Content area
Although individual epidemiological investigations have suggested associations between residential exposure to electromagnetic fields (EMFs) and childhood leukaemia, overall the findings have been inconclusive. Several of these studies do, however, lend themselves to application of the meta-analysis technique. For this purpose we carried out searches using MEDLINE and other sources, and 14 case-control studies and one cohort study were identified and evaluated for epidemiological quality and included in the meta-analysis. Relative risk estimates were extracted from each of the studies and pooled. Separate meta-analyses were performed on the basis of the assessed EMF exposure (wiring configuration codes, distance to power distribution equipment, spot and 24-h measures of magnetic field strength (magnetic flux density) and calculated magnetic field). The meta-analysis based on wiring configuration codes yielded a pooled relative risk estimate of 1.46 (95% confidence interval (CI) = 1.05-2.04, P = 0.024) and for that for exposure to 24-h measurements of magnetic fields, 1.59 (95% CI = 1.14-2.22, P = 0.006), indicating a potential effect of residential EMF exposure on childhood leukaemia. In most cases, lower risk estimates were obtained by pooling high-quality studies than pooling low-quality studies. There appears to be a clear trend for more recent studies to be of higher quality. Enough evidence exists to conclude that dismissing concerns about residential EMFs and childhood leukaemia is unwarranted. Additional high-quality epidemiological studies incorporating comparable measures for both exposure and outcomes are, however, needed to confirm these findings and, should they prove to be true, the case options for minimizing exposure should be thoroughly investigated to provide definitive answers for policy-makers.
Full text
Although individual epidemiological investigations have suggested associations between residential exposure to electromagnetic fields (EMFs) and childhood leukaemia, overall the findings have been inconclusive. Several of these studies do, however, lend themselves to application of the meta-analysis technique. For this purpose we carried out searches using MEDLINE and other sources, and 14 case-control studies and one cohort study were identified and evaluated for epidemiological quality and included in the meta-analysis. Relative risk estimates were extracted from each of the studies and pooled. Separate meta-analyses were performed on the basis of the assessed EMF exposure (wiring configuration codes, distance to power distribution equipment, spot and 24-h measures of magnetic field strength (magnetic flux density) and calculated magnetic field). The meta-analysis based on wiring configuration codes yielded a pooled relative risk estimate of 1.46 (95% confidence interval (CI) = 1.05-2.04, P= 0.024) and for that for exposure to 24-h measurements of magnetic fields, 1.59 (95% Cl = 1.14-2.22, P= 0.006), indicating a potential effect of residential EMF exposure on childhood leukaemia. In most cases, lower risk estimates were obtained by pooling high-quality studies than pooling low-quality studies. There appears to bea clear trend for more recent studies to be of higher quality. Enough evidence exists to conclude that dismissing concerns about residential EMFs and childhood leukaemia is unwarranted. Additional high-quality epidemiological studies incorporating comparable measures for both exposure and outcomes are, however, needed to confirm these findings and, should they prove to be true, the case options for minimizing exposure should be thoroughly investigated to provide definitive answers for policy-makers.
Keywords: electromagnetic fields; leukaemia, radiation induced; meta-analysis; risk factors.
Introduction
In 1979 Wertheimer & Leeper (1) published the first report that showed an increased risk of cancer mortality among children living near electrical wiring configurations and which were consistent with the presence of high currents. Although many studies linking exposure to electromagnetic field, (EMFs) with health effects have been conducted, there is still much debate over whether this exposure at the levels that occur in domestic settings can cause cancer, particularly childhood leukaemia. Some workers have recently called for an end to further research on exposure to magnetic fields (2), whereas others believe that abandoning research on this topic is premature (3, 4).
