Sleep disturbances in Taiwanese patients with Parkinson's disease

Abstract Introduction Sleep disturbance is a common nonmotor symptom of Parkinson's disease (PD) and strongly affects patients’ quality of life. The relationship between excessive daytime sleepiness (EDS) and nighttime problems remains uncertain. Arguments persist regarding the risk factors for sleep disturbance among patients with PD. Furthermore, the prevalence of EDS appears to be lower in Asian countries. Herein, we conducted the study to describe the characteristics of sleep problems in a sample of Taiwanese PD patients and delineate the difference with reported sleep disturbances in Caucasian PD patients from the literature. Methods Patients with PD were recruited from the outpatient clinic of a tertiary medical center and were evaluated using standardized assessment protocols, including the Parkinson's Disease Sleep Scale (PDSS), the Pittsburgh Sleep Quality Index (PSQI), Epworth Sleepiness Scale (ESS), and 39‐Item Parkinson's Disease Questionnaire (PDQ‐39). Results A total of 225 patients with PD were recruited. The mean age of patients with PD was 65.7 years old and the mean disease duration was 8.18 years. Among the patients, 53.8% were defined as poor sleepers (PSQI > 5) and 26.3% had EDS. Seventy‐one percent of the poor sleepers used hypnotic medications. The poor sleepers were worse in the scores of Unified Parkinson's Disease Rating Scale (UPDRS), PDSS, and the PDQ‐39, and received higher levodopa daily dosage. A PDSS score of <126 indicate that a patient with PD was a poor sleeper. EDS was positively correlated with advanced Hoehn and Yahr stages and use of dopamine agonists but not with levodopa daily dosage and levodopa equivalent daily dosage. Conclusions Taiwanese patients with PD had a lower prevalence of EDS compared with the literatures reported in Caucasian patients. We identify and suggest that PDSS total score, rather than subcategory items, should be used to predict poor sleep among patients with PD.

Furthermore, few studies have focused on comparing the prevalence of sleep disturbance specifically in Asian and Caucasian populations. Intriguingly, although the reported percentages of poor sleepers in Asian patients are equal to or slightly higher than those among Western populations, the prevalence of EDS is lower in Asian than in Caucasian patients. In the Western countries, such as France, the Netherlands, Switzerland, Canada, and the United States, the percentage of poor sleepers ranged from 58.8% to 63% and EDS ranged from 40.6% to 57% (Brodsky, Godbold, Roth, & Olanow, 2003;Hobson et al., 2002;Louter, Munneke, Bloem, & Overeem, 2012;Poryazova, Benninger, Waldvogel, & Bassetti, 2010;Ratti et al., 2015;Zhu, van Hilten, & Marinus, 2016). In Asia, EDS was found to be considerably less prevalent among patients with PD. In China, Thailand, Singapore, and in the only report regarding Taiwan, the percentages of poor sleepers ranged from 37% to 64.5%, whereas EDS ranged from 15.1% to 32.3% (Chen et al., 2015;Setthawatcharawanich et al., 2014;Tan et al., 2002;Yu et al., 2015). From these perspectives, a larger scale survey is warranted to elucidate ethnic differences in sleep disturbances among patients with PD.
The aims of our study were to investigate the prevalence of daytime and nighttime sleep problems among Taiwanese patients with PD, to explore the predictive factors for sleep disturbance, and to correlate sleep disorders with clinical demographics and motor dysfunction among patients with PD.

| MATERIAL AND METHODS
This was a cross-sectional, questionnaire-based interview investigation. Consecutive patients with PD were recruited from the neurology outpatient clinic of a tertiary medical center in Taiwan We collected demographic data and medication prescribed for each patient on the day of the interview when they visited the outpatient clinic. The severity of PD was evaluated using the UPDRS and the Hoehn and Yahr stage (H&Y stage), and both scores were determined without stopping medication. Nighttime and daytime sleep problems were evaluated using the Pittsburgh Sleep Quality Index (PSQI) (Buysse, Reynolds, Monk, Berman, & Kupfer, 1989) and the Epworth Sleepiness Scale (ESS), respectively (Chen et al., 2002;Johns, 1991).
We evaluated the sleep problems that might be relevant to PD symptoms by using the Parkinson's Disease Sleep Scale (PDSS) (Chaudhuri et al., 2002), and the quality of life of PD patients was assessed using the 39-Item Parkinson's Disease Questionnaire (PDQ-39) (Jenkinson, Fitzpatrick, Peto, Greenhall, & Hyman, 1997).
Following the generally accepted criteria, we use PSQI > 5 to identify PD patients as poor sleepers (Buysse et al., 1989) and ESS ≥ 10 to identify those with daytime sleep disorders (EDS) (Chen et al., 2002;Johns, 1991).
The PDSS is a 15-item questionnaire validated in 2002 to identify nocturnal symptoms in patients with PD. The PDSS initially suggested eight categories to address the following domains: overall quality of night's sleep (item 1), sleep onset and maintenance insomnia (items 2 and 3), nocturnal restlessness (items 4 and 5), nocturnal psychosis (items 6 and 7), nocturia (items 8 and 9), nocturnal motor symptoms (items 10-13), sleep refreshment (item 14), and daytime dosing (item 15) ( Table 1). With advancing understanding of nonmotor symptoms in patients with PD during the past decades, some items of the PDSS could be grouped together and represented by a single domain. For example, numbness (item 10) and pain (item 11) were originally grouped into nocturnal motor symptoms; however, these two items were actually nocturnal nonmotor symptoms of patients with PD. Nocturia (item 8) is autonomic manifestation of PD, and nocturnal restlessness (items 4 and 5) were manifestations of sleep disturbance among patients with PD. Both were closely related to the disease per se of PD.
Some researchers also regrouped different items into special domains to explore aspects of nocturnal disturbance of specific interest (Ray Chaudhuri et al., 2012;Trenkwalder et al., 2011). Therefore, we categorized individual item into five domains to illustrate the different characteristics in the 15 items, and to correlate each domain with nighttime and daytime sleep problems (Table 1).

