Impact of a Ketogenic Diet on Sleep Quality in People with Relapsing Multiple Sclerosis

Abstract

Background:

Sleep disturbance in MS is common and can significantly impair overall quality of life. The ketogenic diet (KD) associates with improved sleep quality in people living with epilepsy and may have similar benefits when used within MS; however, the impact of a KD on sleep in this population remains poorly defined.

Methods:

Forty-five patients with relapsing MS enrolled into a 6-month KD intervention trial and completed self-reported assessments of sleep quality and sleep disorder symptoms prior to diet initiation and while on diet, using the Epworth Sleepiness Scale (ESS) and Sleep Disorders Symptom Checklist-25 (SDS). Participants who did not complete sleep assessments at baseline and 6-months were excluded from analysis. In addition to sleep metrics, data collection included anthropometrics and MS-related fatigue scores.

Results:

Thirty-nine of 45 (87%) participants completed the required sleep assessments. There was a mean reduction in ESS score of 1.90 (95% CI [−2.85, −0.94], p<0.001). Total SDS score decreased at 6-months on KD (−4.4, 95% CI [−7.1, −1.7], p=0.002), with improvements noted in insomnia (−1.55, 95% CI [−2.66, −0.43], p=0.008), obstructive sleep apnea (−0.91, 95% CI [−1.57, −0.25], p=0.008), and restless leg syndrome screening scores (−1.00, 95% CI [−1.95, −0.051], p=0.04). Sleep duration was unchanged on KD.

Conclusion:

KD associates with improvements in daytime sleepiness, independent of sleep duration, and common comorbid sleep disorders in people living with relapsing MS. The findings herein support the benefits of KD on sleep quality and highlight the potential role of dietary therapeutics for sleep disorders in neurological disease.

1. INTRODUCTION

Multiple sclerosis (MS) is an inflammatory neurodegenerative disease that associates with a wide range of symptoms and comorbid conditions that interfere with daily functioning. Approximately 50% of persons with MS (pwMS) have some type of sleep disturbance, which contributes to reduced quality of life.^1–3 Specifically, insomnia, restless leg syndrome (RLS), poor sleep quality, and excessive daytime sleepiness (EDS) are strong predictors of fatigue – a symptom that affects up to 90% of pwMS and significantly contributes to reduced quality of life.⁴ The most common sleep disorders in pwMS include obstructive sleep apnea (OSA), RLS, and insomnia.^2,3

The effects of disordered sleep on metabolic, vascular, and immune function may further contribute to disease progression, poor quality of life, and elevated risk for comorbidities in pwMS.⁵ Specifically, sleep deprivation is associated with an increase in pro-inflammatory cytokines (e.g. IL-1, TNF-α), activation of resident cerebral astrocytes and microglia, and promotion of pro-inflammatory T cell lineages, all of which are implicated in the pathophysiology of MS.⁵ Serum pro-inflammatory cytokine levels generally decrease during sleep, and sleep promotes anti-inflammatory regulatory T cell (Treg) function.⁶ Further, IL-1, TNF-α, and the pro-inflammatory adipocytokine, leptin, have been identified as regulators of the sleep cycle.^6,7

Dietary intake influences immune function through direct effects of macro- and micronutrients and via the gut-immune interface.⁸ Diets that mimic a fasting state show evidence of reduced pro-inflammatory cytokines, reduced serum Th1 and Th17 cells, and increased Treg cells in an MS mouse model.⁹ Ketogenic diets (KDs) are very low carbohydrate diets that are high in fatty acids and adequate in protein, mimicking a nutritional fasting state through the production of biologically-active ketone bodies. We have previously shown that the modified Atkins KD (KD^MAD) is well-tolerated in pwMS, resulting in improved body composition and quality of life in addition to reduced fatigue, depression, and serum leptin.¹⁰ Moreover, KDs may positively influence mechanisms underlying MS neurodegeneration.^11,12 Importantly, KDs are sustainable outside of a clinical trial and induce positive dietary pattern changes.¹³

Growing data suggests that KDs improve sleep quality and induce beneficial alterations in the sleep cycle. For example, KDs have been linked with an improved proportion of rapid eye movement (REM) sleep and overall sleep quality in children with epilepsy and reduced sleep complaints in people with migraine.^14,15 Further, a recent study showed that KDs associate with improved sleep quality and reduced daytime sleepiness in a small cohort of minimally-disabled pwMS.¹⁶ The effect of KDs on common sleep disorders in pwMS remains underexplored. Given the existing data suggesting a link between a KD and sleep, we sought to evaluate the impact of a 6-month KD^MAD intervention on sleep disorder symptoms in a well-characterized cohort of pwMS.

