Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Recall bias in the assessment of exposure to mobile phones

Abstract

Most studies of mobile phone use are case–control studies that rely on participants’ reports of past phone use for their exposure assessment. Differential errors in recalled phone use are a major concern in such studies. INTERPHONE, a multinational case–control study of brain tumour risk and mobile phone use, included validation studies to quantify such errors and evaluate the potential for recall bias. Mobile phone records of 212 cases and 296 controls were collected from network operators in three INTERPHONE countries over an average of 2 years, and compared with mobile phone use reported at interview. The ratio of reported to recorded phone use was analysed as measure of agreement. Mean ratios were virtually the same for cases and controls: both underestimated number of calls by a factor of 0.81 and overestimated call duration by a factor of 1.4. For cases, but not controls, ratios increased with increasing time before the interview; however, these trends were based on few subjects with long-term data. Ratios increased by level of use. Random recall errors were large. In conclusion, there was little evidence for differential recall errors overall or in recent time periods. However, apparent overestimation by cases in more distant time periods could cause positive bias in estimates of disease risk associated with mobile phone use.

Introduction

Concerns about the possible adverse health effects of radio frequency (RF) fields used in mobile telephony have prompted a number of epidemiological studies, primarily focusing on central nervous system tumours (Muscat et al., 2000, 2002; Inskip et al., 2001; Johansen et al., 2001; Auvinen et al., 2002; Warren et al., 2003; Christensen et al., 2004, 2005; Lonn et al., 2004, 2005, 2006; Hardell et al., 2005; Schoemaker et al., 2005; Hepworth et al., 2006; Takebayashi et al., 2006; Schuz et al., 2006a, 2006b; Hours et al., 2007a; Klaeboe et al., 2007; Lahkola et al., 2007; Schlehofer et al., 2007). Most are case–control studies that rely on participants’ reports of past mobile phone use as surrogate for RF exposure assessment. Error in recalled phone use is a major concern in these studies, particularly if it differs between cases and controls (Auvinen et al., 2006; Inyang et al., 2008).

Differential errors in recall of exposure can attenuate, strengthen or even invert a true association, or induce a spurious association (Drews and Greenland, 1990; Armstrong et al., 1992; White, 2003). Cases may overreport their exposure, perhaps seeking an explanation for their disease (Drews and Greenland, 1990; Infante-Rivard and Jacques, 2000). In brain tumour studies, cases may also recall exposure with greater random error, as the tumour may affect their memory. Previous validation studies have shown substantial error in healthy volunteers’ short-term recall of phone use (Parslow et al., 2003; Samkange-Zeeb et al., 2004; Berg et al., 2005; Morrissey, 2007; Vrijheid et al., 2006a; Hours et al., 2007b). However, there is as yet no direct evidence of differential recall error in studies of mobile phone use and brain tumours.

INTERPHONE is a large multinational case–control study of brain tumour risk and mobile phone use (Cardis and Kilkenny, 1999; Cardis et al., 2007). Here, we present results of an INTERPHONE validation study, which aims to quantify recall error among cases and controls by comparing traffic and/or billing records collected from network operators with self-reports of mobile phone use.

Methods

INTERPHONE study methods have been published elsewhere (Cardis et al., 2007). Briefly, cases were between 30 and 59 years of age, diagnosed with a first primary glioma, meningioma, acoustic neuroma, or parotid gland tumour, and resident in specified regions of 13 participating countries. Controls were randomly selected from the same source populations and matched to cases by age, sex, and region.

Eligibility Criteria for the Validation Study

In six INTERPHONE study centres in three countries (Australia — Melbourne and Sydney; Canada — Montreal, Ottawa, and Vancouver; and Italy — Rome) it was feasible to collect traffic or billing records from network operators for subjects who consented (following approval by Local Ethics Review Boards). Subjects were eligible for inclusion if they were regular mobile phone users at the time of interview and holder of the subscription. Details of data collection protocols are summarised in Table 1. In Canada and Italy, subjects were included only if they did not share their phone with others for more than 25% of total call time. In Italy and Australia, proxy respondents were not included.

