This pilot meta-analysis based on few reports has suggested a weak association between cellphone use and the risk of acoustic neuroma. These findings will need confirmation by analysis of more research reports. The results of the expanded study will be used as the basis for planning a long-term prospective study of the risk of cancer among long-term users of cell phones.

Introduction

Acoustic neuroma is a benign, slow growing tumor of the 8^th cranial nerve. It presents with otological symptoms (tinnitus, vertigo, and hearing loss) and signs (abnormal hearing tests, facial numbness and weakness, and papilledema). Larger tumors can cause symptoms and signs due to compression of the brain stem and involvement of the facial and trigeminal nerves. Men are affected more than women. Presentation is above 30 years. Elderly patients without serious symptoms and signs are left untreated but are followed up for complications. Micro surgery, radio surgery or a combination of the two may be used in treatment.

Electromagnetic radiation is a known cause of cancer. The localized radio frequency microwave energy emitted by cell phones has been suspected as a cause of brain malignancies. This is an issue of public concern because cellphone use is increasing very rapidly in Brunei and other countries. Evidence indicates that the incidence of brain tumors has been rising in the recent past when cellphone use became very popular. Hardell et al found a significant increase of +0.80% in the incidence of all brain tumors taken together for the 1960-1998 [1]. The risk of cancer in association with cellphone use has also been observed to rise in the same period. Hardell et al 2003 in a computation of annual risk increase by treating exposure as a continuous variable showed increase of risk with time the annual risk increase being 1.04 (1.01 – 1.08) [2].

Evidence of the relation between radio frequency electromagnetic fields and brain tumors has been contradictory. Some authors found no relation while others found the evidence to be weak and unconvincing [3,4,5]. The studies reviewed below show weak or insignificant association. Considering all brain tumors together, Hardell et al. 2002 in a study of 588 cases and 581 controls found the following odds ratios with 95% confidence intervals for analog cellphones 1.13 (0.86 – 1.48); for cordless phones 1.13 (0.85 – 1.50); and for digital cellphones OR = 1.59 (1.05 – 2.41). Ispilateral use increased the risk[6]. Hardell et al 2002 in a study of 1617 cases and 1617 controls found the risk for short term exposure to analog telephones to be OR = 1.3 (1.02 – 1.6) and for long term exposure OR = 1.8 (1.1 – 2.9). The risk was higher on the same side of regular cellphone use. There was no significant risk from cordless or digital cellphones[7]. Hardell at al 2004 reported the overall risk of using analog telephones to be OR = 1.31 (1.04-1.64). The risk increased to OR = 1.65 (1.19 – 2.30) for ipsilateral use. Risk was highest among the 20-29 age group with the ipsilateral risk being OR = 5.91 (0.63 – 55). This age group experienced a raised ipsilateral risk if the latency period was over 5 years with OR = 8.17 (0.94-71) for analog phones[8]. Lonn et al 2004 in a study of 148 cases and 604 controls found the risk of acoustic neuroma from mobile phone use to be OR = 1.0 (0.6 – 1.5) for short term use and OR = 1.9 (0.9 – 4.1) for long term use. The risk was increased on the same side as regular phone use[9]. Hardell et al 2006 in a study of 317 cases and 692 controls found the following risks for various cellphones and durations of use. The risk for analog cellphones was OR = 2.6 (1.5-4.3) for short term use and OR = 3.5 (2.0 – 6.4) for long term use. The respective risks for digital cell phones were OR=1.9 (1.3-2.7 and OR=3.6 (1.7-7.5) and for cordless phones OR=2.1 (1.4-3.0) and OR= 2.9 (1.6-5.2). Multivariate analysis showed all three phone types to be associated with increased risk[10].

