Stang, Andreas1; Anastassiou, Gerasimos2; Ahrens, Wolfgang1,3; Bromen, Katja1; Bornfeld, Norbert2; Jöckel, Karl-Heinz1
Although it is a rare disease, uveal melanoma of the eye is the most common primary intraocular malignancy in adults, with an incidence rate of up to 1.0 per 100,000 person-years (age-standardized, world standard) in Europe. 1 Only a few consistent risk factors have been identified for this disease. One set of uncommon risk factors includes predisposing diseases such as dysplastic nevus syndrome, atypical ocular nevi, and ocular and oculardermal melanocytosis. 2,3 Another set of host risk factors comprises ancestry, light skin, and iris pigmentation. 3–6 In addition, a number of environmental factors are weakly or inconsistently associated with uveal melanoma. Occupation may be also relevant and may include chemical work, 7,8 arc welding, 8 and agriculture and farming work. 9,10
Few epidemiologic studies dealing with electromagnetic radiation and uveal melanoma have been published. The majority of these studies are exploratory and are based on job and industry titles only. In a cancer registry study in England and Wales, Swerdlow 11 found an increased risk for eye cancer among electrical and electronics workers in particular. This finding could not be reproduced in a small case-control study conducted by Gallagher et al.4 Holly et al8 found a twofold increased risk of uveal melanoma among subjects with occupational exposure to microwaves and radar units.
There is currently much uncertainty about the role, if any, of radiofrequency transmitted by radio sets or mobile phones in human carcinogenesis. The assessment of the potential association of radiofrequency radiation (RFR) and cancer risk is hampered by uncertainties about effective electromagnetic frequency ranges, the lack of a clear biological mechanism, and difficulties of exposure assessment. To date, no study has been published concerning intraocular melanoma in relation to occupational exposures to electromagnetic fields, including RFR as transmitted by radio sets or wireless telephones.
Electromagnetic waves with frequencies of 300 kHz to 300 GHz are called RFR. Typical occupational sources transmitting RFR in Germany include walkie-talkies in the military and security services and in plants; radio sets on ships, transporters, freight trains, and police cars; and wireless phones, including cellular phones [C-net, 450–465 MHz (since the 1990s); D-net, 890–960 MHz; and E-net, 1710–1800 MHz] and cordless phones (800–1900 MHz) with different modulation types. 12,13
Here, we report findings from a hospital-based and population-based case-control study that used the same standardized questionnaires about potential risk factors, including electromagnetic radiation, for uveal melanoma. The studies produced findings for several potential risk factors. This report concerns the association of RFR with the risk for uveal melanoma.
Subjects and Methods
Between 1994 and 1997, our research group participated in a multinational population-based case-control study on occupational risk factors for eight rare cancers including uveal melanoma, mycosis fungoides, cancer of the small intestine and bones, cancer of the bile duct and breast in males, thymoma, and testicular cancer, the latter only in Germany. During the field phase of this study we started an additional hospital-based case-control study at the Division of Ophthalmology, University of Essen, using the same questionnaire and study personnel. The University of Essen is a referral center for tumors of the eye and treats about 250–300 patients with uveal melanoma per year. The methodologic features of the two studies are presented in Table 1.
Controls in the population-based study were selected randomly from mandatory lists of residence that cover the total population of the local districts. The lists of residence are regarded as the most complete sampling frame for population-based studies in Germany. 15 For each case, about four potential controls matching on sex, age (5-year groups), and region of residence (five regions) were selected and contacted by letters, phone calls, and home visits if necessary. The number of controls was determined by the stratum-specific maximum number of any of the eight rare cancers included in the study. The matching resulted in matching ratios for the uveal melanoma study of up to ten controls per case, because other cancers were more common than uveal melanoma. Owing to the inclusion of testicular cancer in Germany, the number of matching male controls in the uveal melanoma study exceeded that of matching female controls. We excluded 372 control interviews from the analysis of the population-based study because these controls matched to the other rare cancer cases only.
The major diagnoses among the hospital controls were retinal detachment and defects (32%), degeneration of the macula (18%), retinal vascular occlusion (14%), diabetic retinopathy (10%), and others (26%). No hospital control suffered from an occupational accident involving the eye.
Data were collected from study participants by interviewers who were trained specifically for this project and were monitored throughout the study to ensure uniform quality of the questionnaire data. They were unaware of specific study hypotheses as they administered structured questionnaires. Interviews took about 70 minutes (median).
The interview topics included medical history, phenotypic characteristics, life-style factors, details of the lifetime occupational history, and occupational sources of electromagnetic radiation. These sources included working close to high-voltage electric lines, electrical machines, complex electrical environments (that is, control rooms or computer rooms), and visual terminal displays. We added questions about radar units and RFR to the electromagnetic radiation section of the international questionnaire that enabled us to estimate the risk of uveal melanoma associated with these sources within the German study part. Exposed subjects were asked about the starting and ending year of exposure and details of the electromagnetic radiation sources.
