A Prospective Cohort Study of Adolescents’ Memory Performance and Individual Brain Dose of Microwave Radiation from Wireless Communication

Background: The potential impact of microwave radiofrequency electromagnetic fields (RF-EMF) emitted by wireless communication devices on neurocognitive functions of adolescents is controversial. In a previous analysis, we found changes in figural memory scores associated with a higher cumulative RF-EMF brain dose in adolescents. Objective: We aimed to follow-up our previous results using a new study population, dose estimation, and approach to controlling for confounding from media usage itself. Methods: RF-EMF brain dose for each participant was modeled. Multivariable linear regression models were fitted on verbal and figural memory score changes over 1 y and on estimated cumulative brain dose and RF-EMF related and unrelated media usage (n=669–676). Because of the hemispheric lateralization of memory, we conducted a laterality analysis for phone call ear preference. To control for the confounding of media use behaviors, a stratified analysis for different media usage groups was also conducted. Results: We found decreased figural memory scores in association with an interquartile range (IQR) increase in estimated cumulative RF-EMF brain dose scores: −0.22 (95% CI: −0.47, 0.03; IQR: 953 mJ/kg per day) in the whole sample, −0.39 (95% CI: −0.67, −0.10; IQR: 953 mJ/kg per day) in right-side users (n=532), and −0.26 (95% CI: −0.42, −0.10; IQR: 341 mJ/kg per day) when recorded network operator data were used for RF-EMF dose estimation (n=274). Media usage unrelated to RF-EMF did not show significant associations or consistent patterns, with the exception of consistent (nonsignificant) positive associations between data traffic duration and verbal memory. Conclusions: Our findings for a cohort of Swiss adolescents require confirmation in other populations but suggest a potential adverse effect of RF-EMF brain dose on cognitive functions that involve brain regions mostly exposed during mobile phone use. https://doi.org/10.1289/EHP2427

. Derivation of the individual durations of near field exposure situations and mean incidence fields of environmental far field sources. Table S2. Parameters used for the derivation of the cumulative near field dose. Individual exposure to a source is calculated via multiplying the individual time per exposure scenario with its respective SAR value. The brain and whole body dose is consecutively calculated as the sum of the individual exposure of the near-field sources. The table below summarizes the mean values used for the individual RF-EMF dose calculation for all HERMES participants who took part in baseline and follow-up investigations. Note that the duration refers to the average daily device use between baseline and follow-up. Table S3. Simulated RF frequency bands with frequency range and center frequency. Figure S1. Spider plot displaying the profile of the five different media use patterns (Foerster and Röösli 2017) on 11 different media use variables. Values are relative and relate to the ratio between the group mean and the overall mean of the whole sample on the raw questionnaire scores (e.g. a peak of the "Gaming" group at 3 relates to a three times (300 %) higher daily gaming duration compared to the whole sample whereas a value of 0.5 relates to a 50 % lower value on the respective variable. Figure S2. Forest plots of the meta-analysis over the linear exposure estimates for the brain dose calculated separately for the right side users of the five use groups. Coefficients relate to the change in score per interquartile range of the whole sample RF-EMF dose (953 mJ/kg/day).

Derivation of the Cumulative RF-EMF brain gray matter dose 1. Numeric simulations of brain gray matter specific absorption rates (SAR)
This section outlines the methodology used to estimate the specific absorption rate (SAR) in adolescents' brains' grey matter. The SAR quantifies the rate at which radio-frequency power is absorbed in (biological) tissue. This quantity cannot be measured in living subjects and therefore needs to be simulated. These simulations rely on human body models or phantoms obtained using imaging techniques such as magnetic resonance imaging. In this study, the heterogeneous phantoms Billie and Louis from the 'virtual population' are used to obtain SAR values for female and male adolescents, respectively (Gosselin et al. 2014). Billie is an eleven-years-old girl with a height of 1.49 m and a mass of 34 kg. Louis is a male of fourteen years old with a height of 1.68 m and a mass of 49.7 kg. The human body consists out of different tissues that have different properties seen by electromagnetic waves, so called dielectric properties. These properties depend on the frequency of the electromagnetic waves. The phantoms were assigned dielectric parameters from the Gabriel database (Gabriel et al. 1996) corresponding to the frequencies listed in Table S3.

