bmjopen.bmj.com · for peer review only subjective symptoms related to gsm radiation 3 the gsm...
TRANSCRIPT
For peer review only
Subjective Symptoms Related to GSM Radiation from Mobile Phone Base Stations
Journal: BMJ Open
Manuscript ID: bmjopen-2013-003836
Article Type: Research
Date Submitted by the Author: 16-Aug-2013
Complete List of Authors: Gomez-Perretta, Claudio; La Fe, Hospital, Research Center Navarro, Enrique; University of Valencia, Applied Physics Segura, Jaume; University of Valencia, Computer Science Portolés, Manuel; La Fe, Hospital, Research Center
<b>Primary Subject Heading</b>:
Public health
Secondary Subject Heading: Occupational and environmental medicine, Epidemiology
Keywords: PUBLIC HEALTH, OCCUPATIONAL & INDUSTRIAL MEDICINE, EPIDEMIOLOGY, Health & safety < HEALTH SERVICES ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open on A
pril 19, 2021 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2013-003836 on 30 Decem
ber 2013. Dow
nloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
Subjective Symptoms Related to GSM Radiation from Mobile Phone
Base Stations
Corresponding autor:
Enrique A. Navarro
Departamento de Física Aplicada, Universitat de València
C/. Dr. Moliner 50,
46100 Burjassot, Valencia, Spain
Email: [email protected]
Authors
Claudio Gómez-Perretta (1)
Enrique A. Navarro (2)
Jaume Segura (3)
Manuel Portolés (1)
(1) Research Center, Universitary Hospital La Fe, Avda Campanar s/n, 46009 Valencia, Spain
(2) Departament of Applied Physics, Universitat de València, C/. Dr. Moliner 50, 46100 Burjassot
(València), Spain.
(3) Departament of Computer Sciences, ETSE- Universitat de València, Avda Universitat s/n,
46100 Burjassot (València), Spain.
Running title
Subjective Symptoms Related to GSM Radiation
Keywords: Public health; GSM; Microwave Sickness.
Word Count: 3674
Page 1 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
Abstract
Objectives
This article focuses on the possible effects of exposure to radiofrequencies emitted by mobile
phone base stations. The possible influence of other variables such as demographic data and use of
electronic devices was also analysed. Since concerns have been raised about possible health
consequences caused by the emitted microwaves, in the present work we analyse if these
symptoms might be related to fear to the exposure, and therefore having a subjective basis.
Method
Binary logistic regression was used as the main statistical model because our target variables were
dichotomised variables (with or without symptoms) and the independents were either categorical or
continuous. This analysis was carried out on a regular basis and bootstrapping [1000 bootstrap
replications, 95% percentile method, and simple sampling] was used to provide more accurate
standard errors and confidence intervals. Various covariates were taken into account: age; gender;
radiofrequency electromagnetic fields (RF EMFs) from GSM technology; use of computers; use of
mobile (cellular) phones as well as the level of worry about the mobile phone base stations (BSs).
Results
The univariate logistic regression indicated that most of the symptoms were from weak to moderate
related with GSM exposure. Within the moderately affected were: lack of appetite and
concentration, irritability and sleep troubles. Change in -2 log likelihood showed similar results.
Moreover, gender did not influence the results, with the oldest having lighter difficulties for
hearing and walking. Besides, the use of mobile phones was strongly associated with lack of
appetite and moderately with vertigo, while worry about the BSs was strongly related with sleep
troubles.
Page 2 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
3
The GSM variable (RF EMF exposure) remained significant in the multivariate analysis, this
means that the incidence of the symptoms was due to GSM exposure independently of the co-
analyzed variables.
Changes in -2 log likelihood showed similar results. The bootstrapped values were similar to the
asymptotic standard errors and confidence intervals.
Conclusion
This new study partially confirms our preliminary results. Accordingly, we observe that incidence
of most of the symptoms was related to GSM levels – independently of the demographical
variables and some possible risk factors. Also, anxiety about adverse effects from exposure to RF
EMF, despite it is strongly related with sleep disturbances, does not influence the direct
association between GSM exposure and sleep problems. Other covariates such as age and use of
cellular were associated with hearing difficulties, walking difficulties and lack of appetite,
respectively. Lack of appetite, concentration, irritability and sleep troubles, are found moderately
associated with GSM exposure. Finally, the use of mobile phones was strongly associated with
lack of appetite and moderately with vertigo. The bootstrap results were similar – thereby
confirming the adequacy of our conventional analysis.
ARTICLE SUMMARY:
Article Focus: 1) It was stated by some authors that microwave syndrome would be related to anxiety
associated to fear and visual perception of cellular phone towers. In this manuscript we demonstrated that
although anxiety per se perturbs sleep, sleep disorder continued to be related with GSM exposure after
adjusting for these concerns. 2) It was showed that symptoms of the microwave sickness were moderately
related with this radiation, reinforcing that may exists some underlying biological mechanism behind these
symptoms. 3) This work presents a re-analysis of our previous study [4], using a more robust statistical
method.
Key Messages: 1) Existence of the microwave syndrome. 2) The symptoms were related to very low GSM
levels against Spanish and international guidelines [35], [36].
Strengths: We use a robust statistical analysis with a highly homogeneous sample in a homogeneous
environment. Limitations: We work with old data. The environmental radiation at that time (2001) was
mainly associated to GSM.
Page 3 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
4
Authors' contributions
CGP was one of the principle researchers responsible for design, conduct, and writing the
manuscript. CGP also made the main statistical contributions. EAN was also one of the main
researchers responsible for design and acquisition of RF EMF data, writing, and final review of
manuscript. JS contributed mainly to processing of RF EMF data, while MP was responsible for
design and final review of manuscript.
Page 4 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
5
The health risk due to exposure to microwave (MW), continues to be discussed today. The study
that led to this debate was initiated after verification that the U.S. embassy in Moscow was being
subjected to such radiation from 1953 to May 1975 [1]. Recently, a review of that episode [2]
reopened the debate about the potential harmfulness of these MW. In 2010, a review about this
topic [3], found that eight of the 10 studies evaluated through PubMed, reported increased
prevalence of adverse neurobehavioral symptoms or cancer in populations living at distances < 500
meters from base stations.
None of the studies reported exposure above accepted international guidelines, suggesting that
current guidelines may be inadequate in protecting the health of human populations. Thus, emerges
the need to reevaluate our pioneering work in this field, trying to improve it, adding new
procedures and data. To our understanding, few articles have addressed the possible association
between microwave sickness and microwave exposure from GSM (Global System for Mobile
Communications) base stations (BSs) since the publication of our first study [4]. Chronologically,
Santini et al., [5] and Gadzicka et al., [6] reported differences in the distance-dependent prevalence
of symptoms such as headache, impaired concentration, and irritability. Later, a study carried out
in Austria [7] showed a positive association between the measured electric field (GSM 900/1800)
in bedrooms and headaches, cold hands and feet, and difficulties in concentrating. An Egyptian
study [8] showed a prevalence of neurological complaints, such as headache, memory changes,
dizziness, tremors, depressive symptoms, and sleep disturbances among subjects more directly
exposed to GSM signals from BSs.
The symptoms reported by all the above cited authors belong to those attributed to the
microwave syndrome [9]. However, one article [10] using personal monitored data from GSM-
UMTS frequency bands did not find any statistical association in adults. More recently, the same
authors observed no association in children [11], contradictory results in children and adolescents
[12], and concluded that the few observed significant associations were not causal but rather
Page 5 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
6
occurred by chance. Blettner et al. [13] reported in Phase 1 of their study more health problems
closer to base stations, but in Phase 2 [14] they concluded that measured EMF emissions were not
related to adverse health effects.
Other researchers focused their work on the possible existence of subjects with sensitivity to
GSM or UMTS signals according to psychological, cognitive, or autonomic assessment. These
researchers used short-term exposure (only 30-50 minutes) under laboratory conditions [15-19]
and revealed a large disparity between subjects. Recently, a study measuring several biological
stress markers [20] found that RF-EMF emitted by mobile phone base stations from 5.2 µW/m2 to
2126.8 µW/m2 increased cortisol and salivary alpha-amylase, while IgA concentration was not
significantly modified
In 2010, the Selbitz study [21], described a significant dose-response relationship in symptoms
related with sleep, mood, joints, infections, skin condition, as well as neurological cardiovascular,
visual, and auditory systems, and the gastrointestinal tract.
In the work of Danker-Hopfe et al. [22] it was not evident the existence of short-term
physiological effects of EMF on sleep quality, however it was stated that the presence of BSs
per se (not the EMF) may have a negative impact on sleep quality.
In 2012, a Polish study did not present correlation between the electric field strength and the
frequency of subjective symptoms however showed a correlation between subjective symptoms
and the distance to BSs, [23] . A study carried out in Egypt [24] revealed that exposure to EMF
emitted either from mobile phones or BSs had significant effects on pituitary-adrenal axis. More
recently, the work developed in Iran [25] indicated that symptoms such as nausea, headache,
dizziness, irritability, discomfort, nervousness, depression, sleep disturbance, memory loss and
lowering of libido were statistically significant in the inhabitants living near the BSs (<300m
distances) compared to those living far from the BSs (>300m).
Page 6 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
7
In our cross-sectional analysis [4], 11 out of 16 symptoms showed significantly higher scores in
the group with the maximum exposure level. The symptoms are included in the microwave
syndrome. We also reported significant correlation coefficients between the measured electric
field and fourteen out of sixteen symptoms.
A review [26] recently established several conditions for epidemiological studies to be eligible
for introduction in general analysis: eligible studies must quantify exposure using objective
measures (such as distance to the nearest BS, spot, or personal exposure measurements in a
specific frequency range); possible confounders must be considered; and the selection of the
study population must be clearly free of bias in terms of exposure and outcomes.
Accordingly, in this re-analysis of our previous study [4], possible confounders were included in
addition to the specific RF EMF measurements made in 2001 (covering the specific range between
900 and 1800 MHz). So, we co-analyzed the effects of other variables such as socio-demographic
data and use of electronic devices. Lately, concern about being damaged by the radiation of the
antenna was also analyzed.
In the present work a more robust statistical method indifferent to the assumption of normality was
employed. Re-sample method or bootstrapping is based on inferences rather than making
assumptions about the population, treating a given sample as the population. Consequently, we
could extrapolate our results to the entire population from which the sample was obtained and
reduce sample size effect.
The present study re-evaluates our initial work using a new methodology, analyzing confounding
effects, in order to test more deeply the reliability of those results. Interestingly, this period was
free of other sources of RF such as WIFI or UMTS etc. or the massive use of mobile phones,
enabling a specific study of GSM technology.
Finally, the suitability of the size of the sample was analyzed.
Page 7 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
8
METHODS
Study design
The study was carried out in La Ñora, a small village in the south-east of Spain. Two BS masts,
each about 30 m height were sited at different positions to provide GSM-900-1800 coverage. The
GSM 900 BS was positioned not earlier than 1997 while the GSM 1800 BS was built in December
1999.
The main demographic characteristics of the sample and the use of electronic devices was
recollected through a Spanish-language questionnaire [5]. All of the subjects were of the same
ethnic origin, shared similar family income levels and general standard of living, and were born in
La Ñora or nearby. All the residents in the study were living in the village before the erection of
both BSs. All of the residents were at home for more than eight hours a day for at least six days a
week and normally slept at home.
The core of the questionnaire was a symptom checklist for estimating the frequency of 15 health-
related symptoms attributed to microwave sickness. These symptoms are: fatigue; irritability;
headaches; nausea; loss of appetite; sleep disorders; depressive tendency; dizziness; concentration
difficulties; memory loss; skin lesions; visual and hearing deficiencies; walking difficulties; and
cardiovascular problems. The frequency was quantified as: never suffer = 0; sometimes = 1; often
= 2; and very often = 3.
The percentage of residents who reported electrical transformers less than 10 m from their home
was 21.6%; while a 42% reported high voltage power lines less than 100 m from home. Finally,
40% of residents reported a TV transmitter within a radius of around 4 km.
The questionnaire included a statement that its purpose was health research, that the data gathered
would be confidential, and that the study was approved by the Ethical Committee of the University
of Valencia in accordance with the Declaration of Helsinki (http://www.wma.net/e/policy/b3.htm).
Page 8 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
9
215 questionnaires were randomly distributed through seventeen streets which represented
practically the entire village. 150 were collected being the difference due of not being at home or
refusal to fulfil the questionnaire.
Then, 101 RF EMF measurements in bedrooms were made in 2001. The 49 missing residents were
due mainly because, refusal to give admittance for taking measurements, residents not being at
home for the scheduled measurement appointment and having serious healthy problems.
A further review was conducted in 2010. Consequently, two residents were removed from the
primitive survey, one because of alcohol abuse and another because pregnancy. Moreover, eleven
subjects were eliminated because there was a change (it was raining) in the usual dry weather
conditions when registering their respective broadband measurements.
So, the reanalysis of the dataset, which is the main focus of this paper, was finally performed with
88 subjects (45 female and 43 male) instead of the 101 analysed in 2001.
Concerns about microwave exposure
66 subjects out of 88 were reached by telephone in February 2012 following two questions:
a) ¿Were you worry about the masts (BSs) at the time when it was erected?
b) ¿Did you believe its radiation (BSs) could damage your health?
In all cases those who were worried about the masts were concerned about health consequences. 27
subjects (40, 9 %) responded no and 39 (59, 1%) responded yes. Responses were not related to age
(ANOVA test), sex (lambda statistic) or subjective distance to BS (Somers´d statistic).
Exposure assessment
Broadband measurements were made on two Saturdays in February and March 2001 from 11 am to
7pm with a portable electrical field (400 MHz – 3 GHz) detector (Nuova Elettronica Model LX-
1435). This meter was calibrated with an HP-8510C network analyser inside an anechoic chamber
at the University of Valencia. During the bedroom exposure assessment the electric field probe was
held for approximately five minutes about one meter from the walls and 1.2 meters above the
Page 9 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
ground – and moved around a circle of 0.25 m radius, orientating the antenna in different directions
to obtain the maximum electrical field strength above the bed.
To check the intensity of TV and radio channels, as well as the number of working channels for the
GSM-900-1800 BSs, measurements of the spectral power density were carried out with a probe
antenna and a portable spectrum analyser.
The probe was mounted on a linen phenolic tripod 1.2 m above the ground. The location of the
probe was the same on both days – on a hill next to the village and 20 m from the BS. With the
spectrum analyser we scanned the frequency bands and the levels were averaged for six minutes.
The measurement of the spectrum was similar on both days – with a difference in the peak
estimation (channel carriers) of about 1 dB.
Statistical analysis
Demographical data was analysed using the Mann-Whitney one-way variance analysis and Chi
square test.
The main statistical analysis was made using binary logistic regression (mode enter) carried out on
a regular basis and subsequent bootstrapping [1000 bootstrap replications, 95% percentile method
and simple sampling] [27] that could provide more accurate standard errors and confidence
intervals. After producing (1000) bootstrap replicates θ b of an estimator θ, the bootstrap standard
error was the standard deviation of the bootstrap replicates:
SE (θ) = √ [∑ (θb - θ) **2 / (r-1)] b=1� r where θ is the mean of the θb. Because of our small sample size, a non-parametric confidence
interval for the estimate (mean) was constructed from the quartiles of the bootstrap sampling
distribution of θ. The 95% percentile interval (θ (lower) <θ< θ (upper) is shown, and where the θb
are the r-ordered bootstrap replicates: lower = 0.025 × r (sample 25) and upper = 0.975 × r (sample
975).
Page 10 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
The 15 health-related symptoms described above constituted the dichotomous dependent variables.
Then, univariate analysis was performed for each symptom and for each of the predictors variables:
exposure to base station (µW/m2 as a natural logarithmic) and age were used as continuous
variables; while gender, computer use >2 h/ day, mobile phone use > 20 minutes per day and worry
about the antennae were used as dichotomous variables. The covariates with predictive value were
considered for the multivariate analysis. Thus possible confounder effects were evaluated.
In all cases, change in -2 log likelihood, odds ratio, 95%-confidence intervals (95%-CI), as well as
the probability value (p-value) were calculated. For all tests, a p-value below 0.05 was considered
significant.
The dependent variables (health-related symptoms) given in four ordinal categories (0=never,
1=sometimes, 2=often; 3=very often) were dichotomised (0,1 = 0 versus 2, 3 = 1).
We used the GSM exposure (the measurement of RF EMF in the bedroom) as a continuous
variable because it is well recognised that categorisation of continuous variables introduces major
problems in the analysis and interpretation of models derived in a data-dependent fashion [28, 29,
30].
We chose exposure values in the logarithmic form because these values are well grouped around
their median; while the raw values showed a high dispersion of values, with two outliers and ten
extreme values (data not shown).
Confounding was assessed by adding the potential confounding variable to the model and making a
subjective decision as to whether or not the coefficient of the variable of interest, ORs of GSM
exposure, had changed substantially. A 10 per cent variation was accepted as a considerable change
Possible interactions between covariates were also evaluated.
The maximum number of covariates included in each multivariate analysis was calculated
following this formula [31]. Let π be the smallest of the proportions of negative or positive cases
Page 11 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
in the population and k the number of covariates, then the minimum number of cases to include
is:
N = 10 k / π
Goodness-of-fit tests such as the classification table, the Hosmer-Lemeshow statistic, ROC
curves, Cox and Snell’s, and Nagelkerke’s Pseudo R square measures were used. The Wald
statistic was also evaluated to test the significance of individual independent variables.
Moreover, possible multicolinearity was also tested.
For statistical analysis SPSSTM for Windows was used.
Because an exposure assessment for transformers, high voltage power lines, and radio or TV
transmitters based on self-estimated distances would not produce a reliable exposure estimate, it
was decided to omit these covariates in the analysis.
RESULTS
Demographical data and the percentage of users of personal computers and mobile phones were
analysed. The sample age distribution was 42, 17 ± 17, 61 and [15–81] interval, while 13,6 % used
regularly computers (6,6 % were women and 7,0% men) and 23,9 % mobile phones (8,8 %
corresponded to women and 15,1 % to men). In the total sample, 51,1 % were women (mean age =
45.08 years (SD =17.98) ; [interval=15-81] and 48,9% men (39.12 (16.88) [15-75]. No differences
related with age, use of mobile phones or computers between sexes were found.
The univariate logistic regression indicated that age was inversely associated with irritability [OR =
0.97 (0.95-0.99)] and that the oldest had the greatest difficulties hearing [1.03 (1.01-1.06)] and
walking [1.04 (1.01-1.07)] . However, gender clearly did not influence the outcome of any
dependent variable. Use of mobile phones was linked with lack of appetite and vertigo, while
worry about the radiation from BSs was associated with sleep troubles (Table 1).
Page 12 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
Most of the symptoms were related with GSM exposure, especially: fatigue, irritability, lack of
appetite, sleep troubles, depression, and lack of concentration. Change in -2 log likelihood showed
similar results (Table 2).
ROC curves of each one of the logistic regression models (GSM exposure vs. each symptom)
oscillated between 0.65 and 0.87 (Table 3). Headaches (0, 84), nausea (0,86), appetite (0,87), and
vascular problems (0,85) showed the highest values, while memory (0,67), skin (0,67), and visual
disturbances (0,65) showed the lowest ones. The Hosmer and Lemeshow test indicated that most
analyses showed no significance p- values. The exceptions were fatigue (0,003), depression (0,003)
and vertigo (0,03). In the majority of the cases the models predicted better specificity than
sensitivity. Only in the case of headaches and sleep disorder, sensitivity prevailed over specificity
(Table 3 – classification table). In the extreme case, skin and vascular problems showed null or
minimum sensitivity, but 100 % specificity. Nagelkerke pseudo R-squared, showed acceptable
coefficients with the exception of the symptoms related with vertigo and skin problems (Table 3).
The lowest threshold cut-off values of GSM were for headaches, sleep, irritability and attention;
while vascular and skin problems showed the highest threshold cut-off values (Table 3).
The influence of other covariates on the GSM ORs coefficients, such as age, cellular use and
concern about the antennae were always less than 10% (Tables 4,).
There was no observed multicollinearity among variables. Kappa values according to factor
analysis were always lower than 2 and well below the critical value of 30.
Finally, no interactions between covariates were observed.
Standard errors and confidence intervals obtained by re-sampling were similar to those calculated
from the asymptotic approximation (Tables 4, 5). There was a small bias or difference between
the average bootstrap coefficients (not shown) and the respective estimates obtained from the
original sample.
Page 13 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
DISCUSSION
This new study partially confirms our preliminary results. So, the most of the symptoms were
related to GSM levels – independently of the demographical variables and some possible risk
factors.
Related to microwave radiation, the spectral power density analysis maintained that the most
important contribution to broadband measurements was from GSM 900/1800, and the main
variability of the measurements between different places was due to a different coverage of the
GSM 900/1800 signals, i.e. spatial variability. This was further supported by the fact that the
antenna used was fairly insensitive to frequencies below 400 MHz. Therefore, the radio channels
80-110 MHz were not a significant part of the broadband measurements. Moreover, the narrow
band measurements showed TV channels with substantially lower intensities than the GSM
900/1800 signals. The effects from these exposures will therefore not confound the effects of BSs.
Moreover, some authors [7] found that the only relevant contribution to the variance of the high
microwave exposure was from BSs – up to 93% of variance. Moreover, at the time of our study the
GSM signal was almost invariable in time because there were very few calls. The main
contribution was made from the broadcast channels working almost constantly throughout the day.
Short-range evaluations of exposure could be acceptable for describing a 24-hour period and the
measurements were made in bedrooms – a location where the subjects were assumed to spend
significant periods of time.
However, some subjects were mobile phone users at the time of this study and exposure from a
mobile phone during a phone call is much higher than that received from BSs. Nevertheless, some
authors [7] stated about that there is no a priori argument why these lower levels should have no
effect on the presence of a widespread use of mobile telephones. Exposure from a BS will be at a
low but almost constant level for many hours a day and especially at night.
Page 14 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
While GSM exposure was associated with most of the symptoms, walking difficulties and hearing
loss was correlated only with age. Also, age remained slightly inversely associated with irritability.
Those who were users of cellular phones were more prone to complain of loss of appetite and
vertigo, while those who expressed worry about the BSs, were associated with having sleep
problems. This later finding was in concordance with two other articles [7, 14, 22]. However,
worry about the BSs was unrelated with age, gender or subjective distance to BS. This agrees with
article [32] claiming that there was no significant association between symptom occurrence and
actual distance to BSs or powerlines.
Contrarily, some authors indicated that opponents of mobile phone towers generally do not express
anxieties about electromagnetic field exposure, indicating that the risk rating is comparable with
other perceived common hazards in the modern world [33].
None of the analysed covariates behaved as confounders. So, the relationship between GSM
exposure with irritability, sleep troubles, lack of appetite and vertigo remained significant despite
the introduction of the above covariates.
When the conventional multivariate analysis was tested using bootstrapping it was observed that
the standard error and confidence intervals obtained by re-sampling were similar to those
calculated from asymptotic approximation and this supports the adequacy of our conventional
analysis. Our sample, chosen at random, would qualify to represent the population from which it
came.
The model appeared generally well adjusted while the cut-off values could constitute good
guidance for predicting the threshold of symptom appearance.
Page 15 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
CONCLUSIONS
This new study partially confirms our preliminary results about microwave sickness resulting from
exposure to emissions from GSM mobile phone BSs. So, fatigue, irritability, lack of appetite, sleep
troubles, depression, and lack of concentration were especially related with GSM exposure.
These results were independent of the main socio-demographical variables, other EMF exposures
and the worry about being irradiated. Nevertheless, we confirm that apprehension about modern
technology could predict some symptom, specially related with sleep problems.
Our results agree with those who claimed that by distorting perceptions of risk, disproportionate
precaution might paradoxically lead to illness that would not otherwise occur [34]. However,
health changes related with GSM exposure seems to occur unrelated with those fears. Finally, this
exposure was very low in that period and also very low against Spanish recommendations [35] and
international guidelines [36].
Recommendations
We subscribe the guidelines observed by other authors [37] to follow the principle of prevention
as long as the non-thermal effects are not considered in any official standard. This includes
exposure minimization in the frame of the technical feasibility to guarantee a significantly
reduction in the long term radiation exposures of cellular phone towers in residential areas.
Epidemiological and clinical studies should be continuing to observe possible health changes in
the population. Finally, clear information about the correct use of the newer electronic devices
should be implemented.
Page 16 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
Competing interests
The authors declare that they have no competing interests.
ACKNOWLEDGEMENTS
We would like to express our gratitude to Angeles Martinez Gomez for her assistance during the
field work in La Ñora; as well as the Spanish Ministry of Science and Technology for grant FIT
number 070000-2002-58.
Page 17 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
REFERENCES
1. Lilienfeld AM, Tonascia J, Tonascia S, Libauer CH, et al. Foreign Service Health Status
Study, Evaluation of health status of foreign service and other employees from selected
Eastern European posts. Baltimore: Dept Epidemiology, School of Hygiene and Public
Health, The Johns Hopkins University; 1978.
2. Elwood JM. Microwaves in the cold war: the Moscow embassy study and its interpretation.
Review of a retrospective cohort study. Environ Health 2012; Nov 14;11:8.
3. Khurana VG, Hardell L, Everaert J, Bortkiewicz A, et al. Epidemiological evidence for a
health risk from mobile phone base stations. Int J Occup Environ Health 2010; 16(3):263-7.
4. Navarro EA, Segura J, Portolés M, Gómez-Perretta C. The Microwave Syndrome: A
Preliminary Study in Spain. Electromagn Biol Med 2003; 22:161 – 169.
5. Santini R, Santini P, Danze J, Le Ruz P, et al. Study of the health of people living in the
vicinity of mobile phone base stations: I. Influences of distance and sex. Pathol Biol 2002;
50:369 – 373.
6. Gadzicka E, Bortkiewicz A, Zmyslony M, Szymczak W, et al.. Assessment of subjective
complaints reported by people living near mobile phone base stations [Abstract].
BiuletynPTZE Warszawa. 2006; 14:23-26.
7. Hutter H, Moshammer H, Wallner P, Kundi M. Subjective symptoms, sleeping problems,
and cognitive performance in subjects living near mobile phone base stations. Occup Environ
Med 2006; 63:307–313.
8. Abdel-Rassoul G, El-Fateh OA, Salem MA, Michael A, Farahat F, et al. Neurobehavioral
effects among inhabitants around mobile phone base stations. Neurotoxicology 2007;
28:434-440.
Page 18 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
1
9. Johnson-Liakouris AG. Radiofrequency (RF) Sickness in the Lilienfeld Study: an effect of
modulated microwaves? Arch Environ Health 1998;53:236-238.
10. Thomas S, Kühnlein A, Heinrich S, Praml G, et al. Personal exposure to mobile phone
frequencies and well-being in adults: a cross-sectional study based on dosimetry.
Bioelectromagnetics 2008; 29:463-470.
11. Kühnlein A, Heumann C, Thomas S, Heinrich S, et al. Personal exposure to mobile
communication networks and well-being in children--a statistical analysis based on a
functional approach. Bioelectromagnetics 2009; 30:261-9.
12. Heinrich S, Thomas S, Heumann C, von Kries R, et al. Association between exposure to
radiofrequency electromagnetic fields assessed by dosimetry and acute symptoms in children
and adolescents: a population based cross-sectional study. Environ Health 2010; 9:75.