2. METHODS

2.1. Ethics Approval

This study involves human participants and was approved by the University of Virginia (UVA) Institutional Review Board for Health Sciences Research (study ref ID: 20877, date: 08/14/2018). All participants provided informed consent prior to the start of any study-related procedures.

2.2. Study Participants

Sixty-five patients with relapsing MS¹⁷ were enrolled into a prospective, intention-to-treat KD^MAD intervention. From this cohort, the latter 45 participants were asked to complete exploratory assessments of sleep quality and sleep disorder symptoms. To be eligible for this trial, participants had to demonstrate disease stability on current disease modifying therapy (i.e. no clinical relapses and/or new/enlarging T2-hyperintense lesions on MRI) for ≥6 months prior to enrollment. Participants needed to have an Expanded Disability Status Scale (EDSS) score of ≤6.0 to be included. Participants were excluded if they had progressive MS, had been on a KD in the 6 months prior to enrollment, were pregnant/planning a pregnancy, or were underweight by Centers for Disease Control (CDC) criteria.¹⁸ Eligible patients were identified and contacted for participation through the UVA MS clinical database.

2.3. Study Procedures

For detailed study procedures regarding the KD trial protocol, please refer to the primary trial publications.^10,19 In brief, the study entailed assessments at baseline (pre-diet) and three study visits (on-diet) at 1, 3, and 6 months. At baseline, participants met with the study dietitian, who provided education on initiating and maintaining a KD^MAD. As part of KD^MAD teaching, participants were instructed to restrict net carbohydrates to <20 grams/day and were encouraged to increase healthy fat intake. Adherence to KD^MAD was assessed using daily urine ketone test strips. Participants were required to send a dated photograph of their daily ketone test strip to the study team. Participants were adherent for that study day if the photographed strip demonstrated evidence of ketosis. A ketone-negative strip or absence of dated photographic evidence of ketosis were marked as non-adherent days.

Sleep disorder screening assessments were performed at baseline (pre-diet) and at 3 and 6 months on-diet. Only those participants who completed the study and sleep assessments at both baseline and 6 months were included for analysis. Assessment tools included the Epworth Sleepiness Scale (ESS) and the Sleep Disorders Symptom Checklist-25 (SDS). The ESS is validated to evaluate average sleep propensity as a marker of disordered sleep.²⁰ Participants are asked to rate their chances of falling asleep during each of 8 activities on a 4-point scale (0–3). Scores range from 0 to 24, and an ESS score of >10 denotes excessive daytime sleepiness (EDS).²¹ Given that pwMS are at risk for multiple types of sleep disorders, we utilized the SDS to screen for possible sleep pathology that may prove responsive to a KD. The SDS is a 25-item tool that screens for multiple sleep disorders including insomnia, obstructive sleep apnea (OSA), restless leg syndrome (RLS)/ periodic limb movement disorder (PLMD), circadian rhythm disorders (advanced sleep phase and delayed sleep phase syndromes), narcolepsy, and multiple parasomnias (nightmare disorder, night terror disorder, REM sleep behavior disorder, and sleep-related temporomandibular joint (TMJ) dysfunction).²² Using questions grouped by disorder, participants are asked to identify how often they experience each symptom on a 5-point scale, ranging from “never” (0) to “>5 times/week” (4). We performed our analysis using the sum of the items in each group to reflect each participant’s sub-score, with larger sums indicating a greater likelihood of having a given sleep disorder.^22,23 Positive screens for sleep disorders were determined based on cut-point values obtained from receiver operating characteristic (ROC) curve inflections.²² Ultimately, any sub-score of ≥3 points (experiencing a symptom ≥3 times per week) was determined to be a positive screen for all sleep disorders, except insomnia (where a sub-score of ≥5 indicated a positive screen). Importantly, a positive screen on the SDS is not diagnostic but rather suggests which patients should undergo formal evaluation for a given sleep disorder.

