For a list of the members of the QUALIFEX-team please see the Acknowledgments section.
Conceived and designed the experiments: MR CBF. Performed the experiments: EM PF. Analyzed the data: EM PF MR. Contributed reagents/materials/analysis tools: EM PF JF MR. Wrote the paper: EM PF JF CBF MR. Maintained and calibrated the measurement devices: JF. Contributed to the interpretation of the data analysis: EM PF JF CBF MR.
The authors have declared that no competing interests exist.
There is persistent public concern about sleep disturbances due to radiofrequency electromagnetic field (RF-EMF) exposure. The aim of this prospective cohort study was to investigate whether sleep quality is affected by mobile phone use or by other RF-EMF sources in the everyday environment.
We conducted a prospective cohort study with 955 study participants aged between 30 and 60 years. Sleep quality and daytime sleepiness was assessed by means of standardized questionnaires in May 2008 (baseline) and May 2009 (follow-up). We also asked about mobile and cordless phone use and asked study participants for consent to obtain their mobile phone connection data from the mobile phone operators. Exposure to environmental RF-EMF was computed for each study participant using a previously developed and validated prediction model. In a nested sample of 119 study participants, RF-EMF exposure was measured in the bedroom and data on sleep behavior was collected by means of actigraphy during two weeks. Data were analyzed using multivariable regression models adjusted for relevant confounders.
In the longitudinal analyses neither operator-recorded nor self-reported mobile phone use was associated with sleep disturbances or daytime sleepiness. Also, exposure to environmental RF-EMF did not affect self-reported sleep quality. The results from the longitudinal analyses were confirmed in the nested sleep study with objectively recorded exposure and measured sleep behavior data.
We did not find evidence for adverse effects on sleep quality from RF-EMF exposure in our everyday environment.
In the last two decades, emerging wireless technologies like mobile or cordless phones have led to increasing exposure to radiofrequency electromagnetic fields (RF-EMF) in everyday life
Several randomized, double blind studies addressed the question whether short-term RF-EMF exposure affects sleep measures such as brain activity recorded by means of electroencephalography (EEG). Most of the studies were conducted in a laboratory setting applying well controlled exposure conditions mimicking a mobile phone handset exposure during 30 to 45 minutes
Mobile and cordless phones produce a relatively high exposure to the head but not to the rest of the body as EMF is rapidly decreasing with distance
Thus, there is an urgent need for a prospective cohort study on sleep quality addressing all aspects of RF-EMF exposure in our everyday life, which includes exposure to environmental far-fields (e.g. mobile phone base stations) and exposure to sources close to the body localized to the head (mobile and cordless phone use). The aim of this study was to investigate a possible association between different objective RF-EMF exposure surrogates and self-reported sleep quality in a large sample (longitudinal study) and to check the consistency of the results in a subsample with measured RF-EMF exposure and measured sleep behavior data (nested sleep study). Main characteristics of these two study components are presented in
Study characteristics | Longitudinal study | Nested sleep study |
Number of participants | 955 |
119 |
Outcomes |
|
|
- sleep duration | ||
- sleep disturbances | - sleep efficiency | |
|
||
- restfulness of sleep | ||
- wellbeing in the morning | ||
Exposure measures |
|
|
- mobile phone use | - everyday life exposure to all sources (during one typical working day) | |
- cordless phone use | - night-time exposure to all sources in the bedroom | |
|
- fixed site transmitter exposure in the bedroom | |
- mobile phone use | ||
|
||
- everyday life exposure to all sources | ||
- night-time exposure to all sources in the bedroom | ||
- fixed site transmitter exposure in the bedroom | ||
Type of data analysis |
|
|
- cohort analysis | - random effect regression models with a 1-day lag autocorrelation term | |
- change analysis |
After exclusion of nightshift workers (n = 89) and users of sleeping drugs (n = 81).
1 person was excluded because of sleeping drug consumption during all 14 nights.
Ethical approval for this study was received from the Ethical Commission of Basel on March 19th, 2007 (EK: 38/07). Written informed consent was obtained from the participants of the nested sleep study and of the participants of the longitudinal study for providing the mobile phone operator data.
For the present study, we invited 3763 residents from the Basel area (Switzerland) randomly selected from communal population registries. Eligible participants were between 30 and 60 years old, Swiss residents or people who lived in Switzerland for at least five years. A baseline survey was conducted in May 2008 and the follow-up in May 2009. Information was collected on sleep quality, possible confounders and relevant exposure predictors including use of mobile and cordless phones. Exclusion criteria for the analyses of sleep data presented in this paper were regular usage of sleeping pills and night shift working either at the baseline or follow-up survey.
