Re-analysis of risk for glioma in relation to mobile telephone use: comparison with the results of the Interphone international case-control study.

The long-awaited Interphone study on use of mobile phones and the risk of brain tumour was recently published. It was coordinated by International Agency for Research on Cancer (IARC) and included 16 research centres from 13 countries. Results for cases aged 30–59 years of age diagnosed during study periods of 2–4 years between 2000 and 2004 were presented. Our research group has published results for brain tumour risk and long term use of mobile phones. In contrast to Interphone, we also included use of cordless phones. Radiofrequency emissions from a cordless phone are in the same magnitude as from a digital mobile phone, as discussed in our publications and recently shown also by Redmayne et al. Moreover, cordless phones are used for longer calls. Including such use in the ‘unexposed’ group as in the Interphone study would bias the odds ratio (OR) towards unity. We have previously compared study methods and results in our investigations with those in the Interphone group. Due to the lack of information and any discussion of the Interphone findings in relation to our results, it is pertinent to use the same criteria as in Interphone for our case–control studies on glioma. Our inclusion period was 1997–2003 and we give results for all glioma for the same age group, 30–59 years as in Interphone (Table 1), and glioma located in the temporal lobe (Table 2). Overall results are also presented for our studies as well as inclusion of the youngest subjects 20–29 years and in one analysis including use of cordless phones among the unexposed. We have also re-analysed our material with the same cumulative exposure time as in the Interphone study, i.e. 41640 h, whereas we before had 42000 h as highest exposure. In Appendix 2, in the Interphone paper, analysis was restricted to users with lowest category of use as reference in each category. There might be a ‘healthy mobile phone user’ effect among the controls that participated, similar to a ‘healthy worker effect’ in occupational studies. Thus, the analysis in Appendix 2 would be justified to correct for the lower prevalence of mobile phone use among controls that refused to participate than among included controls in Interphone. As can be seen in Table 1, our results in the same age group as in Interphone, 30–59 years, are similar as in Appendix 2 for latency 510 years and cumulative use 51640 h. Unfortunately, Interphone did not give results for laterality analysis in Appendix 2. Interestingly, our results for cumulative use in the age group 30–59 years are similar to Interphone results. Furthermore, in both studies highest ORs were found for ipsilateral use. We found higher risks if the age group 20–29 years was included. This is in agreement with our previous publication showing highest risk for persons that started use of mobile or cordless phone before the age of 20 years. Thus, excluding that age group from the final Interphone seems to have biased the risk towards unity. We examined the results if we considered use of cordless phone as involving no exposure to microwaves, which yielded lower ORs indicating that excluding such use, as in Interphone, would also bias the risk towards unity. Table 2 gives the results for glioma in the temporal lobe. Similarly, as for overall findings, risk estimates were lower in our studies when we restricted the age group to 30–59 years and considered use of cordless phone as no exposure. No results were given in Appendix 2 in the Interphone publication for glioma in the temporal lobe. The participation rate in the Interphone study was only 64% for glioma cases and 53% for controls, i.e. much lower than in the studies from the Hardell group, 90% of cases with malignant brain tumour and 89% of the controls. Furthermore, we used a self-administered questionnaire that was supplemented over the phone. This was done without knowing whether it was a case or a control. Low-participation rate may create selection bias, and not blinding as to case or control status may give observational bias, especially in a study with such vague definition of cut-off for exposure as Interphone. Published by Oxford University Press on behalf of the International Epidemiological Association