Sleep and Biological Aging

Sleep is often presented as a pillar of recovery, but its role extends far beyond the feeling of rest or the next day’s performance. In adults, it plays a role in immune, metabolic, neuroendocrine, and cardiovascular regulation. From a long-term health perspective, the question is therefore not merely whether sleep improves alertness or reduces fatigue, but whether it is part of broader biological trajectories capable of influencing how our organs age.

With the rapid development of biological clock models, this idea has become easier to test. These models estimate biological age based on data from magnetic resonance imaging, plasma proteomics, or metabolomics. When an organ or system exhibits a biological age higher than what chronological age suggests, this is referred to as a biological age discrepancy. Several studies have already demonstrated a U-shaped relationship between sleep duration and various phenotypes based on these biological clocks. But do these relationships pertain only to the brain, or do they reflect a more systemic signature, visible across multiple organs and at various biological levels? And are they similar in women and men? Thus, in middle-aged and older adults, is self-reported sleep duration non-linearly associated with the biological aging clocks of multiple organs, and do these associations translate into risks of systemic diseases, mortality, and late-life depression?

The study conducted

To answer these questions, an international team of researchers developed an approach called Sleep Chart using large population cohorts, primarily the UK Biobank (adults aged 37 to 84). Sleep duration was measured via questionnaire and based on a simple question asking how many hours of sleep participants averaged over 24 hours, including naps (the study did not measure sleep via polysomnography or actigraphy in the main analysis).

The main analysis focused on 23 biological aging clocks derived from three data sets. Some clocks were based on MRI, with indicators for the brain, heart, liver, pancreas, spleen, adipose tissue, or kidneys. Others were derived from plasma proteomics, i.e., circulating proteins associated with different organs. Finally, a third set used plasma metabolomics, specifically to capture digestive, hepatic, immune, endocrine, or metabolic dimensions. The goal was not to test a single organ, but to map the relationship between sleep and biological age across the entire body.

To model these relationships, the authors used generalized additive models. The advantage of this approach was that it did not impose a linear relationship between sleep duration and biological aging in advance. The model could therefore detect a flat, increasing, decreasing, or U-shaped relationship. The analyses were adjusted for numerous factors, including age, sex, weight, height, waist circumference, body mass index, blood pressure, assessment center, and disease status. Extreme values were limited by focusing the analysis on sleep durations between 4 and 10 hours.

The authors then classified sleep into three categories: short (< 6 hours), normal (6–8 hours), and long (> 8 hours). This classification was used to analyze genetic correlations with diseases, the risk of developing clinical diagnoses, all-cause mortality, and possible mediating pathways to two subtypes of late-onset depression.

Results & Analyses

The main result shows that the relationship between sleep duration and biological aging is not simply linear, but follows a U-shaped curve. Of the 23 circadian clocks studied, 9 showed a statistically significant non-linear association with sleep duration. In these clocks, the shortest and longest durations were associated with a higher biological age gap, while the lowest values fell within an intermediate range.

This intermediate range was not a single figure. The observed minimums varied by organ, technology used, and sex. Overall, the smallest biological age gaps fell between 6.4 and 7.8 hours of sleep. For the brain proteomic clock, which showed the strongest U-shaped association, the minimum was estimated at around 7.8 hours for women and 7.7 hours for men. For the endocrine metabolomic clock, the minimums were shorter, around 6.7 hours for women and 6.1 hours for men. For the brain clock derived from MRI, they were close to 6.5 hours for both sexes. This variation suggests that there is no single universal optimal duration, but that extremes in duration are associated with a higher biological burden across multiple systems.

The most pronounced signals involved several biological levels. On the proteomic side, U-shaped associations emerged for the brain, lung, liver, immune system, and skin clocks. On the metabolomic side, the main signal concerned the endocrine clock. In terms of imaging, the clocks of the brain, adipose tissue, and pancreas also followed a U-shaped relationship. Additional analyses of 720 imaging phenotypes, 342 organ-enriched proteins, and 107 organ-associated metabolites reinforced this idea of a systemic rather than strictly cerebral phenomenon.