Meta-analysis is a quantitative approach for systematically combining the results of previous studies in order to arrive at conclusions that cannot be drawn from the results of any one study alone. Although it has been applied most often to combine the results of randomized trials, use of meta-analysis is not confined to the synthesis of information from experimental studies. A large number of studies that involve meta-analysis of nonexperimental data have been published in recent years, although such use of the technique is less accepted than it is in the analysis of data from clinical trials (5). Meta-analyses of observational epidemiological studies have also previously been carried out to examine the relationship between residential EMF exposure and childhood leukaemia (6-11). In general, such analyses have shown a significant increased risk of childhood leukaemia when residential exposure is assessed through the use of wiring configuration codes (a categorical exposure rating scheme based on wire size and distance from the residence), whereas the association with other related markers of exposure, such as proximity to power fines and calculated magnetic fields from power fines, appears less evident.
Several well-conducted epidemiological studies on the association between EMFs and childhood leukaemia were published after the above-mentioned meta-analyses appeared. The purpose of the present investigation was to reassess the risk of childhood leukaemia associated with residential EMF exposure in the fight of these more recent publications. In so doing we hoped to be able to provide answers to the following questions.
Could the association between wire codes and childhood leukaemia be confirmed?
Is there an association between childhood leukaemia and other markers of residential EMF exposure?
What is the overall quality of existing studies and is there a relationship between the quality of studies and the magnitude of risk?
What recommendations can we make for further studies, if warranted?
Materials and methods
Identification of relevant studies
Studies pertaining to the relationship between EMFs and childhood leukaemia were identified using a MEDLINE search of the medical literature pubfished in English over the period 1966-98. Copies of the relevant articles were obtained and reviewed to identify additional references.
To be eligible for inclusion in the metaanalyses, studies had to satisfy the following criteria: be primary studies, not reanalyses or reviews; be of case-control, cohort, or cross-sectional design; examine residential-based exposures through wiring configuration codes, distance to power distribution equipment, spot and 24-h measures of magnetic field strength (magnetic flux density), and calculated magnetic fields; examine childhood leukaemia; report an odds ratio and its variance or sufficient data to estimate them; be in English, and be published before January 1999.
Quality assessment
Each article was blinded with regard to authors, institution, and journal. The articles were read and scored for quality by two independent readers using a system that incorporates elements of methods developed by Chalmers et al. (12), Longnecker et al. (13), Morris et al. (14) and Villari et al. (15). The criteria employed are shown in the results section below. A quality score was calculated as the percentage of applicable criteria that were met in each study. Items concerned with efforts to minimize potential bias were given twice the weight of those evaluating data analysis.
Data extraction
A number of different methods were used to measure EMF exposure in the studies examined. In the absence of a unique criterion, not all studies examining EMF exposure and childhood leukaemia risk could be included in a single analysis. Exposure assessment methods were aggregated into the following categories: wiring configuration codes; distance to power distribution equipment; spot and 24-h measures of magnetic field strength; calculated indices of magnetic field strength using distance to power distribution equipment; and historic load data. To aggregate exposure categories across studies, we dichotomized exposure strata using the cut-off points that were common to most of the studies. More specifically, in studies where the exposure status was assessed through the wiring code or distance criterion, subjects were considered exposed if they were living in homes with high current configurations or at a distance of less than 50 m from any electrical sources, respectively. Similarly, in studies where the assessment of exposure was made via spot, 24-h measurements or historical calculation of magnetic field strength, a cut-off point of 0.2 pT was used.
For each blinded study, data were extracted in a contingency table format, and odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Data extraction was carried out independently by two readers. After completing data extraction, the two readers met to resolve any differences and arrive at a consensus.
Statistical pooling
The random effects model described by DerSimonian & Laird (16) was used to combine the collected values. This procedure yields a single estimate of the OR for leukaemia in children exposed to EMFs compared with nonexposed children. It also enables testing the homogeneity across the individual studies (using the Q statistics); the heterogeneity, if not zero, is then incorporated into the pooled variance estimate.