The focus of the present review are case control studies relating cellphone use to acoustic neuroma. Hardell et al 2003 in a study of 1429 cases and 1470 controls found the risk of acoustic neuroma among analog telephone users to be OR = 4.4 (2.1-9.2)[2]. Hardell et al 2005 in a case control study of 84 acoustic neuroma cases found the risk from analog phones to be OR = 4.2 (1.8 -10) for the short term and OR = 8.4 (1.6-45) for long term exposure. The risk for digital phones was OR = 2.0 (1.05 – 3.8). Cordless phones did not show increased risk. Multivariate analysis showed analog phones to be an independent risk factor for acoustic neuroma[11]. Schoemaker et al 2005 in a study of 678 cases and 3553 controls found no increased risk between regular cellphone use and acoustic neuroma this applying even if the analysis was carried out separately for analog and digital cellphones. Risk was increased for the same side and for long term exposure OR 1.8 (1.1 – 3.1)[12]. Takebayashi et al 2006 in a study of 101 cases and 339 matched controls found no association between cellphone use and acoustic neuroma. There was no association between risk and cumulative years of cellphone use[13]. Hardell et al 2006 in an analysis of 2 pooled case control studies with 1254 cases and 2162 controls found the risk of acoustic neuroma to be OR = 2.9 (2.0 – 4.3) for analog cellphones, OR = 1.5 (1.1 – 2.1) for digital cellphones, and OR = 3.8 (1.4 – 10) for cordless phones. The risk for analog cellphones increase if exposure was >15 years to OR = 3.8 (1.4 – 10). Multivariate analysis showed use of analog cellphones to be an independent risk factor for acoustic neuroma[14]. Schlehofer et al 2007 in a study of 97 cases and 194 matched controls found the risk of acoustic neuroma from regular mobile phone use to be OR = 0.67 (0.38-1.19)[15].

We can conclude from the literature survey above that studies relating cellphone use and brain cancers in general are either negative or show a weak association but the trend to increasing risk with longer duration of cellphone use is very clear. This indicates that the risk may exist but is not detected due to 3 methodological defects explain the results: duration of follow up not sufficient, inaccurate measurement of the level of exposure and biases of response and recall[16].

The present study is a review of recent studies on cell phone use and acoustic neuroma. The objective of this preliminary study is to derive an estimate of acoustic neuroma risk by combining data from a few case control epidemiological studies. This is a pilot study that will be extended to include more studies as soon facilities for extensive literature search are available. All these efforts will culminate in the design and execution of a long-term prospective study in Brunei of the relation between cellphone use and risk of various malignancies. Brunei has an advantage for such a study because of ease of follow up in a small population.

Methods

Five case control studies from the Interphone international collaborative study of the association between cell-phone use and cancer were identified with the help of PUBMED. The studies were all carried out using the same protocol so they had similar design and analytic methods. Tables 1 and 2 summarize the salient features of each research report. The odds ratio with 95% confidence intervals was abstracted from each report. Other essential data abstracted were: type of cell phone used, years of cellphone use <10=short, >10= long), and number of study subjects. The inverse variance meta analytic method was used compute a pooled odds ratio over several studies by summation of the odds ratios of individual studies each being weighted by the inverse of its variance. OR_p = ∑ w_i OR_{i /} ∑ w_i where OR_p = pooled odds ratio, w_i = weighting which is the inverse of the variance of the odds ratio. The 95% Confidence Intervals were computed using the standard error  S(OR_p) = 1/ sqrt{∑ w_i.} . Heterogeneity was tested using χ = ∑ w_i (OR_i - OR_p)² where w_i = 1/S_i² . All computations were carried out using log-transformed data.

Results

Tests for heterogeneity were negative so pooled effect easures were computed. There was no strong, consistent, and significant association between cell phone use and acoustic neuroma in the short term (less than 10 years of use). The data did however suggest increasing risk with long-term use, use of analog cell phones as compared to digital phones, and disease on the same side of the head as the cell phone is usually held. For research reports without specification of the type of cellphone, the pooled effect estimates (95% confidence limits) were OR_p = 0.9 (0.7, 1.0) for short term use and OR_p = 1.6 (1.1, 2.2) for long term cellphone use. The pooled effect measures for analog cellphones were OR_p = 3.1 (2.2, 4.4) for short term use and OR_p = 4.3 (2.2, 8.1) for long term use. The pooled effect measure for digital cellphone use in the short term was 1.6 (0.51, 4.9). Data was not available for long term digital cellphone use.

Discussion

The data suggests association between use of analog cellphones with acoustic neuroma. The association is significant for analog cellphone short term follow up. It is stronger for long term analog cellphones on longer term follow up but does not reach significance due to the large variance based on few research reports. Analysis of more research reports is needed to confirm these findings.

