Effects of Everyday Radiofrequency Electromagnetic-Field Exposure on Sleep Quality: A Cross-Sectional Study

Abstract

The aim of this cross-sectional study was to investigate the association between exposure to various sources of radiofrequency electromagnetic fields (RF EMFs) in the everyday environment and sleep quality, which is a common public health concern. We assessed self-reported sleep disturbances and daytime sleepiness in a random population sample of 1,375 inhabitants from the area of Basel, Switzerland. Exposure to environmental far-field RF EMFs was predicted for each individual using a prediction model that had been developed and validated previously. Self-reported cordless and mobile phone use as well as objective mobile phone operator data for the previous 6 months were also considered in the analyses. In multivariable regression models, adjusted for relevant confounders, no associations between environmental far-field RF EMF exposure and sleep disturbances or excessive daytime sleepiness were observed. The 10% most exposed participants had an estimated risk for sleep disturbances of 1.11 (95% CI: 0.50 to 2.44) and for excessive daytime sleepiness of 0.58 (95% CI: 0.31 to 1.05). Neither mobile phone use nor cordless phone use was associated with decreased sleep quality. The results of this large cross-sectional study did not indicate an impairment of subjective sleep quality due to exposure from various sources of RF EMFs in everyday life.

Full text

Effects of Everyday Radiofrequency Electromagnetic-Field Exposure on
Sleep Quality: A Cross-Sectional Study
Evelyn Mohler,
a,b
Patrizia Frei,
a,b
Charlotte Braun-Fahrla¨nder,
a,b
Ju¨rg Fro¨hlich,
c
Georg Neubauer,
d
Martin Ro¨o¨sli
a,b,1
and the Qualifex Team
a
Swiss Tropical and Public Health Institute, Basel, Switzerland;
b
University of Basel, Basel, Switzerland;
c
Laboratory for Electromagnetic Fields
and Microwave Electronics, ETH Zurich, Switzerland; and
d
Safety and Security Department, Austrian Institute of Technology, Austria
Mohler, E., Frei, P., Braun-Fahrla¨ nder, C., Fro¨hlich, J.,
Neubauer, G. and Ro¨o¨ sli, M. Effects of Everyday Radiofre-
quency Electromagnetic-Field Exposure on Sleep Quality:
A Cross-Sectional Study.
The aim of this cross-sectional study was to investigate the
association between exposure to various sources of radiofre-
quency electromagnetic fields (RF EMFs) in the everyday
environment and sleep quality, which is a common public health
concern. We assessed self-reported sleep disturbances and
daytime sleepiness in a random population sample of 1,375
inhabitants from the area of Basel, Switzerland. Exposure to
environmental far-field RF EMFs was predicted for each
individual using a prediction model that had been developed
and validated previously. Self-reported cordless and mobile
phone use as well as objective mobile phone operator data for the
previous 6 months were also considered in the analyses. In
multivariable regression models, adjusted for relevant confound-
ers, no associations between environmental far-field RF EMF
exposure and sleep disturbances or excessive daytime sleepiness
were observed. The 10%most exposed participants had an
estimated risk for sleep disturbances of 1.11 (95%CI: 0.50 to
2.44) and for excessive daytime sleepiness of 0.58 (95%CI: 0.31
to 1.05). Neither mobile phone use nor cordless phone use was
associated with decreased sleep quality. The results of this large
cross-sectional study did not indicate an impairment of
subjective sleep quality due to exposure from various sources
of RF EMFs in everyday life g2010 by Radiation Research Society
INTRODUCTION
The possible effects of radiofrequency electromagnet-
ic-field (RF EMF) exposure on health-related quality of
life are of public health concern (1–3). The most often
reported complaints related to RF EMFs are impair-
ments of sleep quality (4, 5).
Several studies investigated the effect of short-term
RF EMF exposure on sleep measures in a laboratory
setting, applying real and sham exposure randomly
under well-controlled exposure conditions (6–8). Objec-
tive sleep measures derived from electroencephalogra-
phy (EEG) were used in these laboratory studies.
Overall, these studies showed no consistent association
between RF EMF exposure and objective sleep mea-
sures (i.e. sleep architecture), but small differences for
different frequency ranges in the EEG were observed
repeatedly after exposure to RF EMFs. The primary
aim of laboratory studies is to identify a possible
biological mechanism of the effect of RF EMF exposure
on sleep, if any exists. In general, laboratory studies are
conducted with a relatively small number of participants
and therefore have limited statistical power to investi-
gate subjective sleep quality. Moreover, the unfamiliar
environment of a sleep laboratory may prevent detection
of subtle effects of RF EMFs on sleep quality, as has
been reported by several individuals.
Epidemiological studies allow the examination of the
association between RF EMFs and subjective sleep
quality in a large population sample. The main challenge
is to perform an appropriate exposure assessment. Until
now, only a few studies were conducted. In early studies,
associations between RF EMF exposure and subjective
well-being or sleep quality were observed (9, 10).
However, in these studies, simple exposure proxies like
self-reported distance to mobile phone base stations
were used, which have been demonstrated to be
inadequate (11, 12). Information bias was also of
concern in these studies and might have influenced the
results. Additionally, selection bias might affect results
in such cross-sectional studies if participation is related
to both health and exposure status (13, 14). More recent
studies on RF EMF exposure and sleep quality used
spot measurements in the bedroom for exposure
classification (15, 16). No differences in sleep quality
(Pittsburgh Sleep Quality Index) or in other health
outcomes (headache, SF-36 and health complaint list)
were observed between individuals with high and low
exposures. Although more sophisticated exposure as-
sessment methods were used in these studies, it still is not
1
Address for correspondence: Swiss Tropical and Public Health
Institute, P.O. Box, 4002 Basel, Switzerland; e-mail: martin.roosli@
unibas.ch.
RADIATION RESEARCH
0033-7587/10 $15.00
g2010 by Radiation Research Society.
All rights of reproduction in any form reserved.
DOI: 10.1667/RR2153.1
0

clear how well such spot measurements represent long-
term exposure to various sources of RF EMFs in our
everyday environment. For these reasons, in our study,
we used personal RF EMF exposure measurements and
modeling of fixed-site transmitters (e.g. mobile phone
base stations and broadcast transmitter) to develop a
method to assess individual exposure (17).