We avoided pooling results obtained through the use of different exposure criteria, and instead carried out separate meta-analyses based on exposure assessment methods, using the cut-off points specified above. Studies that used more than one method of ENIF measurement were included in more than one meta-analysis. In order to test all decision rules that we used in extracting OR data, we performed additional meta-analyses using, for each exposure assessment method, data that in the single studies gave the lowest OR (best scenario) and the highest OR (worst scenario).
Finally, studies were also divided into two groups and analysed according to their quality score: the potential impact of the quality of studies on the results was assessed by comparing pooled results from studies with scores above the median to studies whose scores were equal to or below the median.
Results
Literature search
The literature search identified 14 case-control studies (1, 17-29) and one cohort study (30) that investigated the relationship between residential magnetic-field exposures and childhood leukaemia and which met our inclusion criteria. Of these studies, 10 employed only one method of EMF measurement (1, 17, 18, 20, 22, 24, 26, 27, 29, 30), three used two methods (19, 25, 28), and two studies used three methods (21, 23). Thus, six studies were available for the meta-analysis evaluating exposure through wiring configuration codes (1, 17,19,21, 25, 27) and five studies for the meta-analysis in which the exposure was determined by spot measures (18, 19, 21, 23, 28), whereas meta-analyses based on distance from electrical sources (20, 23, 24, 26), 24-h measurements of magnetic fields (21, 25, 28, 29) and calculated magnetic fields (22, 23, 26, 30) comprised four studies each.
The inclusion of the studies by Tomenius (18) and Feychting & Ahlbom (23) in the meta-analysis based on spot measurements of magnetic fields was problematic, since a partial overlap of their data cannot be excluded. Therefore, we included in our main analysis only the Feychting & Ahlbom study, testing the impact of this choice in a sensitivity analysis.
Magnetic field strength was not dichotomized at the 0.2 pT level in four studies reporting spot or 24-h measures and calculated indices (Tomenius (18) = 0.3 gT; London et al. (21) = 0.125 and 0.264 @iT for spot and 24-h measurements, respectively; Olsen et al. (22) = 0.25 gT; Tynes & Haldorsen (26) = 0.14 gT). In order to include in the meta-analyses all available data, we assumed that the exposure cut-off points used in these studies were comparable to the 0.2 [tT exposure.
Quality assessment
Table 1 shows the results of the quality scoring procedure. The potential for selection bias may be a concern in the individual studies considered. Although 11 of the 14 case-control studies used population-based cancer registries to identify cases, there are a number of important concerns about the control selection. Only five studies used a population register to identify controls. If the ascertainment of cases as well as the population register is complete, these studies should be free of selection biases. In most studies, however, controls were selected in less desirable ways, such as using regional birth certificate files, random digit dialing or other cancer cases. Response rates for cases and controls were frequently less than 75%, particularly in studies requiring subject interviews or magnetic field measurements.
Misclassification bias cannot be considered negligible. Although few studies specified disease criteria or clearly stated that cancer diagnosis was validated bv histology or some other gold standard, diseases like leukaemia are subject to relatively little misclassification (false negatives are unlikely given the severity of the disease, and false positives are unlikely given the medical scrutiny of suspected cases). By contrast, exposure misclassification is a pervasive concern in case-control studies of the effect of exposure to EMFs. Unfortunately, not all studies made serious efforts to collect as much of the exposure data as possible while being unaware of the case-control status of the subjects. Use of such a procedure does not ensure the absence of errors but makes it highly probable that errors would be independent of case or control status and therefore that the results would be biased towards the null.
Information bias associated with failure to consider confounding variables may have been more of a problem. With the exception of age (100% of studies), sex (93%) and socioeconomic status (87%), less than half of the studies considered potentially confounding variables such as traffic density (40%), parental occupational exposure to EMFs (20%) or other competing carcinogenic exposures (40%). Although attempts made to adjust for those variables have produced evidence against the presence of substantial confounding, such attempts are severely hampered by the scarcity, of established or even strongly suspected causes of childhood leukaemia.
