SEARCH

SEARCH BY CITATION

Keywords:

  • delta frequency;
  • demodulation;
  • electroencephalogram;
  • electromagnetic field;
  • magnetic field;
  • mobile phone;
  • non-rapid eye movement;
  • radio frequency;
  • rapid eye movement;
  • spindle frequency;
  • theta frequency

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Studies have repeatedly shown that electroencephalographic power during sleep is enhanced in the spindle frequency range following radio frequency electromagnetic field exposures pulse-modulated with fundamental frequency components of 2, 8, 14 or 217 Hz and combinations of these. However, signals used in previous studies also had significant harmonic components above 20 Hz. The current study aimed: (i) to determine if modulation components above 20 Hz, in combination with radio frequency, are necessary to alter the electroencephalogram; and (ii) to test the demodulation hypothesis, if the same effects occur after magnetic field exposure with the same pulse sequence used in the pulse-modulated radio frequency exposure. In a randomized double-blind crossover design, 25 young healthy men were exposed at weekly intervals to three different conditions for 30 min before sleep. Cognitive tasks were also performed during exposure. The conditions were a 2-Hz pulse-modulated radio frequency field, a 2-Hz pulsed magnetic field, and sham. Radio frequency exposure increased electroencephalogram power in the spindle frequency range. Furthermore, delta and theta activity (non-rapid eye movement sleep), and alpha and delta activity (rapid eye movement sleep) were affected following both exposure conditions. No effect on sleep architecture and no clear impact of exposure on cognition was observed. These results demonstrate that both pulse-modulated radio frequency and pulsed magnetic fields affect brain physiology, and the presence of significant frequency components above 20 Hz are not fundamental for these effects to occur. Because responses were not identical for all exposures, the study does not support the hypothesis that effects of radio frequency exposure are based on demodulation of the signal only.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Mobile phones have become a ubiquitous part of modern telecommunications technology, and work by transmitting and receiving radio frequency electromagnetic fields (RF EMF) that are partially absorbed by the head or body during use. A number of studies have now consistently shown that this pulse-modulated RF EMF emitted by mobile phone handsets can alter brain physiology. Specifically, electroencephalographic (EEG) spectral power in the alpha (8–12 Hz) and spindle (∼12–14 Hz) frequency ranges is enhanced both during and following pulse-modulated RF EMF exposure, particularly during non-rapid eye movement (NREM) sleep (Borbély et al., 1999; Huber et al., 2000, 2002; Loughran et al., 2005, 2012; Regel et al., 2007b; Schmid et al., 2012). Importantly, it has also been shown that it is this pulse modulation of the RF EMF signal that is critical to induce these effects on the EEG, as continuous wave exposure at the same intensities did not induce any changes on the EEG (Huber et al., 2002; Regel et al., 2007a). However, it should also be noted that some studies have failed to find any effects on the EEG during sleep (e.g. Fritzer et al., 2007; Mann et al., 1998; Wagner et al., 1998, 2000), while others have reported effects in different frequency ranges (e.g. Lowden et al., 2011). Despite these different reports, methodological issues in the latter mentioned studies make comparability difficult. However, in the aforementioned studies where an effect of RF EMF exposure on the EEG during sleep was found, the effect remained relatively consistent.

In addition to effects observed on the EEG, several studies have also explored the possibility of RF EMF having an influence on cognitive performance. However, although effects have been reported regarding reaction time and accuracy of performance, there remain large inconsistencies and numerous studies have failed to observe any exposure-related effects at all (Regel and Achermann, 2011). Therefore, unlike effects on the EEG, there remains no clear evidence for exposure-related effects on cognition (Valentini et al., 2010).

Despite previous research showing an influence of pulse-modulated RF EMF on the sleep EEG, the mechanisms behind these exposure-induced changes still remain unclear. Furthermore, there is no supporting evidence based on previous studies that this effect is related to health consequences, such as alterations to sleep quality (e.g. Huber et al., 2002; Loughran et al., 2005; Regel et al., 2007b; Schmid et al., 2012). Of particular interest is which pulse modulation components of the RF EMF signal are responsible for the alterations of EEG spectral power observed during sleep. We previously showed that the 14-Hz pulse modulation component, which is in the physiological sleep spindle frequency range, is one potential mediator of these sleep EEG alterations (Schmid et al., 2012). However, further results also indicated that the specificity of the pulse modulation may not be the most important factor, as other modulations outside of the spindle frequency range also appeared to have an influence on the sleep EEG. Additionally, a possible involvement of higher harmonics of the pulsed modulation cannot be ruled out as contributing to the effects on the EEG.

