The International Agency for Research on Cancer (IARC) of the World Health Organization (WHO) launched the INTERPHONE study in 2000. As the largest case-control study on the use of mobile phones and brain tumours, INTERPHONE collected and evaluated epidemiological data and information on the use of mobile phones from 13 countries (Australia, Canada, Denmark, Finland, France, Germany, Israel, Italy, Japan, New Zealand, Norway, Sweden and the UK).

The participants in the INTERPHONE study were between 30 and 59 years old. The study included 2,708 cases of glioma, 2,409 meningiomas and 1,105 acoustic neuromas in the period from 2000 to 2004, as well as an equal number of matched healthy comparison subjects (controls). In the study design of case-control studies, cases and corresponding controls are selected in such a way that they are comparable with regard to possible confounding factors such as age and gender in a procedure known as matching. The main analysis showed no statistically significant increase in risk for these brain tumours overall, but a subgroup analysis revealed an increased risk of glioma^[1] and acoustic neuroma^[2] in people with a very high total call duration.

At about the same time, a retrospective cohort study was set up in Denmark to address the same question. To this end, data from Danish mobile phone operators was linked with data from the Danish Cancer Registry. For the observation periods 1996-2001 and 2002-2007, no increase in the risk of brain tumours was observed in over 420,000 mobile phone users.^{[3, 4]}

In another cohort study from the United Kingdom (the Million Women Study), no increased risk of glioma or meningioma was observed based on self-reported mobile phone use^[5]. Although an increased risk of acoustic neuroma was observed in the first follow-up to this study^[5], this increase was no longer evident in a further study based on a longer follow-up^[6]. This was also confirmed in the most recent follow-up, whose results were published in 2022.^[7]

In contrast to the results of these cohort studies, evaluations of Swedish case-control studies showed increased risks of glioma and acoustic neuroma after only a few hours of cumulative mobile phone use and a short time after the first use of a mobile phone^{[8, 9, 10]}. Internationally, the results of the Swedish case-control studies on brain tumours have been cast in a critical light, particularly with regard to a number of methodological weaknesses that limit their evidential value^[11].

In 2011, the IARC categorised RF-EMF as “possibly carcinogenic” (group 2b), referring explicitly to the results of the case-control studies from Sweden^{[8, 9, 10]} and INTERPHONE^{[1, 2]}. This classification by the IARC means that such a risk is not probable but cannot be excluded because of individual indications.

For the COSMOS international prospective cohort study, participants aged 18 and older were recruited from Denmark, Finland, Sweden, the Netherlands and the United Kingdom. The recruitment strategy varied depending on the country – e.g. via lists from mobile network operators, electoral rolls or existing cohort studies.

All study participants completed a baseline questionnaire between 2007 and 2012. This included questions on mobile phone use, including when regular use (≥ 1 call per week) began, the average use per week (number of calls and duration), and the proportion of hands-free calls over a three-month period. Where participants had given their consent, data was also requested from mobile network operators regarding the number and duration of calls over the same period. In addition, participants were asked about their past mobile phone and hands-free device use from 1985 onwards (for periods of five years) in order to record changes in usage behaviour over time.

To obtain data on the occurrence of brain tumours in the study participants, the personal identifying data of cohort members was compared with information stored in the national cancer registries regarding actual cancer cases (“comparison with cancer registry data”). For the present evaluations, account was taken of the incidence of glioma, meningioma and acoustic neuroma.

Hazard ratios (HRs) and associated 95 % confidence intervals (CIs) were calculated in order to assess a possible association between glioma, meningioma and acoustic neuroma and the use of mobile phones.

An HR of 1 means that the risk of developing a brain tumour is the same in two groups of mobile phone users who differ in terms of their duration or intensity of use over the follow-up period, i.e. regardless of how many years the mobile phone is used for or how intensively.

An HR of > 1 or < 1, on the other hand, means that the risk of developing a brain tumour is higher or lower in the study group than in the reference group.

The HRs were calculated using Cox regression models and adjusted for country, gender, level of education, and marital status. The data from Denmark, Finland, the Netherlands and Sweden was analysed together. Due to legal restrictions preventing the sharing of data from the UK outside the country, data from the UK was analysed separately and then combined with the results from the other four countries by a meta-analysis.

The observation period (follow-up period) was defined as being from the date of the baseline survey plus six months until the first date of either a brain tumour diagnosis, death, emigration or the reference date of the cancer registry comparison (between the end of 2014 or 2017, depending on the participating country).

Risk analysis: cumulative usage time and cumulative call duration

The cumulative usage time of mobile phones in years was included in the risk analyses. This was calculated from the year in which regular use began to the time of the first baseline survey, and the times were divided into three groups: ≤9 years, 10 - 14 years and ≥ 15 years.

Deviating from this, the years since the start of regular use were divided into two groups, < 15 years and ≥ 15 years, for acoustic neuromas because the number of cases was lower than for gliomas, for example.

In addition, the analysis included an examination of the cumulative call time on a mobile phone up to the start of the study, both as a continuous and as a categorical variable.

Based on the distribution of the cohort, the call time categories were broken down into the following groups: below the median (reference category), 50th to < 75th percentile, and ≥ 75th percentile. Specifically for gliomas, an analysis was carried out with the 90th percentile in order to take account of particularly intensive use. Insufficient case numbers meant it was not possible to make detailed analyses for tumour types other than gliomas.