Participants completed 2-week sleep diaries prior to each study visit. Diaries denoted bedtime, estimated sleep latency, wake onset time, sleep duration, number and duration of nocturnal awakenings, number and duration of naps, and number of caffeinated beverages ingested each calendar day. To describe nocturnal awakenings, average Wake After Sleep Onset (WASO) was calculated by dividing the total number of minutes spent awake after falling asleep by the number of days in each diary. For sleep diary analyses, only participants that completed both baseline and 6-month diaries with at least 12 recorded days in each diary were included. Participants who consistently worked nightshift were included for all analyses except “bedtime” and “waketime.” Participants were excluded from sleep diary analysis if there was a switch between dayshift and nightshift work between study visits.

As a measure of body composition, participants underwent air displacement plethysmography (BOD POD) to quantify lean mass versus fat mass at baseline and at 6 months on KD. Fasting blood work was completed at study visits, including leptin levels.

Patient-reported outcomes (PROs) were completed at baseline and at on-diet study visits. Fatigue mpact and severity were assessed with Modified Fatigue Impact Scale (MFIS) and MS Fatigue Severity Scale (MSFSS).^24,25 Higher MFIS and MSFSS scores indicate greater impact and severity of fatigue, respectively. Other PROs included assessments of depression (Beck’s Depression Inventory 1A (BDI)) and quality of life (MS Quality of Life-54 (MSQoL)). Higher BDI scores indicate greater degrees of depression, while lower MSQoL scores indicate poorer quality of life.^26,27

2.4. Statistical Analyses

For continuous outcomes, significance tests and confidence intervals for changes from baseline at 6 months were based on the paired t-test. Nearly identical results were obtained using the nonparametric Wilcoxon signed ranks test. For binary variables, such as the proportion of participants scoring 11 or above on the ESS, changes from baseline were assessed using statistical methods for matched binary outcomes.²⁸ Pearson correlations were used to estimate the association between changes in sleep outcomes, dietary outcomes, change in leptin, and quality of life measures. Similar results were obtained using Spearman rank correlations. For outcomes measured at baseline, 3, and 6 months, repeated measures models were used. Contrasts within the repeated measures model were used to make specific comparisons of changes from baseline at 3 and 6 months. Analyses were carried out in SAS 9.4.

3. RESULTS

3.1. General Characteristics:

Of the 45 participants who were asked to complete sleep assessments, 39 (87%) completed the required baseline (pre-diet) and 6-month (on-diet) assessments. Of the participants who lacked 6-month assessments, 4 were lost to follow up, 1 withdrew due to a separate medical condition, and 1 completed baseline assessment but never began the KD intervention. Baseline ESS and SDS scores were not significantly different between those who did (n=39) and those who did not (n=6) complete 6-month assessments, except for baseline insomnia score, which was significantly greater in the latter group (Appendix).

Demographics for the study cohort (n=39) are shown in Table 1. Additionally, we show the change in potentially relevant sleep quality-related measures from baseline to 6-months on KD, including anthropometrics, indices of fatigue (MFIS and MSFSS), depression (BDI), quality of life (MSQoL-54), and serum leptin.

Table 1. Clinical characteristics of the study cohort.

Anthropometric, patient-reported, and clinical outcomes for participants using intention-to-treat analysis pre- and post-KD intervention.

	Baseline (pre-KD)	Mean change (95% CI)	P-value
Number of Participants	39	---
Age (years) median [range]	40 [20 – 54]	---
Sex, n female (%)	33 (85%)	---
Race
White, n (%)	33/39 (85%)	---
Black, n (%)	6/39 (15%)	---
Ethnicity
Hispanic, n (%)	3/39 (8%)	---
BMI, mean ± SD, kg/m 2	31.9 ± 5.3	−3.5 (−4.2, −2.8)	<0.001
Total Fat Mass, mean ± SD, kg	37.6 ± 12.8	−8.1 (−9.8, −6.3)	<0.001
EDSS, median [range]	2.50 [1.0 – 6.0]	−0.4 (−0.54, −0.23)	<0.001
MFIS, mean ± SD	33.5 ± 19.0	−15.4 (−20.3, −10.4)	<0.001
MSFSS, mean ± SD	22 ± 8.7	−3.4 (−5.3, −1.6)	0.001
BDI, mean ± SD	8.8 ± 5.1	−4.2 (−5.7, −2.6)	<0.001
MS Quality of Life-54, mean ± SD
Mental Sub-score	70.9 ± 17.1	11.2 (5.3, 17.2)	<0.001
Physical Sub-score	66.7 ± 15.8	12.0 (7.5, 16.5)	<0.001
Leptin, mean ± SD, ng/mL	25.6 ± 16.8	−13.0 (−16.8, −9.1)	<0.001

