The present study investigated whether temporal trends in glioma morbidity provide evidence of possible risks from mobile phone use and tested the compatibility of the number of cases observed with different risk scenarios based on the assumption that mobile phone use increases the risk of glioma.

For this purpose, incidence rates from the cancer registries of Denmark, Finland, Norway, and Sweden were used. These data were chosen for the study because health care, cancer diagnostics, and cancer registration are of particularly high quality in these countries.

A) Description of the development of glioma morbidity over time

In a descriptive evaluation, Deltour and Schüz described the development of age-standardised glioma incidence rates between 1979 and 2016 and used models (joinpoint analysis with log-linear models) to investigate whether there are points in time at which glioma morbidity changes. They evaluated the morbidity separately according to sex, age at diagnosis, and type of glioma (glioblastoma, other high-grade gliomas, low-grade gliomas).

B) Examination of the compatibility of different risk scenarios with the frequencies of disease cases observed

To test the plausibility of different risk scenarios, Deltour and Schüz conducted three different types of analyses. In addition to the data available on the disease, they also needed "data" on mobile phone use. Because no data on this are available for the population in the four countries considered over the observation period, the proportions of mobile communications users and sub-groups such as intensive users were estimated via the corresponding proportions from three different studies ("Nordic Interphone", "COSMOS-Denmark", and "COSMOS-France").

In contrast to other study designs (e.g. case-control studies), in the present study, it was not known for the individual glioma cases or for the controls (individuals not suffering from glioma) whether a mobile phone was used. Only the total number of cases in each calendar year and the number of individuals in the age groups considered in each calendar year were known. Based on the number of individuals in an age group and the distributions on mobile phone use from the aforementioned studies, the proportions of exposed and unexposed individuals were estimated for each age group for each year.

Because individual data was not used for these evaluations, this is considered an "ecological approach". The risk scenarios investigated are characterised by the exposure distribution chosen from other studies (Interphone Nordic, COSMOS-Denmark, or COSMOS-France), by the user group (all users vs intensive users of various intensities), and by the assumed induction period. This refers to the period between the first mobile phone use and the onset of the disease (0, 5, 10, 15, or 20 years).

Deltour and Schüz used only data from men for these analyses because they used mobile phones considerably more often than women in the early days of mobile telephony in the 1990s.

B 1) Estimation of the risk of glioma from mobile phone use

In this evaluation, Deltour and Schüz estimated the ratio between the risk of disease of mobile phone users and the risk of disease of individuals who did not use mobile phones (i.e. the relative risk) for the different hypothetical risk scenarios. To do this, they used a model that includes the assumption that any change in the morbidity between 1979 and 2016 was due solely to mobile phone use.

This evaluation was conducted for two age groups: Age of onset of 40–59 years or 60–69 years. The age group 40–59 years was chosen because it fits relatively well with the study population of the Interphone study, which included individuals whose disease was diagnosed between the ages of 30 and 59 years. The second age group, 60–69 years, was chosen because men of this age are presumably the group of individuals who were the first to increasingly use mobile phones.

B 2) Power analysis

In a simulation study, Deltour and Schüz investigated how likely it is that a significantly increased risk of glioma can be detected if the various hypothetical risk scenarios in the study are valid. For example, the level of the assumed risk or the number of individuals in the group under consideration is relevant here. Again, the statistical model was used which includes the assumption that any change in the morbidity between 1979 and 2016 was due solely to mobile phone use. These analyses were carried out only for men aged 40–59 years because the longest period between exposure and tumour development can be modelled for them.

B 3) Analysis of standardised incidence rates (SIR analysis)

With another simulation study, Deltour and Schüz tested the compatibility of different risk scenarios with the case numbers observed. They compared the number of cases observed with the number of cases that would be expected if a certain risk scenario were valid. This evaluation was done without assuming that any change in morbidity was due to mobile phone use. Instead, these analyses assumed that 50% and 75% of the change in morbidity was due to other unknown factors. Such factors could include advances in medical diagnostics or improvements in cancer registration.

These analyses were conducted for the age groups 40–59 years and 60–69 years.

A) Description of the development of glioma morbidity over time

From 1979 to 2016, 28,015 new cases of glioma were registered in men and 20,630 in women in the four countries considered: Denmark, Finland, Norway, and Sweden.

• Among those in the younger and middle age groups (ages 20–39 years and 40–59 years, respectively), the glioma morbidity remained relatively constant for both women and men between 1979 and 2016.
• Among the 60–69 year-olds, the rate increased significantly between 1979 and 2016: for men by 0.6% per year (95% confidence interval: 0.4% to 0.9%) and for women by 0.4% per year (0.2% to 0.7%).
• Among 70–84 year-olds, the course changed several times. For example, the morbidity among men in this age group increased by 7.3% (−0.4% to 15.6%) per year from 1979 to 1984, decreased by 4.5% (−11.0% to 1.9%) per year from 1984 to 1990, and increased by 3.1% (2.6% to 3.5%) per year from 1990 to 2016. For women, there was a weak increase of 0.2% (−1.0% to 3.0%) per year between 1979 and 1994 and a statistically significant increase of 2.8% (2.3% to 3.0%) per year between 1994 and 2016.
• For the different types of gliomas, the course of the morbidity differed between 1990 and 2016. Whilst the incidence rates of glioblastomas increased, the morbidity of other high-grade gliomas decreased, and those of low-grade gliomas remained approximately stable.

