Disaggregating the effects of daytime and nighttime light exposures on obesity, overweight, prostate and breast cancer morbidity worldwide

ABSTRACT

Hormone-dependent cancers and overweight/obesity are not necessarily linked but might have similar underlying causes, such as circadian disruption, lack of physical activity, and unhealthy nutrition. Several empirical studies also attribute the rise in these types of morbidity to vitamin D deficiency, linked in turn to insufficient sunlight exposure. Other studies place an emphasis on melatonin (MLT) hormone suppression, associated with artificial light at night (ALAN) exposure. Yet no studies, carried out to date, have attempted to determine which of these environmental risk factors is associated stronger with the morbidity types in question. In this study, we aim to narrow this knowledge gap by analyzing data available for 100+ countries worldwide, while controlling ALAN and solar radiation exposure estimates by several potential confounders, such as GDP_pc, GINI inequality index and unhealthy food consumption. As the study reveals, all the morbidity types under analysis are significantly and positively associated with ALAN exposure estimates (p < 0.05), while solar radiation appears to be significantly associated with prostate cancer rates only (p < 0.05), but not with breast cancer or overweight/obesity rates (p > 0.1). To the best of our knowledge, this study is the first that separates the effects of ALAN and daylight exposures on the abovementioned types of morbidity.

KEYWORDS:

• Artificial light at night (ALAN)
• vitamin D
• solar radiation
• prostate cancer (PC)
• breast cancer (BC)
• overweight/obesity

Introduction

Rates of hormone-dependent cancers, such as breast cancer (BC) in women and prostate cancer (PC) in men, and overweight and obesity in both genders are on the rise worldwide (see Figure 1). Although these types of morbidity are not necessarily linked, they might have similar underlying causes, such as aging (Cooper et al. 2021; Sung et al. 2021; WHO 2022a), lack of physical activity (Friedenreich and Orenstein 2002; Rippe and Hess 1998), and unhealthy food consumption (Barsouk et al. 2020; Behrens et al. 2018; Mohammadbeigi et al. 2018).

Figure 1. Global changes in breast and prostate cancer incidence ASRs (a) and overweight/obesity prevalence rates, % (b).

Source: Diagrammed using data from the Global Health Data Exchange GHD (2019) and WHO (2022b) respectively.

Notes: Overweight is defined as BMI ≥ 25, while obesity is defined as BMI ≥ 30; the diagrams feature age-standardized rates (ASRs) are for adults only.

Empirical studies have also linked these types of morbidity to several environmental risk factors, such as insufficient sunlight exposure, linked, in turn, to vitamin D deficiency (Alfredsson et al. 2020; Peterlik et al. 2009), while other studies emphasize the effect of prolonged exposure to artificial light at night (ALAN) (Haim and Portnov 2013; Park et al. 2019).

Vitamin D is a seco-steroidal prohormone, produced endogenously when sunlight strikes the skin and triggers vitamin D synthesis (Jeon and Shin 2018). The synthesis of this hormone plays an important role in the body’s functioning, by affecting calcium and phosphor homeostasis (Bellavia et al. 2016), reducing hypertension (Pilz et al. 2009), improving immune function (Ponsonby et al. 2005) and lowering the risk of inflammation (Agrawal and Yin 2014).

In an attempt to explain the observed rise in the above types of morbidity, the authors of several studies place an emphasis on the impact of ALAN, associated with melatonin (MLT) suppression and circadian disruption (Reiter et al. 2007; Smith et al. 2004; Stevens et al. 2014). The main mechanisms behind this association are disruption of the biological clock function (Haim and Portnov 2013; Stevens and Rea 2001), and inhibition of the MLT hormone secretion due to ALAN exposure (Blask et al. 2005; Haim and Portnov 2013).

According to Reiter et al. (2007), the MLT hormone has antioxidative properties, and as a result, suppression of the nocturnal MLT production changes the estrogen receptor affinity and increases susceptibility to different diseases, including hormone-dependent cancers, such as BC and PC (Blask et al. 2005; Figueiro 2017; Haim and Portnov 2013; Momenimovahed and Salehiniya 2019). Recent empirical studies lend support to such a possibility, albeit most of the evidence comes from ecological and semi-ecological research (Rybnikova et al. 2015, 2017).

It should be noted, however, that sunlight and ALAN are interlinked, as the lack of sunlight is often compensated by ALAN, especially in northern latitudes (Haim and Portnov 2013). Yet, which of the abovementioned environmental risk factors, i.e., excessive ALAN exposure, potentially resulting in circadian disruption and MLT suppression, or insufficient exposure to sunlight, resulting in vitamin D-deficiency, is associated stronger with human morbidity remains largely unclear, as, to the best of our knowledge, no study, carried out to date, has attempted to separate the effect of these factors on different types of human morbidity. The present study aims to narrow this knowledge gap, by contrasting sunlight and ALAN exposure data with BC and PC morbidity and overweight/obesity data available for 100+ countries worldwide, while controlling for potential confounders, such as per capita GDP, GINI inequality index, and unhealthy food intake.

The paper starts with a summary of previous studies linking vitamin D deficiency and excessive exposure to ALAN with different types of morbidity. A description of the study method and data sources follows this introduction. Next, the results of the analysis are introduced and the required public policy responses are discussed.

Summary of background studies

Vitamin D deficiency and its potential health effects

Several empirical studies investigated the association between vitamin D deficiency and specific health effects. Thus, Yao et al. (2015) compared vitamin D levels in 3867 obese participants and 9342 healthy individuals, showing a significant difference in the prevalence of vitamin D deficiency among obesity and control groups, with the odds ratio (OR) being equal to 3.43 (95% CI = 2.33–5.06). In another study, Syed et al. (2019) enrolled 117 diabetes type 2 patients, whose 25-OH-Vit D levels were calculated. Of this number, 83 patients (76.14%) were found to be either vitamin D-deficient (60.6%) or vitamin D-insufficient (15.6%), while only 26 patients (23.9%) were vitamin D sufficient. The study also revealed a statistically significant difference in the vitamin D levels between diabetics and healthy subjects (p = 0.015), while a statistically significant association between vitamin D levels and BMI was also found (p = 0.018).

According to another study by Der et al. (2014), who examined the medical records of army veterans diagnosed with PC in the southeastern part of the US between October 1, 1999, and April 30, 2012, patients with vitamin D deficiency showed significantly smaller ORs of survival. In a separate study, Park et al. (2015) conducted a large case–control study in Korean women, comparing serum-25 hydroxyvitamin D 25(OH)D levels in 3634 BC patients and 17 133 controls. As the study revealed, women, who had vitamin D levels of less than 20 ng/mL, exhibited a 27% increase in the risk of developing BC, as compared to women with vitamin D levels of 20 ng/mL and above.

ALAN exposure and its known health effects

In two separate studies, Rybnikova et al. (2015, 2017) examined the association between BC and PC incidence and ALAN exposure estimated using data from 180 countries worldwide. The researchers compared age-standardized rates (ASRs) of these cancers with multi-annual ALAN measurements performed by nighttime satellites and considered potentially different latency periods. After the whole set of countries was disaggregated into geographic clusters of countries with similar development levels, the results showed a positive association between ASRs of both types of cancer and ALAN exposure estimates, lagged by a 10-year period.

The association between ALAN and obesity forms a separate strand of research that reveals the association between ALAN exposure and excessive body mass in humans. The biological mechanisms behind such an association include the inhibition of MLT production due to ALAN exposure (Garaulet et al. 2010) and suppression of MLT rhythmicity, which disrupts metabolic function and may lead to obesity (Kennaway et al. 2002). Another mechanism that links ALAN exposure to obesity is sleep disorders that influence circadian rhythm through changes in carbohydrate metabolism and endocrine function (Laposky et al. 2008). ALAN exposure may also lead to changes in eating behavior, leading to obesity, due to shifting the time of food intake from daytime to the nighttime phase (Arble et al. 2009; Fonken et al. 2010; Rybnikova et al. 2016).