13. Blettner M, Schlehofer B, Breckenkamp J, Kowall B, et al Mobile phonebase stations and
adverse health effects: Phase 1 of a populationbased,cross-sectional study in Germany. Occup
Environ Med.2009;66:118-123.
14. Berg-Beckhoff G, Blettner M, Kowall B, Breckenkamp J, et al. Mobile phone base stations
and adverse health effects: phase 2 of a cross-sectional study with measured radio frequency
electromagnetic fields. Occup Environ Med 2009; 66:124-130.
15. Augner C, Florian M, Pauser G, Oberfeld G, et al. GSM Base Stations: Short-Term Effects
on Well-Being. Bioelectromagnetics 2009; 30:73-80.
16. Eltiti S, Wallace D, Ridgewell A, Zougkou K, et al. Does short-term exposure to mobile
phone base station signals increase symptoms in individuals who report sensitivity to
electromagnetic fields? A double-blind randomized provocation study. Environ Health
Perspect 2007;115:1603-1608.
17. Eltiti S, Wallace D, Ridgewell A, Zougkou K, et al. Short-term exposure to mobile phone
base station signals does not affect cognitive functioning or physiological measures in
Page 19 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
individuals who report sensitivity to electromagnetic fields and controls. Bioelectromagnetics
2009;30:556-63.
18. Furubayashi T, Ushiyama A, Terao Y, Mizuno Y, et al. Effects of short-term W-CDMA
mobile phone base station exposure on women with or without mobile phone related
symptoms. Bioelectromagnetics 2009; 30:100-113.
19. Zwamborn AP, Vossen SH, van Leersum BJ. Effects of global communication system
radio-frequency fields on well being and cognitive functions of human subjects with and
without subjective complaints. Netherlands Organisation for Applied Scientific Research
(TNO) 2003;148:1-89.
20. Augner C, Hacker GW, Oberfeld G, Florian M, et al. Effects of exposure to GSM mobile
phone base station signals on salivary cortisol, alpha-amylase, and immunoglobulin A.
Biomed Environ Sci 2010;23:199-207.
21. Eger H, Jahn M. Spezifische Symptome und Mobilfunkstrahlung in Selbitz (Bayern)
Evidenz für eine Dosiswirkungsbeziehung. Umwelt·medizin·gesellschaft 2010; 2: 130-139.
22. Danker-Hopfe H, Dorn H, Bornkessel C, Sauter C. Do mobile phone base stations affect
sleep of residents? Results from an experimental double-blind sham-controlled field study.
Am J Hum Biol. 2010; 5:613-8
23. Bortkiewicz A, Gadzicka E, Szyjkowska A, et al. Subjective complaints of people living
near mobile phone base stations in Poland. Int J Occup Med Environ Health 2012 ;25(1):31-
40
24. Eskander EF, Estefan SF, Abd-Rabou AA. How does long term exposure to base stations
and mobile phones affect human hormone profiles? Clin Biochem 2012 ; 45(1-2):157-61
25. Shahbazi-Gahrouei D, Karbalae M, Moradi HA, et al. Health effects of living near mobile
phone base transceiver station (BTS) antennae: a report from Isfahan, Iran. Electromagn Biol
Med. 2013 Jun 19.
Page 20 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
26. Röösli M, Frei P, Mohler E, Hug K. Systematic review on the health effects of exposure to
radiofrequency electromagnetic fields from mobile phone base stations. Bull World Health
Organ 2010; 88: 887–896.
27. Efron B, Tibshirani R: An Introduction to the Bootstrap. Boca Raton, FL: Chapman &
Hall/CRC 1993; ISBN 0-412-04231-2.software
28. Maxwell SE, Delaney HD. Bivariate median-splits and spurious statistical significance.
Psychological Bulletin 1993; 113:181–190.
29. Royston P, Altman DG, Sauerbrei W.Dichotomizing continuous predictors in multiple
regression: a bad idea. Stat. Med 2006; 25:127-141.
30. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous confounding
variable is categorized in logistic regression analyses. Stat. Med 2004; 23:1159–1178.
31. Peduzzi P, Concato J, Kemper E, Holford TR, et al. A simulation study of the number of
events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373-1379.
32. Baliatsas C, van Kamp I, Kelfkens G, et al. Non-specific physical symptoms in relation to
actual and perceived proximity to mobile phone base stations and powerlines. BMC Public
Health 2011; 11:421
33. Hutter H, Moshammer H, Wallner P, Kundi M. Public perception of risk concerning
celltowers and mobile phones. Soz Präventivmed 2004;49: 62–66.
34. Coggon D. Health Risks from Mobile Phone Base Stations. Occup Environ Med 2006;
63:298–299.
35. Real Decreto 1066/2001 Telecomunicaciones/TelefoniaMovil/RD1066_2001.pdf
http://www.mityc.es/esES/OficinaVirtual/Documents/SE%20
36. International Commission on Non-Ionizing Radiation Protection: Guidelines for limiting
exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz).
Health Phys 1998; 74: 494–522.
Page 21 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
37. Haumann T, Münzenberg U, Maes W, Sierck P. HF-Radiation levels of GSM cellular
phone towers in residential areas. 2nd International Workshop on Biological effects of
EMFS. October 2002; Rhodes (Greece) Vol 1:327-333.
Page 22 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
TABLE LEGENDS
Table 1. . Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Table 2. ORs & 95% CIs of GSM exposure for predicting clinical symptoms. Change in -2 log likelihood is also shown.
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (ln). Cut-off values of exposure to microwaves when the predicted probability of symptoms is more than 50%. The data are presented in ascending order. Table 4. Adjusted GSM variable coefficients: Asymptotic standard errors (SE) and 95% confidence intervals (CI). Statistics for r = 1000 bootstrapped binary logistic regression and 95% percentile confidence intervals are also shown for comparison. Univariate coefficients for worry about BSs (a) and GSM exposure (b); and (c) Multivariate GSM exposure coefficient adjusted for concern about the BSs.
Table 5. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Page 23 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
Table 1. Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Symptom/
Variable
Worry about BSs (1) Computer (not use) Mobile (not use) OR (95%CI) OR (95%CI) OR (95%CI)
Fatigue 0.67 0.23-1.90 2.62 0.76-9.04 1.56 0.56-4.35 Irritability 1.13 0.43-3.03 1.56 0.45-5.34 2.62 0.94-7.33 Headaches 1.75 0.62-4.94 1.39 0.34-5.58 1.56 0.51-4.83 Nausea 0.68 0.18-2.24 0.34 0.04-2.84 1.43 0.44-4.67 Lack of appetite 1.05 0.33-3.40 3.16 0.87-11.44 4.28** 1.43-12.78 Trouble sleeping 3.12* 1.10-8.86 0.55 0.16-1.88 0.74 0.27-2.02 Depression 1.06 0.39-2.93 0.81 0.22-2.93 1.03 0.38-2.84 Lack of concentration 0.92 0.35-2.47 1.11 0.33-3.76 2.79 0.99-7.80 Memory loss 1.71 0.62-4.75 0.41 0.10-1.64 1.35 0.50-3.61 Skin alterations 0.74 0.23-2.35 φ φ 0.63 0.16-2.45
Visual disturbances 1.31 0.48-3.60 0.77 0.21-2.77 1.63 0.60-4.39 Vertigo 0.61 0.20-1.91 0.77 0.19-3.10 2.90* 1.04-8.07 Vascular alteration 0.96 0.27-3.43 1.48 0.35-6.17 2.04 0.65-6.41 Hearing problems 0.59 0.20-1.70 0.77 0.19-3.10 0.48 0.15-1.60 Walking difficulty 0.60 0.20-1.79 φ φ 0.42 0.11-1.60 * p<0.05; ** p<0.01; (1) Not worried, as reference code (2) & (3) Not device use, as reference code, φ any subject affected using computer Table 2. ORs & 95% CIs of GSM exposure for predicting clinical symptoms. Change in -2 log likelihood is also shown.
Symptom
OR / CI
Change in -2 Log likelihood
Fatigue 1.39*** (1.14-1.70) 11.74***
Irritability 1.51*** (1.23-1.85) 19.36***
Headaches 1.43** (1.15-1.78) 12.32***
Nausea 1.38** (1.09-1.73) 8.3**
Appetite 1.58*** (1.23-2.03) 16.31***
Sleep 1.49*** (1.20-1.84) 16.38***
Depression 1.41*** (1.16-1.72) 13.99***
Concentration 1.54*** (1.25- 1.89) 20.75***
Memory 1.27** (1.06- 1.52) 7.29**
Skin 1.24* (1.001-1.54) 4.08*
Visual 1.23 * (1.03-1.46) 5.30*
Vertigo 1.36** (1.11-1.66) 10.14***
Vascular 1.32* (1.05-1.64) 6.30*
Hearing 0.96 (0.80-1.15)
Walking 0.95 (0.78-1.15)
* p<0.05; ** p<0.01; *** p<0.001
Page 24 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (ln). Cut-off values of exposure to microwaves when the predicted probability of symptoms is more than 50%. The data are presented in ascending order. Symptom ROC curves
Area
Classification table
Pseudo-R**2
(1)
Cut-off (Ln GSM)
Cut-off GSM (µW/m2)
SSV SPF AV
Headaches 0.84*** 0.90 0.23 0.72 0.41 0.57 1.77 Sleep 0.78*** 0.82 0.66 0.76 0.28 1.66 5.26
Attention 0.78*** 0.67 0.72 0.69 0.28 3.61 36.97 Irritability 0.76*** 0.67 0.73 0.71 0.26 3.61 36.97 Memory 0.67** 0.54 0.77 0.67 0.11 4.99 146.94
Depression 0.75*** 0.46 0.76 0.65 0.20 5.22 184.93 Visual 0.65* 0.24 0.83 0.60 0.08 5.91 368.71 Fatigue 0.73*** 0.22 0.90 0.69 0.18 6.53 685.4 Vertigo 0.74*** 0.16 0.87 0.67 0.19 6.53 685.4 Appetite 0.87*** 0.40 0.94 0.85 0.43 7.31 1495.18 Nausea 0.86*** 0.46 0.93 0.87 0.38 7.31 1495.18
Vascular 0.85*** 0.20 1.0 0.90 0.34 8.02 3041.18 Skin 0.67* 0.00 1.00 0.81 0.072 9.06 8604.15
* p<0.05, ** p<0.01, *** p<0.001 (1) Nagelkerke
Page 25 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
Table 4. Adjusted GSM variable coefficients: Asymptotic standard errors (SE) and 95% confidence intervals (CI). Statistics for r = 1000 bootstrapped binary logistic regression and 95% percentile confidence intervals are also shown for comparison. Univariate coefficients for worry about BSs (a) and GSM exposure (b); and (c) Multivariate GSM exposure coefficient adjusted for concern about the BSs.
Symptom Covariates Normal (φ)))) Bootstrap
OR SE 95% CI
Bias SE Sig. 95% percentile
Lower Upper Lower Upper Irritability Age & Mw 1.47*** 0.1 1.20 1.81 0.02 0.1 .002 1.23 1.94
Appetite UCel & Mw 1.53** 0.1 1.19 1.99 0.03 0.1 .001 1.24 2.22
Vertigo UCel & Mw 1.32** 0.1 1.08 1.62 0.01 0.1 .003 1.11 1.68
Sleepless Afraid (a) 3.12* 0.5 1.10 8.86 0.02 0.6 .03 1.12 10.87
Sleepless Mw (b) 1.69*** 0.1 1.27 2.24 0.05 0.2 .001 1.34 2.83
Sleepless Affraid -Mw (c) 1.64*** 0.1 1.22 2.19 0.06 0.7 .001 1.28 2.59
φ Binary logistic regression (enter) Ucel: cellular use; Mw: GSM exposure
Page 26 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation
2
Table 5. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Symptom B
Bootstrap Normal
Bias SE 95% percentile
intervals SE 95% CI
Lower Upper Lower Upper Fatigue 0.329 0.012 0.097 0.155 0.539 0.102 0.128 0.529 Irritability 0.411 0.016 0.110 0.241 0.670 0.104 0.207 0.615 Headache 0.358 0.022 0.139 0.149 0.688 0.113 0.137 0.578 Nausea 0.319 0.013 0.124 0.099 0.590 0.118 0.088 0.550 Appetite 0.456 0.026 0.134 0.264 0.784 0.128 0.205 0.707 Sleep 0.396 0.022 0.124 0.193 0.690 0.109 0.181 0.610 Depression 0.346 0.012 0.102 0.174 0.583 0.100 0.151 0.541 Attention 0.429 0.020 0.118 0.254 0.711 0.106 0.222 0.636 Memory 0.237 0.009 0.098 0.057 0.448 0.091 0.058 0.415 Skin 0.217 0.008 0.110 0.011 0.451 0.110 0.001 0.433 Visual 0.203 0.004 0.093 0.037 0.398 0.090 0.026 0.379 Hearing -0.05 -0.002 0.089 -0.219 0.143 0.093 -0.228 0.135 Vertigo 0.306 0.010 0.101 0.127 0.530 0.102 0.107 0.505 Walking -0.05 -0.006 0.098 -0.265 0.120 0.098 -0.246 0.138 Vascular 0.274 0.010 0.109 0.084 0.520 0.114 0.051 0.497
Page 27 of 27
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation from Mobile Phone Base Stations: A cross-sectional study
Journal: BMJ Open
Manuscript ID: bmjopen-2013-003836.R1
Article Type: Research
Date Submitted by the Author: 29-Oct-2013
Complete List of Authors: Gomez-Perretta, Claudio; La Fe, Hospital, Research Center Navarro, Enrique; University of Valencia, Applied Physics Segura, Jaume; University of Valencia, Computer Science Portolés, Manuel; La Fe, Hospital, Research Center
<b>Primary Subject Heading</b>:
Public health
Secondary Subject Heading: Occupational and environmental medicine, Epidemiology
Keywords: PUBLIC HEALTH, OCCUPATIONAL & INDUSTRIAL MEDICINE, EPIDEMIOLOGY, Health & safety < HEALTH SERVICES ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open on A
pril 19, 2021 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2013-003836 on 30 Decem
ber 2013. Dow
nloaded from
For peer review only
1
1
Subjective Symptoms Related to GSM Radiation from Mobile Phone
Base Stations: A cross-sectional study
Corresponding autor:
Enrique A. Navarro
Departamento de Física Aplicada, Universitat de València
C/. Dr. Moliner 50,
46100 Burjassot, Valencia, Spain
Email: [email protected]
Authors
Claudio Gómez-Perretta (1)
Enrique A. Navarro (2)
Jaume Segura (3)
Manuel Portolés (1)
(1) Research Center, Universitary Hospital La Fe, Avda Campanar s/n, 46009 Valencia, Spain
(2) Departament of Applied Physics, Universitat de València, C/. Dr. Moliner 50, 46100 Burjassot
(València), Spain.
(3) Departament of Computer Sciences, ETSE- Universitat de València, Avda Universitat s/n,
46100 Burjassot (València), Spain.
Running title
Subjective Symptoms Related to GSM Radiation
Keywords: Public health; GSM; Microwave Sickness.
Page 1 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
2
2
Abstract
Objectives: This paper is a cross-sectional study, a reanalysis of the data from Navarro et al,
2003 (10), in which health symptoms related to microwave exposure from mobile phone base
stations (BS) were explored. We analyze if these symptoms might be related to fear to the
exposure, and therefore having a subjective basis.
Method: Since weather circumstances could influence the exposure, we restricted the data to
those measured under similar weather conditions. The participants with known illness in 2003 were
at this time disregarded: 88 subjects instead of 101 (in 2003) were analyzed. A statistical method
indifferent to the assumption of normality was now employed: Binary logistic regression for
modelling a binary response, e.g. having fatigue (1) or not (0), introducing the exposure as a
predictor variable. This analysis was carried out on a regular basis and bootstrapping [95%
percentile method] to provide more accurate confidence intervals.
Results: The symptoms more related with exposure were: Lack of appetite [odds ratio (OR)] =
1.6, 95% confidence interval (95%CI) = 1.2-2.0, lack of concentration [OR = 1.5, 95% CI = 1.25-
1.89], irritability [OR = 1.51, 95% CI = 1.23-1.85] and sleep troubles [OR = 1.5, 95% CI = 1.2-1.8].
Change in -2 log likelihood showed similar results. Concerns about the BS were strongly related
with sleep troubles [OR = 3.1, 95% CI = 1.1-8.9]. The exposure variable remained statistically
significant in the multivariate analysis. The bootstrapped values were similar to the asymptotic
confidence intervals
Conclusion: This study confirms our preliminary results. We observe that incidence of most of
the symptoms was related to exposure levels – independently of the demographical variables and
some possible risk factors. Also, concern about adverse effects from exposure, despite it is strongly
related with sleep disturbances, does not influence the direct association between exposure and
sleep.
Page 2 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
3
3
ARTICLE SUMMARY:
Article Focus: 1) It was stated by some authors that microwave syndrome would be related to
anxiety associated to fear and visual perception of cellular phone towers. In this manuscript we
demonstrated that although anxiety per se perturbs sleep, sleep disorder continued to be related
with GSM exposure after adjusting for these concerns. 2) It was showed that symptoms of the
microwave sickness were moderately related with this radiation, reinforcing that may exists some
underlying biological mechanism behind these symptoms. 3) This work presents a re-analysis of
our previous cross-sectional study [10], using a more robust statistical method.
Key Messages: 1) Existence of the microwave syndrome. 2) The symptoms were related to very
low GSM levels against Spanish and international guidelines [40], [41].
Strengths: We use a robust statistical analysis with a highly homogeneous sample in a
homogeneous environment. Limitations: We work with old data. The environmental radiation at
that time (2001) was mainly associated to GSM.
Page 3 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
4
4
Introduction
The health risk due to exposure to radiofrequency electromagnetic fields (RF EMF) continues to be
discussed today. The study that led to this debate was initiated after verification that the U.S.
embassy in Moscow was being subjected to such radiation from 1953 to May 1975 [1]. Recently, a
review of that episode [2] reopened the debate about the potential harmfulness of RF EMF. The
increasing number of base stations (BSs) on masts and buildings increased public awareness. This
issue promoted scientific research in order to know to what extent the low-intensity
electromagnetic fields might affect the health of humans and other organisms [3, 4]. Furthermore,
the term electromagnetic hypersensitivity (EHS) has been introduced recently, the term EHS was
related to subjects attributing symptoms to exposure to electromagnetic fields (EMFs) [5-8]. In
2010, a review about this topic [9], found that eight of the 10 studies evaluated through PubMed,
reported increased prevalence of adverse neurobehavioral symptoms or cancer in populations
living at distances < 500 meters from base stations.
None of the studies reported exposure above accepted international guidelines, suggesting that
current guidelines may be inadequate in protecting the health of human populations. Thus, emerges
the need to reevaluate our pioneering work in this field, trying to improve it, adding new
procedures and data. To our understanding, few articles have addressed the possible association
between microwave sickness and microwave exposure from GSM (Global System for Mobile
Communications) base stations (BSs) since the publication of our first study [10]. Chronologically,
Santini et al., [11] and Gadzicka et al., [12] reported differences in the distance-dependent
prevalence of symptoms such as headache, impaired concentration, and irritability. Later, a study
carried out in Austria [13] showed a positive association between the measured electric field
(GSM 900/1800) in bedrooms and headaches, cold hands and feet, and difficulties in
concentrating. An Egyptian study [14] showed a prevalence of neurological complaints, such as
Page 4 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
5
5
headache, memory changes, dizziness, tremors, depressive symptoms, and sleep disturbances
among subjects more directly exposed to GSM signals from BSs.
The symptoms reported by all the above cited authors belong to those attributed to the
microwave syndrome [15]. However, one article [16] using personal monitored data from GSM-
UMTS frequency bands did not find any statistical association in adults. More recently, the same
authors observed no association in children [17], contradictory results in children and adolescents
[18], and concluded that the few observed significant associations were not causal but rather
occurred by chance. Blettner et al. [19] reported in Phase 1 of their study more health problems
closer to base stations, but in Phase 2 [20] they concluded that measured EMF emissions were not
related to adverse health effects.
Other researchers focused their work on the possible existence of subjects with sensitivity to
GSM or UMTS signals according to psychological, cognitive, or autonomic assessment. These
researchers used short-term exposure (only 30-50 minutes) under laboratory conditions [21-23]
and revealed a large disparity between subjects. Recently, a study measuring several biological
stress markers [24] found that RF-EMF emitted by mobile phone base stations from 5.2 µW/m2 to
2126.8 µW/m2 increased cortisol and salivary alpha-amylase, while IgA concentration was not
significantly modified.
In 2010, the Selbitz study [25], described a significant dose-response relationship in symptoms
related with sleep, mood, joints, infections, skin condition, as well as neurological cardiovascular,
visual, and auditory systems, and the gastrointestinal tract.
In the work of Danker-Hopfe et al. [26] it was not evident the existence of short-term
physiological effects of EMF on sleep quality, however it was stated that the presence of BSs
per se (not the EMF) may have a negative impact on sleep quality.
In 2012, a Polish study did not present correlation between the electric field strength and the
frequency of subjective symptoms however showed a correlation between subjective symptoms
Page 5 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
6
6
and the distance to BSs, [27] . A study carried out in Egypt [28] revealed that exposure to EMF
emitted either from mobile phones or BSs had significant effects on pituitary-adrenal axis. More
recently, the work developed in Iran [29] indicated that symptoms such as nausea, headache,
dizziness, irritability, discomfort, nervousness, depression, sleep disturbance, memory loss and
lowering of libido were statistically significant in the inhabitants living near the BSs (<300m
distances) compared to those living far from the BSs (>300m).
In our cross-sectional analysis [10], 11 out of 16 symptoms showed statistically significant
higher scores in the group with the maximum exposure level. The symptoms are included in the
microwave syndrome. We also reported statistically significant correlation coefficients between
the measured electric field and fourteen out of sixteen symptoms.
A review [30] recently established several conditions for epidemiological studies to be eligible
for introduction in general analysis: Eligible studies must quantify exposure using objective
measures (such as distance to the nearest BS, spot, or personal exposure measurements in a
specific frequency range); possible confounders must be considered; and the selection of the
study population must be clearly free of bias in terms of exposure and outcomes.
Accordingly, in this re-analysis of our previous study [10], possible confounders were included in
addition to the specific RF EMF measurements made in 2001 (covering the specific range between
900 and 1800 MHz). So, we co-analyzed the effects of other variables such as socio-demographic
data and use of electronic devices. Lately, concern about being damaged by the radiation of the
antenna was also analyzed.
The new statistical approach tested the possible influences of other variables, such as demographic
data, and the use of electronic devices. Moreover, since some concerns have been raised about
possible health consequences caused by the emitted microwaves, we analysed if these symptoms
might be related to fear to the exposure. Also, since some subjects refused to allow measurements
at their homes, we analyzed if symptoms status or subjective distance to the BS, could be a bias of
Page 6 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
7
7
participation in the study. Interestingly, this period was free of other sources of RF such as WIFI or
UMTS etc. or the massive use of mobile phones, enabling a specific study of GSM technology.
Finally, the suitability of the size of the sample was analyzed.
Page 7 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
8
8
METHODS
Study design
We chose a small urban area with mixed rural characteristics, with low environmental pollution
(more agricultural than industrial), with no major differences in socio-economic characteristics
through the region (excluding big cities), same ethnicity (white Caucasian) and language (Spanish);
and having mobile phone communication for at least two years. Therefore, La Nora was chosen
because it met these requirements, common with the rest of this region of the south-east of Spain,
being a place with features of small city, near the capital (Murcia) and rural environment, without
any particular issue, such as a relevant health problem or particular environmental risk.
Consequently, La Ñora could be representative of small urban areas in east Spain with fewer than
20,000 inhabitants, and rural areas, which account for 19.8% of the population and 35.9% of the
territory.
Two BS masts, each about 30 m height were sited at different positions to provide GSM-900-1800
coverage. The GSM 900 BS was positioned not earlier than 1997 while the GSM 1800 BS was
built in December 1999.
The main demographic characteristics of the sample and the use of electronic devices was
recollected through a Spanish-language questionnaire [11]. All of the subjects were of the same
ethnic origin, shared similar family income levels and general standard of living, and were born in
La Ñora or nearby. All the residents in the study were living in the village before the erection of
both BSs. All of the residents were at home for more than eight hours a day for at least six days a
week and normally slept at home.
The core of the questionnaire was a symptom checklist for estimating the frequency of 15 health-
related symptoms attributed to microwave sickness. These symptoms are: fatigue; irritability;
headaches; nausea; loss of appetite; sleep disorders; depressive tendency; dizziness; concentration
difficulties; memory loss; skin lesions; visual and hearing deficiencies; walking difficulties; and
Page 8 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
9
9
cardiovascular problems. The frequency was quantified as: never suffer = 0; sometimes = 1; often
= 2; and very often = 3.
The percentage of residents who reported electrical transformers less than 10 m from their home
was 21.6%; while a 42% reported high voltage power lines less than 100 m from home. Finally,
40% of residents reported a TV transmitter within a radius of around 4 km.
The questionnaire included a statement that its purpose was health research, that the data gathered
would be confidential, and that the study was approved by the Ethical Committee of the University
of Valencia in accordance with the Declaration of Helsinki (http://www.wma.net/e/policy/b3.htm).
215 questionnaires were randomly distributed through seventeen streets which represented
practically the entire village. Election of the subjects' residences was performed using a Street Map
of the village. 150 were collected being the difference due of not being at home (31) or refusal to
fulfil the questionnaire (34).
Then, 101 RF EMF measurements in bedrooms were made in 2001. The missing (49) residents
were due to refusal to give admittance for taking measurements (16), residents not being at home
for the scheduled measurement appointment (10) and having serious health problems (23).
However, some changes are now being introduced in this re-analysis. 13 subjects included in the
original study have now been eliminated, two subjects (one because of alcohol abuse and another
because pregnancy) to increase the requirement on health criteria and eleven to increase the
homogeneity of the RF EMFs measurements, there was a change (it was raining) in the usual dry
weather conditions when their respective broadband measurements were registered.
The reanalysis of the dataset, which is the main focus of this paper, was finally performed with 88
subjects (45 female and 43 male) instead of the 101 analysed in 2001.
Page 9 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
10
10
Concerns about microwave exposure
66 subjects out of 88 were reached by telephone in February 2012 following two questions:
a) Were you worried about the masts (BSs) at the time when they were erected?
b) Did you believe its radiation (BSs) could damage your health?
In all cases those who were worried about the masts were concerned about health consequences. 27
subjects (40, 9 %) responded no and 39 (59, 1%) responded yes. Responses were analyzed related
to age (ANOVA test), sex (lambda statistic), and subjective distance to BS (Somers´d statistic).
Exposure assessment
Broadband measurements were made on two Saturdays in February and March 2001 from 11 am to
7pm with a portable electrical field (400 MHz – 3 GHz) detector (Nuova Elettronica Model LX-
1435). This meter was calibrated with an HP-8510C network analyser inside an anechoic chamber
at the University of Valencia. During the bedroom exposure assessment the electric field probe was
held for approximately five minutes about one meter from the walls and 1.2 meters above the
ground – and moved around a circle of 0.25 m radius, orientating the antenna in different directions
to obtain the maximum electrical field strength above the bed.