In the written questionnaire of the baseline and the follow-up questionnaire, we used seven items of the Epworth Sleepiness Scale
Due to the unknown mechanism of radiofrequency electromagnetic radiation on biological organisms, we used six different exposure surrogates to assess far field exposure and exposure from sources operating close to the body. With respect to local exposure to the head (close to body exposure), we asked participants in the written questionnaire about their average mobile and cordless phone use per week during the past six months. Informed consent was also sought from participants to obtain their mobile phone connection data for the previous six months of each survey from the three Swiss mobile phone network operators (operator data).
For far field exposure, we used a three-dimensional geospatial propagation model in which average RF-EMF from fixed site transmitters (mobile phone base stations and broadcast transmitters) was modeled for the apartment of each study participant
From the responders of the baseline cohort survey, 120 participants were selected for a nested sleep study. We did not recruit persons with children less than two years, people who had experienced a long distance flight within the last three weeks, people with severe illnesses, people who regularly consumed sleeping pills and shift workers. We used our exposure prediction model to oversample highly exposed persons to maximize the exposure range in the nested sleep study.
In the participants of the nested sleep study, sleep behavior was measured by means of a wrist actigraphic device (AW7, Cambridge Neurotechnology) with an epoch length of 15 seconds during two weeks. Participants were asked to wear this device on the non-dominant wrist during two weeks and were advised to press an event marker when trying to fall asleep or getting up. They also received a sleep diary, which they had to fill in every morning and every evening. This diary was based on the sleep diary suggested by the German Society of Sleep Medicine (
Actigraphic data were analyzed using the software provided by the manufacturer. A study assistant checked the night data for artifacts and the diary data were systematically used for data quality control. Nights in which participants forgot to wear the actigraphic device were replaced with the data from the sleep diary. We excluded from the data analysis nights during which a switching from daylight saving time to regular time and vice versa took place, nights when participants slept at another place or nights with sleeping pill consumption. Two sleep parameters were extracted from the actigraphic measurements: total sleep duration and sleep efficiency. Definitions of these parameters are given in
Parameter | Definition | Data sources | Min. | Median | 90th perc. | Max. | |
|
Epworth sleepiness scale (ESS) | Excessive daytime sleepiness | Questionnaire | 0 | Baseline: 5 | 10 | 19 |
Follow-up: 4 | 9 | 21 | |||||
Sleep disturbance score | Difficulties with falling asleep, fitful sleep, waking phases during night and awaking too early in the morning | Questionnaire | 0 | Baseline: 5 | 9 | 12 | |
Follow-up: 5 | 8 | 12 | |||||
|
Total sleep duration [h] | Time from sleep onset to sleep end excluding waking phases | Actigraphy | 4.8 | 7.1 | 8.0 | 9.7 |
Sleep efficiency [%] | Percentage of time in bed with the intention to sleep that a person sleeps | Actigraphy | 79.0 | 91.2 | 95.1 | 96.9 | |
Restfulness of sleep | How restful was your sleep? 1 “very restless sleep” to 5 “very restful sleep” | Sleep diary | 2.8 | 4 | 4.6 | 5 | |
Well-being in the morning | How do you feel now? 1 “depressed” to 6 “easygoing” | Sleep diary | 1.5 | 4.5 | 5.8 | 6 |
For the sleep study, all estimates are given for the level of the individual (i.e. average over two weeks).
