REPORT TO ARPANSA ON EPIDEMIOLOGICAL STUDIES

Elwood: SC4: Epidemiological studies of cancer Page 1 of 32

EPIDEMIOLOGICAL STUDIES OF RADIOFREQUENCY EXPOSURES ANDHUMAN CANCER.

Paper prepared for SC4 of IEEE, May 2002

J. Mark Elwood

MD, DSc, FRCPC.

Director, National Cancer Control Initiative;

Professor, Monash University and University of Melbourne

© J Mark Elwood 2002

[email protected]

1 Rathdowne St, Carlton 3053 (Melbourne)

Australia

Fax 61-3-9635-5320

DECLARATION

This work is based partly on two reports I have published. One published paper [1] reviews

epidemiological studies up to 1999. The second has been published in 2002 as Annex 3 in the

Australia and New Zealand Radiation Protection Standard for radiofrequencies [2]. I authored

this report, with input and comments from the members of the standards working party. The

current document summarized some material presented in these documents and adds

comments on more recent studies and further discussion. The views expressed in this paper

are mine, and do not necessarily represent the views of any other groups.

mailto:[email protected]


EPIDEMIOLOGICAL STUDIES OF RADIOFREQUENCY EXPOSURES ANDHUMAN CANCER. ..........................................................................................1

Principles of epidemiology .....................................................................................................................3Table 1: A HIERARCHY OF TYPES OF EVIDENCE RELEVANT TO HUMAN HEALTHSTUDIES .............................................................................................................................................4Criteria used in assessing causality ......................................................................................................5TABLE 2: GENERALLY ACCEPTED CRITERIA FOR CAUSALITY IN EPIDEMIOLOGICALSTUDIES .............................................................................................................................................6

Methods used in this review...................................................................................................................8

Epidemiological studies of cancer up to 1999.......................................................................................8

Epidemiological studies of cancer published since 1999. ....................................................................9Further study of cancer in relationship to radio and television transmitters.........................................9Case-control study of brain cancer in Israel. ........................................................................................9Cohort study of Canadian police officers...........................................................................................10Cohort study of mortality of US Motorola employees .......................................................................10Table 3 part 1: Radiofrequency emissions and cancers: recent cohort studies of military and civilianoccupational groups and cell phone subscribers ................................................................................12Table 3 part 2: Radiofrequency emissions and cancers: recent cohort studies of military and civilianoccupational groups and cell phone subscribers, continued...............................................................13Further follow up of US Navy cohort. ...............................................................................................14Cohort study of plastic-ware manufacturing workers exposed to radiofrequency sealers. ................15Other occupational studies .................................................................................................................15Study of occupational exposures of parents related to neuroblastoma in offspring ...........................16

Studies of cancer in association with the use of cellular telephones .................................................16Overall mortality of cell phone users .................................................................................................16General population cohort study of cellular telephone users in Denmark..........................................17

Case-control studies of brain tumors and cell phone use ..................................................................18Table 4: Case-control studies of brain cancers and use of mobile phones ........................................19Study in Sweden: Hardell et al. ..........................................................................................................20Study in the U.S. : Muscat et al..........................................................................................................20Further U.S. study: Inskip et al. .........................................................................................................21Study of brain (and salivary gland) cancers in Finland. .....................................................................22Summary analysis for brain cancer and cell phone use......................................................................22Studies of ocular melanoma and mobile phone use ...........................................................................22

Discussion ..............................................................................................................................................23Brain cancers......................................................................................................................................24Leukaemia..........................................................................................................................................24Figure 1. Summary of results (relative risk or odds ratio and 95% confidence limits) of studiesassessing leukaemia in association with radiofrequency exposures...................................................25Implications of epidemiological studies for exposure standards........................................................26

Conclusions ...........................................................................................................................................27

Reference List .......................................................................................................................................28


Principles of epidemiology

Epidemiology is “the study of the distribution and determinants of disease in humanpopulations” [3], p.1. It is the science which studies the causes of disease in human free-living populations, in contrast to studying causal mechanisms in experimental animals or cellsystems.

Very occasionally, where a particular causal agent is the only (or almost the only) cause of aspecific disease and has a very clear and strong effect, a causal relationship can be establishedon the basis of one, or only a few, well-conducted studies; examples include occupationalstudies of asbestos exposure, and the studies of those affected by radiation from the atomicbombs in Japan in 1945. Much more commonly, however, the causes of a disease areestablished by the cumulative evidence provided by a large number of different studies, ratherthan by one particular study. If an association is seen between a possible causal factor and adisease (for example, between exposure to radiofrequencies and the development of cancer) acareful evaluation of the extent and quality of the studies showing that association isnecessary, before concluding that there is likely to be a cause and effect relationship, orwhether the associations seen are more likely to be due to other factors.

Studies in human populations, unlike experimental studies in a laboratory, are limited to whatcan be done ethically and logistically in free-living human subjects. Thus the exactitude of thedata collected, and the ability to isolate the effects of one factor from those of other factors,are usually less controllable than they are in a laboratory situation. In contrast,epidemiological studies, unlike laboratory studies, are directly relevant to causation of diseasein human individuals and populations, and can assess ‘real life’ exposures, which are oftenmore complex than those used in the laboratory.

As with any science, the results of epidemiological studies, whether they show an associationor not, will often be affected by limitations of the study design or analysis. The results may beinfluenced by errors or bias in the data, the influence of other relevant factors, or by chancevariation. These all have to be assessed carefully before the study can be interpreted asshowing a cause and effect relationship, or giving good evidence against such a relationship.There are well-established principles which assist in interpreting epidemiological data.

There are several major types of study (Table 1). The strongest evidence to assess a cause andeffect relationship comes from an experimental study, in which subjects deliberately exposedto a certain factor can be compared to similar subjects not exposed (for example, in trials ofimmunization, consenting subjects can be randomly allocated to receive the immunization ornot). Obviously the experimental design cannot be applied to potential hazards unless thehazard can be reduced in a controlled way. The best possible studies to assess potentialhazards are studies in which individuals are selected for a study and specific information iscollected on the suspected causal factor, the disease outcome, and (most importantly) otherrelevant factors which could be related to the disease outcome. Studies comparing healthoutcomes in two or more groups with different exposures are cohort studies (for example,comparing smokers with non-smokers). Studies comparing subjects with a particular diseaseto an unaffected control group are case-control studies (for example, studies of lung cancerpatients and unaffected persons assessing differences in past smoking). These are the methodsby which most recognized causes of human cancer have been identified (such as smoking,asbestos, ionizing radiation, and so on). Usually, a large number of such studies needs to becompleted before a consensus can be reached on a particular causal situation.

For general exposures to radiofrequencies, the studies of individuals are limited to a fewcohort studies of certain groups (military personnel, or occupational groups) whose exposurelevels are likely to be very different to the general population, and several small case-control


studies of particular types of cancer, which have poor measures of radiofrequency exposureand limited data on other relevant factors. Because of the great interest in mobile phonesafety, the studies of cell phone exposures are actually rather stronger in their design. There isa large prospective cohort study of an entire general population in Denmark assessing theincidence of all types of cancer in cell phone users. There are now several case-control studiesof brain cancer and mobile phone use which are sophisticated in their methodology and havereasonably detailed information on cell phone usage and on other relevant factors, althoughthey are limited to the information that can be gained from records or retrospectivequestioning. The main limitations of the studies of cell phone use is one of time; cell phoneshave not been in widespread use for long enough for any long term effects to bedemonstrable.

Table 1: A HIERARCHY OF TYPES OF EVIDENCE RELEVANT TO HUMANHEALTH STUDIES

Level 1: Randomized trials (experimental studies)

Level 2: Cohort or case-control studies

Level 3: Other comparative studies, including ecological studies

Level 4: Case studies, descriptive studies, professional experience, etc.

In addition to the “level” of evidence, the consistency or otherwise of the results, the quality and detail of thestudies, and the relevance of the studies to the particular situation and question of interest need to beassessed in order to determine the ability of a study to demonstrate cause and effect. Derived from [4].

A third type of study is generally acknowledged as being much weaker - that is, much harderto interpret clearly in terms of cause and effect. This is the ecological study, wherepopulation groups (instead of individuals) are studied and a comparison is made of thefrequencies of disease in groups with different exposure levels. Several of the studies relevantto the radiofrequency exposure issue fall into this category, for example, the studies ofcancers in relationship to TV or radio transmitters in the UK and in Australia. This type ofstudy is rarely regarded as definitive. It should lead, however, to more definitive studies ofthe cohort and case-control type, which are based on observations of selected individuals.

All these types of studies are comparative studies, with control groups, of the exposure in freeliving human subjects. In general, studies of humans which lack an appropriate control group,such as clinical series, are weaker. Studies which are based on a pre-suspected group or‘cluster’ of cases of disease have particular weaknesses. They are generally regarded only aspreliminary observations which have to be re-assessed by one of the study types describedabove. Animal and in vitro experimental evidence is often of high internal validity, but thereare usually substantial questions about its relevance to intact humans. Where epidemiologicalevidence is unclear or is lacking, experimental evidence may be the main way to judgewhether the potential exists for health effects of a certain exposure. Elsewhere in this report,aspects of radiofrequency exposures such as tissue penetration and photon energy are


discussed; these are relevant to judging the possibility of health effects from the knowncharacteristics of the exposure.

Criteria used in assessing causality

Epidemiological studies usually involve measuring the association between an exposure (suchas radiofrequencies) and an outcome (such as cancer). Usually, the results are expressed interms of relative risk; for example, a relative risk of 1.8 means that the rate of cancer is 1.8times as high, or 80% increased, in the exposed group. This measures the association; butfurther assessment is needed to conclude that it is due to causation.

Criteria have been developed which are generally accepted both for the assessment of anindividual study, and of the totality of evidence derived from a number of studies (Table 2).The first process in assessing whether a particular study gives a valid cause and effectassessment is to see if alternative, non-causal, explanations can be reasonably excluded. (Thislogic in fact applies to all science, including laboratory studies). These non-causal factors are[4]:

1. Observation bias in the observations which have been made. For example, in a studybased on an interview recall of exposures, people affected with cancer may be moreready to recall and report a previous exposure (such as an accidental exposure toradiofrequency sources) than people who have not had cancer. If this bias occurs,even if there is no true relationship between the exposure and cancer, the study willshow an (incorrect) positive association (which may be statistically significant).

2. The effect of other relevant factors, sometimes known by the term “confounding”.For example, if users of mobile phones smoked more than other people, an associationbetween mobile phone use and lung cancer would result.

