Sleep and risk of parkinsonism and Parkinson's disease: a population-based study

Abstract

Sleep disturbances may signal presence of prodromal parkinsonism, including Parkinson's disease. Whether general sleep quality or duration in otherwise healthy subjects is related to the risk of parkinsonism remains unclear. We hypothesized that both worse self-reported sleep quality and duration, as well as a longitudinal deterioration in these measures, are associated with the risk of parkinsonism, including Parkinson's disease. In the prospective population-based Rotterdam Study, we assessed sleep quality and duration with the Pittsburgh Sleep Quality Index in 7726 subjects (mean age 65 years, 57% female) between 2002 and 2008, and again in 5450 subjects between 2009 and 2014. Participants were followed until 2015 for a diagnosis of parkinsonism and Parkinson's disease. Outcomes were assessed using multiple modalities: interviews, physical examination, and continuous monitoring of pharmacy records and medical records of general practitioners. We used Cox regression to associate sleep, and changes in sleep over time, with incident parkinsonism and Parkinson's disease, adjusting for age, sex, education and smoking status. Over 64 855 person-years in 13 years of follow-up (mean: 8.4 years), 75 participants developed parkinsonism, of whom 47 developed Parkinson's disease. We showed that within the first 2 years of follow-up, worse sleep quality {hazard ratio (HR) 2.38 per standard deviation increase [95% confidence interval (CI 0.91-6.23)]} and shorter sleep duration [HR 0.61 per standard deviation increase (95% CI 0.31-1.21)] related to a higher risk of parkinsonism. Associations of worse sleep quality [HR 3.86 (95% CI 1.19-12.47)] and shorter sleep duration [HR 0.48 (95% CI 0.23-0.99)] with Parkinson's disease were more pronounced, and statistically significant, compared to parkinsonism. This increased risk disappeared with longer follow-up duration. Worsening of sleep quality [HR 1.76 per standard deviation increase (95% CI 1.12-2.78)], as well as shortening of sleep duration [HR 1.72 per standard deviation decrease (95% CI 1.08-2.72)], were related to Parkinson's disease risk in the subsequent 6 years. Therefore, we argue that in the general population, deterioration of sleep quality and duration are markers of the prodromal phase of parkinsonism, including Parkinson's disease.

Keywords: PSQI; cohort; epidemiology; sleep.

Figure 1

Associations of sleep quality and duration with risk of parkinsonism and Parkinson’s disease, per cumulatively increasing duration of follow-up. The associations of (A) sleep quality and (B) sleep duration with incident parkinsonism and Parkinson’s disease are shown for cumulatively increasing follow-up duration within the study timeframe. HR estimates were obtained from multivariate Firth’s penalized Cox regression models by censoring all participants still at risk at 2, 4, 6, 8 and 10 years after baseline, and after the total follow-up of 13 years. HR estimates were adjusted for age at baseline, sex, educational level and smoking status, are expressed per standard deviation increase of (A) worse sleep quality, or (B) longer sleep duration, and are plotted at a (A) logarithmic (base 2) scale and (B) a linear scale. PD = Parkinson’s disease.

Figure 2

Associations of PSQI component scores with risk of parkinsonism and Parkinson’s disease, per cumulatively increasing duration of follow-up. The associations of the PSQI components (A) quality, (B) latency, (C) efficiency, (D) disturbances, (E) sleep medication, and (F) daytime dysfunction with incident parkinsonism and Parkinson’s disease are shown for cumulatively increasing follow-up duration within the study timeframe. HR estimates were obtained from multivariate Firth’s penalized Cox regression models by censoring all participants still at risk at 2, 4, 6, 8 and 10 years after baseline, and after the total follow-up of 13 years. Estimates are adjusted for age at baseline, sex, educational level, and smoking status, are expressed per category increase in component score, and are plotted at different logarithmic (base 2) scales per component. For parkinsonism analyses, we included following numbers of participants: (A) 7716, (B) 7718, (C) 7473, (D) 6840, (E) 7725, (F) 7689 (samples were five to seven participants smaller for analyses on Parkinson’s disease). To ensure sufficient (>10%) observations in each category, we combined scores 2 and 3 for components quality, latency and efficiency, and scores 1, 2 and 3 for components disturbances, medication and daytime dysfunction. PD = Parkinson’s disease.