Table 1 Details of study protocols for collection of mobile phone use data from network operators.

Interview Data

Trained interviewers interviewed all subjects using the INTERPHONE computer-assisted personal interview (Cardis et al., 2007). The mobile phone section included an “event history calendar”, structured according to factors that may change patterns of phone use, and a comprehensive catalogue of mobile phone pictures to aid in phone identification. The amount of phone use was reported for each mobile phone used; if use changed within the use period of one phone, multiple periods were reported. For each period, questions were asked concerning average duration and number of calls made or received and the network operator. Average call time was reported by call or by day, week, or month. Interviewers rated the apparent quality of respondents’ memory.

Operator Data

Mobile phone network operators were given consent forms and identifying details (name and subscription phone numbers) for consenting subjects; they then provided records for all traceable subscriptions. All major operators in the three countries participated, although in Canada one of the largest operators could not provide records for an appreciable proportion (40%) of subjects. Australian operators were able to provide data on outgoing calls only. All records included the date, time, and duration of each voice call. Records were collected from the date they became accessible (Table 1) or the start date of the subscription, whichever was the latest, until the date of interview or the stop date of the subscription, whichever was the earliest. In Italy, data were collected up to diagnosis date of cases and corresponding reference date of controls for 50% of subjects. Operator data were often incomplete, as many subscriptions, and particularly earlier parts of subscriptions, could not be traced due to difficulties in linking phone numbers and problems in accessing archived records.

Matching Data Periods

Periods of reported phone use were matched to operator records for each subject. Because of the incompleteness of operator records, periods were only included in analyses if data from both the subject and the operator were available, and the operator reported by the subject matched the operator that provided data for the period in question.

For each matching period of mobile phone use, two phone use indicators were calculated: the average number of calls per month and the average duration (in minutes) of calls per month. For many subjects, the data contained multiple matching periods, sometimes with gaps between periods for which no matching data could be found. Matching periods of less than 1 week were excluded, as were subjects with less than 31 days of matching data for all periods combined (6 subjects). Operator data in Australia were multiplied by 2 to correct for the missing incoming calls in the operator data, as interview data were only available for incoming and outgoing calls combined.

Statistical Analysis

Analyses compared self-reported and operator-recorded number and duration of calls. Analyses were performed for phone use in all matching periods accumulated up to the date of the interview and for all use accumulated up to 1 year before the date of interview (the INTERPHONE study analysis excludes phone use less than 1 year before the reference date, defined as date of diagnosis of the case in each matched set). To evaluate whether recall deteriorated with increasing time before interview, data were analysed within periods of ≤1 year, 1–2 years, 2–3 years, 3–4 years, and >4 years before the date of interview.

Agreement between categories of number and duration of calls was tested using the weighted κ-statistic (Landis and Koch, 1977). Quintiles of the operator-recorded phone use distribution were used to categorise the data.

The level of agreement between self-reported and operator-recorded phone use on a continuous scale was measured by the ratio of reported to recorded phone use. The mean of this ratio represents the average level of over- (ratio>1) or under (ratio<1) estimation, and its standard deviation provides a measure of the variation between individuals and, therefore, of the random error in recall. All analyses used log-transformed data as phone use variables have highly skewed distributions. For presentation of results, the mean and 95% confidence bounds of the log-transformed data were exponentiated to the arithmetic scale. Multivariate linear regression was used to test for differences in the log-ratio between cases and controls, adjusting for the effects of country, age (<40 years, 40–49 years, ≥50 years), sex, and time period before interview (see above). Regression analyses were further conducted including explanatory variables from the interview: reported lifetime cumulative number of calls and call time (in deciles), reported years-since-start-of-use of a mobile phone, reported number of phones used, the option used for the reporting of call time (per call or per day/week/month), education (four categories), and quality of memory as rated by the interviewer. Reported lifetime cumulative number of calls and call time, and years-since-start-of-use are the main exposure variables in the full INTERPHONE analyses and the quantification of recall errors in categories of these variables is therefore of particular interest. Analyses were repeated excluding Australia, because of its lack of operator data for incoming calls. In all regression analyses, we included random effects accounting for multiple, correlated periods within one subject (Stata command XTREG; StataCorp., 2005).