Mean change= 6 month – baseline. KD= ketogenic diet; EDSS= Expanded Disability Status Scale; MFIS= Modified Fatigue Impact Scale; MSFSS= and MS Fatigue Severity Scale; BDI= Beck’s Depression Inventory 1A.

3.2. Excessive Daytime Sleepiness:

There was a reduction in mean ESS score at 6 months on-diet of 1.9 points (95% CI [−2.85, −0.94], p<0.001) (Table 2). The frequency of EDS (i.e. ESS score >10) decreased from 15.4% to 7.7% (95% CI [−19.7, 3.2], p=0.170) (Figure 1, Table 3). Of the 6 participants who were EDS positive at baseline, 4 had normalization of the ESS score. A single participant who was EDS negative (ESS=8) at baseline converted to EDS positive (ESS=11) at 6 months. The last 2 participants remained EDS positive at 6 months. Mean change in BMI was not statistically different between those who had resolution of EDS and those who did not (−4.7 vs −3.3 kg/m², p=0.31).

Table 2.

Sleep assessment scores of the study cohort.

	Baseline mean ± SD	6-Month mean ± SD	Mean change (95% CI)	P-value
ESS Score	6.64 ± 3.88	4.74 ± 3.08	−1.90 (−2.85, −0.94)	<0.001
SDS Total Score	22.7 ± 11.4	18.3 ± 10.2	−4.4 (−7.1, −1.7)	0.002
SDS Insomnia Score	7.19 ± 3.69	5.64 ± 3.61	−1.55 (−2.66, −0.43)	0.008
SDS OSA Score	3.21 ± 3.72	2.29 ± 2.78	−0.91 (−1.57, −0.25)	0.008
SDS RLS/PLMD Score	3.62 ± 3.27	2.61 ± 3.06	−1.00 (−1.95, −0.051)	0.040
SDS Narcolepsy Score	0.72 ± 1.69	0.62 ± 1.29	−1.10 (−0.55, 0.35)	0.648
SDS Parasomnia Score	2.56 ± 2.38	2.10 ± 2.27	−0.46 (−0.95, 0.03)	0.065
SDS Circadian Rhythm Disorder Score	1.26 ± 1.29	1.33 ± 1.42	0.08 (−0.47, 0.63)	0.778

ESS= Epworth Sleepiness Scale; SDS= Sleep Disorders Symptom Checklist-25; OSA= obstructive sleep apnea; RLS= restless leg syndrome; PLMD= periodic limb movement disorder; SD= standard deviation.

Figure 1. Proportion of positive screens for sleep disorders for the study cohort per the ESS and SDS assessments.

Values represent % of the cohort with error bars representing standard error. Cutoffs were derived from receiver operating characteristic (ROC) curve inflections per Klingman et al.¹⁷ EDS= excessive daytime sleepiness; SDS= Sleep Disorders Symptom Checklist-25; OSA= SDS obstructive sleep apnea score; RLS= SDS restless leg syndrome score; CRD= SDS circadian rhythm disorder score. EDS was defined as an Epworth Sleepiness Scale (ESS) score of >10.

Table 3. Proportion of positive screens for sleep disorders for the study cohort.

Values represent % of the cohort with standard error. SDS insomnia sub-score represents the sum of questions 3–6, OSA – the sum of questions 10–14, RLS/PLMD – the sum of questions 15–17, narcolepsy – the sum of questions 18–20, parasomnia – the sum of questions 21–24, and circadian rhythm disorder – the sum of questions 7 and 8. Cutoffs were derived from receiver operating characteristic (ROC) curve inflections per Klingman et al.¹⁷