Due to the unknown biological mechanism, it is
unclear which aspect of exposure is relevant for sleep
disturbances, if there are any. It is conceivable that
exposure at the head, caused mainly by mobile and
cordless phones, is most relevant (close to body sources).
Alternatively, environmental sources like exposure from
mobile phone base stations or broadcast transmitter,
which in general cause lower but continuous whole-body
exposures, might play a role (far-field environmental RF
EMF exposure). RF EMF exposure might cause
symptoms immediately, or the accumulated radiation
might be more important. Additionally, psychological
aspects appear to be important. Previous studies showed
that subjective well-being and sleep quality can be
impaired in people from concern or expectations if they
think they are highly exposed to various sources of RF
EMFs (3) (also called a nocebo effect).
The primary aim of this cross-sectional study was to
evaluate whether environmental RF EMF exposure is
associated with self-reported sleep quality. We also
evaluated whether sleep quality is affected by other RF
EMF exposure surrogates such as night exposure or use
of mobile or cordless phones.
METHODS
In May 2008, 4000 questionnaires entitled ‘‘environment and
health’’ were sent out to people aged between 30 to 60 years who were
randomly selected from the population registries of the city of Basel
(Switzerland) and from five communities in the surroundings of Basel.
To minimize noneligibility due to language difficulties, only Swiss
residents or people living in Switzerland for at least 5 years were
selected. A reminder letter was sent out 3 weeks after the first
invitation for participation. Nonresponders were contacted by phone
6 to 10 weeks after the first questionnaires were sent out, and they
were asked a few key questions. Ethical approval for the study was
received from the Ethical Commission of Basel on March 19, 2007
(EK: 38/07).
Written Questionnaire
The questionnaire addressed three issues: (1) sleep quality and
general health status; (2) exposure-relevant characteristics and
behaviors (17) such as owning a mobile phone, a cordless phone,
and/or a wireless LAN and duration of cordless phone use and mobile
phone use; and (3) socio-demographic factors such as age, gender,
education, marital status and additional confounders like body mass
index (BMI), physical activity, smoking behaviors and alcohol
consumption.
Excessive Daytime Sleepiness and Self-Reported Sleep Disturbances
To assess subjective sleep quality, we used two sleep outcomes.
Daytime sleepiness was determined by the Epworth Sleepiness Scale
(ESS), which assigns values ranging from 0 (no daytime sleepiness) to
21 (very excessive daytime sleepiness) (18). We calculated the ESS
scores and created a new binary variable according to a previous
study on insomnia indicating excessive daytime sleepiness (ESS score
over 10) (19).
General subjective sleep quality was assessed by using four
standardized questions from the Swiss Health Survey 2007 (20).
The four questions on subjective sleep quality in the Swiss Health
Survey asked about the frequency of difficulty in falling asleep, fitful
sleep, waking phases during night, and waking too early in the
morning using a four-point Likert scale with categories ‘‘never’’,
‘‘rare’’, ‘‘sometimes’’ and ‘‘most of the time’’. Out of these four
questions, a binary sleep quality score (SQS) was calculated by adding
up all items (ranging from 0 to 12) and defining a score of eight as
having sleep disturbances (20).
Exposure Assessment
Our main hypothesis was that environmental whole-body exposure
in everyday life may affect sleep quality. We developed a model for
predicting personal exposure to environmental RF EMFs on the
power flux density scale in mw/m
2
(17) in which we measured personal
RF EMF exposure of 166 volunteers from our study area by means of
a portable EME Spy 120 exposure meter. Volunteers carried the
exposimeter and filled in an activity diary for 1 week (21). The
exposimeter measured 12 different frequency bands of RF EMFs
ranging from FM radio (frequency modulation; 88–108 MHz), TV
(television, 174–223 MHz and 470–830 MHz), Tetrapol (terrestrial
trunked radio police; 380–400 MHz), uplink in three frequency ranges
(communication from mobile phone handset to base station; 880–915,
1710–1785, 1920–1980 MHz), downlink in three frequency ranges
(communication from mobile phone base station to handset; 925–960,
1805–1880, 2110–2170 MHz), DECT (digital enhanced cordless
telecommunications; 1880–1900 MHz), and W-LAN (wireless local
area network; 2400–2500 MHz). In addition, we developed a three-
dimensional geospatial propagation model in which the average RF
EMF from fixed-site transmitters (e.g., mobile phone base stations
and broadcast transmitters) was modeled for the study region (in- and
outside of buildings) (22, 23). Based on this geospatial propagation
model and on data from the exposimeter measurements, the relevance
of potential predictors on exposure was examined in multivariable
non-linear regression models. The following exposure-relevant factors
were identified and included in the prediction model for environ-
mental exposure in everyday life (17): owning a mobile phone, owning
a wireless LAN at home, having the DECT base station in the
bedroom, having a cordless phone at the place where one spends the
most of their time during the day, house characteristics (window
frame and type of house wall), hours per week in public transport and
cars, percentage full-time equivalent spent at an external workplace,
and exposure from fixed-site transmitters at home computed by the
geospatial propagation model (22, 23).
To estimate exposure during the night, a separate night prediction
model was developed. Ownership of a cordless phone base station in
the bedroom, wireless LAN in the bedroom, house characteristics
(type of house wall and window frame), and the modeled value of
fixed-site transmitters were included in this specific prediction model.
We used the above-mentioned geospatial propagation model for
modeling exposure from fixed-site transmitters at home (22) in mW/
m
2
as well as in percentage of the ICNIRP (International Commission
on Non-Ionizing Radiation Protection) (24) reference level according
to method of Thomas et al. (28).
Finally, with respect to local exposure to the head, we used self-
reported use of mobile and cordless phones per week as reported in
the written questionnaire. Informed consent was also sought from
participants to obtain operator data for their mobile phone use for
the last 6 months from the three Swiss mobile phone network
operators.
0MOHLER ET AL.

Sensitivity Analysis
To evaluate a nocebo effect and information bias (which is also of
concern in this area of research), we asked participants about their
subjective exposure. They had to estimate their exposure compared to
the Swiss population and to indicate whether they felt they were
equally, less or more exposed in comparison to the average of the
Swiss population. Geo-coded data were available for all study
participants. This allowed us to calculate the distance from their
residence to the next mobile phone base station as an additional
exposure surrogate.