Reduction of measurement errors

To reduce measurement errors in the estimation of cumulative call time, various approaches were tested within the framework of COSMOS. This testing was performed in simulation studies based on self-reported data by study participants in combination with objective information from mobile network operators, and the estimation was carried out using the simple regression calibration method^[18]. Information on the use of hands-free devices when making phone calls was taken into account and deducted from the relevant usage time for the analyses. The use of mobile phones for other purposes such as text messaging was not included in the exposure assessment. According to the authors, exposure of the head during this type of mobile phone use is negligible. Missing values for mobile phone usage in a particular year were replaced (imputed) based on the next available years.

COSMOS comprises 264,574 study participants and 1.8 million person-years in the first follow-up, with a mean observation period of 7.12 years. Of all study participants, 30.5 % started using mobile phones at least 15 years before the baseline survey. At 64 %, there were more women than men in the cohort, with men tending to use mobile phones more frequently. During the observation period, 149 gliomas, 89 meningiomas and 29 acoustic neuromas occurred.

No statistical association between mobile phone use and brain tumour development

No statistically significant association (i.e. a significant increase or decrease) was found between the incidence of brain tumours and the intensity of mobile phone use.

• The HR per 100 hours of cumulative call time was:
• 1.00 (95 % CI 0.98 - 1.02) for gliomas
• 1.01 (95 % CI 0.96 - 1.06) for meningiomas and
• 1.02 (95 % CI 0.99 - 1.06) for acoustic neuromas.

The categorical analysis also showed no statistically significant association for all three endpoints when comparing participants with a high cumulative call time with people with a low cumulative call time (for gliomas, for example, ≥ 1,062 hours vs. < 464 hours). Analyses using country-specific exposure categories for cumulative call time did not yield different results.

The analyses of the cumulative duration of mobile phone use and brain tumours also showed no statistically significant increase in risk estimates.

The HR after ≥15 years of use compared to ≤9 years of use since the start of regular mobile phone use is 0.97 (95 % CI 0.62 - 1.52) for gliomas and 1.24 (95 % CI 0.60 - 2.59) for meningiomas. For acoustic neuromas, the HR for ≥ 15 years since first use was 0.76 (95 % CI 0.33 - 1.73), based on 10 cases in the high exposure category compared to 18 cases in the reference category (<15 years of use).

Strengths

The COSMOS study has a number of strengths. Firstly, with over 260,000 study participants and over 1.8 million person-years, COSMOS is a large cohort study with a long follow-up. This enables a robust analysis of a potential connection between the use of mobile phones and the occurrence of brain tumors. The large proportion of people with a long period of use of ≥ 15 years (30.5 %) and people with intensive use (≥ 1,062 hours of cumulative call time with the phone held to the head, 25 %) enable reliable analyses, especially for these groups of participants. This was only possible to a limited extent in previous studies, which meant that there were scientific uncertainties in the evaluation of long-term and intensive mobile phone use. The new findings of the COSMOS study contribute significantly to reducing these uncertainties, especially for gliomas with evaluations based on 149 cases. The number of cases for meningiomas and acoustic neuromas was lower, which somewhat limits the significance of these results in comparison.

Another strength of COSMOS is the prospective design of the study. The exposure was determined well before the onset of the disease. Compared to retrospective case-control studies, COSMOS is therefore less susceptible to recall bias, which can systematically lead to a misclassification of the exposure of cases compared to controls. This in turn can influence the effect estimate and thus considerably limit the strength of the results. The extensive questionnaire-based data collection on the duration since the first regular use of mobile phones, the duration of use and the use of hands-free devices and the validation by combining it with the objective information provided by the mobile phone operators also contributes to greater reliability of the exposure data collected. By asking about usage behavior in the years before the baseline survey, changes in usage behavior over the years were also taken into account, which can be viewed as positive.

In addition, regression calibration methods were able to reduce the risk of non-differential misclassification of exposure status. Nevertheless, the authors state that a certain degree of non-differential misclassification cannot be ruled out when using self-reported data.

Another positive aspect is the transparency and traceability of the evaluation through the publication of the entire study analysis plan in the appendix of the publication and the extensive consideration of relevant confounders in the risk analyses, such as the socioeconomic status of the participants.

Limitations

The limitations of this study include the previously mentioned limited statistical power due to the small number of cases of meningiomas and acoustic neuromas compared to gliomas. This leads to uncertainty for these partial results.

There is also a risk of non-differential misclassification of exposure status through self-reported data on mobile phone use. Despite extensive efforts by the authors to correct this, there remain further limitations in exposure assessment, which are critically discussed by the authors themselves. Firstly, the assessment of cumulative call duration is limited to usage behavior up to the time of the baseline survey. Exposure after that is currently not assessed, but will be included in future follow-ups. Nevertheless, it can be assumed that due to the long latency period for brain tumors, more recent exposures could have a smaller impact.

Uncertainties in the exposure estimate

In addition, the exposure measure of the cumulative call duration is only a fairly rough measure of the actual RF-EMF exposure in the head, because this depends on many other influencing factors (e.g. phone type, mobile phone standard used, how the phone is held on the head, reception conditions).

While a comparatively high exposure of the head is to be expected with many 2G phone calls, the transmission power of mobile phones that handle calls using newer mobile phone standards is in many cases significantly lower. This means that phone calls with more recent or current mobile phone generations are associated with a lower average head exposure than was the case at the time of recruitment for the COSMOS study.

On the other hand, exposure variations cannot be represented by the approach used. The variations in energy absorbed in the brain from RF-EMF exposure during mobile phone calls, which are influenced by the connection quality between the phone and the base station, pose a challenge that is difficult to measure in large epidemiological studies and is not necessarily reflected by the cumulative call time.