	Baseline proportion (% ± SE)	6-Month proportion (% ± SE)	Difference (95% CI)	P-value
EDS (ESS>10)	15.4 ± 5.8	7.7 ± 4.3	−7.7 (−19.7, 3.2)	0.170
SDS Insomnia (sub-score ≥5)	79.5 ± 6.5	51.3 ± 8.0	−28.2 (−44.0, −12.4)	<0.001
SDS OSA (sub-score ≥3)	43.6 ± 7.9	28.2 ± 7.2	−15.4 (−26.7, −4.1)	0.008
SDS RLS/PLMD (sub-score ≥3)	56.4 ± 7.9	30.8 ± 7.4	−25.6 (−44.1, −7.2)	0.006
SDS Narcolepsy (sub-score ≥3)	10.3 ± 4.9	10.3 ± 4.9	0 (−7.1, 7.1)	1.00
SDS Parasomnia (sub-score ≥3)	46.2 ± 8.0	38.5 ± 7.8	−7.7 (−20.8, 5.4)	0.249
SDS Circadian Rhythm Disorder (sub-score ≥3)	15.4 ± 5.8	20.5 ± 6.5	5.1 (−12.2, 22.5)	0.562

EDS= excessive daytime sleepiness; ESS= Epworth Sleepiness Scale; SDS= Sleep Disorders Symptom Checklist-25; OSA= obstructive sleep apnea; RLS= restless leg syndrome; PLMD= periodic limb movement disorder.

3.3. Sleep Disorder Symptoms:

There was a reduction in the mean total SDS score at 6 months on-diet (−4.4 points, 95% CI [−7.1, −1.7], p=0.002) (Table 2). Several SDS sub-scores were reduced at 6 months compared to baseline, including mean insomnia (−1.55, 95% CI [−2.66, −0.43], p=0.008), OSA (−0.91, 95% CI [−1.57, −0.25], p=0.008), and RLS (−1.00, 95% CI [−1.95, −0.051], p=0.04) scores (Figure 2, Table 2). There were no significant differences in mean sub-score change for narcolepsy, parasomnias, or circadian rhythm disorders.

Figure 2. Changes in SDS sub-scores from baseline to 6 months on-diet.

For graphical purposes, this plot represents the average change as a proportion of the baseline average. For example, the average insomnia score at baseline is 7.2 and the average change is −1.54, so the average change is: −1.54 / 7.2 = −0.21. SDS= Sleep Disorders Symptom Checklist-25; OSA= obstructive sleep apnea; RLS= restless leg syndrome; PLMD= periodic limb movement disorder.

The proportion of positive SDS screens for insomnia (sub-score ≥5) was reduced at 6 months on-diet compared to baseline (−28.2%, 95% CI [−44.0, −12.4], p<0.001) (Figure 1, Table 3). The same was true for OSA (sub-score ≥3) (−15.4%, 95% CI [−26.7, −4.1], p=0.008) and RLS (sub-score ≥3) (−25.6%, 95% CI [−44.1, −7.2], p=0.006). There was no difference in positive screens for narcolepsy, parasomnias, or circadian rhythm disorders (sub-scores ≥3) at 6 months.

3.4. Sleep Diaries:

Mean sleep duration, assessed by sleep diaries, was not different from baseline to 6 months on-diet (Table 4). We observed a small reduction in mean WASO (i.e., minutes spent awake after sleep onset per calendar day) from baseline to 3 months on-diet (−8.49 min/day, 95% CI [−15.2, −1.8], p=0.015), though this was not significant at 6 months (−7.06 min/day, 95% CI [−14.8, 0.07], p=0.074). Compared to baseline, there were marginally fewer mean naps per day at 6-months on diet (−0.1 naps/day, 95% CI [−0.13, −0.01], p=0.028). There was also a trend toward reduced caffeine intake at 3 months (−0.25 drinks/day, 95% CI [−0.54, 0.03], p=0.081) but this finding did not persist at 6 months (−0.06 drinks/day, 95% CI [−0.34, 0.23], p=0.687). Compared to baseline, there were no differences in bedtime, wake onset time, or nap length at 3 or 6 months on-diet (Table 4).

Table 4.

Sleep diary data for the study cohort.