Nonresponder Analyses
To evaluate the extent of potential selection bias in our study,
nonresponder interviews were conducted to gather information on
general health status, socio-demographic factors and exposure-
relevant behaviors and factors. One month after the reminder letter
was sent out, we tried to contact all nonresponders. Information on
age, gender and geo-coded addresses was available for all 4000
persons.
We calculated ‘‘selection bias factors’’ for different exposure
proxies (i.e., owning a mobile phone, a cordless phone and/or a W-
LAN and distance to the next mobile phone base station) using the
Greenland method (25) as was done by Vrijheid et al. (26). For these
calculations we assumed that data from nonresponder phone
interviews are representative for all nonresponders. Dividing the
observed odds ratio by the bias factor yields the correct unbiased
association between exposure and outcome. A bias factor of 1.0
indicates that there is no bias.
Statistical Analyses
For binary outcomes (ESS score and SQS), logistic regression
models with three groups of exposure levels for all exposure proxies
(,50th percentile, 50th to 90th percentile, .90th percentile) were
performed. Mean average RF EMF exposures were calculated in
mW/m
2
and converted to V/m. In addition, linear regression models
were computed using the continuous score of both sleep scales.
Separate analyses were done for each of the four questions of the
Swiss Health Survey.
The models were adjusted for age, sex, body mass index (BMI),
stress perception, physical activity, smoking habits, alcohol con-
sumption, self-reported disturbance due to noise, living in urban or
suburban areas, belief in health effects due to RF EMF exposure,
education and marital status. Use of mobile and cordless phones was
included in all models as an independent exposure measure. Missing
values in the confounder variables were replaced with values of either
the most common category (categorical variables) or with the mean
value (linear variables) to ensure that all analyses were performed
with an identical data set for the ESS and the SQS, respectively. Most
missing values in confounder variables were observed in self-reported
disturbance of noise [33 missing out of 1212 observations (2.7%)].
Stratified analyses and testing for interaction were done for people
reporting as electrohypersensitive (EHS). We defined EHS individuals
as those reporting as ‘‘electrohypersensitive’’ or those reporting
adverse effects due to RF EMFs.
All statistical analyses were carried out using STATA 10.1
(StataCorp, College Station, TX).
RESULTS
Study Participants
Of the 4000 persons participating in the study, 237
were excluded due to noneligibility because of severe
disabilities (n527), death (n51), incorrect addresses (n
536), absence during study time (n573), or language
problems (n5100). A total of 1375 people completed
the questionnaire. Detailed information on the response
rate is illustrated in Fig. 1. Users of sleeping pills (n5
81) as well as night shift workers (n582) were excluded
from all the analyses. The final analyses thus included
1212 participants. Due to missing values in exposure
variables (mobile phone and cordless phone use) and in
sleep quality scores (ESS and SQS), 1129 study
participants remained for the analyses of excessive
daytime sleepiness and 1163 study participants remained
for the analyses of self-reported sleep disturbances.
Characteristics of all study participants are listed in
Table 1. The mean age (standard deviation) of study
participants was 46 (9) years, and 39%of all responders
lived in the city of Basel. There were more female (58%)
than male participants. Ninety percent reported that
they had a good or very good health status, which was
comparable to the general Swiss population (87%).
2
The
majority was married (60%) and of normal weight (BMI
,25) (62%).
Seventy-eight percent of the study participants re-
ported that they believed that there are people who
develop adverse health effects due to RF EMF exposure,
18.2%assigned their own adverse health effects as
being due to RE EMF exposure, and 8.1%reported that
they were ‘‘electrohypersensitive’’. Due to overlapping,
20.9%of our study population was electrohypersensitive
according to our definition.
2
National Statistical Institute (Switzerland) 2007; http://www.bfs.
admin.ch/bfs/portal/de/index/themen/14/02/01/key/01.html.
FIG. 1. Schematic illustration of the study design and response
rate.
RF EMF EXPOSURE AND SLEEP QUALITY 0

Level of Exposure
The predicted everyday life mean and median
exposure was 0.18 V/m for all the included study
participants. The cut-off point for 90th percentile was
0.21 V/m. The maximum predicted value was 0.33 V/m.
The mean predicted exposure during the night was
0.06 V/m (median: 0.02 V/m, cut-off 90th percentile:
0.09 V/m, maximum: 0.33 V/m), and the mean exposure
through fixed-site transmitters (geospatial propagation
model) was 0.08 V/m (median: 0.04 V/m, cut-off 90th
percentile: 0.12 V/m, maximum: 0.62 V/m). The mean
level of exposure from fixed-site transmitters was 0.15%
of the ICNIRP reference level. On average, study
participants reported using their mobile phones
62.8 min per week and their cordless phones 75.1 min
per week. Informed consent for objective data on mobile
phone use from the network operators was obtained
from 470 study participants. Those who gave informed
consent reported that they used their mobile phone
46.5 min per week, while the operator data showed a
mobile phone use of 28.8 min per week (27). The
Spearman rank correlation was 0.76 (95%CI: 0.71–0.83)
for self-reported mobile phone use and the operator
data.
The majority (64%) of the participants estimated that
their exposure was similar to the average for the Swiss
population, while 29%believed they were less exposed
and 7%believed they were more exposed.
Excessive Daytime Sleepiness (ESS score)
The prevalence of excessive daytime sleepiness (ESS
score .10) was 29.5%. the results of the logistic
regression models for crude and adjusted odds ratios
(OR) are presented in Table 2. No statistically signifi-
cant association between excessive daytime sleepiness
and various exposure surrogates was observed. The
analysis showed a tendency toward excessive daytime
sleepiness for the highest-exposed group through fixed-
site transmitters, although it was not statistically
significant. This finding was confirmed when exposure
TABLE 1
Characteristics and Results of Statistical Comparison of all Study Participants (including nonresponders)
Participants
(n51212)
a
Percent
Nonresponders
(n52388) Percent P value
Age (years) 0.05
30–40 319 26 719 30
41–50 421 35 829 35
51–60 472 39 840 35
Sex ,0.05
Female 706 58 1190 50
Male 506 42 1198 50
Distance to the next mobile phone base
station (percentage closer than 50 m) 45 4165 7,0.05
Health status
b,c
,0.05
Very good 445 37 215 34
Good 636 53 302 48
Half-half 107 986 14
Bad 12 118 3
Very bad 3 081
Educational level
b,c
0.171
None 79 756 9
Apprenticeship 591 49 320 51
Higher education/University 542 45 255 40
Owning a mobile phone
b,c
,0.05
Yes 1049 87 572 90
No 163 13 60 10
Owning a cordless phone
b,c
0.176
Yes 994 82 537 85
No 213 18 96 15
Owning wireless LAN
b,c
0.931
Yes 492 41 259 41
No 709 59 370 59
a
After exclusion of nightshift workers (n582) and users of sleeping drugs (n581).
b
Nonresponder data only for a subsample of 634 nonresponders who answered a short nonresponder interview by phone (numbers in
nonresponder analyses can vary due to missing data).
c
Data may not sum up to 100%due to missing data.