The data quality was high being collected under a uniform INTERPHONE protocol. The studies were also similar in design and data collection because they largely used the same protocol. Lack of detailed raw data prevented use of the Mantel-Haenszel method and sparsity of the data prevented control for confounding. Use of self-reported questionnaires had limitations in accurate measurement of the total duration of use, frequency of use every day, position in which the cell phone is used, type and power of the phone used. More accurate exposure information can be obtained from the billing records of cell phone subscriber companies which have detailed automated data on times of calls, duration of the calls, type of phone and strength of the radiation energy emitted. It is however doubtful that these companies will cooperate because of business self-interest. A study in Denmark found that there was a fair agreement between self-reported cellphone use and subscriber data. Risk measures based on the two exposure measurements were not very different from one another. Each of the 2 methods has its limitations[17].  The fair agreement between the 2 methods is good news because we can rely on self-reported use that we can get easily instead of trying to obtain subscriber information that is not easily accessible. Exposure assessment is the weak link in studies of the association between cellphone use and cancer. Self reported use of cellphones is unreliable for duration of exposure. The relationship between duration of use and strength of the electromagnetic field is not known. In view of these limitations prospective studies will be needed to settle the questions under study [18].

CONCLUSION

The current analysis has not showed a strong, consistent, or conclusive evidence of a link between cell phone use and acoustic neuroma although the data suggests such a link. Definitive answers will be obtained from studies of longer-term prospective studies because cancer has a long induction period.

TABLE #1: STUDIES WITH NO MENTION OF THE TYPE OF PHONE

TABLE #2: STUDIES THAT GAVE SEPARATE DATA FOR ANALOG AND DIGITAL CELLPHONES

TABLE #3: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR SHORT-TERM EXPOSURES (TYPE OF PHONE NOT KNOWN)

Computation of pooled estimate

OR_p =  exp {-21.7918/153.7689) = exp {-0.14171} = 0.8679

S(OR_p) =  1 / sqrt ∑ w_i  = 1 /sqrt(153.7689) = 1/12.4004 = 0.0806

LB(OR_p) = exp{-0.14171 – 0.0806*1.96} = exp{-0.14171 – 0.1580} = exp(-0.29971) = 0.7410

UB(OR_p) = exp{-0.14171 +  0.0806*1.96} = exp{-0.14171 +  0.1580}= exp (0.01629) = 1.01

Testing for heterogeneity

Heterogeneity was tested using χ = ∑ w_i (OR_i - OR_p)² where w_i = 1/S_i² with all data log-transformed for the computations

χ = [13.9333 (-0.3147- -0.14171)²] + [120.7584 (-0.1053 --0.14171)² + [18.3098 (0- -0.14171)² + [0.7674 (-0.40048 - -0.14171) ²]

χ = [13.9333*0.0299] + [120.7584*0.001326] + [18.3098*0.02008] + [0.7674*0.066961]

χ = 0.4166  +  0.16013 + 0.3677 + 0.0514

χ = 0.99586 (ns)

TABLE #4: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR LONG-TERM EXPOSURES (TYPE OF PHONE NOT KNOWN)

Computation of the pooled estimate including Schlehofer et al 2007

OR_p =  exp {3.7527/54.9751) = exp {0.06826} = 1.07

S(OR_p) = 54.9751^-2 = 0.1349

LB(OR_p) = exp{0.06826 –0.1349*1.96} = exp{0.06826 – 0.2644} = exp {-0.1961} = 0.8219

UB(OR_p) =  exp{0.06826 + 0.1349*1.96} = exp{ 0.06826 + 0.2644} = exp{0.3327} = 1.3947

Computation of the pooled estimate excluding Schlehofer et al 2007

OR_p = exp (13.5287/30.5645) = exp (0.4426) = 1.5567

S(OR_p) =  30.5645^-2 = 0.1809

LB(OR_p) = exp{0.4426 –0.1809*1.96} = exp{0.4426  – 0.3546} = exp {0.088} = 1.09

UB(OR_p) =  exp{0.4426  + 0.1809*1.96} = exp{0.4426 +0.3546 } = exp{0.7972} = 2.219

Test for heterogeneity including Schlehofer et al.

