Table 1.
Studies on the relationship between artificial light at night and diabetes risk
| Study | Setting | Population demographics | Type of ALAN exposure | Assessment of diabetes outcomes | Main results and conclusions | Confounding factors |
|---|---|---|---|---|---|---|
| Outdoor ALAN | ||||||
| Zheng et al. (2023) [29] | Cross-sectional study | Adult (aged ≥18; mean age, 42.7) | Method: DMSP | FPG ≥126 mg/dL or 2hPG ≥200 mg/dL or HbA1c ≥6.5% or previously diagnosed diabetes | The highest quintile of LAN was associated with increased prevalence of diabetes compared to the lowest quintile (PR, 1.28; 95% CI, 1.03–1.60) | Age, sex, education, smoking, drinking, physical activity, family history of diabetes, household income, residential area, taking antihypertensive or lipid-lowering medications, and BMI |
| n=98,658 (men: 50.9%) | Intensity: quintiles (Q1–5) and continuous variable | |||||
| Exposure: year 2010 | ||||||
| Time: NA | ||||||
| Wavelength: NA | ||||||
| Xu et al. (2023) [30] | Prospective cohort study (UK Biobank) | Adults (aged 37–73; mean age, 55.9) | Method: DMSP | ICD-10 codes (E11) by hospital inpatients records | The highest quantile was related to increased diabetes risk compared to the lowest quantile (HR, 1.14; 95% CI, 1.02–1.27) | Age, sex, ethnicity, education level, economic activity, household income, smoking, alcohol frequency, physical activity level, sedentary time, shift work, health diet score, population density, housing score, income score, air pollution level, noise at night, and PRS score |
| Intensity: quartiles (Q1–4) | ||||||
| n=283,374 (men: 48.2%) | Exposure: average value of annual outdoor LAN during follow-up | The exposure-response curve of incident T2DM increases monotonically, with a plateau at the highest level (P for non-linear=0.014) | ||||
| Time: NA | ||||||
| Wavelength: NA | ||||||
| Sorensen et al. (2020) [31] | Cross-sectional study | Adults (aged >18; median age, 40 for women and 29 for men) | Method: NTLI by satellite data from DMSP | FPG (continuous) | There was no association between NTLI and FPG (β=–0.002; 95% CI, –0.1 to 0.1) | Age, sex, caste (India), religion, marital status, and survey season |
| n=5,328 (men: 54%) | Intensity: log-scaled, continuous variable | |||||
| Exposure: NA | ||||||
| Time: NA | ||||||
| Wavelength: NA | ||||||
| Indoor ALAN | ||||||
| Obayashi et al. (2014) [36] | Cross-sectional study | Elderly individuals (aged ≥60; mean age, 72.7) | Method: ambulatory light meter | Diabetes (based on medical history, current diabetes treatment, or FPG ≥126 mg/dL, or HbA1c ≥6.5%) | Prevalent diabetes (OR, 1.72; 95% CI, 1.12–2.64) | Sex, BMI, sleep duration |
| n=513 (men: 46.4%) | Intensity: average light intensity (continuous variable) | The evening light exposure increased from 17.5 (25%) to 37.6 lux (75%), associated with a 51.2% increase in prevalent diabetes. | ||||
| Exposure: 2 consecutive days | ||||||
| Time: evening (4-hr before bedtime) | ||||||
| Wavelength: NA | ||||||
| Obayashi et al. (2020) [37] | Prospective cohort study (median, 42 months FU) | Elderly individuals (≥60 years old) | Method: portable light meter | Diabetes (based on medical history, current diabetes medications, or HbA1c ≥6.5%) | The incident rate for diabetes was higher in the LAN group than in the dark group (IRR, 3.19; 95% CI, 1.38–7.47; P=0.007) | Age, sex, smoking, drinking, education, household income, BMI, hypertension, caloric intake, bedtime and rise time, daytime physical activity, and daytime light exposure |
| n=678 | Intensity: LAN (average ≥5 lux) vs. dark (average <5 lux; reference) | |||||
| Exposure: 2 consecutive bedtime nights | ||||||
| Time: from bedtime and rise time | ||||||
| Wavelength: NA | ||||||
| Xu et al. (2022) [38] | Cohort study (1 year FU) | Young adults (aged 16–22; mean age, 18.8) | Method: portable illuminance meter | HOMA-IR, CM risk score (continuous var.; calculated from the score sum of waist circumference, blood pressure, lipid profile, HOMA-IR) | Higher average ALAN exposure (all night) was associated with an increase in CM risk score (β=1.47; 95% CI, 0.69–2.25) | Age, sex, BMI, smoking, drinking status, socioeconomic status, physical activity, food addiction, sleep duration, sugar-sweetened beverages consumption, screen time |
| n=484 (men: 38.7%) | Intensity: (1) 1-hr and 4-hr average intensity of post-bedtime light; (2) average light intensity from bedtime to rising time; (3) 1-hr and 2-hr average intensity of pre-awake light (continuous variable) | Exposure time: post-bedtime light exposure (within 1 to 4 hr) was associated with increased HOMA-IR | ||||
| Exposure: 7 consecutive days | ||||||
| Time: night (between bedtime and awake) | ||||||
| Wavelength: NA | ||||||
