Introduction

The development of electric light has revolutionized daily living with dramatically altered light conditions at night1. The primary advantage of artificial light at night (LAN) is the visibility that it offers at night, which provides convenience and safety for human activity and enhances economic growth by facilitating productivity through shift work. Due to rapid economic growth and urbanization, the coverage of outdoor LAN is also increasing. Consequently, nighttime exposure to outdoor LAN from various light sources has become a universal emerging environmental factor, impacting over 80% of the global population2,3.

Melatonin is currently regarded as a gold standard for assessing circadian rhythm, as it directly reflects the rhythm in the suprachiasmatic nucleus (SCN)4. The external exogenous of light-dark exposures modulates the central circadian clock in the SCN5. Melatonin secreted from the pineal gland is also light-sensitive. It is stimulated by darkness and suppressed during the daylight to promote sleep by influencing the circadian rhythm6. The pathological mechanisms may include circadian disruption, which leads to poor sleep and suppressed melatonin secretion7. Increasing the intensity or duration of LAN exposure can significantly affect circadian activity and metabolism. A study conducted on rodents revealed that exposure to 5 lx during nighttime resulted in obesity accompanied by elevated daytime food intake and impaired glucose tolerance8. Another study found that rats exposed to dim LAN at an intensity of ~ 2 lx altered daily patterns of locomotor activity and suppressed or shifted many rhythms in the daily patterns of glucose and lipid metabolism9. Besides light intensity, the wavelength of LAN exposure also plays a crucial role in regulating metabolism. Melatonin suppression by monochromatic light is predominantly driven by melanopsin10. Melanopsin, which is expressed in distinct types of intrinsically photosensitive retinal ganglion cells (ipRGCs), is most sensitive to short wavelengths of light at 480 nm4. Research had shown that exposure to monochromatic light at 460 nm for 2 h in the late evening significantly suppresses melatonin secretion while under the same intensity, exposure timing and duration, but at a wavelength of 550 nm such effects were not observed11. Additionally, some other light properties, including color temperature, flicker, spatial distribution, and the type of light also affected human physiology12.

The population-based study found that disruption of human sleep and circadian rhythms due to LAN exposure were linked to an increased risk of various illnesses, including metabolic disorders, mental disorders, cancer, and cardiovascular disease13,14,15. Zheng et al. used data from a national survey to demonstrate the positive associations between outdoor LAN exposure level and blood glucose, insulin resistance, and diabetes prevalence16. Another study observed a monotonically increasing risk of obesity induced by outdoor LAN exposure17. Additionally, LAN exposure had been found to be significantly associated with elevated nighttime blood pressure in elderly individuals18.

Currently, no population-based study has ever assessed the effects of outdoor LAN exposure on blood lipids in humans. It is essential to explore (1) which types of blood lipids may be influenced by outdoor artificial LAN exposure and (2) whether there is an association between outdoor LAN exposure and dyslipidemia. We aimed to investigate the associations between chronic exposure to outdoor artificial LAN and the prevalence of abnormal lipid profiles and dyslipidemia, utilizing data from a national survey of the general population in China.

Materials and methods

Study population and design

The China Health and Retirement Longitudinal Study (CHARLS) is a prospective and nationally representative population-based cohort study. The baseline survey was conducted between June 2011 and March 2012. A 4-stage, stratified, cluster sampling method was used in this survey to recruit a nationally representative sample from 150 counties/districts of 28 provinces across China. Details on this project were presented in a previous study19. All the participants signed informed consent voluntarily. All procedures in this survey were granted by the Institutional Review Board at Peking University (IRB00001052-11015).

Our study is a cross-sectional study based on 17,708 middle-aged and older adults at the baseline of the CHARLS survey between 2011 and 2012. A total of 10,894 participants were selected in this cohort study after excluding those with missing data on lipid detection (n = 6204) and those taking lipid-lowering medicine (n = 610). Individual information on age, sex, residence in urban or rural areas, education level, medical history and medication use, smoking status, drinking status, and household income was obtained from the standardized questionnaire. Height and body weight were measured by trained investigators based on a standardized protocol. Blood samples were collected and measured for blood glucose, low-density lipoprotein-cholesterol (LDL-cholesterol), high-density lipoprotein-cholesterol levels (HDL-cholesterol), triglyceride, and total cholesterol. High LDL-cholesterol was defined as LDL-cholesterol ≥ 160 mg/dl; high triglyceride was defined as triglyceride ≥ 200 mg/dl; low HDL-cholesterol was defined as HDL-cholesterol < 40 mg/dl; and high total cholesterol was defined as total cholesterol ≥ 240 mg/dl. Dyslipidemia was defined as high LDL-cholesterol, high triglyceride, low HDL-cholesterol, or high total cholesterol20. Data on the annual average fine particulate matter (PM2.5) exposure in 2011 were obtained from the Global Annual PM2.5 Grids, which combines aerosol optical depth retrievals from multiple satellite algorithms, including the NASA MODerate resolution Imaging Spectroradiometer Collection 6.1 (MODIS C6.1), Multi-angle Imaging Spectro Radiometer Version 23 (MISRv23), MODIS Multi-Angle Implementation of Atmospheric Correction Collection 6 (MAIAC C6), and the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) Deep Blue Version 421. Participants residing at each study site were assigned the same mean PM2.5 value at that site.