Exposure to all relevant sources of radiofrequency electromagnetic fields was measured with the EME SPY120 (Satimo, Courtaboeuf, France). Exposure measures were taken every 90 seconds during the first week of the measurement period (two weeks). The exposure meter device (exposimeter) was placed in the sleeping room near the bed and the head of the participants. During one typical working day participants were requested to wear the exposimeter to estimate their daytime exposure. Mean exposure values were calculated for measurements in the sleeping room during the night, for fixed site transmitter measurements in the sleeping room and for measurements during the day on which the exposimeter was carried around. Mean values were calculated using regression on order statistics, which allows for nondetects
In the longitudinal study, the association between exposure and outcome was calculated by means of linear regression models. We conducted two different analyses: I) A cohort analysis, where we assessed the association between exposure at baseline and the change in self-reported sleep quality within one year. Three exposure categories were defined a priori for each exposure metric: <50th percentile, 50th to 90th percentile, >90th percentile. II) A change analysis, where we examined whether the change in exposure between baseline and follow-up resulted in a change in self-reported sleep quality. For the change analysis we compared the participants with the 20% largest exposure increase and decrease between baseline and follow-up survey with all other participants who experienced a smaller or no change of exposure between baseline and follow-up survey (reference group). All models were adjusted for age, sex, body mass index, stress level, physical activity per week, smoking status, alcohol consumption, education level, marital status, degree of urbanity, belief in health effects due to RF-EMF exposure, noise annoyance and for moving house between the two surveys. About 20% of the participants in each survey reported to be electro-hypersensitive (EHS) or reported that they thought that they developed detrimental health symptoms due to electromagnetic pollution in everyday life
In the nested sleep study, we used a random intercept mixed regression model with an autocorrelation term of one-day lag to analyze the association between sleep measures and RF-EMF exposure. All models were adjusted for sex, age, smoking status, body mass index, weekday, percent fulltime equivalent, educational level, presence of a bed partner, weekday and the diary-based variables bedtime, alcohol intake within 4 hours before going to bed, physical activity during the day, and sleeping during the day (more information on the confounders is given in the footnote in
Linear multilevel model |
|||||
Exposure range [mW/m2] | n (individuals) |
n (nights) | Coeff. | (95%-CI) | |
|
|||||
|
|||||
<median | 0.00 to 0.11 | 60 | 777 | 0.00 | |
50.–90. percentile | 0.11 to 0.42 | 48 | 616 | 0.07 | (−0.18;0.32) |
>90. percentile | 0.45 to 16.69 | 11 | 158 | 0.19 | (−0.21;0.60) |
|
|||||
<median | 0.00 to 0.03 | 60 | 763 | 0.00 | |
50.–90. percentile | 0.03 to 0.12 | 48 | 624 | 0.16 | (−0.09;0.41) |
>90. percentile | 0.12 to 2.18 | 11 | 164 | 0.16 | (−0.24;0.56) |
|
|||||
<median | 0.00 to 0.01 | 60 | 778 | 0.00 | |
50.–90. percentile | 0.02 to 0.06 | 48 | 622 | 0.07 | (−0.17;0.32) |
>90. percentile | 0.08 to 1.39 | 11 | 151 | 0.00 | (−0.43;0.43) |
|
|||||
|
|||||
<median | 0.00 to 0.11 | 60 | 777 | 0.00 | |
50.–90. percentile | 0.11 to 0.42 | 48 | 616 | 1.21 | (−0.02;2.44) |
>90. percentile | 0.45 to 16.69 | 11 | 158 | 0.43 | (−1.54;2.41) |
|
|||||
<median | 0.00 to 0.03 | 60 | 763 | 0.00 | |
50.–90. percentile | 0.03 to 0.12 | 48 | 624 | 0.80 | (−0.41;2.01) |
>90. percentile | 0.12 to 2.18 | 11 | 164 | −0.67 | (−2.60;1.27) |
|
|||||
<median | 0.00 to 0.01 | 60 | 778 | 0.00 | |
50.–90. percentile | 0.02 to 0.08 | 48 | 622 | 0.80 | (−0.40;1.99) |
>90. percentile | 0.10 to 1.40 | 11 | 151 | −1.04 | (−3.11;1.02) |
adjusted for: age, percent fulltime equivalent, bedtime (derived from diary) (all linear), sex, body mass index (<25, ≥25), smoking status, weekday (weekend vs. workday), presence of a bed partner, alcohol intake within 4 hours before going to bed (diary), physical activity during the day (diary), sleeping during the day (diary) (all binary), and educational level (3 categories).
The division into the exposure categories was done on the individual level.
All statistical analyses were carried out using STATA 10.1 (StataCorp, College Station, TX, USA).
In total, 1375 participants filled in the baseline questionnaire in 2008 and 1125 subjects filled in the follow-up questionnaire one year later (response rate 82%). 170 participants were excluded from the longitudinal analyses due to night shift working (89 participants) and consumption of sleeping pills (81 participants). The analyses of our longitudinal study were therefore performed with 955 subjects. Detailed information on the characteristics of the study participants are described in
|
% |
|
|
|
Age (years) | ||||
30–40 | 224 | 24 |
|
|
41–50 | 329 | 34 |
|
|
51–60 | 402 | 42 |
|
|
Sex | ||||
Female | 578 | 61 |
|
|
Male | 377 | 39 |
|
|
Health status |
||||
Very good | 323 | 34 |
|
|
Good | 530 | 56 |
|
|
Half-half | 83 | 9 |
|
|
Bad | 8 | 1 |
|
|
Very bad | 1 | <1 |
|
|
Educational level |
||||
None | 51 | 5 |
|
|
Apprenticeship | 456 | 48 |
|
|
Higher education/University | 448 | 47 |
|
|
Self-reported electromagnetic hypersensitivity |
||||
Yes | 195 | 20 |
|
|
No | 760 | 80 |
|
|
Owning a mobile phone |
||||
Yes | 909 | 95 |
|
|
No | 41 | 4 |
|
|
Owning a cordless phone |
||||
Yes | 800 | 84 |
|
|
No | 150 | 16 |
|
|
Owning wireless LAN |
||||
Yes | 390 | 41 |
|
|
No | 558 | 59 |
|
|
Data may not sum up to 100% due to missing data.