We used the graphical method of Bland and Altman (1995, 1999) to assess the relationship between level of use and level of error. Following this method, the ratio of the two measures (reported to recorded phone use) was plotted against their average. In all, 95% limits of agreement were calculated as the mean ratio±2 SD.

Results

Matching self-reported and operator-recorded data were available for 212 cases and 296 controls (Table 2). Overall, 50% of controls and 54% of cases were men; 27% of controls and 26% of cases were <40 years old, and 40% of controls and 36% of cases were ≥50 years. These subjects represented 27% of controls (23 in Australia, 21 in Canada, and 43 in Italy) and 23% of cases (15 in Australia, 26 in Canada, and 40 in Italy) included as regular mobile phone users 1 year before the reference date in the full INTERPHONE analyses population in these countries (Table 2). They did not differ greatly from the full analysis population by age, sex, time since start of phone use, or lifetime cumulative phone use (not shown). On average, the matching data periods were longest in Australia and shortest in Canada; they were somewhat longer for controls than cases in all three countries.

Table 2 Characteristics of cases and controls included in analyses of the validation study.

Agreement Overall

There was complete agreement between categories of self-reported and operator-recorded number of calls for 37% of cases and 39% of controls (Table 3); another 41% of both cases and controls showed agreement within adjacent categories. Weighted κ-values were 0.45 for cases and 0.47 for controls. Similarly, call duration showed perfect agreement for 41% of cases and of controls and agreement within adjacent categories for another 32% of cases and 34% of controls. Weighted κ-values for call duration were 0.45 for both cases and controls.

Table 3 Crosstabulations of numbers and cumulative duration of calls between self-reports and operator records.

Median, minimum, and maximum numbers of calls and duration of calls per month as reported in the questionnaires and recorded by the operators are shown in Table 4 for cases and controls in each country. On average, both cases and controls underestimated the number of calls they made by a factor of 0.81 (95% CI: 0.71, 0.93 for cases; 0.73, 0.91 for controls) (Table 5). Call duration was overestimated by around 40% on average by cases and controls (ratio 1.39, 95% CI: 1.18, 1.67 for cases; ratio 1.40, 95% CI 1.21, 1.60 for controls). Ratios were slightly higher when only data up to 1 year before the interview were analysed or when Australia was excluded. There was large variation among individuals of similar magnitude for cases and controls as illustrated by the 95% limits of agreement; these ranged for number of calls from 0.10 to 6.4 for cases and 0.12 to 5.6 for controls (Figure 1), and for duration of calls from 0.10 to 18.8 in cases and 0.12 to 16.1 for controls (Figure 2). The ratio of reported to recorded use increased strongly with increasing level of use for both cases and controls and for both number and duration of calls, showing underestimation at lower levels and overestimation at higher levels of use (Figures 1 and 2).

Table 4 Number and cumulative duration of calls from self-reports and operator records, by country and case–control status.
Table 5 Ratio of self-reported to operator-recorded mobile phone use.
Figure 1
figure1

Bland–Altman plot: ratio of self-reported to recorded numbers of calls against mean numbers of calls (log transformed); lines indicate the mean ratio, the 95% limits of agreement (±2*SD), and the regression line; for (a) cases and (b) controls.

Figure 2
figure2

Bland–Altman plot: ratio of self-reported to recorded duration of calls against mean duration of calls (log transformed); lines indicate the mean ratio, the 95% limits of agreement (±2*SD), and the regression line; for (a) cases and (b) controls.

Ratios for number of calls differed between countries in both cases and controls, due mainly to lower ratios in Australia, but there were no significant differences between cases and controls in any country. For duration of calls, cases in Italy tended to overreport more than controls, whereas in Canada controls tended to overreport more than cases.

For number of calls, the mean ratio and its standard deviation differed little between the main types of tumours. Glioma cases showed a slightly higher mean level of overreporting of duration of calls (ratio=1.54) than meningioma (ratio=1.38) and other (ratio=1.23) cases, but confidence intervals largely overlapped.

