Artificial Light at Night=
and
Obesity: Does the Spread of Wireless Information and Communication Technolo=
gy
Play a Role?
Nataliya A. Rybnikova1,*, Boris A. Portnov1
Received 01 December 2016, Accepted 21
December 2016
Abstract
Purpose: Artificial light at night (AL=
AN)
may influence body mass via different mechanisms, including the suppression=
of
the melatonin production by direct ALAN exposure, and by causing a shift in=
the
food intake, due to activities enabled by ALAN. In the present paper, we
attempt to separate these potential mechanisms by introducing measures of
Internet and cellular phones use that reflect the proliferation of informat=
ion
and communication technologies (ICT) in individual countries of the world. =
Methods: In addition to ICT measures, =
our
multivariate regression models link country-specific obesity rates with amb=
ient
ALAN levels, captured by satellite sensors, with per capita GDP, urbanizati=
on,
birth rate, food consumption and regional differences.
Results: As the analysis reveals, in
low-income countries, the number of cellular phone subscriptions emerged as=
a
statistically significant predictor of obesity rates (t>2.386; P<0.05=
),
while in the rest of the world countries, obesity was found positively rela=
ted
to ALAN (t>2.251; P<0.05), but not to Internet and cellular phone use
(P>0.1).
Conclusion: We explain these differenc=
es
by the fact that Internet and cellular phones are widely spread in low reso=
urce
countries, and their use may become an effective circadian disruptor, despi=
te
relatively low ambient ALAN levels, attributed to limited economic resources
available.
<=
span
lang=3DEN-US style=3D'font-size:10.0pt'>Keywords: Artific=
ial
light at night (ALAN), Internet usage, Cellular subscriptions, Obesity
prevalence rate
1.
Introduction
As
recent empirical studies show, ALAN may influence body mass via different
impact mechanisms, including the suppression of melatonin production, chang=
ing
the metabolic function [1, 2] and shifting the time of food intake by movin=
g it
from active to the rest phase [3, 4]. However, to the best of our knowledge,
none of these studies attempted to separate these potential impact mechanis=
ms,
an issue we attempt to address in this paper.
In
a recent study [5], the authors tested the extent to which artificial
light-at-night (ALAN) helps to explain the obesity prevalence rates observe=
d in
more than 80 countries worldwide and obtained from the World Health
Organization. As the study revealed, population-weighed ALAN levels, obtain=
ed
from the U.S. Defense Meteorological Satellite Program, were positively
associated with female (t=3D2.739, P<0.01) and male (t=3D2.658, P<0.0=
1)
obesity rates, reported for individual countries of the world [6].
However,
satellite images, used in [5], do not represent the whole picture. The matt=
er
is that, due to a rapid proliferation of cellular telephones and Internet
technology, people worldwide are currently exposed to ALAN, not captured by=
any
satellite. Although ipads, cellular phones, PC, gaming devices do not emit =
much
light, they are very effective in melatonin suppression, not only because t=
hey
are kept close to the eyes, but also due to the ¡°blue¡± short wavelength l=
ight
they emit [7]. In addition, cellular phones and Internet, which are widely
spread even in least developed countries and often used after the natural d=
usk
and at night, may serve as distinctive measures of circadian disruption on
their own right, especially in places, where people have no opportunities f=
or
nighttime outdoor activities enabled by ALAN in more developed countries.
2.
Methods
In
this reanalysis of the ALAN-obesity rate association, we retain the general
research setting and modelling approach used in [5], but add the computers =
and
cellular phones usage to the list of explanatory variables for country-spec=
ific
obesity rates, considering that the use of such devices can be a proxy for
indoor ALAN exposure not captured by satellite sensors.
Relevant
data on the percent of Internet users and the number of cellular subscripti=
ons
per 100 persons were obtained from the International Telecommunication Union
[8] and shown in Fig. 1.
(A)
(B)
Fig.
1. Worldwide rates of: (A) Internet usage (% of total population) and (B)
cellular phone subscriptions (number per 100 persons)
3. Results
The results of analysis are reported in Table 1,
both for females and males (Table 1)[2].