| Statistical analysis
Statistical analysis of the data was performed using SPSS version 22.0. All clinical characteristics are presented as mean ± SD. Pearson's chi-squared tests were performed to compare categorical variables, and independent sample t tests were used to compare continuous variables between two groups. H&Y stage of two groups was compared using Mann-Whitney U tests. Logistic regression analyses were conducted to determine the risk factors and predictors of sleep disturbance in PD patients. Statistical significance was set at p < .05.
Among the 225 recruited patients, 121 (53.8%) were defined as poor sleepers based on PSQI global scores of >5. Table 2 summarizes comparisons of the clinical characteristics of good and poor sleepers. Poor sleepers had significantly higher scores in the UPDRS total (p < .001), UPDRS part I (p < .001), UPDRS part II (p < .001), UPDRS part III (p = .004), and PDQ-39 (p < .001), lower scores in the PDSS (p < .001), and more levodopa daily dosage (p = .012). The percentage of patients using hypnotic drugs was higher in poor sleepers (p = .008).
Logistic regression was used to correlate and predict the risk factors of poor sleep among patients with PD. Variables with p < .05 were selected, which included UPDRS total score, PDSS, PDQ-39, levodopa daily dosage, and hypnotic drug use. Of the studied variables, PDSS total score had the strongest effect on poor sleepers (p < .001, OR = 0.909). The receiver operating characteristic (ROC) curve indicated that a PDSS score of <126 could predict poor sleep among patients with PD (<126.25, sensitivity 89.4%, 1-specificity 37.2%; Figure 1). Further analysis of the average score for each item in our subgrouping of PDSS showed that poor sleepers had lower scores for sleep quality (8.95 ± 1.16 vs. 6.58 ± 2.50, p < .001), nighttime PD symptoms (8.61 ± 0.96 vs. 7.88 ± 1.44, p < .001), nocturnal psychosis (9.44 ± 1.01 vs. 8.80 ± 1.74, p < .001), and daytime sleepiness (8.89 ± 2.00 vs. 7.79 ± 2.91, p = .001). However, no difference was observed in the nocturnal motor symptoms of good and poor sleepers (9.60 ± 1.12 vs. 9.2 ± 1.45, p = .21; Figure 2 (Table 4). Sixty-three percent of patients in our cohort used hypnotic drugs which included benzodiazepines, Z drugs (zolpidem, zopiclone, and zaleplon), antidepressants, and neuroleptics. No differences were found in the ratio of hypnotic drug use (p = .851) and each category of hypnotic drug between patients with and without EDS (Table 4). We found no significant difference among the subgroups of PDSS, except with the daytime sleepiness subgroup (8.91 ± 1.81 vs. 6.55 ± 3.50, p < .001; Figure 3).