0MOHLER ET AL.

was calculated as a percentage of the ICNIRP reference
level (adjusted OR for the 90th percentile: 1.62; 95%CI:
0.99–2.64). This finding was confirmed when exposure
was calculated as a percentage of the ICNIRP reference
level (adjusted OR for the 90th percentile: 1.62; 95%CI:
0.99–2.64). Similar results were found for linear regres-
sion models (data not shown).
Based on interaction tests, we found no indication
that RF EMF exposure affects EHS individuals
differently than non-EHS individuals (P.0.05 for all
exposure surrogates).
Self-Reported Sleep Disturbances (SQS)
Problematic sleep disturbances were reported by 9.8%
of respondents. There was no evidence that having sleep
disturbances was influenced by everyday life exposure,
exposure through fixed-site transmitters or exposure
during the night (Table 3). The OR for the top decile of
exposed individuals according to the percentage of the
ICNIRP reference value was 0.95 (95%CI: 0.47 to 1.90).
Mobile phone and cordless phone use showed no
statistically significant effects on having sleep distur-
bances, but tendencies toward fewer sleep disturbances
with increased use of a mobile phone could be seen in
the logistic (Table 3) and linear regression models (data
not shown). However, analysis of a subsample with
objective mobile phone operator data did not show such
a tendency (Table 3).
The separate analyses of each item on the sleep quality
score (falling asleep, fitful sleep, waking phases during
night, waking up early in the morning) revealed no
exposure–response association (data not shown). Inter-
action tests and stratified analyses for EHS and non-
EHS individuals showed no difference between the two
subgroups.
Sensitivity Analysis
An association between self-reported sleep quality and
self-estimated exposure could indicate the presence of
information bias or a nocebo effect, or rather the
development of symptoms due to concerns. In our study,
we found some indications for the presence of a nocebo
effect (Table 4). People reporting to be less exposed to
mobile phone base stations in comparison to the average
population are less likely to suffer from excessive
daytime sleepiness (Table 4). Correspondingly, people
who lived closer than 50 m to the closest mobile phone
base station had a higher risk for excessive daytime
sleepiness, although it was not statistically significant.
Self-reported sleep disturbances were increased in people
claiming to be more exposed in comparison to the
average population. These trends were most pronounced
TABLE 2
Association between Excessive Daytime Sleepiness (Epworth Sleepiness Scale) and Different Exposure Surrogates
[odds ratios (OR) and 95%CI of the three exposure categories]
Excessive daytime sleepiness (n51129)
Exposure categories
,50th percentile 50th–90th percentile .90th percentile
No. of
cases
a
OR
No. of
cases
a
OR 95%CI
No. of
cases
a
OR 95%CI
Far-field exposure
Everyday life exposure
Crude 180 1.00 153 1.10 (0.84–1.43) 25 0.77 (0.47–1.24)
Adjusted
b
180 1.00 153 1.14 (0.83–1.57) 25 0.58 (0.31–1.05)
Exposure during night
Crude 174 1.00 149 1.14 (0.87–1.48) 35 1.06 (0.68–1.65)
Adjusted
b
174 1.00 149 1.05 (0.76–1.43) 35 1.21 (0.74–1.98)
Exposure through fixed-site transmitters
Crude 170 1.00 142 1.07 (0.82–1.40) 46 1.86 (1.21–2.85)
Adjusted
b
170 1.00 142 1.02 (0.74–1.39) 46 1.52 (0.93–2.50)
Close-to-body exposure
Mobile phone use (self-reported)
Crude 210 1.00 106 1.18 (0.89–1.57) 32 1.05 (0.69–1.64)
Adjusted
b
210 1.00 106 1.24 (0.91–1.70) 32 1.03 (0.62–1.69)
Mobile phone use (operator data)
c
Crude 65 1.00 152 1.11 (0.72–1.70) 14 1.26 (0.63–2.54)
Adjusted
b
65 1.00 152 1.30 (0.82–2.07) 14 0.91 (0.39–2.11)
Cordless phone use (self-reported)
Crude 178 1.00 165 1.27 (0.98–1.65) 13 1.44 (0.71–2.90)
Adjusted
b
178 1.00 165 1.30 (0.99–1.72) 13 1.65 (0.72–3.50)
a
Indicates number of people in the corresponding exposure group with an Epworth sleepiness score over 10.
b
Adjusted for age, body mass index, sex, physical activity, alcohol consumption, smoking habits, stress perception, urban/suburban, marital
status, educational level, noise perception, belief in health effects due to radiofrequency electromagnetic-field exposure.
c
For a subsample of 453 subjects who consented to obtain data from the operator.
RF EMF EXPOSURE AND SLEEP QUALITY 0

for self-estimated exposure to a mobile phone base
station. Subjective exposure was not correlated to
modeled mobile phone base station radiation (Spearman
correlation coefficient: –0.01) or total everyday life
exposure (Spearman correlation coefficient: 0.13).
Nonresponder Analysis
To evaluate a possible selection bias, we compared
responders of the questionnaire with nonresponders. The
nonresponder analyses, comparing all 1212 participants
included in our analyses with the 2388 nonresponders,
showed small differences between study participants and
nonresponders (Table 1). Nonresponders were generally
younger, and the participation rate for women was higher
than for men. The distance between the closest mobile
phone base station and place of residence was smaller for
the responders. Some of the nonresponder information
was available only for the nonresponders who partici-
pated in the telephone interviews (n5634): Participants
in these telephone interviews were more likely to be
an owner of a mobile phone (90%) than full study
participants (87%). Study participants who filled in the
questionnaire were somewhat healthier than nonrespond-
ers. No difference was observed in educational level in
owning a wireless LAN or cordless phone. The prevalence
of nonresponders (telephone interviews) who reported
that they were ‘‘electrohypersensitive’’ was 16%.Inthe
full study only 8%answered yes to the corresponding
question (P,0.0001).
