Use of mobile phones and progression of glioma incidence since 1979

Study snapshot

Abstract

Use of mobile phones and progression of glioma incidence since 1979 My notes: Although the title of this IARC report is in German, the report is available in English. The report's summary, "no indications of a detectable effect of mobile phones have been found," seems misleading because it is inconsistent with the report's final conclusion, namely "An increased risk in the 10% heaviest mobile phone users was an exception to this general situation, as it remained plausible." The report's bottom line: "This ecological data is not sufficient to dismiss every potential mobile phone related risk scenario, but suggests that the risk – if it exists - would be very small, only occur after very long latency periods of several decades, or only affect small subgroups within glioma patients." If only a portion of the population has a genetic susceptibility to brain cancer in the presence of microwave radiation as appears to be the case with thyroid cancer (Luo et al., 2020), that could explain why the odds ratios obtained for brain cancer risk from case-control studies of heavy, long-term mobile phone users over- predict glioma incidence in the overall population based upon tumor registry data. * Luo J, Li H, Deziel NC, Huang H, Zhao N, Ma S, Nie X, Udelsman R, Zhang Y. Genetic susceptibility may modify the association between cell phone use and thyroid cancer: A population-based case-control study in Connecticut. Environmental Research. 2020 Mar;182:109013. doi: 10.1016/j.envres.2019.109013. Deltour I, Schuz J. Nutzung von Mobiltelefonen und Verlauf der Gliom-Inzidenz seit 1979: Vorhaben 3618S00000 (FM 8867). International Agency for Research on Cancer. Jun 2022. Summary 1.1 Introduction In the Nordic countries, the sharp increase in the use of mobile phone occurred in the mid-1990s among adults; thus, time trends in glioma incidence rates (IR) may provide information about possible risks associated with mobile phone use. We investigated time trends in IR of glioma, and compared IR and observed number of cases to those that would be expected under a range of hypothetical mobile phone risk scenarios, encompassing risk levels reported in published case-control studies. 1.2 Methods We analyzed age standardised IR of glioma in Denmark, Finland, Norway, and Sweden among adults 20-84 years old, using data from national cancer registries and population data covering the period 1979-2016, using a log linear joinpoint analysis. Exposure distribution of use and of high level of use were obtained from self-reported information in the Nordic Interphone, the Cosmos-Denmark and the Cosmos-France datasets. Based on analytical epidemiological studies, we considered various scenarios according to which mobile phone use would hypothetically increase the glioma risk. We quantified compatibility, or absence of compatibility between the observed data and the risk scenarios by projecting incidence rates of glioma of men aged 40-69 years old under these scenarios and comparing them with the observed incidence rates in the Nordic countries. 1.3 Results Glioma IR increased regularly with annual percent change (APC) of 0.6 (95% confidence interval (CI) 0.4-0.7) in men and 0.3 (95%CI 0.2-0.5) in women in the period 1979-2016. There were hardly any changes in IR among men and women below age 59. In men and women in their sixties, IR increased by 0.6 (95%CI 0.4-0.9) in men and 0.4 (95%CI 0.2-0.7) in women, regularly for the whole period of observation, while IR among 70-84 years old increased very markedly, with APC of 3.1 (95%CI 2.6-3.5) among men and 2.8 (95%CI 2.3-3.3) among women over at least the last 2 decades of observation. Very few risk scenarios appeared compatible with the observed data using standardised incidence ratios analyses. The risk scenarios that appeared compatible involved either long latencies (20 years), or very low risks (RR = 1.08); in these projections, risks that would be limited to mobile phone heavy users were not compatible with the observed number of cases. 1.4 Discussion IR time trends did not demonstrate breakpoints in their secular evolution in the last 20 years. Virtually all the reported results from the case-control studies with a positive association between mobile phone use and glioma risk were shown to be implausible in our simulations comparing them with the observed incidence rates, implying that biases and errors have likely distorted their findings; very low risks at the population level, and risks after very long latencies remained plausible. Simulations were based on high quality case registration, which is a strength, while the uncertainties in the exposure information and the limited information about some of the model’s assumptions were limitations. Altogether, this study confirms and reinforces conclusions made previously, that no indications of a detectable effect of mobile phones have been found. Excerpts ... We analyzed the time trends in the incidence rates of glioma among adults aged 20 to 84 years of the Nordic countries from 1979 to 2016 (step 1 of the work description). Then, we addressed the question whether the observed time trends and observed number of cases were statistically different from the one we would observe if we assumed that the use of mobile phones caused glioma, so if we assumed that there was a true causal association (step 2 and step 3 of the work description). Within this, we delineated the levels of risks and the duration of induction periods that would not be compatible with the observed time trends and numbers of cases in this population (step 3 of the work description). We also discussed these findings in light of some of the elevated OR found in the literature. The study tested the consistency between risks that have been reported and the effect they would have had at the level of the population, had they been true. Noteworthy, the study was not meant to dismiss every single hypothetical association, as it would most likely always be possible to devise a pattern of risk that would fit the data.... This study was based on 28,015 male and 20,630 female glioma cases diagnosed from 1979 to 2016 in Denmark, Finland, Norway and Sweden (called “the Nordic countries” in the following). In 2016, the number of glioma cases was 1,724 in a population of 19.7 million adults aged 20–84 years. Over the last 10 years of data, Sweden accounted for 38% of the population and of the cases; of the remainder, Denmark, Finland, and Norway had populations of similar size. The age-standardized incidence rates were higher in men (9.1 per 100000 person years) than in women (6.1 per 100 000 person years), and higher with increasing age. All countries had comparable rates; Norway had slightly higher rates, while Finland had slightly lower rates in both sexes (Table 2 and Table 3). Joinpoint analyses described in paragraph 6.1 showed that overall, the trends were smooth: glioma rates increased by 0.6% (95% CI 0.4%-0.7%) per year in men and 0.3% (95% CI 0.2%-0.5%) per year in women over the period 1979-2016 in the Nordic countries combined (Table 4 and Table 5), and in each country separately except for a marked increase in 1979-1984 in Swedish men (APC about 6%). For the younger age groups (20-39 and 40- 59 years old), the time trends were smooth and did not demonstrate