The study was approved by the Danish ethical committee system (KF 01–075/96), the Danish Data Protection Board (1996–1200–121, 2009–41–3886), and the Danish Ministry of Justice (Jnr. 1996–760–0219). In accordance with Danish law informed consent was not obtained as the study was entirely register-based and did not involve biological samples from, or contact with study participants.
We conducted this study within the population of Denmark during the period 1987–2004. All Danish citizens are assigned a unique personal identification number at birth by the Danish Central Population Registry (CPR), which keeps complete information about the date of birth, gender, vital status, migration and current and former addresses of all Danes 
. This identification number is applied universally in all contacts within the health care systems in Denmark allowing individuals to be tracked over time in and across all Danish administrative and health registers.
Ascertainment of MS Cases
MS cases were identified from the Danish Multiple Sclerosis Registry 
established in 1956. The register contains information on more than 90% of all MS patients in Denmark since 1949 and is considered to have a validity of 94%. For each patient the medical records have been evaluated and the year of the first symptom has been assessed. Furthermore, the first symptom(s) for each patient in the register are recorded, and multiple simultaneous symptoms are allowed. For the present paper, only definite and probable diagnoses according to the Poser criteria 
Records of all (723 421) mobile phone subscriptions in Denmark during the period 1982 (when this service was established) until the end of 1995 were obtained from the Danish network operatoros. Details of the cleaning process of these data has been reported previously 
: Briefly, 200 507 corporate subscriptions were deleted and 102 828 records were deleted for other reasons including errors in matching variables and duplicate records (persons with multiple subscriptions) leaving a cohort of 420 086 private mobile phone subscription holders. Since handheld mobile phones first became available in Denmark in 1987, all subscription periods were left truncated to 1 January 1987, further deleting four persons from the dataset. The unexposed population was obtained by subtracting the number of MS-cases and exposed persons from the Danish population count by age and gender for each year of the study. Only persons, who turned 18 years before January 1st
1996, i.e., end of exposure period, were included in the analysis.
Risk of MS Among Subscription Holders
In analyses of risk of MS and presenting symptoms among mobile phone subscription holders, symptom free subjects entered the study population on 1 January 1987 or age 18 years, whichever occurred latest. Follow-up ended at date of MS diagnosis, age 65 years, death, emigration from Denmark or 31 December 2004, whichever came first. Exposed person time was further categorized based on duration of follow-up since date of first subscription (<1, 1–3, 4–6, 7–9, 10–12, 13+ years). In a subanalysis, we used July 1st in the year of the first recorded symptom(s) as endpoint. Due to the retrospective nature of debut data follow-up was, for this analysis, terminated on 31 December 2000 allowing four years until 2004 to identify patients with first symptoms in 2000 or earlier. In these analyses, we excluded 13 573 subscription holders who obtained their subscription after age 65 years, 30 with MS symptoms before age 18 years, 366 with symptoms before 1987 and 142 diagnosed with MS before getting a subscription, yielding a population of 405 971 mobile phone subscription holders.
Risk of Death Among MS-patients
In analyses of risk of death among MS patients using mobile phone, MS patients diagnosed between age 18 and 65 years in 1980 or later entered the study population at date of diagnosis or 1 January 1987, whichever came latest. Follow-up ended at date of death, emigration from Denmark or on 31 December 2004, whichever came first. A total of 7420 MS-patients met the inclusion criteria, of whom 717 were subscription holders. For these patients, exposed person-years were cumulated from date of diagnosis or date of subscribing, whichever came last, and subdivided into five categories (<1, 1–3, 4–6, 7–9, 10+ years).
Likelihood of Subscribing for a Mobile Phone Among MS Patients
To evaluate a potentially reverse association, we analysed MS as an explanatory factor for obtaining a subscription in the 5 050 MS patients with a first symptom between age 18 and 65 years in the period 1980 to 1995. Entry and exit criteria were as for the main analysis, except that subjects were censored at the date of subscription acquisition, or on 31 December 1995, and not at the time of the MS-diagnosis. Time after first MS-symptom was subdivided according to diagnostic status (1st symptom, diagnosis), and years since first symptom and since diagnosis (<1, 1–3, 4–6, 7–9, 10+ years).
Log-linear Poisson regression analysis was used to compute incidence and mortality rate ratios (IRRs or MRRs) for MS diagnosis, MS debut and risk of death in MS patients among mobile phone subscription holders compared to non-subscribers. The analyses were adjusted for gender, age (in incidence analyses: 18–29, 30–39, 40–49 and 50–65 years; in mortality analyses:18–29, 30–39, 40–49, 50–59, 60–69, 70–79 and 80+ years) and individual calendar year (1987 to 2004 by increments of 1 year). Subjects were allowed to change between categories of covariates and exposure variables over time. When analysing risk of death, years since diagnosis was included as a linear covariate.
The presenting symptoms of MS among subscribers and nonsubscribers were compared by analysing MS diagnoses with different initial symptoms as competing risks in a Poisson model as above with independent gender, age and period dependency for each symptom.
For date variables with missing day values, the 15th of the respective month was used and when the month was missing, July 1st was used. As only the year of first symptom was available, the date of first symptom was set to July 1st of the given year or the actual date of diagnosis, whichever came first. The statistical analyses were performed in SAS 9.1.