In our selection bias factor, we found a bias factor of
0.79 for owning a mobile phone, 0.70 for owning a
cordless phone, 0.95 for owning a W-LAN, and 1.33 for
living within 50 m from a mobile phone base station.
Thus we expect that in our study the exposure–response
association for mobile and cordless phone use tends to be
biased downward whereas the exposure–response associ-
ation for fixed-site transmitter tends to be biased upward.
DISCUSSION
The aim of this study was to investigate the
association between various RF EMF exposure surro-
gates and self-reported sleep quality. Neither everyday-
life environmental RF EMF exposure nor exposure
during night through fixed-site transmitters or from
mobile and cordless phones was associated with
excessive daytime sleepiness or with having sleep
disturbances. We found some indication for nocebo
effects and information bias; this means that persons
who assumed that they were exposed more than the
average for the Swiss population reported that they
suffered often, although not statistically significantly so,
TABLE 3
Association between Self-Reported Sleep Disturbances (Sleep Quality Score) and Different Exposure Surrogates
[odds ratios (OR) and 95%CI of the three exposure categories]
Self-reported sleep disturbances (n51163)
Exposure categories
,50th percentile 50th–90th percentile .90th percentile
No. of
cases
a
OR
No. of
cases
a
OR 95%CI
No. of
cases
a
OR 95%CI
Far-field exposure
Everyday life exposure
Crude 98 1.00 68 0.91 (0.65–1.28) 14 0.87 (0.48–1.60)
Adjusted
b
98 1.00 68 1.11 (0.72–1.70) 14 1.11 (0.50–2.44)
Exposure during night
Crude 88 1.00 76 1.14 (0.81–1.50) 16 1.01 (0.57–1.80)
Adjusted
b
88 1.00 76 1.30 (0.85–1.98) 16 1.29 (0.66–2.53)
Exposure through fixed-site transmitters
Crude 88 1.00 77 1.15 (0.82–1.62) 15 0.94 (0.52–1.69)
Adjusted
b
88 1.00 77 1.16 (0.76–1.75) 15 1.09 (0.53–2.22)
Close-to-body exposure
Mobile phone use (self-reported)
Crude 124 1.00 41 0.71 (0.49–1.05) 13 0.71 (0.38–1.30)
Adjusted
b
124 1.00 41 0.67 (0.43–1.02) 13 0.64 (0.31–1.28)
Mobile phone use (operator data)
c
Crude 42 1.00 30 0.91 (0.54–1.51) 5 0.60 (0.22–1.62)
Adjusted
b
42 1.00 30 1.57 (0.89–2.78) 5 1.03 (0.32–3.30)
Cordless phone use (self-reported)
Crude 102 1.00 66 0.80 (0.57–1.12) 8 1.51 (0.67–3.40)
Adjusted
b
102 1.00 66 0.71 (0.49–1.03) 8 1.11 (0.44–2.78)
a
Indicates number of people in the corresponding exposure group with a sleep quality score over 8.
b
Adjusted for age, body mass index, sex, physical activity, alcohol consumption, smoking habits, stress perception, urban/suburban, marital
status, educational level, noise perception, belief in health effects due to radiofrequency electromagnetic-field exposure.
c
For a subsample of 453 subjects who consented to obtain data from the operator.
0MOHLER ET AL.

from sleep disturbances than participants who felt that
they were equally exposed as the average of the Swiss
population.
Strengths
Our study is based on a large sample size. To our
knowledge, our study used the most comprehensive
exposure assessment method to date by considering
exposure-relevant behavior and characteristics (predic-
tion model) as well as modeling RF EMFs from fixed-
site transmitters with a geospatial model (22). All
relevant exposure sources of everyday life were included
in the prediction model, and the feasibility and
reproducibility of this exposure assessment method
could be demonstrated (17). Using prediction models
for exposure assessment instead of conducting spot or
personal measurements, as has been done in other
studies (15, 16, 28), is time- and cost-saving for large
study populations and is expected to better represent all
sources of RF EMF exposure in everyday life.
We included several exposure surrogates in our study.
This allowed us to check for consistency and biological
plausibility, because no biological mechanism has been
established. In particular, we included both close-to-
body sources and far-field sources. In addition to self-
reported mobile phone use, we considered objective
operator data on mobile phone use for a subsample who
gave consent.
Limitations
The cross-sectional study design is one of our main
limitations, in particular with respect to EHS individu-
als. EHS individuals may tend to avoid known sources
of RF EMF exposure and are therefore expected to be
less exposed. If so, a cross-sectional study, where
outcome and exposure are measured at the same time,
could not capture an increased risk. It could even result
in observation of a protective effect from exposure
(although this was not the case in our study).
Conversely, people who did not attribute their own
symptoms to EMF exposure were not expected to avoid
exposure sources. Thus our cross-sectional study should
reveal an association in nonhypersensitive individuals, if
one is present, because RF EMF exposure is relatively
constant over a few months (21). This means that
present exposure is also representative of exposure a few
months before. In this regard, it is also relevant that self-
estimated exposure actually is not correlated to true
exposure. This indicates that most persons are not aware
of their most relevant exposure sources. Unawareness of
the exposure status implies that information bias is
unlikely in our study.
In our study, we did not take polysomnographic sleep
measures. We were mainly interested in self-reported
data on sleep quality and well-being, because a decrease
in self-perceived sleep quality due to RF EMF exposure
is the most often stated concern of the population (3, 5).