Agreement by Period before Interview

Mean ratios of reported to recorded number and duration of calls increased significantly with increasing time before interview among cases, but only minimally for number of calls and not for call duration among controls (Table 6). The trends persisted after adjustment for cumulative lifetime phone use or years-since-start-use, and were present in all case groups (data not shown). The trends were apparent in all three countries, although very few Canadian subjects had data in periods more than 2 years before the interview (data not shown). The comparison of ratios over time may be somewhat misleading as few subjects had data in all periods. Analyses were, therefore, also carried out for the restricted group of Australian and Italian subjects who contributed data to each of the first four time periods. In Australia, trends in the restricted group were similar to the overall group (Table 6). In Italy, the restricted case group showed a trend only for number of calls, not for duration.

Table 6 Ratio of self-reported to operator-recorded numbers and duration of calls by time period before interview.

Agreement by Reported Lifetime Use and Time since Start of Use

Strong trends of increasing ratios of reported to recorded use were found by categories of reported lifetime cumulative number of calls and call time (the main exposure variables in INTERPHONE), ranging from a high level of underreporting in lower categories of use to a high level of overreporting in the upper categories (Table 5). These trends were similar for cases and controls. Trends were similar for all case groups (not shown). Ratios also increased with increasing years-since-start-of-use for number and duration of calls (Table 5), and with reported number of phones used (not shown).

Agreement by Other Variables

The ratio of self-reported to recorded call duration did not differ appreciably by the option chosen to report call duration (per call or per day/week/month), for either cases or controls (not shown). Furthermore, no material differences were observed between sexes, age groups, education categories, or by quality of memory as rated by the interviewer (not shown).

Discussion

This validation study is the first to evaluate recall error in mobile phone use for cases and controls. Self-reports were compared with records of phone use kept by network operators, the assumed “gold standard”. It confirms earlier reports that recall of mobile phone use is often inaccurate (Parslow et al., 2003; Samkange-Zeeb et al., 2004; Berg et al., 2005; Vrijheid et al., 2006a; Hours et al., 2007b). Values of the κ-statistic showed only moderate agreement between reported and actual phone use. No more than about 40% of subjects were able to correctly estimate their phone use in categories defined by quintiles. On average, cases and controls underestimated the number of calls they made by 23% and overestimated call duration by 40%. These recall errors are similar to those observed in our validation study of healthy volunteers in 11 countries studying recall over only 6 months: 9% underestimation of number of calls and 42% overestimation of duration (Vrijheid et al., 2006a). The two studies were also similar in subjects’ tendencies to underreport lower levels and overreport higher levels of phone use, and in their levels of random error.

Of particular importance in assessing recall bias is the extent to which recall errors, systematic or random, differed between cases and controls. The systematic and random errors we observed were similar for cases and controls for phone use accumulated over the entire period of the validation study, as well as for phone use during the time periods closest to the interview. Results were somewhat inconsistent between countries; only in Italy did cases overreport more than controls. Our observations were limited, nonetheless, to use within about the past 4–5 years, whereas the main exposure variables in the full INTERPHONE analysis accumulate phone use over the entire period of use for each person.

For all case groups, but not for controls, we observed the ratio of reported to recorded duration and number of calls to increase with increasing time before the interview. These trends were based on few cases with long-term data and were somewhat inconsistent between countries, but they could cause positive bias in measures of disease risk associated with mobile phone use in the intermediate-to-distant past. The reasons for these findings are unclear. Poorer recall of more distant phone use by cases is conceivable if, for example, long-term memory in brain tumour cases was more affected than short-term memory. There is considerable literature on memory impairment and cognitive dysfunction in people with brain tumours (Salander et al., 1995; Weitzner and Meyers, 1997; Tucha et al., 2000), but evidence concerning long-term memory is limited. In any case, memory problems in brain tumour cases would be expected to lead to more random error rather than more overreporting. Random errors were, in fact, similar between the different tumour types and, although somewhat higher levels of overestimation of call duration were observed in brain tumour cases than in other cases, the trend to greater overreporting in earlier time periods was similar in cases of all tumour types. Rumination bias, whereby cases consciously or unconsciously overestimate their phone use in an effort to explain their disease, might account for these results, but it is unclear why such a bias would affect reporting of distant phone use more than recent phone use.