Therefore, the objective of the current study was to test if a signal without significant harmonics above 20 Hz is sufficient to influence the EEG. In addition, we wanted to investigate whether applying a magnetic field with the same pulse sequence as applied in the pulse-modulated RF exposure would result in similar changes to the EEG. The latter is an attempt to test the demodulation hypothesis, i.e. that the effects are induced by the demodulated RF signal (corresponding to the low-frequency envelope of the transmitted signal) via an electrically non-linear structure inside the living human brain. Demodulation occurring in the brain has been proposed as one possible mechanism for effects of pulse-modulated RF EMF (Bawin and Adey, 1976). In electronic circuits demodulation is usually achieved with semiconductors; however, the potential ability of brain structures, cell membranes or larger molecules to demodulate RF EMF signals has to date not been shown (for overview, see Kowalczuk et al., 2010). A demodulated RF EMF is composed of the low-frequency modulation component of the RF signal. The magnetic field chosen in the current study to test for effects of the low-frequency modulation on the EEG induces both components in the tissues, i.e. the electric and magnetic fields.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Subjects

Twenty-five healthy male volunteers aged between 20 and 26 years [mean age 23.2 ± 0.4 years (±SEM)] participated in the study. Participants were recruited through advertisements at the University of Zurich and ETH Zurich. All participants were right-handed, non-smokers, free of medication, and had moderate mobile phone use (<2 h per week, average = 48.5 ± 6.7 min week−1). Participants were free of sleep complaints, and a screening night was used to exclude any suffering from sleep apnea, nocturnal myoclonus or sleep efficiency <80%. Starting 3 days before the study, participants were required to abstain from caffeine and alcohol, and maintain a regular sleep–wake schedule of 8 h of sleep that corresponded with the scheduled laboratory sleep times. Compliance was verified using breath alcohol tests, wrist-worn actimeters and a sleep diary. In addition, excessive physical exercise and all mobile phone calls were prohibited on study days. The cantonal ethical committee approved the study protocols, and participants gave their written informed consent and were recompensed for participation.

Study procedure

The study was undertaken in the sleep laboratory of the Institute of Pharmacology and Toxicology, University of Zurich. A randomized, double-blind, crossover design was employed in which participants spent six study nights (three exposure nights at weekly intervals, each preceded by an adaptation night) in the sleep laboratory. Participants were exposed for 30 min directly prior to their scheduled bedtime. Exposure times were staggered as only two participants could be exposed simultaneously, therefore bedtimes were either 22:40–06:40 h or 23:20–07:20 h, depending on whether the participant received the early (22:00–22:30 h) or late (22:40–23:10 h) exposure. This procedure resulted in a 10-min time window between the end of exposure and the start of recording (i.e. lights out). Following each exposure, participants were asked whether they were able to perceive a field (using a five-point scale including the following possible answers: ‘don’t know, no not at all, yes left, yes right, yes on both sides’), and completed questionnaires regarding current mood state and well-being both prior to bedtime and upon waking up in the morning.

Exposure conditions

Each participant underwent three different exposure conditions, which were: (i) pulse-modulated RF signal of 900 MHz [specific absorption rate (SAR) = 2 W kg−1]; (ii) pulsed magnetic field; (iii) control condition without field (sham). Both pulsed signals had a basic modulation frequency of 2 Hz and a peak-to-average ratio of 4 in pulse amplitude. In order to reduce the higher harmonics the pulse structures were smoothened by applying a Gaussian low-pass filter (−3 dB at 20 Hz), reducing the spectral power above 20 Hz by more than a factor of 10 and by ∼10 000 at 50 Hz (for details on signal characteristics, see Fig. 1). Power for the 2- and 8-Hz components was equivalent (Fig. 1).