Heterogeneity was tested using χ = ∑ w_i (OR_i - OR_p)² where w_i = 1/S_i² with all data log-transformed for the computations

Χ = [9.5663(0.08618 – 0.06826) ²] + [14.3143(0.5878 – 0.06826) ²] + [6.6839(0.6419 - 0.06826) ²] + [24.4106 (-.4005- 0.06826) ²]

X = [9.5663*0.0003211] + [14.3143*0.2699]+ [6.6839*0.3291] + [24.4106*0.2197]

X =  0.003071 + 3.8634  + 2.1997 + 5.3630

X = 11.43 (s)

Test of heterogeneity excluding Schlehofer et al

OR_p = exp (13.5287/30.5645) = exp (0.4426) = 1.5567

S(OR_p) =  30.5645^-2 = 0.1809

X = [9.5663(0.08618-0.4426)²]  + [14.3143 (0.5878 – 0.4426) ² + [6.6839 (0.6419-0.4426) ²]

X = [9.5663*0.1270] + [14.3143*0.0211] + [6.6839*0.0397]

X = 1.21 + 0.3020 + 0.2653

X =  1.7773 (ns)

TABLE #5: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR SHORT TERM EXPOSURE TO ANALOG PHONES

Computation of the pooled odds ratio

OR_p = exp{ 35.4151/31.4447} = exp{1.1263} = 3.084

S(OR_p) = 1/sqrt(31.4447) = 1/5.6076 = 0.1783

LB(OR_p) = exp{1.1263–0.1783*1.96} = exp{1.1263–0.3495} = exp{0.7768} = 2.175

UB(OR_p) =  exp {1.1263 + 0.1783*1.96} = exp {1.1263 + 0.3495} = exp{1.4758} = 4.375

Test of heterogeneity

X= [5.2269 (1.4351-1.1263)²] + [26,2178 (1.0647–1.1263) ²]

X= [5.2269*0.0954] + [26.2178*0.00379456]

X= 0.4986  + 0.0995

X =  0.5981 (ns)

TABLE #6: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR SHORT TERM EXPOSURE TO DIGITAL PHONES

Computation of the pooled odds ratio

OR_p = exp{1.4206/3.0425} = exp{0.46692} = 1.595

S(OR_p) = {sqrt(3.0425)}^-1 = 0.57330

LB(OR_p) = exp{0.46692 –0.5733*1.96} = exp{0.46692 – 1.1237} = exp {-0.65678} = 0.5185

UB(OR_p) = exp {0.46692 + 0.5733*1.96} = exp{0.46692 + 1.1237} = exp {1.59062} = 4.9068

Test of heterogeneity

X = [0.6507(0.6931-0.66692)² + [2.3918(0.4054 – 0.46692) ²]

X = [0.6507*0.0006854] + [2.3918*0.003785]

X = 0.00044598978  +  0.009052963

X=  0.009498..(ns)

TABLE #7: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR LONG TERM EXPOSURE TO ANALOG PHONES

Computation of the pooled estimate

OR_p = exp {13.5153/9.3038} = exp{1.4527} = 4.274

S(OR_p) = 1/sqrt(9.3038) = 1/3.05 = 0.3278

LB(OR_p) = exp{1.4527-0.3278*1.96} = exp{1.4527-0.6425} = exp{0.8102} = 2.248

UB(OR_p) = exp{1.4527 + 0.3278*1.96} = exp{1.4527+0.6425} = exp{2.0925} = 8.105

Test for heterogeneity

X = [1.3802(2.1282-1.4527)² + [7.9236(1.3350–1.4527) ²]

X = [1.3802*0.4563] + [7.9236*0.013853]

X =  0.6297 + 0.10976

X=  0.73946 (ns)

TABLE #8: COMPUTATION OF THE COMBINED EFFECT ESTIMATE FOR SHORT TERM EXPOSURE IRRESPECTIVE OF THE TYPE OF PHONE

1.      Hardell L, Hansson Mild K, Sandstrom M, Carlberg M, Hallquist A, Pahlson A. Vestibular schwannoma, tinnitus and cellular telephones. Neuroepidemiology. 2003 Mar-Apr;22(2):124-9. For all other brain tumors taken together, the incidence significantly increased yearly by +0.80% (CI 0.59-1.02) for the time period 1960-1998, although the increase was only significant for benign tumors other than VS during 1960-1979.