TABLE 4
Sensitivity Analysis to Evaluate the Possible Extent of Information Bias and Nocebo Effect: Association between
Sleep Quality (excessive daytime sleepiness and self-reported sleep disturbances) and Subjectiveoˆ Exposure
Excessive daytime sleepiness (n51129)
Subjective exposure categories
equal
a
lower higher
No. of cases
b
OR No. of cases
b
OR 95%CI No. of cases
b
OR 95%CI
Subjective exposure to all sources
Crude 239 1.00 96 0.80 (0.60–1.06) 23 0.87 (0.52–1.47)
Adjusted
c
239 1.00 96 0.78 (0.56–1.09) 23 0.84 (0.41–1.71)
Subjective exposure to mobile phone base station
Crude 243 1.00 85 0.71 (0.53–0.95) 30 0.98 (0.62–1.59)
Adjusted
c
243 1.00 85 0.67 (0.48–0.95) 30 0.83 (0.44–1.59)
Excessive daytime sleepiness (n51129)
.50 m #50 m
No. of cases
b
OR No. of cases
b
OR 95%CI
Distance to mobile phone base station (geo-coded)
Crude 340 1.00 - - - 18 1.90 (1.00–3.59)
Adjusted
c
340 1.00 - - - 18 2.06 (0.96–4.41)
a
Reference group includes also ‘‘don’t know’’ and missing values.
b
Indicates number of people in the corresponding exposure group with an Epworth sleepiness score over 10 or a sleep quality score over 8,
respectively.
c
Adjusted for age, body mass index, sex, physical activity, alcohol consumption, smoking habits, stress perception, urban/suburban, marital
status, educational level, noise perception, believe in health effects due to radiofrequency electromagnetic-field exposure.
RF EMF EXPOSURE AND SLEEP QUALITY 0

Subjectively perceived sleep quality is relevant to health
because it is an established factor that influences
personal well-being (29). Collecting more sophisticated
sleep measures using electroencephalography (EEG)
would require considerable additional effort in this
large study population, and such an unfamiliar mea-
surement procedure could mask subtle effects on self-
perceived sleep quality.
The participation rate for the full study (whole
questionnaire data) was 37%and was therefore lower
than we had expected and lower than in the study of
Ku¨hnlein et al. (30) and similar to that of Thomas et al.
(28). In recent years, a decreasing response rate has been a
commonly observed phenomenon in epidemiological
research (31). In our study people might have declined
because we asked them to give their informed consent to
provide objective data about their mobile phone use from
the mobile phone operator companies. People may have
felt that it was an invasion of their privacy. The main
concern in having a low participation rate is selection
bias. We made considerable effort to evaluate potential
bias from nonparticipation. To be able to assess the risk
of selection bias, we performed nonresponder interviews,
and data on age, gender and geo-codes were available for
all 4000 persons. We were concerned that people
attributing their sleep disturbances to mobile phone base
stations or to RF EMFs in general would be more
motivated to participate in our survey (32, 33). If these
people live closer to a mobile phone base station than the
average population, this could result in a bias, because
distance is one parameter of our exposure prediction
model. Interestingly, we found indications of the opposite
but yielding the same possible bias: Study participants
generally were healthier than nonresponders, and the
proportion of persons living close to a mobile phone base
station (,50 m) was smaller for participants than
nonparticipants. Thus our selection bias modeling yielded
a selection bias factor of 1.33 for living within 50 m of a
mobile phone base station. According to this selection
bias modeling our observed exposure–response associa-
tions for fixed site transmitter may be biased upward.
Conversely, our exposure–response associations for
mobile and cordless phone use may be biased downward.
Interpretation
The prevalence of excessive daytime sleepiness in our
study was similar to previous studies in which 32.4%
reported suffering from excessive daytime sleepiness
(34). Prevalence of sleep disturbances was in our study
even lower (9.8%) than observed in a study of a Swiss
working population (20), where 19%of a relatively
young Swiss working population suffered from disorders
of initiating and maintaining sleep.
We found no consistent evidence that RF EMF
exposure is associated with subjective sleep quality. Our
findings contradict early studies that used self-estimated
distance to mobile phone base stations as exposure
proxy (9, 10). This approach has been shown to be
inappropriate for exposure estimation (12, 14, 35).
Moreover, these early studies without objective exposure
measures are likely to be affected by nocebo effects since
we found some indication for such a bias in our study
when using self-estimated exposure measures that were
poorly correlated to true exposure levels. This was
particularly pronounced with respect to self-estimated
mobile phone base station radiation.
Our prediction models are developed and validated on
the power flux density scale (mW/m
2
). In our prediction
Self-reported sleep disturbances (n51163)
Subjective exposure categories
equal
a
lower higher
No. of cases
b
OR No. of cases
b
OR 95%CI No. of cases
b
OR 95%CI
116 1.00 49 0.92 (0.64–1.32) 15 1.23 (0.67–2.27)
116 1.00 49 1.05 (0.68–1.64) 15 1.47 (0.62–3.49)
109 1.00 47 1.08 (0.74–1.56) 24 1.99 (1.20–3.30)
109 1.00 47 1.16 (0.74–1.82) 24 1.61 (0.76–3.43)
Self-reported sleep disturbances (n51163)
.50 m #50 m
No. of cases
b
OR No. of cases
b
OR 95%CI
171 1.00 - - - 9 1.53 (0.72–3.25)
171 1.00 - - - 9 1.13 (0.41–3.04)
TABLE 4
Extended
0MOHLER ET AL.

model for everyday life exposure, we added up contribu-
tions from different sources on the power flux density
scale, based on the assumption that effects are not
dependent on frequency. It has also been speculated in
other studies that effects in the low-dose range maybe
dependent on frequency, and another study weighted the
exposure contributions according to the ICNIRP refer-
ence level (28). However, for exposure from a fixed-site
transmitter, where we were able to compare both scales,
we found a very high correlation (Spearman 50.96), and
the results of the epidemiological analyses were similar.
This suggests that choice of the exposure scale is not
crucial unless the effect is very frequency specific.
Our findings are in line with more recent cross-
sectional studies on subjective sleep quality that used
spot measurements in the bedroom for exposure
assessment (15, 16). This is probably an acceptable
exposure proxy for environmental RF EMF exposure
during the night, but it does not capture exposure during
the day or exposure to close-to-body sources that one
might be exposed to prior to sleep. However, such
exposure may be relevant: Several studies indicated that
exposure to a mobile phone prior to sleep affects EEG
during the night (7, 8, 36, 37).
In addition to the cross-sectional studies on self-
reported sleep quality and RF EMF exposure at home,
two studies investigated sleep behavior at home using an
experimental approach and recording polysomno-
graphic sleep measures. In a German study of 394
individuals living within 500 m of a mobile phone base
station, polysomnographic measures were recorded
during five consecutive nights. A transportable mobile
phone base station (GSM 900 and 1800) was installed
and randomly turned on and off.
3
Leitgeb et al. (38)
recruited 43 volunteers who reported to be EHS.
Polysomnography was applied during 9 nights (3 control
nights, 3 nights with sham shielding, and 3 nights with
true shielding). In both studies, polysomnographic
measures were not related to exposure.