Errors in recall varied between countries, possibly because of differences in study methods. For example, the types of subscriptions included differed — mainly prepaid in Italy, only contract subscriptions in the other countries — as did the level of completeness of operator data, the time period over which data were collected, and the inclusion of outgoing calls and shared users (Table 1). The correction factor of two used to account for missing incoming calls in Australia may have influenced the absolute levels of error in this country, but we think it is unlikely to have affected the comparisons of main interest in this study, that is, those between cases and controls. Analyses of Italian and Canadian data sets did not show significant differences in the ratio of outgoing-to-incoming calls between cases and controls, or between sexes and age groups (not shown).

This study included about a quarter of cases and controls who used mobile phones in the three participating countries. This subsample did not differ from the full set of study participants on sex, age, time since start of phone use, and level of phone use. We do not know whether reasons for agreeing to participate in the validation study may have been related to recall of phone use. It is possible, for example, that cases with more severe diseases and perhaps worse memory, were less likely to participate. Furthermore, proxy interviews in Italy were not included in the validation study, although they represented a large proportion of case interviews there (44% of all glioma cases) and may have had worse recall.

The “gold standard” used in this study may itself have been subject to error due to errors in personal identifiers used for linkage, erroneous or missing traffic or billing records, and so on; we have no way of quantifying such errors. We did, however, exclude all uncertain matches between operator and interview data, as well as phone users for whom the operator data might not correctly reflect actual use, such as those who did not hold the subscription for the phone they used and, even though incompletely, those who shared their phone with another user for an important fraction of its use time.

If there is a real association between mobile phone use and brain tumour risk, the nondifferential random errors that dominate this validation study would be expected to bias risk estimates towards the null (no effect), and increase their uncertainty, making it more likely that real associations would not be detected (Armstrong, 1990; Armstrong et al., 1992; Vrijheid et al., 2006b). However, such errors do not normally induce spurious associations if there is no true association between exposure and outcome (Armstrong, 1990). Systematic under- or overestimation of phone use would bias estimates of risk per unit of exposure, respectively, upwards or downwards. The nondifferentially increasing systematic recall errors by level of use observed in this study would be expected to weaken any “true” dose–response. Differential errors observed in longer term phone use could induce positive bias. We have observed in a simulation study that, in the presence of large random errors, systematic errors made very little additional impact on assumed risk estimates for continuous exposure indices (cumulative hours of phone use), even when relatively extreme systematic errors were modelled and when the systematic errors simulated differed between cases and controls (Vrijheid et al., 2006b).

Conclusion

Most studies of mobile phone use have used a case–control approach and rely on participants’ reports of past phone use for their exposure assessment. They are, therefore, subject to recall bias. This study confirms earlier reports that recall of the amount of phone use is imperfect and prone to moderate systematic and substantial random error. Overall, there was little evidence that recall errors differed between cases and controls, but findings were not entirely consistent between countries. Apparent trends towards increasing overreporting with increasing time since interview (up to 4–5 years) were observed for cases, but not controls. This could cause positive bias in measures of disease risk associated with mobile phones use in the intermediate-to-distant past, and requires consideration in the interpretation of case–control studies of possible health effects of mobile phones.

References

  1. Armstrong B.G. The effects of measurement errors on relative risk regressions. Am J Epidemiol 1990: 132: 1176–1184.

    CAS  Article  Google Scholar 

  2. Armstrong B.K., White E., and Saracci R. Principles of Exposure Measurement in Epidemiology. Oxford University Press, Oxford, 1992.

    Google Scholar 

  3. Auvinen A., Hietanen M., Luukkonen R., and Koskela R.S. Brain tumors and salivary gland cancers among cellular telephone users. Epidemiology 2002: 13: 356–359.