Thus, in a recent study, Park et al. (2019) examined the association between exposure to ALAN while sleeping with the risk of obesity in women aged 35 to 74 yo. The research included 43 722 healthy, unpregnant women with no background diseases, such as cancer or CVD, who did not work night shifts and were not daytime sleepers. Several parameters, such as height, weight, hip and waist circumference, were examined, and BMI values were calculated. The analysis showed that women, who were more exposed to ALAN, tended to have higher mean BMI, larger waist circumference, waist-to-hip ratio, and waist-to-height ratio.

In a separate study, Rybnikova et al. (2016) examined the association between ALAN and female and male overweight and obesity prevalence rates in 80 countries worldwide using obesity and overweight data available for 2012. The ambient ALAN exposures were estimated using the U.S. Defense Meteorological Satellite Program (US-DMSP) data. The analysis showed a positive relationship between ALAN and overweight/obesity in both men and women (p < 0.05).

Research hypothesis

As well established, vitamin D deficiency, on the one hand, and overweight/obesity and hormone-dependent cancers, on the other, are associated (see inter alia, Der et al. 2014; Park et al. 2015; Syed et al. 2019; Yao et al. 2015). In other studies, researchers demonstrate the association between ALAN exposure and the above-mentioned types of morbidity (Park et al. 2019; Rybnikova et al. 2015, 2016, 2017). However, what factor – that is, either sunlight or ALAN exposure – helps to explain better the country-wide incidence of the abovementioned morbidity types is largely unclear. Therefore, we hypothesize:

H₀:

Both types of the environmental exposures in question affect the morbidity types under analysis, either equally or with similar strength.

H₁:

One of the abovementioned exposure factors – i.e., either ALAN or solar radiation – has a stronger impact on the morbidity types in question than the other, and the effect of this factor remains statistically significant upon controlling for potential confounders, such as GDP_pc, societal inequality, unhealthy food consumption, and the counterpart factor, that is, either ALAN or solar radiation.

Data sources and research method

Obesity and overweight data

Data on the adult overweight and obesity rates for individual countries worldwide were obtained as ASRs from the World Health Organization database (WHO 2022b), separately for females and males. In the database, these data are available for the period between 1975 and 2016 for 196 countries worldwide. The prevalence rates of obesity were calculated as BMI ≥ 30, while overweight rates were calculated for 25 ≤ BMI <30, both as % of a country’s total adult population.

Cancer data

BC and PC incidence data for the study were obtained from the Global Health Data Exchange database (GHD 2019) as ASRs. In this database, BP and PC data are available for 204 countries worldwide, for the period between 1990 and 2016.

Solar radiation

Data on solar radiation were obtained from the Global Solar Atlas database (Agency et al. 2021). Because in large countries, actual levels of solar radiation tend to differ across different geographic areas and localities (see Figure 2b), the following normalization formula, proposed by Kloog et al. (2008), was used to calculate the population-weighted solar radiation exposure rates:

(1)

Figure 2. ALAN emissions and solar radiation worldwide.

(a) ALAN flux in 2014 (nanowatts/cm²/sr); (b) Solar radiation (kWh/m²).

Sources: Mapped using data from NOAA (2020) and Agency et al. (2021), respectively.

Notes: The raster resolution is ~920 m per pixel for ALAN and 250 m per pixel for solar radiation.

where and are population size of and solar radiation level in location j (that is, a map’s pixel) of country i, respectively.

Using this formula makes it possible to calculate country-wide averages of solar radiation exposure by giving greater weights to more populated areas and smaller weights to less populated areas. For the calculation, data on population distribution worldwide were obtained from the LandScan Database (Rose et al. 2021), where they are reported as population counts for 1 × 1 km grid cells. The calculation was performed in the ArcGIS^TM10.x software, using its zonal statistics tool (Esri 2016).

An example of population-weighted solar radiation levels by countries worldwide for the year 2016, calculated using Formula (1), is shown in Figure 3b, while solar radiation maps for other years and inter-year changes are reported in Appendix A (see Figure A1a-e).

Figure 3. Population-weighted ALAN and solar radiation levels by countries worldwide.

(a) Population-weighted ALAN levels (nanowatts/cm2/sr) in 2014; (b) Population-weighted solar radiation levels (kWh/m²) in 2016.

Source: Mapped using data from NOAA (2020) and Agency et al. (2021).

Note: Both ALAN and solar radiation averages are grouped into five classes using the Jenks natural breaks method. Due to the fact that actual exposures depend on population distribution within countries, a population-weighting procedure was implemented following Kloog et al. (2008) and using data from Rose et al. (2021) (see text for explanations).

ALAN data

ALAN data for the present analysis were downloaded from the U.S. Defense Meteorological Satellite Program (US-DMSP) database for the period between 1996/97 and 2013, and, after 2013, from the Suomi National Polar Partnership Visible Infrared Imaging Radiometer Suite (SNPP-VIIRS) website (NOAA 2020). An example of the ALAN data used in the study is shown in Figure 2a.

For the analysis, the ALAN data available for the year 2006 were matched with the year-2016 cancer data, while ALAN data for the year 2007 were matched with cancer data for the year-2017 and so on, to account for 10-year latency in BC and PC cancer incidence (Etzioni et al. 1998; Howard 2012). Concurrently, for obesity, whose latency is expected to be shorter, i.e., ~1–2 years (cf., e.g., Rybnikova et al. 2016), ALAN data for the year 2004 were matched with the year 2006 obesity/overweight data, and so on. As with solar radiation (see Subsection 4.3), a normalization formula similar to (1) was used to account for local geographic patterns of population distribution in individual countries:

(2)

where and are population size and average ALAN emission from pixel j of country i, respectively.

As with solar radiation, the calculations were performed in the ArcGIS^TM10.x software, using its zonal statistics tool (Esri 2016). An example of population-weighted ALAN levels by countries worldwide for 2014, calculated using Formula (2) is shown in Figure 3a, while ALAN maps for other years and inter-year change are reported in Appendix B (see Figure B1)

Socio-economic confounders

Information on several country-level socio-economic confounders was obtained from the following sources:

• Information on the average per capita GDP (in current US$) and GINI social inequality index was obtained from the World Bank database (WB 2022; WB and OECD 2018), where these data are available for the period between 1960 and 2019 for 257 countries worldwide (GDP_pc) and 175 countries (GINI), respectively.

• Information on healthy/unhealthy food consumption was obtained from the UN Food and Agriculture Organization (Ritchie and Roser 2017). In this database, oils, fats and sugar are classified as unhealthy, while cereals and grains, pulses, starchy roots, fruit and vegetables are classified as healthy (Bruso 2018). In the database, the food consumption data are available for the period between 1961 and 2019 for 166 countries worldwide.

As demonstrated in previous studies, diet and nutrition are important for maintaining good health throughout the life course (GHD 2019; Micha et al. 2017; Nishida et al. 2004; Vieira et al. 2013). The role of nutrition as a major determinant of chronic non-communicable diseases, such as cardio-metabolic diseases, obesity, type 2 diabetes, cancer, dental diseases, and osteoporosis is also well established (Micha et al. 2017; Nishida et al. 2004; WHO 2003, 2021). Concurrently, per capita GDP (GDP_pc) is an important indicator of population wealth, potentially reflecting differences in lifestyle, in general, and local differences in diet and healthcare quality. In particular, in several previous studies, GDPpc was found to be associated with breast and prostate cancers (Kelsey and Gammon 1991; Rybnikova et al. 2015, 2017) and with average BMI (Kark and Rasmussen 2005; Miech et al. 2006). Due to these considerations, the variables in question were included into the present analysis as potential confounders. Descriptive statistics of the research variables are reported in Table 1.

Table 1. Descriptive statistics of the research variables.