strong increases at any point in time during the period 1979 to 2016 in any country among men (Table 6), and women (Table 7). Below the age of 60, incidence rates were generally stable over the whole period (Figure 1, Table 6 and Table 7). Among people aged 60-69 years old, incidence rates increased gradually by 0.6% in men and 0.4% in women per year, and these regular increases with no joinpoint were observed in every country and at a very similar rate in both sexes, except among Swedish women, whose rates showed a slight decrease. Irregular patterns were observed among the persons aged 70-84 years old at the beginning of the observation period, while for at least the last 12 years of observation, all countries showed highly increasing rates. Exceptions to this general pattern were noted among the Finnish males and the Norwegian females, in which an increase was seen at the beginning of the observation period that lasted at least 21 years. The analysis by subgroups of tumour types could be performed only for the period 1990-2016 for reasons of data availability: in Sweden, a separate code for glioblastoma did not exist prior to 1993, and very few of the tumours which had been diagnosed during the period 1990-1992 were retrospectively coded into this code. Indeed, cancer registries are continuously updated when additional information becomes available on an earlier diagnosis, for example. Among men and women, the rates of glioblastomas increased in the last years of observations, while the rate of other high-grade gliomas decreased (Table 8 and Table 9). Rates of low grade gliomas were relatively stable in all countries since the mid 1990’s except in Denmark, where substantial increases were noted towards the end of the period of observation, albeit non-significant.... When examining the trends by subtypes, glioblastoma generally increased while other high grade gliomas decreased, and low grade glioma were stable in the most recent period, except in Denmark where low grade glioma rates increased among men and women in the last 3 years of observation. In Sweden, the rates of glioblastoma underwent most changes, namely the increase in glioblastoma rates in Sweden in the years after the introduction of that code by the cancer registry, since a new code is not mandatorily fully used immediately after it is introduced.... To sum up, our simplified and more sophisticated analyses appeared to indicate that the small increase in IR of men age 40-59 and the marked increase in RR of men aged 60-69 were generally not compatible with the same mobile phone related risks increases. When models in which the totality of the IR increases were assumed to be associated with mobile phone effects, a RR of 1.31 that would start 20 years after first using a mobile phone was borderline compatible between these 2 age groups, while all other induction periods (0, 5, 10, 15 years) or heavy users risk scenarios produced RR estimates and CI which did not overlap between the 2 age groups when the same exposure distribution was considered. When half of the IR increases were attributed to other factors, none of the mobile phone related risks scenarios were compatible with the data, in the SIR analyses (assuming the same risk in both age groups). When most (75%) of the IR increases were attributed to other factors, then small excess risks (RR= 1.08 applying to all users after 10 years) or risks after long latencies (RR = 1.3 applying to all users after 20 years) were compatible with the observed incidence rates and exposure distributions that we assumed. Further work on these scenarios could shed more light on the remaining uncertainties. Of note, scenarios of risks limited to heavy users groups did not appear compatible with the observed number of cases in these analyses.... Our simulation study is not free of assumptions. The induction period relating mobile phone use and glioma risk, if such an association exists, is unknown, so is the magnitude of the risk, and the real patterns may be more complex than the scenarios that we simulated. In addition, there are several factors that we did not account for. The coverage of the Nordic cancer registries was not complete, but some 1.5% to 10% of the malignant tumours were missed in this age group. In Sweden, it has been estimated that completeness would not have changed over the period 1998-2014, while completeness might have improved in other countries. We modelled that other, yet to be discovered, risk factors of the disease as well as improvement in its detection and reporting had a smooth, gradual impact, over the period 19792016, which is consistent with the gradually increasing IR. We used 3 sources of information on the use of mobile phones, all self-reported, to evaluate the prevalence of use and heavy use up to 2002, 2008 or 2016 and extrapolated the prevalences for the periods and age groups for which no data was available, based on the trends observed in the other age groups. The use of hands-free devices was not accounted for, although this was not frequent in these populations (data not shown). In conclusion, it is difficult to demonstrate the absence of risk, in real life condition, and assumptions about the impact of the improvement of diagnosis tools, treatment and registration changes over time were used in our simulations. However, based both on the observed IR and the simulations, we reiterate and strengthen our previous conclusion that, the risk, should one exist, ought to be lower or occur after a longer induction period or act on a smaller population, or a combination of these, than most of the level of risk that have been reported in previously published case-control studies. Conclusions In this project we projected incidence rates of glioma under various scenarios of mobile phone-associated increased glioma risks, and compared them with the observed incidence rates in the Nordic countries. The comparison was carried out on the data of men aged 40 to 69 years. The modelled scenarios included risk increases reported from analytical epidemiological studies, which were all of case-control design. Most of those results were shown to be implausible in our simulations, implying that biases and errors in the self-reported use of mobile phones have likely distorted their findings. An increased risk in the 10% heaviest mobile phone users was an exception to this general situation, as it remained plausible. Results of cohort studies showing no association were compatible with observed incidence rates. We also studied what hypothetical mobile phone- related risks were conceivable if the changes in incidence rates in 40-59 year old and 60-69 year old men were fully attributable to mobile phone use. The fact that we observed different hypothetical risks in these two age groups while research at present has not suggested that older men should have higher risk related to mobile phone use than younger men, does not align with the assumption that mobile phone exposures caused the incidence rate trends. This ecological data is not sufficient to dismiss every potential mobile phone related risk scenario, but suggests that the risk – if it exists - would be very small, only occur after very long latency periods of several decades, or only affect small subgroups within glioma patients. Open access report: doris.bfs.de 2022063033222/4/BfS_2022_3618S00000.pdf

