As we were unable to reach our target of 264 participants, we could not guarantee complete counterbalancing of order of exposures across sessions for the double-blind tests. Chi-square analysis revealed that there were no significant differences between the groups and order of exposure for the double-blind tests; however, almost half the sensitive group received the UMTS exposure first (45.5%) compared with the GSM first (27.3%) or sham first (27.3%). The order of exposure was more evenly distributed for the control group, with 35.1% receiving sham first, 36.0% receiving GSM first, and 28.9% receiving UMTS first.
The sensitive group (M = 46.1, SD = 13.5) was significantly younger than the control group [M = 54.5, SD = 15.23; t-test (t) (174) = −3.51, P < 0.01], with equal numbers of males and females in each group (sensitive: male 57.1%; control: male 57.5%; χ2 (1) = 0.002, P > 0.05). Significantly more controls (38.3%) reported having a chronic illness compared with sensitive participants [21.4% ; χ2 (1) = 4.94, P < 0.05], although there were no differences between the groups among the five most commonly reported chronic illnesses: high blood pressure, underactive thyroid, high cholesterol, asthma, and arthritis (χ2’s (1) < 4.5, P’s > 0.01). Bonferroni corrections were applied to all multiple comparisons to reduce the likelihood of familywise alpha errors.
Visual analog scales
The data for the VAS were skewed, mainly due to individuals reporting close to the end points. The data were therefore transformed into normal distributions using the square root transformation. The relaxation VAS was reversed from the others; therefore, it was transformed using the reflect and square root transformation [SQRT(10-X)]. A 3 (condition: sham, GSM, UMTS) X 2 (group: sensitive, control) mixed analysis of variance (ANOVA) was performed on the transformed data for each VAS separately for the open provocation and double-blind tests (see for means, standard errors, and F-and t-test values). For the open provocation, all VAS resulted in a significant main effect for group, with sensitive participants reporting higher levels of anxiety, tension, arousal, discomfort, and fatigue than controls, whereas controls reported higher levels of relaxation than sensitive participants. The main effect for condition [F’s (2,346) > 10.04, p’s < 0.001] and the interaction between condition and group was significant for all VAS except fatigue. Paired sample t-tests showed a significant difference between sham and GSM and between sham and UMTS conditions for sensitive participants but not for controls. Sensitive individuals reported higher levels of anxiety, tension, arousal, and discomfort and lower levels of relaxation during the GSM and UMTS conditions compared with the sham condition.
Descriptives and statistical tests for the VAS from the open provocation and double-blind tests for sensitive and control participants by exposure.
The results for the double-blind data were similar, with a significant main effect of group for all VAS and of condition for anxiety, tension, and arousal [F’s(2,312) > 3.00, p’s ≤ 0.05]. Of more interest, there were significant conditions by group interactions for anxiety, tension, arousal, and relaxation. Paired samples t-tests revealed higher levels of arousal during the UMTS compared with sham condition for the sensitive group only, as shown in . A problem in interpreting this significant effect is that a larger proportion of sensitive individuals received the UMTS compared with GSM or sham exposure in session 2 (the first of the 50-min double-blind conditions). Examination of the data showed that regardless of exposure condition, sensitive participants had a significantly higher degree of arousal during session 2 (M = 3.03) compared with session 3 [M = 2.34; t(43) = 2.64, p < .025], whereas there was no difference between sessions 3 and 4 [M = 2.32; t(43) = 0.47, p > 0.05]. To further test if there was a significant effect of the UMTS exposure on arousal when order of exposure was held constant, separate 2 (condition: UMTS, sham) X 2 (group: sensitive, control) between-subjects ANOVAs were performed for each session (see for mean, standard error, and F values). The main effect for condition and group by condition interaction was not significant for all three sessions. These results indicate that the apparent increase in arousal with UMTS exposure was attributable to the higher proportion of sensitive individuals who received UMTS in session 2 (45.5%). It is important to note that regardless of type of exposure or session, all the VAS scores fell within the lower “not at all” end of the scale.
Descriptives and statistical tests for level of arousal by session by group.
The majority of control individuals reported experiencing no symptoms in any condition; therefore, Wilcoxon signed-ranks tests were performed on the total number of symptoms reported and total symptom scores (see for medians and Z scores). During the open provocation the sensitive group reported more symptoms and a higher total symptom score during the GSM and UMTS conditions compared with sham. The control group reported more symptoms during the UMTS compared with sham, but not for GSM compared with sham. During the double-blind tests there was no difference between active and sham exposures in either the total number of symptoms or the total symptom score for either group. Sensitive participants reported more symptoms than did controls, but this was not related to exposure condition.
Medians and Z-scores for total number of symptoms and total symptom score from open provocation and double-blind tests for sensitive and controls by exposure.
Inspection of the physiological data revealed that it was skewed for all measurements except the HR (M). Square root transformations were applied to the BVP (SD), SC (M), and SC (SD). A logarithmic transformation was applied to the HR (SD) to form normal distributions (see means, standard errors, and F values). The BVP (M) did not lend itself to transformation or analysis because of low kurtosis values. The data were analyzed using a 3 (condition: sham, GSM, UMTS) X 2 (group: sensitive, control) mixed ANOVA for the open provocation and double-blind tests. There was no difference between active and sham conditions regardless of type or even knowledge of exposure for either group. There was, however, a significant between-group difference in SC, with sensitive participants having higher SC (M and SD) responses during the open provocation and double-blind tests. The HR (SD) was also significantly higher in the sensitive group compared with that in the control group during the open provocation test. No other comparisons were significant.
Descriptives and statistical tests for physiological measures for sensitive and control participants by exposure during open provocation and double-blind tests.
Participants made “on”/”off” judgments during both the 5-min and 50-min double-blind exposures. Sensitive participants had an accuracy rate of 55.2% during the 5-min tests (d′ = −0.08, sensitivity = 66.4%, specificity = 32.7%) and 59.8% during the 50-min tests (d′ = 0.20, sensitivity = 69.3%, specificity = 40.9%). The control group had an accuracy rate of 51.4% during the 5-min tests (d′ = 0.10, sensitivity = 51.7%, specificity = 50.8%) and 50.1% during the 50-min tests (d′ = 0.06, sensitivity = 48.0%, specificity = 54.3%). See for ROC curves and 95% confidence intervals (CI). For each group the 95% CI on the ROC curves includes the diagonal axis, implying that participant performance for each group did not differ from chance. Only two sensitive and five control participants were able to correctly identify all six “on”/ “off” judgments, while no one correctly distinguished between the GSM and UMTS signal 100% of the time.
ROC curve and 95% CI values for “on”/”off” judgments for sensitive and control participants. (A) ROC curve for the 5-min double-blind sessions. (B) ROC curve for the 50-min double-blind sessions.