Variable	Unit of measurement	Year	Min	Max	Mean	SD
Overweight	Prevalence ASR (%)	1998	9.5	68.6	37.6	15.6
		2016	18.3	77.6	47.5	15.4
		1998–2016, %^a	92.6	13.1	26.5	−1.0
Obesity	Prevalence ASR (%)	1998	0.6	35.3	11.5	6.8
		2016	2.1	47.3	18.2	8.7
		1998–2016, %^a	250.0	34.0	58.8	26.8
Prostate cancer	ASR per 100,000	2006	7.3	123.1	44.0	30.6
		2016	8.0	123.1	46.0	28.3
		2006–2016, %^a	8.9	0.1	4.4	−7.5
Breast cancer	ASR per 100,000	2006	13.1	114.6	45.4	26.9
		2016	14.0	111.7	47.7	23.9
		2006–2016, %^a	6.3	−2.5	5.1	−11.4
GDP	US$	1996	145.2	52,571	6,364	10,729
		2006	194.7	89,739	10,273	16,499
		2014	371.3	118,823	14,038	20,653
		1996–2006, %^a	34.1	70.7	61.4	53.8
		1996–2014, %^a	155.8	126.0	120.6	92.5
GINI	%	1996	23.1	69.5	40.2	9.8
		2006	24.4	64.2	39.1	8.5
		2014	24.0	63.0	37.8	7.7
		1996–2006,%^a	5.5	−7.6	−2.8	−13.9
		1996–2014, %^a	3.8	−9.4	−6.1	−21.5
Unhealthy food consumption	(kcal/capita/day)	1996	49.0	1,540.0	642.6	311.1
		2006	123.0	1,506.0	694.6	302.6
		2014	149.0	1,499.0	709.8	299.8
		1996–2006, %^a	151.0	−2.2	8.1	−2.7
		1996–2014, %^a	204.1	−2.7	10.5	−3.6
ALAN	nanowatts/cm2/sr	1996	-	59.5	24.3	15.0
		2006	-	58.9	24.6	14.1
		2014	3.0	61.1	24.8	13.8
		1996–2006,%^a	-	−1.0	1.1	−6.1
		1996–2014,%^a	-	2.6	2.1	−7.7
Solar radiation	kWh/m^2	1998	0.0	6.3	4.6	1.1
		2006	0.0	6.3	4.6	1.1
		2016	0.0	6.3	4.6	1.1
		1998–2016, %^a	45.6	−0.1	0.0	−0.5
		2006–2016, %^a	−64.6	−0.1	0.0	−0.3

^aInter-year change, %.

Statistical analysis

The collected data (see Subsections 4.1-4.5) were co-integrated and analyzed using multiple-regression tools (“Linear Regression -IBM Documentation” n.d.). To account for disease latency (see Subsection 4.4), for BC and PC, 11 valid annual observation series were generated, spanning from 2006 to 2016 and containing 131 countries each (i.e., an 11 × 131 matrix). Concurrently, in the obesity/overweight data set, longer series with 19 annual observations were generated for 131 countries (from 1998 to 2016), considering a shorter ALAN latency period expected for these health conditions (see Subsection 4.4). Since solar radiation changes only slightly from year to year (Lindsey 2009), there was no need to lag solar radiation exposure estimates for either cancer or overweight/obesity.

The following generic equation was used to measure the strength of associations between the health effects under analysis and their potential determinants:

(3)

where HE_i,j,k is the health effect under analysis (k = 1 … 4 - either overweight, obesity, BC, or PC) observed in country i during year j; n = time lag; SOC = vector of country-level socio-economic confounders; REG = vector of regional dummies that reflects regional commonalities, not captured by the variables directly controlled by the analysis, that is, GDP_pc, societal inequality, and unhealthy food consumption; ALAN_ij is population weighed nighttime light intensity in country i during period j; SOL_i is population weighed solar radiation intensity in country i; b₀, b₁, b₂, β, and λ are regression coefficients; and is a random error term.

If our research hypothesis (see Section 3) is correct, only one of the abovementioned environmental factors – that is, either ALAN or solar radiation – should emerge, upon controlling for potential confounders, as a statistically significant predictor of the morbidity types under analysis at the accepted statistical significance level (p < 0.05).

Results

General trends

The worldwide geographic patterns of the morbidity types under analysis are illustrated in Figure 4, separately for BC (Figure 4a), PC (Figure 4b), obesity (Figure 4c), and overweight (Figure 4d).

Figure 4. Morbidity rates by countries worldwide in 2016.

(a) Breast cancer (BC) incidence rate per 100,000 (ASR); (b) Prostate cancer (PC) incidence rate per 100,000 (ASR); (c) Obesity prevalence among adults (ASR, %); (d) Overweight prevalence among adults (ASR, %).

Source: Mapped using data from GHD (2019) and WHO (2022b). The rates are grouped into five classes using the Jenks natural breaks method.

Notes: Other years and inter-year changes are reported in Appendixes C-F.

Figure 4. (Continued).

As evidenced in Figure 4a, the highest ASRs of BC are found in North America, Europe, Australia, and Oceania (78.27–111.71 cases per 100,000), while the lowest BC ASRs are observed in Central America, Africa, and Southeast Asia (15.37–26.16 cases per 100,000). Similar geographic patterns are also exhibited by PC, with the highest rates of this disease found in the United States, Northern Europe, Australia, and Oceania (85.79–122.90 cases per 100,000) and its lowest rates are recorded in Central America, Africa, the Middle East and most of Asia (7.96–25.59 cases per 100,000) (see Figure 4b).

Characteristically, the geographic patterns of overweight and obesity are somewhat different, with the highest rates of these health conditions found in North and South America, the Middle East, North Africa, Australia, and Oceania (26–47% for obesity, and 53.3–77.6% for overweight, respectively). Concurrently, the lowest rates of these health conditions are observed in Central Africa and Southeast Asia: 2–7% for obesity and 18.3–23.3% for overweight (see Figure 4c,d).

As further demonstrated in Figures 3a and 4a, the geographic patterns of BC and ALAN exhibit considerable similarities. In particular, the highest BC rates (Figure 4a) are found in the regions, where ALAN levels are highest (Figure 3a). Concurrently, the places with the highest levels of solar radiation (e.g., Central America and Central Africa; see Figure 3b) typically exhibit relatively low BC rates (Figure 4a). This indicates that both factors (i.e., ALAN and solar radiation) appear to be associated with BC, albeit the directions of these associations might differ being positive for ALAN-BC and negativefor solar radiation-BC.

With a notable exception of the Middle East and Northern Africa, the geographic patterns of ALAN (Figure 3a) and PC (Figure 4b) are also similar. In particular, the highest PC rates are observed in North America, Russia, and Northern Europe, that is, in the regions where solar radiation is relatively scant (see Figures 4b and 3b). As it is also demonstrated in Figure 4c,d and Figure 3a, the geographic patterns of overweight and obesity are similar to those of ALAN, with these health conditions being more common in the regions with high ALAN exposures, viz., North and South America, the Middle East, North Africa, Australia, and Oceania (see Figure 3a).

Regression analysis

The full set of models estimated for BC, PC, overweight, and obesity is reported in Appendix G (see Tables G1–G18). Concurrently, changes in the t-statistics of the ALAN and solar radiation variables in the models, estimated for different years between 1998 and 2016 (overweight and obesity) and for 2006–2016 (BC and PC), are reported in Figure 5.

Figure 5. T-statistics of the ALAN and solar radiation variables, estimated for separate years between 1998 and 2016 for different types of morbidity.

(a) BC; (b) PC; (c) Obesity; (d) Overweight.

Notes: Based on the models reported in Appendix G, Tables G1-G18. Dotted red lines indicate a 0.01 two-tailed significance level, dotted black lines indicate a 0.05 significance level (two-tailed).