    Article  Google Scholar 

  4. Auvinen A., Toivo T., and Tokola K. Epidemiological risk assessment of mobile phones and cancer: where can we improve? Eur J Cancer Prev 2006: 15: 516–523.

    Article  Google Scholar 

  5. Berg G., Schuz J., Samkange-Zeeb F., and Blettner M. Assessment of radiofrequency exposure from cellular telephone daily use in an epidemiological study: German validation study of the international case–control study of cancers of the brain — INTERPHONE-Study. J Expo Anal Environ Epidemiol 2005: 15: 217–224.

    Article  Google Scholar 

  6. Bland J.M., and Altman D.G. Comparing methods of measurement: why plotting difference against standard method is misleading. Lancet 1995: 346: 1085–1087.

    CAS  Article  Google Scholar 

  7. Bland J.M., and Altman D.G. Measuring agreement in method comparison studies. Stat Methods Med Res 1999: 8: 135–160.

    CAS  Article  Google Scholar 

  8. Cardis E., and Kilkenny M. International case–control study of adult brain, head and neck tumours: results of the feasibility study. Rad Prot Dos 1999: 83: 179–183.

    Article  Google Scholar 

  9. Cardis E., Richardson L., Deltour I., Armstrong B., Feychting M., and Johansen C., et al. The INTERPHONE Study: design, epidemiological methods, and description of the study population. Eur J Epidemiol 2007: 22: 647–664.

    Article  Google Scholar 

  10. Christensen H.C., Schuz J., Kosteljanetz M., Poulsen H.S., Boice Jr J.D., and McLaughlin J.K., et al. Cellular telephones and risk for brain tumors: a population-based, incident case–control study. Neurology 2005: 64: 1189–1195.

    Article  Google Scholar 

  11. Christensen H.C., Schuz J., Kosteljanetz M., Poulsen H.S., Thomsen J., and Johansen C. Cellular telephone use and risk of acoustic neuroma. Am J Epidemiol 2004: 159: 277–283.

    Article  Google Scholar 

  12. Drews C.D., and Greenland S. The impact of differential recall on the results of case–control studies. Int J Epidemiol 1990: 19: 1107–1112.

    CAS  Article  Google Scholar 

  13. Hardell L., Carlberg M., and Hansson M.K. Case–control study on cellular and cordless telephones and the risk for acoustic neuroma or meningioma in patients diagnosed 2000–2003. Neuroepidemiology 2005: 25: 120–128.

    Article  Google Scholar 

  14. Hepworth S.J., Schoemaker M.J., Muir K.R., Swerdlow A.J., van Tongeren M.J., and McKinney P.A. Mobile phone use and risk of glioma in adults: case-control study. Br Med J 2006: 332: 883–887.

    Article  Google Scholar 

  15. Hours M., Bernard M., Montestrucq L., Arslan M., Bergeret A., and Deltour I., et al. [Cell phones and risk of brain and acoustic nerve tumours: the French INTERPHONE case–control study]. Rev Epidemiol Santé Publique 2007a: 55: 321–332.

    CAS  Article  Google Scholar 

  16. Hours M., Montestrucq L., Arslan M., Bernard M., El Hadjimoussa H., and Vrijheid M., et al. [Validation des outils utilisés pour la mesure de la consommation téléphonique mobile dans l’étude INTERPHONE en France]. Evironnement, Risques & Santé 2007b: 6: 101–109.

    Google Scholar 

  17. Infante-Rivard C., and Jacques L. Empirical study of parential recall bias. Am J Epidemiol 2000: 152: 480–486.

    CAS  Article  Google Scholar 

  18. Inskip P.D., Tarone R.E., Hatch E.E., Wilcosky T.C., Shapiro W.R., and Selker R.G., et al. Cellular-telephone use and brain tumors. N Engl J Med 2001: 344: 79–86.

    CAS  Article  Google Scholar 

  19. Inyang I., Benke G., McKenzie R., and Abramson M. Comparison of measuring instruments for radiofrequency radiation from mobile telephones in epidemiological studies: implications for exposure assessment. J Exp Sci Environ Epidemiol 2008: 18 (2): 134–141.

