Mobile Phone Use and Risk of Brain Neoplasms and Other Cancers

Prospective Study

Victoria S Benson; Kirstin Pirie; Joachim Schüz; Gillian K Reeves; Valerie Beral; Jane Green


Int J Epidemiol. 2013;42(3):792-802. 

In This Article

Materials and Methods

Study Design, Data Collection and Follow-up

During the period 1996–2001, 1.3 million middle-aged women were recruited through the UK National Health Service (NHS) Breast Screening Programme into the Million Women Study (see Supplementary data at IJE online), completing a postal questionnaire about sociodemographic, medical and lifestyle factors. The study population is resurveyed approximately every 3–4 years. Full details of the study design and methods are described elsewhere[6,7] and all questionnaires can be viewed at Questions on mobile phone use were asked in 1999–2005, and again in 2009.

All study participants have a unique NHS number, and are followed via record linkage (using this number and other personal details) to the NHS Central Register. Cancer registrations (including non-invasive tumours of the CNS, and those of uncertain behaviour) and deaths are routinely notified to the study investigators; this information includes the date of each such event, with tumour site and morphology coded using the 10th revision of the International Classification of Diseases (ICD-10),[8] and the third edition of the International Classification of diseases for Oncology (ICD-O).[9]

Information on incident vascular disease during follow-up was obtained through linkage to Hospital Episodes Statistics (HES) in England and to Scottish Morbidity Records (SMR); these agencies provided dates and ICD-10 diagnosis codes for inpatient and day-patient hospital admissions.

All study participants gave written consent to taking part in the study, and ethical approval was provided by the Oxford and Anglia Multi-Centre Research Ethics Committee. Access to hospital admissions data was approved by the Information Centre for Health and Social Care (England) and the Information Services Division (Scotland).

Exposure Variables

Women in the study have been asked twice about mobile phone use. In a survey conducted between 1999 and 2005 (to which about 65% of women recruited in 1996–2001 replied), women were asked: 'About how often do you use a mobile phone?', and given three options to respond: 'never', 'less than once a day', 'every day'; and 'For how long have you used one?' (participants were asked to provide total years of use). The responses to these questions provided baseline exposure data for analyses. In 2009, study participants were asked 'How much do you talk on a mobile phone?' (average minutes per week) and 'How long have you used a mobile phone?' (in years). This information is currently available for a random sample of 31 110 women who had also responded to the questions on mobile phone use at baseline, and although it was not used to define exposure status for analyses, it allowed assessment of the repeatability of use of mobile phones reported earlier. Of the women who reported at baseline that they had used a mobile phone, 77% of those reporting ever use (13 437/17 647) and 92% of those reporting daily use (1702/1852) also reported use for at least 1 minute per week at follow up, an average of 8.8 years later. In women reporting use at both surveys, duration of use reported at the later survey was consistent, on average, with that estimated to have accrued by that time on the basis of duration reported at baseline, assuming continued use between surveys. Approximately half (49%) of those who reported no phone use at baseline reported using a mobile phone in 2009.


The main outcomes examined here are registered cancers or non-invasive tumours occurring after the date that the baseline questionnaire was completed. Results are reported for incident intracranial tumours of the CNS: ICD-10 C70, C71, C72.1–5, C75.1–3, D32.0, D33.0–3, D35.2–4, D42.0, D43.0–3 and D44.3–5; and where possible, CNS tumours were further classed by site and morphology as glioma (ICD-O 9380–9481), meningioma (ICD-O 9530–9539), pituitary tumours (C75.1, D35.2, and D44.3) and acoustic neuroma (D33.3, ICD-O 9560). Results are also reported for all invasive cancer (C00-C97, excluding non-melanoma skin cancer C44), and separately for 18 invasive cancer sites, 16 of which had accrued over 500 incident cases during follow-up and 2 others (eye and thyroid) for comparison with reports by others. The 18 cancer sites were defined as follows: 'other head and neck' (ICD-10 C00–14, C30–32, i.e. excluding CNS, eye and thyroid), oesophagus (C15), stomach (C16), colon (C18), rectum (C19–20), pancreas (C25), lung (C34), melanoma (C43), breast (C50), endometrium (C54), ovary (C56), kidney (C64), bladder (C67), eye (C69), thyroid (C73), non-Hodgkin lymphoma (C82–85), myeloma (C90) and leukaemia (C91–95).