As evidenced in Tables G1–G18, the estimated models provide a reasonably good fit, with the coefficients of determination ranging between R² = 0.639 (PC in the year 2012) and R² = 0.844 (BC in the year 2007). The estimated models also exhibit a high degree of generality, with F-statistics ranging from 29.475 (p < 0.01) in the model estimated for obesity (Appendix G Table G4) to 82.231 (p < 0.01) in the model estimated for BC (Appendix G Table G13) (p < 0.01).

Notably, in all the models estimated for BC, the main factors affecting this type of morbidity are ALAN (+) and GDPpc (+). In several models, unhealthy food consumption also emerges as a positive predictor of BC at a 0.05 probability level (see Appendix G Tables G13–G15). Concurrently, significant determinants of PC in most models are as follows: ALAN (+), GDPpc (+), GINI (+) and solar radiation (-) (p < 0.05), while the factors significantly associated with obesity and overweight are: ALAN (+), GDPpc(-) and unhealthy food consumption (+) (p < 0.05) (see Appendix G Tables G1–G18).

As further evidenced in Figure 5, ALAN has a statistically significant and positive effect on all types of morbidity under analysis (p < 0.05), and this influence does not change substantially over the years. Concurrently, when introduced jointly with ALAN, the solar radiation variable has a significant negative effect on PC only (p < 0.05), while its effects on BC, overweight and obesity are insignificant (p > 0.1).

Factor contribution test

The relative contribution of different predictors to obesity, overweight, BC and PC models, estimated by a consequent variable ablation, is reported in Figure 6. The test was run using the best-performing models, estimated for each morbidity type based on its R², that is, Models 1, 22, 44, and 55 from Tables G1, G7, G13, and G14. During the test, each variable was removed from the model consequently, and change in the coefficient of determination, R², was monitored. Then, the variable in question was returned to the model and another variable was omitted, and so on.

Figure 6. Relative contribution of different factors to obesity (a), overweight (b), BC (c), and PC (d).

Notes: The factors are ranked based on Model 1 and Model 22 for obesity and overweight (see Appendixes, Tables G1 and G7) and Model 44 and Model 55 for BC and PC (see Appendixes, Tables G13 and G16); the variables are sorted in the order of diminishing contribution, with the relative contribution of the strongest variable conditionally set to 100%.

As evidenced in Figure 6, ALAN contributes most to the obesity and overweight models, while solar radiation is found to have a minor contribution to obesity (Figure 6a) and overweight (Figure 6b). Concurrently, as shown in Figure 6c, GDP_pc, South America, Africa, and Oceania are the main contributors to BC, followed by ALAN, unhealthy food consumption, and GINI. Notably, solar radiation contributes the least to BC, while Asia and GDP_pc contribute the most to PC (Figure 6d), followed by solar radiation and ALAN.

Discussion

Exposure to ALAN affects the internal cyclicity that occurs in the SCN cells (Aulinas 2019; Haim and Portnov 2013). Moreover, it affects MLT secretion through different pathways. The first – visual – pathway leads from the eye retina to the occipital region of the brain that produces a visual sensation (Rea et al. 2005; Van Bommel 2006), while the second (more recently discovered) – non-visual pathway – is pathway, in which the radiant energy travels to other parts of the brain where no visual sensation is produced (Zaidi et al. 2007). According to Reiter et al. (2007), MLT has antioxidative properties. As a result, suppression of MLT affects the biological function resulting in changing estrogen receptor affinity and increasing susceptibility to different diseases, including hormone-dependent cancers, such as BC and PC (Blask et al. 2005; Figueiro 2017; Haim and Portnov 2013). Moreover, ALAN is associated with obesity (Rybnikova et al. 2016); fatigue (Martin et al. 2012); poor sleep quality (Gabinet and Portnov 2021, 2022; Smolensky et al. 2015), and general stress (Haim and Portnov 2013). Concurrently, vitamin D, produced in the skin in response to exposure to sunlight radiation (Jeon and Shin 2018) and metabolized in the liver and other tissues (Bikle 2000), improves the immune function (Ponsonby et al. 2005) and lowers the risk of inflammation (Agrawal and Yin 2014). As revealed by previous studies, the lack of vitamin D is associated with obesity (Yao et al. 2015); diabetes (Hussain Gilani et al. 2019); PC (Xie et al. 2017), and BC (Pineda-Moncusí et al. 2018). However, which of these explanations is more valid remains largely unclear, as in previous studies, the effects of ALAN and sunlight exposure on human health have been, most often, investigated separately, that is, independently from each other.

In the present study, we attempted to narrow this knowledge gap, by contrasting sunlight-ALAN exposure estimates with cancer morbidity and overweight/obesity data available for 131 countries worldwide, while controlling for potential confounders, including GDP_PC, GINI inequality index, and unhealthy food intake. Data for the study were obtained from multiple sources, including WHO (2022b), NOAA (2020), World Bank (WB 2022; WB and OECD 2018), and others.

In order to account for geographic patterns of population distribution, a population weighting procedure was applied to ALAN and solar radiation data, enabling us to assign greater weights to locations with larger population counts. The cointegrated data were mapped and analyzed using multiple statistical tools. To account for disease latency, ALAN lags for BC and PC were set to 10 years, following the estimates reported in Rybnikova et al. (2015, 2017). Concurrently, ALAN data for obesity and overweight were lagged by 2 years, considering a shorter latency period is expected for these health conditions.

As the study revealed, the highest ASRs of BC are found in North America, Europe, Australia, and Oceania, while the lowest BC ASRs are observed in Central America, Africa, and Southeast Asia. Similar geographic patterns are also exhibited by PC, with the highest rates of this disease found in the United States, Northern Europe, Australia, and Oceania and its lowest rates recorded in Central America, Africa, the Middle East, and most of Asia. The geographic patterns of overweight and obesity are found to be somewhat different, with their highest rates found in North and South America, the Middle East, North Africa, Australia, and Oceania, while their lowest rates are observed in Central Africa and Southeast Asia.

As further revealed in the analysis, the geographic patterns of BC/PC and ALAN exhibit considerable similarities, with the highest BC and PC rates observed in the regions, where ALAN levels are highest. The geographic patterns of overweight and obesity are also similar to those of ALAN, with these health conditions being more common in the regions with high ALAN exposures, viz., North and South America, the Middle East, North Africa, Australia and Oceania. Moreover, all the morbidity types under analysis were found to be significantly and positively associated with ALAN (p < 0.05), but not with solar radiation exposure estimates, which was found to affect PC rates only (p < 0.05), but not BC, overweight and obesity (p > 0.1).

These results thus provide partial support for our H1, according to which only one environmental risk factor – i.e., either ALAN or solar radiation – was expected to affect the morbidity types under analysis. In particular, the hypothesis in question was upheld for BC, overweight, and obesity (p < 0.05), but not for PC, for which both ALAN and solar radiation emerged as statistically significant (p < 0.05).

The coefficients of determination in the estimated models (R² = 0.643–0.844) are similar to those obtained by Rybnikova et al. (2015, 2016, 2017) and Park et al. (2019) that link ALAN exposure to BC, PC, and obesity/overweight rates, attributing this association to MLT suppression and circadian disruption, caused by prolonged ALAN exposure (Reiter et al. 2007; Smith et al. 2004; Stevens et al. 2014).

However, our results differ from those reported by Yao et al. (2015), Syed et al. (2019), and Park et al. (2015) who attribute BC, obesity, and overweight rates to vitamin D deficiency, linked, in turn, to insufficient sunlight exposure. This difference in findings may be due to the fact that these studies analyzed only one factor (i.e., solar radiation), while another factor (ALAN) was overlooked, which might have led to spurious results. Another possible reason for the differences in findings might be attributed to the fact that the studies in question were focused on specific population groups (i.e., patients in hospitals and specialized medical institutions) which limits their generalizability.

Several limitations of the present analysis need to be mentioned. First and foremost, the current study is ecological research based on average country-level data. A well-known limitation of ecological studies is potential misclassification of exposure, as people who suffer from a given disease might not be the same people who were exposed to a particular risk factor (Hart 2011). To address this issue, we implemented a population weighting procedure, which allowed us to obtain more accurate ALAN and solar radiation exposure estimates. Nevertheless, future studies should attempt to assign different weights to different countries, based on, e.g., their population size or land area.