3. Apparent associations may be due to chance variation. This is assessed by statisticalmethods, which should be applied once observation bias and confounding have beendealt with.

These same influences have to be assessed in the interpretation of studies which show noassociation, that is, the results give similar rates of disease in exposed and unexposedsubjects. A confounding factor can disguise a true association: for instance, an increased riskdue to an occupational hazard may be disguised by the generally better health of peopleselected for employment: the ‘healthy worker effect’; this bias can be dealt with by comparingthe workers exposed to the suspected hazard with other workers in the same general situation,but not exposed to that hazard. The size of the study is important; small studies can only showeffects which are large. Another problem is the specification of the exposure; for example, ifthe hazardous effect is restricted to a particular wavelength range, a study in which exposureis defined as any radiofrequency exposure will have reduced ability to detect an effect.However, these studies do have some advantages. They can assess large numbers of subjectswithout the biases of requiring consent and depending on individual recall, and unlike forexample some occupational studies they have the potential to separate out exposure toradiofrequencies as distinct from exposure to ELF fields.


TABLE 2: GENERALLY ACCEPTED CRITERIA FOR CAUSALITY INEPIDEMIOLOGICAL STUDIES

1 Non-causal explanations to be excluded:

Bias in observations

Effects of other factors (confounding)

Chance variation

2 Features expected if causation applies:

Correct time relationship

Strong association

Dose-response effect

Specificity of effect, by causal factor and outcome

Plausibility: biological mechanism; analogy

Coherence with distribution of causal factor and outcome

Consistency of results, both within and among studies

From: [5], [4], and others.

After excluding non-causal explanations, the next process is to look for specific featureswhich would be expected if a biological cause and effect relationship applies. Such criteriaare often called the Bradford Hill criteria [5]; they are used by many multidisciplinaryinternational groups in the assessment of cause and effect in health studies. They include anappropriate time relationship, which is logically essential: a reasonable strength of therelationship; and a dose-response relationship. These are helpful mainly in making it easier todetect, and allow for, observation bias and confounding; for example, if a study reports asmall relative risk, for example less than 1.5, it may be difficult to ensure that such biases canbe excluded. Criteria of specificity of effect, plausibility, and coherence are sometimes useful.

Consistency is the most important criterion and is assessed in two ways: as consistency withina study, and, the most important criterion of all, consistency among various studies. In thegreat majority of situations the development of a consensus amongst the scientific communityon whether a particular agent causes (for example) cancer is based on a consideration of theconsistency of evidence from a large number of studies of different designs and in differentpopulations, which overall produce a substantial body of evidence. This requires that allrelevant studies be considered. This is made more difficult by the effects of publication bias,that is, not all studies have an equal chance of being published; studies which have negativeresults, are in accord with conventional assumptions and therefore are not news worthy, or incontrast give unexpected results which are not accepted by reviewers, may have difficultybeing published.

The main result is usually expressed as a measure of association, the relative risk, which is therisk of disease in people exposed to the factor under consideration, as a ratio of the risk inthose people not exposed. For example, a relative risk of 1.5 means that the study isestimating that people exposed to the factor under consideration have 1.5 times the diseaserisk of those not exposed; this could also be expressed as a 50% increase; a relative risk of 1means that there is no association, and a relative risk of less than one equates to a protective


effect. This result (the relative risk) is the size of the association provided by the study. Theaccuracy or statistical precision of that estimate is shown by confidence limits. These areusually expressed as “95% confidence limits”, meaning that in statistical terms there is a 95%probability (95 chances in 100) that the true result will be within that range. A small study,because it is imprecise, will have wide confidence limits. A larger study will have narrowerconfidence limits, that is, the estimate is much more precise. If the confidence limits includethe value of 1.0, the study is said to be “not statistically significant”, in other words, it is stillcompatible (at the selected probability value) with no association and a relative risk of 1.0. Ifthe confidence limits are all higher than 1.0, it means that the study shows an increased risk ora positive association which in technical terms is “statistically significant”. The choice of95% is arbitrary; the 99% confidence interval will be wider, the 90% narrower.

If radiofrequencies do cause a disease like cancer, a good study will show this by giving arelative risk greater than one. If the study is large enough, the 95% confidence limits will alsobe above one: a hypothetical example would be a relative risk of 1.5, with limits of 1.2 to 1.8.This result would be described as showing an increased risk, which is statistically significant.Even this result does not mean that a cause and effect relationship has been shown: thatdepends on whether the study is free of biases in the data used, and on whether otherexplanations such as the effects of related factors have been taken into account.

If, on the other hand, radiofrequencies do not cause (or prevent) the disease, a good study willgive a relative risk close to one. However, it is unlikely that the relative risk will be preciselyone, because of the impossibility of collecting perfectly accurate data and having noinfluences of other factors, and also because of the effects of chance variation. The 95%confidence limits will usually include the value of 1.0: a hypothetical example would be arelative risk of 1.1, with limits of 0.8 to 1.3. This result would be described as showing noincreased risk (or only a small increased risk), which is not statistically significant. A studywith a relative risk of for example 3.0 with confidence limits of 0.5 to 18.0 is howeverdifficult to interpret as it gives a non-significant result, but shows an association;fundamentally, the study is very imprecise as it is too small.

The reported relative risk and its confidence limits depend on the association seen, the size ofthe study and the statistical methods used. They do not assess whether the observations havebeen collected without bias, or whether the association is due to factors other than the onesuspected, except where these have been dealt with in the study design or analysis. Theseissues have to be addressed by a careful review of the study.

It is impossible to prove, with absolute certainty, the absence of an effect. To prove withcertainty that radiofrequency energy, or any other aspect of the human environment, iscompletely safe is impossible; as to do so requires proof of the absence of any associationbetween exposure to radiofrequencies and any one of an infinite number of health outcomes.This logical difficulty is expressed in the general approach of epidemiology, and science ingeneral, which accepts as "fact" not something which has been proven with absolute certainty,but as the best current explanation of the available results of scientific studies. Scientificstudies are designed not to give “proof”, but are designed to disapprove or “falsify” thecurrent hypothesis or accepted viewpoint on an issue. If well performed scientific studies ofstrong design are carried out and fail to disprove the hypothesis, the hypothesis becomesstronger, that is gains more validity and is more likely to be true, but it never reaches the pointof being “proven” with absolute certainty.

If the balance of the available evidence overall is that health effects have not beendemonstrated, despite some studies of reasonable quality having been done, then thelikelihood that radiofrequency exposures are safe is increased. The evidence pointing to safety


may well be sufficient so that the community will accept the evidence as sufficient to allownormal activities based on the assumption of safety.

It follows from this that a claim that health effects may occur, even if they have not beenshown in any study, will always be true. But because it is always true, it is not very helpful.The claim that health effects may exist is of no value unless it is based on some evidenceeither of the existence of such effects, or of other scientific evidence which make such effectslikely, rather than just conjecture or assumption.

Methods used in this review

As noted above, this paper is based on two previously published works, and repeats somematerial presented there [1,2]. A number of studies have appeared since those reviews and areincluded here. This review is based on published reports on studies of radiofrequencyemissions and cancer, identified by literature searches using Medline from 1988 to April2002, supplemented by references found in previous reports and other studies. No review canbe complete. Some earlier studies may not have been included, and very recent work may alsobe missing, as published work is often not indexed on these databases until several weeks ormonths after publication. The review is restricted to radiofrequencies, rather than other partsof the electromagnetic spectrum. It does not include individual case reports, studies withoutany comparison group, or studies based only on large routinely collected data sets.

Studies published up to 1999 are reviewed in detail in Elwood [1], and will be brieflysummarized here. Epidemiological studies relating radiofrequency exposures and cancershave also been reviewed in the reports by ICNIRP [6], the Royal Society of Canada [7], andthe U.K. committee chaired by Stewart [8], and in publications by Bergqvist [9] amongstothers.

Epidemiological studies of cancer up to 1999

The studies fall into five groups: studies of clusters of cases; studies of general populationsexposed to television, radio and similar sources; studies of occupational groups; case controlstudies; and studies of users of cell phones. Cluster studies are inherently difficult to interpretbecause of the impossibility of assessing the effects of chance variation if the study isperformed after a cluster has been identified in an anecdotal way. Cluster studies should beregarded as raising a hypothesis, which can then be tested in further studies. This has beendone in regard to the Sutton Coldfield radio and transmitter in the United Kingdom, where acluster of leukaemias and lymphomas in adults living close to the transmitter was noted,although the authors correctly conclude that no causal inference can be drawn from a clusterinvestigation alone [10]. In response to this, these authors carried out studies of thedistribution of other types of cancer around the Sutton Coldfield transmitter, and studies of alltypes of cancer around 20 other transmitters in the United Kingdom, in an appropriatehypothesis testing investigation [11]. In general this showed negative results, although a weaktrend towards a decrease in rates of adult leukaemia with increasing distance from thetransmitter was seen, of borderline statistical significance. The trend was inconsistent in thatthere was no excess risk living closest to the transmitter. The authors suggested that if thisreflected a true association, a simple radial decline exposure model was not sufficient toexplain it, and regarded their studies as giving only weak support to the previous cluster basedhypothesis.


In a study in Sydney, Australia [12], Hocking et al. showed increased incidence and mortalityrates of childhood leukaemia in the aggregate of three local authority areas close to a VHF-TV transmitter, compared to a number of areas further away. A further analysis by individuallocal government area showed that the excess applied only to one of the three inner areas[13]. An earlier study of childhood cancer in San Francisco showed no geographicalassociation with a transmitter described as a microwave tower [14].

There have been several studies of occupational groups. A study in the Polish military showedsubstantial excesses of total cancer and of several sub-types of cancer [15], but questions havebeen raised which suggest a severe bias in the exposure information in that study [1,8,9], andthe results are inconsistent with those of other studies. An study based on US Navyservicemen showed no clear increase in cancer in personnel likely to have had substantialexposure to radar, although the control group were also likely to have been exposed to someextent [16]; this study has recently been updated and is discussed later. Studies of USamateur radio operators showed an excess in one of nine types of leukaemia assessed,although they may have had other types of exposure [17]. A study of female radio andtelegraph operators working at sea showed an excess of breast cancer and uterine cancer, andagain the influence of other confounding factors may be relevant [18]. A detailed study ofelectrical workers in Quebec and France showed an excess of lung cancer, but their exposureswere not primarily to radiofrequencies [19].

