06/03/2026
Most people think their bedroom is dark enough. The metabolic data suggests otherwise, and the threshold where measurable effects show up is lower than almost any standard sleep environment.
Mason and colleagues (2022, Proceedings of the National Academy of Sciences) ran a controlled laboratory study at Northwestern University. Twenty healthy adults slept for one night under one of two conditions: a dimly lit room (less than 3 lux, effectively dark) or a moderately lit room (100 lux, roughly the brightness of a hallway nightlight or a bedside lamp left on). The 100 lux night produced measurable changes the next morning. Insulin resistance was higher. Nighttime heart rate was elevated. Heart rate variability was reduced, indicating sympathetic nervous system activation. One night of moderate ambient light during sleep was enough to push cardiometabolic markers in an unfavorable direction.
The follow-up question is whether that single-night signal translates into long-term disease risk. Obayashi and colleagues (2020, Sleep Medicine) had already provided population-level data suggesting it does. Their HEIJO-KYO cohort of 678 elderly Japanese adults without diabetes at baseline had bedroom light intensity measured objectively over consecutive nights. After a median 42 months of follow-up, 19 participants developed diabetes. The 128 participants whose bedrooms averaged 5 lux or more had an incidence rate ratio of 3.74 (95% CI 1.55 to 9.05) compared with the 550 participants whose bedrooms averaged below 5 lux. When the cutoff was lowered to 3 lux, the relationship remained significant at 2.74x.
To put 5 lux in context. Direct sunlight measures around 50,000 lux. A bright office is around 500 lux. A living room in the evening is around 50 lux. A hallway nightlight is around 10 lux. Five lux is roughly what reaches your bedroom from a streetlight through closed curtains, or from an LED display across the room, or from a phone face-up on the nightstand. It is well below the threshold of "feels dark" to most people. The Obayashi cohort had 19 percent of participants exceeding it.
The mechanism that Mason's data points to is sympathetic activation. Even modest ambient light during sleep keeps the sympathetic nervous system more engaged, which manifests as higher heart rate, lower heart rate variability, and impaired insulin sensitivity the next morning. Repeated chronically across years, this is a plausible pathway to the diabetes signal Obayashi observed at the population level.
Three caveats are worth being explicit about. First, Mason 2022 is a small acute study (n=20, one night). The within-subject signal is robust, but extrapolating to chronic effects requires interpretation. Second, Obayashi 2020 is observational cohort data. The 3.74x incidence rate ratio is an association, not proof of causation, and despite adjustment for known confounders, residual confounding is always possible in observational designs. Third, the Obayashi cohort was specifically elderly Japanese adults (mean age 70.6). Generalization to younger populations and other ethnicities is uncertain, though the underlying circadian and sympathetic biology is conserved across humans.
What this means in practice. The lever is mechanical, not pharmacological. Pull the curtain. Cover the LED clock. Move the phone out of the room or face down. Tape over standby lights on electronics. The cost is minutes of effort and no money. The Mason RCT shows the next-morning signal from a single moderately lit night. The Obayashi cohort shows the population-level diabetes association. Together they are not proof, but they are aligned, and the practical lever is cheap.
Most "dark" bedrooms aren't actually dark. Five lux is the threshold below which a single epidemiological signal disappears, and most bedrooms with any electronic devices, any uncovered windows, or any hallway light bleed are well above it. If you wake up and can see across the room before turning anything on, you are probably above the threshold.
Mason et al., Proceedings of the National Academy of Sciences, 2022
Obayashi et al., Sleep Medicine, 2020