Table 1. Factors affecting
female & male obesity prevalence rates in individual countries of the w=
orld
(dependent variables – female & male obesity prevalence rates, % =
of adult
population; method – OLS regression)
|
Predictors |
Model 1 |
Model 2 |
Model 3 |
Model 4 |
|||||||
|
Ba |
(t)b |
Ba |
(t)b |
Ba |
(t)b |
Ba |
(t)b |
||||
|
|
2.796 |
(4.761)*** |
2.154 |
(3.252)*** |
2.492 |
(2.692)*** |
1.762 |
(1.583) |
|||
|
|
-3.24E-05 |
(-4.550)*** |
-3.15E-05 |
(-4.478)*** |
-8.49E-06 |
(-0.920) |
-8.62E-06 |
(-0.937) |
|||
|
|
0.022 |
(3.214)*** |
0.019 |
(2.687)*** |
0.019 |
(1.811)* |
0.017 |
(1.699)* |
|||
|
|
-0.037 |
(-3.087)*** |
-0.026 |
(-1.965)* |
-0.047 |
(-2.089)** |
-0.035 |
(-1.431) |
|||
|
|
0.002 |
(2.739)*** |
0.002 |
(2.426)** |
0.003 |
(2.658)*** |
0.003 |
(2.331)** |
|||
|
|
0.001 |
(2.393)** |
0.001 |
(2.480)** |
0.001 |
(2.077)** |
0.001 |
(2.226)** |
|||
|
|
-0.002 |
(-3.739)*** |
-0.002 |
(-3.681)*** |
-0.003 |
(-2.772)*** |
-0.003 |
(-2.781)*** |
|||
|
|
2.990 |
(5.939)*** |
3.094 |
(6.189)*** |
3.148 |
(4.928)*** |
3.241 |
(5.048)*** |
|||
|
|
-2.530 |
(-7.419)*** |
-2.409 |
(-7.043)*** |
-3.025 |
(-6.289)*** |
-2.883 |
(-5.825)*** |
|||
|
|
– |
– |
0.006 |
(2.005)** |
– |
– |
0.005 |
(1.172) |
|||
|
|
129 |
129 |
84 |
84 |
|||||||
|
|
0.727 |
0.736 |
0.679 |
0.685 |
|||||||
|
|
0.709 |
0.716 |
0.644 |
0.646 |
|||||||
|
|
(39.948)*** |
(36.850)*** |
(19.802)*** |
(17.842)*** |
|||||||
|
|
0.923 |
0.912 |
1.097 |
1.094 |
|||||||
Notes: BMI¡Ã30 cut-off point for obesity rate (W=
HO,
2014). *Indicates a 0.1 two-tailed significance level; **Indicates a 0.05
significance level; ***Indicates a 0.01 significance level. a Unstandardized
regression coefficient; b t-statistic; c Standard error of the estimate.
Model 1: Basic female model reported in Rybnikov=
a et
al (2016).
Model 2: Female model with the number of cellular
phone subscriptions added as an additional predictor.
Model 3: Basic male model reported in Rybnikova =
et
al (2016).
Model 4: Male model with the number of cellular
phone subscriptions added as an additional predictor.
We decided to look into the differences between =
male
and female models further and reran the analysis separately for countries w=
ith
different income levels. Our motivation was as follows: by contrast to fema=
le
obesity rates, which are available for a relatively large pool of 129
countries, data on male obesity rates are currently available only for
countries with relatively high income levels (84 countries).
For the country breakdown we used the World Bank
income classification, which denotes countries with annual GDP per capita of
less than US $1000 as low-income (LI) countries [9]. Concurrently, countries
with more than $1000 per capita per annum we refer to as above-low-income (=
ALI)
countries. The analysis is performed for female obesity rates only, because=
, as
we previously mentioned, data on obesity rates in males are unavailable for=
LI
countries. The results of re-analysis are reported in Table 2.