There have been several case control studies of particular types of cancer, in whichradiofrequencies have been one of usually a large number of potential exposure factors whichhave been addressed. One study showed an association between likely radiofrequencyexposures and brain cancers in US Air Force personnel [20]. A study in US civilians showedan excess only for the combination of radiofrequency exposures and other electrical orelectronic job exposures, but not with radiofrequency alone [21]. Other studies show excesseswhich are inconsistent in terms of the method of collecting the information, or are non-significant or open to problems of multiple testing [22-25].

Epidemiological studies of cancer published since 1999.

Further study of cancer in relationship to radio and television transmitters

A further study on cancer incidence in residents living close to the Sutton Coldfieldtransmitter in England [26] was carried out using cancer data for the years 1987-94, and thesame methods as in the earlier studies. The only site showing a marginally significant declinewith distance was leukaemia in male children, based on 15 cases including only one withintwo kilometers distance. There were small increases in risk in several types of adultleukaemia, but no significant declines in risk with distance. The findings of the originalSutton Coldfield study were not replicated.

Case-control study of brain cancer in Israel.

In this study [27], 139 patients with primary brain tumors in Israel from 1987 to 1991 werecompared to controls in terms of lifetime occupational history, assessing many occupationalcategories. Amongst several categories, ‘electric and electronics manufacture, andcommunication’ is given, with 8 cases only, and no significant increased risk. For malignantbrain tumors, based on only 4 cases, the odds ratio was 2.2 (95% confidence interval 0.5 to9.3). Another breakdown separating out ‘telephone and radio operators and electricians’ give


a risk of 1.2 for all brain tumors based on three cases (95% confidence interval 0.3 to 5.2).This small study is basically uninformative for radiofrequencies.

Cohort study of Canadian police officers

This study [28] does not include any data on radiofrequency exposure, but is relevant to anearlier cluster study of testicular cancer in police officers [29]. In Ontario, for 20,601 maleofficers, the overall cancer incidence ratio, compared to the general population, was 0.9 (90%confidence interval 0.83 – 0.98); there was a reduced rate of lung cancer (0.66), and anincreased rate of melanoma (1.45, 90% limits 1.10-1.88). The rate of testicular cancer wasnon-significantly increased (relative risk 1.3, 90% limits 0.89 to 1.84), based on 23 cases.There was no information on the use of radar equipment.

Cohort study of mortality of US Motorola employees

A cohort study of mortality has been conducted by Morgan et al. [30] of all US Motorolaemployees with at least six months employment at any time between 1 January 1976 and 31December 1996, with follow up to 31 December 1996. This study included 195,775 workers,of whom 44% were women, and of whom 6,296 died during the follow up period.

Likely radiofrequency (RF) exposures from job positions were based on the business sector,work site, job description, and calendar period; each of 9,724 job titles were classified intoone of four exposure groups in terms of likely RF exposure, described as background, low,moderate and high. RF exposure sources were classified into different groups in terms ofpower, from background exposure up to 50+ W, and the relative level of likely radiofrequency exposure for the four groups defined above was given as 0, 1, 6 and 100. Examplesof classification of jobs are given; unexposed workers included administrative and supportpersonnel, low RF exposure included assemblers and operators not directly involved with RFtechnologies, moderate RF exposures included those who routinely used hand-held radios orworked with RF product development, and high RF exposure included technicians, testers andengineers involved with RF product testing.

In the analysis, worker’s exposure assignments were classified in three different ways: interms of their usual assignment relating to the job they held longest while at Motorola, theirpeak assignment reflecting the job with the highest expected RF level, and a cumulativeexposure score based on the summation of the RF level multiplied by the duration ofemployment for each job throughout the employee’s work history at Motorola.

A comparison of the mortality of the workforce with the mortality rates expected for a generalUS population, showed a mortality ratio for all causes of 0.66, and for all cancers of 0.78,both significantly reduced (Table 3). This is characteristic of the “healthy worker effect”. Of60 specific causes of death assessed, the highest standardized mortality ratio (SMR) was 1.28,and only five of the 60 were greater than one. For all employees, SMR’s for cancers of thelymphatic / haemopoetic system, and also those of the central nervous system, were bothsignificantly reduced from the expected rates, with SMR’s of 0.77 (95% confidence limits0.67 – 0.89) and 0.60 (limits 0.45 – 0.78) respectively. SMR analyses were also carried outfor the 24,621 subjects who were classified as moderate to high RF exposure by peakexposure classification, which showed somewhat lower SMRs for cancers of the centralnervous system and brain cancer (SMR 0.53, limits 0.21 – 1.09), and for all lymphomas andleukaemias (SMR 0.54, limits 0.33 – 0.83).


The more powerful analyses are the comparisons within the Motorola employees, comparingthose with higher radiofrequency exposures with the lower exposed or unexposed categories.Comparisons were based on each of the usual exposure and the peak exposure classifications,comparing the categories of high, moderate, and low exposures to the 'no exposure' group.Results are also presented looking at duration of exposure, latency (that is allowing for a lagtime between the first time of exposure and death), and looking at men and women separately.

Detailed analyses are presented for cancers of the brain, all lymphatic and haemopoeticcancers, leukaemia, non-Hodgkin’s lymphoma, and Hodgkin’s disease. None of the resultssuggested any increased risk. The relative risk for the high exposure category, based on usualexposure, for brain cancer was 1.07 (95% confidence limits 0.32 to 2.66), for lymphatic andhaemopoetic cancers was 0.70 (limits 0.27 to 1.47), for leukaemia was 0.99 (limits 0.39 to2.09), and for non-Hodgkin’s lymphoma was 0.58 (limits 0.12-1.74). For Hodgkin’s diseasethere were no cases in the highest exposure category, but for those in the high and moderateexposure categories for usual exposure the relative risk was 1.37 (limits 0.36-3.94) based onthree cases. There was no excess risk comparing those above the median exposure with thosewith no exposure (relative risk 0.95).

The authors point out that this study is limited by the qualitative job exposure matrix (ratherthan the ideal of having actual exposure measurements on each subject). It is also limited bythe relatively young age of the cohort, with the result that the numbers of deaths from specificcauses are small, despite the large size of the occupational group. They conclude that “Thelack of elevated mortality risk for brain cancers and all lymphatic/haemopoetic cancerscombined suggests that occupational RF exposure, at the frequencies and field levelsexperienced within this cohort, are not associated with an increased risk for these diseases”(p. 124). They state “the occupational RF levels amongst Motorola workers are lower thanmilitary and plastics manufacturing workers” (p.126). They conclude that their findings arenot compatible with excess risks of 3 or greater for brain cancers, lymphomas or leukaemias,and note “We did not observe indications of excess relative risk, but we cannot rule out thepossibility of potential effects in the range of 1.5-2.0 relative risk” (p.126).

These results do not suggest any general increased mortality risk, and show no evidence of anincrease in any specific cancer, although a small increase (or decrease) cannot be excluded.The exposure information is very limited; the likely exposures of the various groups ofworkers are not defined and no estimates of their exposures are given in the paper. Thenegative results of this study would be of considerable value if it were known that some of thegroups had substantial RF exposure, but if their level of exposure were well within currentstandards, the negative result is less informative. If an effect were specific to a particular typeof radiofrequency exposure, the study would have less ability to detect it.

Despite these limitations, which are shared by all other studies yet done, this is a strong study.It shows no association of occupational RF exposure with brain cancers or with lymphatic /haemopoetic cancers, and although numbers are small, the other results do not suggest anysubstantial increased risk. This is a more detailed study than most of the other occupationalstudies, and includes both men and women. Even a study of this size cannot confidentlyexclude a modest increased risk of specific cancers which occur in relatively small numbers,although it can confidently exclude increases in total mortality or from major causes such asall cancer. The study is sufficiently powerful to reasonably exclude a substantial excess ofleukaemia or lymphoma in about ten years from radiofrequency exposure in these workers.This time interval is not long enough to exclude an incidence effect, but it does providesubstantial evidence against a short-term promotion effect, such as has been suggested bysome animal experiments.


Table 3 part 1: Radiofrequency emissions and cancers: recent cohort studies of military and civilian occupationalgroups and cell phone subscribers

Cells which are blank are those for which no information was given in the study

Study, reference US Navy[16]

US Navy furtherfollow up [31]

Danish cell phone subscribers [32] Motorola employees [30]

First author, date Robinette, 1996 Groves, 2002 Johansen, 2001 Morgan, 2000

Exposure microwave (radar) microwave (radar) Cell phone radiofrequency

Ascertainment service records service records Phone subscription records employment records

Exposed group military: male military: male General population phone subscribers all employees

Frequency MHz 800-2000 30-800Outcome data mortality Mortality Incidence mortalityExposed group for RR's hazard no. 5001+ High radar exposure Subscribers

Outcome: Men Women All employees vs.external controls

Highest exposurevs. unexposed

Total deaths, all causes 1.23 (0.98 - 1.52) 0.87 (0.83 – 0.90) 0.66 (0.64-0.67)All cancer 1.44 (0.96 - 2.07) 0.80 (0.74 – 0.87) 0.86 (0.83-0.90) 1.03 (0.95-1.13) 0.78 (0.75-0.82)

All lymphatic &hematopoetic

1.64 (0.70 - 3.25) 0.77 (0.67-0.89) 0.70 (0.27-1.47)

All leukaemia 1.48 (1.01 – 2.17) 0.97 (0.76-1.21) 1.07 (0.43-2.20) 0.82 (0.65-1.02) 0.99 (0.39-2.09)

Acute myeloid leukaemia 1.81 (0.87 – 3.78)Acute lymphatic leukaemia 0.87 (0.23 – 3.26)Acute non-lymphoidleukaemiaChronic myeloid leukaemia 1.55 (0.50 – 4.75)Chronic lymphaticleukaemia

1.08 (0.44 – 2.66)

Non-lymphocyticleukaemia

1.82 (1.05 – 3.14)

Hodgkin's disease 0.88 (0.58-1.29) 1.18 (0.24-3.43) 1.14 (0.3-1.8) 1.37 (0.36-3.94)Non- Hodgkin's lymphoma 0.93 (0.77-1.13) 1.04 (0.52-1.86) 0.58 (0.12-1.74)

All lymphoma /lymphosarcoma

0.91 (0.68 – 1.22)

Other lymphatic cancerMultiple myeloma

Brain / nervous system 0.65 (0.43 – 1.01) 0.95(0.79– 1.12) 1.03 (0.62-1.61) 0.60 (0.45-0.78) 1.07 (0.32-2.66)Brain: astrocytomaBrain: glioblastoma


Table 3 part 2: Radiofrequency emissions and cancers: recent cohort studies of military and civilian occupationalgroups and cell phone subscribers, continued

Study, reference US Navy[16]