Occupational exposure to RFR devices was assessed by asking subjects, “Did you use radio sets, mobile phones, or similar devices at your workplace for at least several hours per day?” Subjects who answered “yes” were asked about the beginning and ending year of exposure and about the way the RFR source was carried with the subjects. This question often helped to identify the source of RFR, because subjects gave details on the RFR sources.
The reference date for the exposure was the date of diagnosis (cases), date of first contact letter (population controls), and date of diagnosis (hospital controls). Exposures after the reference date were not considered. All information on the RFR exposure and on corresponding occupational periods (including production processes, job tasks, etc) were reviewed for all subjects who reported being exposed to RFR devices (pooled analysis; 18 cases and 55 controls). With the case-control status masked, subjects were rated by one of the authors (A. S.) as (a) exposed only to radio receivers that do not transmit RFR and therefore unexposed, (b) exposed to walkie-talkies and radio sets, (c) possibly exposed to mobile phones, and (d) probably or certainly exposed to mobile phones.
Subsequently, all subjects were reviewed and rated under similar circumstances by another author (K. H. J.), who was not aware of the other ratings. All subjects with discrepant ratings were evaluated and the discrepancy was resolved. The interrater agreement for the ratings (a–d), as expressed by weighted kappa, was 0.78 [95% confidence interval (95% CI) = 0.67–0.88].
We applied conditional logistic regression models to calculate odds ratios (ORs) and 95% CIs. 16,17 If cell sizes across the strata were five or lower, we calculated exact OR estimates and exact 95% CI. 18 Matching factors for the conditional analysis were age (5-year groups), sex, and geographic area [population-based study, Bremen, Essen, Hamburg, Saarbrücken, and the federal state of Saarland without Saarbrücken; hospital-based study, Ruhrarea, non-Ruhrarea southern part (<100,000 or ≥100,000 inhabitants), and non-Ruhrarea northern part (<100,000 or ≥100,000 inhabitants)]. After we estimated the ORs for the separate studies, we pooled the two studies and ran the same conditional logistic models taking the matching into account.
First, we estimated ORs without assumptions about latency periods or exposure duration. Second, we estimated ORs on the basis of exposures that started at least 5 years before the reference year. Third, we calculated ORs on the basis of exposures with durations of at least 3 years before the reference year. We used dummy coding to control for potential confounding by social class with “9 years at school or less” as the reference category.
Tumors involved the choroid (98%), iris (1%), and unknown parts of the uveal tract (1%). The proportion of face-to-face interviews in the hospital-based study was lower than in the population-based study. In addition, the proportion of male controls was larger in the population-based study because the control group was used for several rare cancers and its size was determined by the most frequent cancers in each stratum. Compared with the population-based study, cases of the hospital-based study had a slightly higher and controls a slightly lower socioeconomic status as measured by the number of years at school (Table 2).
Table 3 presents the exposure prevalences and OR estimates for the different sources of electromagnetic radiation. The highest exposure prevalences were seen for electrical machines and visual display terminals. Only the effect estimates for the exposure to radio sets or mobile phones were consistently elevated across the studies. The remaining effect estimates either were close to unity or were noninformative owing to low exposure prevalences (radar units).
Table 4 presents the OR estimates for uveal melanoma based on the RFR exposure rating for both studies and the pooled analysis. Overall, 46 controls (9.7%) and 16 cases (13.6%) were rated as ever having been exposed to any kind of RFR at their jobs for at least a 6-month duration and several hours per day. The lifetime prevalences were lower for both cases and controls in the hospital-based study. Only one female case and one female control (both in the hospital-based study) reported having been exposed to any RFR at their jobs. These women were rated as “probably or certainly exposed to mobile telephone.” Occupational exposures to radio sets included walkie-talkies in the military and security services and in plants and radio sets on ships, transporters, freight trains, and police cars. The median year of start of exposure was 1975 (25% percentile, 1963; 75% percentile, 1983). Twelve subjects were rated as “possibly exposed to mobile phones” because occupational information was scant or equivocal with regard to mobile phone exposure. The median year of start of exposure was 1992 (25% percentile, 1988; 75% percentile, 1996).
Subjects probably or certainly exposed to mobile phones typically worked in the service sector as real estate agents, tax consultants, or sales representatives. The median year of start of exposure was 1993 (25% percentile, 1990; 75% percentile, 1995) and was considerably later compared with subjects exposed to radio sets. The comparatively late onset of exposure to mobile phones may be explained by the recent widespread introduction of mobile phones to the German market.
