?
Hypothesis:
Patients with MS who exhibit positive changes in sleep parameters will report lower levels of subjective fatigue on the FSMC after 12 months compared with baseline.
Hypothesis:
Patients with MS who exhibit positive changes in sleep parameters will demonstrate reduced fatigability after 12 months compared with baseline.
Hypothesis:
Patients with MS who exhibit positive changes in sleep parameters will demonstrate improved performance on the SDMT, VLMT, and BVMT-R after 12 months compared with baseline.
Hypothesis:
Patients with MS who exhibit negative changes in sleep parameters will report higher levels of subjective fatigue on the FSMC after 12 months compared with baseline.
Hypothesis:
Patients with MS who exhibit negative changes in sleep parameters will demonstrate increased fatigability after 12 months compared with baseline.
Hypothesis:
Patients with MS who exhibit negative changes in sleep parameters will demonstrate poorer performance on the SDMT, VLMT, and BVMT-R after 12 months compared with baseline.
1.10. Development and training of an AI model for automated analysis of polysomnography data.
Exploratory analysis:
How effective is a trained AI model compared to human experts in diagnosing sleep disorders and grading their severity?
Hypothesis:
AI-based analysis of polysomnography data can identify sleep disorders in MS patients more rapidly than manual evaluation.
Objectives of the Substudy
The substudy exclusively uses data collected within the framework of the main study. No additional examinations will be conducted. The targeted sample size for the cross-sectional analyses within the substudy is N = 100, and N = 77 for the longitudinal follow-up.
2.1 Characterization of sleep disorders in MS patients with fatigue undergoing high-efficacy (category 2/3) versus low-efficacy (category 1) therapy according to the current S2k guideline (3rd version) of the German Society of Neurology, and comparison of relevant sleep parameters as well as prevalence rates of sleep disorders between these two groups.
Exploratory analysis: Comparison of sleep-related parameters between patients receiving high- versus low-efficacy therapy.
Exploratory analysis: Which specific types of sleep disorders occur in MS patients with fatigue in the different therapy groups, how frequently do they occur, and what is the effect size of possible differences in their distribution?
2.2 Characterization of sleep disorders in patients with relapsing-remitting MS (RRMS) and fatigue undergoing Ocrelizumab therapy compared to other high-efficacy therapies, and comparison of key sleep parameters as well as prevalence rates of sleep disorders between these two groups.
Exploratory analysis: Comparison of sleep-related parameters between RRMS patients treated with Ocrelizumab versus other high-efficacy therapies.
Exploratory analysis: Which specific types of sleep disorders occur in RRMS patients with fatigue in the above-mentioned therapy groups, how frequently do they occur, and what is the effect size of potential differences in their distribution?
2.3 Investigate how objectively measured sleep parameters, fatigue, and cognition change over 12 months in patients undergoing high- versus low-efficacy therapy.
Exploratory analysis: Do the above-mentioned objectively measured sleep parameters, fatigue, and cognition develop differently between baseline and follow-up depending on therapy efficacy (high vs. low efficacy)?
2.4 Investigate how objectively measured sleep parameters, fatigue, and cognition change over 12 months in RRMS patients undergoing Ocrelizumab versus other high-efficacy therapies.
Exploratory analysis: Do the above-mentioned objectively measured sleep parameters, fatigue, and cognition develop differently between baseline and follow-up depending on therapy (Ocrelizumab vs. other high-efficacy therapies)?
2.5 Investigate how changes in sleep parameters are associated with changes in subjective motor and cognitive fatigue as well as objective cognitive performance after 12 months in patients undergoing high- versus low-efficacy therapy.
Exploratory analysis: Can objective sleep parameters derived from polysomnography be used to identify predictors of changes in motor and cognitive fatigue as well as cognitive performance at follow-up in the respective groups (high- vs. low-efficacy therapy)?
2.6 Investigate how changes in sleep parameters are associated with changes in subjective motor and cognitive fatigue as well as objective cognitive performance after 12 months in RRMS patients undergoing Ocrelizumab compared to other high-efficacy therapies.
Exploratory analysis: Can objective sleep parameters derived from polysomnography be used to identify predictors of changes in motor and cognitive fatigue as well as cognitive performance at follow-up in the respective groups (Ocrelizumab vs. other high-efficacy therapies)?