	Baseline mean ± SD	3 Months Mean ± SD	3 Month - Baseline (95% CI)	P-value	6-Month Mean ± SD	6 Month - Baseline (95% CI)	P-value
Sleep Duration (hours) per night	7.36 ± 0.83	7.62 ± 0.94	0.25 (−0.03, 0.54)	0.082	7.58 ± 0.83	0.23 (−0.06, 0.52)	0.117
Bedtime (hours from 12:00 AM)	23.0 ± 0.80	23.0 ± 0.90	−0.05 (−0.32, 0.23)	0.730	22.9 ± 0.96	−0.14 (−0.37, 0.09)	0.217
Wake Onset Time (hours from 12:00 AM)	7.49 ± 1.08	7.46 ± 1.20	−0.01 (−0.25, 0.23)	0.961	7.39 ± 1.13	−0.63 (−0.36, 0.19)	0.533
Nocturnal Awakenings (# per night)	0.52 ± 0.62	0.29 ± 0.49	−0.19 (−0.35, −0.02)	0.027	0.37 ± 0.57	−0.15 (−0.32, 0.02)	0.074
Nights with Nocturnal Awakening (# over 2-week diary**)	0.35 ± 0.31	0.21 ± 0.27	−0.11 (−0.20, −0.02)	0.013	0.26 ± 0.34	−0.09 (−0.18, 0.01)	0.089
WASO (average minutes per night)	29.1 ± 35.7	18.4 ± 29.3	−8.49 (−15.2, −1.8)	0.015	22.0 ± 36.7	−7.06 (−14.8, 0.07)	0.074
Nap Length (average minutes per nap)	47.8 ± 46.7	38.5 ± 52.5	−12.5 (−31.9, 7.0)	0.202	43.4 ± 62.2	−4.4 (−29.8, 21.0)	0.728
Naps (# per day)	0.13 ± 0.18	0.12 ± 0.22	−0.01 (−0.09, 0.06)	0.704	0.07 ± 0.11	−0.07 (−0.13, −0.01)	0.028
Caffeine per day (drinks/day)	1.48 ± 1.42	1.33 ± 0.94	−0.25 (−0.54, 0.03)	0.081	1.43 ± 1.08	−0.06 (−0.34, 0.23)	0.687

^**Several diaries contained 12 or 13 days of data, and therefore the number of days recorded was used as the denominator. WASO= time spent awake after sleep onset; SD= standard deviation.

3.5. Correlation Analyses:

Demographic factors (age, sex) and changes in body composition (ΔBMI and Δfat mass) were not correlated with outcome measures of interest (ΔESS, ΔSDS or sub-scores, ΔWASO, and Δnaps/day). Improvement in ESS directly correlated with improvement in SDS total score (r=0.36, p=0.03) and RLS sub-score (r=0.36, p=0.03). Reductions in ESS scores moderately correlated with improvements in fatigue severity (0.40, p=0.012), depression scores (r=0.40, p=0.011), and quality of life mental (r=−0.44, p=0.005) and physical (r=−0.53, p=0.001) scores. Reduction in the total SDS score correlated with improvement in fatigue impact (r=0.40, p=0.011), depression (r=0.49, p=0.002), and quality of life physical score (r=−0.39, p=0.012). Selected sleep diary outcomes (ΔWASO and Δnaps/day) were not correlated with any of the fatigue, depression, or quality of life PROs.

Higher leptin levels correlated with higher ESS scores at baseline and 6 months on diet (r=0.41, p=0.005; r=0.47, p=0.002). Change in leptin moderately correlated with change in RLS sub-score (r=0.41, p=0.009).

4. DISCUSSION:

Herein, we demonstrate that a KD leads to improvements in daytime sleepiness and screening symptoms of insomnia, OSA, and RLS in people with relapsing MS. Our findings support recently published data demonstrating that a KD associates with improved sleep quality and reduced EDS in a small cohort of pwMS with minimal disability (EDSS <2.5).¹⁶ Our study expands on this data in a larger MS cohort exhibiting more significant neurological disability (EDSS ≤6.0, median baseline EDSS 2.50 [range 1.0–6.0]). We provide novel data to support KD-induced improvements in symptoms associated with OSA and RLS - two of the most common sleep disorders in pwMS.^2,3 Further, these improvements are independent of sleep duration.