To check the intensity of TV and radio channels, as well as the intensity of working channels and
broadcast channels for the GSM-900-1800 BSs, measurements of the spectral power density were
carried out with a probe antenna and a portable spectrum analyser.
The probe was mounted on a linen phenolic tripod 1.2 m above the ground. The position of the
probe was the same on both days – on a hill next to the village and 20 m from the BS. With the
spectrum analyser we scanned the frequency bands and the levels were averaged for six minutes.
The measurement of the spectrum was similar on both days – with a difference in the peak
estimation (channel carriers) of about 1 dB.
The broadband measured exposure was almost invariable during the time interval of the
measurements. It changed with the position or place but it did not change in time, this could be
Page 10 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
11
11
related with a low intensity of the traffic channels (a very low number of phone calls) related
with the higher and constant intensity of the broadcast channel [10].
Statistical analysis
Demographical data was analysed using the Mann-Whitney one-way variance analysis and Chi
square test. Also, differences between groups were performed through the analysis of variance
(ANOVA), and the analysis of covariance (ANCOVA).
The main statistical analysis was made using binary logistic regression (mode enter) carried out on
a regular basis and subsequent bootstrapping [1000 bootstrap replications, 95% percentile method
and simple sampling] [31] that could provide more accurate standard errors and confidence
intervals. After producing (1000) bootstrap replicates θ b of an estimator θ, the bootstrap standard
error was the standard deviation of the bootstrap replicates:
SE (θ) = √ [∑ (θb - θ) **2 / (r-1)] b=1� r where θ is the mean of the θb. Because of our small sample size, a non-parametric confidence
interval for the estimate (mean) was constructed from the quartiles of the bootstrap sampling
distribution of θ. The 95% percentile interval (θ (lower) <θ< θ (upper) is shown, and where the θb
are the r-ordered bootstrap replicates: lower = 0.025 × r (sample 25) and upper = 0.975 × r (sample
975).
The 15 health-related symptoms described above constituted the dichotomous dependent variables.
Then, univariate analysis was performed for each symptom and for each of the predictors variables:
exposure to base station (µW/m2 as a natural logarithmic) and age were used as continuous
variables; while gender, computer use >2 h/ day, mobile phone use > 20 minutes per day and worry
about the antennae were used as dichotomous variables. The covariates with predictive value were
considered for the multivariate analysis. Thus possible confounder effects were evaluated.
Page 11 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
12
12
In all cases, change in -2 log likelihood, odds ratio, 95%-confidence intervals (95%-CI), as well as
the probability value (p-value) were calculated. For all tests, a p-value below 0.05 was considered
statistically significant.
The dependent variables (health-related symptoms) given in four ordinal categories (0=never,
1=sometimes, 2=often; 3=very often) were dichotomised (0,1 = 0 versus 2, 3 = 1).
We used the GSM exposure (the measurement of RF EMF in the bedroom) as a continuous
variable because it is well recognised that categorisation of continuous variables introduces major
problems in the analysis and interpretation of models derived in a data-dependent fashion [32, 33,
34].
We chose exposure values in the logarithmic form because these values are well grouped around
their median; while the raw values showed a high dispersion of values, with two outliers and ten
extreme values (data not shown).
Confounding was assessed by adding the potential confounding variable to the model and making a
subjective decision as to whether or not the coefficient of the variable of interest, ORs of GSM
exposure, had changed substantially. A 10 per cent variation was accepted as a considerable change
Possible interactions between covariates were also evaluated.
The maximum number of covariates included in each multivariate analysis was calculated
following this formula [35]. Let π be the smallest of the proportions of negative or positive cases
in the population and k the number of covariates, then the minimum number of cases to include
is:
N = 10 k / π
Goodness-of-fit tests such as the classification table, the Hosmer-Lemeshow statistic, ROC
curves, Cox and Snell’s, and Nagelkerke’s Pseudo R square measures were used. The Wald
statistic was also evaluated to test the significance of individual independent variables.
Moreover, possible multicolinearity was also tested.
Page 12 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
13
13
With the predicted probability scores derived from the regression analysis, receiver operating
characteristic (ROC) curves were constructed for all symptoms or modalities in order to analyze
sensitivity and specificity levels. For each curve, the best cut-offs for GSM exposure that
maximizes (sensitivity + specificity) were also calculated.
For statistical analysis we used the Statistical Package for Social Sciences, version 21.0 (IBM
SPSS Inc., Chicago, IL, USA) for Windows.
Because an exposure assessment for transformers, high voltage power lines, and radio or TV
transmitters based on self-estimated distances would not produce a reliable exposure estimate, it
was decided to omit these covariates in the analysis.
RESULTS
Demographical data and the percentage of users of personal computers and mobile phones were
analysed. The mean age was 42, 17 years (standard deviation ± 17, 61, interval 15–81), while 13,6
% used regularly computers (6,6 % were women and 7,0% men) and 23,9 % mobile phones (8,8 %
corresponded to women and 15,1 % to men). In the total sample, 51,1 % were women (mean age =
45.08 years (SD =17.98) ; [interval=15-81] and 48,9% men (39.12 (16.88) [15-75]. No differences
related with age, use of mobile phones or computers between sexes were found.
The univariate logistic regression indicated that age was inversely associated with irritability [OR =
0.97, 95%CI = 0.95-0.99] and that the oldest had the greatest difficulties hearing [OR= 1.03, 95%
CI=1.01-1.06] and walking [OR= 1.04, 95%CI = 1.01-1.07]. However, gender clearly did not
influence the outcome of any dependent variable. Use of mobile phones was linked with lack of
appetite and vertigo, while worry about the radiation from BSs was associated with sleep troubles
(Table 1). However, concerns about the radiation from BSs were not related to age (ANOVA test),
sex (lambda statistic) or subjective distance to BS (Somers´d statistic).
Page 13 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
14
14
Most of the symptoms were related with GSM exposure, especially: fatigue, irritability, lack of
appetite, sleep troubles, depression, and lack of concentration. Change in -2 log likelihood showed
similar results (Table 2).
ROC curves of each one of the logistic regression models (GSM exposure vs. each symptom)
oscillated between 0.65 and 0.87 (Table 3). Headaches (0, 84), nausea (0,86), appetite (0,87), and
vascular problems (0,85) showed the highest values, while memory (0,67), skin (0,67), and visual
disturbances (0,65) showed the lowest ones. The Hosmer and Lemeshow test indicated that most
analyses showed no significance p- values. The exceptions were fatigue (0,003), depression (0,003)
and vertigo (0,03). In the majority of the cases the models predicted better specificity than
sensitivity. Only in the case of headaches and sleep disorder, sensitivity prevailed over specificity
(Table 3 – classification table). In the extreme case, skin and vascular problems showed null or
minimum sensitivity, but 100 % specificity. Nagelkerke pseudo R-squared, showed acceptable
coefficients with the exception of the symptoms related with vertigo and skin problems (Table 3).
The lowest threshold cut-off values of GSM were for headaches, sleep, irritability and attention;
while vascular and skin problems showed the highest threshold cut-off values (Table 3).
The influence of other covariates on the GSM ORs coefficients, such as age, cellular use and
concern about the antennae were always less than 10% (Tables 2,).
There was no observed multicollinearity among variables. Kappa values according to factor
analysis were always lower than 2 and well below the critical value of 30.
Finally, no interactions between covariates were observed.
Standard errors and confidence intervals obtained by re-sampling were similar to those calculated
from the asymptotic approximation (Table 4). There was a small bias or difference between the
average bootstrap coefficients (not shown) and the respective estimates obtained from the
original sample.
Page 14 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
15
15
There were not health differences between those who permitted an exposure measurement (88 in
our last model) and those uninterested in RFs measures (26). ANOVA showed that the group with
RF EMFs measures were more prone to complain of memory loss ( F = 5,07 ; p = 0,027); while
subjects without EMF measures showed more skin problems ( F = 10,66; p = 0,001). There were
not health differences between those groups when using age as covariate. Square partial eta
measured 0 % contribution of the willingness participation variable to symptoms that correlated
with age such as: irritability, headaches, walking difficulties and hearing loss. Also, there was not
relationship between subjective distance to the BS and willingness to participate (Pearson Chi-
square = 2,80, df = 1, p=0,094).
DISCUSSION
In the present re-analysis a more robust statistical method was employed, indifferent to the
assumption of normality. In order to reduce the limitation of sample size effect and extrapolate
our results to the entire population from which the sample was obtained, re-sample method or
bootstrapping was used.
This new study partially confirms our preliminary results, being the most of the symptoms related
to GSM levels – independently of the demographical variables and some possible risk factors.
Related to microwave radiation, the spectral power density analysis maintained that the most
important contribution to broadband measurements was from GSM 900/1800, and the main
variability of the measurements between different places was due to a different coverage of the
GSM 900/1800 signals, i.e. spatial variability. This was further supported by the fact that the
antenna used was fairly insensitive to frequencies below 400 MHz. Therefore, the radio channels
80-110 MHz were not a significant part of the broadband measurements. Moreover, the narrow
band measurements showed TV channels with substantially lower intensities than the GSM
900/1800 signals. The effects from these exposures will therefore not confound the effects of BSs.
Page 15 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
16
16
Moreover, some authors [13] found that the only relevant contribution to the variance of the high
microwave exposure was from BSs – up to 93% of variance. Moreover, at the time of our study the
GSM signal was almost invariable in time because there were very few calls. The main
contribution was made from the broadcast channels working almost constantly throughout the day.
Short-range evaluations of exposure could be acceptable for describing a 24-hour period and the
measurements were made in bedrooms – a location where the subjects were assumed to spend
significant periods of time.
However, some subjects were mobile phone users at the time of this study and exposure from a
mobile phone during a phone call is much higher than that received from BSs. Nevertheless, some
authors [13] stated about that there is no a priori argument why these lower levels should have no
effect on the presence of a widespread use of mobile telephones. Exposure from a BS will be at a
low but almost constant level for many hours a day and especially at night.
While GSM exposure was associated with most of the symptoms, walking difficulties and hearing
loss was correlated only with age. Also, age remained slightly inversely associated with irritability.
Those who were users of cellular phones were more prone to complain of loss of appetite and
vertigo, while those who expressed worry about the BSs, were associated with having sleep
problems. This later finding was in concordance with two other articles [13, 20, 26]. However,
worry about the BSs was unrelated with age, gender or subjective distance to BSs. This agrees
with an article [36] claiming that there was no statistically significant association between
symptom occurrence associated with perceived proximity to BSs, psychological components,
socio-demographic characteristics, and actual distance to BSs or powerlines.
Some authors indicated that opponents of mobile phone towers generally do not express anxieties
about electromagnetic field exposure, indicating that the risk rating is comparable with other
perceived common hazards in the modern world [37].
Page 16 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
17
17
None of the analysed covariates behaved as confounders. So, the relationship between GSM
exposure with irritability, sleep troubles, lack of appetite and vertigo remained statistically
significant despite the introduction of the above covariates.
When the conventional multivariate analysis was tested using bootstrapping it was observed that
the standard error and confidence intervals obtained by re-sampling were similar to those
calculated from asymptotic approximation and this supports the adequacy of our conventional
analysis. Our sample, chosen at random, would qualify to represent the population from which it
came.
The model appeared generally well adjusted while the cut-off values could constitute good
guidance for predicting the threshold of symptom appearance.
We cannot truly rule out that their inhabitants were more; equal or less worried than elsewhere in
this region, since we cannot provide the percentage of worriers about the BS masts in La Ñora
compared with other places of its environment. However, information about this issue spread to
all over this region at this time and La Ñora circumstances are common with the rest of small
urban and rural areas. The sample was randomly selected but a participation bias cannot be rule
out since the most part of our participants expressed fear for the existence of the BSs and this could
contribute to their participation in the study. Also, it is also possible to speculate that the percentage
of the subjects who refused to participate were for opposite reasons (indifference about the BSs). In
this regard, neither health status nor subjective distance to the BS explained willingness to
participate in the study.
Concerns about the radiation from BSs were not related to age, sex, or subjective distance to BSs.
This agrees with statements from some authors [13] that living near a base station does not make
more people generally fearful, but people that generally worry about the fields would express
stronger fears when they live close by.
Page 17 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
18
18
Nevertheless, irrespective of these explanations, there seem to be effects of exposure that occur
independent of the fear of the subjects, since controlling for fear did not change the association
between exposure and symptoms. However, biological grounds explaining non-thermal effects
have not been clearly established. Recently, it has been described that voltage-gated calcium
channels are essential to the beneficial or adverse responses to microwave EMFs, nanosecond EMF
pulses, and static electrical and magnetic fields [38].
In summary, the results of this study indicate that effects of very low but long lasting exposures to
emissions from mobile telephone base-stations on well-being cannot be ruled out. The effects
fulfilled almost completely the symptoms described within the microwave syndrome. Finally,
unraveling the causal pathways would be better done with experimental study design.
Page 18 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
19
19
CONCLUSIONS
This new study partially confirms our preliminary results about microwave sickness resulting from
exposure to emissions from GSM mobile phone BSs. So, fatigue, irritability, lack of appetite, sleep
troubles, depression, and lack of concentration were especially related with GSM exposure.
These results were independent of the main socio-demographical variables, other EMF exposures
and the worry about being irradiated. Nevertheless, we confirm that apprehension about modern
technology could predict some symptom, specially related with sleep problems.
Our results agree with those who claimed that by distorting perceptions of risk, disproportionate
precaution might paradoxically lead to illness that would not otherwise occur [39]. However,
health changes related with GSM exposure seems to occur unrelated with those fears. Finally, this
exposure was very low in that period and also very low against Spanish recommendations [40] and
international guidelines [41].
Recommendations
We subscribe the guidelines observed by other authors [42] to follow the principle of prevention
as long as the non-thermal effects are not considered in any official standard. This includes
exposure minimization in the frame of the technical feasibility to guarantee a significantly
reduction in the long term radiation exposures of cellular phone towers in residential areas.
Epidemiological and clinical studies should be continuing to observe possible health changes in
the population. Finally, clear information about the correct use of the newer electronic devices
should be implemented.
Authors' contributions
CGP was one of the principle researchers responsible for design, conduct, and writing the
manuscript. CGP also made the main statistical contributions. EAN was also one of the main
Page 19 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
20
20
researchers responsible for design and acquisition of RF EMF data, writing, and final review of
manuscript. JS contributed mainly to processing of RF EMF data, while MP was responsible for
design and final review of manuscript.
Page 20 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
21
21
ACKNOWLEDGEMENTS
We would like to express our gratitude to Angeles Martinez Gomez for her assistance during the
field work in La Ñora; as well as the Spanish Ministry of Science and Technology for grant FIT
number 070000-2002-58.
Authors' contributions
CGP was one of the principle researchers responsible for design, conduct, and writing the
manuscript. CGP also made the main statistical contributions. EAN was also one of the main
researchers responsible for design and acquisition of RF EMF data, writing, and final review of
manuscript. JS contributed mainly to processing of RF EMF data, while MP was responsible for
design and final review of manuscript.
Funding
This work was supported by Spanish Ministry of Science and Technology for grant FIT number
070000-2002-58.
Competing Interest
There are no competing interests
Page 21 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
22
22
REFERENCES
1. Lilienfeld AM, Tonascia J, Tonascia S, Libauer CH, et al. Foreign Service Health Status
Study, Evaluation of health status of foreign service and other employees from selected
Eastern European posts. Baltimore: Dept Epidemiology, School of Hygiene and Public
Health, The Johns Hopkins University; 1978.
2. Elwood JM. Microwaves in the cold war: the Moscow embassy study and its interpretation.
Review of a retrospective cohort study. Environ Health 2012; Nov 14;11:8.
3. Belyaev IY. Non-thermal Biological Effects of Microwaves. Microwave Review. 2005a;
11:13-29
4. Belyaev IY. Non-thermal Biological Effects of Microwaves: Current Knowledge,
Further Perspective, and Urgent Needs. Electromagn. Biol. Med. 2005b ; 24:375-403.
5. Röösli M, Moser M, Baldinini Y, Meier M, Braun-Fahrländer C. Symptoms of ill health
ascribed to electromagnetic field exposure—a questionnaire survey. Int. J. Hyg. Environ.
Health. 2004 ; 207:141–150.
6. Frick U, Mayer M, Hauser S, Binder H , Rosner R , Eichhammer P. Entwicklung eines
deutschprachigen Messinsntrumentes für ‘‘Elektrosmog-Beschwerden’’. Umwelt Med.
Forsch. Prax. 2006 ; 11:103–113.
7. Eltiti S, Wallace D, Ridgewell A, Zougkou K, Russo R, Sepulveda F, et al. Does short-term
exposure to mobile phone base station signals increase symptoms in individuals who report
sensitivity to electromagnetic fields? A double-blind randomized provocation study. Environ
Health Perspect 2007;115:1603-1608.
8. Eltiti S, Wallace D, Ridgewell A, Zougkou K, et al. Short-term exposure to mobile phone
base station signals does not affect cognitive functioning or physiological measures in
Page 22 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
23
23
individuals who report sensitivity to electromagnetic fields and controls. Bioelectromagnetics
2009;30:556-63.
9. Khurana VG, Hardell L, Everaert J, Bortkiewicz A, et al. Epidemiological evidence for a
health risk from mobile phone base stations. Int J Occup Environ Health 2010; 16(3):263-7.
10. Navarro EA, Segura J, Portolés M, Gómez-Perretta C. The Microwave Syndrome: A
Preliminary Study in Spain. Electromagn Biol Med 2003; 22:161 – 169.
11. Santini R, Santini P, Danze J, Le Ruz P, et al. Study of the health of people living in the
vicinity of mobile phone base stations: I. Influences of distance and sex. Pathol Biol 2002;
50:369 – 373.
12. Gadzicka E, Bortkiewicz A, Zmyslony M, Szymczak W, et al.. Assessment of subjective
complaints reported by people living near mobile phone base stations [Abstract].
BiuletynPTZE Warszawa. 2006; 14:23-26.
13. Hutter H, Moshammer H, Wallner P, Kundi M. Subjective symptoms, sleeping problems,
and cognitive performance in subjects living near mobile phone base stations. Occup Environ
Med 2006; 63:307–313.
14. Abdel-Rassoul G, El-Fateh OA, Salem MA, Michael A, Farahat F, et al. Neurobehavioral
effects among inhabitants around mobile phone base stations. Neurotoxicology 2007;
28:434-440.
15. Johnson-Liakouris AG. Radiofrequency (RF) Sickness in the Lilienfeld Study: an effect of
modulated microwaves? Arch Environ Health 1998;53:236-238.
16. Thomas S, Kühnlein A, Heinrich S, Praml G, et al. Personal exposure to mobile phone
frequencies and well-being in adults: a cross-sectional study based on dosimetry.
Bioelectromagnetics 2008; 29:463-470.
Page 23 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
24
24
17. Kühnlein A, Heumann C, Thomas S, Heinrich S, et al. Personal exposure to mobile
communication networks and well-being in children--a statistical analysis based on a
functional approach. Bioelectromagnetics 2009; 30:261-9.
18. Heinrich S, Thomas S, Heumann C, von Kries R, et al. Association between exposure to
radiofrequency electromagnetic fields assessed by dosimetry and acute symptoms in children
and adolescents: a population based cross-sectional study. Environ Health 2010; 9:75.
19. Blettner M, Schlehofer B, Breckenkamp J, Kowall B, et al Mobile phonebase stations and
adverse health effects: Phase 1 of a populationbased,cross-sectional study in Germany. Occup
Environ Med.2009;66:118-123.
20. Berg-Beckhoff G, Blettner M, Kowall B, Breckenkamp J, et al. Mobile phone base stations
and adverse health effects: phase 2 of a cross-sectional study with measured radio frequency
electromagnetic fields. Occup Environ Med 2009; 66:124-130.
21. Augner C, Florian M, Pauser G, Oberfeld G, et al. GSM Base Stations: Short-Term Effects
on Well-Being. Bioelectromagnetics 2009; 30:73-80.
22. Furubayashi T, Ushiyama A, Terao Y, Mizuno Y, et al. Effects of short-term W-CDMA
mobile phone base station exposure on women with or without mobile phone related
symptoms. Bioelectromagnetics 2009; 30:100-113.
23. Zwamborn AP, Vossen SH, van Leersum BJ. Effects of global communication system
radio-frequency fields on well being and cognitive functions of human subjects with and
without subjective complaints. Netherlands Organisation for Applied Scientific Research
(TNO) 2003;148:1-89.
24. Augner C, Hacker GW, Oberfeld G, Florian M, et al. Effects of exposure to GSM mobile
phone base station signals on salivary cortisol, alpha-amylase, and immunoglobulin A.
Biomed Environ Sci 2010;23:199-207.
Page 24 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
25
25
25. Eger H, Jahn M. Spezifische Symptome und Mobilfunkstrahlung in Selbitz (Bayern)
Evidenz für eine Dosiswirkungsbeziehung. Umwelt·medizin·gesellschaft 2010; 2: 130-139.
26. Danker-Hopfe H, Dorn H, Bornkessel C, Sauter C. Do mobile phone base stations affect
sleep of residents? Results from an experimental double-blind sham-controlled field study.
Am J Hum Biol. 2010; 5:613-8
27. Bortkiewicz A, Gadzicka E, Szyjkowska A, et al. Subjective complaints of people living
near mobile phone base stations in Poland. Int J Occup Med Environ Health 2012;25:31-40
28. Eskander EF, Estefan SF, Abd-Rabou AA. How does long term exposure to base stations
and mobile phones affect human hormone profiles? Clin Biochem 2012 ; 45(1-2):157-61
29. Shahbazi-Gahrouei D, Karbalae M, Moradi HA, et al. Health effects of living near mobile
phone base transceiver station (BTS) antennae: a report from Isfahan, Iran. Electromagn Biol
Med. 2013 Jun 19.
30. Röösli M, Frei P, Mohler E, Hug K. Systematic review on the health effects of exposure to
radiofrequency electromagnetic fields from mobile phone base stations. Bull World Health
Organ 2010; 88: 887–896.
31. Efron B, Tibshirani R: An Introduction to the Bootstrap. Boca Raton, FL: Chapman &
Hall/CRC 1993; ISBN 0-412-04231-2.software
32. Maxwell SE, Delaney HD. Bivariate median-splits and spurious statistical significance.
Psychological Bulletin 1993; 113:181–190.
33. Royston P, Altman DG, Sauerbrei W.Dichotomizing continuous predictors in multiple
regression: a bad idea. Stat. Med 2006; 25:127-141.
34. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous confounding
variable is categorized in logistic regression analyses. Stat. Med 2004; 23:1159–1178.
35. Peduzzi P, Concato J, Kemper E, Holford TR, et al. A simulation study of the number of
events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373-1379.
Page 25 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
26
26
36. Baliatsas C, van Kamp I, Kelfkens G, et al. Non-specific physical symptoms in relation to
actual and perceived proximity to mobile phone base stations and powerlines. BMC Public
Health 2011; 11:421
37. Hutter H, Moshammer H, Wallner P, Kundi M. Public perception of risk concerning
celltowers and mobile phones. Soz Präventivmed 2004;49: 62–66.
38. Pall ML. Electromagnetic fields act via activation of voltage-gated calcium channels to
produce beneficial or adverse effects. J. Cell. Mol. Med 2013; 7:. 958-965
39. Coggon D. Health Risks from Mobile Phone Base Stations. Occup Environ Med 2006;
63:298–299.
40. Real Decreto 1066/2001 Telecomunicaciones/TelefoniaMovil/RD1066_2001.pdf
http://www.mityc.es/esES/OficinaVirtual/Documents/SE%20
41. International Commission on Non-Ionizing Radiation Protection: Guidelines for limiting
exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz).
Health Phys 1998; 74: 494–522.
42. Haumann T, Münzenberg U, Maes W, Sierck P. HF-Radiation levels of GSM cellular
phone towers in residential areas. 2nd International Workshop on Biological effects of EMFS.
October 2002; Rhodes (Greece) Vol 1:327-333.
Page 26 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
27
27
Table 1. Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Symptom/
Variable
Worry about BSs (1) Computer (not use) Mobile (not use) OR (95%CI) OR (95%CI) OR (95%CI)
Fatigue 0.67 0.23-1.90 2.62 0.76-9.04 1.56 0.56-4.35
Irritability 1.13 0.43-3.03 1.56 0.45-5.34 2.62 0.94-7.33
Headaches 1.75 0.62-4.94 1.39 0.34-5.58 1.56 0.51-4.83
Nausea 0.68 0.18-2.24 0.34 0.04-2.84 1.43 0.44-4.67
Lack of appetite 1.05 0.33-3.40 3.16 0.87-11.44 4.28** 1.43-12.78
Trouble sleeping 3.12* 1.10-8.86 0.55 0.16-1.88 0.74 0.27-2.02
Depression 1.06 0.39-2.93 0.81 0.22-2.93 1.03 0.38-2.84
Lack of concentration 0.92 0.35-2.47 1.11 0.33-3.76 2.79 0.99-7.80
Memory loss 1.71 0.62-4.75 0.41 0.10-1.64 1.35 0.50-3.61
Skin alterations 0.74 0.23-2.35 φ φ 0.63 0.16-2.45
Visual disturbances 1.31 0.48-3.60 0.77 0.21-2.77 1.63 0.60-4.39
Vertigo 0.61 0.20-1.91 0.77 0.19-3.10 2.90* 1.04-8.07
Vascular alteration 0.96 0.27-3.43 1.48 0.35-6.17 2.04 0.65-6.41
Hearing problems 0.59 0.20-1.70 0.77 0.19-3.10 0.48 0.15-1.60
Walking difficulty 0.60 0.20-1.79 φ φ 0.42 0.11-1.60 * p<0.05; ** p<0.01; (1) Not worried, as reference code (2) & (3) Not device use, as reference code,φ any subject affected using computer
Page 27 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
28
28
Table 2. ORs & 95% CIs for GSM exposure: increase in risk per increase in ln GSM (µW/m2). Change in -2 log likelihood is also shown.
