br F de Vocht br include mobile phone
F. de Vocht
include mobile phone usage. Regardless, although this study (nor any of the previous ones) can unambiguously in- or exclude mobile phone usage as a contributing causal factor, data are broadly in agreement that the contribution of mobile phone usage in causing GBM or other types of CFTRinh-172 cancer, if any, is likely to be small.
This study has several important limitations. Foremost, the in-ference made about the (causal) effects rely on the assumption that the relationships between the numbers of newly diagnosed cases and the variables in the model on which the counterfactuals are based remain consistent before and after the hypothesised point where an effect from the introduction of mobile phones would be measurable in national statistics.
The assumption that a 10-year lag was the most plausible period between first exposure and when increased risk could be observed in registry data was based on the previous analyses (De Vocht, 2016). Although sensitivity analysis using a 15-year lag showed no evidence of excesses relative to counterfactuals, this may still have been too short. It has, however, been observed that the 10-year lag corresponds to a plausible maximum impact after first introduction of mobile phones in society of several decades (Ahlbom et al., 2009); a pattern that was previously also observed for asbestos and cancers in atomic bomb survivors (Walker, 1984; Furukawa et al., 2009). Although incon-sistencies in coding and coverage could affect the evaluation of time series of cancer incidence (Zada et al., 2012) the UK cancer registry is considered to be of high quality (IARC, 2017); it seems therefore un-likely any coding errors would have significantly impacted on these results. However, improvement in diagnostic techniques and practices (other than the number of scans alone, which was included in these analyses), especially in the elderly, seems a plausible explanation for the observed effects, at least in part, and has happened roughly over the same time period as the introduction of mobile phones into society (Davis et al., 1990; Greig et al., 1990).
Despite its rigorous methodology and testing of a priori specified hypotheses, this remains an ecological study with its known limitations, including the ecological fallacy. Moreover, no quantitative data were available on other potential putative factors, environmental, diagnostic or otherwise, for inclusion in the models. This would be beneficial for future work. Information on laterality would also have been beneficial to further strengthen the inferences (Carlberg and Hardell, 2015), but are similarly not available for these data.
The main strength of this research is the use of Bayesian structural time-series and the use of synthetic counterfactuals, which are a flexible approach to modelling and forecasting of time series and, given certain assumptions, here enable the distinction between the temporal in-creases in newly diagnosed cases of brain cancer subtypes which are nonetheless not different from expected trends and those that are in excess of expected, counterfactual, trends. In addition, this metho-dology uses Bayesian model averaging to make estimation of the counterfactuals relatively insensitive to the specific choice of ex-planatory factors and model specifications. These analytic features provide a stronger causal framework compared to recent analyses of similar data from England relying solely on descriptive analyses of trends (Philips et al., 2018). A further strength is that these analyses are based on high-quality data, which enabled the thorough investigation of temporal trends in the incidence of specific brain cancer subtypes and benign neoplasms in the temporal lobe specifically, as well as of GBM in different anatomical regions of the brain.
This study, in agreement with other data from the UK and else-where, shows that the incidence of glioblastoma multiforme (astro-cytoma grade IV) has increased significantly since the 1980s, especially in the frontal and temporal lobes and cerebellum. However, it further provides evidence that the trend of increasing numbers of newly diag-nosed cases of glioblastoma multiforme in the temporal lobe (but likely Environmental Research 168 (2019) 329–335
in the frontal lobe and cerebellum as well) since the mid-1980s, al-though seemingly consistent with the hypothesis of exposure to radio-frequency radiation from mobile phones being an important putative factor, should to a large extent (if not exclusively) be attributed to another factor or factors; of which improvements in diagnostic tech-niques, especially in the elderly, seems the most plausible. Although these analyses indicate that it is unlikely that exposure to RF from mobile phones is an important putative factor, they also cannot exclude it as a contributing factor completely. It is therefore important to keep monitoring incidence trend data.