ID Number		170
Study Type		Epidemiology
Model		INTERPHONE - Swedish National site 900, 1800 MHz (NMT, GSM, W-CDMA) mobile phone use and correlation with brain, acoustic nerve, parotid gland tumors
Details		Cases of malignant glioma & meningioma in Sweden (29-60 yrs old diagniosed from 200 - 2002) were evaluated for correlations with mobile phone use. The authors report no correlations with mobile phone use for glioma (OR = 0.8; 95% CI 0.6-1.0 based upon 214 cases) or meningioma (OR = 0.7; 95% CI 0.5-0.9 based upon 118 cases). There was no correlation for any subcategory of glioma or any tumor site. ORs did not differ between men and women, and results were not changed by adjusting for family history of cancer or medical X-ray exposure. There was no increased risk of either glioma or meningioma with mobile phone use in rural vs. urban areas. There was also no correlation with laterality (tumors on the same side of the head as primary phone use). In a separate evaluation from the same data from Sweden, patients with acoustic neuroma (n=150) were assessed for possible correlations with mobile phone use. AN = a benign tumor of the acoustic nerve that affects ~1 in 100,000 individual. Controls (n = 600) included healthy individuals without tumors. Data was obtained from Swedish cancer registrar and the network operator's registrar. No distinction was made in the analysis between GSM and analogue phone use - at the time of the study GSM use would be anticipated for the last 5-7 years back with more dated use probably involving NMT or other analogue technologies. There was no excess incidence of acoustic neuroma in users as a whole (OR = 1.0, 95% CI = 0.6-1.5). There was an elevated trend but no statistically significant increase with all users with 10+ years of use (OR = 1.9, 95% CI = 0.9-4.1). However, when ipsilateral acoustic neuromas were evaluated with phone use, a statistical significance was observed with "long duration" users (for 10+ years of use = OR 3.1, 95% CI 1.2-8.4; for 10+ years since first use = OR 3.9, 95% CI 1.6-9.5). The number of cases for the ipsilateral analysis was 11, with 26 controls. The correlation was also not between the tumors and time on the mobile phone in terms of hours or cumulative number of calls, but on the period of time that people in the study had been regular users (regular user = at least one call per week during six months or more). Numerous (5) comments followed publication and questioned the study methods and interpretation of the ipsilateral acoustic neuroma data. A subsequent paper by the same group and using the same data set described a statistically significant correlation between acoustic neuroma and loud noise from machines, power tools, and construction (OR = 1.79; 95% CI: 1.11, 2.89), as well as loud music (OR = 2.25; 95% CI: 1.20, 4.23). The authors also report a tendency for AN to occur with higher incidence on the right side. In a subsequent paper in 2006, the authors report no correlation between regular mobile phone use and malignant (n = 60; OR = 0.7, 95% CI 0.4-1.3) and benign (n = 112; OR = 0.9, 95% CI 0.5-1.5) parotid gland tumors. For benign parotid gland tumors with ipsilateral 10+ yr use, a non-significant increase was reported (OR = 2.6, 95% CI 0.9-7.9, based on 6 cases) with a decreased contralateral risk (OR = 0.3, 95% 0 -2.3, based on 1 case). Overall study findings: OR = 0.8; 95% CI 0.6-1.0 - glioma (no ipsilateral, no 10+ yr), OR = 0.7; 95% CI 0.5-0.9 - meningioma (no ipsilateral, no 10+ yr), OR = 0.7; 95% CI 0.4-1.3 - malignant salivary, OR = 0.9; 95% CI 0.5-1.5 - benign salivary, OR = 1.0; 95% CI 0.6-1.5 - acoustic neuroma, OR = 1.9; 95% CI 0.9-4.1- 10+ yr acoustic neuroma, OR = 3.9; 95% CI 1.6-9.5 - ipsilateral /10+ yr acoustic neuroma.