US Navy furtherfollow up [31]

Danish cell phone subscribers [32] Motorola employees [30]

Outcome data mortality Mortality Incidence mortalityExposed group for RR's hazard no. 5001+ High radar exposure Subscribers

Outcome: Men Women All employees vs.external controls

Highest exposurevs. unexposed

Salivary gland 0.78 (0.31-1.60)Lung 0.73 (0.63 – 0.83) 0.65 (0.58-0.73) 0.87 (0.60-1.22)Larynx 0.81 (0.61-1.06) 1.24 (0.14-4.48)Larynx, lungRespiratory tract 2.20 (1.05 - 4.06) 0.8 (0.7-0.9)Other respiratoryOral cavityPharynx 0.62 (0.42-0.88) 2.43 (0.65-6.22)Oral cavity, pharynx 0.62 (0.35 – 1.08)Esophagus 0.74 (0.53-0.99) 1.53 (0.31-4.46)Stomach 0.78 (0.60-0.99) 0.45 (0.05-1.61) 0.9 (0.7-1.2)Esophagus, stomachColon 0.95 (0.82-1.10) 0.97 (0.61-1.47)Rectum 1.00 (0.84-1.18) 1.13 (0.58-1.98)ColorectalAll digestive organs 0.78 (0.15 - 2.31) 0.8 (0.7-0.9)Liver 0.60 (0.36-0.96) 1.00 (0.11-3.61)Pancreas 0.82 (0.62-1.07) 0.73 (0.23-1.70)Liver, pancreasOther gastro-intestinal 1.12

Prostate 0.91 (0.77-1.06)Kidney 1.03 (0.84-1.24) 1.04 (0.42-2.15)Kidney, prostateBladder 0.97 (0.85-1.11) 1.34 (0.69-2.33)Urinary tractTestis 1.30 (0.35 – 4.89) 1.12 (0.97-1.30)

Female breast 0.99 (0.32-2.32) 1.08 (0.91-1.26) 0.8 (0.7-0.9)Cervix 1.34 (0.95-1.85)Uterus (endometrium) 1.02 (0.60-1.61)Ovary 1.09 (0.70-1.62)All female genital 1.0 (0.8-1.3)

BoneSkin 0.92 (0.85-1.00) 1.00 (0.79-1.24)Melanoma 0.86 (0.72-1.02) 0.80 (0.49-1.22) 0.8 (0.6-1.1)Eye (mainly melanoma) 0.65 (0.28-1.27)Thyroid 1.01 (0.54-1.72) 0.92 (0.25-2.35) 1.1Other cancers 1.17 (0.50 - 2.32) 0.70 (0.62-0.80) 1.07 (0.75-1.50)


Further follow up of US Navy cohort.

A further follow up has been published by Groves et al. [31] of the cohort study of UnitedStates naval personnel, in which 20,000 men with maximum opportunity for exposure to radaremissions were identified and compared to 20,000 subjects with a lower potential forexposure [16]. All had graduated from U.S. Navy technical schools in 1950-54, and hadserved on U.S. navy ships at the time of the Korean war. As in the original study, there werethree job classifications regarded as likely to have low exposure: radioman, radarman, andaviation electrician’s mate. These are likely to have had only casual exposure to radar withexposures well below 1 mW/cm2. The highly exposed groups were electronics technicians,aviation electronics technicians, and fire control technicians. Fire control and electronicstechnicians repaired and maintained radar related to gun fire control and search capacities,with the potential for exposures exceeding 100 mW/cm2, even though usual exposures werewell below 1 mW/cm2. The occupational standard at that time was 10 mW/cm2. The attemptin the earlier study to quantify potential radar exposure by means of hazard number, based onthe type and power of radar units and the number of months spent on ships and in airsquadrons, could not be extended in the later study because the records were not readilyavailable. All the subjects would have been exposed to 60 Hz fields from electricalequipment, and electrician’s mates, who repaired wiring, could have had high levels ofexposure to such ELF fields. In addition the three groups of technicians with high potentialradar exposure, and aviation electrician’s mates, were involved in the maintenance and repairof electrical equipment and would likely have had considerable exposure to solder fumes,chlorinated solvents, and oils and greases. Aviation electronics technicians and aviationelectrician’s mates were involved in the maintenance and use of aircraft radar and electricequipment, and could have had flying duties.

The cohorts of navy personnel assembled in the original study were linked to mortality dataup to 1997 by Department of Veterans Affairs, social security and death records. The finalcohort was 40,581 men, of whom 20,021 were in the three categories regarded as having highpotential exposure to radar. In the entire cohort there were 8393 deaths, 21 percent, andcompared to the US white male death rates the overall mortality ratio (SMR) was significantlyreduced at 0.74. Death rates were similarly reduced for most diseases and most cancers. Thisrepresents a healthy worker (or healthy sailor) effect.

The critical comparison is between the high radar exposure group and the low radar exposuregroup, presented as relative risks which are adjusted for age and age at entry to the navycohort (Table 3). Year of graduation, year of birth, and duration of follow up from graduationonwards were also assessed but found not to be important. The men with high radar exposuregroup had a significantly reduced total mortality rate (RR 0.87), and significantly reduceddeaths rates from cancer (RR 0.80), and within cancers, significantly reduced rates of lungand related cancers (RR 0.73). They also showed significantly reduced risks of diabetes, heartdisease, and non-cancerous lung disease. The increases were seen predominantly in deathsrelated to war injuries (11 in the high exposure group, and 2 in the low exposure group), andair accidents (24 in the low exposure group and 100 in the high exposure group). Withincancers there was a significant excess in all leukaemias, (RR 1.48, with confidence limits of1.01-2.17). Deaths from brain cancer were non-significantly reduced (RR 0.65). Thesefindings contrast to the findings in the first more limited follow-up study, where totalmortality, total cancer mortality, and deaths from all lymphatic and hematopoetic diseaseswas non-significantly increased, and there was a significant excess of deaths from respiratorytract cancer (RR 2.2) which was replaced in the more extensive analysis by a significantdecreased risk.

Further analysis examined each of the three high risk groups and the aviation electrician’smates, compared to the combined group of radiomen and radarmen, being a low radar


exposure group without likely flying duties. This showed that the war related deaths wereconcentrated in aviation electronics technicians, and deaths involving air transport wereconcentrated in the two aviation groups with likely flying duties. In regard to cancers,reductions in cancers were seen in each of the three high exposure groups, being significant inelectronic technicians and aviation electronics technicians. The increase in total leukaemiaswas restricted to the aviation electronics technicians, where it was significant with a relativerisk of 2.6, and the increased risk was seen in most subtypes of leukemia.

A weakness of this study is that actual exposure levels are unknown, and it compares twogroups with estimated high and low potentials of exposure, but radiofrequency exposures mayhave been substantial in both groups. While the generally lower mortality experience of thehigh radar exposure group in this study offers some evidence against there being major effectsof radar exposure, the significantly lower overall death rate is unexplained, as both groups hadhad similar military service selection criteria. There is no information available on lifestylefactors such as smoking, or on the post- navy exposures or occupations of this group. Theincreased leukaemia mortality in the high exposure group is of interest given other results onleukaemias and it stands in contrast to the overall results. It was virtually confined to aviationelectronic technicians, the leukaemia relative risks in the other two high exposure groupsbeing 1.3 and 1.04. The authors comment that aviation electronics technicians may have hadmore inadvertent or accidental exposure as they dealt primarily with radar units in aircraft,and may have been more likely to get into the beam path of an operating radar than were mendealing with shipboard installations. Despite this, they regard the failure to see an excess ofleukaemia in either of the other two highly radar exposed groups as evidence against therebeing a causal link, and also caution about multiple comparisons. Their overall conclusionwas “Radar exposure had very little effect on mortality in this cohort of US Navy veterans”(p.810).

Cohort study of plastic-ware manufacturing workers exposed to radiofrequency sealers.

This study [33] was based on a plastic-ware manufacturing plant in Grosseto, Italy, andcompares operators of radiofrequency sealers (302 women and 4 men), other laborers, andwhite-collar workers. A survey carried out in the 1980’s showed that the recommended

exposure limit of 10 W/m2 equivalent power density was frequently exceeded in this factory

mainly due to high electric field strengths. These workers were also exposed to solvents, andto vinyl chloride monomer, an established carcinogenic agent. The analysis, restricted towomen, is based on only 9 observed deaths amongst radiofrequency sealer operators,compared to 6.3 expected. The excesses were seen in accidents and violence (2 observed, 0.8expected, standardized mortality ratio, SMR, 2.4) and malignant neoplasms, (6 observed, 3expected, SMR 2.0, 95% confidence interval 0.7 to 4.3). The authors’ conclusion is "Thisstudy raises interest in a possible association between exposure to RF radiation and cancerrisk. However, the study power was very small, and the possible confounding effects ofexposure to solvents and vinyl chloride monomer could not be ruled out". The results cannotbe interpreted clearly without further relevant studies.

Other occupational studies

Many studies, of all types, have reported on occupations likely to be exposed mainly to ELFlow frequency fields, but who may have had exposure to radiofrequencies in addition. It isdifficult in these studies to assess radiofrequency exposure specifically, although some resultswould seem relevant. In a cohort study in Sweden with 19 years follow up [34] leukaemia and


brain tumors were assessed in ‘electrically related occupations’, which included categories ofradio and TV assemblers and repairmen. In the whole group the relative risk of leukaemia was1.1, and of brain tumors 1.2, both non-significant, but in the assemblers and repairmen in theradio or TV industry the risk of brain tumors was increased, RR 2.9 (95% limits 0.3-2.5),based on seven cases. A very large number of comparisons were made in this study. A studybased on occupation as recorded in the New Zealand cancer registry [35] showed an increasedrisk of leukaemia (odds ratio 1.62) which was highest in radio and TV repairers (odds ratio7.9, 95% limits 2.2 to 28.1). There are many other similar studies, but they are difficult tointerpret in terms of radiofrequency effects.