Genetic analyses suggest that short sleep and long sleep may not have exactly the same biological architecture. Short sleep appears to have broader associations with cardiovascular, metabolic, musculoskeletal, pulmonary, digestive, neurological, and psychiatric diseases. Meanwhile, long sleep appears to have a profile more focused on neuropsychiatric and brain-related phenotypes. Concurrently, survival analyses in the UK Biobank identified 153 significant associations between categories of abnormal sleep and incident diagnoses, with a predominance of short sleep.

Regarding all-cause mortality, compared to 6 to 8 hours of sleep, short sleep was associated with a higher risk of death, with a hazard ratio of 1.50 and a 95% confidence interval of 1.44 to 1.55. Long sleep was also associated with a higher risk, with a hazard ratio of 1.40 and a confidence interval of 1.36 to 1.44. Although these data are important, they should be interpreted as adjusted associations, not as evidence that changing sleep duration would automatically reduce risk by the same proportion.

Regarding late-onset depression, for short sleep, the models suggest a more direct relationship with subtypes of late-onset depression, with a limited mediating role for MRI biological clocks, except for certain signals involving adipose tissue in particular. For long sleep, the association appeared to be mediated more by biological age clocks, particularly those in the brain and adipose tissue. In one of the models, the brain clock as measured by MRI explained a significant proportion of the total effect. This difference is consistent with two distinct biological scenarios: short sleep could reflect or induce a more immediate physiological stress, via neuroendocrine stress, inflammation, and metabolic dysregulation; long sleep may be more of a marker of frailty, subclinical processes, or physiological compensation.

The authors remain cautious regarding causality. Mendelian randomization analyses do not strongly support the idea that diseases largely cause sleep disturbances, but they do not rule out reverse causality, especially for long sleep. Persistent long sleep duration may be a behavioral pattern, but it may also signal an underlying health condition, chronic fatigue, depression, inflammation, or an undiagnosed condition. The study provides a robust mapping of associations, not a universal prescription.

Practical Applications

The value of this study lies in its suggestion that habitual sleep duration should be considered a systemic marker, particularly among middle-aged adults and seniors. Sleep consistently lasting less than 6 hours should not be casually interpreted as a simple lack of discipline or an adaptation to one’s lifestyle. In this study, it is associated with a higher burden of biological aging across multiple systems and a broad clinical risk profile.

Long sleep warrants an even more nuanced interpretation. Sleeping more than 8 hours is not automatically problematic for every individual, and the study does not support the conclusion that one should deliberately reduce their sleep. However, a sustained increase in the need for sleep, especially when accompanied by fatigue, reduced activity, depressive symptoms, pain, weight gain, or impaired recovery, can become a sign of an underlying medical condition. For a healthcare professional, this may warrant a more in-depth discussion, or even a referral to a specialist when the clinical picture suggests it.

In practice, the most sensible approach is to monitor trends rather than aim for a perfect duration. For adults, a range of 6 to 8 hours appears to be the benchmark used by the study and corresponds to the range where biological age discrepancies were generally the lowest. But the exact range varies by organ and individual. Therefore, other factors must be taken into account: sleep quality, consistency of sleep schedules, nighttime fragmentation, daytime sleepiness, mental state, level of physical activity, light exposure, training load, pain, medications, and medical conditions, etc. Not all of these factors were measured in the study, but they strongly influence the interpretation of reported sleep duration.

In conclusion, consistently too little sleep would be a warning sign, consistently too much sleep would be a signal to be contextualized, and the intermediate range appears to be biologically more favorable at the population level. However, these data do not justify guilt, anti-aging promises, or individualized recommendations based on a single threshold. Sleep appears to be a cross-sectional indicator of biological aging, useful for thinking about health in a systemic way.

Reference

MULTI Consortium; O’Toole CK, Song Z, Anagnostakis F, Yang Z, Tian YE, Duggan MR, Zou C, Leng Y, Cai Y, Bai W, Fu CHY, Rafii MS, Aisen P, Wang G, De Jager PL, Zeng J, Oh HS, Zhou X, Walker KA, Belsky DW, Zalesky A, Simonsick EM, Resnick SM, Ferrucci L, Davatzikos C & Wen J. Sleep chart of biological aging clocks in middle and late life. Nature 2026, Ahead of print.