    Article  Google Scholar 

  20. Johansen C., Boice Jr J., McLaughlin J., and Olsen J. Cellular telephones and cancer — a nationwide cohort study in Denmark. J Natl Cancer Inst 2001: 93: 203–207.

    CAS  Article  Google Scholar 

  21. Klaeboe L., Blaasaas K.G., and Tynes T. Use of mobile phones in Norway and risk of intracranial tumours. Eur J Cancer Prev 2007: 16: 158–164.

    Article  Google Scholar 

  22. Lahkola A., Auvinen A., Raitanen J., Schoemaker M.J., Christensen H.C., and Feychting M., et al. Mobile phone use and risk of glioma in 5 North European countries. Int J Cancer 2007: 120: 1769–1775.

    CAS  Article  Google Scholar 

  23. Landis J.R., and Koch G.G. The measurement of observer agreement for categorical data. Biometrics 1977: 33: 159–174.

    CAS  Article  Google Scholar 

  24. Lonn S., Ahlbom A., Christensen H.C., Johansen C., Schuz J., and Edstrom S., et al. Mobile phone use and risk of parotid gland tumor. Am J Epidemiol 2006: 164: 637–643.

    Article  Google Scholar 

  25. Lonn S., Ahlbom A., Hall P., and Feychting M. Mobile phone use and the risk of acoustic neuroma. Epidemiology 2004: 15: 653–659.

    Article  Google Scholar 

  26. Lonn S., Ahlbom A., Hall P., and Feychting M. Long-term mobile phone use and brain tumor risk. Am J Epidemiol 2005: 161: 526–535.

    Article  Google Scholar 

  27. Morrissey J.J. Radio frequency exposure in mobile phone users: implications for exposure assessment in epidemiologic studies. Radiat Prot Dosimetry 2007: 123: 490–497.

    Article  Google Scholar 

  28. Muscat J.E., Malkin M.G., Shore R.E., Thompson S., Neugut A.I., and Stellman S.D., et al. Handheld cellular telephones and risk of acoustic neuroma. Neurology 2002: 58: 1304–1306.

    CAS  Article  Google Scholar 

  29. Muscat J.E., Malkin M.G., Thompson S., Shore R.E., Stellman S.D., and McRee D., et al. Handheld cellular telephone use and risk of brain cancer. JAMA 2000: 284: 3001–3007.

    CAS  Article  Google Scholar 

  30. Parslow R.C., Hepworth S.J., and McKinney P.A. Recall of past use of mobile phone handsets. Radiat Prot Dosimetry 2003: 106: 233–240.

    CAS  Article  Google Scholar 

  31. Salander P., Karlsson T., Bergenheim T., and Henriksson R. Long-term memory deficits inpatients with malignant gliomas. J Neurooncol 1995: 25: 227–238.

    CAS  Article  Google Scholar 

  32. Samkange-Zeeb F., Berg G., and Blettner M. Validation of self-reported cellular phone use. J Expo Anal Environ Epidemiol 2004: 14: 245–248.

    Article  Google Scholar 

  33. Schlehofer B., Schlaefer K., Blettner M., Berg G., Bohler E., and Hettinger I., et al. Environmental risk factors for sporadic acoustic neuroma (Interphone Study Group, Germany). Eur J Cancer 2007: 43: 1741–1747.

    CAS  Article  Google Scholar 

  34. Schoemaker M.J., Swerdlow A.J., Ahlbom A., Auvinen A., Blaasaas K.G., and Cardis E., 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: 93: 842–848.

    CAS  Article  Google Scholar 

  35. Schuz J., Bohler E., Berg G., Schlehofer B., Hettinger I., and Schlaefer K., et al. Cellular phones, cordless phones, and the risks of glioma and meningioma (Interphone Study Group, Germany). Am J Epidemiol 2006a: 163: 512–520.

    Article  Google Scholar 

  36. Schuz J., Jacobsen R., Olsen J.H., Boice Jr J.D., McLaughlin J.K., and Johansen C. Cellular telephone use and cancer risk: update of a nationwide Danish cohort. J Natl Cancer Inst 2006b: 98: 1707–1713.