image

Figure 1.  (a) Time domain signals. Pulse structure of the applied generic ‘2-Hz signal’ [normalized to 2 W kg−1 specific absorption rate (SAR) for the radio frequency (RF) field] and comparison to the ‘handset-like’ signal (normalized to 1 W kg−1 SAR) described in Huber et al. (2005). Both are based on the GSM frame structure with the fundamental period of 480 ms (2.08 Hz). The ‘2-Hz signal’ was used to modulate the magnetic field (MF) and the RF field. (b) Power spectrum (RF) and amplitude spectrum (MF, ‘2-Hz signal’) of the modulation components of the signals. The power spectrum reflects the side bands of the modulated carrier frequency (RF), and is identical with the spectrum of low-frequency modulation components (envelope). Spectra were calculated with a Fast Fourier Transformation of 40 signal repetitions after application of a Tukey window (α = 0.1). Spectra are plotted on a logarithmic scale (dB); 0 dB corresponds to 1 W kg−1 for RF fields and 142 A m−1 for the MF. Because the MF is represented by the amplitude spectrum, the DC component (0 Hz) is 3 dB lower. The spectral power of the 2-Hz component is 0.51 W kg−1 for the ‘2-Hz signal’, and 0.048 W kg−1 for the ‘handset-like’ signal applied in Huber et al. (2002, 2005). The corresponding values for the 217-Hz component (most prominent component of the GSM signal) are 0 and 0.13 W kg−1, respectively.

Download figure to PowerPoint

The RF signal was modulated with the described signal and applied by patch antennas (SPA860/65/9/0/V; Huber + Suhner, Herisau, Switzerland), the center of which was 42 mm vertically above the ear canal at 115 mm distance from the left side of the head (unilateral exposure of the left hemisphere). Left hemisphere exposure was chosen in order to be comparable to previous studies from our laboratory (Huber et al., 2002; Schmid et al., 2012). The input power was adjusted to obtain a peak spatial SAR of 2 W kg−1 in the head tissue (Fig. 2). For a detailed description of the exposure and dosimetry, see Huber et al. (2003) and Boutry et al. (2008).

image

Figure 2.  Tissue models, estimated SAR [radio frequency (RF) field, left hemisphere exposure] and magnetic field (MF; homogenous exposure) distribution are illustrated (color-coded distribution: 0 dB at 2 W kg−1 for RF and 164 A mrms−1 for MF).

Download figure to PowerPoint

The magnetic field with the described pulsed signal characteristics was generated by a set of Helmholtz-like coils placed on both sides of the exposure system (Fig. 3), resulting in a largely homogenous magnetic field at the level of the subject’s head. The rectangular shaped coils (33 × 39 cm, 2 × 23 windings) on each side of the head were separated by 25.7 ± 0.5 cm, depending on the participant’s anatomy. The ear canal was 8.2 cm below the center line between the coils. The coils were powered by a voltage-controlled current source (VCCS, SPEAG, Zurich, Switzerland). The amplitude was set to a whole-brain time and spatial averaged H-field of 140 A m−1 ± 8% (SD: k = 1) and a peak field of 560 A m−1 (excluding the marginally exposed medulla oblongata). This is equivalent to a B-field of 0.176 and 0.70 mT, respectively. All exposures complied with the international limits according to the ICNIRP guidelines (International Commission on Non-Ionizing Radiation Protection, 1998). Because little is known about the effects of magnetic fields on the EEG, we applied a field strength that is significantly larger than the extremely low frequency (ELF) exposures posed by mobile phones (approximately 10-fold) and close to the limit for the general public (85.7% of the ICNIRP limit).

image

Figure 3.  Exposure system for both conditions: antenna for the RF condition inside the left box and the Helmholtz-like coils for the MF exposure condition.

Download figure to PowerPoint

Polysomnographic recordings

The EEG (C3A2), electrooculogram, electromyogram (EMG; mental or submental) and electrocardiogram (ECG) were continuously recorded during 8 h of nighttime sleep with a polygraphic amplifier Artisan (Micromed, Mogliano Veneto, Italy). Analog signals were filtered by a high-pass filter (EEG: −3 dB at 0.15 Hz; EMG: 10 Hz; ECG: 1 Hz) and an anti-aliasing low-pass filter (−3 dB at 67.2 Hz). The data were sampled at 256 Hz and recorded using the software Rembrandt DataLab (Version 8.0; Embla Systems, Broomfield, CO, USA).