Another limitation is that an increase in exposure to a particular risk factor might not necessarily lead to an immediate increase in morbidity, reflecting the phenomenon known as disease latency (Howard 2012; Rybnikova et al. 2016). To address this issue, we used different latency periods for ALAN exposure estimates. Moreover, outdoor ALAN, which penetrates human habitats during nighttime and is captured by satellites, causes circadian disruption due to nighttime activities it enables, and its negative effects on MLT levels (Haim and Portnov 2013; Stevens and Rea 2001). Concurrently, indoor ALAN exposure, which is captured by indoor ALAN measurements, may also result in the suppression of the MLT hormone and circadian disruption (Haim and Portnov 2013). Although indoor and outdoor ALAN exposures are not necessarily linked, both of them might thus contribute to increased morbidity by becoming endocrine disruptors. The present study is focused, as explicitly mentioned, on the outdoor ALAN exposure, while individual-level studies might be needed to investigate the effect of indoor ALAN exposure on the morbidity types under analysis. Finally, as ecological studies cannot control for potential individual-level cofounders, e.g., individual income, family history, etc., our findings should, therefore, be tested in individual-level follow-up studies.

Conclusion

The present study is an ecological analysis aimed to disentangle the effects of ALAN and solar radiation on three types of morbidity: BC in women, PC in men, and overweight/obesity in both genders. Although the study was performed for countries, not for individuals, it might improve our knowledge about the etiology of the above diseases, thus helping to spearhead public health policies. In particular, as the present study reveals, ALAN exposure is significantly associated with the morbidity types under analysis, while the links of these diseases with solar radiation are weaker. This finding may necessitate regulatory measures that would reduce ALAN exposure, both indoors and outdoors. As Schroer and Hölker (2014) note, there are several ways to minimize the effect of ALAN on natural ecosystems and humans, including the use of dimmers and smart lighting control devices that turn off when they are not needed (Aussat et al. 2022), shielded luminaires that minimize ALAN spillover to neighboring areas (Bara et al. 2020), and limiting lighting level on streets and roads to the levels that take into account personal safety and road safety but not lead to extensive and unneeded lighting (Portnov et al. 2020, 2022; Svechkina et al. 2020). There is also a need for follow-up individual-level studies, aimed to verify our results against individual-level data, while controlling ALAN and sunlight exposure estimates for potential individual-level confounders.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The author(s) reported that there is no funding associated with the work featured in this article.

Appendix A

Figure A1. Population-weighted national averages of solar radiation by countries worldwide, kWh/m².

(a) Year 1998; (b) Year 2006, (c) Year 2016, (d) Percent change between 1998 and 2016; (e) Percent change between 2006 and 2016.

Source: Authors’ calculations applied to Agency et al (2021) - Global Solar Atlas (see text for explanations); Note: Solar radiation average is grouped into five classes using the Jenks natural breaks method. The solar radiation raster is annually invariant. The changes in the maps are due to changes in the population distribution (see text for explanations).

Figure A1. (Continued).

Appendix B

Figure B1. Population-weighted ALAN averages by countries worldwide, nanowatts/cm2/sr.

(a) Year 1996; (b) Year 2006, (c) Year 2014, (d) Percent change between 1996 and 2006; (e) Percent change between 1996 and 2014.

Source: Authors’ calculations applied to NOAA (2020) data (see text for explanations); Note: ALAN averages are grouped into five classes using the Jenks natural breaks method.

Figure B1. (Continued).

Appendix C

Figure C1. BC rates by countries worldwide.

(a) Year 2006; (b) Year 2016; (c) BC rate change between 2006 and 2016, %.

Source: Mapped using GHD (2019) - Global Health Data Exchange; the BC rates are grouped into five classes using the Jenks natural breaks method.

Appendix D

Figure D1. BC rates by countries worldwide.

(a) Year 2006; (b) Year 2016; PC rate change between 2006 and 2016, %.

Source: Mapped using GHD (2019); the PC rates are grouped into five classes using the Jenks natural breaks method.

Appendix E

Figure E1. Overweight ASR by countries worldwide, %.

(a) Year 1998, (b) Year 2016; and (c) Obesity rate change between 1998 and 2016, %.

Source: Mapped using WHO (2022b) - the World Health Organization countrywide database: the obesity rates are grouped into five classes using the Jenks natural breaks method.

Appendix F

Figure F1. Overweight ASR by countries worldwide, %.

(a) Year 1998, (b) Year 2016; (c) Rate change between 1998 and 2016, %.

Source: Mapped using WHO (2022a) - the World Health Organization countrywide database. Overweight rates are grouped into five classes using the Jenks natural breaks method.

Appendix G.

Factors affecting obesity, overweight, BC, PC – Method: linear regression

Table G1. Factors affecting obesity rates in individual countries of the world (Dependent variable – Obesity, % ASR in the years 1998; 1999; 2000 and 2001; explanatory 2-year latency period, unless stated otherwise).

Predictor	Model 1		Model 2		Model 3		Model 4
	1998		1999		2000		2001
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	7.350	3.119***	8.113	3.324***	6.924	2.76***	8.376	3.184***
GDP per capita (US$)	2.5E^−4	−5.37***	2E^−4	−5.31***	2E^−4	−4.773***	2E^−4	−4.63***
GINI	0.003	0.075*	−0.005	−0.1*	−0.006	−0.128*	−0.004	−0.087*
Unhealthy food consumption (kcal/capita/day)	0.003	1.498*	0.004	2.184**	0.004	1.908*	0.003	1.528*
ALAN, nanowatts/cm2/sr	0.319	7.85***	0.306	7.232***	0.318	7.36***	0.322	7.035***
Solar Radiation^c, kWh/m^2	−0.556	−1.188	−0.523	−1.066	−0.285	−0.57	−0.424	−0.804
Regional dummies:
Africa	−2.796	−2.404**	−3.192	−2.679***	−3.014	−2.442**	−3.309	−2.561**
Asia	−4.028	−3.788***	−4.434	−4.052***	−4.218	−3.76***	−4.443	−3.742***
Oceania	13.518	5.697***	13.471	5.474***	13.777	5.488***	13.990	5.344***
No. of obs.	131		131		131		131
R^2	0.726		0.715		0.714		0.699
R^2- adjusted	0.709		0.697		0.695		0.679
F	40.501***		38.335***		38.045***		35.449
SEE^d	3.712		3.842		3.914		4.078

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 1998, 1999, 2000, and 2001, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G2. Factors affecting obesity rates in individual countries of the world (dependent variable – obesity, % ASR in the years 2002; 2003; 2004 and 2005; explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 5		Model 6		Model 7		Model 8
	2002		2003		2004		2005
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	6.749	2.572**	6.841	2.5**	10.838	3.668***	8.183	2.882***
GDP per capita (US$)	0.000	−4.678***	0.000	−4.493***	0.000	−3.744***	0.000	−4.679***
GINI	0.033	0.642*	0.021	0.397*	0.065	1.153*	−0.003	−0.056*
Unhealthy food consumption (kcal/capita/day)	0.003	1.644*	0.004	1.995**	0.007	3.05***	0.005	2.449**
ALAN, nanowatts/cm2/sr	0.341	7.239***	0.328	6.991***	0.274	7.193***	0.319	7.196***
Solar Radiation^c, kWh/m^2	−0.549	−1.031*	−0.424	−0.787*	−2.033	−3.932***	−0.506	−0.887*
Regional dummies:
Africa	−2.653	−2.006**	−3.006	−2.281**	N/A	N/A	−2.998	−2.205**
Asia	−3.781	−3.157***	−4.047	−3.336***	−2.548	−2.441**	−4.405	−3.527**
Oceania	15.250	5.761***	13.848	5.25***	14.802	5.419***	14.513	5.372***
No. of obs.	131		131		131		131
R^2	0.707		0.708		0.667		0.711
R^2- adjusted	0.688		0.689		0.648		0.692
F	36.771***		36.999***		35.26***		37.474***
SEE^d	4.096		4.144		4.461		4.246

