HG-9-91-01 br E mail address frank devocht bristol ac uk br
E-mail address: [email protected]
(SAR) levels that were comparable to the maximum localised SAR levels occurring in human head tissues located near to a mobile phone during a call (Wyde et al., 2016; NTP, 2018a, 2018b); however, there are some queries about these results, including their comparison with human exposures, and the study is yet to be published in peer-reviewed lit-erature (Sienkiewicz et al., 2017). It remains unclear what the biological mechanism for the associa-tion between GBM and exposure to radiofrequency radiation (RF) ra-diation from mobile phones would be, if an association were causal. Absorption of RF power leading to tissue heating has been argued to be the main mechanism by which RF can affect living tissue, but in the case of mobile phones it is unlikely to be sufficient to result in biological effects (Dewhirst et al., 2003). Alternative mechanisms have been proposed and include selective microthermal heating and dielectric heating, as well as non-thermal iron ion-mediated reactions and radical pair mechanisms, as well as several other mechanisms considered less plausible (Sienkiewicz et al., 2017).
Monitoring of trends in HG-9-91-01 tumour incidence rates was high-lighted as an important research area, because it is unclear when any
F. de Vocht
effect, if present, would be observable given the long lag between ex-posure and resulting clinical detection of brain cancers (Samet et al., 2014). Brain cancer incidence, including gliomas and especially an aggressive subtype, astrocytoma grade IV (glioblastoma multiforme [GBM]), have been increasing since the 1980s (Ostrom et al., 2014). A previous study based on English cancer incidence data and a novel causal inference framework based on synthetic counterfactuals in-dicated that the observed increase in malignant neoplasms in the temporal lobe between 1985 and 2014 was in agreement with mobile phone use as an important putative factor, but that for GBM, despite its incidence increasing over time as well, this did not deviate from ex-pected, counterfactual, trends (De Vocht, 2016; De Vocht, 2017). Si-milar trends have been observed elsewhere (Ho et al., 2014; Kim et al., 2015; Ostrom et al., 2014; Zada et al., 2012), and improvements in diagnostics techniques, especially in the elderly, is generally considered the main explanation for the observed increase in incidence, while genetic risk factors and ionizing radiation exposure are known to in-crease the risk and allergic conditions appear to decrease it (Miranda-Filho et al., 2017). A variety of other potential contributing factors have also been hypothesised to additionally explain these patterns, including hormonal contraceptives, hormone replacement therapy, statins, cer-tain infections, a variety of occupational exposures, vitamin D, alcohol, height, BMI, as well as non-ionizing radiation including RF from mobile phones (Kim et al., 2015; Miranda-Filho et al., 2017; Philips et al., 2018).
This paper further expands on previous analyses and utilises the strengths of the Bayesian causal inference work using synthetic coun-terfactuals to explore the likelihood of mobile phone use as the im-portant putative factor explaining the increases in the incidence of GBM in different anatomic brain regions and for specific malignant (other than GBM) and benign (acoustic neuroma and meningioma) subtypes in the temporal lobe specifically (based on the results from the previous study (De Vocht, 2016)).
National annual numbers of newly registered cases of malignant and benign neoplasms in the temporal lobe, based on 4-digit ICD-9 (up to 1995) and ICD-10 (post-1995) and 5-digit ICD-02 morphology codes, were obtained for the years 1985 to 2014 from the UK Office of National Statistics.
Covariates used to construct the counterfactual time series (see Statistical Methodology) included, similar to (De Vocht, 2016), annual incidence of all malignant neoplasms (except for non-melanoma skin cancer), population prevalence of current and never smokers, urbani-zation rate, and the percentage of status 3 records (record failed one or several vital validation checks on fields which are vital for inclusion in ONS tables) as a quality measure of coding, and were obtained from ONS (2016), the Health Survey for England (HSCIC, 2015) and the Worldbank (2016). Annual population estimates and median age of the UK population were replaced by age category (0–24, 25–44, 45–64, 65 + years of age; as well as 75 + and 85 + years of age) specific population estimates to incorporate changes in demographics over the time period. Annual incidence of all brain cancers was included as a covariate and, to also account for changes in medical imaging practice in the UK, the annual total number of X-rays, CT, MRI, ultrasound, fluoroscopies and radio-isotope scans (as one total number) obtained from the NHS for the years 1995–2014 ((Operations), 2014) and ex-trapolated backwards, was also included.