Sleep stages were visually scored for 20-s epochs according to the criteria of Rechtschaffen and Kales (1968), and sleep cycles were defined as per Feinberg and Floyd (1979). Artefact removal was performed as per Schmid et al. (2012). EEG power spectra of consecutive 20-s epochs were calculated (Hanning window, averages of five 4-s epochs, frequency resolution 0.25 Hz), and frequencies between 0.75 and 20 Hz were analyzed.

Mean power density spectra were calculated for NREM and stage 2 sleep (whole-night and for the first four episodes), as well as for rapid eye movement (REM) sleep (whole-night, and for the second, third and fourth episodes). Participants with <5 min of artefact-free REM sleep were excluded from analysis (Schmid et al., 2012). The initial REM sleep episode was not analyzed due to an insufficient number of participants meeting the 5-min criterion (n = 10). The second, third and fourth REM sleep episode analyses had sample sizes of 21, 23 and 22, respectively. Linear mixed models (presuming an identical intraclass correlation for all participants) were used for analysis of EEG spectral power (sas 9.1.3; SAS Institute, Cary, NC, USA), and included the factors ‘Condition’ [RF, magnetic field (MF) and sham], ‘Order of Sham’ (1, 2, 3), and their interaction. Where the factor ‘Condition’ or the interaction ‘Condition × Order’ reached significance (P < 0.05), post hoc paired t-tests were performed. Furthermore, analysis of the time-course of the effect was performed on significant frequency ranges across the first four NREM and REM sleep episodes (excluding the first REM episode) using a mixed model anova including the factors ‘Condition’, ‘Episode’, and their interaction. In addition, a frequency range was only considered to be significant if at least three consecutive frequency bins reached significance.

Sleep variables as derived from visual scoring were also analyzed with a linear mixed model anova, and included the factors ‘Condition’, ‘Order of Sham’, and their interaction. Two subjects were excluded from the sleep variable and EEG analysis, one due to poor signal quality and the other due to long periods of wakefulness, resulting in a sleep efficiency of <80%. Therefore, the final analysis consisted of 23 subjects. In addition, heart rate was assessed by identification of R-R intervals in the ECG, which were averaged for the first four NREM and REM sleep episodes and analyzed as for the EEG power spectral data.

Cognitive tasks

Throughout exposure, participants performed three different cognitive tasks: the simple reaction time task (SRT); choice reaction time task (CRT); and the N-back task (N-back). Details of the cognitive tasks, their application and analysis are described elsewhere (Regel et al., 2007b; Schmid et al., 2012). Reaction times were analyzed using linear mixed model anovas, including the factors ‘Condition’ (RF, MF, sham), ‘Session’ (1, 2), ‘Order of Sham’ (1, 2, 3), and associated interactions. Accuracy of performance was not normally distributed, and therefore non-parametric Wilcoxon Signed Rank tests were performed. In order to account for multiple comparisons, the alpha level for all cognitive task analyses was adjusted to P < 0.015 (Schmid et al., 2012).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Sleep variables and subjective measures

Analysis of visually scored sleep variables showed no differences in sleep architecture between the exposure conditions and sham (Table 1). No condition effects were observed for NREM sleep episode durations or respective stage 2 sleep percentages; however, less REM sleep contributed to the second sleep cycle following the RF exposure condition compared with sham (P < 0.03; Table 2).

Table 1.   Sleep variables derived from visual scoring are provided in minutes (for sleep efficiency as a percentage of total sleep time) and standard error of the mean in parentheses (n = 23)
  ShamRFMF
  1. Time in bed was 8 h. Sleep latency: interval from lights off to the first occurrence of stage 2 (sleep onset); REM sleep latency: interval from sleep onset to the first occurrence of REM sleep; slow-wave sleep: NREM sleep stages 3 and 4. No significant differences between exposure and sham were observed. RF = 2-Hz pulse-modulated radio frequency EMF, MF = 2-Hz pulsed magnetic field.