We evaluated various exposure proxies. Except in a
subgroup analysis with non-sensitive individuals for
excessive daytime sleepiness and cordless phone use, no
statistically significant effects were found. Given the
numerous tests performed, one statistically significant
result can be expected by chance. Similarly, some of the
observed exposure–response tendencies such as the de-
creased occurrence of sleep disturbances for the moderate
user of cordless phones are probably due to chance or may
be affected by selection bias. If there were a true exposure–
response association in our large study population, we
would have expected to see a consistent pattern in terms of
outcome (i.e., similar effects for sleep quality or daytime
sleepiness) or in terms of exposure sources (i.e., similar
effects for close-to-body sources or for environmental
sources). Nevertheless, the cross-sectional design is a
limitation, particularly if one has the hypothesis that
people avoid exposure if they are suffering from sleep
disturbances. In our study we found no evidence for such a
behavior, nor have recent reviews suggested that the ability
to perceive RF EMF exposure actually exists (14, 39).
Overall, we found no indication that RF EMF
exposure in our daily life impairs subjective sleep
quality. In contrast to previous studies on that topic,
we considered all relevant RF EMF sources of the
everyday environment in our exposure assessment
through consideration of various proxies that are
relevant in everyday life.
ACKNOWLEDGMENTS
The study is funded by the Swiss National Science Foundation
(Grant 405740-113595). It is part of the National Research Program
57 ‘‘Non-Ionising Radiation - Health and Environment’’. Martin
Ro¨o¨ sli is supported by the Swiss School of Public Health z
(SSPHz). We thank Alfred Bu¨rgi for calculating the exposure to
fixed site transmitters for each study participant, Matthias Egger,
Niklas Joos and Axel Hettich (QUALIFEX team) for inspiring
discussions, Fabian Trees from the Swiss Federal Statistical Office for
providing the geographical coordinates of the study participants, and
the statistical department of Basel for providing the addresses of the
study participants. Many thanks go to all study participants who
volunteered for the study.
Received: January 2, 2010; accepted: April 13, 2010; published online:
REFERENCES
1. J. Schro¨ttner and N. Leitgeb, Sensitivity to electricity—temporal
changes in Austria. BMC Pub. Health 8, 310 (2008).
2. M. Ro¨o¨sli, M. Moser, Y. Baldinini, M. Meier and C. Braun-
Fahrla¨nder, Symptoms of ill health ascribed to electromagnetic
field exposure—a questionnaire survey. Int. J. Hyg. Environ.
Health 207, 141–150 (2004).
3. M. Blettner, B. Schlehofer, J. Breckenkamp, B. Kowall, S.
Schmiedel, U. Reis, P. Potthoff, J. Schuz and G. Berg-Beckhoff,
Mobile phone base stations and adverse health effects: phase 1 of
a population-based, cross-sectional study in Germany. Occup.
Environ. Med.66, 118–123 (2009).
4. G. Oftedal, J. Wilen, M. Sandstrom and K. H. Mild, Symptoms
experienced in connection with mobile phone use. Occup. Med.
(Lond.) 50, 237–245 (2000).
5. N. Schreier, A. Huss and M. Ro¨o¨sli, The prevalence of symptoms
attributed to electromagnetic field exposure: a cross-sectional
representative survey in Switzerland. Soz. Praventivmed.51, 202–
209 (2006).
6. H. Hinrichs, H. Heinze and M. Rotte, Human sleep under the
influence of a GSM 1800 electromagnetic far field. Somnologie 9,
185–191 (2005).
7. R. Huber, V. Treyer, A. A. Borbely, J. Schuderer, J. M.
Gottselig, H. P. Landolt, E. Werth, T. Berthold and P.
Achemann, Electromagnetic fields, such as those from mobile
phones, alter regional cerebral blood flow and sleep and waking
EEG. J. Sleep Res.11, 289–295 (2002).
3
H. Danker-Hopfe, H. Dorn, C. Sauter and M. Schubert,
Untersuchung der Schlafqualita¨t bei Anwohnern einer Basisstation.
Experimentelle Studie zur Objektivierung mo¨glicher psychologischer
und physiologischer Effekte unter ha¨uslichen Bedingungen. Final
report. Deutsches Mobilfunkforschungsprogramm, 2009.
RF EMF EXPOSURE AND SLEEP QUALITY 0

8. S. J. Regel, G. Tinguely, J. Schuderer, M. Adam, N. Kuster,
H. P. Landolt and P. Achermann, Pulsed radio-frequency
electromagnetic fields: dose-dependent effects on sleep, the
sleep EEG and cognitive performance. J. Sleep Res.16, 253–
258 (2007).
9. R. Santini, P. Santini, J. M. Danze, P. Le Ruz and M. Seigne,
[Investigation on the health of people living near mobile
telephone relay stations: I. Incidence according to distance and
sex]. Pathol. Biol. (Paris) 50, 369–373 (2002).
10. E. A. Navarro, J. Segura, M. Prortoles and C. G. P. deMateo,
The microwave syndrome: a preliminary study in Spain.
Electromagn. Biol. Med.22, 161–169 (2003).
11. C. Bornkessel, M. Schubert, M. Wuschek and P. Schmidt,
Determination of the general public exposure around GSM and
UMTS base stations. Radiat. Prot. Dosimetry 124, 40–47 (2007).
12. J. Schu¨z and S. Mann, A discussion of potential exposure metrics
for use in epidemiological studies on human exposure to
radiowaves from mobile phone base stations. J. Expo. Anal.
Environ. Epidemiol.10, 600–605 (2000).
13. R. de Marco, G. Verlato, E. Zanolin, M. Bugiani and J. W.
Drane, Nonresponse bias in EC Respiratory Health Survey in
Italy. Eur. Respir. J.7, 2139–2145 (1994).
14. M. Ro¨o¨sli, Radiofrequency electromagnetic field exposure and
non-specific symptoms of ill health: a systematic review. Environ.
Res.107, 277–287 (2008).
15. G. Berg-Beckhoff, M. Blettner, B. Kowall, J. Breckenkamp, B.
Schlehofer, S. Schmiedel, C. Bornkessel, U. Reis, P. Potthoff and
J. Schuz, Mobile phone base stations and adverse health effects:
phase 2 of a cross-sectional study with measured radio frequency
electromagnetic fields. Occup. Environ. Med.66, 124–130 (2009).