Statistical analysis was judged appropriate in most studies. Although the majority of studies listed P values and/or 95% CI, very few listed or analysed demographic data. Only two studies provided power calculations.
Quality scores for the individual studies ranged from 0.33 to 0.82, and there was a clear tendency for more recent studies to be of better quality (Fig. 1).
Statistical analysis
Table 2 lists the OR estimates with their respective 95% Cls extracted from each study and shows the results of pooling data from all studies according to the five different methods of exposure assessment. Although 17 of these 23 estimates (73.9%) had ORs greater than 1.00, only 5 (21.7%) were significantly greater than this (P < 0.05). The overall OR estimates of these metaanalyses were always greater than 1.00, indicating a potential effect of residential EMF exposure on childhood leukaemia risk. However, this effect is statistically significant only for meta-analyses based on wiring configuration codes (Pooled OR = 1.46; 95% CI = 1.05-2.04; P = 0.024) and 24-h measurements of magnetic fields (pooled OR = 1.59; 95% Cl = 1.142.22; P= 0. 006). In the meta-analysis of studies in which the exposure assessment was made through spot measurements of magnetic field, the exclusion of the Feychting & Ahlborn study (23) and the inclusion of that by Tomenius (18) decreased the overall OR below 1.00 (pooled OR = 0.94; 95% CI = 0.50-1.77; P = 0.850). Individual studies included in the different meta-analyses appear heterogeneous (P < 0.05) only in the case of exposure assessment performed with wiring configuration codes or by distance.
Even in the presence of predefined and wellestablished criteria, we encountered in many studies more than one possibility for data extraction. For example, in the study by Wertheimer & Leeper (1), exposure assessment was performed both at birth and death addresses, which resulted in OR estimates of 2.28 and 2.97, respectively; Linet et al. (25) calculated results using unmatched analysis as well as analysis of matched case-control pairs, with corresponding OR estimates of 1.19 and 1.39. To test the sensitivity of our results to the different choices of data extraction, we performed separate meta-analyses using, for each exposure assessment method, data that gave the lowest (best scenario) and the highest (worst scenario) OR estimates. The impact of the different choices of data extraction did not seem to be substantial (Table 3). The overall ORs resulting from studies based on wiring configuration codes and 24-h measurements of magnetic fields were always significantly different from 1.00, ranging from 1.33 to 1.58 and from 1.50 to 1.95, respectively. Interestingly, the heterogeneity of studies included in the meta-analysis based on wiring configuration codes, compared with the baseline analysis, tended to decrease, and in the best-scenario hypothesis, became not significant (P < 0.05). In contrast, the heterogeneity of studies in which the exposure was assessed by 24-h FMF measures, although not significant, tended to increase slightly either in the best-case or worst-case scenario.
The quality of the studies does seem to have a substantial impact on our summary estimates, particularly in meta-analyses based on wiring configuration codes, 24-h measurements of magnetic fields, and calculated magnetic fields, since high-quality-- score studies have lower risk estimates than lowquality-score studies (Table 4).
Discussion
By the end of 1998,15 studies had been published that provided relevant data on the association between residential exposure to EMF and childhood leukaemia: nine in Europe, five in the USA, and one in New Zealand. The majority of these studies appeared between 1986 and 1993, and five studies were published in the period 1997-98. M but one were case-control studies, most of which were based on a comprehensive case ascertainment in a geographically defined population. Exposure assessment was based on a variety of methods, including wiring configuration codes, distance to power distribution equipment, spot and 24-h measures of magnetic field strength, and calculated indices using distance to power distribution equipment and historical load data. One-third of the studies employed more than one method of EMF measurement.