Findings		Effects
Status		Completed With Publication
Principal Investigator		Karolinska Institute, Sweden
Funding Agency		EU, INTERPHONE (IARC) coordinated studies, MMF, GSM Association, Research Council, Sweden
Country		SWEDEN
References		Lonn, S et al. Am J Epidemiol , (2006) 164:637-643 Lonn, S et al. Am J Epidemiology., (2005) 161:526-535 Stang, A et al. Epidemiology , (2005) 16:414-415 Hardell, L et al. Epidemiology, (2005) 16:415-415 Thomas, BN et al. Epidemiology , (2005) 16:415-416 Johnston, SA et al. Epidemiology 16(3):416-7., (2005) 16:416-417 Lonn, S et al. Epidemiology, (2005) 16:417-417 Lonn , S et al. Epidemiology, (2004) 15:653-659 Savitz, DA Epidemiology, (2004) 15:651-652 Tarone, RE et al. Epidemiology, (2005) 16:414-
Comments		The data is part of a larger joint Nordic-South UK study now being analyzed. For the acoustic neuroma analysis, the risk in heavy users did not seem to follow an obvious dose response (OR = 1.6 for 10+ years of use [11 cases], OR = 1.9 for 10+ years since first use [14 cases], OR = 1.1 for 450+ hrs of use [25 cases], OR = 1.2 for 7350+ calls [28 cases]) and none were statistically significant. The acoustic neuromas that occurred in the long-term users (10+ years) occurred disproportionately on the side of the head where they reported "generally using" their mobile phones, but for users with only 5-9 years of use there is no sign of such a "sidedness effect" while the size and incidence of the tumors might change between the 10+ and 5-9 groups, the sidedness ratio should not change. With regard to sidedness, 91% of the users reported "generally using" the phone on the same side of their head. However, an answer to this vague question, particularly when being asked about a 10+ year history of use, and for someone who already knows they have an acoustic neuroma on one side of the head, would seem to invite potential "recall bias". The authors acknowledge that this is an alternative explanation for their results. The participation rate was rather high by US experience (93% of cases and 72% of controls). There is also some question regarding potential bias in the selection of both the controls and cases, (only 10% came from the Swedish cancer registry the rest were identified, apparently ad hoc, during the course of the study). This may have introduced a case population that did not necessarily represent the general Swedish population (specifically, in table 1 [page 655] 20% of cases and 22% of controls had no education past elementary school; 24% of cases / 27% of controls only finished the first two years of high school; 18% of cases / 20% of controls finished high school - indicating that 62% of cases / 69% of controls never continued on to higher education, and one-fifth only have a basic, elementary education this is not reflective of the general Swedish population. Controls were not people with other types of benign tumors, but healthy people from the population registry. While a preceding evaluation did suggest ipsilateral tumors were associated with NMT (OR 1.7, 95% CI 1.2-2.3) as well as with GSM (OR 1.3, 95% CI 1.02-1.8), this was total tumors and not benign acoustic neuromas (Hardell et al., 2003). In the conclusion section of the paper the authors discuss possible response bias, recall bias, information bias and selection bias. In addition, small sample bias (statistical bias) is also an issue due to the sample size  this is mentioned at the bottom of the first column of page 656 ("Side-specific estimates were unstable ..."). This can result in unreliable relative risk (RR, or OR) estimates. An indicator of such bias is large confidence intervals (CI) which are larger for the statistically significant findings than for other OR estimates in Tables 2 and 3. The study design is a matched case-control - matching variables are age, sex and residential area of the study subject with 2 controls selected for every case occurrence to account for confounding, but this matched case-control design appears to be abandoned in the tumor laterality statistics  so it is no longer clear what kind of OR measure was used. Theoretical work has reported that if matched binary data are analyzed as if they were unmatched (by using unconditional logistic regression where the matching variables are included as confounders), the resulting estimates of the OR can be biased. The amount of bias is the larger with more confounders present in the study, or more precisely the more parameters you need to represent them in the model. It is not clear whether Lonn and coauthors accounted for this. The results of Lonn et al are not supported by laterality or total tumor data from Muscat et al (1999), Inskip et al (2000), Johanson et al (2001), or Auvinen (2002). A report by Christensen et al., a population-based study based on complete, high-quality registers (Am J Epidemiol, Vol. 159, Pg. 277 - 283, 2004) reported NO INCREASE in acoustic neuroma (report from the Danish National Registry INTERPHONE study group), and they had a large case and control sample size (106 and 212, respectively). The biggest criticism of the study is the exposure assessment as well as the small number of cases and corresponding controls in each dose group (lack of statistical power). Drs. Lonn, Feyching and Ahlbom have a 2004 paper in Occupational Medicine characterizing mobile phone exposure using network data and state in the Introduction section that "While the weight of evidence does not suggest adverse effects, the negative results cannot be considered conclusive. This is particularly true for epidemiological studies, which suffer from severe limitations in the exposure assessment and other methodological problems". However, in the immediate study with only questionnaire data for exposure assessment and a limited number of cases for statistics, they state "the data suggest an increased risk of acoustic neuroma associated with mobile phone use of at least 10 years duration". Finally, acoustic neuromas cause hearing loss, so I would have predicted that most mobile phone users with acoustic neuromas would use the phone on the contralateral side. Immediately following publication in the journal Epidemiology, the story ran in the US media on the Today Show (Eastern USA) featuring Dr. Ted Schwartz of New York Presbyterian Hospital and Louis Slesin of Microwave News. In letter tot he editor replies, Tarone and Inskip criticized assumptions made in unexplained exclusion of 9 cases that "exaggerates" the difference between the Swedish and Danish INTERPHONE results, and the suggestion that Inskip and Tarone's 2001 study supported a positive relationship between tumor laterality and usual side mobile phone use. Stang suggest that the epidemiologic findings "fit together". Hardell and Mild also point out 9 cases excluded without explanation and too few histologically verified cases. Thomas questioned recall bias. Sheila Johnston and Hagen Scherb suggest the acoustic neuroma analysis had only a 39% statistical power to detect an effect, not the 80% recommended by the INTERPHONE Study, and that Lonn did not correct for multiple comparisons problems, and that 2 significantly elevated risk ratios from 37 comparisons is not surprising (1.85 predicted by chance alone). Lonn et al (2005) gathered his data from regional cancer registries in Umea, Stockholm, Goeteborg, and Lund from September 1, 2000 thorugh August 31, 2002, while Hardell et al (2002, 2005,2006) gathered data from registries in Uppsala-Orebro, Stockholm, Linkoping, and Goteborg from January 1997 through June 2000 and extended through December 2003. Since much of this data overlaps, it is odd that such different ORs were obtained. That point could be emphasized.