We show that the KD^MAD intervention is associated with a reduced ESS score of nearly 2 points after 6 months of diet. While data from OSA trials suggest that an ESS change of 2 is a clinically-important improvement in daytime sleepiness²⁹, the clinical significance of this score change in an MS population without formally-diagnosed OSA is unknown. Further, 4 of 6 patients in our cohort had resolution of EDS at 6 months on-diet. This finding was not statistically significant, which may be due, in part, to the relatively small percentage of patients who screened positive for EDS at baseline. The improvement in daytime sleepiness on KD^MAD is further supported by the trend toward decreased caffeine intake and time spent napping while on the KD^MAD. Future studies that aim to primarily study the impact of a KD on sleep quality should consider a target population of pwMS exhibiting an ESS >10 at baseline.

While our study was not designed or equipped to diagnose specific sleep disorders, we found that the KD^MAD was associated with fewer positive screens for insomnia, OSA, and RLS. Rates of positive screens for parasomnias (nightmares, night terrors, REM sleep behavior disorder, and sleep related TMJ dysfunction), narcolepsy, and circadian rhythm disorders were not different after the diet intervention. Overall, this data indicates that KD^MAD may be beneficial as an adjunctive treatment of insomnia, OSA, and RLS in pwMS.

At baseline, there was a high prevalence of sleep-related symptoms in our cohort, with nearly 80% screening positive for insomnia, 43% for OSA, and 56% for RLS. Of note, the improvement in sleep disorder symptoms did not correlate with degree of weight loss or reduction in plethysmography-measured total body fat mass. This is consistent with prior data showing that KDs improve sleep quality, independent of BMI change.³⁰ Interestingly, only 15% of our cohort screened positive for EDS, which is consistent with estimates of EDS prevalence in the general population (10–20%).³¹

The lack of change in sleep duration pre- and on-diet is consistent with recent findings that fatigue and other sleep disorder symptoms are more dependent on sleep quality than sleep duration.⁴ Poor sleep quality, particularly in the setting of RLS, is a significant risk factor for MS fatigue.⁴ Our data supports the hypothesis that treatments aimed at improving sleep quality by treating sleep disorder symptoms will improve daytime sleepiness and fatigue in pwMS.

Potential mechanisms of a KDs’ effect on sleep quality include normalization of the sleep cycle, consolidation of sleep through circadian synchrony, and/or effects on bioenergetic and endocrine pathways.³² KDs associate with increased proportion of slow wave sleep (SWS), decreased REM, and overall improved sleep efficiency in healthy participants.³³ Regarding the connection between bioenergetics and sleep, ketones yield ATP more efficiently than glucose, with evidence supporting increased brain ATP and its breakdown product, adenosine.³⁴ Adenosine acts as an endogenous sleep factor that is strongly associated with SWS.³⁴ Meanwhile, orexin (a neuropeptide secreted by the hypothalamus) levels increase when fasting, lending support to increased orexin production during the metabolically similar state of ketosis.³⁵ Orexin signaling is crucial in promoting wakefulness and also prevents the transition into REM sleep.³⁶ Therefore, KDs may improve sleep quality and reduce EDS through effects on the sleep cycle, mediated by adenosine and orexin.

Leptin is an adipocytokine that is of particular interest in MS due to its ability to stimulate pro-inflammatory cytokine production and leukocyte proliferation.⁷ The observed decrease in leptin is partly explained by the reduction in fat mass in our cohort.¹⁰ Leptin signaling also influences the sleep-wake cycle.⁷ Serum leptin concentrations increase during the first part of the night, promoting sleep through antagonism of orexin. Several studies have demonstrated that sleep deprivation is associated with hypoleptinemia in healthy people – a relationship that is at least partly mediated by changes in the sympathovagal balance. Conversely, multiple studies in people who have, or are at high risk for diabetes, found that sleep deprivation was associated with increased leptin levels.⁷ Importantly, leptin transport across the blood-brain barrier is dependent on receptor expression, and thus, leptin resistance can occur in certain disease states (such as diabetes). Therefore, leptin resistance is one possible explanation for the correlation between leptin and ESS in our study – while serum leptin was elevated at baseline, it was unable to exert its effect to promote sleep. This is consistent with prior data showing that KDs improve leptin resistance and suggests that this mechanism may partly explain the observed improvements in sleep.