16. H. P. Hutter, H. Moshammer, P. Wallner and M. Kundi,
Subjective symptoms, sleeping problems, and cognitive
performance in subjects living near mobile phone base stations.
Occup. Environ. Med.63, 307–313 (2006).
17. P. Frei, E. Mohler, A. Bu¨rgi, J. Fro¨ hlich, G. Neubauer, C.
Braun-Fahrla¨nder and M. Ro¨o¨ sli, A prediction model for
personal radio frequency electromagnetic field exposure. Sci.
Total Environ.408, 102–108 (2009).
18. M. Johns and B. Hocking, Daytime sleepiness and sleep habits of
Australian workers. Sleep 20, 844–849 (1997).
19. K. L. Lichstein, H. H. Durrence, U. J. Bayen and B. W. Riedel,
Primary versus secondary insomnia in older adults: subjective
sleep and daytime functioning. Psychol. Aging 16, 264–271 (2001).
20. B. E. Schmitt, M. Gugger, K. Augustiny, C. Bassetti and B. P.
Radanov, [Prevalence of sleep disorders in an employed Swiss
population: results of a questionnaire survey]. Schweiz. Med.
Wochenschr.130, 772–778 (2000).
21. P. Frei, E. Mohler, G. Neubauer, G. Theis, A. Bu¨rgi, J. Fro¨hlich,
C. Braun-Fahrla¨nder, J. Bolte, M. Egger and M. Ro¨o¨ sli,
Temporal and spatial variability of personal exposure to radio
frequency electromagnetic fields. Environ. Res.109, 779–785
(2009).
22. A. Bu¨rgi, P. Frei, G. Theis, E. Mohler, C. Braun-Fahrla¨nder, J.
Fro¨hlich, G. Neubauer, M. Egger and M. Ro¨o¨ sli, A model for
radiofrequency electromagnetic field predictions at outdoor and
indoor locations in the context of epidemiological research.
Bioelectromagnetics 31, 226–236 (2009).
23. A. Bu¨rgi, G. Theis, A. Siegenthaler and M. Ro¨o¨ sli, Exposure
modeling of high-frequency electromagnetic fields. J. Expo. Sci.
Environ. Epidemiol.18, 183–191 (2008).
24. ICNIRP, Guidelines for limiting exposure to time-varying
electric, magnetic, and electromagnetic fields (up to 300 GHz).
Health Phys.74, 494–522 (1998).
25. S. Greenland and K. J. Rothman, Basic methods for sensitivity
analysis and external adjustment. In Modern Epidemiology, pp.
343–358. Williams & Wilkins, Philadelphia, 1998.
26. M. Vrijheid, L. Richardson, B. K. Armstrong, A. Auvinen, G.
Berg, M. Carroll, A. Chetrit, I. Deltour, M. Feychting and E.
Cardis, Quantifying the impact of selection bias caused by
nonparticipation in a case-control study of mobile phone use.
Ann. Epidemiol.19, 33–41 (2009).
27. E. Mohler, P. Frei, D. Aydin, A. Bu¨ rgi and M. Ro¨o¨sli,
Perso¨nliche Exposition durch hochfrequente elektromagnetische
Felder in der Region Basel (Schweiz): Ein U
¨berblick u¨ ber die
QUALIFEX-Studie. Umweltmed. Forsch. Prax.14, 329–338
(2009).
28. S. Thomas, A. Kuhnlein, S. Heinrich, G. Praml, D. Nowak, R.
von Kries and K. Radon, Personal exposure to mobile phone
frequencies and well-being in adults: a cross-sectional study based
on dosimetry. Bioelectromagnetics 29, 463–470 (2008).
29. G. Jean-Louis, D. F. Kripke and S. Ancoli-Israel, Sleep and
quality of well-being. Sleep 23, 1115–1121 (2000).
30. A. Ku¨hnlein, C. Heumann, S. Thomas, S. Heinrich and K.
Radon, Personal exposure to mobile communication networks
and well-being in children—a statistical analysis based on a
functional approach. Bioelectromagnetics 30, 261–269 (2009).
31. P. Hartge, Participation in population studies. Epidemiology 17,
252–254 (2006).
32. A. Lahkola, T. Salminen and A. Auvinen, Selection bias due to
differential participation in a case-control study of mobile phone
use and brain tumors. Ann. Epidemiol.15, 321–325 (2005).
33. J. Schu¨ z, J. P. Grigat, K. Brinkmann and J. Michaelis,
Residential magnetic fields as a risk factor for childhood acute
leukaemia: results from a German population-based case-control
study. Int. J. Cancer 91, 728–735 (2001).
34. J. A. Walsleben, V. K. Kapur, A. B. Newman, E. Shahar, R. R.
Bootzin, C. E. Rosenberg, G. O’Connor and F. J. Nieto, Sleep
and reported daytime sleepiness in normal subjects: the Sleep
Heart Health Study. Sleep 27, 293–298 (2004).
35. G. Neubauer, M. Feychting, Y. Hamnerius, L. Kheifets, N.
Kuster, I. Ruiz, J. Schuz, R. Uberbacher, J. Wiart and M. Ro¨o¨sli,
Feasibility of future epidemiological studies on possible health
effects of mobile phone base stations. Bioelectromagnetics 28,
224–230 (2007).
36. S. P. Loughran, A. W. Wood, J. M. Barton, R. J. Croft, B.
Thompson and C. Stough, The effect of electromagnetic fields
emitted by mobile phones on human sleep. Neuroreport 16, 1973–
1976 (2005).
37. R. Huber, T. Graf, K. A. Cote, L. Wittmann, E. Gallmann, D.
Matter, J. Schuderer, N. Kuster, A. A. Borbely and P.
Achermann, Exposure to pulsed high-frequency electromagnetic
field during waking affects human sleep EEG. Neuroreport 11,
3321–3325 (2000).
38. N. Leitgeb, J. Schro¨ttner, R. Cech and R. Kerbl, EMF-
protection sleep study near mobile phone base stations.
Somnologie 12, 234–243 (2008).
39. G. J. Rubin, R. Nieto-Hernandez and S. Wessely, Idiopathic
environmental intolerance attributed to electromagnetic fields
(formerly ‘electromagnetic hypersensitivity’): An updated
systematic review of provocation studies. Bioelectromagnetics
31, 1–11 (2010).
0MOHLER ET AL.