Exposure assessment

Outdoor LAN was defined as the excessive and inappropriate utilization of artificial outdoor lighting emitted by various sources, including street lamps, electronic billboards, residential, and commercial buildings22. The intensity of outdoor LAN exposure was obtained from the US Defense Meteorological Satellite Program (DMSP) (https://ngdc.noaa.gov/eog/download.html). The Operational Linescan System (OLS) on the satellites captures low-light imaging data at night worldwide, which is stored at the US National Geophysical Data Center (NGDC). The DMSP-OLS imaging data is transferred to the US National Oceanic and Atmospheric Administration (NOAA) National Geophysical Data Center (NGDC) database. The Global Radiance Calibrated Nighttime Lights, which excludes the sun and moon luminance, clouds, atmospheric lighting, and ephemeral events such as fires, can be used to theoretically indicate the relative intensity of LAN at ground level23. The NGDC provides eight high dynamic range data versions of LAN collected at different times. In the present study, the LAN exposure data collected in 2010 was utilized as it covers an entire year and was the closest to the baseline survey of CHARLS (https://ngdc.noaa.gov/eog/dmsp/download_radcal.html, data version ‘F16_20100111–20101209_rad_v4’)24.

The downloaded LAN data was a type of raster file composed of pixels with the size of the pixel reflecting the resolution of the raster. We processed the LAN data by using ArcGIS (ESRI, Redlands, California). The raw LAN values were converted into units of radiance (nW/cm2/sr). The study sites included urban districts and rural counties. Address geocodes of the study sites were linked to the LAN data via ArcGIS to establish baseline exposure to outdoor LAN. The mean nighttime radiance of outdoor LAN for each study site was calculated. Participants residing at a given site were assigned the same mean radiance of outdoor LAN at that site.

Statistical analysis

Demographic and lipid markers of the study participants are presented as means ± standard deviation or median (quartile) for continuous variables, and number (percentages) for categorical variables. The mean values and percentages across different groups were compared using ANOVA and χ2 tests. To assess the association between the level of outdoor LAN exposure and dyslipidemia prevalence, we did not use odds ratios, as they could strongly overestimate the prevalence ratio (PR)25,26. Cox regression, with a constant time variable assigned to all individuals and robust variance estimates, was employed to calculate the PR per quintile of LAN exposure adjusting for covariates. The covariates included age, sex, education, smoking status, alcohol consumption status, total household income, fine particulate matter, body mass index, fasting glucose, and other lipid markers. To explore potential nonlinear relationships between LAN exposure and lipid markers, as well as prevalent dyslipidemia, linear regression and Cox regression with equal times assigned to all individuals were performed by incorporating a restricted cubic spline function with 3 to 7 knots for LAN exposure levels. The Akaike Information Criteria (AIC) was compared among these models. The model with three knots set at the 5th, 50th, and 95th percentiles, which yielded the lowest AIC, was selected as the final model. Due to the limited missing data for the covariates (< 2%) in the present study, they were excluded from the analysis, and no imputation was performed.

All analyses were performed using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). A 2-sided P < 0.05 was considered statistically significant.