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2002, 2003, 2004, and 2005, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G3. Factors affecting obesity rates in individual countries of the world (dependent variable – obesity, % ASR in the years 2006; 2007; 2008 and 2009; explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 9		Model 10		Model 11		Model 12
	2006		2007		2008		2009
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	8.847	3.07***	6.729	2.245**	8.368	2.739***	9.681	2.861***
GDP per capita (US$)	0.000	−4.479***	0.000	−4.617***	0.000	−3.789***	0.000	−3.234***
GINI	−0.011	−0.195*	0.045	0.788*	−0.001	−0.021*	−0.008	−0.125*
Unhealthy food consumption (kcal/capita/day)	0.007	3.132***	0.007	3.408***	0.006	2.775***	0.007	2.912***
ALAN, nanowatts/cm2/sr	0.296	7.05***	0.361	8.320***	0.318	7.154***	0.262	5.634***
Solar Radiation^c, kWh/m^2	−0.650	−1.144*	−1.223	−2.404**	−0.475	−0.799*	−0.472	−0.722*
Regional dummies:
Africa	−3.267	−2.468**	N/A	N/A	−3.330	−2.417**	−4.052	−2.719***
Asia	−4.612	−3.785***	−2.741	−2.726***	−4.908	−3.891***	−5.885	−4.322***
Oceania	13.945	5.197***	16.394	6.010***	14.622	5.246***	13.879	4.585***
No. of obs.	131		131		131		131
R^2	0.721		0.708		0.714		0.673
R^2- adjusted	0.702		0.692		0.696		0.651
F	39.324***		42.642***		38.154***		31.339***
SEE^d	4.234		4.360		4.387		4.764

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2006, 2007, 2008, and 2009, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G4. Factors affecting obesity rates in individual countries of the world (dependent variable – obesity, % ASR in the years 2010; 2011; 2012 and 2013; explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 13		Model 14		Model 15		Model 16
	2010		2011		2012		2013
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	8.782	2.52**	10.220	2.89***	5.353	1.54*	7.678	2.166**
GDP per capita (US$)	0.000	−3.678***	0.000	−3.385***	0.000	−4.045***	0.000	−3.76***
GINI	0.023	0.339*	−0.001	−0.016*	0.044	0.633*	0.043	0.6*
Unhealthy food consumption (kcal/capita/day)	0.008	3.452***	0.008	3.365***	0.009	3.635***	0.009	3.754***
ALAN, nanowatts/cm2/sr	0.312	6.873***	0.236	5.183***	0.315	6.881***	0.301	6.781***
Solar Radiation^c, kWh/m^2	−1.307	−2.191**	−0.485	−0.72*	−0.883	−1.481*	−1.255	−2.043*
Regional dummies:
Africa	N/A	N/A	−4.548	−2.974***	N/A	N/A	N/A	N/A
Asia	−3.707	−3.163***	−5.878	−4.137***	−3.632	−3.063***	−3.514	−2.915***
Oceania	16.162	5.259***	13.419	4.254***	16.961	5.322***	17.212	5.307***
No. of obs.	131		131		131		131
R^2	0.65		0.659		0.651		0.647
R^2- adjusted	0.63		0.637		0.632		0.627
F	32.609***		29.475***		32.844***		32.243***
SEE^d	4.955		4.978		5.071		5.166

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2010, 2011, 2012, and 2013, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G5. Factors affecting obesity rates in individual countries of the world (dependent variable – obesity, % ASR in the years 2014; 2015 and 2016; explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 17		Model 18		Model 19
	2014		2015		2016
	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	5.502	1.571*	6.310	1.724**	10.502	2.910***
GDP per capita (US$)	0.000	−3.963***	0.000	−3.630***	0.000	−3.463***
GINI	0.048	0.659*	0.050	0.651*	−0.001	−0.014*
Unhealthy food consumption (kcal/capita/day)	0.010	4.049***	0.011	4.571***	0.010	4.327***
ALAN, nanowatts/cm2/sr	0.324	7.123***	0.292	6.58***	0.340	7.621***
Solar Radiation^c, kWh/m^2	−0.937	−1.529*	−1.056	−1.666**	−1.280	−2.062**
Regional dummies:
Oceania	−3.605	−3.016***	−3.692	−2.961***	−3.508	−2.877***
Asia	18.058	5.577***	17.107	5.148***	17.831	5.557***
No. of obs.	131		131		131
R^2	0.658		0.643		0.664
R^2- adjusted	0.638		0.622		0.645
F	33.768***		31.58***		34.793***
SEE^d	5.145		5.300		5.191

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2014, 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G6. Factors affecting obesity rates in individual countries of the world (dependent variable – obesity, % ASR in the years 2015 and 2016 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 20		Model 21
	2015		2016
	B^a	t^b	B^a	t^b
(Constant)	14.650	4.118***	12.925	3.588***
GDP per capita (US$)	0.000	−2.490**	0.000	−2.599**
GINI	−0.019	−0.262*	0.010	0.14*
Unhealthy food consumption (kcal/capita/day)	0.011	4.929***	0.011	4.722***
ALAN, nanowatts/cm2/sr	0.353	6.872***	0.386	8.031***
Solar Radiation^c, kWh/m^2	−1.085	−1.579*	−1.418	−2.32**
Regional dummies:
Africa	−3.402	−2.226**	N/A	N/A
Asia	−4.382	−3.172***	−2.315	−1.986**
Oceania	14.738	4.796***	17.925	5.691***
No. of obs.	131		131
R^2	0.697		0.676
R^2- adjusted	0.678		0.657
F	35.157***		36.649***
SEE^d	4.895		5.102

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G7. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 1998; 1999; 2000 and 2001 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 22		Model 23		Model 24		Model 25
	1998		1999		2000		2001
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	33.886	7.136***	35.994	7.414***	33.389	6.77***	36.61	7.213***
GDP per capita (US$)	0	−4.502***	0	−4.634***	0	−4.095***	−0.0004	−4.161***
GINI	−0.024	−0.285*	−0.043	−0.476*	−0.04	−0.432*	−0.0392	−0.401*
Unhealthy food consumption (kcal/capita/day)	0.004	0.907*	0.006	1.534*	0.005	1.219*	0.00379	0.925*
ALAN, nanowatts/cm2/sr	0.685	8.364***	0.664	7.883***	0.679	7.999***	0.68405	7.738***
Solar Radiation^c, kWh/m^2	−0.1315	−1.392*	−1.331	−1.365*	−0.835	−0.849*	−1.1719	−1.150*
Regional dummies:
Africa	−10.969	−4.68***	−11.536	−4.867***	−11.173	−4.604***	−11.635	−4.667***
Asia	−11.213	−5.234***	−12.044	−5.533***	−11.426	−5.181***	−11.766	−5.136***
Oceania	19.475	4.073***	19.125	3.907***	19.453	3.941***	19.553	3.871***
No. of obs.	131		131		131		131
R^2	0.786		0.777		0.775		0.766
R^2- adjusted	0.771		0.763		0.761		0.751
F	55.859***		53.283***		52.653***		49.969***
SEE^d	7.480		7.641		7.695		7.870