In the present investigation, pooling results arising from the use of different exposure criteria was avoided, and we performed separate meta-analyses for each method of exposure assessment. The metaanalysis based on wiring configuration codes confirmed the 1.5-fold statistically significant excess of childhood leukaemia already documented in previous studies (7-11). The exposure rates among control subjects in the individual studies varied from 6.9% to 50%, and the heterogeneity was found to be statistically significant. The ca. 1.5-fold risk of childhood leukaemia was also reported in metaanalyses based on calculated magnetic fields and 24-h measurements of magnetic fields. This excess risk was significant only in the analysis of studies relying on 24-h measurements of magnetic fields, in which the exposure rate among controls varied from 1.9% to 9.5%, and the heterogeneity was not statistically significant. Meta-analyses based on distance and spot measurements of magnetic fields produced ORs of lower magnitude and not significantly different from unity.
Other meta-analyses of the association between exposure to residential electromagnetic fields and childhood leukaemia have been carried out. For example, Washburn et at. (6), combining results of studies published before 1992, found increased risks for leukaemia, lymphoma, and nervous system cancers, although the risk of lymphoma was not significant. Miller et al. (7), in separate meta-analyses of studies published before 1993 according to EMF measurement methods, documented statistically significant increased risks for childhood leukaemia for wiring configuration codes, and distance and calculated indices, whereas spot measures consistently showed non-significant odds ratios. On the basis of studies published before 1994, Meinert & Michaelis (8) confirmed the significant association between childhood leukaemia and residential EMF exposure measured through wiring configuration codes, whereas no association was found with distance; the meta-analysis of studies in which the ENIF exposure was either measured directly or calculated did not show an increase of childhood leukaemia with higher cut-off points. More recently, \Wartenberg (9, 11) documented that wiring codes and related markers of exposure, such as proximity to power lines and calculated magnetic fields from power tines, were associated with an approximate 1.5-fold excess risk of childhood leukaemia, whereas the evidence of an association with magnetic fields measured directly was not, in the aggregate, supported. None of these previous meta-analyses provided overall risk estimates from studies in which the exposure assessment was performed through 24-h measurements of magnetic fields.
The quality of the individual studies included in our meta-analyses was assessed on the basis of their statistical analyses and the efforts made to minimize potential for selection bias, misclassification bias related to exposure as well as disease, and information bias due to failure to consider potential confounding variables. Items concerned with efforts to minimize potential bias were given twice the weight of items evaluating data analysis. Since there appears to be no "gold standard" at present for EMF measurement, we did not evaluate the operational definition of exposure (e.g., all EMF measurements methods were assumed to be equally valid). It is well known that all quality assessment systems have a subjective component, none have yet been validated, and efforts to correlate quality scores with direction or size of effect have had mixed findings (31, 32). Therefore, we did not use quality scores to determine studies to be included in the meta-analysis or to assign statistical weights.
Despite these limitations, assessment of the quality of the individual studies used in our metaanalysis allowed us to draw the following conclusions:
there has been improvement in study design and reporting, since findings published more recently tended to receive a higher quality rating;
the possibility of selection bias, misclassification bias related to exposure, and information bias related to failure to consider potential confounders cannot be ruled out;
most importantly, pooling high-quality-score studies resulted in lower risk estimates than did pooling low-quality-score studies.
If high-quality studies are more likely to yield valid information than low-quality studies, we can conclude that currently available data do not permit exact quantification of the true excess risk of childhood leukaemia due to residential EMF exposure.
An important limitation of the meta-analytical approach is related to publication bias. This occurs if positive results are more likely to be published than negative ones (33-35). In the case of residential EMF exposure and childhood leukaemia, there is an important factor that may mitigate the tendency for negative findings to be excluded from the published literature. In view of the considerable interest in this topic, it seems unlikely that any investigator would have trouble in getting even a negative study published. Indeed, the great majority of the studies that we included in our meta-analyses reported risk estimates with P values greater than 0.05, suggesting that non-significant results are readily publishable.
Several conclusions may be drawn from the results of this study.