Occupational exposure to radio sets was associated with an increased risk for uveal melanoma in both studies and under different exposure-disease assumptions. The separate analyses for the exposure categories “possibly exposed to mobile phones” and “probably or certainly exposed to mobile phones” were not informative within the studies because the exposure prevalences were too low in these categories. The combined analysis of these two exposure categories showed consistently elevated OR estimates across the studies and across different exposure definitions (Table 4).
The adjustment for socioeconomic status as measured by years at school did not alter our results. Light iris color (blue, gray, green, or light brown) as compared with dark iris color (dark brown) was associated with an increased risk of uveal melanoma (pooled OR = 2.9, 95% CI = 1.5–5.6). People with light hair color (red or blonde) at age 20 years showed an increased risk for uveal melanoma (pooled OR = 1.4, 95% CI = 0.9–2.3). The control for iris and hair color did not change our results substantially (data not shown).
We found an increased risk of uveal melanoma in relation to RFR as transmitted by radio sets and mobile phones. The association between electromagnetic fields and uveal melanoma was limited to RFR. Analyses based on different RFR exposure definitions [minimum exposure duration (3 years) and RFR exposure beginning at least 5 years before the diagnosis of the disease] changed the magnitudes of effect estimates to some extent but did not change our interpretation of the data. Furthermore, we restricted our analyses to face-to-face interviews, to interviews that were rated as reliable by the interviewers, or to both and found that the effect estimates were stable (data not shown). Adjustment for social class as measured by highest degree of schooling did not change our results.
The exposure prevalences for sources of electromagnetic radiation in the control groups were consistently higher in the population-based study as compared with the hospital-based study. After adjusting for age and gender, we found that the prevalence differences in the control group were mainly explained by differences in the sex and age distributions of the control groups (for the population-based and hospital-based study, respectively, the proportions of males were 73% and 51%, and median ages were 56 and 61 years).
The analysis of the German testicular cancer study (269 interviewed cases and 797 interviewed controls), which used the same study methodology and questionnaire as the population-based eye melanoma study, showed OR estimates close to unity for all sources of electromagnetic fields (data not shown). Thus, the RFR finding within the uveal melanoma study is specific for this tumor site.
It is still unclear whether RFR exposures too weak to increase temperature measurably could have biological effects. 19 Biological interaction mechanisms are not necessarily thermal; however, many studies have suggested that RFR exposure at low levels that do not challenge thermoregulation or produce any change in cell temperature may have biological effects, but they have either not been consistently replicated or else their relevance for human health cannot be adequately assessed using information currently available. A hypothesized mode of action is that RFR might promote (that is, speed up) the development of cancer that has been caused by other agents. 20
How could RFR act as a promoter for uveal melanoma? Ocular melatonin is synthesized in the retina and the ciliary body and is also found in the aqueous humor. 21 Experimental studies on cultured human uveal melanoma cells indicate that melatonin inhibits the growth of uveal melanoma cell lines in a dose-dependent manner and therefore has an antiproliferative effect. 22 The following explanation is speculative: if RFR decreased the amount of ocular melatonin, it would promote the development of uveal melanoma.
The results of the present study do not provide a firm indication of an effect on uveal melanoma for several reasons. If cases tend to overreport use of radio sets in their work place relative to controls, the OR estimates would be biased upward. Study subjects, however, were unlikely to be aware of the hypothesis, because no study on the relation between radio set use and uveal melanoma has been previously reported. We selected controls in the hospital-based case-control study from the same division as the cases, with benign eye diseases of the posterior eye segment that are diagnosed in a way similar to that for uveal melanoma. This choice may have reduced recall and interviewer bias. In addition, among the different sources of occupational electromagnetic radiation that were asked, the association was limited to radio sets and mobile phones. If recall bias would explain the association between RFR and uveal melanoma, it is implausible that the relative risk estimates for visual display terminals would not be increased, because there are public concerns regarding display terminals and eye diseases.
Our questionnaire, which dealt with many potential risk factors, not only for uveal melanoma, was not originally designed for a detailed exposure history of RFR. It therefore contained few questions about RFR. We did not explicitly ask the subjects which specific kinds of RFR devices they used (including electromagnetic properties of the devices such as signal frequency, modulation type, etc) and how many hours per day they actively used the RFR devices. Also, we did not study the laterality of the tumor in relation to the hand that usually used the RFR device.