Results

Study population characteristics

There were significant regional variations in the LAN exposure radiance in China (Fig. 1A). Most regions exposed to low-intensity outdoor LAN, while the eastern coastal cities exhibited higher-intensity outdoor LAN. The distribution of outdoor LAN was right-skewed, with a median and interquartile range (IQR) of 6.4 (2.7–13.5) nW/cm2/sr across the board (Fig. 1B). The participants were categorized by the quintiles of outdoor LAN exposure. The medians and IQRs of outdoor LAN for quintiles 1 through 5 were 1.7 (1.0–1.9), 3.3 (2.7–4.0), 6.4 (4.9–7.3), 11.8 (9.8–13.5) and 42.3 (26.4–70.5) nW/cm2/sr, respectively (Fig. 1C). Among all participants, the mean age was 59.1 ± 9.9 years, and 53.4% of participants were women. The characteristics of the study participants by outdoor LAN groups are presented in Table 1. There were no significant differences in age and gender distribution across outdoor LAN categories (P = 0.145 and P = 0.431, respectively). Participants in the higher quintiles of outdoor LAN were more likely to have higher education levels, lower rates of current smoking, and drinking behavior, higher total household income, and higher level of PM 2.5 exposure (all P < 0.01). The levels of BMI, LDL-cholesterol, triglyceride, HDL-cholesterol, and total cholesterol varied significantly across outdoor LAN categories (all P < 0.001, except for total cholesterol which had P = 0.037). The proportions of participants with high LDL-cholesterol, low HDL-cholesterol, and dyslipidemia also differed significantly with elevated LAN levels (all P < 0.01).

Fig. 1
figure 1

Distribution of the participants exposed to outdoor light at night. (A) Sample distribution with colored circles indicates the study sites and LAN exposure intensity. (B) The number of participants exposed to each radiance level of outdoor LAN. (C) The distribution of the radiance categorized by quintiles of outdoor LAN.

Full size image
Table 1 Baseline characteristics by outdoor LAN exposure levels.
Full size table

Outdoor LAN exposure and blood lipids

The exposure-response curve illustrates the associations between outdoor LAN exposure and lipid markers, as well as dyslipidemia (Fig. 2). There were evidently monotonically increasing relationships between outdoor LAN exposure and both LDL-cholesterol and triglyceride (Fig. 2A,B). The association between outdoor LAN exposure and HDL-cholesterol also monotonically decreasing, exhibiting a steeper slope at radiance levels below 25 nW/cm2/sr (Fig. 2C). We did not observe a significant linear or nonlinear association between outdoor LAN exposure and total cholesterol (Fig. 2D). The PRs for the associations between outdoor LAN and dyslipidemia, using different definitions, displayed similar monotonically increasing curves (Fig. 2E,F).

Fig. 2
figure 2

Spline curves for outdoor light at night and lipid markers and dyslipidemia. (A) LDL-cholesterol; (B) triglyceride; (C) HDL-cholesterol; (D) total cholesterol; (E) high LDL-cholesterol, high triglyceride, or low HDL-cholesterol; (F) dyslipidemia. In all analyses, participants experiencing the top 1% of LAN were trimmed for the spline model. LAN was fitted as a smooth term using a restricted cubic spline with three knots. Shading indicates 95% CIs. The reference artificial LAN level was 0.2 nW/cm2/sr. Model was adjusted for age, sex, education, smoking status, drinking status, household total income, fine particulate matter, body mass index, fasting glucose, and mutually other lipid markers. LDL-cholesterol, low-density lipoprotein-cholesterol; HDL-cholesterol, high-density lipoprotein-cholesterol.

Full size image

Table 2 presents the PRs and 95% CIs for the associations between each quintile of outdoor LAN exposure and various adverse lipid profiles. In model 3, we observed significant increased PRs for the associations of outdoor LAN with high LDL-cholesterol (PR = 1.10, 95% CI = 1.04–1.16), high triglyceride (PR = 1.07, 95% CI = 1.03–1.12), and low HDL-cholesterol (PR = 1.11, 95% CI = 1.08–1.16). The increase LAN exposure per quintile was not significantly associated with high total cholesterol (PR = 1.02, 95% CI = 0.97–1.08). Additionally, a per-quintile LAN exposure was positively associated with the prevalence of dyslipidemia in model 3 (PR = 1.06, 95% CI = 1.03–1.09).

Table 2 PRs and 95% CIs for the prevalence of adverse lipid profile associated with per-quintile outdoor LAN exposure.
Full size table

High intensity of outdoor LAN exposure was associated with an increased risk of prevalent dyslipidemia (Fig. 3). In fully adjusted models, the PRs (95% CIs) in quintile 5 of LAN exposure were 1.49 (1.19–1.88) for high LDL-cholesterol and 1.32 (1.09–1.59) for high triglyceride, compared with the lowest quintile (Fig. 3A,B). The PRs (95% CIs) in quintile 3–5 of LAN exposure were 1.36 (1.18–1.57), 1.42 (1.21–1.66), and 1.40 (1.21–1.63) for low HDL-cholesterol compared with the lowest quintile (Fig. 3C). Significant PRs for prevalent dyslipidemia with different definitions were also observed in the highest quintile of LAN exposure compared with the lowest (Fig. 3E,F).