Table 2: Factors affecting
female obesity prevalence rates in LI and ALI countries (dependent variable=
–
female obesity rate, % of adult female population; method – OLS regre=
ssion)
|
Predictors |
Model 5 |
Model 6 |
|||
|
Ba |
(t)b |
Ba |
(t)b |
||
|
(Constant) |
2.943 |
(4.039)*** |
-0.134 |
(-0.087) |
|
|
GDP per capita (current US$) |
-2.59E-05 |
(-3.534)*** |
0.000 |
(-0.346) |
|
|
Urban population (% of total) |
0.013 |
(1.657) |
0.021 |
(1.025) |
|
|
Birth rate (per 1000 persons) |
-0.013 |
(-0.876) |
0.004 |
(0.143) |
|
|
ALAN (dimensionless units) |
0.002 |
(2.251)** |
0.006 |
(0.796) |
|
|
Oils, fats, sugars (calories per day) |
0.001 |
(2.235)** |
0.001 |
(0.437) |
|
|
Roots, tubers (calories per day) |
-0.002 |
(-3.150)*** |
-0.001 |
(-1.676) |
|
|
SIDS in the Pacific |
2.663 |
(5.210)*** |
– |
– |
|
|
Asia |
-2.728 |
(-7.175)*** |
-1.409 |
(-1.924)* |
|
|
Cellular Subscriptions, per 100 persons |
0.002 |
(0.481) |
0.026 |
(2.386)** |
|
|
N of obs. |
103 |
26 |
|||
|
R2 |
0.617 |
0.628 |
|||
|
R2-adjusted |
0.580 |
0.453 |
|||
|
F |
(16.635)*** |
(3.593)** |
|||
|
SEEc |
0.902 |
0.682 |
|||
Notes: BMI¡Ã30 cut-off point for obesity rate (W=
HO,
2014). *Indicates a 0.1 two-tailed significance level; **Indicates a 0.05
significance level; ***Indicates a 0.01 significance level. a Unstandardize=
d regression
coefficient; b t-statistic; c Standard error of the estimate.
Model 5: ALI countries.
Model 6: LI countries.
As Table 2 shows, ALAN emerged as a significant
variable in the model estimated for female obesity rates in ALI countries
(t=3D2.251; P<0.05; see Model 5; Table 2), while the cellular phone
subscription variable is insignificant in this model (t=3D0.481; P>0.1; =
Model
5; Table 2). By contrast, the model estimated for LI countries, reports no
significant association between female obesity rates and ALAN (t=3D0.796;
P>0.1; Model 6; Table 2), while there is a significant and positive
association between cellular phone subscriptions and female obesity rates
(t=3D2.386; P<0.05 in Model 6; Table 2).
4. Discussion and Conclusion<= o:p>
The present analysis is a continuation of the
previous study [5], which reported a positive association between outdoor A=
LAN
and adult overweight and obesity prevalence rates in world countries. In th=
is
study, we retain the modelling approach used in [5], while adding the Inter=
net
usage and cellular subscriptions to the list of explanatory variables in
multivariate regression models for country-specific obesity rates, consider=
ing
these variables as proxies for indoor ALAN exposure. We estimated ALAN-obes=
ity
associations separately for men and women, living in low-income (LI) and
above-low-income (ALI) countries.
As the analysis reveals, in low-income countries,
the number of cellular phone subscriptions is as a statistically significant
predictor of obesity rates (t>2.386; P<0.05), while in the rest of the
world countries, obesity was found positively related to nighttime outdoor
light (t>2.251; P<0.05), but not to Internet and cellular phone use
(P>0.1). We explain these differences by the fact that Internet and cell=
ular
phones are widely spread in low resource countries, and their use may becom=
e an
effective circadian disruptor, despite relatively low ambient ALAN levels,
attributed to limited economic resources available.