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 1998, 1999, 2000, and 2001, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G8. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 2002; 2003; 2004 and 2005 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 26		Model 27		Model 28		Model 29
	2002		2003		2004		2005
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	33.820	6.774***	34.961	6.763***	44.225	8.14***	37.482	7.212***
GDP per capita (US$)	0.000	−4.27***	0.000	−4.1***	0.000	−2.946***	0.000	−4.298***
GINI	0.030	0.301*	−0.009	−0.087*	−0.001	−0.008*	−0.066	−0.64*
Unhealthy food consumption (kcal/capita/day)	0.004	0.927*	0.005	1.191*	0.008	1.855*	0.006	1.586*
ALAN, nanowatts/cm2/sr	0.719	8.029***	0.685	7.734***	0.503	6.586***	0.655	8.071***
Solar Radiation^c, kWh/m^2	−1.409	−1.39*	−1.131	−1.111*	−2.725	−2.348**	−1.151	−1.102*
Regional dummies:
Africa	−10.362	−4.118***	−11.213	−4.505***	−11.668	−4.404***	−11.109	−4.465***
Asia	−10.414	−4.571***	−11.061	−4.825***	−11.666	−4.831***	−11.552	−5.053***
Oceania	21.664	4.302***	18.509	3.714***	16.850	3.253***	19.094	3.861***
No. of obs.	131		131		131		131
R^2	0.772		0.77		0.748		0.774
R^2- adjusted	0.757		0.755		0.732		0.759
F	51.76***		51.079***		45.306***		52.143***
SEE^d	7.791		7.829		8.188		7.772

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2002, 2003, 2004, and 2005, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G9. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 2006; 2007; 2008 and 2009 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 30		Model 31		Model 32		Model 33
	2006		2007		2008		2009
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	39.087	7.432***	36.734	6.864***	38.163	7.105***	40.448	6.844***
GDP per capita (US$)	0.000	−3.915***	0.000	−3.815***	0.000	−3.473***	0.000	−2.932***
GINI	−0.079	−0.753*	−0.013	−0.122*	−0.055	−0.515*	−0.063	−0.547*
Unhealthy food consumption (kcal/capita/day)	0.008	2.171**	0.008	2.046**	0.008	1.98**	0.009	2.243**
ALAN, nanowatts/cm2/sr	0.601	7.847***	0.655	7.94***	0.627	8.024***	0.511	6.299***
Solar Radiation^c, kWh/m^2	−1.460	−1.408*	−1.501	−1.452*	−1.210	−1.156*	−1.226	−1.073*
Regional dummies:
Africa	−11.675	−4.833***	−10.652	−4.342***	−11.503	−4.749***	−12.751	−4.899***
Asia	−11.779	−5.296***	−10.898	−4.903***	−11.878	−5.355***	−13.517	−5.685***
Oceania	17.678	3.61***	19.581	3.944***	18.193	3.712***	16.422	3.106***
No. of obs.	131		131		131		131
R^2	0.777		0.776		0.776		0.739
R^2- adjusted	0.762		0.762		0.761		0.722
F	53.05***		52.912***		52.845***		43.271***
SEE^d	7.727		7.718		7.715		8.320

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2006, 2007, 2008, and 2009, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G10. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 2010; 2011; 2012 and 2013 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 34		Model 35		Model 36		Model 37
	2010		2011		2012		2013
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	40.488	6.844***	41.411	6.816***	35.113	6.06***	38.824	6.68***
GDP per capita (US$)	0.000	−2.876***	0.000	−2.954***	0.000	−3.233***	0.000	−2.938***
GINI	−0.070	−0.605*	−0.059	−0.487*	−0.024	−0.205*	−0.027	−0.230*
Unhealthy food consumption (kcal/capita/day)	0.009	2.134**	0.012	2.774***	0.010	2.364**	0.010	2.452**
ALAN, nanowatts/cm2/sr	0.515	6.32***	0.437	5.596***	0.496	6.179***	0.472	6.092***
Solar Radiation^c, kWh/m^2	−1.068	−0.929*	−1.120	−0.968*	−0.388	−0.352*	−0.985	−0.866*
Regional dummies:
Africa	−12.729	−4.86***	−13.614	−5.182***	−13.003	−5.042***	−12.690	−4.842***
Asia	−13.173	−5.516***	−13.221	−5.418***	−12.826	−5.467***	−12.424	−5.266***
Oceania	17.080	3.218***	14.711	2.715***	17.352	3.221***	17.464	3.216***
No. of obs.	131		131		131		131
R^2	0.735		0.723		0.733		0.729
R^2- adjusted	0.718		0.704		0.716		0.711
F	42.325***		39.730***		41.913***		40.976***
SEE^d	8.363		8.551		8.366		8.426

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2010, 2011, 2012, and 2013, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G11. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 2014; 2015 and 2016 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 38		Model 39		Model 40
	2014		2015		2016
	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	35.397	6.233***	37.106	6.385***	43.628	7.706***
GDP per capita (US$)	0.000	−3.234***	0.000	−2.99***	0.000	−2.849***
GINI	−0.017	−0.141*	−0.027	−0.218*	−0.098	−0.824*
Unhealthy food consumption (kcal/capita/day)	0.011	2.752***	0.012	3.18***	0.011	2.899***
ALAN, nanowatts/cm2/sr	0.507	6.455***	0.449	6.053***	0.525	7.046***
Solar Radiation^c, kWh/m^2	−0.575	−0.521*	−0.606	−0.538*	−1.128	−1.025*
Regional dummies:
Africa	−12.285	−4.776***	−12.849	−4.981***	−11.982	−4.725***
Asia	−12.225	−5.311***	−12.612	−5.381***	−11.882	−5.161***
Oceania	18.496	3.455***	16.483	3.073***	17.533	3.408***
No. of obs.	131		131		131
R^2	0.738		0.729		0.742
R^2- adjusted	0.72		0.711		0.725
F	42.854***		41.004***		43.906***
SEE^d	8.267		8.351		8.120

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2014, 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G12. Factors affecting overweight rates in individual countries of the world (dependent variable – overweight, % ASR in the years 2015 and 2016 explanatory 2-year latency period, unless stated otherwise).

Predictors	Model 41		Model 42
	2015		2016
	B^a	t^b	B^a	t^b
(Constant)	49.453	8.574***	47.221	8.328***
GDP per capita (US$)	0.000	−2.055**	0.000	−1.997***
GINI	−0.094	−0.80*	−0.077	−0.659**
Unhealthy food consumption (kcal/capita/day)	0.015	4.324***	0.012	3.269***
ALAN, nanowatts/cm2/sr	0.610	7.318***	0.593	7.337***
Solar Radiation^c, kWh/m^2	−2.109	−1.892*	−1.410	−1.289*
Regional dummies:
Africa	−11.510	−4.644***	−11.617	−4.62***
Asia	−10.110	−4.512***	−9.824	−4.294***
Oceania	15.333	3.077***	17.744	3.494***
No. of obs.	131		131
R^2	0.755		0.748
R^2- adjusted	0.739		0.732
F	47.005***		45.35***
SEE^d	7.938		8.022

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G13. Factors affecting breast cancer rates in individual countries of the world (dependent variable – breast cancer rate ASR in the years 2006; 2007; 2008 and 2009 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 43		Model 44		Model 45		Model 46
	2006		2007		2008		2009
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	35.441	4.373***	37.575	4.736***	36.585	4.359***	37.661	4.442***
GDP per capita (US$)	0.001	5.485***	0.001	5.469***	0.001	5.9***	0.001	5.437***
GINI	−0.017	−0.116*	−0.035	−0.235*	−0.022	−0.141*	−0.026	−0.157*
Unhealthy food consumption (kcal/capita/day)	0.013	2.121**	0.010	1.748*	0.008	1.444*	0.011	1.834*
ALAN, nanowatts/cm2/sr	0.497	4.079***	0.542	4.519***	0.523	4.234***	0.519	4.103***
Solar Radiation^c, kWh/m^2	−0.243	−0.153*	0.060	0.039*	0.250	0.155*	−0.124	−0.077*
Regional dummies:
Africa	−20.636	−4.544***	−22.295	−5.078***	−21.819	−4.769***	−21.216	−4.601***
Asia	−16.144	−4.206***	−17.952	−4.778***	−16.865	−4.367***	−16.233	−4.103***
South America	−23.570	−4.797***	−25.220	−5.174***	−24.443	−4.782***	−23.543	−4.509***
No. of obs.	131		131		131		131
R^2	0.834		0.844		0.834		0.832
R^2- adjusted	0.823		0.833		0.823		0.821
F	76.678***		82.231***		76.551***		75.678***
SEE^d	11.360		11.086		11.470		11.480