| DISCUSSION
The sleep profiles of 225 consecutive Taiwanese patients with PD in a single medical center were documented. This series had four Furthermore, hypnotic drugs did not increase the occurrence of EDS.
(4) A PDSS total score of <126 was discovered to predict poor sleep among patients with PD.
To the best of our knowledge, this series was the second study of the prevalence of sleep disturbance among Taiwanese patients with PD. Overall, 53.8% of patients were defined as poor sleepers, which was in agreement with the other study of Taiwanese patients by Yu et al. (2015) and similar to Asia and Western series that used the PSQI (Table 3). Although a control group was not recruited in our cohort, another general study of 760 representative Taiwanese individuals between 2010 and 2013 showed that 46.6% were poor sleepers, which is lower than the result obtained from our data (Tai, Wang, & Yang, 2015). Our findings further highlight the importance of increasing the awareness of sleep problems in PD patients.
Medication was an important factor contributing to EDS, and might influence prevalence of EDS. In our series, no differences were observed in the category and dosage of prescribed hypnotics between the EDS and non-EDS groups. The percentages of PD patients receiving benzodiazepines were no difference between the EDS and non-EDS groups in our patient cohort as well as those identified by Zhu et al. (2016) in the Netherland population; however, 43% of patients studied by Zhu et al. had EDS. The ratio of DA utilization in our patients was higher than that reported in the United States (Crispo et al., 2015).
More than 50% of the patients in our series received DAs, which may increase daytime sleepiness. However, the prevalence of EDS among our patients was even lower. Given these findings, ethnicity-specific determinants and genetic predisposition should play some roles in the discrepancy. Some authors have disclosed that patients with the catechol-O-methyltransferase (COMT) Met/Met genotype tend to have higher ESS scores (Frauscher et al., 2004). However, the genotype distribution of COMT Val158Met polymorphism in Caucasians quite differs from that observed in Northeastern Asians (the Val/Val genotype frequencies are 25% and 10%, respectively) (Lachman et al., 1996).  , 2002). However, in our study, a significant correlation was not identified between levodopa dosage and EDS (Table 2), and two other studies also reported no such correlation (Gjerstad, Alves, Wentzel-Larsen, Aarsland, & Larsen, 2006;Kumar, Bhatia, & Behari, 2003). DAs therapy contributes to EDS and sudden somnolence (Hauser, Gauger, Anderson, & Zesiewicz, 2000;Pal, Bhattacharya, Agapito, & Chaudhuri, 2001). Our results confirmed that EDS could be augmented by more DA use and higher DA dosage (  (Table 4). In spite of our findings were in concord with previous reports (Hauser et al., 2000;Pal et al., 2001), the issues of drugs related EDS warrant further in-depth analysis with large sample and well-designed study. It is believed that hypnotics may increase EDS, and medications like amantadine, rasagiline, and selegiline may decrease EDS through the effects of their stimulating metabolite, nevertheless, our analysis could not support the general assumption (Table 4). Benzodiazepines were currently reported to be associated with lower score in SCOPA-SLEEP-Daytime Sleepiness (Zhu et al., 2016). A case-control study described that clonazepam, the most prescribed benzodiazepines in our study, decreased the occurrence of EDS among patients with PD (Shpirer et al., 2006). Except for H&Y stage and DA utilization, we found occurrence of the EDS was not correlated with the UPDRS (disease severity), the PDSS (nocturnal PD symptoms), and the PSQI (Figure 3). Our results supported the notion that EDS represents a separate manifestation of PD and that it is merely dependent on disease rather than on nocturnal disturbance and nighttime sleep problems (Goldman et al., 2013;Suzuki et al., 2008).
Aside from the daytime problem of EDS in patients with PD, nighttime sleep problems also significantly impair the quality of life in PD patients. Identifying risk factors that predict nighttime dysfunction are important for clinical practice. We clarified disease severity (identified by UPDRS), nocturnal PD symptom (identified by PDSS), quality of life (identified by PDQ-39), and medications (levodopa and hypnotics) were correlated with poor sleep in our cohort (Table 2 and Figure 2). We found that the poor sleepers had significant higher UPDRS scores. Our finding contrasted with a community-based study in Norway (Svensson et al., 2012), but is in agreement with a casecontrol study conducted in India (Kumar et al., 2002). In addition, poor sleep was positively correlated with the dosage of levodopa and hypnotics in our and two other studies (Antczak et al., 2013;Verbaan, van Rooden, Visser, Marinus, & van Hilten, 2008). Levodopa can improve sleep efficiency with reduced sleep latency by improving motor scores (Askenasy & Yahr, 1985;Ferreira, Prabhakar, & Kharbanda, 2014;Kales, Ansel, Markham, Scharf, & Tan, 1971). Therefore, instead of higher dosage of levodopa leading to poor sleep, it is possible that complaint of poor sleep made poor sleepers in our cohort receive higher dosage of levodopa than good sleepers. We also discovered that the PDSS had a strong effect on and is a useful predictor of poor sleep among patients with PD. Some researchers have also tried to analyze the correlation of PDSS items and subgroup with poor sleep.
However, these reports have revealed heterogeneous results (Uemura

| Limitation
Our study was designed as a questionnaire-based interview investigation. Lacking of strong objective data such as polysomnography or mean sleep latency tests limited the findings. In addition, patients needed enough mental and physical condition to complete the questionnaires. Therefore, the severity of PD of most patients was not advanced.

ACKNOWLEDGMENTS
We acknowledge Wallace Academic Editing for editing this manuscript. This research was not preregistered in an independent, institutional registry.