Symptom
OR (95%CI) Change in
-2 Log likelihood
Fatigue 1.39***(1.14-1.70) 11.74***
Irritability 1.51***(1.23-1.85) 19.36***
Irritability (adjusted with age) 1.47***(1.20-1.81) -
Headaches 1.43** (1.15-1.78) 12.32***
Nausea 1.38** (1.09-1.73) 8.3**
Lack of Appetite 1.58** (1.23-2.03) 16.31***
Lack of Appetite (adjusted to use of cellular ) 1.53** (1.19-1.99) -
Trouble sleeping 1.49***(1.20-1.84) 16.38***
Trouble sleeping (adjusted to worry to BSs) 1.64***(1.22-2.19) -
Depression 1.41***(1.16-1.72) 13.99***
Concentration 1.54***(1.25- 1.89) 20.75***
Memory 1.27** (1.06- 1.52) 7.29**
Skin 1.24* (1.001-1.54) 4.08*
Visual 1.23 * (1.03-1.46) 5.30*
Vertigo 1.36** (1.11-1.66) 10.14***
Vertigo (adjusted to use of cellular ) 1.32** (1.08-1.62) -
Vascular 1.32* (1.05-1.64) 6.30*
Hearing 0.96 (0.80-1.15)
Walking 0.95 (0.78-1.15)
* p<0.05; ** p<0.01; *** p<0.001
Page 28 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
29
29
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (ln). Cut-off values of exposure to microwaves according to ROC analysis. The data are presented in ascending order. Symptom ROC curves
Area
Classification Table (a)
Pseudo-R**2
(1)
Cut-off (2) (Ln GSM)
Cut-off (2) GSM (µW/m2)
SSV SPF AV
Headaches 0.84*** 0.90 0.23 0.72 0.41 0.57 1.77 Sleep 0.78*** 0.82 0.66 0.76 0.28 1.66 5.26
Attention 0.78*** 0.67 0.72 0.69 0.28 3.61 36.97 Irritability 0.76*** 0.67 0.73 0.71 0.26 3.61 36.97 Memory 0.67** 0.54 0.77 0.67 0.11 4.99 146.94
Depression 0.75*** 0.46 0.76 0.65 0.20 5.22 184.93 Visual 0.65* 0.24 0.83 0.60 0.08 5.91 368.71 Fatigue 0.73*** 0.22 0.90 0.69 0.18 6.53 685.4 Vertigo 0.74*** 0.16 0.87 0.67 0.19 6.53 685.4 Appetite 0.87*** 0.40 0.94 0.85 0.43 7.31 1495.18 Nausea 0.86*** 0.46 0.93 0.87 0.38 7.31 1495.18
Vascular 0.85*** 0.20 1.0 0.90 0.34 8.02 3041.18 Skin 0.67* 0.00 1.00 0.81 0.072 9.06 8604.15
* p<0.05, ** p<0.01, *** p<0.001 (1) Nagelkerke (2) Cut-off (ROC curve) (a) SSV=sensitivity, SPF =specificity, AV=average
Page 29 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
30
30
Table 4. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients: Increase in risk per increase in ln GSM (µW/m2). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Symptom B*
Bootstrap Normal
Bias SE 95% percentile
intervals SE 95% CI
Lower Upper Lower Upper Fatigue 0.329 0.012 0.097 0.155 0.539 0.102 0.128 0.529 Irritability 0.411 0.016 0.110 0.241 0.670 0.104 0.207 0.615 Headache 0.358 0.022 0.139 0.149 0.688 0.113 0.137 0.578 Nausea 0.319 0.013 0.124 0.099 0.590 0.118 0.088 0.550 Appetite 0.456 0.026 0.134 0.264 0.784 0.128 0.205 0.707 Sleep 0.396 0.022 0.124 0.193 0.690 0.109 0.181 0.610 Depression 0.346 0.012 0.102 0.174 0.583 0.100 0.151 0.541 Attention 0.429 0.020 0.118 0.254 0.711 0.106 0.222 0.636 Memory 0.237 0.009 0.098 0.057 0.448 0.091 0.058 0.415 Skin 0.217 0.008 0.110 0.011 0.451 0.110 0.001 0.433 Visual 0.203 0.004 0.093 0.037 0.398 0.090 0.026 0.379 Hearing -0.05 -0.002 0.089 -0.219 0.143 0.093 -0.228 0.135 Vertigo 0.306 0.010 0.101 0.127 0.530 0.102 0.107 0.505 Walking -0.05 -0.006 0.098 -0.265 0.120 0.098 -0.246 0.138 Vascular 0.274 0.010 0.109 0.084 0.520 0.114 0.051 0.497 * beta coefficent (log OR)
Page 30 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
31
31
Page 31 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
1
1
Subjective Symptoms Related to GSM Radiation from Mobile Phone
Base Stations: A cross-sectional study
Corresponding autor:
Enrique A. Navarro
Departamento de Física Aplicada, Universitat de València
C/. Dr. Moliner 50,
46100 Burjassot, Valencia, Spain
Email: [email protected]
Authors
Claudio Gómez-Perretta (1)
Enrique A. Navarro (2)
Jaume Segura (3)
Manuel Portolés (1)
(1) Research Center, Universitary Hospital La Fe, Avda Campanar s/n, 46009 Valencia, Spain
(2) Departament of Applied Physics, Universitat de València, C/. Dr. Moliner 50, 46100 Burjassot
(València), Spain.
(3) Departament of Computer Sciences, ETSE- Universitat de València, Avda Universitat s/n,
46100 Burjassot (València), Spain.
Running title
Subjective Symptoms Related to GSM Radiation
Keywords: Public health; GSM; Microwave Sickness.
Page 32 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
2
2
Abstract
Objectives: This paper is a cross-sectional study, a reanalysis of the data from Navarro et al,
2003 (10), in which health symptoms related to microwave exposure from mobile phone base
stations (BS) were explored. We analyze if these symptoms might be related to fear to the
exposure, and therefore having a subjective basis.
Method: Since weather circumstances could influence the exposure, we restricted the data to
those measured under similar weather conditions. The participants with known illness in 2003 were
at this time disregarded: 88 subjects instead of 101 (in 2003) were analyzed. A statistical method
indifferent to the assumption of normality was now employed: Binary logistic regression for
modelling a binary response, e.g. having fatigue (1) or not (0), introducing the exposure as a
predictor variable. This analysis was carried out on a regular basis and bootstrapping [95%
percentile method] to provide more accurate confidence intervals.
Results: The symptoms more related with exposure were: Lack of appetite [odds ratio (OR)] =
1.6, 95% confidence interval (95%CI) = 1.2-2.0, lack of concentration [OR = 1.5, 95% CI = 1.25-
1.89], irritability [OR = 1.51, 95% CI = 1.23-1.85] and sleep troubles [OR = 1.5, 95% CI = 1.2-1.8].
Change in -2 log likelihood showed similar results. Concerns about the BS were strongly related
with sleep troubles [OR = 3.1, 95% CI = 1.1-8.9]. The exposure variable remained statistically
significant in the multivariate analysis. The bootstrapped values were similar to the asymptotic
confidence intervals
Conclusion: This study confirms our preliminary results. We observe that incidence of most of
the symptoms was related to exposure levels – independently of the demographical variables and
some possible risk factors. Also, concern about adverse effects from exposure, despite it is strongly
related with sleep disturbances, does not influence the direct association between exposure and
sleep.
Page 33 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
3
3
Authors' contributions
CGP was one of the principle researchers responsible for design, conduct, and writing the
manuscript. CGP also made the main statistical contributions. EAN was also one of the main
researchers responsible for design and acquisition of RF EMF data, writing, and final review of
manuscript. JS contributed mainly to processing of RF EMF data, while MP was responsible for
design and final review of manuscript.
ARTICLE SUMMARY:
Article Focus: 1) It was stated by some authors that microwave syndrome would be related to
anxiety associated to fear and visual perception of cellular phone towers. In this manuscript we
demonstrated that although anxiety per se perturbs sleep, sleep disorder continued to be related
with GSM exposure after adjusting for these concerns. 2) It was showed that symptoms of the
microwave sickness were moderately related with this radiation, reinforcing that may exists some
underlying biological mechanism behind these symptoms. 3) This work presents a re-analysis of
our previous cross-sectional study [10], using a more robust statistical method.
Key Messages: 1) Existence of the microwave syndrome. 2) The symptoms were related to very
low GSM levels against Spanish and international guidelines [40], [41].
Strengths: We use a robust statistical analysis with a highly homogeneous sample in a
homogeneous environment. Limitations: We work with old data. The environmental radiation at
that time (2001) was mainly associated to GSM.
Page 34 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
4
4
The health risk due to exposure to radiofrequency electromagnetic fields (RF EMF) continues to be
discussed today. The study that led to this debate was initiated after verification that the U.S.
embassy in Moscow was being subjected to such radiation from 1953 to May 1975 [1]. Recently, a
review of that episode [2] reopened the debate about the potential harmfulness of RF EMF. The
increasing number of base stations (BSs) on masts and buildings increased public awareness. This
issue promoted scientific research in order to know to what extent the low-intensity
electromagnetic fields might affect the health of humans and other organisms [3, 4]. Furthermore,
the term electromagnetic hypersensitivity (EHS) has been introduced recently, the term EHS was
related to subjects attributing symptoms to exposure to electromagnetic fields (EMFs) [5-8]. In
2010, a review about this topic [9], found that eight of the 10 studies evaluated through PubMed,
reported increased prevalence of adverse neurobehavioral symptoms or cancer in populations
living at distances < 500 meters from base stations.
None of the studies reported exposure above accepted international guidelines, suggesting that
current guidelines may be inadequate in protecting the health of human populations. Thus, emerges
the need to reevaluate our pioneering work in this field, trying to improve it, adding new
procedures and data. To our understanding, few articles have addressed the possible association
between microwave sickness and microwave exposure from GSM (Global System for Mobile
Communications) base stations (BSs) since the publication of our first study [10]. Chronologically,
Santini et al., [11] and Gadzicka et al., [12] reported differences in the distance-dependent
prevalence of symptoms such as headache, impaired concentration, and irritability. Later, a study
carried out in Austria [13] showed a positive association between the measured electric field
(GSM 900/1800) in bedrooms and headaches, cold hands and feet, and difficulties in
concentrating. An Egyptian study [14] showed a prevalence of neurological complaints, such as
Page 35 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
5
5
headache, memory changes, dizziness, tremors, depressive symptoms, and sleep disturbances
among subjects more directly exposed to GSM signals from BSs.
The symptoms reported by all the above cited authors belong to those attributed to the
microwave syndrome [15]. However, one article [16] using personal monitored data from GSM-
UMTS frequency bands did not find any statistical association in adults. More recently, the same
authors observed no association in children [17], contradictory results in children and adolescents
[18], and concluded that the few observed significant associations were not causal but rather
occurred by chance. Blettner et al. [19] reported in Phase 1 of their study more health problems
closer to base stations, but in Phase 2 [20] they concluded that measured EMF emissions were not
related to adverse health effects.
Other researchers focused their work on the possible existence of subjects with sensitivity to
GSM or UMTS signals according to psychological, cognitive, or autonomic assessment. These
researchers used short-term exposure (only 30-50 minutes) under laboratory conditions [21-23]
and revealed a large disparity between subjects. Recently, a study measuring several biological
stress markers [24] found that RF-EMF emitted by mobile phone base stations from 5.2 µW/m2 to
2126.8 µW/m2 increased cortisol and salivary alpha-amylase, while IgA concentration was not
significantly modified.
In 2010, the Selbitz study [25], described a significant dose-response relationship in symptoms
related with sleep, mood, joints, infections, skin condition, as well as neurological cardiovascular,
visual, and auditory systems, and the gastrointestinal tract.
In the work of Danker-Hopfe et al. [26] it was not evident the existence of short-term
physiological effects of EMF on sleep quality, however it was stated that the presence of BSs
per se (not the EMF) may have a negative impact on sleep quality.
In 2012, a Polish study did not present correlation between the electric field strength and the
frequency of subjective symptoms however showed a correlation between subjective symptoms
Page 36 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
6
6
and the distance to BSs, [27] . A study carried out in Egypt [28] revealed that exposure to EMF
emitted either from mobile phones or BSs had significant effects on pituitary-adrenal axis. More
recently, the work developed in Iran [29] indicated that symptoms such as nausea, headache,
dizziness, irritability, discomfort, nervousness, depression, sleep disturbance, memory loss and
lowering of libido were statistically significant in the inhabitants living near the BSs (<300m
distances) compared to those living far from the BSs (>300m).
In our cross-sectional analysis [10], 11 out of 16 symptoms showed statistically significant
higher scores in the group with the maximum exposure level. The symptoms are included in the
microwave syndrome. We also reported statistically significant correlation coefficients between
the measured electric field and fourteen out of sixteen symptoms.
A review [30] recently established several conditions for epidemiological studies to be eligible
for introduction in general analysis: Eligible studies must quantify exposure using objective
measures (such as distance to the nearest BS, spot, or personal exposure measurements in a
specific frequency range); possible confounders must be considered; and the selection of the
study population must be clearly free of bias in terms of exposure and outcomes.
Accordingly, in this re-analysis of our previous study [10], possible confounders were included in
addition to the specific RF EMF measurements made in 2001 (covering the specific range between
900 and 1800 MHz). So, we co-analyzed the effects of other variables such as socio-demographic
data and use of electronic devices. Lately, concern about being damaged by the radiation of the
antenna was also analyzed.
The new statistical approach tested the possible influences of other variables, such as demographic
data, and the use of electronic devices. Moreover, since some concerns have been raised about
possible health consequences caused by the emitted microwaves, we analysed if these symptoms
might be related to fear to the exposure. Also, since some subjects refused to allow measurements
at their homes, we analyzed if symptoms status or subjective distance to the BS, could be a bias of
Page 37 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
7
7
participation in the study. Interestingly, this period was free of other sources of RF such as WIFI or
UMTS etc. or the massive use of mobile phones, enabling a specific study of GSM technology.
Finally, the suitability of the size of the sample was analyzed.
Page 38 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
8
8
METHODS
Study design
We chose a small urban area with mixed rural characteristics, with low environmental pollution
(more agricultural than industrial), with no major differences in socio-economic characteristics
through the region (excluding big cities), same ethnicity (white Caucasian) and language (Spanish);
and having mobile phone communication for at least two years. Therefore, La Nora was chosen
because it met these requirements, common with the rest of this region of the south-east of Spain,
being a place with features of small city, near the capital (Murcia) and rural environment, without
any particular issue, such as a relevant health problem or particular environmental risk.
Consequently, La Ñora could be representative of small urban areas in east Spain with fewer than
20,000 inhabitants, and rural areas, which account for 19.8% of the population and 35.9% of the
territory.
Two BS masts, each about 30 m height were sited at different positions to provide GSM-900-1800
coverage. The GSM 900 BS was positioned not earlier than 1997 while the GSM 1800 BS was
built in December 1999.
The main demographic characteristics of the sample and the use of electronic devices was
recollected through a Spanish-language questionnaire [11]. All of the subjects were of the same
ethnic origin, shared similar family income levels and general standard of living, and were born in
La Ñora or nearby. All the residents in the study were living in the village before the erection of
both BSs. All of the residents were at home for more than eight hours a day for at least six days a
week and normally slept at home.
The core of the questionnaire was a symptom checklist for estimating the frequency of 15 health-
related symptoms attributed to microwave sickness. These symptoms are: fatigue; irritability;
headaches; nausea; loss of appetite; sleep disorders; depressive tendency; dizziness; concentration
difficulties; memory loss; skin lesions; visual and hearing deficiencies; walking difficulties; and
Page 39 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
9
9
cardiovascular problems. The frequency was quantified as: never suffer = 0; sometimes = 1; often
= 2; and very often = 3.
The percentage of residents who reported electrical transformers less than 10 m from their home
was 21.6%; while a 42% reported high voltage power lines less than 100 m from home. Finally,
40% of residents reported a TV transmitter within a radius of around 4 km.
The questionnaire included a statement that its purpose was health research, that the data gathered
would be confidential, and that the study was approved by the Ethical Committee of the University
of Valencia in accordance with the Declaration of Helsinki (http://www.wma.net/e/policy/b3.htm).
215 questionnaires were randomly distributed through seventeen streets which represented
practically the entire village. Election of the subjects' residences was performed using a Street Map
of the village. 150 were collected being the difference due of not being at home (31) or refusal to
fulfil the questionnaire (34).
Then, 101 RF EMF measurements in bedrooms were made in 2001. The missing (49) residents
were due to refusal to give admittance for taking measurements (16), residents not being at home
for the scheduled measurement appointment (10) and having serious health problems (23).
However, some changes are now being introduced in this re-analysis. 13 subjects included in the
original study have now been eliminated, two subjects (one because of alcohol abuse and another
because pregnancy) to increase the requirement on health criteria and eleven to increase the
homogeneity of the RF EMFs measurements, there was a change (it was raining) in the usual dry
weather conditions when their respective broadband measurements were registered.
The reanalysis of the dataset, which is the main focus of this paper, was finally performed with 88
subjects (45 female and 43 male) instead of the 101 analysed in 2001.
Page 40 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
10
10
Concerns about microwave exposure
66 subjects out of 88 were reached by telephone in February 2012 following two questions:
a) Were you worried about the masts (BSs) at the time when they were erected?
b) Did you believe its radiation (BSs) could damage your health?
In all cases those who were worried about the masts were concerned about health consequences. 27
subjects (40, 9 %) responded no and 39 (59, 1%) responded yes. Responses were analyzed related
to age (ANOVA test), sex (lambda statistic), and subjective distance to BS (Somers´d statistic).
Exposure assessment
Broadband measurements were made on two Saturdays in February and March 2001 from 11 am to
7pm with a portable electrical field (400 MHz – 3 GHz) detector (Nuova Elettronica Model LX-
1435). This meter was calibrated with an HP-8510C network analyser inside an anechoic chamber
at the University of Valencia. During the bedroom exposure assessment the electric field probe was
held for approximately five minutes about one meter from the walls and 1.2 meters above the
ground – and moved around a circle of 0.25 m radius, orientating the antenna in different directions
to obtain the maximum electrical field strength above the bed.
To check the intensity of TV and radio channels, as well as the intensity of working channels and
broadcast channels for the GSM-900-1800 BSs, measurements of the spectral power density were
carried out with a probe antenna and a portable spectrum analyser.
The probe was mounted on a linen phenolic tripod 1.2 m above the ground. The position of the
probe was the same on both days – on a hill next to the village and 20 m from the BS. With the
spectrum analyser we scanned the frequency bands and the levels were averaged for six minutes.
The measurement of the spectrum was similar on both days – with a difference in the peak
estimation (channel carriers) of about 1 dB.
The broadband measured exposure was almost invariable during the time interval of the
measurements. It changed with the position or place but it did not change in time, this could be
Page 41 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
11
11
related with a low intensity of the traffic channels (a very low number of phone calls) related
with the higher and constant intensity of the broadcast channel [10].
Statistical analysis
Demographical data was analysed using the Mann-Whitney one-way variance analysis and Chi
square test. Also, differences between groups were performed through the analysis of variance
(ANOVA), and the analysis of covariance (ANCOVA).
The main statistical analysis was made using binary logistic regression (mode enter) carried out on
a regular basis and subsequent bootstrapping [1000 bootstrap replications, 95% percentile method
and simple sampling] [31] that could provide more accurate standard errors and confidence
intervals. After producing (1000) bootstrap replicates θ b of an estimator θ, the bootstrap standard
error was the standard deviation of the bootstrap replicates:
SE (θ) = √ [∑ (θb - θ) **2 / (r-1)] b=1� r where θ is the mean of the θb. Because of our small sample size, a non-parametric confidence
interval for the estimate (mean) was constructed from the quartiles of the bootstrap sampling
distribution of θ. The 95% percentile interval (θ (lower) <θ< θ (upper) is shown, and where the θb
are the r-ordered bootstrap replicates: lower = 0.025 × r (sample 25) and upper = 0.975 × r (sample
975).
The 15 health-related symptoms described above constituted the dichotomous dependent variables.
Then, univariate analysis was performed for each symptom and for each of the predictors variables:
exposure to base station (µW/m2 as a natural logarithmic) and age were used as continuous
variables; while gender, computer use >2 h/ day, mobile phone use > 20 minutes per day and worry
about the antennae were used as dichotomous variables. The covariates with predictive value were
considered for the multivariate analysis. Thus possible confounder effects were evaluated.
Page 42 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
12
12
In all cases, change in -2 log likelihood, odds ratio, 95%-confidence intervals (95%-CI), as well as
the probability value (p-value) were calculated. For all tests, a p-value below 0.05 was considered
statistically significant.
The dependent variables (health-related symptoms) given in four ordinal categories (0=never,
1=sometimes, 2=often; 3=very often) were dichotomised (0,1 = 0 versus 2, 3 = 1).
We used the GSM exposure (the measurement of RF EMF in the bedroom) as a continuous
variable because it is well recognised that categorisation of continuous variables introduces major
problems in the analysis and interpretation of models derived in a data-dependent fashion [32, 33,
34].
We chose exposure values in the logarithmic form because these values are well grouped around
their median; while the raw values showed a high dispersion of values, with two outliers and ten
extreme values (data not shown).
Confounding was assessed by adding the potential confounding variable to the model and making a
subjective decision as to whether or not the coefficient of the variable of interest, ORs of GSM
exposure, had changed substantially. A 10 per cent variation was accepted as a considerable change
Possible interactions between covariates were also evaluated.
The maximum number of covariates included in each multivariate analysis was calculated
following this formula [35]. Let π be the smallest of the proportions of negative or positive cases
in the population and k the number of covariates, then the minimum number of cases to include
is:
N = 10 k / π
Goodness-of-fit tests such as the classification table, the Hosmer-Lemeshow statistic, ROC
curves, Cox and Snell’s, and Nagelkerke’s Pseudo R square measures were used. The Wald
statistic was also evaluated to test the significance of individual independent variables.
Moreover, possible multicolinearity was also tested.
Page 43 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
13
13
With the predicted probability scores derived from the regression analysis, receiver operating
characteristic (ROC) curves were constructed for all symptoms or modalities in order to analyze
sensitivity and specificity levels. For each curve, the best cut-offs for GSM exposure that
maximizes (sensitivity + specificity) were also calculated.
For statistical analysis we used the Statistical Package for Social Sciences, version 21.0 (IBM
SPSS Inc., Chicago, IL, USA) for Windows.
Because an exposure assessment for transformers, high voltage power lines, and radio or TV
transmitters based on self-estimated distances would not produce a reliable exposure estimate, it
was decided to omit these covariates in the analysis.
RESULTS
Demographical data and the percentage of users of personal computers and mobile phones were
analysed. The mean age was 42, 17 years (standard deviation ± 17, 61, interval 15–81), while 13,6
% used regularly computers (6,6 % were women and 7,0% men) and 23,9 % mobile phones (8,8 %
corresponded to women and 15,1 % to men). In the total sample, 51,1 % were women (mean age =
45.08 years (SD =17.98) ; [interval=15-81] and 48,9% men (39.12 (16.88) [15-75]. No differences
related with age, use of mobile phones or computers between sexes were found.
The univariate logistic regression indicated that age was inversely associated with irritability [OR =
0.97, 95%CI = 0.95-0.99] and that the oldest had the greatest difficulties hearing [OR= 1.03, 95%
CI=1.01-1.06] and walking [OR= 1.04, 95%CI = 1.01-1.07]. However, gender clearly did not
influence the outcome of any dependent variable. Use of mobile phones was linked with lack of
appetite and vertigo, while worry about the radiation from BSs was associated with sleep troubles
(Table 1). However, concerns about the radiation from BSs were not related to age (ANOVA test),
sex (lambda statistic) or subjective distance to BS (Somers´d statistic).
Page 44 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
14
14
Most of the symptoms were related with GSM exposure, especially: fatigue, irritability, lack of
appetite, sleep troubles, depression, and lack of concentration. Change in -2 log likelihood showed
similar results (Table 2).
ROC curves of each one of the logistic regression models (GSM exposure vs. each symptom)
oscillated between 0.65 and 0.87 (Table 3). Headaches (0, 84), nausea (0,86), appetite (0,87), and
vascular problems (0,85) showed the highest values, while memory (0,67), skin (0,67), and visual
disturbances (0,65) showed the lowest ones. The Hosmer and Lemeshow test indicated that most
analyses showed no significance p- values. The exceptions were fatigue (0,003), depression (0,003)
and vertigo (0,03). In the majority of the cases the models predicted better specificity than
sensitivity. Only in the case of headaches and sleep disorder, sensitivity prevailed over specificity
(Table 3 – classification table). In the extreme case, skin and vascular problems showed null or
minimum sensitivity, but 100 % specificity. Nagelkerke pseudo R-squared, showed acceptable
coefficients with the exception of the symptoms related with vertigo and skin problems (Table 3).
The lowest threshold cut-off values of GSM were for headaches, sleep, irritability and attention;
while vascular and skin problems showed the highest threshold cut-off values (Table 3).
The influence of other covariates on the GSM ORs coefficients, such as age, cellular use and
concern about the antennae were always less than 10% (Tables 2,).
There was no observed multicollinearity among variables. Kappa values according to factor
analysis were always lower than 2 and well below the critical value of 30.
Finally, no interactions between covariates were observed.
Standard errors and confidence intervals obtained by re-sampling were similar to those calculated
from the asymptotic approximation (Table 4). There was a small bias or difference between the
average bootstrap coefficients (not shown) and the respective estimates obtained from the
original sample.
Page 45 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
15
15
There were not health differences between those who permitted an exposure measurement (88 in
our last model) and those uninterested in RFs measures (26). ANOVA showed that the group with
RF EMFs measures were more prone to complain of memory loss ( F = 5,07 ; p = 0,027); while
subjects without EMF measures showed more skin problems ( F = 10,66; p = 0,001). There were
not health differences between those groups when using age as covariate. Square partial eta
measured 0 % contribution of the willingness participation variable to symptoms that correlated
with age such as: irritability, headaches, walking difficulties and hearing loss. Also, there was not
relationship between subjective distance to the BS and willingness to participate (Pearson Chi-
square = 2,80, df = 1, p=0,094).
DISCUSSION
In the present re-analysis a more robust statistical method was employed, indifferent to the
assumption of normality. In order to reduce the limitation of sample size effect and extrapolate
our results to the entire population from which the sample was obtained, re-sample method or
bootstrapping was used.
This new study partially confirms our preliminary results, being the most of the symptoms related
to GSM levels – independently of the demographical variables and some possible risk factors.
Related to microwave radiation, the spectral power density analysis maintained that the most
important contribution to broadband measurements was from GSM 900/1800, and the main
variability of the measurements between different places was due to a different coverage of the
GSM 900/1800 signals, i.e. spatial variability. This was further supported by the fact that the
antenna used was fairly insensitive to frequencies below 400 MHz. Therefore, the radio channels
80-110 MHz were not a significant part of the broadband measurements. Moreover, the narrow
band measurements showed TV channels with substantially lower intensities than the GSM
900/1800 signals. The effects from these exposures will therefore not confound the effects of BSs.
Page 46 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
16
16
Moreover, some authors [13] found that the only relevant contribution to the variance of the high
microwave exposure was from BSs – up to 93% of variance. Moreover, at the time of our study the
GSM signal was almost invariable in time because there were very few calls. The main
contribution was made from the broadcast channels working almost constantly throughout the day.
Short-range evaluations of exposure could be acceptable for describing a 24-hour period and the
measurements were made in bedrooms – a location where the subjects were assumed to spend
significant periods of time.
However, some subjects were mobile phone users at the time of this study and exposure from a
mobile phone during a phone call is much higher than that received from BSs. Nevertheless, some
authors [13] stated about that there is no a priori argument why these lower levels should have no
effect on the presence of a widespread use of mobile telephones. Exposure from a BS will be at a
low but almost constant level for many hours a day and especially at night.
While GSM exposure was associated with most of the symptoms, walking difficulties and hearing
loss was correlated only with age. Also, age remained slightly inversely associated with irritability.
Those who were users of cellular phones were more prone to complain of loss of appetite and
vertigo, while those who expressed worry about the BSs, were associated with having sleep
problems. This later finding was in concordance with two other articles [13, 20, 26]. However,
worry about the BSs was unrelated with age, gender or subjective distance to BSs. This agrees
with an article [36] claiming that there was no statistically significant association between
symptom occurrence associated with perceived proximity to BSs, psychological components,
socio-demographic characteristics, and actual distance to BSs or powerlines.
Some authors indicated that opponents of mobile phone towers generally do not express anxieties
about electromagnetic field exposure, indicating that the risk rating is comparable with other
perceived common hazards in the modern world [37].