Study Design
Study Population
Sample Size:
One aim of the study is to determine the prevalence of sleep disorders in a representative cohort of patients with MS and fatigue. As no precise estimates of prevalence are available, the most conservative assumption of 50% (± 5) was used for sample size calculation, resulting in a target of 382 participants (OpenEpi version 3, 95% confidence interval, reference population approximately 52,000 MS patients in North Rhine-Westphalia).
A matched control group (matched for age, education, and sex) will be recruited prospectively at the sleep laboratory of Kliniken Maria Hilf, with a maximum of n = 382 participants depending on the number of enrolled MS patients.
The AI algorithm will be trained on retrospective anonymized polysomnography data from the past five years (≥ 10,000 cases) at Kliniken Maria Hilf, and subsequently validated using the prospective control cohort.
Study Procedures
Informed Consent Process:
Participants will receive both oral (via study investigators) and written (via a standardized patient information sheet) explanations of the study objectives, procedures, and associated risks in language that is understandable. They will be informed that participation is voluntary and may be withdrawn at any time without disadvantage to subsequent medical care. Written informed consent will be obtained using a standardized form.
The indication for sleep medical evaluation is established prior to informing the patient about the possibility of study participation and is independent of it. Information about participation in the study is provided only after the patient has decided to undergo a sleep medical evaluation.
For AI training, anonymized retrospective data (n ≥ 10,000) obtained as part of routine clinical practice will be analyzed without requiring consent. This does not apply to prospectively enrolled participants, who will be explicitly informed in the consent form about the use of their data for AI training and may opt out of such use.
Assessment of Outcomes:
Medical history:
- Disease history, immunomodulatory and symptomatic therapies, number of relapses, and demographic variables
Clinical examination:
- Expanded Disability Status Scale (EDSS), Nine-Hole Peg Test (NHPT), 25-Foot Walk Test (T25FWT), and hand grip strength measurement (dynamometer)
Neuropsychological tests:
Daytime sleepiness:
- Pupillographic sleepiness test
Validated self-report questionnaires:
Polysomnography (two nights) according to AASM criteria, including:
Multiple Sleep Latency Test (MSLT):
- Only in patients and controls with suspected narcolepsy, hypersomnia, or other disorders of excessive daytime sleepiness
AI training dataset:
Approximately 11 months after the baseline assessment, study participants will be contacted by telephone by the study team and asked to complete the Fatigue Scale for Motor and Cognitive Functions (FSMC) again to assess fatigue.
If no clinically relevant fatigue syndrome is present, study participation will end with this telephone contact.
If at least a mild fatigue syndrome persists, participants will be invited again to the sleep laboratory at Kliniken Maria Hilf. During this follow-up visit (approximately 12 months after baseline), the entire assessment protocol of the baseline examination will be repeated. If participants decline to repeat polysomnography, the follow-up assessment may alternatively consist solely of the neuropsychological and neurological examination.
Study participation will end after the telephone assessment or, if applicable, after the follow-up visit, and no further data collection will take place thereafter.
Overall Study Duration:
The total planned duration of the study is approximately four years.
Subsequent re-contact of already enrolled patients
Risk-Benefit Assessment
The enrolled patients/participants will not derive any direct individual benefit from study participation. The study aims to advance understanding of the etiology of fatigue in multiple sclerosis (MS). However, the examinations may have differential diagnostic or therapeutic relevance, provide insights into novel therapeutic interventions, and thereby contribute to optimizing patient care.
All planned assessments are procedures routinely conducted in clinical practice. Sleep medicine evaluation is clinically indicated due to the presence of a fatigue syndrome.
During the study, participants will be asked to complete several questionnaires. These questionnaires may include personal questions regarding sleep quality, lifestyle, and mental health. Completing these questionnaires could, in rare cases, lead to psychological distress. For this reason, questionnaires will be completed in the presence of a psychologist. Furthermore, participants will be informed in writing that, should psychological distress occur after study participation, they may contact the study team and arrange an outpatient appointment at the Maria Hilf Clinics in Mönchengladbach.
Some assessments involve the use of surface electrodes placed on the participants' skin. These electrodes are applied with a conductive gel designed to optimize signal recording. There is a minimal risk of allergic reactions, which may manifest as redness, itching, or rash.
No further complications or risks associated with the planned procedures are known. Accordingly, the overall risk to study participants is considered very low.
For the purposes of training the AI system, the data to be analyzed were collected routinely as part of standard clinical practice and will be analyzed retrospectively in anonymized form. There is therefore no associated risk to patients.
