Re: Mobile phone use and glioma risk: comparison of epidemiological study results with incidence trends in the United States
We clearly state in the statistical methods that while the figures are based on a specific reference group (males aged 60-64 in the Los Angeles registry), the entire dataset (all ages ≥18 years, sexes, registries) was used to estimate these rates. The reference group was chosen purely for statistical reasons, to minimise the variance of estimates – choosing a different group would make little difference to inference on trends. We also show projected rates in Appendix Tables A5-A10 for all relevant age/gender/ethnic groups and the results shown in the figures broadly support results for the whole study population. Kundi argues that population mobility may invalidate predicted rates derived from the studies of Hardell et al. (1) and Interphone (2). Population mobility undoubtedly varies with age, but since at each age numbers of cancers are related to accurate estimates of population at risk in the areas covered by the SEER registries, risks of brain cancers in any year can be predicted by the odds ratios derived from the studies of Hardell et al. (1) and Interphone (2).
For comparability with the Interphone study (2), and many other studies referred to in the Appendix, we used only the results for mobile phones in Hardell et al. (1), although clearly results would not be very different had we used cordless phone data from Hardell et al. (1) as well or instead. As we state in the Introduction, our objective was to explore the two epidemiological studies that formed the basis of the IARC Working Group classification of cancer risk associated with mobile phones (3); the other papers cited by Hardell et al. are therefore irrelevant.
As we note in the foonote to Table 2, there was some indeterminacy in the periods of latency cited by Hardell et al. (1); however, using the various alternative possibilities suggested by that paper did not markedly change our findings. The results for oligodendroglioma cited in Table A1 are not used in the analysis.
Kundi expresses concern with formula (A2). The issue involves inferring the proportions of people first exposed in the period 1-<5 / 5-<10 / ≥10 years before from the proportion of population that were phone users 1 / 5 / 10 years before, CPy-1 / CPy-5 / CPy-10. We constructed a semi-conservative lower bound to the excess relative risk for the dominating first term in (A2) (see below) when the relative risk RR1l ≥ 1 (which is the case for the studies of Hardell et al. (1) and Interphone (2) (the latter only when we impose the constraint RR≥1)). The bound is easily obtained by assuming that the groups corresponding to the proportions CPy-1, CPy-5 are nested within each other, but that those corresponding to CPy-5, CPy-10 are disjoint. Not subtracting off CPy-10 results in the projected numbers based on Hardell et al. (1) slightly increasing (projected 2008 glioma rates go from 25.5 to 26.3 per 100,000 per year), thereby augmenting the discrepancy with the observed US rate; the increase in the rates relating to projections based on Interphone (2) is even more modest (projected 2008 glioma rates go from 18.2 to 18.3 per 100,000 per year).
We agree that one cannot assume that the proportions of mobile phone use are the same for all US population subgroups. The projected rates for 2008 are heavily dominated by risk for those first using phones in the most recent 5 years (the first term in (A2)), when US usage approached 100% – over 95% of the 44.7% excess relative risk predicted for the US population by the study of Hardell et al. (1) comes from this group – so that heterogeneity in patterns of phone use would have little impact. Our study does not provide confirmatory evidence of a 1.5 to 2-fold increased glioma risk with decade-long regular use as stated by Davis et al.
Reference List
(1) Hardell L, Carlberg M, Hansson MK. Pooled analysis of case-control studies on malignant brain tumours and the use of mobile and cordless phones including living and deceased subjects. Int J Oncol 2011; 38(5):1465-1474.
(2) Brain tumour risk in relation to mobile telephone use: results of the INTERPHONE international case-control study. Int J Epidemiol 2010; 39(3):675-694.
(3) Baan R, Grosse Y, Lauby-Secretan B, El Ghissassi F, Bouvard V, Benbrahim-Tallaa L et al. Carcinogenicity of radiofrequency electromagnetic fields. The Lancet Oncology 2011; 12(7):624-626.