Page 47 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
17
17
None of the analysed covariates behaved as confounders. So, the relationship between GSM
exposure with irritability, sleep troubles, lack of appetite and vertigo remained statistically
significant despite the introduction of the above covariates.
When the conventional multivariate analysis was tested using bootstrapping it was observed that
the standard error and confidence intervals obtained by re-sampling were similar to those
calculated from asymptotic approximation and this supports the adequacy of our conventional
analysis. Our sample, chosen at random, would qualify to represent the population from which it
came.
The model appeared generally well adjusted while the cut-off values could constitute good
guidance for predicting the threshold of symptom appearance.
We cannot truly rule out that their inhabitants were more; equal or less worried than elsewhere in
this region, since we cannot provide the percentage of worriers about the BS masts in La Ñora
compared with other places of its environment. However, information about this issue spread to
all over this region at this time and La Ñora circumstances are common with the rest of small
urban and rural areas. The sample was randomly selected but a participation bias cannot be rule
out since the most part of our participants expressed fear for the existence of the BSs and this could
contribute to their participation in the study. Also, it is also possible to speculate that the percentage
of the subjects who refused to participate were for opposite reasons (indifference about the BSs). In
this regard, neither health status nor subjective distance to the BS explained willingness to
participate in the study.
Concerns about the radiation from BSs were not related to age, sex, or subjective distance to BSs.
This agrees with statements from some authors [13] that living near a base station does not make
more people generally fearful, but people that generally worry about the fields would express
stronger fears when they live close by.
Page 48 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
18
18
Nevertheless, irrespective of these explanations, there seem to be effects of exposure that occur
independent of the fear of the subjects, since controlling for fear did not change the association
between exposure and symptoms. However, biological grounds explaining non-thermal effects
have not been clearly established. Recently, it has been described that voltage-gated calcium
channels are essential to the beneficial or adverse responses to microwave EMFs, nanosecond EMF
pulses, and static electrical and magnetic fields [38].
In summary, the results of this study indicate that effects of very low but long lasting exposures to
emissions from mobile telephone base-stations on well-being cannot be ruled out. The effects
fulfilled almost completely the symptoms described within the microwave syndrome. Finally,
unraveling the causal pathways would be better done with experimental study design.
Page 49 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
19
19
CONCLUSIONS
This new study partially confirms our preliminary results about microwave sickness resulting from
exposure to emissions from GSM mobile phone BSs. So, fatigue, irritability, lack of appetite, sleep
troubles, depression, and lack of concentration were especially related with GSM exposure.
These results were independent of the main socio-demographical variables, other EMF exposures
and the worry about being irradiated. Nevertheless, we confirm that apprehension about modern
technology could predict some symptom, specially related with sleep problems.
Our results agree with those who claimed that by distorting perceptions of risk, disproportionate
precaution might paradoxically lead to illness that would not otherwise occur [39]. However,
health changes related with GSM exposure seems to occur unrelated with those fears. Finally, this
exposure was very low in that period and also very low against Spanish recommendations [40] and
international guidelines [41].
Recommendations
We subscribe the guidelines observed by other authors [42] to follow the principle of prevention
as long as the non-thermal effects are not considered in any official standard. This includes
exposure minimization in the frame of the technical feasibility to guarantee a significantly
reduction in the long term radiation exposures of cellular phone towers in residential areas.
Epidemiological and clinical studies should be continuing to observe possible health changes in
the population. Finally, clear information about the correct use of the newer electronic devices
should be implemented.
Page 50 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
20
20
Competing interests
The authors declare that they have no competing interests.
ACKNOWLEDGEMENTS
We would like to express our gratitude to Angeles Martinez Gomez for her assistance during the
field work in La Ñora; as well as the Spanish Ministry of Science and Technology for grant FIT
number 070000-2002-58.
Page 51 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
21
21
REFERENCES
1. Lilienfeld AM, Tonascia J, Tonascia S, Libauer CH, et al. Foreign Service Health Status
Study, Evaluation of health status of foreign service and other employees from selected
Eastern European posts. Baltimore: Dept Epidemiology, School of Hygiene and Public
Health, The Johns Hopkins University; 1978.
2. Elwood JM. Microwaves in the cold war: the Moscow embassy study and its interpretation.
Review of a retrospective cohort study. Environ Health 2012; Nov 14;11:8.
3. Belyaev IY. Non-thermal Biological Effects of Microwaves. Microwave Review. 2005a;
11:13-29
4. Belyaev IY. Non-thermal Biological Effects of Microwaves: Current Knowledge,
Further Perspective, and Urgent Needs. Electromagn. Biol. Med. 2005b ; 24:375-403.
5. Röösli M, Moser M, Baldinini Y, Meier M, Braun-Fahrländer C. Symptoms of ill health
ascribed to electromagnetic field exposure—a questionnaire survey. Int. J. Hyg. Environ.
Health. 2004 ; 207:141–150.
6. Frick U, Mayer M, Hauser S, Binder H , Rosner R , Eichhammer P. Entwicklung eines
deutschprachigen Messinsntrumentes für ‘‘Elektrosmog-Beschwerden’’. Umwelt Med.
Forsch. Prax. 2006 ; 11:103–113.
7. Eltiti S, Wallace D, Ridgewell A, Zougkou K, Russo R, Sepulveda F, et al. Does short-term
exposure to mobile phone base station signals increase symptoms in individuals who report
sensitivity to electromagnetic fields? A double-blind randomized provocation study. Environ
Health Perspect 2007;115:1603-1608.
8. Eltiti S, Wallace D, Ridgewell A, Zougkou K, et al. Short-term exposure to mobile phone
base station signals does not affect cognitive functioning or physiological measures in
Page 52 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
22
22
individuals who report sensitivity to electromagnetic fields and controls. Bioelectromagnetics
2009;30:556-63.
9. Khurana VG, Hardell L, Everaert J, Bortkiewicz A, et al. Epidemiological evidence for a
health risk from mobile phone base stations. Int J Occup Environ Health 2010; 16(3):263-7.
10. Navarro EA, Segura J, Portolés M, Gómez-Perretta C. The Microwave Syndrome: A
Preliminary Study in Spain. Electromagn Biol Med 2003; 22:161 – 169.
11. Santini R, Santini P, Danze J, Le Ruz P, et al. Study of the health of people living in the
vicinity of mobile phone base stations: I. Influences of distance and sex. Pathol Biol 2002;
50:369 – 373.
12. Gadzicka E, Bortkiewicz A, Zmyslony M, Szymczak W, et al.. Assessment of subjective
complaints reported by people living near mobile phone base stations [Abstract].
BiuletynPTZE Warszawa. 2006; 14:23-26.
13. Hutter H, Moshammer H, Wallner P, Kundi M. Subjective symptoms, sleeping problems,
and cognitive performance in subjects living near mobile phone base stations. Occup Environ
Med 2006; 63:307–313.
14. Abdel-Rassoul G, El-Fateh OA, Salem MA, Michael A, Farahat F, et al. Neurobehavioral
effects among inhabitants around mobile phone base stations. Neurotoxicology 2007;
28:434-440.
15. Johnson-Liakouris AG. Radiofrequency (RF) Sickness in the Lilienfeld Study: an effect of
modulated microwaves? Arch Environ Health 1998;53:236-238.
16. Thomas S, Kühnlein A, Heinrich S, Praml G, et al. Personal exposure to mobile phone
frequencies and well-being in adults: a cross-sectional study based on dosimetry.
Bioelectromagnetics 2008; 29:463-470.
Page 53 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
23
23
17. Kühnlein A, Heumann C, Thomas S, Heinrich S, et al. Personal exposure to mobile
communication networks and well-being in children--a statistical analysis based on a
functional approach. Bioelectromagnetics 2009; 30:261-9.
18. Heinrich S, Thomas S, Heumann C, von Kries R, et al. Association between exposure to
radiofrequency electromagnetic fields assessed by dosimetry and acute symptoms in children
and adolescents: a population based cross-sectional study. Environ Health 2010; 9:75.
19. Blettner M, Schlehofer B, Breckenkamp J, Kowall B, et al Mobile phonebase stations and
adverse health effects: Phase 1 of a populationbased,cross-sectional study in Germany. Occup
Environ Med.2009;66:118-123.
20. Berg-Beckhoff G, Blettner M, Kowall B, Breckenkamp J, et al. Mobile phone base stations
and adverse health effects: phase 2 of a cross-sectional study with measured radio frequency
electromagnetic fields. Occup Environ Med 2009; 66:124-130.
21. Augner C, Florian M, Pauser G, Oberfeld G, et al. GSM Base Stations: Short-Term Effects
on Well-Being. Bioelectromagnetics 2009; 30:73-80.
22. Furubayashi T, Ushiyama A, Terao Y, Mizuno Y, et al. Effects of short-term W-CDMA
mobile phone base station exposure on women with or without mobile phone related
symptoms. Bioelectromagnetics 2009; 30:100-113.
23. Zwamborn AP, Vossen SH, van Leersum BJ. Effects of global communication system
radio-frequency fields on well being and cognitive functions of human subjects with and
without subjective complaints. Netherlands Organisation for Applied Scientific Research
(TNO) 2003;148:1-89.
24. Augner C, Hacker GW, Oberfeld G, Florian M, et al. Effects of exposure to GSM mobile
phone base station signals on salivary cortisol, alpha-amylase, and immunoglobulin A.
Biomed Environ Sci 2010;23:199-207.
Page 54 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
24
24
25. Eger H, Jahn M. Spezifische Symptome und Mobilfunkstrahlung in Selbitz (Bayern)
Evidenz für eine Dosiswirkungsbeziehung. Umwelt·medizin·gesellschaft 2010; 2: 130-139.
26. Danker-Hopfe H, Dorn H, Bornkessel C, Sauter C. Do mobile phone base stations affect
sleep of residents? Results from an experimental double-blind sham-controlled field study.
Am J Hum Biol. 2010; 5:613-8
27. Bortkiewicz A, Gadzicka E, Szyjkowska A, et al. Subjective complaints of people living
near mobile phone base stations in Poland. Int J Occup Med Environ Health 2012;25:31-40
28. Eskander EF, Estefan SF, Abd-Rabou AA. How does long term exposure to base stations
and mobile phones affect human hormone profiles? Clin Biochem 2012 ; 45(1-2):157-61
29. Shahbazi-Gahrouei D, Karbalae M, Moradi HA, et al. Health effects of living near mobile
phone base transceiver station (BTS) antennae: a report from Isfahan, Iran. Electromagn Biol
Med. 2013 Jun 19.
30. Röösli M, Frei P, Mohler E, Hug K. Systematic review on the health effects of exposure to
radiofrequency electromagnetic fields from mobile phone base stations. Bull World Health
Organ 2010; 88: 887–896.
31. Efron B, Tibshirani R: An Introduction to the Bootstrap. Boca Raton, FL: Chapman &
Hall/CRC 1993; ISBN 0-412-04231-2.software
32. Maxwell SE, Delaney HD. Bivariate median-splits and spurious statistical significance.
Psychological Bulletin 1993; 113:181–190.
33. Royston P, Altman DG, Sauerbrei W.Dichotomizing continuous predictors in multiple
regression: a bad idea. Stat. Med 2006; 25:127-141.
34. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous confounding
variable is categorized in logistic regression analyses. Stat. Med 2004; 23:1159–1178.
35. Peduzzi P, Concato J, Kemper E, Holford TR, et al. A simulation study of the number of
events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373-1379.
Page 55 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
25
25
36. Baliatsas C, van Kamp I, Kelfkens G, et al. Non-specific physical symptoms in relation to
actual and perceived proximity to mobile phone base stations and powerlines. BMC Public
Health 2011; 11:421
37. Hutter H, Moshammer H, Wallner P, Kundi M. Public perception of risk concerning
celltowers and mobile phones. Soz Präventivmed 2004;49: 62–66.
38. Pall ML. Electromagnetic fields act via activation of voltage-gated calcium channels to
produce beneficial or adverse effects. J. Cell. Mol. Med 2013; 7:. 958-965
39. Coggon D. Health Risks from Mobile Phone Base Stations. Occup Environ Med 2006;
63:298–299.
40. Real Decreto 1066/2001 Telecomunicaciones/TelefoniaMovil/RD1066_2001.pdf
http://www.mityc.es/esES/OficinaVirtual/Documents/SE%20
41. International Commission on Non-Ionizing Radiation Protection: Guidelines for limiting
exposure to time-varying electric, magnetic, and electromagnetic fields (up to 300 GHz).
Health Phys 1998; 74: 494–522.
42. Haumann T, Münzenberg U, Maes W, Sierck P. HF-Radiation levels of GSM cellular
phone towers in residential areas. 2nd International Workshop on Biological effects of EMFS.
October 2002; Rhodes (Greece) Vol 1:327-333.
Page 56 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
26
26
Table 1. Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Symptom/
Variable
Worry about BSs (1) Computer (not use) Mobile (not use) OR (95%CI) OR (95%CI) OR (95%CI)
Fatigue 0.67 0.23-1.90 2.62 0.76-9.04 1.56 0.56-4.35 Irritability 1.13 0.43-3.03 1.56 0.45-5.34 2.62 0.94-7.33 Headaches 1.75 0.62-4.94 1.39 0.34-5.58 1.56 0.51-4.83 Nausea 0.68 0.18-2.24 0.34 0.04-2.84 1.43 0.44-4.67 Lack of appetite 1.05 0.33-3.40 3.16 0.87-11.44 4.28** 1.43-12.78 Trouble sleeping 3.12* 1.10-8.86 0.55 0.16-1.88 0.74 0.27-2.02 Depression 1.06 0.39-2.93 0.81 0.22-2.93 1.03 0.38-2.84 Lack of concentration 0.92 0.35-2.47 1.11 0.33-3.76 2.79 0.99-7.80 Memory loss 1.71 0.62-4.75 0.41 0.10-1.64 1.35 0.50-3.61 Skin alterations 0.74 0.23-2.35 φ φ 0.63 0.16-2.45
Visual disturbances 1.31 0.48-3.60 0.77 0.21-2.77 1.63 0.60-4.39 Vertigo 0.61 0.20-1.91 0.77 0.19-3.10 2.90* 1.04-8.07 Vascular alteration 0.96 0.27-3.43 1.48 0.35-6.17 2.04 0.65-6.41 Hearing problems 0.59 0.20-1.70 0.77 0.19-3.10 0.48 0.15-1.60 Walking difficulty 0.60 0.20-1.79 φ φ 0.42 0.11-1.60 * p<0.05; ** p<0.01; (1) Not worried, as reference code (2) & (3) Not device use, as reference code,φ any subject affected using computer
Page 57 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
27
27
Table 2. ORs & 95% CIs for GSM exposure: increase in risk per increase in ln GSM (µW/m2). Change in -2 log likelihood is also shown.
Symptom
OR (95%CI) Change in
-2 Log likelihood
Fatigue 1.39***(1.14-1.70) 11.74***
Irritability 1.51***(1.23-1.85) 19.36***
Irritability (adjusted with age) 1.47***(1.20-1.81) -
Headaches 1.43** (1.15-1.78) 12.32***
Nausea 1.38** (1.09-1.73) 8.3**
Lack of Appetite 1.58** (1.23-2.03) 16.31*** Lack of Appetite (adjusted to use of cellular ) 1.53** (1.19-1.99) -
Trouble sleeping 1.49***(1.20-1.84) 16.38***
Trouble sleeping (adjusted to worry to BSs) 1.64***(1.22-2.19) -
Depression 1.41***(1.16-1.72) 13.99***
Concentration 1.54***(1.25- 1.89) 20.75***
Memory 1.27** (1.06- 1.52) 7.29**
Skin 1.24* (1.001-1.54) 4.08*
Visual 1.23 * (1.03-1.46) 5.30*
Vertigo 1.36** (1.11-1.66) 10.14***
Vertigo (adjusted to use of cellular ) 1.32** (1.08-1.62) -
Vascular 1.32* (1.05-1.64) 6.30*
Hearing 0.96 (0.80-1.15)
Walking 0.95 (0.78-1.15)
* p<0.05; ** p<0.01; *** p<0.001
Page 58 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
28
28
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (ln). Cut-off values of exposure to microwaves according to ROC analysis. The data are presented in ascending order. Symptom ROC curves
Area
Classification Table (a)
Pseudo-R**2
(1)
Cut-off (2) (Ln GSM)
Cut-off (2) GSM (µW/m2)
SSV SPF AV
Headaches 0.84*** 0.90 0.23 0.72 0.41 0.57 1.77 Sleep 0.78*** 0.82 0.66 0.76 0.28 1.66 5.26
Attention 0.78*** 0.67 0.72 0.69 0.28 3.61 36.97 Irritability 0.76*** 0.67 0.73 0.71 0.26 3.61 36.97 Memory 0.67** 0.54 0.77 0.67 0.11 4.99 146.94
Depression 0.75*** 0.46 0.76 0.65 0.20 5.22 184.93 Visual 0.65* 0.24 0.83 0.60 0.08 5.91 368.71 Fatigue 0.73*** 0.22 0.90 0.69 0.18 6.53 685.4 Vertigo 0.74*** 0.16 0.87 0.67 0.19 6.53 685.4 Appetite 0.87*** 0.40 0.94 0.85 0.43 7.31 1495.18 Nausea 0.86*** 0.46 0.93 0.87 0.38 7.31 1495.18
Vascular 0.85*** 0.20 1.0 0.90 0.34 8.02 3041.18 Skin 0.67* 0.00 1.00 0.81 0.072 9.06 8604.15
* p<0.05, ** p<0.01, *** p<0.001 (1) Nagelkerke (2) Cut-off (ROC curve) (a) SSV=sensitivity, SPF =specificity, AV=average
Page 59 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
29
29
Table 4. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients: Increase in risk per increase in ln GSM (µW/m2). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Symptom B*
Bootstrap Normal
Bias SE 95% percentile
intervals SE 95% CI
Lower Upper Lower Upper Fatigue 0.329 0.012 0.097 0.155 0.539 0.102 0.128 0.529 Irritability 0.411 0.016 0.110 0.241 0.670 0.104 0.207 0.615 Headache 0.358 0.022 0.139 0.149 0.688 0.113 0.137 0.578 Nausea 0.319 0.013 0.124 0.099 0.590 0.118 0.088 0.550 Appetite 0.456 0.026 0.134 0.264 0.784 0.128 0.205 0.707 Sleep 0.396 0.022 0.124 0.193 0.690 0.109 0.181 0.610 Depression 0.346 0.012 0.102 0.174 0.583 0.100 0.151 0.541 Attention 0.429 0.020 0.118 0.254 0.711 0.106 0.222 0.636 Memory 0.237 0.009 0.098 0.057 0.448 0.091 0.058 0.415 Skin 0.217 0.008 0.110 0.011 0.451 0.110 0.001 0.433 Visual 0.203 0.004 0.093 0.037 0.398 0.090 0.026 0.379 Hearing -0.05 -0.002 0.089 -0.219 0.143 0.093 -0.228 0.135 Vertigo 0.306 0.010 0.101 0.127 0.530 0.102 0.107 0.505 Walking -0.05 -0.006 0.098 -0.265 0.120 0.098 -0.246 0.138 Vascular 0.274 0.010 0.109 0.084 0.520 0.114 0.051 0.497 * beta coefficent (log OR)
Page 60 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
30
30
Page 61 of 61
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Subjective Symptoms Related to GSM Radiation from Mobile Phone Base Stations: a cross-sectional study
Journal: BMJ Open
Manuscript ID: bmjopen-2013-003836.R2
Article Type: Research
Date Submitted by the Author: 16-Nov-2013
Complete List of Authors: Gomez-Perretta, Claudio; La Fe, Hospital, Research Center Navarro, Enrique; University of Valencia, Applied Physics Segura, Jaume; University of Valencia, Computer Science Portolés, Manuel; La Fe, Hospital, Research Center
<b>Primary Subject Heading</b>:
Public health
Secondary Subject Heading: Occupational and environmental medicine, Epidemiology
Keywords: PUBLIC HEALTH, OCCUPATIONAL & INDUSTRIAL MEDICINE, EPIDEMIOLOGY, Health & safety < HEALTH SERVICES ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open on A
pril 19, 2021 by guest. Protected by copyright.
http://bmjopen.bm
j.com/
BM
J Open: first published as 10.1136/bm
jopen-2013-003836 on 30 Decem
ber 2013. Dow
nloaded from
For peer review only
1
1
Subjective Symptoms Related to GSM Radiation from Mobile
Phone Base Stations: a cross-sectional study
Corresponding author:
Enrique A. Navarro
Departamento de Física Aplicada, Universitat de València
C/. Dr. Moliner 50,
46100 Burjassot, Valencia, Spain
Email: [email protected]
Authors
Claudio Gómez-Perretta (1)
Enrique A. Navarro (2)
Jaume Segura (3)
Manuel Portolés (1)
(1) Research Center, University Hospital La Fe, Avda Campanar s/n, 46009 Valencia,
Spain
(2) Department of Applied Physics, Universitat de València, C/. Dr. Moliner 50, 46100
Burjassot (Valencia), Spain.
(3) Department of Computer Sciences, ETSE- Universitat de València, Avda Universitat
s/n, 46100 Burjassot (Valencia), Spain.
Running title
Subjective Symptoms Related to GSM Radiation
Keywords: public health; GSM; microwave sickness.
Page 1 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
2
2
Abstract
Objectives: We performed a reanalysis of the data from Navarro et al., 2003, in which
health symptoms related to microwave exposure from mobile phone base stations (BS)
were explored, including data obtained in a retrospective inquiry about fear of exposure
from BS.
Method: Since weather circumstances can influence exposure, we restricted data to
measurements made under similar weather conditions. Participants with known illness in
2003 were subsequently disregarded: 88 participants instead of 101 (in 2003) were
analysed. A statistical method indifferent to the assumption of normality was employed:
namely, binary logistic regression for modelling a binary response (e.g. suffering fatigue
(1) or not (0)), and so exposure was introduced as a predictor variable. This analysis was
carried out on a regular basis and bootstrapping [95% percentile method] was used to
provide more accurate confidence intervals.
Results: The symptoms most related to exposure were: lack of appetite [odds ratio
(OR)] = 1.58, 95% confidence interval (95%CI) = 1.23-2.03; lack of concentration
[OR = 1.54, 95% CI = 1.25- 1.89]; irritability [OR = 1.51, 95% CI = 1.23-1.85]; and
trouble sleeping [OR = 1.49, 95% CI = 1.20-1.84]. Changes in -2 log likelihood showed
similar results. Concerns about the BS were strongly related with trouble sleeping
[OR = 3.12, 95% CI = 1.10-8.86]. The exposure variable remained statistically significant
in the multivariate analysis. The bootstrapped values were similar to the asymptotic
confidence intervals.
Conclusion: This study confirms our preliminary results. We observed that the
incidence of most of the symptoms was related to exposure levels – independently of the
demographic variables and some possible risk factors. Concerns about adverse effects
Page 2 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
3
3
from exposure, despite being strongly related with sleep disturbances, do not influence the
direct association between exposure and sleep.
ARTICLE SUMMARY:
Article Focus: 1) Some authors stated that the microwave syndrome could be related to anxiety
associated with fear and the visual perception of cell phone towers. In this manuscript we
demonstrate that although anxiety per se disturbs sleep, sleep disorder continued to be related
with GSM exposure after adjusting for these concerns. 2) It is shown that symptoms of
microwave sickness are moderately related with this radiation, reinforcing the idea that some
underlying biological mechanism may be behind these symptoms. 3) This work presents a re-
analysis of our previous cross-sectional study [10] that uses a more robust statistical method.
Key messages: 1) Existence of the microwave syndrome. 2) Symptoms related to very low GSM
levels in comparison with Spanish and international guidelines [40], [41].
Strengths: We used a robust statistical analysis with a highly homogeneous sample in a
homogeneous environment. Limitations: A participation bias cannot be ruled out. The late
query about concerns (as a possible confounder) may render the results less valid.
Page 3 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
4
4
The health risk due to exposure to radiofrequency electromagnetic fields (RF EMF)
continues to be discussed today. The study that led to this debate was initiated after
verification that the U.S. embassy in Moscow was being subjected to such radiation from
1953 to May 1975 [1]. Recently, a review of that episode [2] reopened the debate about
the potential harmfulness of RF EMF. The increasing number of base stations (BSs) on
masts and buildings has increased public awareness. This issue has prompted scientific
research to establish to what extent low-intensity electromagnetic fields may affect the
health of humans and other organisms [3, 4]. Furthermore, the term electromagnetic
hypersensitivity (EHS) has been recently introduced in discussions attributing symptoms
to exposure to electromagnetic fields (EMFs) [5-8]. A review of this topic [9] in 2010
found that 8 of the 10 studies evaluated through PubMed had reported increased
prevalence of adverse neurobehavioral symptoms or cancer in populations living at
distances < 500 meters from base stations.
None of the studies reported exposure above accepted international guidelines, suggesting
that current guidelines may be inadequate in protecting health. Thus, the need emerges to
revaluate our pioneering work in this field in order to add new procedures and data. Few
articles have addressed the possible association between microwave sickness and
microwave exposure from GSM (Global System for Mobile Communications) base
stations (BSs) since the publication of our first study [10]. Chronologically, Santini et al.,
[11] and Gadzicka et al., [12] reported differences in the distance-dependent prevalence of
symptoms such as headache, impaired concentration, and irritability. A later Austrian
study [13] showed a positive association between the measured electrical field (GSM
900/1800) in bedrooms – and headaches, cold hands and feet, and difficulties in
concentration. An Egyptian study [14] showed a prevalence of neurological complaints,
Page 4 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
5
5
such as headache, memory changes, dizziness, tremors, depressive symptoms, and sleep
disturbances among participants directly exposed to GSM signals from BSs.
The symptoms reported by all the above cited authors belong to those attributed to the
microwave syndrome [15]. However, one article [16] using personal monitored data
from GSM-UMTS frequency bands found no statistical association in adults. More
recently, the same authors observed no association in children [17], contradictory results
in children and adolescents [18], and concluded that the few observed significant
associations were not causal but rather occurred by chance. Blettner et al. [19] reported
in Phase 1 of their study more health problems closer to base stations, but in Phase 2 [20]
they concluded that measured EMF emissions were not related to adverse health effects.
Other researchers focused their work on the possible existence of participants with
sensitivity to GSM or UMTS signals according to psychological, cognitive, or
autonomic assessment. These researchers used short-term exposure (only 30-50
minutes) under laboratory conditions [21-23] and revealed a large disparity between
participants. Recently, a study measuring several biological stress markers [24] found that
RF-EMF emitted by mobile phone base stations from 5.2 µW/m2 to 2126.8 µW/m2
increased cortisol and salivary alpha-amylase, while IgA concentration was not
significantly modified.
The Selbitz study [25] in 2010 described a significant dose-response relationship in
symptoms related with sleep, mood, joints, infections, skin condition, as well as
neurological, cardiovascular, visual, and auditory systems, and the gastrointestinal tract.
The existence of short-term physiological effects of EMF on sleep quality was not
evident in the work of Danker-Hopfe et al. [26]; however, it was stated that the presence
of BSs per se (not the EMF) may have a negative impact on sleep quality.
