The study was led by Shengzhi Sun of Capital Medical University in Beijing, China, alongside a team of international colleagues. Their findings arrive at a time when sleep health has ascended to the forefront of public health priorities, particularly in East Asia. In China, the National Health Commission has increasingly identified sleep quantity and quality as a pillar of national wellness, a focus that is reflected in the most recent 2026 China Sleep Health White Paper. That report highlights a growing crisis of sleep deprivation: while the average surveyed Chinese citizen currently manages about 6.97 hours of sleep per night, nearly a quarter of the population—approximately 23 percent—survives on less than 6 hours. Furthermore, the White Paper indicates that only 26 percent of respondents maintain a consistent sleep schedule, with sleep disturbances becoming more prevalent over time. For citizens and policymakers alike, understanding the precise mechanisms by which these patterns translate into disease risk has become a matter of urgent relevance.
Historically, the scientific community has struggled to map the relationship between specific sleep stages and health outcomes in large populations. This difficulty stems largely from the limitations of data collection; most large-scale studies have relied on self-reported sleep measures, which often correlate poorly with objective physiological assessments. People frequently overestimate how long they sleep or fail to accurately recall the frequency of nighttime awakenings. To overcome these hurdles, Sun and his team turned to the UK Biobank, a massive biomedical database and research resource that follows the health and genetics of half a million participants.
The researchers analyzed data from a subset of 95,559 UK Biobank participants who participated in an objective sleep study. These individuals wore wrist-worn accelerometers—high-tech motion sensors—for seven consecutive days and nights. To translate this motion data into meaningful biological information, the team employed a sophisticated deep-learning algorithm capable of distinguishing between different sleep stages, including Rapid Eye Movement (REM) sleep, deep sleep (also known as slow-wave sleep), and light sleep. This technology also allowed them to measure total sleep duration, wakefulness after sleep onset (WASO), and the degree of night-to-night sleep irregularity. By following these participants for a median period of 8.9 years and cross-referencing their sleep data with comprehensive health records, the researchers were able to test for associations across more than 1,000 distinct disease outcomes.
The results of this "phenome-wide association analysis" were striking, identifying 156 significant associations between sleep patterns and incident diseases after applying the rigorous Bonferroni correction to ensure statistical reliability. One of the most significant revelations concerned REM sleep, the stage of sleep typically associated with vivid dreaming and memory consolidation. The study found that greater amounts of REM sleep were associated with a significantly lower risk of 83 different diseases. For every interquartile range increase in REM sleep—approximately 47.6 minutes—the risk of heart failure dropped significantly, with a hazard ratio of 0.74. Even more profound were the links to brain health: higher REM sleep duration was associated with a hazard ratio of 0.54 for dementia and a remarkable 0.20 for Parkinson’s disease, suggesting a potentially protective role for REM sleep against neurodegeneration.