Study of occupational exposures of parents related to neuroblastoma in offspring

In the first detailed study of radiofrequency exposures of parents in relationship to cancer intheir children, 538 subjects diagnosed with neuroblastoma at ages up to 19 years between1992 and 1994 were identified from 139 hospitals in the U.S. and Canada [36]. Telephoneinterviews were conducted with the mother, and the father if available, and with controlsidentified by random digit dialing. Exposures were assessed both by self report and by anassessment by an industrial hygienist based on the description of the occupations held.Exposures were divided into ELF fields, radiofrequency fields, and ionizing radiation.Information on maternal exposures was obtained from 537 mothers of children withneuroblastoma and 503 control mothers. There were no significant associations with maternaloccupational exposures to radiofrequency sources. There were non-significant increased risksseen with occupations for which the industrial hygienist classified RF exposure as ‘probable’(odds ratio 2.8, 95% limits 0.9 to 8.7), and for cell phone exposure (odds ratio 2.1, 95% limits0.4 to 11.0, based on only 7 exposed mothers). Data on 405 fathers of affected children and302 controls also showed no significant associations with occupational exposures to RF. Dataare also presented on individual types of equipment. There were odds ratios over 2.0 seen forsome exposures, such as radio transmitters of under seven watts power (but not higher powertransmitters), radar, and high frequency arc welding machines. But these were all based onsmall numbers and were not statistically significant; there is also the issue of multipleexposures being assessed. Low frequency field exposure was assessed in more detail, withoutany clear associations being seen.

Studies of cancer in association with the use of cellular telephones

Overall mortality of cell phone users

In the U.S., a cohort of over 255,000 persons who were customers of a telephone company in1993-94 in four urban areas were identified from telephone company records [37]. Of these,65% were men, and the median age was 42 years in men, 41 in women. Deaths in one year,1994, were obtained by data linkage. The object was to compare death rates for customerswith ‘portable’ phones (cell phones) with rates for customers with ‘mobile’ phones, whichhere means the older type of transportable bag phones with the antenna separate from thehand piece, on the basis that the ‘portable’ phone (the modern cell phone) will have morehead exposure to radiofrequencies. This study was published to show the methods forproposed further studies. The age-specific death rates were similar for users of the two typesof telephones. For customers with accounts at least 3 years old, the ratio of mortality rates in1994 for ‘portable’ telephone users, compared with transportable telephone users, was 0.86(90% confidence interval 0.47-1.53). The numbers of deaths due to brain tumors and


leukaemias were small, but there was no increased risk with greater use of hand held phones[38]. However, the short follow up time does not allow assessment of longer term effects.

General population cohort study of cellular telephone users in Denmark.

Johansen et al. [32] carried out a prospective cohort study in Denmark, using thecomputerized files of the two Danish operating companies. From a total of over 720,000subscribers some 200,000 corporate customers had to be excluded because information onindividuals was not available, and after further exclusions because of errors in name, address,duplications, etc. there were 420,095 cellular telephone subscribers identified, being 80.3% ofthe original list of residential subscribers. Follow up was from the date of first subscription upto December 31, 1996, and rates were compared to national rates adjusted for age, sex andcalendar period. Of the total cohort, most were men (357,000), most were aged 18 – 29 at firstsubscription, and the years of first subscription were 1982 to 1995, with 70% being in 1994-95 and 23% in 1991-93; 58% used a digital GSM system at first subscription, with theremainder having an analogue NMT system.

The standardized incidence ratios are presented by gender, and for all cancers were 0.86 (95%confidence limits 0.83 – 0.90) in men, and 1.03 (confidence limits 0.95–1.13) in women,based on 2876 and 515 cases of cancer respectively (Table 3). For men, the incidence ratios ofmost smoking related cancers were reduced, while testicular cancer was non-significantlyelevated (incidence ratio 1.12, 95% limits 0.97–1.30). For women, the variations were greateras they were based on smaller numbers, and there were no significant differences; theincidence ratio for breast cancer was 1.08 (limits 0.91–1.26). Tumors of the central nervoussystem, and leukaemia, were examined in more detail. The overall incidence ratio, both sexescombined, was 1.0 for each of these, and there were no trends apparent with latency up to 5 ormore years, with age at first subscription, and no differences seen between analogue anddigital telephones. There was no association with site of tumor within the brain, with tumorsof the temporal lobe having an incidence ratio of 0.86, frontal lobe 1.11, and parietal lobe0.48, all non-significant. There was no increase in salivary gland tumors or leukaemia.

There was no control for socioeconomic status or other covariates, and the pattern ofincidence ratios is consistent with a distribution of mobile phone use characterized by highersocioeconomic status, and as a correlate, a lower rate of smoking. The study was not able toassess intensity of use, as records on number of calls made or length of call were not useable,and the follow-up was up to 15 years, although the average period of follow-up was only 3.1years. However, it provides considerable evidence against any large increase in risk within afew years of use.

The authors comment that “Conceivably, the latency may be too brief to detect an early stageeffect or an effect on the more slowly growing brain tumors. Moreover our study maycurrently have too few heavy users to exclude with confidence a carcinogenic effect on braintissue following intensive, prolonged use of cellular telephones. On the other hand, if RFexposure is assumed to act by promoting the growth of an underlying brain lesion, then theintense recent use, as currently experienced by large numbers of our cohort, might be of moreimportance than latency or long-term use considerations.” In an accompanying editorial,Park notes [32] that the study is strong because of its population base and size, and commentsthat the evidence suggesting that radiofrequencies could have a carcinogenic effect is veryslim, making an analogy with previous concerns about low frequency fields which wereallayed by a high quality case control study.


In correspondence, Hocking [39] emphasized the exclusion of corporate customers and thelack of information on intensity of use, and also suggested that the increased risk of testicularcancer (relative risk = 1.12) could be related to exposure by carrying a phone on the belt. Theauthors responded that corporate customers may be an important high exposure group, butany bias produced by their exclusion would be small, and it is unlikely there would be anysubstantial radiofrequency exposure from cellular phones worn on a belt or in a pocket [40].Godward et al. [41] questioned the use of the whole population reference group rather than anunexposed group, which could lead to an underestimate of effect, and also emphasized thelimited data on exposure intensity, dose response effects, and socio-economic status, and thelimited length of follow-up. The authors responded that the underestimation of effect by thechoice of control group would be very small, and agreed with the limitations in terms oflength of follow-up. They argue that confounding by socio-economic status would be unlikelyto be a major issue in Denmark, although linkage to such information is planned in the future[42] and point out that the study had sufficient power to rule out moderate or high risks withina short follow-up period [40]. Hardell and Mild [43] asked for specific analyses for tumors ofthe temporal and occipital lobe, after a 5 year latency period, distinguishing analogue fromdigital phones. The authors comment [42] that even in this large study of 420,000 subjects, ananalysis stratified by subsite, latency period and type of telephone would have insufficientnumbers to be informative.

Case-control studies of brain tumors and cell phone use

There have been several such studies published (Table 4).


Table 4: Case-control studies of brain cancers and use ofmobile phones

Firstauthor

Reference Year Numbers of: Overall association: odds ratio, 95% limits

Association for high exposure: odds ratio, 95% limits

Site within brain

cases Controls OR lower upper Exposure OR lower upper Site OR lower upper

Hardell [44], [45] 1999,2000

209 425 0.98 0.69 1.41 latency>10yrs;use >968 hrs

1.06 0.33 3.4 Temporaland occipital;same side use

2.6 1.0 6.7

Muscat [46] 2000 469 422 0.85 0.6 1.2 >10 hrs/month 0.7 0.3 1.4 Temporal 0.9 0.5 1.7

Inskip [47] 2001 782 799 1.0 0.6 1.5 >500 hrs 0.7 0.2 1.1 Temporal 0.8 0.5 1.4

Auvinen [48] 2002 398 1990 1.3 0.9 1.8 >2 yr. 1.5 0.9 2.5

Combined analysis by fixed effect meta-analysismethod:

1.02 0.85 1.23 1.08 0.75 1.57no significantheterogeneity.


Study in Sweden: Hardell et al.

In this study [44,45], 209 subjects with pathologically verified brain tumors living in twoareas in 1994-96 were included, with 425 controls from the Swedish Population Register,matched for sex, age and study region. Exposure was assessed by questionnairessupplemented by telephone interviews. Of 262 cases identified, 209 (80%) are in the study,but only 198 (76%) are included in the detailed tables. Ever-use of a cellular telephoneshowed no association, (odds ratio 0.98, 95% confidence interval 0.69 – 1.41); Table 4. Dose-response assessment and use of different tumor induction periods gave no associations, evenat the highest level of use and latency period (over 968 hours of use, and over 10 years). Innormal use of a hand held cell phone, the maximum energy absorption is in areas of thetemporal lobe, and also the adjacent occipital, frontal and parietal lobes [49]. An analysisrestricted to tumors occurring in the temporal or occipital lobe of the brain, and on the sameside as the reported use of the cellular phone gave non-significantly increased risks, based on8 and 5 cases respectively. This comparison comes from a table involving 26 comparisons.

In a later paper based on the same study [45] the authors show a marginally significantincreased risk for tumors in the temporal, occipital, or temporoparietal regions, where cellphone use was on the same side: relative risk 2.62 (95% confidence limits 1.02 – 6.71) aftermultivariate analysis. They also show several other factors as showing statistically significantassociations: occupation as a physician, in laboratory work, or in the chemical industry, andexposure to diagnostic radiology of the head and neck region. The authors state that anincreased risk was found only for use of the analogue system, but they had few data on digitalGSM phones.

In this study, the main comparison assessing use of cell phones in regard to brain tumorsgives no association, with virtually identical rates of cell phone use in brain tumor patientsand in controls. The result showing an elevated risk for tumors near the ear on the same sideas the use of the telephone is based on one of many comparisons, and on only 13 of the 209subjects. While of interest, this result needs to be regarded as a reasonable hypothesis, whichneeds assessment by further studies. The Stewart Report [8] concludes that the results of thisstudy could easily have occurred by chance. Others have compared the numbers of casesincluded with the numbers registered in the stated areas and times, finding a large discrepancythat suggests the study had incomplete ascertainment of cases [50]; a reply from the authors[51] does not clearly deal with the issue.

Study in the U.S. : Muscat et al.

In a larger study, Muscat et al. [46] compared patients with primary brain cancer identified atfive referral centers in the U.S. to inpatient controls in the same hospital with either benignconditions or cancer, excluding lymphoma or leukaemia. Controls were matched by hospital,age, sex, race, and month of admission. There were 469 cases, being 82% of those approachedfor interview, but 70% of all those eligible. The response rate in the controls was 90%.

The primary question was whether patients had ever used a hand-held cellular telephone on aregular basis, defined as having had a subscription to a cellular telephone service. Relativerisks by the number of years of use (up to 4 or more), number of hours per month (up to 10 ormore), and number of cumulative hours (up to 480 or more), showed no excess risks and nosignificant trends. The relative risk in the highest exposure groups by each measure ofintensity of exposure was 0.7; and a non-parametric regression curve showed that most highusage groups had a slightly reduced relative risk.