Page 5 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
6
6
A Polish study in 2012 did not show correlation between electrical field strength and
frequency of subjective symptoms; however, it showed a correlation between subjective
symptoms and the distance to BSs [27]. A study carried out in Egypt [28] revealed that
exposure to EMF emitted either from mobile phones or BSs had significant effects on the
pituitary-adrenal axis. More recently, work developed in Iran [29] indicated that
symptoms such as nausea, headache, dizziness, irritability, discomfort, nervousness,
depression, sleep disturbance, memory loss and lowering of libido were statistically
significant in people living near BSs (<300m distances) compared to those living far from
the BSs (>300m).
In our cross-sectional analysis [10], 11 out of 16 symptoms showed statistically
significant higher scores in the group with the maximum exposure level. The symptoms
are included in the microwave syndrome. We also reported statistically significant
correlation coefficients between the measured electrical field and 14 out of 16
symptoms.
A review [30] recently established several conditions for epidemiological studies to be
eligible for introduction in general analysis: eligible studies must quantify exposure
using objective measures (such as distance to the nearest BS, spot, or personal exposure
measurements in a specific frequency range); possible confounders must be considered;
and the selection of the study population must be clearly free of bias in terms of
exposure and outcomes.
Accordingly, in this re-analysis of our previous study [10], possible confounders were
included in addition to the specific RF EMF measurements made in 2001 (covering the
specific range between 900 and 1800 MHz). Therefore, we co-analysed the effects of
other variables such as socio-demographic data and the use of electronic devices. Concern
about being damaged by radiation from antennas was also analysed.
Page 6 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
7
7
The new statistical approach tested the possible influences of other variables, such as
demographic data, and the use of electronic devices. Moreover, since some concerns have
been raised about possible health consequences caused by the emitted microwaves, we
analysed if these symptoms might be related to fear of exposure. As some participants
refused to allow measurements in their homes, we analysed if symptom status or
subjective distance to the BS could be a bias of participation in the study. Interestingly,
this period was free of other sources of RF such as WIFI or UMTS etc. or the massive use
of mobile phones, enabling a specific study of GSM technology. Finally, the suitability of
the size of the sample was analysed.
Page 7 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
8
8
METHODS
Study design
We chose a small urban area with mixed rural characteristics: low levels of environmental
pollution (more agricultural than industrial); no major differences in socio-economic
characteristics throughout the region (excluding large cities); similar ethnicity (white
Caucasian) and language (Spanish); and with mobile phone communication operative for
at least two years. La Ñora was chosen because it had the features of a small city, and was
located near the capital (Murcia) in a rural environment without any particular health or
environmental problems. Consequently, La Ñora was representative of small urban areas
in eastern Spain with fewer than 20,000 inhabitants – such rural areas accounting for
19.8% of the population and 35.9% of the territory in Spain.
Two BS masts, each about 30 m height were sited at different positions to provide GSM-
900-1800 coverage. The GSM 900 BS was positioned not before 1997 while the GSM
1800 BS was built in December 1999.
Data regarding the main demographic characteristics of the sample and their use of
electronic devices was collected through a Spanish-language questionnaire [11]. All of
the participants were of the same ethnic origin, shared similar family income levels and
general standard of living, and were born in La Ñora or nearby. All the residents in the
study were living in the village before the erection of both BSs. All of the residents were
at home for more than eight hours a day for at least six days a week and normally slept at
home.
The core of the questionnaire was a symptom checklist for estimating the frequency of 15
health-related symptoms attributed to microwave sickness. These symptoms were:
fatigue; irritability; headaches; nausea; loss of appetite; sleep disorders; depressive
tendency; dizziness; concentration difficulties; memory loss; skin lesions; visual and
Page 8 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
9
9
hearing deficiencies; walking difficulties; and cardiovascular problems. The frequency
was quantified as: never suffer = 0; sometimes = 1; often = 2; and very often = 3.
The percentage of residents who reported electrical transformers less than 10 m from their
home was 21.6%; while 42% reported high voltage power lines less than 100 m from
home. Finally, 40% of residents reported a TV transmitter within a radius of around 4 km.
The questionnaire included a statement that its purpose was health research, that the data
gathered would be confidential, and that the study was approved by the ethical committee
of the University of Valencia in accordance with the Declaration of Helsinki
(http://www.wma.net/e/policy/b3.htm).
Some 215 questionnaires were randomly distributed through 17 streets representing
practically the entire village. The houses were selected using a street map of the village.
In total 150 questionnaires were collected with the remainder being uncollected because
nobody was at home (31); or there was a refusal by the householder to complete the
questionnaire (34).
During 2001, 101 RF EMF measurements in bedrooms were made. The other (49)
residents refused admittance for taking the measurements (16), were not at home for the
scheduled measurement appointment (10), or had serious health problems (23).
However, some changes are now being introduced in this re-analysis. Thirteen of the
participants included in the original study have now been eliminated: two participants
were eliminated (one regarding alcohol abuse and another regarding pregnancy) to
increase the requirement on health criteria; and 11 to increase the homogeneity of the RF
EMFs measurements because there was a change (it was raining) in the usual dry weather
conditions when the respective broadband measurements were registered.
Page 9 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
10
10
The re-analysis of the dataset, which is the main focus of this paper, was finally
performed with 88 participants (45 female and 43 male) instead of the 101 analysed in
2001.
Concerns about microwave exposure
Sixty-six of the 88 participants were reached by telephone in February 2012 and asked
two questions:
a) Were you worried about the masts (BSs) when they were erected?
b) Did you believe their radiation (BSs) could damage your health?
In all cases, those who were worried about the masts were concerned about health
consequences. Twenty-seven participants (40. 9 %) responded no, and 39 (59.1%)
responded yes. Responses were analysed relative to age (ANOVA test), sex (lambda
statistic), and subjective distance to BS (Somers’ D statistic).
Exposure assessment
Broadband measurements were made on two Saturdays in February and March 2001 from
11 am to 7pm with a portable electrical field (400 MHz – 3 GHz) detector (Nuova
Elettronica Model LX-1435). This meter was calibrated with an HP-8510C network
analyser inside an anechoic chamber at the University of Valencia. During the bedroom
exposure assessment the electric field probe was held for approximately five minutes
about one meter from the walls and 1.2 meters above the ground – and moved around a
circle of 0.25 m radius, orientating the antenna in different directions to obtain the
maximum electrical field strength above the bed.
To check the intensity of TV and radio channels, as well as the intensity of working
channels and broadcast channels for the GSM-900-1800 BSs, measurements of the
Page 10 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
11
11
spectral power density were carried out with a probe antenna and a portable spectrum
analyser.
The probe was mounted on a linen phenolic tripod 1.2 m above the ground. The position
of the probe was the same on both days – on a hill next to the village and 20 m from the
BS. With the spectrum analyser we scanned the frequency bands and the levels were
averaged for six minutes. The measurement of the spectrum was similar on both days –
with a difference in the peak estimation (channel carriers) of about 1 dB.
The measured broadband exposure was almost invariable during the time interval of the
measurements. Exposure changed with the position or place but it did not change over
time, and this could be related with a low intensity of traffic (few phone calls) and the
high and constant intensity of the broadcast channel [10].
Statistical analysis
Demographic data was analysed using the Mann-Whitney one-way variance analysis and
Chi square test. Differences between groups were performed through variance
(ANOVA) and covariance analysis (ANCOVA).
The main statistical analysis was made using binary logistic regression (mode enter)
carried out on a regular basis with subsequent bootstrapping [1000 bootstrap replications,
95% percentile method and simple sampling] [31] to provide more accurate standard error
and confidence intervals. After producing (1000) bootstrap replicates θ b of an estimator
θ, the bootstrap standard error was the standard deviation of the bootstrap replicates:
SE (θ) = √ [∑ (θb - θ) **2 / (r-1)] b=1� r where θ is the mean of the θb. Because of our small sample size, a non-parametric
confidence interval for the estimate (mean) was constructed from the quartiles of the
bootstrap sampling distribution of θ. The 95% percentile interval (θ (lower) <θ< θ (upper)
Page 11 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
12
12
is shown, and where the θb are the r-ordered bootstrap replicates: lower = 0.025 × r
(sample 25) and upper = 0.975 × r (sample 975).
The dependent variables (health-related symptoms) given in four ordinal categories
(0=never, 1=sometimes, 2=often; 3=very often) were dichotomised (0,1 = 0 versus 2, 3 =
1).
The 15 health-related symptoms described above constituted the dichotomous dependent
variables. Univariate analysis was then performed for each symptom and for each of the
predictor variables: exposure to base station (µW/m2 as a natural logarithmic) and age
were used as continuous variables; while gender, computer use >2 h/ day, mobile phone
use > 20 minutes per day, and worry about the antennae were used as dichotomous
variables. The covariates with predictive value were considered for the multivariate
analysis. Thus possible confounder effects were evaluated.
In all cases, changes in -2 Log likelihood, odds ratio, 95%-confidence intervals (95%-CI),
as well as the probability value (p-value) were calculated. For all tests, a p-value below
0.05 was considered statistically significant.
We used the GSM exposure (the measurement of RF EMF in the bedroom) as a
continuous variable because it is recognised that categorisation of continuous variables
introduces major problems in the analysis and interpretation of models derived in a data-
dependent fashion [32, 33, 34].
We chose exposure values in the logarithmic form because these values are well grouped
around their median; while the raw values showed a high dispersion of values, with two
outliers and ten extreme values (data not shown).
Confounding was assessed by adding the potentially confounding variable to the model
and making a subjective decision as to whether or not the coefficient of the variable of
Page 12 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
13
13
interest, ORs of GSM exposure, had changed substantially. A 10 per cent variation was
accepted as a considerable change.
Possible interactions between covariates were also evaluated.
The maximum number of covariates included in each multivariate analysis was
calculated following this formula [35]. Let π be the smallest of the proportions of
negative or positive cases in the population and k the number of covariates, then the
minimum number of cases to include is:
N = 10 k / π
Goodness-of-fit tests such as the classification table, the Hosmer-Lemeshow statistic,
ROC curves, Cox and Snell’s, and Nagelkerke’s Pseudo R square measures were used.
The Wald statistic was also evaluated to test the significance of individual independent
variables. Moreover, possible multicolinearity was also tested.
With the predicted probability scores derived from the regression analysis, receiver
operating characteristic (ROC) curves were constructed for all symptoms or modalities in
order to analyse sensitivity and specificity levels. For each curve, the best cut-offs for
GSM exposure that maximises (sensitivity + specificity) were also calculated.
For statistical analysis we used the Statistical Package for Social Sciences, version 21.0
(IBM SPSS Inc., Chicago, IL, USA) for Windows.
Because an exposure assessment for transformers, high voltage power lines, and radio or
TV transmitters based on self-estimated distances would not produce a reliable exposure
estimate, it was decided to omit these covariates in the analysis.
RESULTS
Demographic data and the percentage of users of personal computers and mobile phones
were analysed. The mean age was 42. 17 years (standard deviation ± 17. 61, interval 15–
Page 13 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
14
14
81). Women totalled 51.1 % (mean age = 45.08 years (SD =17.98) ; [interval=15-81] and
48.9% were men (mean age = 39.12 years, SD=16.88) [15-75]. Some 13.6 % regularly
used computers and 23.9 % mobile phones.
No differences related with age, and use of mobile phones or computers were found
between the sexes.
The univariate logistic regression indicated that age was inversely associated with
irritability [OR = 0.97, 95%CI = 0.95-0.99] and that the oldest had the greatest difficulties
hearing [OR= 1.03, 95% CI=1.01-1.06] and walking [OR= 1.04, 95%CI = 1.01-1.07].
However, gender clearly did not influence the outcome of any dependent variable. Use of
mobile phones was linked with lack of appetite and vertigo, while worry about the
radiation from BSs was associated with trouble sleeping (Table 1). However, concern
about radiation from BSs was unrelated to age (ANOVA test), sex (lambda statistic), or
subjective distance to BS (Somers’ D statistic).
Most of the symptoms were related with GSM exposure, especially: fatigue, irritability,
lack of appetite, trouble sleeping, depression, and lack of concentration. Change in -2 log
likelihood showed similar results (Table 2). Figure 1 shows the distribution of EMF
measurements throughout the sample.
ROC curves for each of the logistic regression models (GSM exposure vs. each symptom)
oscillated between 0.65 and 0.87 (Table 3). Headaches (0.84), nausea (0.86), appetite
(0.87), and vascular problems (0.85) showed the highest values, while memory (0.67),
skin (0.67), and visual disturbances (0.65) showed the lowest values. The Hosmer and
Lemeshow test indicated that most analyses showed no significant p-values. The
exceptions were fatigue (0.003), depression (0.003) and vertigo (0.03). In the majority of
the cases, the models predicted better specificity than sensitivity. Only in the case of
headaches and sleep disorder, did sensitivity prevail over specificity (Table 3 –
Page 14 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
15
15
classification table). In the extreme case, skin and vascular problems showed null or
minimum sensitivity and100 % specificity. Nagelkerke pseudo R-squared, showed
acceptable coefficients with the exception of the symptoms related with vertigo and skin
problems (Table 3).
Threshold cut-off values of GSM for sleep, attention, irritability, and memory are also
shown (Table 3). The remaining cut-off values were not considered since sensitivity or
specificity was reported at below 0.50 %.
The influence of other covariates on the GSM ORs coefficients, such as age, cellular
use, and concern about the BS were always less than 10% (Tables 2).
There was no observed multi-collinearity among variables. Kappa values according to
factor analysis were always lower than 2 and well below the critical value of 30.
Finally, no interactions between covariates were observed.
Standard errors and confidence intervals obtained by re-sampling were similar to those
calculated from the asymptotic approximation (Table 4). There was a small bias or
difference between the average bootstrap coefficients (not shown) and the respective
estimates obtained from the original sample.
There were no global health differences between those who permitted a bedroom
exposure measurement (88 in our last model) and those who refused RF measurements
(26), and these results were unaltered when using age as a covariate. Square partial eta
measured a 0% contribution of the willing participation variable to symptoms that
correlated with age such as: irritability, headaches, walking difficulties, and hearing loss.
There was no relationship between subjective distance to the BS and willingness to
participate (Pearson Chi- square = 2.80, df = 1, p=0.094).
Page 15 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
16
16
However, ANOVA showed that the group with recorded RF EMF levels was more prone
to complain of memory loss ( F = 5.07 ; p = 0.027); while participants without EMF
measures showed more skin problems ( F = 10.66; p = 0.001).
DISCUSSION
In the present re-analysis a more robust statistical method was employed that was
indifferent to the assumption of normality. To reduce the limitation of the sample size
effect and extrapolate our results to the entire population from which the sample was
obtained, a re-sample method or bootstrapping was used.
This new study partially confirms our preliminary results – namely, that most of the
symptoms related to GSM levels independently of the demographical variables and some
possible risk factors. Related to microwave radiation, the spectral power density analysis
maintained that the most important contribution to broadband measurements was from
GSM 900/1800, and the main variability of the measurements between different places
was due to a different coverage of the GSM 900/1800 signals, i.e. spatial variability. This
was further supported by the fact that the antenna used was fairly insensitive to
frequencies below 400 MHz. Therefore, the radio channels 80-110 MHz were not a
significant part of the broadband measurements. Moreover, the narrow band
measurements showed TV channels with substantially lower intensities than the GSM
900/1800 signals. The effects from these exposures will therefore not confound the effects
of BSs. Moreover, some authors [13] found that the only relevant contribution to the
variance of the high microwave exposure was from BSs – up to 93% of variance.
Moreover, at the time of our study the GSM signal was almost invariable in time because
there were very few calls. The main contribution was made from the broadcast channels
working almost constantly throughout the day. Short-range evaluations of exposure could
Page 16 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
17
17
be acceptable for describing a 24-hour period and the measurements were made in
bedrooms – a location where the participants were assumed to spend significant periods
of time.
However, some participants were mobile phone users at the time of this study and
exposure to a mobile phone during a phone call is much higher than that received from
BSs. Nevertheless, some authors [13] stated about that there is no a priori argument why
these lower levels should have no effect on the presence of a widespread use of mobile
telephones. Exposure to a BS will be at a low but almost constant level for many hours of
the day and especially at night.
While GSM exposure was associated with most of the symptoms, walking difficulties and
hearing loss was correlated only with age. Age also remained slightly inversely associated
with irritability. Users of cellular phones were more prone to complain of loss of appetite
and vertigo, while those who expressed worry about the BSs, were associated with sleep
problems. This later finding was in concordance with two other articles [13, 20, 26].
However, worry about the BSs was unrelated with age, gender, or subjective distance to
BSs. This agrees with an article [36] claiming that there was no statistically significant
association between symptom occurrence associated with perceived proximity to BSs,
psychological components, socio-demographic characteristics, and distance to BSs or
power lines.
Some authors indicated that opponents of mobile phone towers generally do not express
anxieties about electromagnetic field exposure, indicating that the risk rating is
comparable with other commonly perceived hazards in the modern world [37].
None of the analysed covariates behaved as confounders. The relationship of GSM
exposure with irritability, sleep troubles, lack of appetite, and vertigo remained
statistically significant despite the introduction of the above covariates.
Page 17 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
18
18
When the conventional multivariate analysis was tested using bootstrapping it was
observed that the standard error and confidence intervals obtained by re-sampling were
similar to those calculated from asymptotic approximation and this supports the
adequacy of our conventional analysis. Our sample, chosen at random, represents the
population from which it came.
The model appeared generally well adjusted while the cut-off values could constitute
good guidance for predicting the threshold of symptom appearance.
We cannot truly state that residents were more worried; equally worried, or less worried
than elsewhere in this region, since we cannot provide the percentage of those worried
about the BS masts in La Ñora compared with other nearby places. However,
information about this issue was widespread in this region at the time, and the
circumstances at La Ñora were shared with most other small urban and rural areas. The
sample was randomly selected but a participation bias cannot be ruled out since most of
our participants expressed fear regarding BSs and this could contribute to their
participation in the study. It is also possible to speculate that the percentage of participants
who refused to participate did so for the opposite reasons (indifference about BSs). In this
regard, neither health status nor subjective distance to the BS explained a willingness to
participate in the study.
Concerns about radiation from BSs were not related to age, sex, or subjective distance to
BSs. This agrees with statements from several authors [13] that living near a base
station does not make people generally fearful, but people that generally worry about
fields express stronger fears when they live close to a station.
Nevertheless, irrespective of these explanations, there seems to be effects of exposure that
occur independently of the fear felt by the participants, since controlling for fear did not
change the association between exposure and symptoms. However, the late query about
Page 18 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
19
19
concerns (as a possible confounder) may render the results less valid. In contrast to our
findings, note that biological grounds explaining non-thermal effects have not been
clearly established. Recently, it has been stated that voltage-gated calcium channels are
essential to the beneficial or adverse responses to microwave EMFs, nanosecond EMF
pulses, and static electrical and magnetic fields [38].
In summary, the results of this study indicate that effects of very low but long lasting
exposure to emissions from mobile telephone base stations on well-being cannot be ruled
out. The effects almost completely matched the symptoms described within the
microwave syndrome. Finally, unravelling the causal pathways would be best performed
with an experimental study design.
Page 19 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
20
20
CONCLUSIONS
This new study partially confirms our preliminary results about microwave sickness
resulting from exposure to emissions from GSM mobile phone BSs. Fatigue, irritability,
lack of appetite, sleep troubles, depression, and lack of concentration were especially
related with GSM exposure.
These results were independent of the main socio-demographic variables, other EMF
exposures, and anxiety about being irradiated. Nevertheless, we confirm that
apprehension about modern technology could predict some symptoms, especially those
related with sleep problems.
Our results agree with those who claimed that by distorting perceptions of risk,
disproportionate precaution might paradoxically lead to illness that would not otherwise
occur [39]. However, health changes related with GSM exposure seem to occur in a
manner unrelated with those fears. Finally, exposure was very low during the period and
also very low in comparison with Spanish recommendations [40] and international
guidelines [41].
Recommendations
We subscribe to the guidelines observed by other authors [42] in following the principle
of prevention while the non-thermal effects are not considered in any official standard.
This includes exposure minimisation within the limits of technical feasibility to
guarantee a significant reduction in long term radiation exposure to cellular phone
towers in residential areas. Epidemiological and clinical studies should continue to
observe possible health changes in the population. Finally, clear information about the
correct use of newer electronic devices should be implemented.
Page 20 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
21
21
ACKNOWLEDGEMENTS
We would like to express our gratitude to Angeles Martinez Gomez for her assistance
during the fieldwork in La Ñora; as well as the Spanish Ministry of Science and
Technology for grant FIT number 070000-2002-58.
Authors' contributions
CGP was mainly responsible for designing and writing the manuscript. CGP also made
the main statistical contribution. EAN was one of the researchers responsible for the
design and acquisition of RF EMF data, as well as writing and reviewing the
manuscript. JS contributed mainly to processing the RF EMF data, while MP was
responsible for the design and final review of the manuscript.
Funding
This work was funded by the Spanish Ministry of Science and Technology for grant
FIT number 070000-2002-58.
Competing interests
The authors declare that they have no competing interests.
Data Sharing
The data used in this statistical analysis can be obtained from Dr. Claudio Gómez-
Perretta upon request by email ([email protected]).
Page 21 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
22
22
REFERENCES
1. Lilienfeld AM, Tonascia J, Tonascia S, et al. Foreign Service Health Status Study,
Evaluation Of Health Status Of Foreign Service and Other Employees from Selected
Eastern European Posts. Baltimore: Dept Epidemiology, School of Hygiene and
Public Health, The Johns Hopkins University; 1978.
2. Elwood JM. Microwaves in the cold war: the Moscow embassy study and its
interpretation. Review of a retrospective cohort study. Environ Health 2012; Nov
14;11:8.
3. Belyaev IY. Non-thermal Biological Effects of Microwaves. Microwave Review.
2005a; 11:13-29
4. Belyaev IY. Non-thermal Biological Effects of Microwaves: Current Knowledge,
Further Perspective, and Urgent Needs. Electromagn. Biol. Med. 2005b ; 24:375-
403.
5. Röösli M, Moser M, Baldinini Y, et al. Symptoms of ill health ascribed to
electromagnetic field exposure – a questionnaire survey. Int. J. Hyg. Environ.
Health. 2004 ; 207:141–150.
6. Frick U, Mayer M, Hauser S, et al. Entwicklung eines deutschprachigen
Messinsntrumentes für ‘‘Elektrosmog-Beschwerden’’. Umwelt Med. Forsch. Prax.
2006 ; 11:103–113.
7. Eltiti S, Wallace D, Ridgewell A, et al. Does short-term exposure to mobile phone
base station signals increase symptoms in individuals who report sensitivity to
electromagnetic fields? A double-blind randomised provocation study. Environ
Health Perspect 2007;115:1603-1608.
Page 22 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
23
23
8. Eltiti S, Wallace D, Ridgewell A, et al. Short-term exposure to mobile phone base
station signals does not affect cognitive functioning or physiological measures in
individuals who report sensitivity to electromagnetic fields and controls.
Bioelectromagnetics 2009;30:556-63.
9. Khurana VG, Hardell L, Everaert J, et al. Epidemiological evidence for a health
risk from mobile phone base stations. Int J Occup Environ Health 2010; 16(3):263-
7.
10. Navarro EA, Segura J, Portolés M, et al. The Microwave Syndrome: A Preliminary
Study in Spain. Electromagn Biol Med 2003; 22:161 – 169.
11. Santini R, Santini P, Danze J, et al. Study of the health of people living in the
vicinity of mobile phone base stations: I. Influences of distance and sex. Pathol Biol
2002; 50:369 – 373.
12. Gadzicka E, Bortkiewicz A, Zmyslony M, et al.. Assessment of subjective
complaints reported by people living near mobile phone base stations [Abstract].
BiuletynPTZE Warsaw. 2006; 14:23-26.
13. Hutter H, Moshammer H, Wallner P, et al. Subjective symptoms, sleeping
problems, and cognitive performance in subjects living near mobile phone base
stations. Occup Environ Med 2006; 63:307–313.
14. Abdel-Rassoul G, El-Fateh OA, Salem MA, et al. Neurobehavioral effects among
inhabitants around mobile phone base stations. Neurotoxicology 2007; 28:434-440.
15. Johnson-Liakouris AG. Radiofrequency (RF) Sickness in the Lilienfeld Study: an
effect of modulated microwaves? Arch Environ Health 1998;53:236-238.
16. Thomas S, Kühnlein A, Heinrich S, et al. Personal exposure to mobile phone
frequencies and well-being in adults: a cross-sectional study based on dosimetry.
Bioelectromagnetics 2008; 29:463-470.
Page 23 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
24
24
17. Kühnlein A, Heumann C, Thomas S, et al. Personal exposure to mobile
communication networks and well-being in children--a statistical analysis based on
a functional approach. Bioelectromagnetics 2009; 30:261-9.
18. Heinrich S, Thomas S, Heumann C, et al. Association between exposure to
radiofrequency electromagnetic fields assessed by dosimetry and acute symptoms in
children and adolescents: a population based cross-sectional study. Environ Health
2010; 9:75.
19. Blettner M, Schlehofer B, Breckenkamp J, et al Mobile phonebase stations and
adverse health effects: Phase 1 of a populationbased,cross-sectional study in
Germany. Occup Environ Med.2009;66:118-123.
20. Berg-Beckhoff G, Blettner M, Kowall B, et al. Mobile phone base stations and
adverse health effects: phase 2 of a cross-sectional study with measured radio
frequency electromagnetic fields. Occup Environ Med 2009; 66:124-130.
21. Augner C, Florian M, Pauser G, et al. GSM Base Stations: Short-Term Effects on
Well-Being. Bioelectromagnetics 2009; 30:73-80.
22. Furubayashi T, Ushiyama A, Terao Y, et al. Effects of short-term W-CDMA
mobile phone base station exposure on women with or without mobile phone related
symptoms. Bioelectromagnetics 2009; 30:100-113.
23. Zwamborn AP, Vossen SH, van Leersum BJ. Effects of global communication
system radio-frequency fields on well being and cognitive functions of human
subjects with and without subjective complaints. Netherlands Organisation for
Applied Scientific Research (TNO) 2003;148:1-89.
24. Augner C, Hacker GW, Oberfeld G, et al. Effects of exposure to GSM mobile
phone base station signals on salivary cortisol, alpha-amylase, and immunoglobulin
A. Biomed Environ Sci 2010;23:199-207.
Page 24 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
25
25
25. Eger H, Jahn M. Spezifische Symptome und Mobilfunkstrahlung in Selbitz
(Bayern) Evidenz für eine Dosiswirkungsbeziehung. Umwelt·medizin·gesellschaft
2010; 2: 130-139.
26. Danker-Hopfe H, Dorn H, Bornkessel C, et al. Do mobile phone base stations
affect sleep of residents? Results from an experimental double-blind sham-
controlled field study. Am J Hum Biol. 2010; 5:613-8
27. Bortkiewicz A, Gadzicka E, Szyjkowska A, et al. Subjective complaints of people
living near mobile phone base stations in Poland. Int J Occup Med Environ Health
2012;25:31-40
28. Eskander EF, Estefan SF, Abd-Rabou AA. How does long term exposure to base
stations and mobile phones affect human hormone profiles? Clin Biochem 2012 ;
45(1-2):157-61
29. Shahbazi-Gahrouei D, Karbalae M, Moradi HA, et al. Health effects of living near
mobile phone base transceiver station (BTS) antennae: a report from Isfahan, Iran.
Electromagn Biol Med. 2013 Jun 19.
