Elsevier

Annals of Epidemiology

Volume 19, Issue 1, January 2009, Pages 33-41.e1
Annals of Epidemiology

Quantifying the Impact of Selection Bias Caused by Nonparticipation in a Case–Control Study of Mobile Phone Use

https://doi.org/10.1016/j.annepidem.2008.10.006Get rights and content

Purpose

To quantitatively assess the impact of selection bias caused by nonparticipation in a multinational case–control study of mobile phone use and brain tumor.

Methods

Non-response questionnaires (NRQ) were completed by a sub-set of nonparticipants. Selection bias factors were calculated based on the prevalence of mobile phone use reported by nonparticipants with NRQ data, and on scenarios of hypothetical exposure prevalence for other nonparticipants.

Results

Regular mobile phone use was reported less frequently by controls and cases who completed the NRQ (controls, 56%; cases, 50%) than by those who completed the full interview (controls, 69%; cases, 66%). This relationship was consistent across study centers, sex, and age groups. Lower education and more recent start of mobile phone use were associated with refusal to participate. Bias factors varied between 0.87 and 0.92 in the most plausible scenarios.

Conclusions

Refusal to participate in brain tumor case–control studies seems to be related to less prevalent use of mobile phones, and this could result in a downward bias of around 10% in odds ratios for regular mobile phone use. The use of simple selection bias estimation methods in case–control studies can give important insights into the extent of any bias, even when nonparticipant information is incomplete.

Introduction

Declining levels of participation in case–control studies have led to increasing concerns about selection bias, more specifically nonparticipation bias 1, 2, 3, 4, 5. Studies evaluating the risk of brain and other tumors associated with the use of mobile phones are no exception. Participation proportions in published mobile phone case–control studies have ranged from 50% to over 90% for cases and 45% to 70% for controls 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20. It has been posited that odds ratios (OR) less than one for regular mobile phone use reported in some studies may be explained by the combination of low participation rates and unrepresentative respondents 9, 12, 14, 16, 17, 20. Fragmentary information from some sources indicates that people who refuse participation in these studies have different phone use patterns from those who do participate 7, 9, 13, 16, 21. Against this background, it is important to assess the likelihood of bias, both by evaluating participation rates and characteristics of participants and nonparticipants 2, 22, 23, 24, 25, 26, 27 and by quantitatively assessing the impact of nonparticipation on risk estimates 3, 28, 29, 30, 31.

We evaluate whether participation in a multinational case–control study of brain and other intracranial tumors (INTERPHONE) (19) was related to use of a mobile phone, and estimate the potential for selection bias.

Section snippets

Study Design

The methods of the INTERPHONE study have been published elsewhere (19). Briefly, eligible cases were between 30 and 59 years of age, diagnosed with a first primary glioma, meningioma, or acoustic neuroma, and resident in study regions of 16 centers in 13 participating countries (Australia, Canada, Denmark, Finland, France, Germany, Israel, Italy, Japan, New Zealand, Norway, Sweden, and the UK). Controls were selected randomly from the same source populations and matched to cases by age, sex,

Results

Participation proportions for this study have been published elsewhere (19). Participation by cases (overall = glioma, 64%; meningioma, 78%; acoustic neuroma, 82%) was higher than by controls (overall = 53%) in each study center. There was little difference in participation by age and sex apart from slightly lower participation by male controls and by older female glioma cases (19). The most common reasons for nonparticipation by controls were refusal (30%) and inability to be traced (13%) (

Discussion

The results of this study suggest that refusal to participate in a large, multi-national, case–control study of brain and other intracranial tumors is related to less frequent regular use of a mobile phone and that this relationship is consistent across cases and controls, study centers, sex, and age groups. Less education and more recent start of mobile phone use were also associated with refusal to participate. Several explanations should be considered, as well as the impact of these findings

References (36)

  • H.C. Christensen et al.

    Cellular telephone use and risk of acoustic neuroma

    Am J Epidemiol

    (2004)
  • S. Lonn et al.

    Mobile phone use and the risk of acoustic neuroma

    Epidemiology

    (2004)
  • H.C. Christensen et al.

    Cellular telephones and risk for brain tumors: a population-based, incident case-control study

    Neurology

    (2005)
  • S. Lonn et al.

    Long-term mobile phone use and brain tumor risk

    Am J Epidemiol

    (2005)
  • M.J. Schoemaker et al.

    Mobile phone use and risk of acoustic neuroma: results of the INTERPHONE case-control study in five North European countries

    Br J Cancer

    (2005)
  • S.J. Hepworth et al.

    Mobile phone use and risk of glioma in adults: case-control study

    Br Med J

    (2006)
  • S. Lonn et al.

    Mobile phone use and risk of parotid gland tumor

    Am J Epidemiol

    (2006)
  • J. Schuz et al.

    Cellular phones, cordless phones, and the risks of glioma and meningioma (INTERPHONE study group, Germany)

    Am J Epidemiol

    (2006)
  • Cited by (65)

    • An international prospective cohort study of mobile phone users and health (COSMOS): Factors affecting validity of self-reported mobile phone use

      2018, International Journal of Hygiene and Environmental Health
      Citation Excerpt :

      Overall, the balance of evidence does not suggest an excess risk, with studies on mobile phone use and cancer, primarily brain tumours, mostly reporting risk estimates close to unity (AGNIR, 2012; Ahlbom et al., 2009; Pettersson et al., 2014; Schoemaker et al., 2005; Swerdlow et al., 2011; Lahkola et al., 2006; Frei et al., 2011; Interphone Study Group, 2010), though some have reported increased risk of brain tumours among the heaviest mobile phone users when considering long-term (>10 years) use (Interphone Study Group, 2010; Coureau et al., 2014; Hardell et al., 2013; Hardell and Carlberg, 2015; The INTERPHONE Study Group, 2011). However, the majority of these cancer studies are limited by their reliance on subjective, self-reported measures of telephone use in the past (Schoemaker et al., 2005; Interphone Study Group, 2010; Coureau et al., 2014; Hardell et al., 2013; Hardell and Carlberg, 2015; The INTERPHONE Study Group, 2011; Takebayashi et al., 2006; Hardell et al., 2011; Mortazavi et al., 2007; Soderqvist et al., 2008) which are prone to substantial recall error (Parslow et al., 2003; Vrijheid et al., 2006; Vrijheid et al., 2009a; Berg et al., 2005; Samkange-Zeeb et al., 2004; Pettersson et al., 2015), and are case-control studies (Interphone Study Group, 2010; Hardell et al., 2013; Hardell and Carlberg, 2015; The INTERPHONE Study Group, 2011; Hardell et al., 2011) which are also prone to recall and selection bias (Mann, 2003; Schulz and Grimes, 2002). Evidence for potential effects of mobile phone use on other health outcomes (e.g. headaches, migraines, fatigue, cognition, sleep disturbance, dizziness, hearing loss, etc) is largely based on cross-sectional studies, with inconsistent results (AGNIR, 2012; Roosli and Hug, 2011; Frei et al., 2012; Seitz et al., 2005; Baliatsas et al., 2012; Baliatsas et al., 2015).

    • Investigation of bias related to differences between case and control interview dates in five INTERPHONE countries

      2016, Annals of Epidemiology
      Citation Excerpt :

      Possible sources of bias in INTERPHONE including selective nonparticipation and recall bias have been investigated. In an analysis of nonresponse questionnaires (people who did not participate were asked to answer a brief questionnaire), nonparticipation was associated with a lower prevalence and more recent start of cellular telephone use [11]. This was estimated to bias ORs downward by approximately 10% for all users.

    View all citing articles on Scopus
    View full text