Answering yes to either “Are you electro hypersensitive?” or “Do you think that you develop detrimental health symptoms due to electromagnetic pollution in everyday life?”
In the nested sleep study, age and gender distribution were comparable with participants of the longitudinal study (
Exposure at baseline | Change (between baseline and follow-up) | |||
Close to body exposure | ||||
|
<Median | 0.00 to 0.23 | Decrease | −11.67 to −0.15 |
50th–90th percentile | 0.23 to 3.50 | No relevant change | −0.13 to 0.15 | |
>90th percentile | 3.50 to 17.5 | Increase | 0.15 to 17.50 | |
<Median | 0.00 to 0.15 | Decrease | −2.85 to −0.18 | |
50th–90th percentile | 0.16 to 1.30 | No relevant change | −0.17 to 0.04 | |
>90th percentile | 1.33 to 8.61 | Increase | 0.04 to 1.49 | |
|
<Median | 0.00 to 0.35 | Decrease | −9.27 to −0.58 |
50th–90th percentile | 0.93 to 4.67 | No relevant change | −0.35 to 0.58 | |
>90th percentile | 9.33 to 9.33 |
Increase | 0.87 to 9.33 | |
Far field exposure | ||||
|
<Median | 0.00 to 0.12 | Decrease | −0.14 to −0.02 |
50th–90th percentile | 0.12 to 0.17 | No relevant change | −0.02 to 0.03 | |
>90th percentile | 0.17 to 0.41 | Increase | 0.03 to 0.18 | |
|
<Median | 0.00 to 0.00 | Decrease | −0.23 to −0.00 |
50th–90th percentile | 0.00 to 0.04 | No relevant change | −0.00 to 0.00 | |
>90th percentile | 0.05 to 0.40 | Increase | 0.00 to 0.23 | |
|
<Median | 0.00 to 0.01 | Decrease | −0.16 to −0.00 |
50th–90th percentile | 0.01 to 0.05 | No relevant change | −0.00 to 0.00 | |
>90th percentile | 0.05 to 1.43 | Increase | 0.00 to 0.62 |
For the change analysis we compared the participants with the 20% largest exposure increase and decrease between baseline and follow-up survey with all other participants, who experienced a smaller or no change of exposure between baseline and follow-up survey (no relevant change).
n = 389 at baseline (cohort analyses) and n = 245 at follow-up (change analyses).
equal values due to the use of categories in the questionnaire.
Measured exposure levels of the nested sleep study are presented in
Median daytime sleepiness and sleep disturbances scores per individual at baseline and follow-up are presented in
An increase in score refers to an increase in daytime sleepiness. * indicates statistical significance. All models are adjusted for age, body mass index, stress level, physical activity, noise annoyance (all linear), sex, alcohol consumption, belief in health effects due to RF-EMF exposure, smoking status, degree of urbanity, moving house between the two surveys (all binary), educational level, marital status (categorical). a) for a subsample of 363 (225) subjects who consented that we receive data from the operator at baseline (follow-up). b) In the change analysis a decrease and increase in exposure refers to the participants with the 20% largest exposure decrease and increase between baseline and follow-up survey. No relevant change includes all other participants, who experienced a smaller or no change of exposure (reference group).
An increase in score refers to an increase in sleep disturbances. * indicates statistical significance. Confounders see
Measured arithmetic mean sleep duration per individual was 6.9 hours (h) during weekdays (range: 4.9 h to 9.4 h) and 7.8 h during weekends (range: 4.5 h to 11.9 h) (
This study did not find indications for an association between typical levels of RF-EMF exposure in an everyday environment and self-reported sleep disturbances or excessive daytime sleepiness considering an exposure period of one year. These results were confirmed in a subsample of 119 study participants with data on sleep behavior measured with actigraphic devices and measured RF-EMF exposure.