2.       Hardell L, Mild KH, Carlberg M. Further aspects on cellular and cordless telephones and brain tumours. Int J Oncol. 2003 Feb;22(2):399-40.71429 cases; 1470 controls. For acoustic neurinoma OR=4.4, 95% CI=2.1-9.2 was calculated among analogue cellular telephone users. When duration of use was analysed as a continuous variable in the total material, the risk increased per year for analogue phones with OR=1.04, 95% CI=1.01-1.08

3.   Heynick LN, Johnston SA, Mason PA. Radio frequency electromagnetic fields: cancer, mutagenesis, and genotoxicity. Bioelectromagnetics. 2003;Suppl 6:S74-100. Overall, however, the preponderance of published epidemiologic and experimental findings do not support the supposition that in vivo or in vitro exposures to such fields are carcinogenic. Published 2003 Wiley-Liss, Inc.

4. Moulder JE, Erdreich LS, Malyapa RS, Merritt J, Pickard WF, Vijayalaxmi. Cell phones and cancer: what is the evidence for a connection? Radiat Res. 1999 May;151(5):513-31. Overall, the existing evidence for a causal relationship between RF radiation from cell phones and cancer is found to be weak to nonexistent.

5.   Moulder JE, Foster KR, Erdreich LS, McNamee JP. Mobile phones, mobile phone base stations and cancer: a review. Int J Radiat Biol. 2005 Mar;81(3):189-203. Overall, a weight-of-evidence evaluation shows that the current evidence for a causal association between cancer and exposure to RF energy is weak and unconvincing.

6.   Hardell L, Mild KH, Carlberg M. Case-control study on the use of cellular and cordless phones and the risk for malignant brain tumours. Int J Radiat Biol. 2002 Oct;78(10):931-6. 588 cases; 581 controls. ipsilateral use of analog phone increased risk of malignant brain tumor.

7.   Hardell L, Hallquist A, Mild KH, Carlberg M, Pahlson A, Lilja A. Cellular and cordless telephones and the risk for brain tumours. Eur J Cancer Prev. 2002 Aug;11(4):377-86. 1617 cases; 1617 controls. Analog short OR = 1.3 (1.02-1.6) long OR = 1.8 (1.1-2.9). No clear association was found for digital or cordless telephones. Risk higher on the same side. Highest risk was with acoustic neuroma OR = 3.5 (1.8-6.8)

8.   Hardell L, Mild KH, Carlberg M, Hallquist A.  Cellular and cordless telephone use and the association with brain tumors in different age groups. Arch Environ Health. 2004 Mar;59(3):132-7. 1,429 (88%) cases and 1,470 (91%) controls. Use of analog cellular telephones OR for brain tumors of 1.31 (1.04-1.64) but increasing for ipsilateral to OR = 1.65 (1.19-2.30). The authors found the highest risk for the 20-29-yr age group, with OR = 5.91 (0.63-55) for ipsilateral use of analog phones. The highest risks were associated with >5-year latency period in the 20-29-yr age group for analog phones OR = 8.17 (0.94-71), and cordless phones (OR = 4.30 (1.22-15)

9.   Lonn S, Ahlbom A, Hall P, Feychting M. Mobile phone use and the risk of acoustic neuroma. Epidemiology. 2004 Nov;15(6):653-9. 148 cases; 604 controls. For acoustic neuroma short term OR = 1.0 (0.6=1.5). Long term OR = 1.9 (0.9-4.1). For the same side OR = 3.9 (1.6 – 9.5).

10.Hardell L, Carlberg M, Mild KH. Case-control study of the association between the use of cellular and cordless telephones and malignant brain tumors diagnosed during 2000-2003. Environ Res. 2006 Feb;100(2):232-41. 317 cases (88%) and 692 controls (84%). analog cellular short (OR) of 2.6 (1.5-4.3), long OR=3.5 (2.0 – 6.4). digital cellular telephones short OR=1.9 (1.3-2.7 long OR=3.6 (1.7-7.5). Cordless telephones yielded short OR=2.1 (1.4-3.0) long OR= 2.9 (1.6-5.2). In multivariate analysis, all three phone types studied showed an increased risk.