B 1) Estimation of the risk of glioma from mobile phone use

Considering all users and the age group 40–59 years, relative risks ranged from 1.05 (exposure distribution from the Nordic Interphone study, induction period 15 years) to 1.11 (exposure distribution from Cosmos-France, induction period 20 years). The risks for intensive users were higher and ranged up to 1.75 (exposure distribution of Cosmos-Denmark, duration of use before 2003 > 1,640 h). For the age group 60–69 years, the risks ranged from 1.21 (exposure distribution from the Nordic Interphone study, induction period 0 years) to 1.54 (exposure distribution from Cosmos-France, induction period 20 years). Also for this age group, the risk estimates for intensive users were higher and ranged up to 6.89 (exposure distribution from the Nordic Interphone study, duration of use before 2003 > 1,640 h).

A comparably high relative risk for both age groups resulted only for an induction period of 20 years when using the Cosmos-France exposure distribution. This risk was 1.3.

B 2) Power analysis

The power analysis showed that such a pooled analysis with data from four nations has sufficient size and thus enough statistical power to show even slightly increased risks.

B 3) Analysis of standardised incidence rates (SIR analysis)

The SIR analysis showed that for the age group of 40–59 year-old men, risk scenarios with hypothetical relative risks up to 1.5 are compatible with the case numbers observed. For the age group of 60–69 year-old men, risk scenarios with higher relative risks are compatible with the case numbers observed. For both age groups, low relative risks were compatible with the data for scenarios with shorter induction periods, whilst higher relative risks were more compatible with the data for long induction periods.

In the scenarios in which it was assumed that 50% of the change in morbidity would be due to other unknown factors, none of the risk scenarios were consistent with the data for both age groups. In the scenarios in which it was assumed that 75% of the change in morbidity would be due to other unknown factors, three scenarios were consistent with data for both age groups: assuming the exposure distribution of Nordic Interphone, a relative risk of 1.08 for a 10-year induction period and a relative risk of 1.3 for a 20-year induction period; assuming the exposure distribution of Cosmos-Denmark, also a relative risk of 1.08 for a 10-year induction period.

For intensive users, none of the risk scenarios for both age groups is consistent with the data.

Overall, Deltour and Schüz conclude that their results do not support a detectable effect of mobile phone use on the risk of glioma. A central reason for this is that the course of incidence rates over the last 20 years does not show any corresponding "breakpoints". In their opinion, the increases observed in the age groups 60–69 years and 70–84 years cannot be explained by mobile phone use. They think that more plausible causes for the increases observed are advances in diagnostics and improvements in the registration of gliomas in some of the countries considered over the last 20 years. These may have a particular impact on incidence rates in older persons.

Another reason for their conclusion is that in their simulations, almost all results from case-control studies that observed a positive association between mobile phone use and glioma risk turned out to be implausible. In this context, Deltour and Schüz assess risk scenarios that are not compatible with data from both age groups as implausible because there is no evidence from studies to date that a possible risk of glioma from mobile phone use could depend on age.

According to Deltour and Schüz, minor increases in risk and an increase in risk after long latency periods of several decades are compatible with the data. They see limitations to the study because of uncertainties regarding exposure and limited knowledge regarding the required modelling assumptions (e.g. actual latency).

Overall, the temporal development of glioma morbidity described in the study does not speak for an increase in the risk of glioma through the use of mobile phones. If mobile phone use did increase the risk of glioma, a considerable increase in the incidence of glioma would have been expected from the 1990s onwards because mobile phone use increased sharply from that time onwards. However, in the period 1990–2016, no point in time at which the increases would have changed noticeably was identified. Moreover, the increase was relatively similar for men and women, although there were differences in the frequency of use between the sexes, especially in the early years of mobile telephony. The increases observed in older persons are presumably attributable to the improved detection and more complete registration of glioma diseases.

The results of the modelling and simulations show that various scenarios with increased risks would be compatible with the data. However, these simulations are based on some critical assumptions that limit their validity. For example, because a lack of more recent usage figures, the data on mobile phone usage is based on projections from 2002 and 2009. However, it can be assumed that mobile phone use has changed considerably; this should be taken into account in further considerations and future studies.

Overall, the results of the present study do not provide robust evidence for an increase in incidence rates that could be associated with widespread mobile phone use. However, a possible increase in risk, especially with long latency periods or for certain subgroups, cannot be completely ruled out with the present study.

Because of its design, the study can make only a limited contribution to clarifying the question of whether there is a causal relationship between the use of mobile phones and the risk of developing glioma. This is due to the nature of the "ecological" study design; individual data on both disease and mobile phone use are missing. In particular, temporal changes as a result of other factors such as diagnostic accuracy or completeness of data collection may not be adequately taken into account and thus may lead to "spurious correlations". Cohort studies, in particular, are considerably more informative here but also considerably more costly.

Deltour I, Auvinen A, Feychting M, Johansen C, Klaeboe L, Sankila R, Schüz J. Mobile phone use and incidence of glioma in the Nordic countries 1979-2008: consistency check. Epidemiology. 2012 Mar;23(2):301-7. doi: 10.1097/EDE.0b013e3182448295. PMID: 22249239.