The use of cellular phones and Internet has beco=
me
increasingly common, even in less developed countries. However, these devic=
es
are an unqualified ALAN source by satellite sensors, and any analysis which
does not take the use of these devices into account, may lead to biased
results. The matter is that these devices emit short wavelength light,
nighttime exposure to which is known to lead to an effective suppression of=
the
rhythmicity of melatonin production and may become an effective circadian
disruptor on its own, potentially influencing body mass by shifting the tim=
e of
food intake. Moreover, low quality cellular phones and tablets are often so=
ld
in underdeveloped countries, where mass consumers cannot afford more expens=
ive
devices. The screens of such low quality devices are likely to have less
protection against shortwave (blue) light, compared to higher quality devic=
es
sold in more developed countries. The use of such low quality devices may
result in a stronger effect on melatonin suppression and circadian disrupti=
on.
A stronger effect of cellular phone use on women
than on men, which our study also revealed (see Models 2 and 4 in Table 1),=
may
be due to more extensive cellular phone use by females or, alternatively, to
higher susceptibility of females to harmful effects associated with frequent
cellular phone use. However, this hypothesis needs to be confirmed by furth=
er
analysis, which should be performed separately for low-income and
above-low-income countries, and, separately, for men and women. However,
possibilities for such an analysis are currently limited due to limited data
availability.
Several limitations of the study should be menti=
oned.
First and foremost, the present study is a geographical analysis and its
results cannot be considered as a proof of causality. In addition, as we
previously mentioned, data on obesity prevalence rates are available for wo=
men
and not available for men in low income countries. For that reason, our
findings need to be confirmed by further studies, as data on obesity preval=
ence
and indoor ALAN proxies become available for more countries and regions.
5. Acknowledgement
The
authors are grateful to the Israel Ministry of Science, Technology and Space
for supporting this study.
References
[1] Kennaway, D. J., Voults=
ios,
A., Varcoe, T. J., & Moyer, R. W. (2002). Melatonin in mice: rhythms,
response to light, adrenergic stimulation, and metabolism. American Journal=
of
Physiology-Regulatory, Integrative and Comparative Physiology, 282(2),
R358-R365.
[2] Korkmaz, A., Topal, T.,
Tan, D. X., & Reiter, R. J. (2009). Role of melatonin in metabolic
regulation. Reviews in Endocrine and Metabolic Disorders, 10(4), 261-270.
[3] Arble, D. M., Bass, J.,
Laposky, A. D., Vitaterna, M. H., & Turek, F. W. (2009). Circadian timi=
ng
of food intake contributes to weight gain. Obesity, 17(11), 2100-2102.
[4] Fonken, L. K., Workman,=
J.
L., Walton, J. C., Weil, Z. M., Morris, J. S., Haim, A., & Nelson, R. J.
(2010). Light at night increases body mass by shifting the time of food int=
ake.
Proceedings of the National Academy of Sciences, 107(43), 18664-18669.
[5] Rybnikova, N. A., Haim,=
A.,
& Portnov, B. A. (2016). Does artificial light-at-night exposure contri=
bute
to the worldwide obesity pandemic?. International Journal of Obesity.
[6] World Health Organizati=
on
(WHO) Obesity and overweight: Available at: http://www.who.int/mediacentre/=
factsheets/
fs311/en/. Accessed 22 Feb 2015.
[7] Haim, A., & Portnov=
, B.
A. (2013). Light pollution as a new risk factor for human breast and prosta=
te
cancers (p. 168). Dordrecht: Springer.
[8] International Telecommu=
nication
Union (ITU) Statistics. Available at: http://www.itu.int/en/ITU-D/ Statisti=
cs/ Pages/stat/default.aspx.
Accessed 11 Feb 2016.
[9] The World Bank (WB) GDP=
per
capita. Available at: http://data.worldbank.org/indicator/NY.GDP.PCAP. CD.
Accessed 24 May 2014.
N.A. Rybnikova
B.A. Portnov / International Journal of Sustainable Lighting 35 (2016) 16-2=
0
N.A. Rybnikova B.A. Portnov / International Journal of
Sustainable Lighting 35 (2016) 16-20 http://doi.org/1=
0.22644/ijsl.2016.35.1.016