11. Hardell L, Carlberg M, Hansson Mild 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(3):120-8. The association between cellphone use and various tumors was investigated in a case control study with 413 cases (305 meningiomas, 84 acoustic neuromas, 24 others) and 692 controls. For acoustic neuroma, analogue phones gave OR = 4.2, 95% CI = 1.8-10 increasing to OR = 8.4, 95% CI = 1.6-45 with a >15-year latency period. Digital phones yielded OR = 2.0, 95% CI = 1.05-3.8, whereas for cordless phones OR was not significantly increased. In the multivariate analysis, analogue phones represented a significant risk factor for acoustic neuroma.

12. Schoemaker MJ, Swerdlow AJ, Ahlbom A, Auvinen A, Blaasaas KG, Cardis E, Christensen HC, Feychting M, Hepworth SJ, Johansen C, Klaeboe L, Lonn S,McKinney PA, Muir K, Raitanen J, Salminen T, Thomsen J, Tynes T. Mobile phone use and risk of acoustic neuroma: results of the Interphone case-control study in five North European countries. Br J Cancer. 2005 Oct 3;93(7):842-8. 678 cases and 3553 controls. There was no relation between regular cellphone use and acoustic neuroma OR = 0.9 (0.7, 1.1). There was no increased risk is the analysis was carried out for analog and digital phones separately. There was no increased risk if the exposures were duration of use, lifetime cumulative hours of use, number of calls. There was an increased risk for a tumor on the same side of the head if cellphone use extended to 10 years or more OR 1.8 (1.1, 3.1) but the evidence was not strong.

13.  Takebayashi T, Akiba S, Kikuchi Y, Taki M, Wake K, Watanabe S, Yamaguchi N. Mobile phone use and acoustic neuroma risk in Japan.Occup Environ Med. 2006 Dec;63(12):802-7. The studies selected were part of an international collaborative effort and followed a common protocol. The study recruited 101 cases of acoustic neuroma aged 30-69 and residing in Tokyo. It recruited 339 controls matched for age, sex, and residency. There was no association between cellphone use and risk of acoustic neuroma. There was no association between cumulative years of use or cumulative call times with acoustic neuroma.

14.  Hardell L, Carlberg M, Hansson Mild K. Pooled analysis of two case-control studies on the use of cellular and cordless telephones and the risk of benign brain tumours diagnosed during 1997-2003. Int J Oncol. 2006 Feb;28(2):509-18. Results of 2 pooled case control studies 1254 cases abd 2162 controls. For acoustic neuroma analog phones OR = 2.9 (2.0 – 4.3), digital phones OR = 1.5 (1.1 – 2.1), cordless phones OR = 1.5 (1.04 – 2.0). For analog phones with latency >15 years OR = 3.8(1.4 – 10). In multivariate analysis analog was associated with acoustic neuroma.

15.  Schlehofer B, Schlaefer K, Blettner M, Berg G, Böhler E, Hettinger I, Kunna-Grass K, Wahrendorf J, Schüz J; Interphone Study Group. Environmental risk factors for sporadic acoustic neuroma (Interphone Study Group, Germany).  Eur J Cancer. 2007 Jul;43(11):1741-7. 97 case of AN and 194 matched controls. Acoustic neuroma ionizing radiation OR = 0.91 (0.51-1.61). Acoustic neuroma for regular mobile phone use OR = 0.67 (0.38 – 1.19)

16. Kundi M, Mild K, Hardell L, Mattsson MO. Mobile telephones and cancer--a review of epidemiological evidence. J Toxicol Environ Health B Crit Rev. 2004 Sep-Oct;7(5):351-84. All studies have some methodological deficiencies: (1) too short duration of mobile phone use to be helpful in risk assessment, (2)exposure was not rigorously determined, and (3) there is a possibility of recall and response error in some studies.

17. Schuz J, Johansen C. A comparison of self-reported cellular telephone use with subscriber data:agreement between the two methods and implications for risk estimation Bioelectromagnetics. 2007 Feb;28(2):130-6.. Institute of Cancer Epidemiology

18. Auvinen A, Toivo T, Tokola K. Epidemiological risk assessment of mobile phones and cancer: where can we improve? Eur J Cancer Prev. 2006 Dec;15(6):516-23.