First, an association between residential EMF exposure and childhood leukaemia may exist. This possibility is supported by the statistically significant risk estimates obtained by pooling results of studies in which the exposure was assessed not only indirectly with markers such as wiring configuration codes but also through direct measurements of magnetic field for at least a 24-h period.
Second, the magnitude of this excess risk, if any, is at present unknown, given the possibility of selection bias, exposure misclassification, and the existence of confounding variables in the individual studies.
Third, there appears to be a clear trend for the more recent publications to be of better quality. If this trend continues, new good-quality studies can be expected in the future.
Finally, enough evidence exists to lead us to conclude that dismissing concerns about EMF and childhood leukaemia is unwarranted. What is required is the publication of new state-of-the-art epidemiological studies that incorporate comparable measures for both exposure and outcomes in order to facilitate future meta-analyses. If this excess risk of childhood leukaemia is confirmed, we should thoroughly investigate, also from a cost-effectiveness point of view, possible options for minimizing exposure in order to provide definitive answers for policy-makers.
Resume
Meta-analyse de la relation entre leucemie de 1'enfant et exposition A des champs electromagnetiques du fait du lieu de residence
Les etudes epidemiologiques consacrees a la relation entre leucemie cle 1'enfant et exposition a des champs electromagnetiques du fait du lieu cle residence sont suggestives sans toutefois kre concluantes. Un certain nombre d'entre elles se pretent cependant a la technique cle la meta-analyse. Nous avons effectue une recherche bibliographique au moyen cle MEDLINE et d'autres sources de donnees pour retenir 14 etudes cas-temoins et une etude de cohorte dont nous avons MW la qualite epidemiologique et que nous avons soumises ensuite 6 une meta-analyse. Nous avons tire cle chacune d'entre elles une estimation du risque relatif et nous avons reuni ces valeurs. Une meta-analyse distincte a ete effectuee selon le mode d'6valuation du champ electromagnetique (code cle configuration du bobinage, distance au centre cle distribution, mesures ponctuelles ou sur 24 h de l'intensite du champ magnetique (densite de flux magnetique) ou determination cle ce champ par le calcul). La meta-analyse basee sur les codes cle configuration a donne une estimation combinee du risque relatif egale a 1,46 (intervalle de confiance (IC) @ 95% = 1,05-2,04, p = 0,024); celle qui prenait en compte la mesure du champ sur 24 h donnant une valeur de 1,59 (IC a 95 % = 1,14-2,22, p = 0,006 ) et indiquant donc la possibilite d'une relation entre exposition au champ magnetique et leucemie chez l'enfant). Dans la plupart des cas, le regroupement des resultats des etudes de tres bonne qualite a donne une estimation du risque plus faible que dans le cas des 6tudes de qualite mediocre. On constate une nette tendance a I'amelioration de la qualite dans les etudes recentes. On a en tout cas suff isam ment de preuves pour pouvoir conclure qu'il West pas justifie de faire bon marche des craintes qui se sont exprimees au sujet du risque de leucemie chez les enfants exposes a des champs electromagnetiques du fait de leur lieu de residence. 11 est necessaire d'effectuer d'autres etudes epidemiologiques de tres bonne quake basees sur des mesures comparables de 1'exposition et de son resultat pour pouvoir confirmer ces observations. Dans l'eventualit6 d'une confirmation de cet exces de risque, il faudrait etudier minutieusement les possibilites de reduction de 1'exposition afin que les decideurs puissent disposer cle conclusions definitives.