    Article  Google Scholar 

  37. StataCorp. STATA Statistical Software: Release 9.0. Stata Press, College Station, Texas, 2005.

  38. Takebayashi T., Akiba S., Kikuchi Y., Taki M., Wake K., and Watanabe S., et al. Mobile phone use and acoustic neuroma risk in Japan. Occup Environ Med 2006: 63: 802–807.

    CAS  Article  Google Scholar 

  39. Tucha O., Smely C., Preier M., and Lange K.W. Cognitive deficits before treatment among patients with brain tumours. Neurosurgery 2000: 47: 324–333.

    CAS  Article  Google Scholar 

  40. Vrijheid M., Cardis E., Armstrong B.K., Auvinen A., Berg G., and Blaasaas K.G., et al. Validation of short-term recall of mobile phone use for the Interphone Study. Occup Environ Med 2006a: 63: 237–243.

    CAS  Article  Google Scholar 

  41. Vrijheid M., Deltour I., Krewski D., Sanchez M., and Cardis E. The effects of recall errors and of selection bias in epidemiologic studies of mobile phone use and cancer risk. J Expo Sci Environ Epidemiol 2006b: 16: 371–384.

    Article  Google Scholar 

  42. Warren H.G., Prevatt A.A., Daly K.A., and Antonelli P.J. Cellular telephone use and risk of intratemporal facial nerve tumor. Laryngoscope 2003: 113: 663–667.

    Article  Google Scholar 

  43. Weitzner M.A., and Meyers C.A. Cognitive functioning and quality of life in malignant glioma patients: a review of the literature. Psycooncology 1997: 6: 169–177.

    CAS  Article  Google Scholar 

  44. White E. Design and interpretations of studies of differential exposure measurement error. Am J Epidemiol 2003: 157: 380–387.

    Article  Google Scholar 

Download references

Acknowledgements

We thank the network operators in all countries and the subjects who kindly accepted to participate in the study. We also acknowledge study investigators and staff who made nonauthorship contributions to the validation study: Monika Moissonnier (IARC, Lyon, France), Enrichetta Barbieri and Cristiano Tesei (ISS, Rome), Louise Nadon (Montreal), Tracey McPhail (Melbourne), Matthew Carroll (Sydney). This study was conducted with funding from the European Fifth Framework Program, ‘‘Quality of Life and Management of Living Resources’’ (contract QLK4-CT-1999901563), and the International Union against Cancer (UICC). The UICC received funds for this purpose from the Mobile Manufacturers’ Forum and GSM Association. Provision of funds to the INTERPHONE study investigators via the UICC was governed by agreements that guaranteed INTERPHONE's complete scientific independence. The terms of these agreements are publicly available at http://www.iarc.fr/ENG/Units/RCAd.html/. The Australian Centre was supported by the National Health and Medical Research Council (EME Grant 219129); Bruce K. Armstrong is supported by a University of Sydney Medical Foundation Program grant and Julianne Brown by an Australian Postgraduate Award. The Cancer Council NSW and The Cancer Council Victoria provided most of the infrastructure for the project in Australia. The Canada–Montréal data collection was funded by a grant (project MOP-42525) from the Canadian Institutes of Health Research (CIHR). The other Canadian centres were supported by a university–industry partnership grant (project POP-200102UOP-UI-90605) from CIHR, the latter including partial support from the Canadian Wireless Telecommunications Association. D. Krewski is the NSERC/SSHRC/McLaughlin Chair in Population Health Risk Assessment at the University of Ottawa.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Martine Vrijheid.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Vrijheid, M., Armstrong, B., Bédard, D. et al. Recall bias in the assessment of exposure to mobile phones. J Expo Sci Environ Epidemiol 19, 369–381 (2009). https://doi.org/10.1038/jes.2008.27

Download citation

Keywords

  • bias
  • recall
  • case–control studies
  • cellular phone
  • brain neoplasms

Further reading

Search

Quick links