Our exposure rating is another limitation of the study. About 12 subjects (19% of all exposed subjects) were rated as only “possibly exposed to mobile phones” because the information was scanty or equivocal. The similarity of the median start year of exposure for “subjects possibly exposed” (1992) and “subjects probably or certainly exposed to mobile phones” (1993) is a rough indicator of validity of the assignment. Low exposure prevalences of mobile phone use and almost no female cases and controls exposed further limit our study. Two other methodologic aspects might have influenced our results. Nonresponse in the population-based control group might have biased our results if nonparticipation was associated with a higher prevalence of radio set or mobile phone use. In addition, the selection criteria for the hospital-based control group might have biased our results if the prevalence estimates of RFR exposure are underestimated.
1. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J, eds. Cancer in Five Continents. vol. VII. IARC Scientific Pub. No. 143. Lyon: International Agency for Research on Cancer, 1997.
2. Jensen OA, Movin M, Müller J. Malignant melanoma of the choroid in an infant with the dysplastic naevus syndrome. Acta Ophthalmol 1987; 65: 91–100.
3. Seddon JM, Gragoudas ES, Glynn RJ, Egan KM, Albert DM, Blitzer PH. Host factors, UV radiation, and risk of uveal melanoma: a case-control study. Arch Ophthalmol 1990; 108: 1274–1280.
4. Gallagher RP, Elwood JM, Rootman J, Spinelli JJ, Hill GB, Threlfall WJ, Birdsell JM. Risk factors for ocular melanomas: Western Canada Melanoma Study. J Natl Cancer Inst 1985; 74: 775–778.
5. Tucker MA, Shields JA, Hartge P, Augsburger J, Hoover RN, Fraumeni JF Jr. Sunlight exposure as risk factor for intraocular malignant melanoma. N Engl J Med 1985; 313: 789–792.
6. Holly EA, Aston DA, Char DH, Kristiansen JJ, Ahn DK. Uveal melanoma in relation to ultraviolet light exposure and host factors. Cancer Res 1990; 50: 5773–5777.
7. Albert DM, Puliafito CA, Fulton AB, Robinson NL, Zakov ZN, Dryja TP, Smith AB, Egan E, Leffingwell SS. Increased incidence of choroidal malignant melanoma occurring in a single population of chemical workers. Am J Ophthalmol 1980; 89: 323–337.
8. Holly EA, Aston DA, Ahn DK, Smith AH. Intraocular melanoma linked to occupations and chemical exposures. Epidemiology 1996; 7: 55–61.
9. Saftlas AF, Blair A, Cantor KP, Hanrahan L, Anderson HA. Cancer and other causes of death among Wisconsin farmers. Am J Ind Med 1987; 11: 119–129.
10. Ajani UA, Seddon JM, Hsieh CC, Egan KM, Albert DM, Gragoudas ES. Occupation and risk of uveal melanoma. Cancer 1992; 70: 2891–2900.
11. Swerdlow AJ. Epidemiology of eye cancer in adults in England and Wales, 1962–1977. Am J Epidemiol 1983: 118: 294–300.
12. Valberg PA. Radio frequency radiation (RFR): the nature of exposure and carcinogenic potential. Cancer Causes Control 1997; 8: 323–332.
13. Bernhardt JH. Gesunheitliche Aspekte des Mobilfunks. Dtsch Ärzteblatt 1999; 96: A845–A852.
14. Slattery ML, Edwards SL, Caan BJ, Kerber RA, Potter JD. Response rates among control subjects in case-control studies. Ann Epidemiol 1995; 5: 245–249.
15. Stang A, Ahrens W, Jöckel KH. Control response proportions in population-based case-control studies in Germany. Epidemiology 1999; 10: 181–183.
16. Breslow NE, Day NE. Statistical Methods in Cancer Research. vol. 1. The Analysis of Case-Control Studies. IARC Scientific Pub. No. 32. Lyon: International Agency for Research on Cancer, 1980.
17. SAS Institute. Changes and Enhancements through Release 6.11. Cary, NC: SAS Institute, 1996.
18. Cytel Software Corp. LogXact. Cambridge, MA: Cytel, 1993.
19. Rothman KJ, Chou CK, Morgan R, Balzano Q, Guy AW, Funch DP, Preston-Martin S, Mandel J, Steffens R, Carlo G. Assessment of cellular telephone and other radio frequency exposure for epidemiologic research. Epidemiology 1996; 7: 291–298.
20. International Commission on Non-Ionizing Radiation Protection. Health issues related to the use of hand-held radiotelephones and base transmitters. Health Phys 1996; 70: 587–593.
21. Martin XD, Malina HZ, Brennan MC, Hendrickson PH, Lichter PR. The ciliary body: the third organ found to synthesize the indoleamines in humans. Eur J Ophthalmol 1992; 2: 67–72.
22. Hu DN, McCormick SA, Roberts JE. Effects of melatonin, its precursors and derivatives on the growth of cultured human uveal melanoma cells. Melanoma Res 1998; 8: 205–210.
© 2001 Lippincott Williams & Wilkins, Inc.