Incident vascular disease endpoints were defined as first hospital admission with a primary diagnosis of stroke (ICD-10 I60–69) or ischaemic heart disease (ICD-10 I20–25).

Statistical Analyses

Potentially eligible for these analyses were 866 525 women who responded to the study survey conducted between 1999 and 2005. Of these, 14 387 were excluded because they completed a version of the survey which did not include the question on mobile phone use, and 11 981 because they did not answer the question asked on mobile phone use. Analyses also excluded 48 531 women with a CNS tumour or any other invasive cancer [other than non-melanoma skin cancer (C44)] registered before baseline, and 6 women who reported having the inherited disorder neurofibromatosis (Q85.0) (which is associated with a high risk of neurological tumours).

Analyses for vascular disease additionally excluded women with a history of vascular disease (diagnosis of and/or treatment for heart disease or for stroke before baseline, either self-reported or identified from hospital admission data).

Cox regression models (taking attained age as the underlying time variable) were used to obtain adjusted relative risks (RRs) and 95% confidence intervals (CIs) for each of the endpoints of interest in relation to mobile phone use. Duration of use of a mobile phone was treated as a time-dependent variable, incrementing duration for each year of follow-up.

Eligible women contributed woman-years from the date they answered the baseline questions about mobile phone use until the date of diagnosis with the tumour or disease of interest, date of death or the end of follow-up, whichever was earliest. In analyses of cancer and CNS tumour outcomes, censoring was done at first cancer or CNS tumour diagnosis at any site: for stroke and ischaemic heart disease censoring was done at first diagnosis of either condition. The last date of follow-up for analyses of tumour incidence was 31 December 2009 for all 10 regions (corresponding to 10 cancer registries), except for the North West (Mersey) region and Scotland, where it was 31 December 2008 (as registrations were incomplete after that date). For vascular disease incidence, hospital admissions data were available in England until 31 March 2008 and in Scotland until 31 December 2008, with follow-up ending on these dates.

All analyses were stratified by quintiles of socioeconomic status (based on the Townsend deprivation index[10]), geographical region of residence (10 regions corresponding to the areas covered by the cancer registries) and age at baseline (<53, 53–55, 56–58, … , 78–80, 80+ years). Analyses were additionally adjusted for height (<160, 160–164.9, ≥165 cm), body mass index (<25, 25–29.9, ≥30 kg/m2), smoking (never, past, current 1–14 cigarettes per day, current ≥15 cigarettes per day), alcohol intake (none, <10, ≥10 g per day), duration of strenuous exercise (<0.5 h, 0.5–1 h, ≥1 h per week) and use of menopausal hormone therapy (never, past, current). For each adjustment and stratification variable, missing values were assigned to a separate category. For analyses of CNS tumours, sensitivity analyses were carried out excluding the first 3 years of follow-up (because pre-clinical disease may affect reporting of mobile phone use, or may cause women to change their mobile phone use) and, separately, excluding women who completed the baseline questionnaire in 1999 or 2000, because the prevalence of use of mobile phones increased rapidly over the next few years; non-users who completed the baseline questionnaire in 1999–2000 were more likely than non-users reporting in 2001–05 to have started to use a mobile phone over the follow-up period to 2009.

Where summary estimates are given combining our results with those from the Danish prospective study of mobile phone use,[4,5,11] study-specific results were combined using the method of inverse variance least squares.

National trends in the incidence of acoustic neuroma (ICD-10 code D33.3) in England were examined for the years from 1998 to 2008 by calculating annual age-standardized incidence rates per 100 000 men and women aged 20–79 years, using data on tumour incidence and population estimates from the Office for National Statistics.[12] All analyses were performed using Stata version 12.0.