Comments on "A critical review of epidemiologic studies of radiofrequency exposure and human cancers".

Comments on Elwood's article: A critical review of epidemiologic studies of radiofrequency exposure and human cancers. Environ Health Perspect 107(suppl 1):155-168 (1999).


Comments on "A Critical
Review of Epidemiologic Studies of Radiofrequency Exposure and Human Cancers" Elwood (1) made some important omissions in his critical review of radiofrequency radiation (RFR) and cancer.
Elwood referred extensively to the report by myself and others (2) on childhood leukemia in proximity to television (TV) towers in Sydney, Australia. He noted that the relative risk (RR) for childhood leukemia incidence was 1.58 [95% confidence interval (CI), 1.1-2.
3)] and more so for mortality 2.3 (CI, 1.4-4.0). He then referred to the studies of Dolk et al. (3,4) of cancer near TV/ultrahigh frequency (UHF) transmitters in the United Kingdom as negative studies with regard to our study. In our letter (5), in which we commented on Dolks' UK studies, we pointed out that Dolks' studies did not examine for mortality. Elwood (1) did not discuss our novel finding of greater risk for mortality than incidence, which is suggestive of adverse survival, or that this observation has not been negated by other studies.
Elwood (1) also discussed a paper by McKenzie et al. (6), which was a reanalysis of our original data. Neither McKenzie et al. (6) nor Elwood (1) mentioned that the original hypothesis was that the group of three municipalities that immediately surround the TV towers would differ from the next six municipalities surrounding the towers (ring) with regard to leukemia. We treated the municipalities in each ring as a group, and we reported tests of homogeneity (p = 0.10 for incidence and p = 0.13 for mortality) between the inner municipalities in the original paper, which is shown in detail in our rebuttal letter (7). That there were some differences between the three municipalities is to be expected. However, it violates the original hypothesis to disaggregate the three inner municipalities, thus ignoring their homogeneity, to retrospectively conduct individual comparisons.
Elwood (1) contrasted the U.S. Naval Study by Robinette et al. (8), which apart from lung cancer found no excess cancer, with the Polish Military Study by Szmigelski (9), which found an excess of cancer at several sites including esophagus and bowel, as well as lymphohematopoietic and brain cells. Elwood (1) suggested that a systematic bias arose in the Polish study when data were collected on RFR exposure on cancer cases. However, all jobs had been previously measured and classified as exposed or nonexposed to RFR. All new cancer cases were individually reassessed regarding exposures. It is not obvious where the bias arose.
Elwood (1) noted that a weakness in the U.S. Naval study (8) is that it compared groups with high and low (> or < 1.0 mW/cm2) exposures and lacked an unexposed group to assess if the low-exposure group was truly unaffected. This is more than a weakness because both high and low exposure groups took recreation on decks where they were exposed to RFR, occasionally up to 1 mW/cm2 according to Robinette et al. (8). This is important given Szmigelski's finding of effects occurring at < 0.1 mW/cm2, and may explain the null findings of the U.S. Naval study.
Also, Szmigelski (9) stated that exposures were 150-3,500 MHz, whereas the U.S. Naval study simply stated that microwave radar was > 300 MHz. The importance of this difference is that the lower frequencies (150-300 MHz) in the Polish study (9) include wavelengths that have much greater coupling with the body, which in turn may contribute to a different spectrum of cancer sites.
Early in his paper, Elwood (1) noted that there is evidence that RFR may be a promoter of cancer. However, he did not consider the implications of this when discussing the study of brain tumors by Thomas et al. (10). Thomas et al. (10) found an increased risk of brain tumors (RR 2.3) in individuals who had both been exposed to RFR and worked in electronics, which would have likely caused exposure to solvents and fumes. A promotional effect of RFR is consistent with this observation.
Finally, in "Acknowledgments" Elwood mentioned that his paper was "stimulated by a request from Telecom New Zealand for a review of this topic." He did not mention that 2 months before submission of the paper, he had appeared as the major witness for Telecom NZ in a court case regarding placement of a mobile phone tower beside a primary school (11) (1). In regard to his own study (2), I put more emphasis on the incidence than the mortality results for several reasons. The interpretation of the mortality results is more complex, requiring control for confounding by prognostic factors (such as stage at diagnosis and precise age) as well as by risk factors for incidence. The difference between the relative risks for incidence and for mortality is not statistically significant, and of course the two results are not independent. The incidence results are also more useful because they can be compared with those of another study. The discussion in the paper by Hocking et al. (2) is almost all on the incidence relationship. The suggestion that radiofrequency radiation (RFR) exposure is related to adverse survival is a new hypothesis generated from these results and, as far as I know, has not been assessed in other studies. The comparison of the two studies of childhood leukemia in Sydney, Australia (2-4), involves a comparison of concepts. In his letter, Hocking claims that the original hypothesis for these studies was that the leukemia rate in the three areas close to the TV towers would be different from the rate in the six areas farther away; as stated in my review (1), his statistical analysis depends on this comparison. However, in my opinion, the original hypothesis is epidemiologicalwhether there is an increased cancer incidence (and mortality) in children exposed to RFR from TV towers; this is given as the objective in the first paper by Hocking et al. (2). The use of a statistical design that compares two sets of areas is one way to assess this. This approach is not unreasonable but ignores the information provided by the comparison of each individual area. Such data are relevant to the assessment of the consistency of any association, which is an important aspect in assessing causality. I was surprised that the results by individual municipality, which Hocking et al. had available, were not given in the original paper (4), as I believe they affect the interpretation. The subsequent analysis showed that the excess was seen in only one of the three areas close to the TV towers (3). Because of statistical variability, this does not rule out the general association seen by Hocking et al., but it shows inconsistency and weakens the argument that the association seen is caused by RFR from the TV towers rather than from any other cause.
In the Polish military study (5), the published report states that information on possible carcinogenic factors and RFR exposure was available for cancer cases from hospital records, in addition to data from other sources available for all personnel. This raises the possibility of systematic bias, as some information on exposure is available only for affected subjects. This potential bias has been noted independently in another detailed epidemiologic review (6). In regard to the U.S. Navy study (7), Hocking emphasizes the major weakness of the study, which I have noted. I agree that this study is very limited in exposure information.
In the case-control study of brain cancers, Thomas et al. (8) found a significant excess risk in electronics workers with no exposure to RFR, and no excess risk in those exposed to RFR who were not electronics workers. There was an increased risk in electronics workers who were also exposed to RFR, but this risk was lower than the risks for all electronics workers. Although this may be consistent with some complex promotional effect, the more parsimonious explanation is that the increased risk in electronics workers is due to some exposure other than RFR.
In his letter, Hocking refers to a New Zealand environment court case (9) that concerned a proposed Telecom cell phone transmitter site near a school. I appeared as an expert witness for Telecom, and he appeared as a witness for the school. My published review (1) was developed at the same time as my written evidence, but was not submitted until after the case in order to benefit from legal review as well as from scientific peer review. The legal hearing has resulted in a detailed judgment in favor of Telecom (9). In his judgment, Judge Jackson commented on each of the several expert witness submissions. He noted that "Elwood's evidence was carefully constructed and balanced" (y.
In summary, although the points raised by Hocking are worthy of note, I do not agree that any of them represent "important omissions" in my review paper.