  2. MF, magnetic field; REM, rapid eye movement; RF, radio frequency.

Total sleep time (min)462.1 (1.3)461.2 (2.1)459.4 (2.4)
Sleep latency (min)7.6 (1.1)8.0 (1.8)8.5 (2.0)
REM sleep latency (min)72.1 (2.8)72.6 (3.2)70.8 (3.0)
Waking after sleep onset (min)3.2 (0.8)4.0 (1.0)5.2 (1.4)
Stage 2 (min)190.7 (5.3)189.4 (4.9)191.8 (5.8)
Slow-wave sleep (min)136.4 (5.9)140.6 (6.2)133.3 (5.4)
REM sleep (min)124.4 (3.9)117.6 (3.8)121.3 (4.8)
Movement time (min)6.5 (0.5)6.8 (0.5)7.3 (0.5)
Sleep efficiency (%)96.3 (0.3)96.1 (0.4)95.6 (0.5)
Table 2.   Amount of NREM, REM and stage 2 sleep per sleep cycle
 ShamRFMF
  1. Durations of non-rapid eye movement sleep (NREMS) and rapid eye movement sleep (REMS) (min) and stage 2 sleep (as a percentage of NREMS) are provided with standard error of the mean in parentheses (n = 23). Following the RF condition, less REM sleep contributed to the second sleep cycle compared with sham (*P < 0.03). No further condition-dependent effects were observed.

  2. MF, magnetic field; RF, radio frequency.

Cycle 1
 NREMS (min)72.1 (2.8)72.6 (3.2)70.8 (3.0)
 REMS (min)13.7 (1.4)16.4 (2.3)14.6 (2.5)
 Stage 2 (%)21.1 (1.7)22.6 (2.2)21.5 (1.8)
Cycle 2
 NREMS (min)71.4 (3.2)70.9 (3.0)71.2 (2.6)
 REMS (min)32.8 (3.2)24.6 (1.7)*27.9 (3.1)
 Stage 2 (%)44.2 (2.9)43.9 (4.0)45.0 (3.6)
Cycle 3
 NREMS (min)74.4 (3.4)72.6 (3.3)71.2 (2.5)
 REMS (min)31.0 (2.4)32.2 (2.6)33.2 (3.9)
 Stage 2 (%)66.0 (3.7)61.3 (2.8)69.6 (3.8)
Cycle 4
 NREMS (min)60.9 (1.4)62.7 (2.2)62.4 (1.8)
 REMS (min)44.8 (4.8)47.6 (4.7)50.4 (4.6)
 Stage 2 (%)76.6 (3.8)74.1 (4.7)71.3 (3.8)

Spectral analysis of the NREM sleep EEG revealed increased power in the spindle frequency range following exposure [Condition, F2,46 > 4.53, P < 0.02 (NREM sleep); F2,46 > 3.46, P < 0.04 (stage 2 sleep); Fig. 4]. Post hoc comparisons revealed that this increase was only present following the RF condition [P < 0.03 (NREM sleep), P < 0.04 (stage 2 sleep): 13.75–15.25 Hz; Fig. 4]. In both NREM and stage 2 sleep, the maximum increase in spectral power (16%) was seen at 14 Hz. In addition to these changes in the spindle frequency range, increased spectral power in the delta and theta ranges was also observed [Condition, F2,46 > 3.27, P < 0.05 (NREM sleep); F2,46 > 4.04, P < 0.03 (stage 2 sleep); Fig. 4]. The effect was present following both exposure conditions between 1.25 and 9.0 Hz [P < 0.05 (NREM and stage 2 sleep)].

image

Figure 4.  (a) Average relative electroencephalogram (EEG) power density spectra (C3A2, 0.25-Hz bins, n = 23) for non-rapid eye movement sleep (NREMS) across the entire night [2-Hz pulse-modulated radio frequency (RF) EMF: black; 2-Hz pulsed magnetic field (MF): gray]. Power in the exposure conditions is expressed relative to sham (sham = 1.0). Statistical analysis revealed an increase of spectral power in the spindle frequency range (13.75–15.25 Hz) following RF exposure, and an increase in the delta/theta frequency range following both conditions (1.25–9.0 Hz). Spectra for stage 2 sleep were very similar and therefore for clarity are not shown. F-values of the linear mixed model anova (factor ‘Condition’) are shown for NREM and stage 2 (S2) sleep (P < 0.05, black bars; P < 0.1, open bars). Significant frequency bins (paired t-test) of both exposure conditions (RF, MF) compared with sham are indicated by triangles (P < 0.05, filled triangles; P < 0.1, open triangles; for NREMS and S2). (b) Temporal evolution of EEG spectral power changes in NREMS (mean ± SEM) following the RF (black) and MF (gray) conditions are illustrated relative to sham (sham = 1.0) for the two affected frequency ranges (delta/theta range: 1.25–9.0 Hz, upper panel; spindle range: 13.75–15.25 Hz, lower panel). Episode midpoints are provided in hours after lights off. Significant differences from sham are indicated by triangles (linear mixed model anova followed by post hoc paired t-tests performed separately for each frequency range; P < 0.05, filled triangles; P < 0.1, open triangles).