Resumen
Exposicion a campos electromagneticos en zonas de residencia y leucemia infantil: metanalisis
Las diversas investigaciones epidemiol6gicas realizadas sobre la relacion entre la exposicion a campos electromagneticos (CEM) en zonas de residencia y la leucemia infantil se han saldado con indicios no concluyentes. Varios de esos trabajos, sin embargo, se prestan a ser estudiados mediante tecnicas de metanalisis. En el presente metanalisis se incluyeron 14 estudios de casos y testigos y un estudio de cohortes identificados a travis de MEDLINE y de otras fuentes, previa evaluaci6n de su calidad epidemiol6gica. Se combinaron las estimaciones del riesgo relativo obtenidas en cada uno de los estudios y se realizaron diversos metano lisis basados en las distintas evaluaciones de la exposicion a los CEM (c6digos de configuraci6n del cableado, distancia a las instalaciones de distribuci6n de la energia, mediciones puntuales y de 24 horas de la potencia de los campos electromagneticos (densidad de flujo magnetico), campos magneticos calculados). El metan6lisis basado en los c6digos de configuracion del cableado arroj6 una estimaci6n del riesgo relativo de 1,46 (IC95%: 1,05 2,04, p = 0,024), y en el caso de las mediciones de 24 horas de los campos electromagneticos se obtuvo un valor de 1,59 (IC95%: 1,14-2,22, p = 0,006), lo que indica una posible relacion entre la exposicion a CEM en zonas cle residencia y la aparici6n de leucemia infantil. En la mayoria cle los casos las combinaciones cle estudios cle alta calidad dieron estimaciones del riesgo mas bajas que las combinaciones cle estudios cle baja calidad. Se observa una clara tendencia a una mayor calidad en los estudios mas recientes. Existen datos suficientes para concluir que no puede descartarse una relacion entre la influencia cle los campos electromagneticos en la zona cle residencia y la leucemia infantil. Es necesario realizar nuevos estudios epiclemiologicos de gran calidad, con mediciones comparables tanto cle la exposici6n como cle los resultados, para corroborar estos resultados. Si se confirmara el posible exceso cle riesgo, habria que investigar a fondo las alternativas para reducir al minimo la exposicion y proporcionar respuestas definitivas a los formuladores de politicas.
References
Wertheimer N, Leeper E. Electrical wiring configurations and childhood cancer. American journal of epidemiology, 1979, 109: 273-284.
Campion EW. Power lines, cancer, and fear. New England journal of medicine, 1997, 337: 44-46.
ELF-EMF European Feasibility Study Group. Need for a European approach to the effects of extremely low-frequency electromagnetic fields on cancer. Scandinavian journal of work, environment and health, 1997, 14: 23-25.
Wartenberg D. Leukaemia and exposure to magnetic field. New Englandjournal of medicine, 1997, 337: 1471,
Blettner Met al. Traditional reviews, meta-analyses, and pooled analyses in epidemiology. Internationa/journal of epidemio/ogy, 1999, 28: 1-9.
Washburn EP et aL Residential proximity to electricity transmission and distribution equipment and risk of childhood leukaemia, childhood lymphoma, and childhood nervous system tumors: systematic review, evaluation, and meta-analysis. Cancer causes and control, 1994, 5: 299-309.
Miller MA et al. Variation in cancer risk estimates for exposure to powerline frequency electromagnetic fields: a meta-analysis comparing ENIF measurement methods. Risk analysis, 1995, 15:281-287.
Meinert R, Michaelis J. Meta-analyses of studies on the association between electromagnetic fields and childhood cancer. Radiation and environmental biophysics, 1996, 35: 11-18.
National Research Council. Possible health effects of exposure to residential electric and magnetIc fields. Washington, DC, National Academy Press, 1997,
. Angelillo IF, Ferreira G, Pavia M @Exposure to electromagnetic fields and effects on human health: a meta-analysis]. Proceedings of the 38th National Meeting of the Italian Public Health Association (Fiuggi, Italy, 27-30 September 1998). Annali di ig ine, medicina preventIva e di comunita, 1998, 10 (Suppl. 1 ): 115-141 (in Italian, summary in English).
Wartenberg D. Residential magnetic fields and childhood leukaemia: a meta-analysis. American journal of public health, 1998,88:1787-1794.