Study participation will be discontinued if any of the exclusion criteria described above occur, particularly if participants withdraw their informed consent.
Statement on Medical Justifiability:
Taking the above risk-benefit assessment into account, medical justifiability is clearly established.
Biometry
The statistical analyses will be conducted using multivariate procedures with appropriate statistical software (R, SPSS).
8.1. Analysis Plan:
Objective 1.1.
The following sleep-related parameters will be quantified and compared with established normative data: sleep latency, sleep efficiency, time in bed, total sleep time, REM latency, sleep stage transitions, sleep periods, arousals, apnea/hypopnea index, and periodic limb movement index. In addition, the prevalence of established sleep disorders will be assessed according to the criteria of the International Classification of Sleep Disorders (ICSD): restless legs syndrome, sleep apnea, insomnia, REM sleep behavior disorder, and narcolepsy. Statistical analyses will include descriptive reports of absolute and relative frequencies of abnormalities in the parameters listed above, as well as prevalence rates of established sleep disorders.
Objective 1.2.
First, the influence of the sleep-related parameters described above on fatigue will be examined using multiple regression. Separate models will be computed for overall fatigue, motor fatigue, and cognitive fatigue. The following covariates will be controlled: disease duration, age, sex, and EDSS. Second, the statistical analyses will include parametric or non-parametric comparisons (depending on data distribution) of (a) total FSMC score, (b) FSMC cognitive fatigue subscale, and (c) FSMC motor fatigue subscale between patients with and without abnormalities in the sleep-related parameters described above, as well as patients with one of the above-mentioned diagnosed sleep disorders.
Objective 1.3.
Subjective fatigue scores obtained from the FSMC will be compared with objective measures of fatigability. Objective fatigability will be operationalized as the difference in hand grip strength (Hand Grip Strength Test) and the change in reaction times in the TAP "Alertness" subtest, each assessed before and after the neuropsychological test battery. Correlation analyses and linear regression models will be used to examine associations between subjective fatigue and objectively measured fatigability. The goal is to identify potential discrepancies between subjective experience and measurable changes, as well as systematic influencing factors.
Objective 1.4.
This analysis will examine whether objectively measured fatigability-assessed via changes in hand grip strength and the TAP "Alertness" subtest before and after neuropsychological testing-differs between patients with and without sleep-related abnormalities. Both motor and cognitive fatigability will be considered. Analyses will be performed using parametric or non-parametric procedures depending on data distribution. Covariates such as age, sex, disease duration, and EDSS will be controlled to exclude potential confounding influences.
Objective 1.5.
Cognitive performance will be operationalized as follows:
Statistical analyses will include:
Objective 1.6.
Cognitive performance will be operationalized in line with 8.3, and analyses will be performed analogously.
Fatigue severity will be operationalized as follows:
Statistical analyses will involve parametric or non-parametric comparisons of mean (sub)scores on the FSMC, Hand Grip Strength Test, and TAP between MS patients with the respective sleep disorders and non-MS control patients with the same sleep disorders, depending on data distribution.
Objective 1.7.
Parametric or non-parametric comparisons of sleep-related parameters will be performed between the respective groups, depending on data distribution. Prevalence rates of different sleep disorders across MS subtypes will be compared using logistic regression, with disease duration, age, sex, and disability level (EDSS) as covariates.
Objective 1.8.
Longitudinal data will be analyzed using mixed-effects models to examine changes in the above-mentioned parameters over time while accounting for intra-individual variability. The models will be adjusted for potential confounding factors, including age, sex, Expanded Disability Status Scale (EDSS) score, disease duration, and comorbid depressive or anxiety disorders.
Objective 1.9.
Mixed-effects models will be used to examine predictors of motor and cognitive fatigue, fatigability, and cognitive decline at follow-up. Both objectively measured sleep parameters derived from polysomnography and patient-reported outcomes (PROs) will be included in the analyses.
Objective 1.10.
Training phase: Supervised learning with manually scored polysomnography data (n ≥ 10,000) from first-night diagnostic recordings.
Validation phase: Application of the trained algorithm to data from new control patients (n ~ 300). Depending on results, either proceed to step 3 or return to step 1.
Testing phase: Application of the algorithm to data from MS patients (n ≥ 382).