To the best of our knowledge this is the first longitudinal study investigating the association between RF-EMF exposure and self-reported sleep quality in a large population sample using objectively recorded exposure data and data on sleep behavior measured with actigraphic devices. The cohort design allows for more robust conclusions, particularly because participation rate in the follow-up survey was rather high (82%). Therefore, in the present cohort and change analyses of the longitudinal study selection bias is expected to be of minor concern.
We applied a comprehensive exposure assessment method. All RF-EMF sources relevant in our everyday environment are included in the model and also personal exposure relevant behaviors are considered. The prediction models of the longitudinal study are based on extensive measurements with personal dosimetric devices. For the development of these prediction models we used weekly measurements of 166 persons and conducted a validation study by repeating the exposure measurements in 31 study participants 21 weeks later on average. In this validation study agreement between personal measurements and the prediction model for everyday exposure was found to be good (Spearman rank correlation: 0.75 (95%-CI 0.53–0.87), sensitivity: 0.67 and specificity 0.96)
The subjective sleep parameters in the longitudinal study might be considered a weakness of this study. However, we used standardized questions to assess daytime sleepiness and sleep disturbances. Subjectively perceived sleep quality is an established factor influencing personal well-being and is thus health relevant
With respect to self-reported outcome measures, information bias may be of concern if study participants are aware of their exposure status. For instance, individuals who consider themselves as exposed to mobile phone base station radiation may claim to suffer more often from sleep disturbances. There is some evidence from laboratory trials that more symptoms are reported in open provocations where participants were aware about the exposure status than in subsequent double blind provocations
We did not find an association between self-reported sleep quality and prolonged exposure to RF-EMF. Our findings are in line with results of cross-sectional surveys about RF-EMF exposure and self-reported sleep quality, which used spot measurements to assess exposure
We conducted a large number of analyses because in the absence of a known biological mechanism in the low dose range, it was unclear which aspect of exposure might be relevant for sleep disturbances, if any at all. We simultaneously took into account exposure from sources close to the body, producing high, localized and short-term exposures, as well as sources further away, which typically cause lower, more homogenous long-term exposures. Since mobile phone base stations are the EMF source people in Switzerland are most concerned about
Given the absence of an observed association, non-differential exposure misclassification may be of concern. For such ubiquitously distributed exposure sources, some exposure misclassification is unavoidable although we have put considerable effort in validating our methods. Non-differential exposure misclassification is expected to shift the regression coefficients towards zero if there is a true association. Nevertheless, assuming there is a true association, we would expect to see a non-significant exposure-response pattern consistently pointing towards an association. However, this was not observed in our study neither in the direction of a harmful nor in the direction of a beneficial effect. For interpretation of this and similar studies on symptoms, a “healthy communicator effect” may be relevant. Healthy communicator effect refers to the possibility that healthy people may use more often wireless communication devices and thus may be more exposed than ill people. It can thus be considered an analogy to the well known healthy worker effect.
In our study we observed relatively low far-field exposure levels. The levels were far below current standard limits
We found no evidence that individuals who reported to react sensitively to EMF (electromagnetic hypersensitivity) were more vulnerable to RF-EMF exposure than the rest of the population. This is in line with reported randomized double blind provocation studies addressing short term effects
Our longitudinal study captured a latency period of one year. It is not clear whether such a period is sufficient for sleep effects to manifest. Thus, we cannot completely rule out that our study has missed sleep effects that occur after prolonged exposure duration. However, most individuals who reported sleep disturbances in relation to mobile phone base station exposure claimed that such symptoms have occurred within a few days or weeks after a new exposure source was put into operation
Overall, we did not find an association between self-reported sleep quality and everyday RF-EMF levels from various sources over one year. By applying a longitudinal design and using objective exposure and measured outcome data, this study increases evidence for the true absence of an effect of everyday RF-EMF exposure on sleep quality.
We thank Fabian Trees from the Swiss Federal Statistical Office for providing the geographical coordinates of the study participants and the statistical department of Basel for providing the addresses of the study participants. We are grateful for the mobile phone connection data from the three Swiss network operators. Many thanks go to all study participants who volunteered for the study.
• Georg Neubauer, Austrian Institute of Technology, Safety and Security Department, A-2444 Seibersdorf, Austria • Matthias Egger, Institute of Social and Preventive Medicine, University of Bern • Alfred Bürgi, ARIAS umwelt.forschung.beratung, Bern, Switzerland • Axel Hettich, Air Quality Management Agency of Basel, Switzerland