30. Röösli M, Frei P, Mohler E, et al. Systematic review on the health effects of
exposure to radiofrequency electromagnetic fields from mobile phone base stations.
Bull World Health Organ 2010; 88: 887–896.
31. Efron B, Tibshirani R: An Introduction to the Bootstrap. Boca Raton, FL: Chapman
& Hall/CRC 1993; ISBN 0-412-04231-2.software
32. Maxwell SE, Delaney HD. Bivariate median-splits and spurious statistical
significance. Psychological Bulletin 1993; 113:181–190.
33. Royston P, Altman DG, Sauerbrei W.Dichotomizing continuous predictors in
multiple regression: a bad idea. Stat. Med 2006; 25:127-141.
Page 25 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
26
26
34. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous
confounding variable is categorized in logistic regression analyses. Stat. Med 2004;
23:1159–1178.
35. Peduzzi P, Concato J, Kemper E, et al. A simulation study of the number of events
per variable in logistic regression analysis. J Clin Epidemiol 1996; 49:1373-1379.
36. Baliatsas C, van Kamp I, Kelfkens G, et al. Non-specific physical symptoms in
relation to actual and perceived proximity to mobile phone base stations and
powerlines. BMC Public Health 2011; 11:421
37. Hutter H, Moshammer H, Wallner P, et al. Public perception of risk concerning
celltowers and mobile phones. Soz Präventivmed 2004;49: 62–66.
38. Pall ML. Electromagnetic fields act via activation of voltage-gated calcium
channels to produce beneficial or adverse effects. J. Cell. Mol. Med 2013; 7:. 958-
965
39. Coggon D. Health Risks from Mobile Phone Base Stations. Occup Environ Med
2006; 63:298–299.
40. Real Decreto 1066/2001 Telecomunicaciones/TelefoniaMovil/RD1066_2001.pdf
http://www.mityc.es/esES/OficinaVirtual/Documents/SE%20
41. International Commission on Non-Ionizing Radiation Protection: Guidelines for
limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up
to 300 GHz). Health Phys 1998; 74: 494–522.
42. Haumann T, Münzenberg U, Maes W, et al. HF-Radiation levels of GSM
cellular phone towers in residential areas. 2nd International Workshop on Biological
effects of EMFS. October 2002; Rhodes (Greece) Vol 1:327-333.
Page 26 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
27
27
Page 27 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
28
28
Table 1. Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Symptom/
Variable
Worry about BSs (1) Computer (not use) Mobile (not use) OR (95%CI) OR (95%CI) OR (95%CI)
Fatigue 0.67 0.23-1.90 2.62 0.76-9.04 1.56 0.56-4.35 Irritability 1.13 0.43-3.03 1.56 0.45-5.34 2.62 0.94-7.33 Headaches 1.75 0.62-4.94 1.39 0.34-5.58 1.56 0.51-4.83 Nausea 0.68 0.18-2.24 0.34 0.04-2.84 1.43 0.44-4.67 Lack of appetite 1.05 0.33-3.40 3.16 0.87-11.44 4.28** 1.43-12.78 Trouble sleeping 3.12* 1.10-8.86 0.55 0.16-1.88 0.74 0.27-2.02 Depression 1.06 0.39-2.93 0.81 0.22-2.93 1.03 0.38-2.84 Lack of concentration 0.92 0.35-2.47 1.11 0.33-3.76 2.79 0.99-7.80 Memory loss 1.71 0.62-4.75 0.41 0.10-1.64 1.35 0.50-3.61 Skin alterations 0.74 0.23-2.35 φ φ 0.63 0.16-2.45
Visual disturbances 1.31 0.48-3.60 0.77 0.21-2.77 1.63 0.60-4.39 Vertigo 0.61 0.20-1.91 0.77 0.19-3.10 2.90* 1.04-8.07 Vascular alteration 0.96 0.27-3.43 1.48 0.35-6.17 2.04 0.65-6.41 Hearing problems 0.59 0.20-1.70 0.77 0.19-3.10 0.48 0.15-1.60 Walking difficulty 0.60 0.20-1.79 φ φ 0.42 0.11-1.60 * p<0.05; ** p<0.01; (1) Not worried, as reference code (2) & (3). No device use, as reference code,φ any subject affected using computer
Page 28 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
29
29
Table 2. ORs & 95% CIs for GSM exposure: increase in risk per increase in Log GSM (µW/m2). Changes in -2 Log likelihood is also shown. The third column represents the ORs for a 10 fold increase in GSM (Log10 GSM)
Symptom OR (95%CI)
Change in -2 Log likelihood OR (95%CI)
Fatigue 1.39***(1.14-1.70) 11.74*** 2.13***(1.34-3.83) Irritability 1.51***(1.23-1.85) 19.36*** 2.58***(1.61-4.12) Irritability (adjusted with age)
1.47***(1.20-1.81) - 2.44***(1.52-3.94)
Headaches 1.43** (1.15-1.78) 12.32*** 2.28**(1.37-3.78) Nausea 1.38** (1.09-1.73) 8.3** 2.09**(1.23-3.55) Lack of Appetite 1.58** (1.23-2.03) 16.31*** 2.86***(1.60-5.09) Lack of Appetite (adjusted to cellular use)
1.53** (1.19-1.99) - 2.68***(1.48-4.84)
Trouble sleeping 1.49***(1.20-1.84) 16.38*** 2.49***(1.52-4.08) Trouble sleeping (adjusted to worry to BSs)
1.64***(1.22-2.19) - 3.11***(1.59-6.09)
Depression 1.41***(1.16-1.72) 13.99*** 2.22***(1.42-3.48) Concentration 1.54***(1.25- 1.89) 20.75*** 2.68***(1.67-4.32) Memory 1.27** (1.06- 1.52) 7.29** 1.73**(1.14-2.60) Skin 1.24* (1.001-1.54) 4.08* 1.65*(1.01-2.71) Visual 1.23 * (1.03-1.46) 5.30* 1.59*(1.06-2.40) Vertigo 1.36** (1.11-1.66) 10.14*** 2.02**(1.28-3.20) Vertigo (adjusted to cellular use)
1.32** (1.08-1.62) - 1.91**(1.20-3.04)
Vascular 1.32* (1.05-1.64) 6.30* 1.88*(1.12-3.14) Hearing 0.96 (0.80-1.15) 0.90(0.59-1.37) Walking 0.95 (0.78-1.15) 0.88(0.57-1.37) * p<0.05; ** p<0.01; *** p<0.001
Page 29 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
30
30
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (Log=Ln). Cut-off values of exposure to microwaves according to ROC analysis. The data is presented in ascending order. Symptom ROC curves
Area
Classification Table (a)
Pseudo-R**2
(1)
Cut-off (2) (Log GSM)
Cut-off (2) GSM (µW/m2)
SSV SPF AV
Headaches 0.84*** 0.90 0.23 0.72 0.41 - 1.77 Sleep 0.78*** 0.82 0.66 0.76 0.28 1.66 5.26 Attention 0.78*** 0.67 0.72 0.69 0.28 3.61 36.97 Irritability 0.76*** 0.67 0.73 0.71 0.26 3.61 36.97 Memory 0.67** 0.54 0.77 0.67 0.11 4.99 146.94 Depression 0.75*** 0.46 0.76 0.65 0.20 - 184.93 Visual 0.65* 0.24 0.83 0.60 0.08 - 368.71 Fatigue 0.73*** 0.22 0.90 0.69 0.18 - 685.4 Vertigo 0.74*** 0.16 0.87 0.67 0.19 - 685.4 Appetite 0.87*** 0.40 0.94 0.85 0.43 - 1495.18 Nausea 0.86*** 0.46 0.93 0.87 0.38 - 1495.18 Vascular 0.85*** 0.20 1.0 0.90 0.34 - 3041.18 Skin 0.67* 0.00 1.00 0.81 0.072 - 8604.15 * p<0.05, ** p<0.01, *** p<0.001 (1) Nagelkerke (2) Cut-off (ROC curve): Only values showing SSV and SPF above 0.5 are reported (a) SSV=sensitivity, SPF =specificity, AV=average
Page 30 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
31
31
Table 4. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients: Increase in risk per increase in Log GSM (µW/m2). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Symptom B*
Bootstrap Normal
Bias SE 95% percentile
intervals SE 95% CI
Lower Upper Lower Upper Fatigue 0.329 0.012 0.097 0.155 0.539 0.102 0.128 0.529 Irritability 0.411 0.016 0.110 0.241 0.670 0.104 0.207 0.615 Headache 0.358 0.022 0.139 0.149 0.688 0.113 0.137 0.578 Nausea 0.319 0.013 0.124 0.099 0.590 0.118 0.088 0.550 Appetite 0.456 0.026 0.134 0.264 0.784 0.128 0.205 0.707 Sleep 0.396 0.022 0.124 0.193 0.690 0.109 0.181 0.610 Depression 0.346 0.012 0.102 0.174 0.583 0.100 0.151 0.541 Attention 0.429 0.020 0.118 0.254 0.711 0.106 0.222 0.636 Memory 0.237 0.009 0.098 0.057 0.448 0.091 0.058 0.415 Skin 0.217 0.008 0.110 0.011 0.451 0.110 0.001 0.433 Visual 0.203 0.004 0.093 0.037 0.398 0.090 0.026 0.379 Hearing -0.05 -0.002 0.089 -0.219 0.143 0.093 -0.228 0.135 Vertigo 0.306 0.010 0.101 0.127 0.530 0.102 0.107 0.505 Walking -0.05 -0.006 0.098 -0.265 0.120 0.098 -0.246 0.138 Vascular 0.274 0.010 0.109 0.084 0.520 0.114 0.051 0.497 * beta coefficent (Log OR)
Page 31 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
32
32
FIGURE CAPTIONS
Figure 1. Distribution of EMF measurement throughout the sample.
Page 32 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
1
1
Subjective Symptoms Related to GSM Radiation from Mobile
Phone Base Stations: a cross-sectional study
Corresponding author:
Enrique A. Navarro
Departamento de Física Aplicada, Universitat de València
C/. Dr. Moliner 50,
46100 Burjassot, Valencia, Spain
Email: [email protected]
Authors
Claudio Gómez-Perretta (1)
Enrique A. Navarro (2)
Jaume Segura (3)
Manuel Portolés (1)
(1) Research Center, University Hospital La Fe, Avda Campanar s/n, 46009 Valencia,
Spain
(2) Department of Applied Physics, Universitat de València, C/. Dr. Moliner 50, 46100
Burjassot (Valencia), Spain.
(3) Department of Computer Sciences, ETSE- Universitat de València, Avda Universitat
s/n, 46100 Burjassot (Valencia), Spain.
Running title
Subjective Symptoms Related to GSM Radiation
Keywords: public health; GSM; microwave sickness.
Page 33 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
2
2
Abstract
Objectives: We performed a reanalysis of the data from Navarro et al., 2003, in which
health symptoms related to microwave exposure from mobile phone base stations (BS)
were explored, including data obtained in a retrospective inquiry about fear of exposure
from BS.
Method: Since weather circumstances can influence exposure, we restricted data to
measurements made under similar weather conditions. Participants with known illness in
2003 were subsequently disregarded: 88 participants instead of 101 (in 2003) were
analysed. A statistical method indifferent to the assumption of normality was employed:
namely, binary logistic regression for modelling a binary response (e.g. suffering fatigue
(1) or not (0)), and so exposure was introduced as a predictor variable. This analysis was
carried out on a regular basis and bootstrapping [95% percentile method] was used to
provide more accurate confidence intervals.
Results: The symptoms most related to exposure were: lack of appetite [odds ratio
(OR)] = 1.58, 95% confidence interval (95%CI) = 1.23-2.03; lack of concentration
[OR = 1.54, 95% CI = 1.25- 1.89]; irritability [OR = 1.51, 95% CI = 1.23-1.85]; and
trouble sleeping [OR = 1.49, 95% CI = 1.20-1.84]. Changes in -2 log likelihood showed
similar results. Concerns about the BS were strongly related with trouble sleeping
[OR = 3.12, 95% CI = 1.10-8.86]. The exposure variable remained statistically significant
in the multivariate analysis. The bootstrapped values were similar to the asymptotic
confidence intervals.
Conclusion: This study confirms our preliminary results. We observed that the
incidence of most of the symptoms was related to exposure levels – independently of the
demographic variables and some possible risk factors. Concerns about adverse effects
Page 34 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
3
3
from exposure, despite being strongly related with sleep disturbances, do not influence the
direct association between exposure and sleep.
Authors' contributions
CGP was mainly responsible for designing and writing the manuscript. CGP also made
the main statistical contribution. EAN was one of the researchers responsible for the
design and acquisition of RF EMF data, as well as writing and reviewing the
manuscript. JS contributed mainly to processing the RF EMF data, while MP was
responsible for the design and final review of the manuscript.
ARTICLE SUMMARY:
Article Focus: 1) Some authors stated that the microwave syndrome could be related to anxiety
associated with fear and the visual perception of cell phone towers. In this manuscript we
demonstrate that although anxiety per se disturbs sleep, sleep disorder continued to be related
with GSM exposure after adjusting for these concerns. 2) It is shown that symptoms of
microwave sickness are moderately related with this radiation, reinforcing the idea that some
underlying biological mechanism may be behind these symptoms. 3) This work presents a re-
analysis of our previous cross-sectional study [10] that uses a more robust statistical method.
Key messages: 1) Existence of the microwave syndrome. 2) Symptoms related to very low GSM
levels in comparison with Spanish and international guidelines [40], [41].
Strengths: We used a robust statistical analysis with a highly homogeneous sample in a
homogeneous environment. Limitations: A participation bias cannot be ruled out. The late
query about concerns (as a possible confounder) may render the results less valid.
Page 35 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
4
4
The health risk due to exposure to radiofrequency electromagnetic fields (RF EMF)
continues to be discussed today. The study that led to this debate was initiated after
verification that the U.S. embassy in Moscow was being subjected to such radiation from
1953 to May 1975 [1]. Recently, a review of that episode [2] reopened the debate about
the potential harmfulness of RF EMF. The increasing number of base stations (BSs) on
masts and buildings has increased public awareness. This issue has prompted scientific
research to establish to what extent low-intensity electromagnetic fields may affect the
health of humans and other organisms [3, 4]. Furthermore, the term electromagnetic
hypersensitivity (EHS) has been recently introduced in discussions attributing symptoms
to exposure to electromagnetic fields (EMFs) [5-8]. A review of this topic [9] in 2010
found that 8 of the 10 studies evaluated through PubMed had reported increased
prevalence of adverse neurobehavioral symptoms or cancer in populations living at
distances < 500 meters from base stations.
None of the studies reported exposure above accepted international guidelines, suggesting
that current guidelines may be inadequate in protecting health. Thus, the need emerges to
revaluate our pioneering work in this field in order to add new procedures and data. Few
articles have addressed the possible association between microwave sickness and
microwave exposure from GSM (Global System for Mobile Communications) base
stations (BSs) since the publication of our first study [10]. Chronologically, Santini et al.,
[11] and Gadzicka et al., [12] reported differences in the distance-dependent prevalence of
symptoms such as headache, impaired concentration, and irritability. A later Austrian
study [13] showed a positive association between the measured electrical field (GSM
900/1800) in bedrooms – and headaches, cold hands and feet, and difficulties in
concentration. An Egyptian study [14] showed a prevalence of neurological complaints,
Page 36 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
5
5
such as headache, memory changes, dizziness, tremors, depressive symptoms, and sleep
disturbances among participants directly exposed to GSM signals from BSs.
The symptoms reported by all the above cited authors belong to those attributed to the
microwave syndrome [15]. However, one article [16] using personal monitored data
from GSM-UMTS frequency bands found no statistical association in adults. More
recently, the same authors observed no association in children [17], contradictory results
in children and adolescents [18], and concluded that the few observed significant
associations were not causal but rather occurred by chance. Blettner et al. [19] reported
in Phase 1 of their study more health problems closer to base stations, but in Phase 2 [20]
they concluded that measured EMF emissions were not related to adverse health effects.
Other researchers focused their work on the possible existence of participants with
sensitivity to GSM or UMTS signals according to psychological, cognitive, or
autonomic assessment. These researchers used short-term exposure (only 30-50
minutes) under laboratory conditions [21-23] and revealed a large disparity between
participants. Recently, a study measuring several biological stress markers [24] found that
RF-EMF emitted by mobile phone base stations from 5.2 µW/m2 to 2126.8 µW/m2
increased cortisol and salivary alpha-amylase, while IgA concentration was not
significantly modified.
The Selbitz study [25] in 2010 described a significant dose-response relationship in
symptoms related with sleep, mood, joints, infections, skin condition, as well as
neurological, cardiovascular, visual, and auditory systems, and the gastrointestinal tract.
The existence of short-term physiological effects of EMF on sleep quality was not
evident in the work of Danker-Hopfe et al. [26]; however, it was stated that the presence
of BSs per se (not the EMF) may have a negative impact on sleep quality.
Page 37 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
6
6
A Polish study in 2012 did not show correlation between electrical field strength and
frequency of subjective symptoms; however, it showed a correlation between subjective
symptoms and the distance to BSs [27]. A study carried out in Egypt [28] revealed that
exposure to EMF emitted either from mobile phones or BSs had significant effects on the
pituitary-adrenal axis. More recently, work developed in Iran [29] indicated that
symptoms such as nausea, headache, dizziness, irritability, discomfort, nervousness,
depression, sleep disturbance, memory loss and lowering of libido were statistically
significant in people living near BSs (<300m distances) compared to those living far from
the BSs (>300m).
In our cross-sectional analysis [10], 11 out of 16 symptoms showed statistically
significant higher scores in the group with the maximum exposure level. The symptoms
are included in the microwave syndrome. We also reported statistically significant
correlation coefficients between the measured electrical field and 14 out of 16
symptoms.
A review [30] recently established several conditions for epidemiological studies to be
eligible for introduction in general analysis: eligible studies must quantify exposure
using objective measures (such as distance to the nearest BS, spot, or personal exposure
measurements in a specific frequency range); possible confounders must be considered;
and the selection of the study population must be clearly free of bias in terms of
exposure and outcomes.
Accordingly, in this re-analysis of our previous study [10], possible confounders were
included in addition to the specific RF EMF measurements made in 2001 (covering the
specific range between 900 and 1800 MHz). Therefore, we co-analysed the effects of
other variables such as socio-demographic data and the use of electronic devices. Concern
about being damaged by radiation from antennas was also analysed.
Page 38 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
7
7
The new statistical approach tested the possible influences of other variables, such as
demographic data, and the use of electronic devices. Moreover, since some concerns have
been raised about possible health consequences caused by the emitted microwaves, we
analysed if these symptoms might be related to fear of exposure. As some participants
refused to allow measurements in their homes, we analysed if symptom status or
subjective distance to the BS could be a bias of participation in the study. Interestingly,
this period was free of other sources of RF such as WIFI or UMTS etc. or the massive use
of mobile phones, enabling a specific study of GSM technology. Finally, the suitability of
the size of the sample was analysed.
Page 39 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
8
8
METHODS
Study design
We chose a small urban area with mixed rural characteristics: low levels of environmental
pollution (more agricultural than industrial); no major differences in socio-economic
characteristics throughout the region (excluding large cities); similar ethnicity (white
Caucasian) and language (Spanish); and with mobile phone communication operative for
at least two years. La Ñora was chosen because it had the features of a small city, and was
located near the capital (Murcia) in a rural environment without any particular health or
environmental problems. Consequently, La Ñora was representative of small urban areas
in eastern Spain with fewer than 20,000 inhabitants – such rural areas accounting for
19.8% of the population and 35.9% of the territory in Spain.
Two BS masts, each about 30 m height were sited at different positions to provide GSM-
900-1800 coverage. The GSM 900 BS was positioned not before 1997 while the GSM
1800 BS was built in December 1999.
Data regarding the main demographic characteristics of the sample and their use of
electronic devices was collected through a Spanish-language questionnaire [11]. All of
the participants were of the same ethnic origin, shared similar family income levels and
general standard of living, and were born in La Ñora or nearby. All the residents in the
study were living in the village before the erection of both BSs. All of the residents were
at home for more than eight hours a day for at least six days a week and normally slept at
home.
The core of the questionnaire was a symptom checklist for estimating the frequency of 15
health-related symptoms attributed to microwave sickness. These symptoms were:
fatigue; irritability; headaches; nausea; loss of appetite; sleep disorders; depressive
tendency; dizziness; concentration difficulties; memory loss; skin lesions; visual and
Page 40 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
9
9
hearing deficiencies; walking difficulties; and cardiovascular problems. The frequency
was quantified as: never suffer = 0; sometimes = 1; often = 2; and very often = 3.
The percentage of residents who reported electrical transformers less than 10 m from their
home was 21.6%; while 42% reported high voltage power lines less than 100 m from
home. Finally, 40% of residents reported a TV transmitter within a radius of around 4 km.
The questionnaire included a statement that its purpose was health research, that the data
gathered would be confidential, and that the study was approved by the ethical committee
of the University of Valencia in accordance with the Declaration of Helsinki
(http://www.wma.net/e/policy/b3.htm).
Some 215 questionnaires were randomly distributed through 17 streets representing
practically the entire village. The houses were selected using a street map of the village.
In total 150 questionnaires were collected with the remainder being uncollected because
nobody was at home (31); or there was a refusal by the householder to complete the
questionnaire (34).
During 2001, 101 RF EMF measurements in bedrooms were made. The other (49)
residents refused admittance for taking the measurements (16), were not at home for the
scheduled measurement appointment (10), or had serious health problems (23).
However, some changes are now being introduced in this re-analysis. Thirteen of the
participants included in the original study have now been eliminated: two participants
were eliminated (one regarding alcohol abuse and another regarding pregnancy) to
increase the requirement on health criteria; and 11 to increase the homogeneity of the RF
EMFs measurements because there was a change (it was raining) in the usual dry weather
conditions when the respective broadband measurements were registered.
Page 41 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
10
10
The re-analysis of the dataset, which is the main focus of this paper, was finally
performed with 88 participants (45 female and 43 male) instead of the 101 analysed in
2001.
Concerns about microwave exposure
Sixty-six of the 88 participants were reached by telephone in February 2012 and asked
two questions:
a) Were you worried about the masts (BSs) when they were erected?
b) Did you believe their radiation (BSs) could damage your health?
In all cases, those who were worried about the masts were concerned about health
consequences. Twenty-seven participants (40. 9 %) responded no, and 39 (59.1%)
responded yes. Responses were analysed relative to age (ANOVA test), sex (lambda
statistic), and subjective distance to BS (Somers’ D statistic).
Exposure assessment
Broadband measurements were made on two Saturdays in February and March 2001 from
11 am to 7pm with a portable electrical field (400 MHz – 3 GHz) detector (Nuova
Elettronica Model LX-1435). This meter was calibrated with an HP-8510C network
analyser inside an anechoic chamber at the University of Valencia. During the bedroom
exposure assessment the electric field probe was held for approximately five minutes
about one meter from the walls and 1.2 meters above the ground – and moved around a
circle of 0.25 m radius, orientating the antenna in different directions to obtain the
maximum electrical field strength above the bed.
To check the intensity of TV and radio channels, as well as the intensity of working
channels and broadcast channels for the GSM-900-1800 BSs, measurements of the
Page 42 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
11
11
spectral power density were carried out with a probe antenna and a portable spectrum
analyser.
The probe was mounted on a linen phenolic tripod 1.2 m above the ground. The position
of the probe was the same on both days – on a hill next to the village and 20 m from the
BS. With the spectrum analyser we scanned the frequency bands and the levels were
averaged for six minutes. The measurement of the spectrum was similar on both days –
with a difference in the peak estimation (channel carriers) of about 1 dB.
The measured broadband exposure was almost invariable during the time interval of the
measurements. Exposure changed with the position or place but it did not change over
time, and this could be related with a low intensity of traffic (few phone calls) and the
high and constant intensity of the broadcast channel [10].
Statistical analysis
Demographic data was analysed using the Mann-Whitney one-way variance analysis and
Chi square test. Differences between groups were performed through variance
(ANOVA) and covariance analysis (ANCOVA).
The main statistical analysis was made using binary logistic regression (mode enter)
carried out on a regular basis with subsequent bootstrapping [1000 bootstrap replications,
95% percentile method and simple sampling] [31] to provide more accurate standard error
and confidence intervals. After producing (1000) bootstrap replicates θ b of an estimator
θ, the bootstrap standard error was the standard deviation of the bootstrap replicates:
SE (θ) = √ [∑ (θb - θ) **2 / (r-1)] b=1� r where θ is the mean of the θb. Because of our small sample size, a non-parametric
confidence interval for the estimate (mean) was constructed from the quartiles of the
bootstrap sampling distribution of θ. The 95% percentile interval (θ (lower) <θ< θ (upper)
Page 43 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
12
12
is shown, and where the θb are the r-ordered bootstrap replicates: lower = 0.025 × r
(sample 25) and upper = 0.975 × r (sample 975).
The dependent variables (health-related symptoms) given in four ordinal categories
(0=never, 1=sometimes, 2=often; 3=very often) were dichotomised (0,1 = 0 versus 2, 3 =
1).
The 15 health-related symptoms described above constituted the dichotomous dependent
variables. Univariate analysis was then performed for each symptom and for each of the
predictor variables: exposure to base station (µW/m2 as a natural logarithmic) and age
were used as continuous variables; while gender, computer use >2 h/ day, mobile phone
use > 20 minutes per day, and worry about the antennae were used as dichotomous
variables. The covariates with predictive value were considered for the multivariate
analysis. Thus possible confounder effects were evaluated.
In all cases, changes in -2 Log likelihood, odds ratio, 95%-confidence intervals (95%-CI),
as well as the probability value (p-value) were calculated. For all tests, a p-value below
0.05 was considered statistically significant.
We used the GSM exposure (the measurement of RF EMF in the bedroom) as a
continuous variable because it is recognised that categorisation of continuous variables
introduces major problems in the analysis and interpretation of models derived in a data-
dependent fashion [32, 33, 34].
We chose exposure values in the logarithmic form because these values are well grouped
around their median; while the raw values showed a high dispersion of values, with two
outliers and ten extreme values (data not shown).
Confounding was assessed by adding the potentially confounding variable to the model
and making a subjective decision as to whether or not the coefficient of the variable of
Page 44 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
13
13
interest, ORs of GSM exposure, had changed substantially. A 10 per cent variation was
accepted as a considerable change.
Possible interactions between covariates were also evaluated.
The maximum number of covariates included in each multivariate analysis was
calculated following this formula [35]. Let π be the smallest of the proportions of
negative or positive cases in the population and k the number of covariates, then the
minimum number of cases to include is:
N = 10 k / π
Goodness-of-fit tests such as the classification table, the Hosmer-Lemeshow statistic,
ROC curves, Cox and Snell’s, and Nagelkerke’s Pseudo R square measures were used.
The Wald statistic was also evaluated to test the significance of individual independent
variables. Moreover, possible multicolinearity was also tested.
With the predicted probability scores derived from the regression analysis, receiver
operating characteristic (ROC) curves were constructed for all symptoms or modalities in
order to analyse sensitivity and specificity levels. For each curve, the best cut-offs for
GSM exposure that maximises (sensitivity + specificity) were also calculated.
For statistical analysis we used the Statistical Package for Social Sciences, version 21.0
(IBM SPSS Inc., Chicago, IL, USA) for Windows.