Chalmers TC et al. A method for assessing the quality of a randomized trial. Controlled clinical trials, 1981, 2: 31-49,
Longnecker MP et al. A meta-analysis of alcohol consumption in relation to risk of breast cancer. Journa/of the Amerlcan Medical Association, 1988, 260: 652-656.
Morris RD et al. Chlorination, chlorination by-products and cancer: a meta-analysis. American journal of public hea/th, 1992, 82:955-963.
Villari P et al. Cesarean section to reduce perinatal transmission of human immunodeficiency virus: a meta-analysis. The online /burna/of current clinical trials, 1993, Jul 8: Doc. No. 74. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled-clin/ca/tri,715, 1986, 7: 177-188.
Fulton JP et al. Electrical wiring configurations and childhood leukaemia in Rhode Island. American journal of epidemiology, 1980,111:292-296.
Tomenius L. SO-Hz electromagnetic environment and the incidence of childhood tumors in Stockholm County. Bioelectromagnetics, 1986, 7: 191-207.
Savitz DA et al. Case-control study of childhood cancer and exposure to 60-Hz magnetic fields. American journal of epidemiology, 1988, 128: 21-38.
Coleman MP et al. Leukaemia and residence near electricity transmission equipment: a case-control study. British journal of cancer, 1989, 60: 793-798.
London SJ et al. Exposure to residential electric and magnetic fields and risk of childhood leukaemia. Americanjournal of epidem/o/ogy, 1991, 134: 923-937.
Olsen JH, Nielsen Schulgen G. Residence near high voltage facilities and risk of cancer in children. British medicalJournal, 1993.307:891-895.
Feychting M, Ahlbom A. Magnetic fields and cancer in children residing near Swedish high-voltage power line. American journal of epidemiology, 1993, 138: 467-481,
Petridou E et al. Age of exposure to infections and risk of childhood leukaemia. British medica/journal, 1993, 307: 774. Linet MS et al. Residential exposure to magnetic fields and acute lymphoblastic leukaemia in children. New Englandjournal of medicine, 1997, 337: 1-7.
Tynes T, Haldorsen T. Electromagnetic fields and cancer in children residing near Norwegian high-voltage power lines. American journal of epidemiology, 1997, 145: 219-226. Petridou E et al. Electrical power lines and childhood leukaemia: a study from Greece, Internationa/journalof cancer, 1997,
73: 345-348.
Michaelis J et al. Combined risk estimates for two German population-based case-control studies on residential magnetic fields and childhood acute leukaemia. Epidemiolqqy 1997,
9: 92-94.
Dockerty JD et al. Electromagnetic field exposures and childhood cancers in New Zealand. Cancer causes and control, 1998,9:299-309.
Verkasalo PK et al. Risk of cancer in Finnish children living close to power lines. gr/ti@h medi@a/journal, 1993, 307: 895-899.
Berlin JA, Colditz GA. A meta-analysis of physical activity in the prevention of coronary hearth disease. American/journal of epidemiology, 19 90, 132: 612 -628.
Emerson JD et all. An empirical study of the possible relation of treatment differences to quality scores in controlled randomized clinical trials. Controlled clinical trials, 1990, 11: 3 30@39-352, Begg CB, Berlin JA. Publication bias and dissemination of clinical research. Journal of National Cancer Institute, 1989,
81: 107-115.
Easterbrook PJ et al. Publication bias in clinical research. lancet, 1991, 337: 867-872.
Dickersin K, Min Yl, Meinert CL. Factors influencing publication of research results: follow-up of applications submitted to two institutional review boards. Journal of American Medical Association, 1992, 267: 374-378.
I.F. Angelillo1 & R Villari2
1 Associate Professor, Chair of Hygiene, Medical School, University Magna Graecia, Via Tommaso Campanella, 88100 Catanzaro, Italy. Correspondence should be addressed to this author (e-mail: igiene@thebrain. net).
2 Assistant Professor, Department of Health and Preventive Sciences, University Federico 11, Naples, Italy.
Copyright World Health Organization 1999