Download figure to PowerPoint

The temporal evolution of the effects in both the spindle and delta/theta frequency ranges (1.25–9.0 Hz, 13.75–15.25 Hz) was analyzed for the first four NREM sleep episodes, and revealed significant effects for both factors ‘Condition’ (F2,253 > 3.26, P < 0.04) and ‘Episode’ (F3,253 > 25.95, P < 0.01). Post hoc analyses revealed that the increase in power in the spindle frequency range was present in both NREM and stage 2 sleep across all four NREM sleep episodes, but only following the RF exposure condition (Fig. 4). The effect reached significance for NREM and/or stage 2 sleep in episodes 1, 3 and 4 (P < 0.04), and was at a trend level for the second episode (P < 0.06). Regarding the delta/theta frequency range, post hoc analyses revealed an increase in power in the third (both exposure conditions) and fourth (MF exposure condition only) NREM sleep episodes.

Spectral analysis of the EEG during REM sleep also showed a significant increase in power in two frequency ranges (0.75–1.5 Hz and 7.75–12.25 Hz) following exposure (Condition; F2,46 > 3.24, P < 0.05; Fig. 5). Post hoc comparisons showed that the increase in EEG power within the alpha range (7.75–12.25 Hz) was present only following the RF condition (P < 0.05), whereas the increase in the lower delta range (0.75–1.5 Hz) was observed following both field conditions (P < 0.04; Fig. 5). For the RF condition, both frequency ranges were significantly increased across all episodes, while in the MF condition significant increases in power were only seen in the second and fourth REM sleep episodes.

image

Figure 5.  (a) Average relative electroencephalogram (EEG) power density spectra (C3A2, 0.25-Hz bins, n = 23) for rapid eye movement sleep (REMS) across the entire night [2-Hz pulse-modulated radio frequency (RF) EMF: black; 2-Hz pulsed magnetic field (MF): gray]. Statistical analysis revealed an increase of spectral power in the alpha frequency range (7.75–12.25 Hz) following RF exposure, and an increase in the lower delta range (0.75–1.5 Hz) following both conditions when compared with sham. F-values of linear mixed model anova (factor ‘Condition’) are shown (P < 0.05, black bars; P < 0.1, open bars). Significant frequency bins for both exposure conditions (RF, MF) compared with sham are indicated by triangles (paired t-tests; P < 0.05, filled triangles; P < 0.1, open triangles). (b) Temporal evolution of EEG spectral power changes (mean ± SEM) following the RF (black) and MF (gray) conditions are illustrated relative to sham (sham = 1.0) for the two affected frequency ranges (second–fourth REM sleep episodes; lower delta range: 0.75–1.5 Hz, upper right panel; alpha range: 7.75–12.25 Hz, lower right panel). Episode midpoints are provided in hours after lights off. Significant differences from sham are indicated by triangles (linear mixed model anova followed by post hoc paired t-tests performed separately for each frequency range; P < 0.05, filled triangles; P < 0.1, open triangles).

Download figure to PowerPoint

The effects on the NREM and REM sleep EEG were further explored in terms of individual variability. For both NREM and REM sleep, and regarding all affected frequency ranges, the majority of participants showed increased EEG spectral power; however, this was not the case for all participants, with some also showing decreases following exposure (detailed individual responses are shown in Fig. 6).

image

Figure 6.  Individual change in whole-night spectral power relative to sham (sham = 0) during non-rapid eye movement sleep (NREMS: 1.25–9.0 Hz, 13.75–15.25 Hz) and rapid eye movement sleep (REMS: 0.75–1.5 Hz, 7.75–12.25 Hz) for both the radio frequency (RF) and magnetic field (MF) exposure conditions. Subjects were sorted according to the increase in the spindle range following the RF condition (A – W, n = 23). Dashed lines indicate a ±5% change from sham. EEG, electroencephalogram.