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2006, 2007, 2008, and 2009, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G14. Factors affecting breast cancer rates in individual countries of the world (dependent variable – breast cancer rate ASR in the years 2010; 2011; 2012 and 2013 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 47		Model 48		Model 49		Model 50
	2010		2011		2012		2013
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	38.639	4.504***	37.252	4.235***	45.022	5.153***	33.767	4.009***
GDP per capita (US$)	0.001	5.131***	0.001	5.419***	0.001	5.547***	0.001	4.654***
GINI	0.021	0.12*	0.049	0.272*	0.067	0.364*	0.072	0.408*
Unhealthy food consumption (kcal/capita/day)	0.009	1.533*	0.009	1.524*	0.011	1.836*	0.012	2.163**
ALAN, nanowatts/cm2/sr	0.551	4.337***	0.512	3.972***	0.423	4.023***	0.518	4.54***
Solar Radiation^c, kWh/m^2	−0.703	−0.336*	−0.562	−0.349*	−2.626	−1.532*	−0.786	−0.494*
Regional dummies:
Africa	−19.952	−4.273***	−20.118	−4.270***	−17.278	−3.584***	−17.482	−3.838***
Asia	−15.562	−3.966***	−14.869	−3.720***	−13.072	−3.258***	−12.640	−3.262***
South America	−22.844	−4.355***	−22.594	−4.174***	−18.813	−3.501***	−20.268	−3.906***
No. of obs.	131		131		131		131
R^2	0.824		0.818		0.804		0.815
R^2- adjusted	0.812		0.806		0.791		0.803
F	71.289***		68.363***		62.447***		67.132***
SEE^d	11.507		11.559		11.711		11.122

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2010, 2011, 2012, and 2013, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G15. Factors affecting breast cancer rates in individual countries of the world (dependent variable – breast cancer rate ASR in the years 2014; 2015 and 2016 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 51		Model 52		Model 53
	2014		2015		2016
	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	32.531	3.726***	33.119	3.705***	31.674	3.594***
GDP per capita (US$)	0.001	4.661***	0.000	3.91***	0.000	3.867***
GINI	0.090	0.496*	0.037	0.204*	0.117	0.642*
Unhealthy food consumption (kcal/capita/day)	0.011	1.87*	0.015	2.439**	0.011	1.968*
ALAN, nanowatts/cm2/sr	0.502	4.404***	0.513	4.235***	0.563	4.871***
Solar Radiation^c, kWh/m^2	−0.313	−0.19*	−0.376	−0.224*	−0.404	−0.241*
Regional dummies:
Africa	−19.323	−4.07***	−17.776	−3.638***	−18.784	−3.96***
Asia	−13.901	−3.441***	−12.909	−3.14***	−14.004	−3.48***
South America	−21.810	−4.52***	−20.332	−3.802***	−22.236	−4.25***
No. of obs.	131		131		131
R^2	0.792		0.784		0.780
R^2- adjusted	0.778		0.770		0.765
F	58.094***		55.332***		53.933***
SEE^d	11.533		11.737		11.605

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2014, 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G16. Factors affecting prostate cancer rates in individual countries of the world (dependent variable – prostate cancer rate ASR in the years 2006; 2007; 2008 and 2009 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 54		Model 55		Model 56		Model 57
	2006		2007		2008		2009
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	30.667	2.69***	32.204	2.799***	31.002	2.556**	32.680	2.656***
GDP per capita (US$)	0.001	5.019***	0.001	4.756***	0.001	4.715***	0.001	4.248***
GINI	0.430	2.128**	0.446	2.132**	0.460	2.065**	0.527	2.278**
Unhealthy food consumption (kcal/capita/day)	0.011	1.265*	0.010	1.103*	0.008	0.878*	0.009	0.951*
ALAN, nanowatts/cm2/sr	0.407	2.296**	0.463	2.578**	0.465	2.479**	0.479	2.479**
Solar Radiation^c, kWh/m^2	−4.996	−2.501**	−5.307	−2.629**	−4.981	−2.323**	−5.795	−2.702***
Regional dummies:
Asia	−22.058	−5.318***	−22.457	−5.342***	−22.235	−5.094***	−22.251	−4.981***
No. of obs.	131		131		131		131
R^2	0.668		0.672		0.649		0.641
R^2- adjusted	0.652		0.656		0.632		0.623
F	41.577***		42.249***		38.216***		36.863***
SEE^d	18.101		18.206		19.122		19.276

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2006, 2007, 2008, and 2009, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G17. Factors affecting prostate cancer rates in individual countries of the world (dependent variable – prostate cancer rate ASR in the years 2010; 2011; 2012 and 2013 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 58		Model 59		Model 60		Model 61
	2010		2011		2012		2013
	B^a	t^b	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	33.913	2.932***	30.647	2.579**	32.579	2.740***	29.168	2.557**
GDP per capita (US$)	0.001	4.249***	0.001	4.168***	0.001	4.605***	0.001	3.824***
GINI	0.610	2.721***	0.690	3.02***	0.816	3.615***	0.786	3.518***
Unhealthy food consumption (kcal/capita/day)	0.012	1.293*	0.010	1.16*	0.016	1.806*	0.010	1.265*
ALAN, nanowatts/cm2/sr	0.389	2.117**	0.417	2.247**	0.169	1.128*	0.430	2.665***
Solar Radiation^c, kWh/m^2	−6.674	−3.265***	−6.653	−3.247***	−7.844	−3.774***	−7.146	−3.571***
Regional dummies:
Asia	−21.442	−5.104***	−20.900	−4.939***	−19.319	−4.585***	−20.234	−5.024***
No. of obs.	131		131		131		131
R^2	0.653		0.647		0.639		0.659
R^2- adjusted	0.636		0.630		0.622		0.642
F	38.906***		37.894***		36.641***		39.91***
SEE^d	18.297		18.251		18.009		17.242

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2010, 2011, 2012, and 2013, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.

Table G18. Factors affecting prostate cancer rates in individual countries of the world (dependent variable – prostate cancer rate ASR in the years 2014; 2015 and 2016 explanatory 10-year latency period, unless stated otherwise).

Predictors	Model 62		Model 63		Model 64
	2014		2015		2016
	B^a	t^b	B^a	t^b	B^a	t^b
(Constant)	31.472	2.712***	31.856	2.699***	31.405	2.656**
GDP per capita (US$)	0.001	3.807***	0.001	3.732***	0.001	3.677***
GINI	0.755	3.308***	0.722	3.176***	0.773	3.35***
Unhealthy food consumption (kcal/capita/day)	0.013	1.615*	0.013	1.574*	0.013	1.648*
ALAN, nanowatts/cm2/sr	0.362	2.328**	0.381	2.32**	0.385	2.44**
Solar Radiation^c, kWh/m^2	−7.469	−3.738***	−7.131	−3.551***	−7.490	−3.684***
Regional dummies:
Asia	−20.304	−5.098***	−20.587	−5.184***	−20.723	−5.194***
No. of obs.	131		131		131
R^2	0.653		0.650		0.650
R^2- adjusted	0.637		0.633		0.633
F	38.966***		38.137***		38.41***
SEE^d	17.193		17.223		17.188

ALAN=artificial light-at-night; SEE=standard error of the estimate. ^aUnstandardized regression coefficient; ^bt-statistic; ^cSolar radiation variable is taken for years 2014, 2015, and 2016, that is, simultaneously, without a latency period; ^dSEE, standard error of the estimate; * Indicates a 0.1 two-tailed significance level; ** indicates a 0.05 significance level; *** indicates a 0.01 significance level.