Competing interests:
No competing interests
23 April 2012
Mark P Little
Senior Scientist
Preetha Rajaraman, Rochelle E Curtis, Susan S Devesa, Peter D Inskip, David P Check, Martha S Linet
National Cancer Institute
Radiation Epidemiology Branch, 6120 Executive Boulevard, Rockville MD 20852-7238, USA
Rapid Response:
Re: Mobile phone use and glioma risk: comparison of epidemiological study results with incidence trends in the United States
We clearly state in the statistical methods that while the figures are based on a specific reference group (males aged 60-64 in the Los Angeles registry), the entire dataset (all ages ≥18 years, sexes, registries) was used to estimate these rates. The reference group was chosen purely for statistical reasons, to minimise the variance of estimates – choosing a different group would make little difference to inference on trends. We also show projected rates in Appendix Tables A5-A10 for all relevant age/gender/ethnic groups and the results shown in the figures broadly support results for the whole study population. Kundi argues that population mobility may invalidate predicted rates derived from the studies of Hardell et al. (1) and Interphone (2). Population mobility undoubtedly varies with age, but since at each age numbers of cancers are related to accurate estimates of population at risk in the areas covered by the SEER registries, risks of brain cancers in any year can be predicted by the odds ratios derived from the studies of Hardell et al. (1) and Interphone (2).
For comparability with the Interphone study (2), and many other studies referred to in the Appendix, we used only the results for mobile phones in Hardell et al. (1), although clearly results would not be very different had we used cordless phone data from Hardell et al. (1) as well or instead. As we state in the Introduction, our objective was to explore the two epidemiological studies that formed the basis of the IARC Working Group classification of cancer risk associated with mobile phones (3); the other papers cited by Hardell et al. are therefore irrelevant.
As we note in the foonote to Table 2, there was some indeterminacy in the periods of latency cited by Hardell et al. (1); however, using the various alternative possibilities suggested by that paper did not markedly change our findings. The results for oligodendroglioma cited in Table A1 are not used in the analysis.
Kundi expresses concern with formula (A2). The issue involves inferring the proportions of people first exposed in the period 1-<5 / 5-<10 / ≥10 years before from the proportion of population that were phone users 1 / 5 / 10 years before, CPy-1 / CPy-5 / CPy-10. We constructed a semi-conservative lower bound to the excess relative risk for the dominating first term in (A2) (see below) when the relative risk RR1l ≥ 1 (which is the case for the studies of Hardell et al. (1) and Interphone (2) (the latter only when we impose the constraint RR≥1)). The bound is easily obtained by assuming that the groups corresponding to the proportions CPy-1, CPy-5 are nested within each other, but that those corresponding to CPy-5, CPy-10 are disjoint. Not subtracting off CPy-10 results in the projected numbers based on Hardell et al. (1) slightly increasing (projected 2008 glioma rates go from 25.5 to 26.3 per 100,000 per year), thereby augmenting the discrepancy with the observed US rate; the increase in the rates relating to projections based on Interphone (2) is even more modest (projected 2008 glioma rates go from 18.2 to 18.3 per 100,000 per year).
We agree that one cannot assume that the proportions of mobile phone use are the same for all US population subgroups. The projected rates for 2008 are heavily dominated by risk for those first using phones in the most recent 5 years (the first term in (A2)), when US usage approached 100% – over 95% of the 44.7% excess relative risk predicted for the US population by the study of Hardell et al. (1) comes from this group – so that heterogeneity in patterns of phone use would have little impact. Our study does not provide confirmatory evidence of a 1.5 to 2-fold increased glioma risk with decade-long regular use as stated by Davis et al.
Reference List
(1) Hardell L, Carlberg M, Hansson MK. Pooled analysis of case-control studies on malignant brain tumours and the use of mobile and cordless phones including living and deceased subjects. Int J Oncol 2011; 38(5):1465-1474.
(2) Brain tumour risk in relation to mobile telephone use: results of the INTERPHONE international case-control study. Int J Epidemiol 2010; 39(3):675-694.
(3) Baan R, Grosse Y, Lauby-Secretan B, El Ghissassi F, Bouvard V, Benbrahim-Tallaa L et al. Carcinogenicity of radiofrequency electromagnetic fields. The Lancet Oncology 2011; 12(7):624-626.
Competing interests: No competing interests