Near field source modeling
In addition, the source of the RF exposure needs to be modeled for the simulation. In this study, the near field source of exposure, the mobile device, was modeled as a dipole antenna. In each frequency band listed in Table S3 a dipole antennas, was tuned to operate next to the body. These dipole antennas represent the mobile phone in the simulation and their dimensions of the dipole antennas can be found in (Aminzadeh et al. 2016).In total, six dipoles are modeled that resonate and emit at the center frequencies of the frequency bands listed in Table S3.
The following three different exposure situations were modelled for near field sources:  Close to the body: During these simulations, the wireless device is assumed to be in the pocket of trousers. A distance between body and dipole of 1 cm was chosen, with the dipoles aligned to the phantom's sagittal and coronal planes.
 Close to the ear: In this exposure scenario, the dipoles are placed at 1 cm from the ear of one of the two phantoms and are rotated over 45° towards the phantom's back in the sagittal plane, which contains the dipoles.
 Hands free kit: In this exposure scenario, the dipoles in the 'close to the ear' scenario are translated 20 cm further away from the ear orthogonal to the sagittal plane which contains the dipole in the 'close to ear' scenario.
Numerical simulations using the finite-difference time-domain (FDTD) technique (Taflove and Hagness 2005) with Sim4Life (ZMT, Zürich, Switzerland) were executed to obtain SAR values under the RF exposure conditions listed above.
During each of the simulations, a dipole was fed a harmonic signal of 1 W at its resonance frequency and emitted RF-EMFs which are partially absorbed in the human body models. The simulations are executed until a steady state was reached. The induced electric fields in each location of the grey matter of the phantoms were then extracted and used to calculate the SAR in the phantoms in each exposure scenario. These SAR values were then averaged over the mass of the grey matter of the phantoms. This resulted in a set of 5 (frequencies ,  Table S2) x 1 (averaging volume, the grey matter) x 2 (phantoms, Billie and Louis) x 3 (exposure conditions) = 30 SAR values, which were used for dose calculations.
The obtained near field SAR-values are normalized to 1 W output power of the dipoles used in the simulations. However, a real source might have another output power. Therefore, the SAR values were multiplied with the output power for respective sources in order to properly rescale them. These output powers were derived from literature (Persson et al. 2012). To obtain the GSM output power we assumed a difference in power control of approximately 23dB to an average UMTS output power of 0.45 mW (Kühn and Kuster 2013;Persson et al. 2012). For WiFi we assumed an average output power of 100 mW with an average duty cycle of 3.5 % assuming 10 % of time watching youtube and 90 % surfing (Joseph et al. 2013;Plets et al. 2015). These values were then averaged over the two phantoms and in the case of GSM also averaged over the two GSM frequencies. The resulting SAR grey matter values are listed in column four of Table S2.

Far field source modeling
The simulations described above are valid for exposure from devices relatively close to the human body. However, subjects are also exposed by sources, which may be devices or a mobile network, further away from the human body. In order to model this exposure, a common technique is to use plane waves incident on a human body model. In this study, we have modelled the far-field exposure and corresponding SAR values using plane-wavesimulations in the FDTD-based simulation software SEMCAD X. These plane waves will have a frequency, an angle of arrival (a direction), an amplitude, and a polarization. Our evaluation covered the frequency range from 50 MHz to 5. The simulations result in specific absorption rate values (SAR) in the grey matter of these two phantoms. These SAR values are averaged over the grey matter of the phantoms. For each phantom and each frequency, 12 SAR values are extracted corresponding to the six directions of incidence and two orthogonal polarizations. These SAR values correspond to an incident electric field strength of 2.45 V/m, which is chosen in the simulation. They are renormalized to mean incident field strengths, see Table S2, obtained using geospatial propagation models (Bürgi et al, 2010).The SAR values are then averaged over these 12 values, the two phantoms, and interpolated to the centre frequencies of the respective telecommunication frequency band. The resulting values are listed in Table S2 for far-field exposure.  Headset use (%) -HERMES questionnaire The answer categories "never", "seldom", "often" and "most of the time/always" were translated to the numerical values 0, 0.25, 0.50 and 1, respectively, and furthermore averaged over baseline and follow-up.

Proportion 3G /WiFi (%) -HERMES questionnaire Mobile phone close to body; (min/day)
Exposure to RF-EMF only while the mobile phone is actively transmitting or sending data (receiving a message, connecting to a mobile phone antenna) or receiving a call. This actual exposure time to be assumed 1 % of total time on body

Computer, laptop and tablet use with WiFi
Daily duration use (min/day)

Source Derivation of estimated variables
Uplink (from other people) -Personal RF-EMF measurements -HERMES questionnaire Estimated via multivariable linear regression modeling calibrated on the mobile phone uplink obtained via personal RF-EMF measurements of 148 participants. Variables from the self-reported baseline questionnaire were stepwise included in order to determine the best fitting model. Subsequently the predicted values from the model were used as uplink estimates for dose calculation for all participants without personal measurement data. The following predictors were used in the final model -Mobile phone operator -Mobile phone on/off during night -Number of Smartphones at home -Daily duration using public transport: train -Daily duration using public transport: bus -Investigation phase ( Radio/Broadcast a Adolescents far-field exposure in school might differed from far-field exposure at home. Adolescents' time in school was assumed to be one fifth of 24 hours on weekdays and modelled values from the place of school were used for this proportion of time. Further far-field exposure might be substantially higher in public transports than elsewhere due to the many people actively engaged with their mobile phones. Average times spent in public transports were derived from personal measurements or the questionnaire. The remaining time neither spent in school nor in public transports was assumed residential time at home. DECT 0.01 6.3*10 -3 0.01 a For calls with the mobile phone on the GSM network the mean of the SARs for the GSM900 and the GSM1800 network was used because there was no differentiation between GSM900 and GSM1800 network in the mobile phone operator data. A headset is assumed to be wired to the phone.  1920-1980 1950 Wi-Fi 2 GHz 2400-2483.5 2450

Overview of the parameters considered for RF-EMF brain gray matter dose
Meta-analysis over five latent classes of media use Figure S1: Spider plot displaying the profile of the five different media use patterns (Foerster and Röösli 2017) on 11 different media use variables. Values are relative and relate to the ratio between the group mean and the overall mean of the whole sample on the raw questionnaire scores (e.g. a peak of the "Gaming" group at 3 relates to a three times (300 %) higher daily gaming duration compared to the whole sample whereas a value of 0.5 relates to a 50 % lower value on the respective variable. Figure S2: Forest plots of the meta-analysis over the linear exposure estimates for the brain dose calculated separately for the right side users of the five use groups. Coefficients relate to the change in score per interquartile range of the whole sample RF-EMF dose (953 mJ/kg/day).