Frei P, Mohler E, Bürgi A, Fröhlich J, Neubauer G, Braun-Fahrländer C, Röösli M; the QUALIFEX team. A prediction model for personal radio frequency electromagnetic field exposure. Sci Total Environ. Ahead of print. Oct 9, 2009.
Assessment of exposure to radiofrequency fields (RF) for epidemiological studies is challenging. There are two principal sources of exposure: 1) sources close to the human body (e.g. mobile phones), which cause relatively high and periodic short-term exposure primarily to the head, and 2) more distant sources (e.g. base stations), which cause lower but continuous whole-body exposure. There are various exposure assessment methods: questionnaires on mobile phone use sometimes in combination with objective data from mobile phone operators, the use of proxy measures (e.g. distance to a mobile phone base station), RF- EMF measurements in homes, 24 hour personal measurements, modeling of mobile phone base stations or broadcast transmitter radiation. It is not clear how well these methods reflect long-term individual exposure from all RF sources.
The aim of this study was to develop and validate a statistical RF-EMF exposure prediction model suitable for the QUALIFEX study (health related quality of life and radio frequency electromagnetic field exposure: prospective cohort study).
Personal EMF measurements by means of portable exposure meters were conducted during one week for 166 volunteers living in the city of Basel (Switzerland). The participants filled in an activity diary and a questionnaire. In the activity diary the participants recorded place of stay and the use of cordless and mobile phones, and in the questionnaire they responded to questions about characteristics of their homes, workplaces, the use of wireless devices, behavioral aspects such as time spent in public transport, and about socio-demographic characteristics. A three-dimensional geospatial propagation model was developed for modeling average RF-EMF from fixed site transmitters (mobile phone base stations) inside and outside of buildings in the study region. To predict personal RF-EMF exposure, a multivariate regression model was developed. The models were evaluated by comparing predicted values with measured values. The agreement between these values was assessed by calculating the Spearman rank correlation coefficient. Exposure misclassification was characterized by the sensitivity and specificity using the measurement results as gold standard. For sensitivity and specificity calculations, both measured and calculated exposures were dichotomized at their 90th percentile. To investigate whether the models can predict mean weekly exposures several weeks later, a validation study was conducted 21 weeks after the first measurements. The validation study included personal RF-EMF measurements in most motivated and reliable and also in some of the highly exposed participants (32 in total) from the first measurement week.
Results and Interpretation
The proportion of variance explained by the final model was 0.52. The agreement between the predicted and the measured RF EMF was characterized by a sensitivity of 0.56 and a specificity of 0.95. The following exposure predictors were identified by the multiple regression analysis: the modeled RF-EMF at home from the propagation model, housing characteristics, ownership of wireless communication devices, and some behavioral aspects. The fact that most predictors (except modeled RF-EMF at home) are derived from questionnaire data implies that exposure can be assessed without extensive personal measurement campaigns. Since people spend considerable proportion of their time at home, it is important to precisely characterize exposure at home; therefore, it is essential to have a geospatial propagation model for the study region. The values of sensitivity and specificity for the validation study were 0.67 and 0.96, respectively. The validation study demonstrated that the models were applicable for assessment of exposure over several months.
The study results suggest feasibility of RF-EMF exposure assessment based on a geospatial propagation model and a questionnaire without extensive measurement campaigns. RF-EMF exposure can be predicted for a longer time period.