Regression calibration of self-reported mobile phone use to optimize quantitative risk estimation in

Abstract

Regression calibration of self-reported mobile phone use to optimize quantitative risk estimation in the COSMOS study Reedijk M, Portengen L, Auvinen A, Kojo K, Heinävaara S, Feychting M, Tettamanti G, Hillert L, Elliott P, Toledano MB, Smith RB, Heller J, Schüz J, Deltour I, Poulsen AH, Johansen C, Verheij R, Peeters P, Rookus M, Traini E, Huss A, Kromhout H, Vermeulen R, Study Group TC. Regression calibration of self-reported mobile phone use to optimize quantitative risk estimation in the COSMOS study. Am J Epidemiol. 2024 May 13:kwae039. doi: 10.1093/aje/kwae039. Abstract The Cohort Study of Mobile Phone Use and Health (COSMOS) has repeatedly collected self-reported and operator-recorded data on mobile phone use. Assessing health effects using self-reported information is prone to measurement error, but operator data were available prospectively for only part of the study population and did not cover past mobile phone use. To optimize the available data and reduce bias, we evaluated different statistical approaches for constructing mobile phone exposure histories within COSMOS. We evaluated and compared the performance of four regression calibration (RC) methods (simple, direct, inverse, and generalized additive model for location, shape, and scale), complete-case (CC) analysis and multiple imputation (MI) in a simulation study with a binary health outcome. We used self-reported and operator-recorded mobile phone call data collected at baseline (2007-2012) from participants in Denmark, Finland, the Netherlands, Sweden, and the UK. Parameter estimates obtained using simple, direct, and inverse RC methods were associated with less bias and lower mean squared error than those obtained with CC analysis or MI. We showed that RC methods resulted in more accurate estimation of the relation between mobile phone use and health outcomes, by combining self- reported data with objective operator-recorded data available for a subset of participants. Excerpts Self-reported duration of mobile phone use (REPORT) at baseline [21] was based on answers to the question: “Over the last three months, on average, how much time per week did you spend talking on a mobile phone?”. The following response options were provided: “< 5 min/week”, “5-29 min/week”, “30- 59 min/week”, “1-3 hours/week”, “4-6 hours/week”, and “>6 hours/week”. In the Netherlands and the UK two further categories of call duration were included, “7-9 hours/week” and “10 or more hours/week”, but these were combined into “>6 hours/week” for the present analyses.... Operator-recorded duration of mobile phone use (RECORD) was collected for all participants who provided consent and had a subscription under their own name. Operator-recorded data were used only when available for all mobile phones that were reported (up to two phones in Denmark, the Netherlands, Sweden, Finland, and up to three in the UK) and for at least three full months at the time the baseline questionnaire was administered. Participants with mobile phones that were also used by others were excluded. Operator-recorded data were available for 21% of participants in Denmark, 76% in Finland, 40% in the Netherlands, 63% in Sweden, and 76% in the UK.... A major issue is how well mobile phone use predicts the exposure of interest, namely radiofrequency electromagnetic fields. While past validation studies have been carried out for the 2nd mobile phone generation, showing fair agreement between amount of use and cumulated emission from the handset [4], fewer data are available on the predictive power of mobile phones of the 3rd, 4th and 5th generations that have and are being used by COSMOS participants.... Table 1 (excerpt) Complete data by country: Denmark (11%: 2993/25912), Finland (70%: 9162/13062), Netherlands (3%: 3039/88466), Sweden (49%: 24881/50678), United Kingdom (58%: 56862/98685) Recorded call duration in minutes/week by country (geometric mean): Denmark (60.6), Finland (81.5), Netherlands (23.4), Sweden (78.3), United Kingdom (46.5) Conclusion This study addressed an important concern in mobile phone research and more generally in environmental epidemiology: how to leverage self-reported exposure estimates that are often available but error-prone, with more objective measurements that may be obtained in only a subset of participants. Our simulation study indicated RC approaches may improve estimation of exposure- outcome relations between mobile phone use and health outcomes within COSMOS. The prospective design and improved exposure assessment within COSMOS compared to that in previous case-control studies are expected to lead to more robust conclusions about possible health effects from use of mobile phones. Conflict of Interest: MF was vice chairman (2012-2020) of the International Commission on Non-Ionizing Radiation Protection, an independent body setting guidelines for non-ionizing radiation protection. She has served as advisor to a number of national and international public advisory and research steering groups concerning the potential health effects of exposure to non-ionizing radiation, currently for the World Health Organization (WHO). HK was the chair of the Committee on Electromagnetic Fields of the Health Council of The Netherlands till 2022. He currently is a member of the WHO Task Group for the Environmental Health Criteria Monograph on RF-EMF. AH is a member of the International Commission on Non-Ionizing Radiation Protection since 2020, and of the Committee on Electromagnetic Fields of the Health Council of The Netherlands, and chairs the Swedish Radiation Safety Authority’s (SSM) Scientific Council on Electromagnetic Fields since 2020. AA currently is a member of the WHO Task Group for the Environmental Health Criteria Monograph on RF-EMF. MBT is currently member of the WHO groups tasked with systematic review of evidence on non-ionizing radiation and health, that is feeding into the Environmental Health Criteria Monograph on RF-EMF. All other authors declare they have no competing financial interests. Open access paper: academic.oup.com