All prospective datasets will be analyzed both manually and using AI. During manual scoring, the time required for evaluation will be recorded. Subsequently, mean evaluation times will be compared using parametric or non-parametric group comparisons. Agreement between manual and automated scoring in the diagnosis of sleep disorders will be assessed using Cohen's kappa.
8.2. Analysis plan of the sub-study:
Objective 2.1.
An exploratory comparison of the sleep-related parameters specified under Objective 1 of the main study will be conducted between patients receiving high-efficacy versus low-efficacy therapy using either a t-test or Mann-Whitney U test (depending on data distribution). P-values will be interpreted with great caution and for exploratory purposes only. The focus of the analysis lies on effect sizes. Therefore, for each of the parameters listed above, Hedges' g including the 95% confidence interval for the group difference will be reported.
In addition, the prevalence of established sleep disorders in the respective groups will be assessed and analyzed based on the criteria of the International Classification of Sleep Disorders, following the same procedure described under Objective 1 of the main study.
The statistical analysis will include descriptive reporting of the absolute and relative frequencies of abnormalities in the above-mentioned sleep-related parameters as well as the prevalence rates of established sleep disorders. Furthermore, frequencies between the two groups (high-efficacy vs. low-efficacy therapy) will be compared using the chi-square test or Fisher's exact test. The primary focus will not be on p-values but rather on effect sizes, which will be reported for the 2×2 contingency tables (abnormal vs. normal by group) as the Phi coefficient.
Objective 2.2
The procedures described under Objective 2.1 will be applied analogously to compare sleep parameters and fatigue between patients treated with Ocrelizumab and those receiving other high-efficacy therapies.
Objective 2.3
Analogous to the procedure described under Objective 1.8 of the main study, linear mixed-effects models will be used to evaluate longitudinal changes in the objectively measured sleep parameters, fatigue, and cognition. Therapy type (high vs. low efficacy) and intra-individual variability will be taken into account.
The models will be adjusted for the following potential confounders: age at baseline, sex, EDSS at baseline, disease duration at baseline, depression at baseline, and anxiety at baseline. These covariates will be included in the model, and the basic models will be compared with more complex models including the covariates. Only those covariates that significantly improve model fit will be retained in the final model. If the model fails to converge, the random slope per participant will be removed.
The longitudinal endpoints are the standardized beta coefficients of the group × time interaction in the linear mixed-effects models, quantifying differences in mean change of the outcome variables.
Objective 2.4
Analogous to the procedure described under Objective 2.3, linear mixed-effects models will be used to evaluate longitudinal changes in the objectively measured sleep parameters, fatigue, and cognition. In contrast to the models described under Objective 2.3, the group factor "high vs. low efficacy" will be replaced by "Ocrelizumab vs. other high-efficacy therapy."
Objective 2.5
Analogous to the procedure described under Objective 2.3, linear mixed-effects models will be constructed to evaluate longitudinal changes in the objectively measured sleep parameters, fatigue, and cognition, accounting for therapy type (high vs. low efficacy) and intra-individual variability.
The following polysomnographic (PSG) parameters will be entered separately into the analyses to examine their influence on fatigue and cognition over time:
These parameters will be included as fixed effects and-if supported by the data and model diagnostics-as random slopes per individual.
The models will be adjusted for the following potential confounders: age at baseline, sex, EDSS at baseline, disease duration at baseline, depression at baseline, and anxiety at baseline. These covariates will be included in the model, and the basic models will be compared with more complex models including the covariates. Only covariates that significantly improve model fit will be retained in the final model.
Objective 2.6
Analogous to the procedure described under Objective 2.5, longitudinal trajectories of sleep parameters, fatigue, and cognitive performance will be compared between patients treated with Ocrelizumab and those receiving other high-efficacy therapies. In contrast to the models described under Objective 2.5, the group factor "high vs. low efficacy" will be replaced by "Ocrelizumab vs. other high-efficacy therapy."
Funding
Parts of this study are funded by the German Multiple Sclerosis Society (DMSG). The DMSG will not have access to data from individual study participants but will receive reports solely on the scientific results.
In addition, a substudy conducted in a subset of participants is funded by Roche Pharma AG. The company will receive only fully anonymized data as well as (interim) reports on the scientific results.
Neither the DMSG nor Roche Pharma AG has any influence on treatment decisions or study participation.
Inclusion Criteria:
Exclusion Criteria:
carolin.balloff@gmail.com+4917622970340
phil.albrecht@gmail.com
carolin.balloff@mariahilf.de+492161893026