| Windred et al. (2024) [39] | Prospective cohort study (2006–2010) | Mild aged individuals (aged 40–69) | Method: light sensor | Diabetes (based on hospital admission records, death register) | Dose-dependent relationship between brighter LAN and higher risk of T2DM (90–100th percentiles, HR, 1.53; 95% CI, 1.32–1.77) | Age, sex, ethnicity, income, material deprivation, education, employment status, smoking/alcohol, healthy diet, physical activity, and urbanicity |
| Intensity: 4 percentile ranges | ||||||
| n=84,790 (men: 42%) | Exposure: 7 consecutive days | |||||
| Time: day and night (0:30 AM–6:00 AM) | ||||||
| Wavelength: NA | ||||||
| Kim et al. (2023) [40] | Cross-sectional study (2007–2010) | Elderly individuals (aged 65–84) | Method: actigraphy | Diabetes (based on FPG >126 mg/dL or taking medications) | Any exposure to light during the night-time was associated with diabetes risk (OR, 2.0; 95% CI, 1.19–3.43) | Age, sex, race, season |
| n=552 | Intensity: average 0 > light value during 5-hr nadir (L5, log-scaled) light value during the 5-hr nadir (LAN) vs. 5-hr period of complete darkness per 24-hr (reference; no-LAN) | |||||
| Exposure: 7 consecutive days | Non- (reference), low- (OR, 1.87; 95% CI, 1.01–3.45), high-LAN (OR, 2.15; 95% CI, 1.17–3.97; P=0.014) | |||||
| Time: during sleep | ||||||
| Wavelength: NA | ||||||
| Mason et al. (2022) [41] | Parallel-group study | Healthy young adults (mean age, 26) | Method: exposure to ALAN (overhead ceiling light bulb) | HOMA-IR | Higher HOMA-IR, poorer sleep quality, higher HR with lower HR variability, higher sympathovagal balance associated with higher 30-min insulin AUC (increased insulin resistance the following morning) under the ALAN condition: mainly by increased SNS activation | NA |
| n=20 (men: 30%) | Intensity: BL (100 lux) vs. DL (<3 lux) | Insulin AUC | ||||
| Exposure: 1-day | ||||||
| Time: during sleep (8-hr) | ||||||
| Wavelength: NA | ||||||
| Grimaldi et al. (2021) [42] -abstract- | Randomized parallel cross-over study | Healthy adults (18–40 years old) | Method: light exposure | Postprandial insulin AUC (OGTT) | Higher insulin AUC under DL than BL condition (P=0.029) | NA |
| n=20 | Intensity: BL (100 lux) during sleep vs. dark (<3 lux) | Higher HR in DL condition was positively correlated with insulin AUC | ||||
| Exposure: 1-day | ||||||
| Time: overnight | ||||||
| Wavelength: NA | ||||||
| Albreiki et al. (2017) [43] | Two-way cross-over design study | Healthy adults (mean age, 22) | Method: exposure to ALAN | Plasma glucose, insulin, HOMA-IR | The total AUC of glucose and insulin levels during the study period was higher in BL than in the DL condition. | Age, BMI |
| n=17 (men: 52.9%) | Intensity: BL (>500 lux) vs. DL (<5 lux, reference) | |||||
| Exposure: 1-day | HOMA-IR levels before and after dinner were comparable between BL and DL condition | |||||
| Time: between evening and night (6:00 PM–6:00 AM, 12-hr) | ||||||
| Wavelength: fluorescent light | ||||||
| Cheung et al. (2016) [44] | Randomized parallel-group study | healthy adults (20–39 years old; median age, 28) | Method: exposure to ALAN in the morning or evening | Plasma glucose, insulin, HOMA-IR | BL exposure in the evening: higher peak glucose and HOMA-IR AUC compared to DL | NA |
| n=19 (men: 42.1%) | Intensity: morning or evening BL (260 lux) vs. DL (<5 lux, reference) | BL exposure in the morning: higher HOMA-IR but no differences in glucose level compared to DL | ||||
| Exposure: 1-day | ||||||
| Time: morning (0.5-hr after awake) or evening (10.5-hr after awake) for 3-hr | The evening exposure group had higher peak glucose than the morning exposure group | |||||
| Wavelength: blue-enriched light (LED) | ||||||
| Chamorro et al. (2021) [45] | Randomize, controlled, cross-over experimental study | Normal-weight adults (mean age, 23.4) | Method: portable LED lamp exposure | Glucose, insulin, C-peptide | No significant differences in morning levels of glucose, insulin, C-peptide, and HOMA-IR between the two groups | NA |
| n=20 | Intensity: DL (<5 lux) vs. total darkness (reference) | |||||
| Exposure: 2 consecutive nights | ||||||
| Time: night (during 8-hr of sleep) | ||||||
| Wavelength: LED |
ALAN, artificial light at night; DMSP, Department of Defense’s Defense Meteorological Satellite Program; NA, not applicable; FPG, fasting plasma glucose; 2hPG, 2-hour prandial glucose; HbA1c, glycosylated hemoglobin; LAN, light at night; PR, prevalence ratio; CI, confidence interval; BMI, body mass index; ICD-10, International Classification of Diseases 10th edition; HR, hazard ratio; T2DM, type 2 diabetes mellitus; PRS, polygenic risk score; NTLI, night-time light intensity; OR, odds ratio; FU, follow-up; IRR, incident rate ratio; HOMAIR, homeostatic model assessment of insulin resistance; CM, cardiometabolic; HR, heart rate; BL, blight light; DL, dim light; AUC, area under the curve; SNS, sympathetic nervous system; OGTT, oral glucose tolerance test; LED, light-emitting diode.