Fig. 3
figure 3

Associations between outdoor light at night and lipid profile with different definitions. (A) high LDL-cholesterol: LDL-cholesterol ≥ 160 mg/dl; (B) high triglyceride: triglyceride ≥ 200 mg/dl; (C) low HDL-cholesterol: HDL-cholesterol < 40 mg/dl; (D) high total cholesterol: total cholesterol ≥ 240 mg/dl; (E) LDL-cholesterol ≥ 160 mg/dl, triglyceride ≥ 200 mg/dl, or HDL-cholesterol < 40 mg/dl; (F) dyslipidemia: high LDL-cholesterol, high triglyceride, low HDL-cholesterol, or high total cholesterol. Model was adjusted for age, sex, education, smoking status, drinking status, household total income, fine particulate matter, body mass index, fasting glucose, and mutually other lipid markers.

Full size image

Discussion

By utilizing a nationwide investigation of the Chinese adult population conducted across 150 study sites covering 31 provinces across China, the present study found that long-term exposure to outdoor LAN was positively associated with levels of LDL-cholesterol, triglyceride, and the prevalence of dyslipidemia, and inversely associated with HDL-cholesterol levels. These associations persisted even after adjusting for several important risk factors. To our knowledge, this study is the first study to examine the relationship between outdoor LAN exposure and the risk of dyslipidemia, as well as lipid markers levels in humans. Our findings provide epidemiological evidence supporting the detrimental effects of outdoor LAN exposure on lipid metabolism.

The world atlas of outdoor LAN using satellite data from the DMSP reveals that extensive areas of the Earth`s surface have lost their natural dark skies. Light pollution is not uniformly distributed; rather, it is concentrated in densely populated and economically active regions27. Many studies used the satellite images of outdoor light as a proxy measure for the actual LAN exposure to assess the intensity of outdoor LAN exposure28. By analyzing nighttime satellite images, several lines of evidence also support the adverse effects of outdoor LAN exposure on metabolism and metabolic-related diseases29,30. For example, an analysis based on a 10-year follow-up study of older adults found that a 60 nW/cm2/sr increase in outdoor LAN exposure was associated with a 10% increased risk of coronary heart disease30. Another study also found that increasing nighttime light intensity was positively associated with cardiovascular risk factors, including BMI, systolic blood pressure, and LDL-cholesterol31. Previous studies have proved that exposure to LAN can increase disease risk by disturbing circadian rhythms, resulting in various alterations in physiological functions and metabolism, such as body temperature regulation, dietary intake behavior, insulin sensitivity, glucose metabolism, and plasma melatonin concentrations32.

For environmental pollution, threshold values are typically established to safeguard public health by maintaining pollutant levels below a specified limit. Our study suggests that the relationship between outdoor LAN exposure and dyslipidemia may follow a linear pattern. This implies that the effect of outdoor LAN on dyslipidemia may not depend on reaching a specific level of light exposure; instead, it may have a continuous impact across different levels of exposure. However, to better understand these associations and to inform public health policies, further research is necessary.

The present study has a significant strength in that the participants are recruited across mainland China with a broad coverage of geographic areas. This broad coverage ensures that the participants are exposed to different intensities of outdoor LAN exposure, providing a representative sample of the population. However, several limitations need to be identified in this study. First, due to its cross-sectional design, our study cannot examine the causal relationship between LAN exposure and the incidence of dyslipidemia. Second, the satellite-derived data serves as a crude representation of LAN exposure for participants residing at the same study site, as they were assigned the same level of LAN exposure. Actual LAN exposure can be influenced by many factors at an individual level, such as indoor light exposure, nighttime activities, and shift work. Third, the presence of a substantial number of participants with missing data on lipid measurements may introduce selection bias. Finally, unmeasured confounders (e.g., lifestyle habits, access to healthcare, shift work, sleep quality, and traffic noise) may also affect on the association between LAN exposure and dyslipidemia. Due to a lack of relevant information, we were unable to control these factors in our analysis.

Conclusion

The current study demonstrates a significant association between higher intensity of outdoor LAN exposure and an abnormal lipid profile, as well as the presence of dyslipidemia. Further prospective research is needed to validate this association, and intervention studies are required to determine whether reducing light pollution could serve as an effective public health strategy for the prevention of dyslipidemia.