In this study, 80% of cell phones used were analogue. The analysis was done separately fordifferent locations of tumors, each compared to all controls with multivariate analysis forconfounders, and showed no significant associations with any site, with the relative risk foroccipital lobe tumors being 0.8, temporal lobe 0.9, parietal lobe 0.8, and frontal lobe 1.1. Sub-division by pathological type showed no significant associations, although the risk forneuroepitheliomatous tumors was 2.1 (95% limits 0.9 - 4.7), based on 35 cases. Informationon the laterality of cellular telephone use was obtained for 56 of the 66 cases with braincancer. Of 41 cases who specified laterality and had a localized tumor, 25 reported ipsilateralrelationships, and 15 contralateral relationships, (P = 0.06). Of the fourteen cases withtemporal lobe cancer that used cellular telephones, 5 were ipsilateral and 9 contralateral (P =0.33).

In summary this study shows no excess risks, even for the specific locations of tumors whichwere highlighted in the previous case control study [44,45]. The interviews were carried outby "health professionals or health professionals in training", which is often not ideal, asdedicated interviewers employed for the purpose are usually more reliable. The interviewslasted about half an hour, which suggests they were fairly superficial. The study covers arestricted time period. However, despite these limitations it is a useful study. The authors'conclusions are "Our data suggest that use of handheld cellular telephones is not associatedwith risk of brain cancer, but further studies are needed to account for longer inductionperiods, especially for slow-growing tumors with neuronal features”.

Further U.S. study: Inskip et al.

A larger U.S. case control study involved 782 patients and 799 hospital controls with non-malignant conditions [47]. Patients had a primary brain cancer diagnosed between 1994 and1998, and 92% of eligible patients agreed to participate, along with 86% of controls, whowere matched by hospital, age, sex, race or ethnic group, and proximity of their residence tothe hospital. A computer assisted personal interview was carried out by a research nurse,using proxy interviews for subjects who were too ill or functionally impaired, which appliedto between 3 and 16% of different categories of cases, and 3% of controls.

The relative risk associated with use of a cellular telephone for more than 100 hours was 1.0(95% limits 0.6 - 1.5) for all brain cancers, and 0.9 for glioma, 1.4 for acoustic neuroma, and0.7 for meningioma; all non-significant. There was no evidence that the risks were higherwith use of 1 hour or more per day, or use for 5 or more years. There was no associationbetween laterality of telephone use and laterality of brain tumor, no increased risk fortemporal, parietal or frontal lobe tumors, and no increased risk with specific subtypes oftumors. In contrast to the study by Muscat et al. the risk for neuroepitheliomatous tumors was0.5 (95% limits 0.1 – 2.0), based on 25 cases. The authors conclude that "These data do notsupport the hypothesis that the recent use of hand-held cellular phones causes brain tumors,but they are not sufficient to evaluate the risks among long-term, heavy users and forpotentially long induction periods".

An accompanying editorial [52] comments that the limitations to the study are that thefindings apply to predominantly analogue phones, do not assess risks which may occur after aconsiderable latency period, and cannot confidently exclude small increases such as relativerisks less than 1.5.


Study of brain (and salivary gland) cancers in Finland.

In a case-control study using record linkage, Auvinen et al. obtained information on all 398subjects with newly diagnosed brain tumors diagnosed in 1996 was obtained from thepopulation based Finnish cancer registry, and for each five age and sex matched controls werefound from the population registry [48]. Computerized linkage for personal subscriptions tothe two cellular network operators in Finland in 1996 provided the exposure information, butno information on actual use of phones was obtained. Potential confounders of place ofresidence, socio-economic status and occupation were obtained from the population registry,without any significant differences being seen between cases and controls. There was a non-significant odds ratio of 1.3 (95% limits 0.9 to 1.8) for ever having held a cell phonesubscription. For the longest duration of subscription, over 2 years, the odds ratio was 1.5(limits 0.9 to 2.5). Fifty percent of the brain tumors were gliomas, which showed a higherand statistically significant association with analogue phone subscription (odds ratio 2.1,limits 1.3 to 3.4), and a weak increase in risk with length of subscription. Subscriptions foranalogue phones averaged 2-3 years, for digital less than one year; there were no increasesseen with digital phones. Control for the available covariates did not change the results. The32 subjects with glioma who had held a cellular phone subscription were similar to subjectswith glioma with no cell phone subscription in terms of histological subtype, lobe of thebrain, and laterality; this suggests the association does not differ by these features, but thisanalysis uses only a small subset of the subjects and is lacking in power. The authorsacknowledge that having a named phone subscription is only an indirect measure of actualuse: before 1996 there were more corporate than private subscriptions in Finland. The resultsare consistent with a small increased risk for brain cancer, but the associations are weak andcould be due to chance variation, misclassification, or uncontrolled confounding. However bybeing registry based and using record linkage rather than interviews this study does avoid theproblems of incomplete response to interviews and recall bias.

The study also included 34 subjects with salivary gland cancers, but this is too few for anymeaningful results. The odds ratio with having had a phone subscription was 1.3, with wideconfidence limits of 0.4 to 4.7.

Summary analysis for brain cancer and cell phone use

The results of the four case control studies can be summarized by a fixed effects meta-analysis method [53], which gives a weighted average result for the four studies (Table 5.6).There is no association seen with total use of cell phones, (combined odds ratio 1.02, 95%limits 0.85 to 1.23), or with the maximum use reported (combined odds ratio 1.08, 95% limits0.75 to 1.57). There is no significant heterogeneity between the studies.

Studies of ocular melanoma and mobile phone use

A case-control study of uveal melanoma assessed occupation in terms of likelyradiofrequency exposure [54]. The analysis combines two small studies; one in five differentregions of Germany with population based controls, and an additional study based on onehospital, with controls seen in the same department. The relevant question on these was “didyou use radio sets, mobile phones, or similar devices at your work place for at least severalhours per day? ”, with further details requested if the reply was ‘yes’.

There was a significant association with radio sets or mobile phones, odds ratio 3.0 (95%confidence limits 1.4 to 6.3). The association was seen both in the population based study


(odds ratio 3.2) and in the hospital based study (odds ratio 2.7). Further analysis showed thatthe elevated risk was similar in those who had been exposed for a short time or for longer.Occupations were categorized as having ‘possible’, or ‘probable or certain’, mobile phoneexposure. The risk for the ‘probable or certain’ category was 4.2 (95% confidence limits 1.2 -14.5), but this was based on only six cases. The odds ratio for those exposed to radio sets was3.3 (95% confidence limits 1.2 - 9.2) based on nine cases; these exposures included walkie-talkies in military and security services, and radio sets on ships, police cars, and similar.Control for iris and hair colour did not change the results substantially, but there was noconsideration of exposure to ultraviolet radiation, which is a risk factor for ocular melanoma[55].

The hypothesis raised by this study was assessed in the context of the Danish cohort study[56]. In this study there were eight cases of ocular cancer observed in cell phone subscriberscompared to an expectation of 13.5, giving a risk ratio 0.59, with 95% confidence limits of0.25 to 1.17, so no increase was seen. The authors also showed that the incidence of ocularmelanoma in Denmark from 1943 to 1996 had been relatively stable, with no increasefollowing the rapid growth of cell phone use since 1982. Their descriptive analysis does showthat in the last time period recorded, 1993-96, the rate was higher than in the precedingperiod, but similar rates had been recorded in some earlier periods. Dolk et al. assessed eyemelanoma around the Sutton Coldfield transmitter, finding no trend with distance, but basedon only 20 cases [10].

Discussion

The epidemiological evidence does not give clear or consistent results which indicate a causalrole of radiofrequency exposures in connection with human cancer. On the other hand, theresults cannot establish the absence of any hazard, other than to indicate that for somesituations any undetected health effects must be small [1].

Cancer has been studied most extensively than other outcomes, but although there are manyindividual associations seen, there is little overall consistency in the results. None of thesestudies give good information on individual levels of exposure. The studies of generalpopulations living near radio or television transmitters relate to radiofrequency exposureslikely to be well below currently accepted standards. The studies of military personnel andoccupational groups may include some exposures beyond general population standards.

Of the individual studies, the general population study in the UK [11] is sufficiently strong toreasonably exclude a geographical pattern with an excess of human cancers in subjects livingclose to large television and radio transmitters, although there is still a possible question inregard to adult leukaemia. The Motorola employees’ study [30] is sufficiently powerful toreasonably exclude a substantial excess of leukemia or lymphoma in about ten years fromradiofrequency exposure in these workers. This time interval is not long enough to exclude anincidence effect, but it does provide substantial evidence against a short-term promotioneffect, such as has been suggested by some animal experiments. However it has a greatweakness in that there is no estimate of the exposures of these workers; the negative resultsare reassuring if the exposures are substantial, but not if the exposures are low. The furtherextension of the U.S. Navy study gives data for up to 40 years after likely radar exposures,which may have been quite high, and shows no increase in most cancers, but a small increasein leukaemia in one of the three high exposure groups [31]. Again, the lack of more preciseexposure data is the main limitation. The ‘control’ group in this study may also have hadlevels of RF exposure higher than a general population. The large population based study of


mobile phone subscribers in Denmark [32] also gives substantial evidence against there beingany short term increases in cancer with typical levels of phone use experienced by residentialsubscribers. None of these large studies can provide good information on the intensities ofexposure experienced by the people studied.

There are now four case control studies published on brain cancer in relationship to personaluse of mobile phones, which show no consistent evidence of any increased risk [44,46-48].The main limitation of these studies is the short time frame.

Brain cancers

The two types of cancer which have been of most concern with respect to radiofrequenciesare brain cancers and leukaemias. The brain because with the use of cell phones it is exposedto a very close source of radiofrequencies, rather than because of any biological reason thattumors in the brain are likely to be caused by radiofrequencies. For radiofrequencies otherthan mobile phones, in adults, cancers of the central nervous system and the brain showedpositive results in the Polish military study and in the small case-control study of brain tumorsin US military, but showed no association in the study of electric utility workers, the Sydneystudy, and no clear association in the Sutton Coldfield study. There was no association withcancers of the brain in childhood in any of the three studies which assessed this. The studiesof mobile phone use are generally negative. However, although the studies of brain cancerhave included reasonable total numbers of cancers, the numbers of cancers occurring in theareas of the brain which receive substantial exposure are small, and so again an effect cannotbe excluded.