Main findings

Glioma incidence rates increased gradually over 1979-2016 with no abrupt changes. Most reported increased risks from case-control studies were incompatible with observed population incidence trends. Only scenarios with very low risk increases or long latency periods (20 years) were compatible with observed data. Increased risk limited to the 10% heaviest mobile phone users remained plausible. Overall, the data suggest any mobile phone related glioma risk is very small, occurs after long latency, or affects small subgroups.

Outcomes measured

• glioma incidence rates

Limitations

• Ecological study design limits causal inference
• Exposure data based on self-reported mobile phone use with uncertainties
• Incomplete cancer registry coverage (1.5%-10% missing cases)
• Assumptions about latency periods and risk magnitude
• Did not account for use of hands-free devices

Related papers

• Time trends in mobile phone use and glioma incidence among males in the Nordic Countries, 1979-2016 2022
• Mobile Phone Use and Time Trend of Brain Cancer Incidence Rate in Korea 2021
• Inferring the 1985-2014 impact of mobile phone use on selected brain cancer subtypes 2016
• Glioma incidence and survival variations by county-level socioeconomic measures 2019
• Mobile phone use and incidence of brain tumour histological types, grading or anatomical location: a population-based ecological study 2018
• Mobile phone and cordless phone use and the risk for glioma - Analysis of pooled case-control studies in Sweden, 1997-2003 and 2007-2009 2015