Because an exposure assessment for transformers, high voltage power lines, and radio or
TV transmitters based on self-estimated distances would not produce a reliable exposure
estimate, it was decided to omit these covariates in the analysis.
RESULTS
Demographic data and the percentage of users of personal computers and mobile phones
were analysed. The mean age was 42. 17 years (standard deviation ± 17. 61, interval 15–
Page 45 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
14
14
81). Women totalled 51.1 % (mean age = 45.08 years (SD =17.98) ; [interval=15-81] and
48.9% were men (mean age = 39.12 years, SD=16.88) [15-75]. Some 13.6 % regularly
used computers and 23.9 % mobile phones.
No differences related with age, and use of mobile phones or computers were found
between the sexes.
The univariate logistic regression indicated that age was inversely associated with
irritability [OR = 0.97, 95%CI = 0.95-0.99] and that the oldest had the greatest difficulties
hearing [OR= 1.03, 95% CI=1.01-1.06] and walking [OR= 1.04, 95%CI = 1.01-1.07].
However, gender clearly did not influence the outcome of any dependent variable. Use of
mobile phones was linked with lack of appetite and vertigo, while worry about the
radiation from BSs was associated with trouble sleeping (Table 1). However, concern
about radiation from BSs was unrelated to age (ANOVA test), sex (lambda statistic), or
subjective distance to BS (Somers’ D statistic).
Most of the symptoms were related with GSM exposure, especially: fatigue, irritability,
lack of appetite, trouble sleeping, depression, and lack of concentration. Change in -2 log
likelihood showed similar results (Table 2). Figure 1 shows the distribution of EMF
measurements throughout the sample.
ROC curves for each of the logistic regression models (GSM exposure vs. each symptom)
oscillated between 0.65 and 0.87 (Table 3). Headaches (0.84), nausea (0.86), appetite
(0.87), and vascular problems (0.85) showed the highest values, while memory (0.67),
skin (0.67), and visual disturbances (0.65) showed the lowest values. The Hosmer and
Lemeshow test indicated that most analyses showed no significant p-values. The
exceptions were fatigue (0.003), depression (0.003) and vertigo (0.03). In the majority of
the cases, the models predicted better specificity than sensitivity. Only in the case of
headaches and sleep disorder, did sensitivity prevail over specificity (Table 3 –
Page 46 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
15
15
classification table). In the extreme case, skin and vascular problems showed null or
minimum sensitivity and100 % specificity. Nagelkerke pseudo R-squared, showed
acceptable coefficients with the exception of the symptoms related with vertigo and skin
problems (Table 3).
Threshold cut-off values of GSM for sleep, attention, irritability, and memory are also
shown (Table 3). The remaining cut-off values were not considered since sensitivity or
specificity was reported at below 0.50 %.
The influence of other covariates on the GSM ORs coefficients, such as age, cellular
use, and concern about the BS were always less than 10% (Tables 2).
There was no observed multi-collinearity among variables. Kappa values according to
factor analysis were always lower than 2 and well below the critical value of 30.
Finally, no interactions between covariates were observed.
Standard errors and confidence intervals obtained by re-sampling were similar to those
calculated from the asymptotic approximation (Table 4). There was a small bias or
difference between the average bootstrap coefficients (not shown) and the respective
estimates obtained from the original sample.
There were no global health differences between those who permitted a bedroom
exposure measurement (88 in our last model) and those who refused RF measurements
(26), and these results were unaltered when using age as a covariate. Square partial eta
measured a 0% contribution of the willing participation variable to symptoms that
correlated with age such as: irritability, headaches, walking difficulties, and hearing loss.
There was no relationship between subjective distance to the BS and willingness to
participate (Pearson Chi- square = 2.80, df = 1, p=0.094).
Page 47 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
16
16
However, ANOVA showed that the group with recorded RF EMF levels was more prone
to complain of memory loss ( F = 5.07 ; p = 0.027); while participants without EMF
measures showed more skin problems ( F = 10.66; p = 0.001).
DISCUSSION
In the present re-analysis a more robust statistical method was employed that was
indifferent to the assumption of normality. To reduce the limitation of the sample size
effect and extrapolate our results to the entire population from which the sample was
obtained, a re-sample method or bootstrapping was used.
This new study partially confirms our preliminary results – namely, that most of the
symptoms related to GSM levels independently of the demographical variables and some
possible risk factors. Related to microwave radiation, the spectral power density analysis
maintained that the most important contribution to broadband measurements was from
GSM 900/1800, and the main variability of the measurements between different places
was due to a different coverage of the GSM 900/1800 signals, i.e. spatial variability. This
was further supported by the fact that the antenna used was fairly insensitive to
frequencies below 400 MHz. Therefore, the radio channels 80-110 MHz were not a
significant part of the broadband measurements. Moreover, the narrow band
measurements showed TV channels with substantially lower intensities than the GSM
900/1800 signals. The effects from these exposures will therefore not confound the effects
of BSs. Moreover, some authors [13] found that the only relevant contribution to the
variance of the high microwave exposure was from BSs – up to 93% of variance.
Moreover, at the time of our study the GSM signal was almost invariable in time because
there were very few calls. The main contribution was made from the broadcast channels
working almost constantly throughout the day. Short-range evaluations of exposure could
Page 48 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
17
17
be acceptable for describing a 24-hour period and the measurements were made in
bedrooms – a location where the participants were assumed to spend significant periods
of time.
However, some participants were mobile phone users at the time of this study and
exposure to a mobile phone during a phone call is much higher than that received from
BSs. Nevertheless, some authors [13] stated about that there is no a priori argument why
these lower levels should have no effect on the presence of a widespread use of mobile
telephones. Exposure to a BS will be at a low but almost constant level for many hours of
the day and especially at night.
While GSM exposure was associated with most of the symptoms, walking difficulties and
hearing loss was correlated only with age. Age also remained slightly inversely associated
with irritability. Users of cellular phones were more prone to complain of loss of appetite
and vertigo, while those who expressed worry about the BSs, were associated with sleep
problems. This later finding was in concordance with two other articles [13, 20, 26].
However, worry about the BSs was unrelated with age, gender, or subjective distance to
BSs. This agrees with an article [36] claiming that there was no statistically significant
association between symptom occurrence associated with perceived proximity to BSs,
psychological components, socio-demographic characteristics, and distance to BSs or
power lines.
Some authors indicated that opponents of mobile phone towers generally do not express
anxieties about electromagnetic field exposure, indicating that the risk rating is
comparable with other commonly perceived hazards in the modern world [37].
None of the analysed covariates behaved as confounders. The relationship of GSM
exposure with irritability, sleep troubles, lack of appetite, and vertigo remained
statistically significant despite the introduction of the above covariates.
Page 49 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
18
18
When the conventional multivariate analysis was tested using bootstrapping it was
observed that the standard error and confidence intervals obtained by re-sampling were
similar to those calculated from asymptotic approximation and this supports the
adequacy of our conventional analysis. Our sample, chosen at random, represents the
population from which it came.
The model appeared generally well adjusted while the cut-off values could constitute
good guidance for predicting the threshold of symptom appearance.
We cannot truly state that residents were more worried; equally worried, or less worried
than elsewhere in this region, since we cannot provide the percentage of those worried
about the BS masts in La Ñora compared with other nearby places. However,
information about this issue was widespread in this region at the time, and the
circumstances at La Ñora were shared with most other small urban and rural areas. The
sample was randomly selected but a participation bias cannot be ruled out since most of
our participants expressed fear regarding BSs and this could contribute to their
participation in the study. It is also possible to speculate that the percentage of participants
who refused to participate did so for the opposite reasons (indifference about BSs). In this
regard, neither health status nor subjective distance to the BS explained a willingness to
participate in the study.
Concerns about radiation from BSs were not related to age, sex, or subjective distance to
BSs. This agrees with statements from several authors [13] that living near a base
station does not make people generally fearful, but people that generally worry about
fields express stronger fears when they live close to a station.
Nevertheless, irrespective of these explanations, there seems to be effects of exposure that
occur independently of the fear felt by the participants, since controlling for fear did not
change the association between exposure and symptoms. However, the late query about
Page 50 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
19
19
concerns (as a possible confounder) may render the results less valid. In contrast to our
findings, note that biological grounds explaining non-thermal effects have not been
clearly established. Recently, it has been stated that voltage-gated calcium channels are
essential to the beneficial or adverse responses to microwave EMFs, nanosecond EMF
pulses, and static electrical and magnetic fields [38].
In summary, the results of this study indicate that effects of very low but long lasting
exposure to emissions from mobile telephone base stations on well-being cannot be ruled
out. The effects almost completely matched the symptoms described within the
microwave syndrome. Finally, unravelling the causal pathways would be best performed
with an experimental study design.
Page 51 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
20
20
CONCLUSIONS
This new study partially confirms our preliminary results about microwave sickness
resulting from exposure to emissions from GSM mobile phone BSs. Fatigue, irritability,
lack of appetite, sleep troubles, depression, and lack of concentration were especially
related with GSM exposure.
These results were independent of the main socio-demographic variables, other EMF
exposures, and anxiety about being irradiated. Nevertheless, we confirm that
apprehension about modern technology could predict some symptoms, especially those
related with sleep problems.
Our results agree with those who claimed that by distorting perceptions of risk,
disproportionate precaution might paradoxically lead to illness that would not otherwise
occur [39]. However, health changes related with GSM exposure seem to occur in a
manner unrelated with those fears. Finally, exposure was very low during the period and
also very low in comparison with Spanish recommendations [40] and international
guidelines [41].
Recommendations
We subscribe to the guidelines observed by other authors [42] in following the principle
of prevention while the non-thermal effects are not considered in any official standard.
This includes exposure minimisation within the limits of technical feasibility to
guarantee a significant reduction in long term radiation exposure to cellular phone
towers in residential areas. Epidemiological and clinical studies should continue to
observe possible health changes in the population. Finally, clear information about the
correct use of newer electronic devices should be implemented.
Page 52 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
21
21
Competing interests
The authors declare that they have no competing interests.
ACKNOWLEDGEMENTS
We would like to express our gratitude to Angeles Martinez Gomez for her assistance
during the fieldwork in La Ñora; as well as the Spanish Ministry of Science and
Technology for grant FIT number 070000-2002-58.
Page 53 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
22
22
REFERENCES
1. Lilienfeld AM, Tonascia J, Tonascia S, Libauer CH, et al. Foreign Service Health
Status Study, Evaluation Of Health Status Of Foreign Service and Other Employees
from Selected Eastern European Posts. Baltimore: Dept Epidemiology, School of
Hygiene and Public Health, The Johns Hopkins University; 1978.
2. Elwood JM. Microwaves in the cold war: the Moscow embassy study and its
interpretation. Review of a retrospective cohort study. Environ Health 2012; Nov
14;11:8.
3. Belyaev IY. Non-thermal Biological Effects of Microwaves. Microwave Review.
2005a; 11:13-29
4. Belyaev IY. Non-thermal Biological Effects of Microwaves: Current Knowledge,
Further Perspective, and Urgent Needs. Electromagn. Biol. Med. 2005b ; 24:375-
403.
5. Röösli M, Moser M, Baldinini Y, Meier M, Braun-Fahrländer C. Symptoms of ill
health ascribed to electromagnetic field exposure – a questionnaire survey. Int. J.
Hyg. Environ. Health. 2004 ; 207:141–150.
6. Frick U, Mayer M, Hauser S, Binder H , Rosner R , Eichhammer P. Entwicklung
eines deutschprachigen Messinsntrumentes für ‘‘Elektrosmog-Beschwerden’’.
Umwelt Med. Forsch. Prax. 2006 ; 11:103–113.
7. Eltiti S, Wallace D, Ridgewell A, Zougkou K, Russo R, Sepulveda F, et al. Does
short-term exposure to mobile phone base station signals increase symptoms in
individuals who report sensitivity to electromagnetic fields? A double-blind
randomised provocation study. Environ Health Perspect 2007;115:1603-1608.
Page 54 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
23
23
8. Eltiti S, Wallace D, Ridgewell A, Zougkou K, et al. Short-term exposure to mobile
phone base station signals does not affect cognitive functioning or physiological
measures in individuals who report sensitivity to electromagnetic fields and
controls. Bioelectromagnetics 2009;30:556-63.
9. Khurana VG, Hardell L, Everaert J, Bortkiewicz A, et al. Epidemiological
evidence for a health risk from mobile phone base stations. Int J Occup Environ
Health 2010; 16(3):263-7.
10. Navarro EA, Segura J, Portolés M, Gómez-Perretta C. The Microwave Syndrome:
A Preliminary Study in Spain. Electromagn Biol Med 2003; 22:161 – 169.
11. Santini R, Santini P, Danze J, Le Ruz P, et al. Study of the health of people living
in the vicinity of mobile phone base stations: I. Influences of distance and sex.
Pathol Biol 2002; 50:369 – 373.
12. Gadzicka E, Bortkiewicz A, Zmyslony M, Szymczak W, et al.. Assessment of
subjective complaints reported by people living near mobile phone base stations
[Abstract]. BiuletynPTZE Warsaw. 2006; 14:23-26.
13. Hutter H, Moshammer H, Wallner P, Kundi M. Subjective symptoms, sleeping
problems, and cognitive performance in subjects living near mobile phone base
stations. Occup Environ Med 2006; 63:307–313.
14. Abdel-Rassoul G, El-Fateh OA, Salem MA, Michael A, Farahat F, et al.
Neurobehavioral effects among inhabitants around mobile phone base stations.
Neurotoxicology 2007; 28:434-440.
15. Johnson-Liakouris AG. Radiofrequency (RF) Sickness in the Lilienfeld Study: an
effect of modulated microwaves? Arch Environ Health 1998;53:236-238.
Page 55 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
24
24
16. Thomas S, Kühnlein A, Heinrich S, Praml G, et al. Personal exposure to mobile
phone frequencies and well-being in adults: a cross-sectional study based on
dosimetry. Bioelectromagnetics 2008; 29:463-470.
17. Kühnlein A, Heumann C, Thomas S, Heinrich S, et al. Personal exposure to mobile
communication networks and well-being in children--a statistical analysis based on
a functional approach. Bioelectromagnetics 2009; 30:261-9.
18. Heinrich S, Thomas S, Heumann C, von Kries R, et al. Association between
exposure to radiofrequency electromagnetic fields assessed by dosimetry and acute
symptoms in children and adolescents: a population based cross-sectional study.
Environ Health 2010; 9:75.
19. Blettner M, Schlehofer B, Breckenkamp J, Kowall B, et al Mobile phonebase
stations and adverse health effects: Phase 1 of a populationbased,cross-sectional
study in Germany. Occup Environ Med.2009;66:118-123.
20. Berg-Beckhoff G, Blettner M, Kowall B, Breckenkamp J, et al. Mobile phone base
stations and adverse health effects: phase 2 of a cross-sectional study with measured
radio frequency electromagnetic fields. Occup Environ Med 2009; 66:124-130.
21. Augner C, Florian M, Pauser G, Oberfeld G, et al. GSM Base Stations: Short-Term
Effects on Well-Being. Bioelectromagnetics 2009; 30:73-80.
22. Furubayashi T, Ushiyama A, Terao Y, Mizuno Y, et al. Effects of short-term W-
CDMA mobile phone base station exposure on women with or without mobile
phone related symptoms. Bioelectromagnetics 2009; 30:100-113.
23. Zwamborn AP, Vossen SH, van Leersum BJ. Effects of global communication
system radio-frequency fields on well being and cognitive functions of human
subjects with and without subjective complaints. Netherlands Organisation for
Applied Scientific Research (TNO) 2003;148:1-89.
Page 56 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
25
25
24. Augner C, Hacker GW, Oberfeld G, Florian M, et al. Effects of exposure to GSM
mobile phone base station signals on salivary cortisol, alpha-amylase, and
immunoglobulin A. Biomed Environ Sci 2010;23:199-207.
25. Eger H, Jahn M. Spezifische Symptome und Mobilfunkstrahlung in Selbitz
(Bayern) Evidenz für eine Dosiswirkungsbeziehung. Umwelt·medizin·gesellschaft
2010; 2: 130-139.
26. Danker-Hopfe H, Dorn H, Bornkessel C, Sauter C. Do mobile phone base stations
affect sleep of residents? Results from an experimental double-blind sham-
controlled field study. Am J Hum Biol. 2010; 5:613-8
27. Bortkiewicz A, Gadzicka E, Szyjkowska A, et al. Subjective complaints of people
living near mobile phone base stations in Poland. Int J Occup Med Environ Health
2012;25:31-40
28. Eskander EF, Estefan SF, Abd-Rabou AA. How does long term exposure to base
stations and mobile phones affect human hormone profiles? Clin Biochem 2012 ;
45(1-2):157-61
29. Shahbazi-Gahrouei D, Karbalae M, Moradi HA, et al. Health effects of living near
mobile phone base transceiver station (BTS) antennae: a report from Isfahan, Iran.
Electromagn Biol Med. 2013 Jun 19.
30. Röösli M, Frei P, Mohler E, Hug K. Systematic review on the health effects of
exposure to radiofrequency electromagnetic fields from mobile phone base stations.
Bull World Health Organ 2010; 88: 887–896.
31. Efron B, Tibshirani R: An Introduction to the Bootstrap. Boca Raton, FL: Chapman
& Hall/CRC 1993; ISBN 0-412-04231-2.software
32. Maxwell SE, Delaney HD. Bivariate median-splits and spurious statistical
significance. Psychological Bulletin 1993; 113:181–190.
Page 57 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
26
26
33. Royston P, Altman DG, Sauerbrei W.Dichotomizing continuous predictors in
multiple regression: a bad idea. Stat. Med 2006; 25:127-141.
34. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous
confounding variable is categorized in logistic regression analyses. Stat. Med 2004;
23:1159–1178.
35. Peduzzi P, Concato J, Kemper E, Holford TR, et al. A simulation study of the
number of events per variable in logistic regression analysis. J Clin Epidemiol 1996;
49:1373-1379.
36. Baliatsas C, van Kamp I, Kelfkens G, et al. Non-specific physical symptoms in
relation to actual and perceived proximity to mobile phone base stations and
powerlines. BMC Public Health 2011; 11:421
37. Hutter H, Moshammer H, Wallner P, Kundi M. Public perception of risk concerning
celltowers and mobile phones. Soz Präventivmed 2004;49: 62–66.
38. Pall ML. Electromagnetic fields act via activation of voltage-gated calcium
channels to produce beneficial or adverse effects. J. Cell. Mol. Med 2013; 7:. 958-
965
39. Coggon D. Health Risks from Mobile Phone Base Stations. Occup Environ Med
2006; 63:298–299.
40. Real Decreto 1066/2001 Telecomunicaciones/TelefoniaMovil/RD1066_2001.pdf
http://www.mityc.es/esES/OficinaVirtual/Documents/SE%20
41. International Commission on Non-Ionizing Radiation Protection: Guidelines for
limiting exposure to time-varying electric, magnetic, and electromagnetic fields (up
to 300 GHz). Health Phys 1998; 74: 494–522.
Page 58 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
27
27
42. Haumann T, Münzenberg U, Maes W, Sierck P. HF-Radiation levels of GSM
cellular phone towers in residential areas. 2nd International Workshop on Biological
effects of EMFS. October 2002; Rhodes (Greece) Vol 1:327-333.
Page 59 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
28
28
Table 1. Univariate ORs & 95% CIs of all clinical symptoms related with various possible confounders.
Symptom/
Variable
Worry about BSs (1) Computer (not use) Mobile (not use) OR (95%CI) OR (95%CI) OR (95%CI)
Fatigue 0.67 0.23-1.90 2.62 0.76-9.04 1.56 0.56-4.35 Irritability 1.13 0.43-3.03 1.56 0.45-5.34 2.62 0.94-7.33 Headaches 1.75 0.62-4.94 1.39 0.34-5.58 1.56 0.51-4.83 Nausea 0.68 0.18-2.24 0.34 0.04-2.84 1.43 0.44-4.67 Lack of appetite 1.05 0.33-3.40 3.16 0.87-11.44 4.28** 1.43-12.78 Trouble sleeping 3.12* 1.10-8.86 0.55 0.16-1.88 0.74 0.27-2.02 Depression 1.06 0.39-2.93 0.81 0.22-2.93 1.03 0.38-2.84 Lack of concentration 0.92 0.35-2.47 1.11 0.33-3.76 2.79 0.99-7.80 Memory loss 1.71 0.62-4.75 0.41 0.10-1.64 1.35 0.50-3.61 Skin alterations 0.74 0.23-2.35 φ φ 0.63 0.16-2.45
Visual disturbances 1.31 0.48-3.60 0.77 0.21-2.77 1.63 0.60-4.39 Vertigo 0.61 0.20-1.91 0.77 0.19-3.10 2.90* 1.04-8.07 Vascular alteration 0.96 0.27-3.43 1.48 0.35-6.17 2.04 0.65-6.41 Hearing problems 0.59 0.20-1.70 0.77 0.19-3.10 0.48 0.15-1.60 Walking difficulty 0.60 0.20-1.79 φ φ 0.42 0.11-1.60 * p<0.05; ** p<0.01; (1) Not worried, as reference code (2) & (3). No device use, as reference code,φ any subject affected using computer
Page 60 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
29
29
Table 2. ORs & 95% CIs for GSM exposure: increase in risk per increase in Log GSM (µW/m2). Changes in -2 Log likelihood is also shown. The third column represents the ORs for a 10 fold increase in GSM (Log10 GSM)
Symptom OR (95%CI)
Change in -2 Log likelihood OR (95%CI)
Fatigue 1.39***(1.14-1.70) 11.74*** 2.13***(1.34-3.83) Irritability 1.51***(1.23-1.85) 19.36*** 2.58***(1.61-4.12) Irritability (adjusted with age)
1.47***(1.20-1.81) - 2.44***(1.52-3.94)
Headaches 1.43** (1.15-1.78) 12.32*** 2.28**(1.37-3.78) Nausea 1.38** (1.09-1.73) 8.3** 2.09**(1.23-3.55) Lack of Appetite 1.58** (1.23-2.03) 16.31*** 2.86***(1.60-5.09) Lack of Appetite (adjusted to cellular use)
1.53** (1.19-1.99) - 2.68***(1.48-4.84)
Trouble sleeping 1.49***(1.20-1.84) 16.38*** 2.49***(1.52-4.08) Trouble sleeping (adjusted to worry to BSs)
1.64***(1.22-2.19) - 3.11***(1.59-6.09)
Depression 1.41***(1.16-1.72) 13.99*** 2.22***(1.42-3.48) Concentration 1.54***(1.25- 1.89) 20.75*** 2.68***(1.67-4.32) Memory 1.27** (1.06- 1.52) 7.29** 1.73**(1.14-2.60) Skin 1.24* (1.001-1.54) 4.08* 1.65*(1.01-2.71) Visual 1.23 * (1.03-1.46) 5.30* 1.59*(1.06-2.40) Vertigo 1.36** (1.11-1.66) 10.14*** 2.02**(1.28-3.20) Vertigo (adjusted to cellular use)
1.32** (1.08-1.62) - 1.91**(1.20-3.04)
Vascular 1.32* (1.05-1.64) 6.30* 1.88*(1.12-3.14) Hearing 0.96 (0.80-1.15) 0.90(0.59-1.37) Walking 0.95 (0.78-1.15) 0.88(0.57-1.37) * p<0.05; ** p<0.01; *** p<0.001
Page 61 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
30
30
Table 3. Goodness-of-fit of the outcome binary response variable related to GSM exposure (Log=Ln). Cut-off values of exposure to microwaves according to ROC analysis. The data is presented in ascending order. Symptom ROC curves
Area
Classification Table (a)
Pseudo-R**2
(1)
Cut-off (2) (Log GSM)
Cut-off (2) GSM (µW/m2)
SSV SPF AV
Headaches 0.84*** 0.90 0.23 0.72 0.41 - 1.77 Sleep 0.78*** 0.82 0.66 0.76 0.28 1.66 5.26 Attention 0.78*** 0.67 0.72 0.69 0.28 3.61 36.97 Irritability 0.76*** 0.67 0.73 0.71 0.26 3.61 36.97 Memory 0.67** 0.54 0.77 0.67 0.11 4.99 146.94 Depression 0.75*** 0.46 0.76 0.65 0.20 - 184.93 Visual 0.65* 0.24 0.83 0.60 0.08 - 368.71 Fatigue 0.73*** 0.22 0.90 0.69 0.18 - 685.4 Vertigo 0.74*** 0.16 0.87 0.67 0.19 - 685.4 Appetite 0.87*** 0.40 0.94 0.85 0.43 - 1495.18 Nausea 0.86*** 0.46 0.93 0.87 0.38 - 1495.18 Vascular 0.85*** 0.20 1.0 0.90 0.34 - 3041.18 Skin 0.67* 0.00 1.00 0.81 0.072 - 8604.15 * p<0.05, ** p<0.01, *** p<0.001 (1) Nagelkerke (2) Cut-off (ROC curve): Only values showing SSV and SPF above 0.5 are reported (a) SSV=sensitivity, SPF =specificity, AV=average
Page 62 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
31
31
Table 4. Statistics for r = 1000 bootstrapped binary logistic regression (GSM exposure coefficients: Increase in risk per increase in Log GSM (µW/m2). Asymptotic standard errors (SE) and 95% confidence intervals (CI) are also shown for comparison.
Symptom B*
Bootstrap Normal
Bias SE 95% percentile
intervals SE 95% CI
Lower Upper Lower Upper Fatigue 0.329 0.012 0.097 0.155 0.539 0.102 0.128 0.529 Irritability 0.411 0.016 0.110 0.241 0.670 0.104 0.207 0.615 Headache 0.358 0.022 0.139 0.149 0.688 0.113 0.137 0.578 Nausea 0.319 0.013 0.124 0.099 0.590 0.118 0.088 0.550 Appetite 0.456 0.026 0.134 0.264 0.784 0.128 0.205 0.707 Sleep 0.396 0.022 0.124 0.193 0.690 0.109 0.181 0.610 Depression 0.346 0.012 0.102 0.174 0.583 0.100 0.151 0.541 Attention 0.429 0.020 0.118 0.254 0.711 0.106 0.222 0.636 Memory 0.237 0.009 0.098 0.057 0.448 0.091 0.058 0.415 Skin 0.217 0.008 0.110 0.011 0.451 0.110 0.001 0.433 Visual 0.203 0.004 0.093 0.037 0.398 0.090 0.026 0.379 Hearing -0.05 -0.002 0.089 -0.219 0.143 0.093 -0.228 0.135 Vertigo 0.306 0.010 0.101 0.127 0.530 0.102 0.107 0.505 Walking -0.05 -0.006 0.098 -0.265 0.120 0.098 -0.246 0.138 Vascular 0.274 0.010 0.109 0.084 0.520 0.114 0.051 0.497 * beta coefficent (Log OR)
Page 63 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
32
32
FIGURE CAPTIONS
Figure 1. Distribution of EMF measurement throughout the sample.
Page 64 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from
For peer review only
Distribution of EMF measurement throughout the sample 106x90mm (300 x 300 DPI)
Page 65 of 65
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml
BMJ Open
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960
on April 19, 2021 by guest. P
rotected by copyright.http://bm
jopen.bmj.com
/B
MJ O
pen: first published as 10.1136/bmjopen-2013-003836 on 30 D
ecember 2013. D
ownloaded from