Validity of at home model predictions as a proxy for personal exposure to radiofrequency electromagnetic fields from mobile phone base stations
Astrid L. Martens, John F.B. Bolte, Johan Beekhuizen, Hans Kromhout, Tjabe Smid, Roel C.H. Vermeulen. Validity of at home model predictions as a proxy for personal exposure to radiofrequency electromagnetic fields from mobile phone base stations. Environmental Research. 142:221-226. Oct 2015. doi:10.1016/j.envres.2015.06.
Highlights
- • Inadequate RF-EMF exposure assessment hinders current epidemiological studies.
- • We compared modelled at home exposure levels to RF-EMF with personal measurements.
- • Meaningful ranking of personal RF-EMF can be achieved but there is misclassification.
- • Large sample sizes are required for sufficient power to study RF-EMF health effects.
Background
Epidemiological studies on the potential health effects of RF-EMF from mobile phone base stations require efficient and accurate exposure assessment methods. Previous studies have demonstrated that the 3D geospatial model NISMap is able to rank locations by indoor and outdoor RF-EMF exposure levels. This study extends on previous work by evaluating the suitability of using NISMap to estimate indoor RF-EMF exposure levels at home as a proxy for personal exposure to RF-EMF from mobile phone base stations.
Methods
For 93 individuals in the Netherlands we measured personal exposure to RF-EMF from mobile phone base stations during a 24 h period using an EME-SPY 121 exposimeter. Each individual kept a diary from which we extracted the time spent at home and in the bedroom. We used NISMap to model exposure at the home address of the participant (at bedroom height). We then compared model predictions with measurements for the 24 h period, when at home, and in the bedroom by the Spearman correlation coefficient (rsp) and by calculating specificity and sensitivity using the 90th percentile of the exposure distribution as a cutpoint for high exposure.
Results
We found a low to moderate rsp of 0.36 for the 24 h period, 0.51 for measurements at home, and 0.41 for measurements in the bedroom. The specificity was high (0.9) but with a low sensitivity (0.3).
Discussion
These results indicate that a meaningful ranking of personal RF-EMF can be achieved, even though the correlation between model predictions and 24 h personal RF-EMF measurements is lower than with at home measurements. However, the use of at home RF-EMF field predictions from mobile phone base stations in epidemiological studies leads to significant exposure misclassification that will result in a loss of statistical power to detect health effects.
http://bit.ly/1DhSIZM
RF-EMF exposure from mobile phone base stations (in the Netherlands) contributes ~13% to total environmental RF-EMF exposure (Bolte and Eikelboom, 2012). This contribution may vary by location and by age groups due to differences in behavioural patterns.
The use of models to predict personal exposure to RF-EMF has limitations due to the large spatial variation in RF-EMF levels in combination with subject movement patterns. Misclassification can lead to significant problems in epidemiological studies that look at an association between RF-EMF exposure and possible health effects, as potential health effects might not be detected due to lack of power and attenuated effect sizes. However, there are currently no alternatives for geospatial models to predict exposure for large scale epidemiological studies. Some improvements might be made by modelling additional locations where participants spend a lot of time like work or school, but future studies are necessary to assess the potential added value of this approach. It should be noted that detailed location information of the participants within buildings such as schools and offices are needed to reliable model RF-EMF exposure due to the large spatial variation in RF-EMF levels. This information is often not readily available, making it difficult to include these locations in estimating total exposure. When we stratified our analyses by the subjects that did not work during the measurement data (n=57) and subjects that did work during the measurement day we observed a slightly higher Spearman correlation for subjects who didn’t work (not worked: rsp=0.39, worked: rsp=0.32). Note that the low to moderate association between modelled exposure to RF-EMF from mobile phone base stations and measured personal exposure is similar to the accuracy found for other environmental pollutants, most notably air pollution (e.g. Nethery et al. 2008; Van Roosbroeck et al. 2008). Despite the presence of misclassification, a large number of air pollution studies have found health effects, although the type of exposure and health effects expected for air pollution are very different than for RF-EMF. When epidemiological studies have a sufficient sample size it should be possible to pick up potential health effects of RF-EMF exposure using NISMap.
Conclusion
This study evaluated the use of NISMap to predict personal exposure to RF-EMF from mobile phone base stations. The results indicate that a meaningful ranking of personal RF-EMF can be achieved, even though the correlation between model predictions and 24 h personal RF-EMF measurements is lower than with at home measurements. Our results indicate significant misclassification of participants, although in part our low Spearman correlations and sensitivity parameters can be explained by the inherent measurement error in the personal RF-EMF measurements. Exposure misclassification, assuming a classical error structure, leads to loss of power and can lead to attenuation of effect sizes (Armstrong, 1998). The main implication of our findings is therefore that epidemiological studies of health risks from far field RF-EMF will need a large number of participants in order to have sufficient power for detecting potential health effects. Ideally we would use more accurate methods of exposure assessment, but such methods (personal measurements, modelling multiple locations where the participants spend a lot of time, or including behavioural characteristics and other RF-EMF sources in the exposure model) are often expensive or require information that is not readily available.
--
Joel M. Moskowitz, Ph.D., Director
Center for Family and Community Health
School of Public Health
University of California, Berkeley
Electromagnetic Radiation Safety
Website: http://www.saferemr.com
Facebook: http://www.facebook.com/SaferE
News Releases: http://pressroom.prlog.org/
Twitter: @berkeleyprc
No comments:
Post a Comment