MS-related fatigue is a predictor of unemployment and reduced QoL, and thus, there is strong interest in developing treatments to target this considerable problem.³⁷ The strong connection between sleep quality and fatigue in pwMS is widely accepted.^2,4 Further, our previously published data demonstrated a significant improvement in MS-related fatigue in pwMS on a KD.¹⁰ In the current study, reduced daytime sleepiness correlated with improved fatigue severity. Improvement in SDS total score correlated with reduced fatigue impact. While the fatigue benefit observed on KD is partly explained by improvements in sleep quality, there may be a relationship between ketosis and improved fatigue that is independent of sleep quality. Moreover, poor sleep quality is also associated with worsened depression.³⁸ Our data demonstrates that reduced daytime sleepiness and sleep disorder symptoms are associated with improved depression scores. Conversely, improvement in depression scores strongly correlate with improvements in the total sleep disorder symptoms score. This is consistent with prior data supporting the bi-directional relationship between sleep and depression.³⁹ While inferences regarding causality are not feasible in the current study, KDs may improve depression in pwMS through its effect on sleep, or vice versa. Additionally, KDs may have direct effects on both depression and sleep.

We present these results amid growing interest in the application of dietary interventions in MS. In tandem to the observed relationship between diet and MS symptoms, there are likely direct dietary effects on inflammation and overall brain health. Notably, dietary intake influences the gut microbiome, which in turn, influences host immune function via effects on the intestinal barrier, regulation of endogenous factors, and modulation of effector and regulator T cell lineages.⁴⁰ The microbiome also influences nervous system function through neural, endocrine, and metabolic pathways - a network known as the “gut-brain axis.”^41,42 KDs have been shown to reduce oxidative stress, up-regulate gene expression of anti-inflammatory molecules, and reduce Th1 and Th17 pro-inflammatory responses - effects that may be mediated by the KD influence on the microbiome.^9,43–45 These pathophysiological findings, combined with the positive safety/ tolerability profile, promotion of weight loss, and benefits for MS symptoms, explain why KDs have become popular in this arena.

The main limitation of this study is the lack of polysomnography (PSG) to formally diagnose sleep disorders (particularly OSA, PLMS/PLMD, parasomnia, and narcolepsy) in those that screened positively on self-report questionnaires. It is important to note that our study was not designed to diagnose individual sleep disorders but rather explore patterns of sleep pathology and their response to a diet modification in pwMS. The current findings underscore the need for PSG to confirm the reported benefits of KD on sleep quality and specific sleep disorders in pwMS. Further, our study lacked an age, sex, and BMI-matched control-diet group of pwMS who have a similar prevalence of sleep disorder symptoms. This study was restricted to clinically-stable relapsing pwMS and thus, our findings may not be generalizable to a population with actively relapsing or progressive MS.

In conclusion, the findings from this observational study demonstrates that a KD associates with improvements in daytime sleepiness, independent of sleep duration, in mild-to-moderately disabled people with relapsing MS. This effect may be due to reduced sleep disorder symptoms, reduced nighttime awakenings, and/or improved sleep efficiency/normalization of the sleep cycle. Randomized controlled trials that utilize objective measures of disordered sleep, such as PSG and actigraphy, are needed to confirm these results to fully ascertain the impact of KDs in the treatment of MS-related sleep disorders.

Highlights:

• Ketogenic diet reduces daytime sleepiness in people with multiple sclerosis (MS)

• Ketogenic diet associates with improved screening scores for common sleep disorders in MS

• Improvements in sleep disorder screening are not explained by sleep duration change or weight loss

• Sleep quality improvements may be explained by reduced awakenings and sleep cycle normalization

Appendix.

Sleep metrics for subjects who did and did not complete the 6-month dietary intervention.

	N		Baseline mean ± SD		P-value
	Completed	Did Not Complete	Completed	Did Not Complete
ESS Score	39	6	6.6 ± 3.9	7.3 ± 4.6	0.69
SDS Total Score	36	4	20.3 ± 10.4	31.8 ± 21.6	0.37
SDS Insomnia Score	38	5	7.2 ± 3.7	11.6 ± 4.8	0.02
SDS OSA Score	38	6	3.2 ± 3.7	2.2 ± 2.1	0.59
SDS RLS/PLMD Score	39	6	3.6 ± 3.3	4.7±3.4	0.47

ESS= Epworth Sleepiness Scale; SDS= Sleep Disorders Symptom Checklist-25; OSA= obstructive sleep apnea; RLS= restless leg syndrome; PLMD= periodic limb movement disorder; SD= standard deviation.