Download figure to PowerPoint

Analysis of heart rate showed no exposure-related alterations in either NREM or REM sleep. No significant differences between the exposure conditions were found for measures of mood, well-being or subjective sleep quality. Additionally, subjects were not able to perceive the applied fields.

Cognitive tasks

Session effects (1st versus 2nd half of exposure) were observed in all tasks independent of exposure conditions. For reaction speed a significant condition effect was observed in the SRT (Condition, F2,125 = 12.82, P < 0.015) and the 3-back (Condition, F2,125 = 6.56, P < 0.015). Post hoc analysis revealed increased reaction speed only in the SRT following the MF exposure condition in both sessions (P < 0.015). Performance accuracy was largely unaffected by exposure, except for a trend level increase in the 1-back task (Wilcoxon Signed Rank test, session 2, P < 0.03) during the RF condition.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

Consistent with previous research, the current results provide further evidence that pulse-modulated RF EMF influences brain physiology. Additionally, pulsed MFs also appear to have an effect on brain physiology. On the other hand, the current study does not support the hypothesis that effects are based on demodulation of the signal alone as spindle frequency activity was only affected following the pulsed RF EMF.

The RF EMF condition, which included only low-frequency pulse modulation components with significantly attenuated higher harmonics, led to a significant increase in EEG spectral power in both NREM and stage 2 sleep in the spindle frequency range. Similar to our previous study (Schmid et al., 2012), the presence of an effect with yet another different pulse modulation scheme provides further support for the idea that the specificity of the pulse modulation is not the most important factor in inducing effects on the EEG. That is, pulsed signals with fundamental components at 2, 8, 14 or 217 Hz induce similar effects. Therefore, any pulse-modulated RF EMF scheme in a frequency range that is close to biologically relevant rhythms may be sufficient to induce changes in the spindle frequency range of the EEG during sleep (Schmid et al., 2012). Interestingly, the modulation scheme used in the current study also induced effects in the delta and theta frequency ranges, whereas in most previous studies the effect was restricted to the spindle (and in some cases, alpha) frequency range (Borbély et al., 1999; Huber et al., 2000, 2002; Loughran et al., 2005, 2012; Regel et al., 2007b; Schmid et al., 2012). It is not clear why this low-frequency modulation scheme led to changes in delta/theta activity, particularly as these same low-frequency components were also present in the majority of previous studies where effects were not observed. However, it could be that subtle differences in exposure regimes, in this case particularly the significant enhancement of the 2- and 8-Hz component compared with previous studies (10.2 times higher than in the handset-like signal; Fig. 1) and the attenuation of higher harmonics, may be sufficient to induce effects in different EEG frequency ranges.

Similarly, the pulsed MF exposure also induced effects in the delta and theta frequency ranges during NREM and stage 2 sleep, but not the spindle frequency range. Therefore, it appears that changes in the spindle frequency range are only induced by pulse-modulated RF EMF exposure. This finding does not support the demodulation hypothesis, because if the effect was related to demodulation of the signal within the brain then similar effects on the EEG would be expected from both exposure conditions applied. On the other hand, the current results suggest that both low-frequency pulse-modulated RF EMF and low-frequency pulsed MF can alter brain electrical activity in a similar way in the lower frequency ranges. Overall, the findings illustrate that other mechanisms must also be involved and that demodulation alone is not sufficient.