AI evidence extraction

Main findings

In simulations using baseline self-reported and operator-recorded call duration data, simple, direct, and inverse regression calibration methods produced parameter estimates with less bias and lower mean squared error than complete-case analysis or multiple imputation. The authors conclude that regression calibration can improve estimation of relations between mobile phone use and health outcomes by combining self-reported with operator-recorded data available for a subset.

Outcomes measured

• Performance of exposure assessment/statistical methods (bias, mean squared error) for estimating exposure–outcome relations in simulations with a binary health outcome

Limitations

• Operator-recorded data were available only for a subset of participants and required consent and subscription under participant’s own name; participants with phones used by others were excluded.
• Operator-recorded data were used only when available for all reported phones and for at least three full months at baseline.
• The evaluation included a simulation study with a binary health outcome rather than reporting specific health-effect results.
• The abstract notes uncertainty about how well mobile phone use predicts the exposure of interest (RF-EMF), especially for 3rd–5th generation phones.

Suggested hubs

• who-icnirp (0.6)
  Conflict-of-interest statement notes author roles with ICNIRP and WHO RF-EMF task groups.

{
    "study_type": "other",
    "exposure": {
        "band": "RF",
        "source": "mobile phone",
        "frequency_mhz": null,
        "sar_wkg": null,
        "duration": "Self-reported weekly call duration over last 3 months at baseline (2007–2012); operator-recorded call duration available for ≥3 full months at baseline for a subset"
    },
    "population": "COSMOS cohort participants from Denmark, Finland, the Netherlands, Sweden, and the UK (baseline 2007–2012)",
    "sample_size": null,
    "outcomes": [
        "Performance of exposure assessment/statistical methods (bias, mean squared error) for estimating exposure–outcome relations in simulations with a binary health outcome"
    ],
    "main_findings": "In simulations using baseline self-reported and operator-recorded call duration data, simple, direct, and inverse regression calibration methods produced parameter estimates with less bias and lower mean squared error than complete-case analysis or multiple imputation. The authors conclude that regression calibration can improve estimation of relations between mobile phone use and health outcomes by combining self-reported with operator-recorded data available for a subset.",
    "effect_direction": "unclear",
    "limitations": [
        "Operator-recorded data were available only for a subset of participants and required consent and subscription under participant’s own name; participants with phones used by others were excluded.",
        "Operator-recorded data were used only when available for all reported phones and for at least three full months at baseline.",
        "The evaluation included a simulation study with a binary health outcome rather than reporting specific health-effect results.",
        "The abstract notes uncertainty about how well mobile phone use predicts the exposure of interest (RF-EMF), especially for 3rd–5th generation phones."
    ],
    "evidence_strength": "moderate",
    "confidence": 0.7399999999999999911182158029987476766109466552734375,
    "peer_reviewed_likely": "yes",
    "keywords": [
        "COSMOS",
        "mobile phone use",
        "self-report",
        "operator-recorded data",
        "regression calibration",
        "measurement error",
        "exposure assessment",
        "simulation study",
        "RF-EMF"
    ],
    "suggested_hubs": [
        {
            "slug": "who-icnirp",
            "weight": 0.59999999999999997779553950749686919152736663818359375,
            "reason": "Conflict-of-interest statement notes author roles with ICNIRP and WHO RF-EMF task groups."
        }
    ]
}

Comments

Log in to comment.

No comments yet.