Leukaemia

Leukaemias have been of great concern in relationship to extremely low frequency (ELF)electric and magnetic fields, from power lines and electrical appliances. There have beenmany more studies of ELF fields than of radiofrequency fields, and the studies are morecomprehensive and more detailed [57,58]. For example there have been many studies ofchildhood leukaemia in relationship to ELF fields which involve taking direct measurementsover 24 or 48 hours in the home, or in one study asking children to wear a recording devicefor a similar period [59]. Although the general consensus is that no definite association hasbeen shown, the several available epidemiological studies of childhood leukaemia and ELFfields taken together do show an association of a modest increase in risk with exposures toresidential ELF fields which are more intense than is usual, with the excess risk being seenwith exposures to fields of over 0.3 or 0.4 microtesla (3 or 4 milligauss) [60-62]. As a resultthe International Agency for Research in Cancer has classified ELF fields as a “possiblecarcinogen”, (classification 2B in their system), along with over 200 other exposures onwhich there is some animal or epidemiological evidence of potential risk [63].

The results for leukaemia are also where the epidemiological studies of radiofrequencyexposures are the most difficult to interpret. Two of the three cluster studies which have beendescribed relate to leukaemia, although the Hawaiian study related to childhood leukaemia,and the Sutton Coldfield to adult leukaemia. The results of all the other studies in terms ofleukaemia are shown in Fig. 5.4. Shown is the main result for the highest exposure group,odds ratio or relative risk, and 95% confidence limits: this simplification of the results servesto show that there is little consistency, but there are several results showing some association.


Figure 1. Summary of results (relative risk or odds ratio and 95% confidence limits) ofstudies assessing leukaemia in association with radiofrequency exposures.

Results for the highest exposure group in each study.

Key to studies. A = Adults C = ChildrenA-UK = UK 21 transmitter Studies [11] C-UK = UK 21 transmitter study [11]A-SC = Sutton Coldfield 1987-94 [26] C-SC = Sutton Coldfield 1987-94 [26]A-SYD = Sydney [64] C-SH = Sydney study (Hocking et al.) [12]A-USN = US Navy [31] C-SMA = Sydney study (Mackenzie et al. all areas)

[13]A-PM= Polish Military [15] C-SMX = Sydney study (Mackenzie et al.,

excluding Lane Cove) [13]A-USR = US amateur radio operators [17] C-SF = San Francisco [14]A-NR = Norwegian radio operators [18]A-CF= Canada-France utility workers [19] A-MO= Motorola employees [30]

The general population studies of cancers in relationship to residence near TV and radiotransmitters are not totally negative. The study of 20 transmitters in the UK to follow up theSutton Coldfield cluster showed a weak and irregular association, but with a significant trendtowards leukaemia rates increasing closer to the transmitters, although the risk in thesepopulations living closest was not elevated [11]. In the later study of the Sutton Coldfieldsituation there was no significant trend but there was a small non-significant increase in riskin the closest group [26]. Although most attention in the North Sydney study by Hocking andcolleagues was on childhood leukaemia, the results also showed some excess of adultleukaemia in the three areas closest to the transmitters, which was significant for acutelymphatic leukaemia [64]. Whereas the situation in children in North Sydney has been

0.1

1

10

100

A-UK

A-SC

A-SY

DA-

USN

A-PM

A-US

RA-

NRA-

CFA-

MO C-UK

C-SC

C-SH

C-SM

AC-

SMX

C-SF

RR and 95% limits


explored further by detailed analysis, that has not been done for adult leukaemia. It would berelevant to see if the excess of adults is concentrated in one area, as is the excess in children.In children, the further studies of childhood leukaemia in North Sydney showed the increaseto be inconsistent, being seen in only one of three municipalities close to the transmitter, butno other explanation for it has been found [13]. The UK 21 transmitter study and the laterstudy of Sutton Coldfield do not confirm any risk in children, but are too weak in terms ofnumbers to exclude a risk. The study in San Francisco however shows a slightly reduced riskcloser to the transmitter, although no confidence limits are given in the paper. In terms of theoccupational studies, a most interesting finding is that there is a significant excess of deathsfrom leukaemia in the servicemen likely to have been exposed heavily to radar in the US navystudy [31]. There is a small excess of leukaemia, significant for one type, in the study ofamateur radio operators. However the studies of Norwegian women radio operators on ships,Canadian and French electric utility workers, Motorola employees, and mobile phonesubscribers in Denmark all show no excess risk. If the hypothesis was that radiofrequencieswere related to leukaemia with a long latent period, such as over 10 years, the studies ofMotorola employees and Danish mobile phone users would not be relevant as they relate onlyto a short time of follow up.

There have been to date no detailed case-control studies on leukaemia, using adequatemeasures of exposure, in regard to radiofrequencies. The case-control studies of generalexposures to radiofrequencies, as distinct from mobile phone use, are generally very weak intheir methodology and radiofrequencies has been simply one of may potential exposuresstudied. For low frequency ELF fields, it is the case-control studies of childhood leukaemiawhich have been able to provide the greatest detail on exposures and on other relevant factors,even though there are still difficulties as it is past exposure that is relevant. High quality case-control studies with great attention to the estimation of exposure and documentation of otherrelevant factors are now in progress to assess further any link between brain cancer and cellphone use, including those in the World Health Organization International EMF project:www.who.int/peh-emf.

There are likely to be some such studies in regard to childhood leukaemia andradiofrequencies, and other cancers and radiofrequencies. The case-control study ofneuroblastoma is a high quality study using good methods, but the paper which has beenpublished concentrates on ELF fields and parental occupation and has much less informationon radiofrequency fields. A major problem is that if there is indeed a real association between,for example, childhood leukaemia and ELF fields an apparent association with radiofrequencyfields may occur simply from a correlation between exposure to ELF fields and exposure toradiofrequency fields. This is already a major problem in occupational studies, as mostoccupational exposures to radiofrequency fields will also involve exposures to ELF fields,and often also to chemical exposures related to electrical and electronics work.

Implications of epidemiological studies for exposure standards

Epidemiological studies primarily relate to the question of whether there is or is not anincreased risk of disease in human populations exposed to the suspect agent. The studiesinclude some which assess likely low levels of exposure, well within current standards, aswell as some which may be assessing irregular higher exposure levels; in none of the studiesis detailed exposure information available. Therefore, the epidemiological work is not directlyhelpful in defining a particular level of radiofrequency exposure which could be hazardous.Equally, the epidemiological evidence does not lead to an argument for any particular changesin currently accepted exposure standards.

http://www.who.int/peh-emf


The epidemiological studies reviewed here do not suggest that currently accepted exposurestandards, such as that of ICNIRP, need to be revised downwards. The overall conclusionfrom the literature is that no detrimental health effects have been observed consistently instudies which are assessing exposure levels which are likely to be within the current standardsor which may have occasionally been beyond those standards, for example in the occupationalstudies. As is expected in any area of work where there are numbers of studies, some makingmultiple observations, there are some positive associations reported: but overall these aremore likely to be due to chance variation, biases in the observations made in the study, or theeffects of other related factors, than due to a causal association with radiofrequencyexposures.

The negative experimental evidence on markers of serious effects, for example in vivo and invitro indicators of carcinogenesis also argue against there being any cancer causing effects atvery low levels of exposure [7,8]. This would apply to the levels of exposure characteristic ofgeneral population exposures from mobile phone base transmitter sites, where typicallyexposures are below one percent of the current ICNIRP standard.

The exposures to the head in users of mobile phones are considerably higher, but the currentepidemiological studies do not suggest any increased risk of brain tumors, but are based onshort follow up times, and include few tumors in the exposed area of the brain.

Conclusions

The epidemiological information available at the time of writing leads to a conclusion that thetotal evidence does not suggest any increased risk of cancer with the exposures toradiofrequencies that have been assessed. Many of the studies individually are weak and, as aconsequence their results cannot be clearly interpreted in terms of cause and effect. Theevidence is weak in regard to its inconsistency, the weak design of many of the studies, thelack of detail on actual exposures, the limitations of the studies in their ability to deal withother likely factors, and in some studies there may be biases in the data used. While thecurrent epidemiological evidence justifies further research to clarify the situation, there is noconsistent evidence of any substantial effect on human cancer causation. In summary, theepidemiological evidence falls short of the strength and consistency of evidence which isrequired to come to a reasonable conclusion that radiofrequency emissions are a cause ofhuman cancer.

However, for the same reasons the studies are not able to confidently exclude that possibility.The cancers for which further work may be most valuable are leukaemia in both adults andchildren on the basis of the results of studies to date, and brain cancer in relationship tomobile phone use because of the dose considerations rather than the epidemiological results,which so far are reassuring.


Reference List

1. Elwood JM. (1999). A critical review of epidemiologic studies of radiofrequencyexposure and human cancers. Environ Health Perspect ;107:155-68.

2. Australian Radiation Protection and Necular Safety Agency. (2002). Radiation ProtectionStandard. Maximum exposure levels to radiofrequency fields - 3kHz to 300 GHz.Melbourne: ARPANSA. p. 1-128.

3. MacMahon B, Pugh TF. (1970). Epidemiology: principles and methods, 1 ed. Boston:Little Brown.

4. Elwood JM. (1998). Critical Appraisal of Epidemiological Studies and Clinical Trials, 2ed. Oxford: Oxford Univ Press. p. 1-448.

5. Hill AB. (1965). The environment and disease: association or causation? Proc R Soc Med;58:295-300.

6. International Commission on Non-ionizing Radiation Protection (ICNIRP). (1998).Guidelines for limiting exposure to time-varying electric, magnetic, and electromagneticfields (up to 300 GHz). Health Phys ;74:494-522.

7. Expert panel on the potential health risks of radiofrequency fields from wirelesstelecommunication devices. (1999). A review of the potential health risks ofradiofrequency fields from wireless telecommunication devices. Royal Society of Canada.1-149. Royal Society of Canada.

8. Independent Expert Group on Mobile Phones. (2000). Mobile phones and health. Oxford:IEGMP. p. 1-160.

9. Bergqvist U.(1997). Review of epidemiological studies, In: Kuster N, Balzano Q, Lin JC,editors. Mobile Communications Safety. London: Chapman & Hall. p. 147-70.

10. Dolk H, Shaddick G, Walls P, Grundy C, Thakrar B, Kleinschmidt L et al. (1997). Cancerincidence near radio and television transmitters in Great Britain 1. Sutton Coldfieldtransmitter. Am J Epidemiol ;145:1-9.

11. Dolk H, Elliott P, Shaddick G, Walls P, Thakrar B. (1997). Cancer incidence near radioand television transmitters in Great Britain 2: All high power transmitters. Am JEpidemiol ;145:10-17.

12. Hocking B, Gordon IR, Grain HL, Hatfield GE. (1996). Cancer incidence and mortalityand proximity to TV towers. Med J Aust ;165:601-05.