In regards to REM sleep, unlike the majority of previous research, there was a widespread enhancement of EEG spectral power following the RF EMF exposure condition. An increase of mean EEG power density across a wide range of frequencies (1–20 Hz) during REM sleep was reported in one early study by Mann and Röschke (1996), although several subsequent studies were unable to replicate this initial report of exposure-induced EEG alterations during REM sleep (Borbély et al., 1999; Huber et al., 2000, 2002; Regel et al., 2007b; Wagner et al., 1998, 2000). Despite this, in our most recent previous study we did observe some small changes to the REM sleep EEG, although there was no clear pattern and the effect seemed to be randomly scattered at different frequencies (Schmid et al., 2012). It is unclear why the specific modulation scheme used in the current study would induce effects on the REM sleep EEG given that this has generally not been seen previously. It may be that limiting pulse modulation to low frequencies and eliminating higher harmonics produces a more concentrated and biologically relevant exposure in terms of EEG frequencies, leading to a more widespread effect across several frequency ranges in both sleep states. However, this is the first time that a well-controlled and relatively large study has shown a clear and significant effect on the REM sleep EEG, therefore further studies are required in order to confirm the presence of an influence of EMF exposure on the EEG during REM sleep.

Although effects on the sleep EEG have now been repeatedly shown across several well-controlled studies (Borbély et al., 1999; Huber et al., 2000, 2002; Loughran et al., 2005, 2012; Regel et al., 2007b; Schmid et al., 2012), the time-course of the effect remains variable. Indeed, in the current study the effects were generally seen across the night but did not always reach significance. The current time-course also differed from our most recent study (Schmid et al., 2012). One possible explanation for this discrepancy would be that different exposure parameters, such as the frequencies applied, as well as differences in the intensity, distribution, duration or peak power of exposure may lead to alterations in the time-course of the effect. However, another possible explanation would be that the effect is not constant throughout the night and continually changes, and therefore may not always reach a level of significance, particularly given the individual variability in response. Despite this, the current and previous observations all provide evidence for an effect that outlasts exposure, and in some cases by up to several hours.

The individual variability in response to exposure, with the majority of participants showing an increase in spectral power following exposure, is again an interesting observation that is in line with our previous study (Schmid et al., 2012). A recent study specifically addressed this issue, indicating that the influence of RF EMF on human brain activity during sleep varied across individuals, with participants who showed an increase in a first study tending to respond in a similar manner when re-tested (Loughran et al., 2012). This issue also highlights the importance of using a sufficient sample size in studies addressing the potential influence of EMF, as with small samples this overall increase in spectral power would likely go undetected.

Consistent with previous research, neither exposure condition led to any changes in sleep architecture. Regarding cycles, it was found that less REM sleep contributed to the second sleep cycle following the RF exposure condition. The potential meaning of such a small change in REM sleep distribution is unclear, and as this is the first time such an effect has been observed following RF exposure independent replication is required in order to verify the presence of such an effect. It should be noted that one previous study reported exposure-related effects on the macrostructure of sleep (Lowden et al., 2011); however, due to methodological differences interpretation and comparison with the current results remains difficult. In addition to effects on sleep architecture, other measures such as mood and well-being were not influenced by exposure, and given that no reliable changes were observed in sleep architecture it can also be assumed that there are no exposure-induced effects on sleep quality (for example, duration, latency and efficiency of sleep as derived from standard polysomnographic measures of sleep). Similarly, no clear exposure-related effects on cognition were observed, with only one effect seen relating to reaction speed in the SRT task following MF exposure. The lack of consistent effects on performance adds to the increasing evidence that cognition is likely not influenced, at least not in a repeatable or measureable manner, by mobile phone-type exposures (Valentini et al., 2010).

Overall, our results add to the now established effect of a pulse-modulated RF EMF on brain physiology. In addition, they provide further support for pulse modulation being the critical component in inducing effects, and that the effect does not have a very high specificity in regards to the frequency domain of the modulation scheme applied. The results also provide a first indication of low-frequency modulation being capable of inducing an effect across a broader range of frequencies, regardless of what type of EMF the modulation is combined with. Despite these repeatable effects on brain activity, the related underlying mechanisms remain unknown and warrant further investigation.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References

The authors thank Dominik Benz, Katrin Stadelmann, Sarah Münst and Iva Jelezarova for their assistance in the sleep laboratory, Steven Geinitz for statistical advice, Karl Wüthrich and Dr Roland Dürr for technical support, and Dr Myles Capstick for comments on the manuscript. This study was supported by the Swiss National Science Foundation (National Research Programme 57: ‘Non-Ionizing Radiation – Health and Environment’).

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflict of Interest
  9. References