13. McKenzie DR, Yin Y, Morrell S. (1998). Childhood incidence of acute lymphoblasticleukaemia and exposure to broadcast radiation in Sydney - a second look. Aust N Z JPublic Health ;22:360-67.

14. Selvin S, Schulman J, Merrill DW. (1992). Distance and risk measures for the analysis ofspatial data: a study of childhood cancers. Soc Sci Med ;34:769-77.

15. Szmigielski S. (1996). Cancer morbidity in subjects occupationally exposed to highfrequency (radiofrequency and microwave) electromagnetic radiation. Sci.Total.Environ;180:9-17.

16. Robinette CD, Silverman C, Jablon S. (1980). Effects upon health of occupationalexposure to microwave radiation (radar). Am J Epidemiol ;112:39-53.

17. Milham S. (1988). Increased mortality in amateur radio operators due to lymphatic andhematopoietic malignancies. Am J Epidemiol ;127:50-54.

18. Tynes T, Hannevik M, Andersen A, Vistnes AI, Haldorsen T. (1996). Incidence of breastcancer in Norwegian female radio and telegraph operators. Cancer Causes Control;7:197-204.

19. Armstrong B, Theriault G, Guenel P, Deadman J, Goldberg M, Heroux P. (1994).Association between exposure to pulsed electromagnetic fields and cancer in electricutility workers in Quebec, Canada, and France. Am J Epidemiol ;140:805-20.

20. Grayson JK, Lyons TJ. (1996). Cancer incidence in United States Air Force aircrew,1975-89. Aviat Space Environ Med ;67:101-04.

21. Thomas TL, Stolley PD, Stemhagen A, Fontham ETH, Bleeker ML, Stewart PA et al.(1987). Brain tumour mortality risk among men with electrical and electronic jobs: acase-control study. J Natl Cancer Inst ;79:233-38.

22. Hayes RB, Brown LM, Pottern LM, Gomez M, Kardaun JWPF, Hoover RN et al. (1990).Occupation and risk for testicular cancer: a case-control study. Int J Epidemiol ;19:825-31.

23. Demers PA, Thomas DB, Rosenblatt KA, Jimenez LM, McTiernan A, Stalsberg H et al.(1991). Occupational exposure to electromagnetic fields and breast cancer in men. Am JEpidemiol ;134:340-47.

24. Cantor KP, Stewart PA, Brinton LA, Dosemeci M. (1995). Occupational exposures andfemale breast cancer mortality in the United States. J.Occup.Environ.Med.37:336-48.

25. Holly EA, Aston DA, Ahn DK, Smith AH. (1996). Intraocular melanoma linked tooccupations and chemical exposures. Epidemiology ;7:55-61.


26. Cooper D, Hemmings K, Saunders P. (2001). Re: "Cancer incidence near radio andtelevision transmitters in Great Britain I. Sutton Coldfield transmitter II. All high powertransmitters". Am J Epidemiol ;153:202-04.

27. Kaplan S, Etlin S, Novikov I, Modan B. (1997). Occupational risks for the developmentof brain tumors. Am J Ind Med ;31:15-20.

28. Finkelstein MM. (1998). Cancer incidence among Ontario police officers. Am J Ind Med;34:157-62.

29. Davis RL, Mostofi FK. (1993). Cluster of testicular cancer in police officers exposed tohand-held radar. Am J Ind Med ;24:231-33.

30. Morgan RW, Kelsh MA, Zhao K, Exuzides KA, Heringer S, Negrete W. (2000).Radiofrequency exposure and mortality from cancer of the brain andlymphatic/hematopoietic systems. Epidemiology ;11:118-27.

31. Groves FD, Page WF, Gridley G, Lisimaque L, Stewart PA, Tarone RE et al. (2002).Cancer in Korean war navy technicians: mortality survey after 40 years. Am J Epidemiol;155:810-18.

32. Johansen C, Boice JD, Jr., McLaughlin JK, Olsen JH. (2001). Cellular telephones andcancer - a nationwide cohort study in Denmark. J Natl Cancer Inst ;93:203-07.

33. Lagorio S, Rossi S, Vecchia P, de Santis M, Bastianini L, Fusilli M et al. (1997).Mortality of plastic-ware workers exposed to radiofrequencies. Bioelectromagnetics;18:418-21.

34. Törnqvist S, Knave B, Ahlbom A, Persson T. (1991). Incidence of leukaemia and braintumours in some "electrical occupations". Br J Ind Med ;48:597-603.

35. Pearce N, Reif J, Fraser J. (1989). Case-control studies of cancer in New Zealandelectrical workers. Int.J.Epidemiol. 18:55-59.

36. De Roos AJ, Teschke K, Savitz DA, Poole C, Grufferman S, Pollock BH et al. (2001).Parental occupational exposures to electromagnetic fields and radiation and the incidenceof neuroblastoma in offspring. Epidemiology ;12:508-17.

37. Rothman KJ, Loughlin JE, Funch DP, Dreyer NA. (1996). Overall mortality of cellulartelephone customers. Epidemiology ;7:303-05.

38. Dreyer NA, Loughlin JE, Rothman KJ. (1999). Cause specific mortality in cellulartelephone users. JAMA ;282:1814-14.

39. Hocking B. (2001). Re: Cellular telephones and cancer-a nationwide cohort study inDenmark. J Natl Cancer Inst ;93:877-78.


40. Johansen C, Boice JD, Jr., McLaughlin JK, Olsen JH. (2001). Response: Cellulartelephones and cancer - a nationwide cohort study in Denmark. J Natl Cancer Inst;93:878-79.

41. Godward S, Sandhu M, Skinner J, McCann J. (2001). Re: Cellular telephones and cancer-a nationwide cohort study in Denmark. J Natl Cancer Inst ;93:878-78.

42. Johansen C, Boice JD, Jr., McLaughlin JK, Olsen JH. (2001). Response: Cellulartelephones and cancer - a nationwide cohort study in Denmark. J Natl Cancer Inst;93:952-53.

43. Hardell L, Mild KH. (2001). Re: Cellular telephones and cancer-a nationwide cohortstudy in Denmark. J Natl Cancer Inst ;93:952-53.

44. Hardell L, Näsman Å, Påhlson A, Hallquist A, Hansson Mild K. (1999). Use of cellulartelephones and the risk for brain tumours: a case-control study. Int J Oncol ;15:113-16.

45. Hardell L, Nasman A, Pahlson A, Hallquist A. (2000). Case-control study on radiologywork, medical x-ray investigations, and use of cellular telephones as risk factors for braintumors. MedGenMed ;May 4:1-11.

46. Muscat JE, Malkin MG, Thompson S, Shore R, Stellman S, McRee D et al. (2000).Handheld cellular telephone use and risk of brain cancer. JAMA ;284:3001-07.

47. Inskip PD, Tarone RE, Hatch EE, Wilcosky TC, Shapiro WR, Selker RG et al. (2001).Cellular-telephone use and brain tumors. N Engl J Med ;344:79-86.

48. Auvinen A, Hietanen M, Luukkonen R, Koskela RS. (2002). Brain tumors and salivarygland cancers among cellular telephone users. Epidemiology ;13:356-59.

49. Rothman KJ, Chou C-K, Morgan R, Balzano Q, Guy AW, Funch DP et al. (1996).Assessment of cellular telephone and other radio frequency exposure for epidemiologicresearch. Epidemiology ;7:291-98.

50. Ahlbom A, Feychting M. (1999). Re: Use of cellular phones and risk of brain tumours: acase-control study. Int J Oncol ;15:1045-45.

51. Hardell L, Nasman A, Pahlson A, Hallquist A, Hansson Mild K. (1999). In reply: Use ofcellular phones and the risk of brain tumours: a case-control study. Int J Oncol ;15:1045-47.

52. Trichopoulos D, Adami H-O. (2000). Cellular telephones and brain tumors. N Engl J Med.

53. Greenland S. (1987). Quantitative methods in the review of epidemiologic literature.Epidemiol Rev ;9:1-30.


54. Stang A, Anastassiou G, Ahrens W, Bromen K, Bornfeld N, Jockel K-H. (2001). Thepossible role of radiofrequency radiation in the development of uveal melanoma.Epidemiology ;12:7-12.

55. Inskip PD. (2001). Frequent radiation exposures and frequency-dependent effects: theeyes have it. Epidemiology ;12:1-4.

56. Johansen C, Boice JD, McLaughlin JK, Christensen HC, Olsen JH. (2002). Mobilephones and malignant melanoma of the eye. Br J Cancer ;86:348-49.

57. Linet M, Hatch EE, Kleinerman RA, Robison LL, Kaune WT, Friedman DR et al. (1997).Residential exposure to magnetic fields and acute lymphoblastic leukemia in children. NEngl J Med ;337:1-7.

58. UK Childhood Cancer Study Investigators. (1999). Exposure to power-frequencymagnetic fields and the risk of childhood cancer. Lancet ;354:1925-31.

59. McBride ML, Gallagher RP, Thériault G, Armstrong BG, Tamaro S, Spinelli JJ et al.(1999). Power-frequency electric and magnetic fields and risk of childhood leukemia inCanada. Am J Epidemiol ;149:831-42.

60. Advisory Group on Non-ionising radiation. (2001). ELF Electromagnetic fields and therisk of cancer. Report of an Advisory Group on Non-ionising radiation., vol. 1. Didcot:National Radiological Protection Board. p. 1-179.

61. Ahlbom A, Day N, Feychting M, Roman E, Skinner J, Dockerty J et al. (2000). A pooledanalysis of magnetic fields and childhood leukemia. Br J Cancer ;83:692-98.

62. Greenland S, Sheppard AR, Kaune WT, Poole C, Kelsh MA. (2000). A pooled analysis ofmagnetic fields, wire codes, and childhood leukemia. Epidemiology ;11:624-34.

63. IARC Working Group on the Evaluation of Carcinogenic Risk to Humans. (2002). IARCMonographs on the evaluation of carcinogenic risks to humans. Non-ionizing radiation,part 1: Static and extremely low-frequency (elf) electric and magnetic fields, vol. 80.Lyon, France: World Health Organization, International Agency for Research on Cancer.p. 1-429.

64. Hocking B, Gordon I, Hatfield G, Grain H. (1998). Re: Cancer incidence near radio andtelevision transmitters in Great Britain. I. Sutton Coldfield transmitter. II. All highpower transmitters. Am J Epidemiol ;147:90-91.

REPORT TO ARPANSA ON EPIDEMIOLOGICAL STUDIES

Documents