EXPOSURE TO EXTREMELY LOW-FREQUENCY MAGNETIC FIELDS IN LOW- AND MIDDLE-INCOME COUNTRIES: AN OVERVIEW

Abstract To compare extremely low-frequency magnetic field (ELF-MF) exposure in the general population in low- and middle-income countries (LMICs) with high-income countries (HIC), we carried out a systematic literature search resulting in 1483 potentially eligible articles; however, only 25 studies could be included in the qualitative synthesis. Studies showed large heterogeneity in design, exposure environment and exposure assessment. Exposure assessed by outdoor spot measurements ranged from 0.03 to 4μT. Average exposure by indoor spot measurements in homes ranged from 0.02 to 0.4μT. Proportions of homes exposed to a threshold of ≥0.3μT were many times higher in LMICs compared to HIC. Based on the limited data available, exposure to ELF-MF in LMICs appeared higher than in HIC, but a direct comparison is hampered by a lack of representative and systematic monitoring studies. Representative measurement studies on residential exposure to ELF-MF are needed in LMICs together with better standardisation in the reporting.


INTRODUCTION
The International Agency for Research on Cancer (IARC) classified exposure to extremely lowfrequency magnetic fields (ELF-MF) as possibly carcinogenic to humans (group 2B) in 2001 (1) ; this assessment has been more recently confirmed by the European Commission's Scientific Committee on Emerging and Newly Identified Health Risks (SCENIHR) (2) . The IARC and SCENIHR classifications are mainly based on epidemiological studies of childhood leukemia. So far, experimental studies on animals failed to convincingly confirm an increased risk for leukemia (3) , and no plausible biophysical mechanisms have yet been identified (4,5) . A potential risk of ELF-MF for childhood leukemia has been investigated in high-income countries (HICs) in numerous studies in over 30 years of research (6) , two pooled analyses from 2000 showed significantly increased risks when children exposed to daily average ELF-MF of ≥0.4 μT (7) and ≥0.3 μT (8) were compared to children in the reference group with daily average exposure ≤0.1 μT. For children exposed to ≥0.4 μT, the relative risk was 2.0 (95% confidence interval (CI) 1.3-3.1) (7) and for children exposed to ≥0.3 μT, the odds ratio (OR) was 1.7 (95% CI 1.2-2.3) (8) .
The most recent pooled analysis, which included studies published between 2000 and 2010 also suggested an increased risk for children exposed to daily average ELF-MF levels of ≥0.3 μT (OR 1.44, 95% CI 0.88-2.36), although not statistically significant (9) . Studies included in the mentioned pooled analyses were all from HICs like Germany (10) , UK (11) or Sweden (12) , with the exception of data from one study from Brazil (13) . One percent or less of the children were categorised in the highest exposure group in the pooled analysis of Ahlbom and colleagues (7) . This shows that exposure to higher levels of ELF-MF is uncommon in HICs. Due to a different state of technical development, housing conditions and legal regulations, exposure levels in the general population and children in lowand middle-income countries (LMICs) could be different. In addition, LMICs typically have a higher proportion of children in the population. Taken together, more children in LMICs compared to HICs may be exposed to higher levels of ELF-MF as a potential risk factor to leukemia. Therefore, it is key to collect high-quality information on exposure to ELF-MF in these countries.
Although of the worldwide 1.958 billion children (age 0-14 years old), 1.758 billion live in LMICs (14) , we were not aware of any overview of exposure levels to ELF-MF in LMICs in the general population including children. For this reason, we compiled the first overview of studies from LMICs by applying techniques of a systematic review. In the following, we describe their main characteristics and results and discuss our findings in the context of results from measurement surveys conducted in HIC.

Eligibility
All studies listed in PubMed and Web of Science (WoS) which included information on exposure to ELF-MF and reported on LMICs were eligible for inclusion. Specifically, we included studies reporting on exposure to ELF-MF with frequencies up to 300 Hz (1) , which reported measured or calculated magnetic fields for the general population, for children or for areas close to residential areas. Studies reporting on distance to power lines or substations as an exposure metric were also included. We excluded studies on occupational exposure or exposure from specific devices, cars or trains. We did not apply any restrictions in terms of health outcomes, study design, language, study size or time period. LMICs were defined in accordance with the definition of the World Bank (WB), based on the gross national income per capita in 2018. The LMICs were classified into three groups: low-income countries, lower middle income countries and upper middle income countries, as defined by the WB (15) .

Search strategy
Electronic search was performed in PubMed and WoS. For the search in PubMed and WoS, we used a combination of keywords for ELF-MF and the name of each LMICs. A more detailed description of the search terms is included in 'Supplementary A1'.
In addition to the systematic approach of the searches in PubMed and WoS, we used further extensive search techniques to identify all potentially relevant articles for this review: we conducted an informal survey among WHO-experts for electromagnetic fields (using a respective email distribution list) whom we asked for studies relevant to our research question; we included 'snowballing' methods, that is, screening the reference lists of the included studies for additional relevant articles, and; we systematically examined the issues of the last 2 years of the journal that published the majority of the included articles.

Study selection and data extraction
After removing duplicates, all titles and abstracts identified by the search were screened for relevance. Full texts of potentially eligible articles were reviewed for inclusion. Studies fulfilling all eligibility criteria were included. Study selection was done by one reviewer (DB). Relevant data were extracted by one reviewer (DB) using a predefined data extraction form, recording information on the author, year of publication, country, study type, information on measurement device and exposure assessment technique, and major results of the exposure assessment. When needed, ELF-MF units were converted from Gauss to Tesla, and if results were only provided graphically, values were abstracted from graphs, whenever possible (16,17) .

Study selection
The systematic search in PubMed and WoS was done in October 2019. The flowchart for the process of study selection is presented in Figure 1. After removing duplicates from the initial search, 1483 articles were selected for the title and abstract screening, out of which 92 articles were retained for full-text screening. Evaluation of the 92 full texts left 23 original studies. Reasons for exclusion of full texts were: occupational exposure (n = 31), investigated exposure was not ELF-MF (n = 11), review articles not on exposure in LMICs (n = 8), the country the study was conducted in did not belong to LMICs (n = 6), studies on methodological aspects of measurements (n = 6) or other reasons (n = 7). Four of these latter seven articles reported on measurements in cars, trains or of specific devices, two on simulation studies and one on an in-vitro experimental study.
We added two articles that we identified through our additional extensive searches. One of these (18) was identified by checking the issues of the last 2 years (November 2017-October 2019) of the journal Radiation Protection Dosimetry. The other article has been obtained through an informal survey among experts in the field (19) . Therefore, our overview included a total of 25 studies.
For the remainder of our overview, we considered separately studies with personal measurements (n = 1), with spot measurements (outdoor spot measurements (n = 8), indoor spot measurements (n = 6) and studies that reported separately on both (n = 4)) and studies with other exposure assessment methods (n = 4) ( Table 1). One study (28) with indoor and outdoor measurements was excluded from assignment to these categories as the authors did not report separate results from the measurement of indoor and outdoor exposure, and another study was excluded because we could not extract the exposure data from the figures with the required accuracy (30) .

Personal measurements
One study reported 24-h personal measurements for 128 pregnant women in China (21) . Measurements were carried out using the EMDEX Lite meter as a measurement device, located at the waist in the daytime and next to the beds while sleeping. Measurements were taken every 4 s. A median of the time-weighted average of 0.06 μT was observed in these women.

Outdoor spot measurements
Characteristics and major findings of all 12 studies with reported measurements of outdoor exposure to ELF-MF are displayed in Table 2. Four studies (16,19,20,26) reported measured values in the vicinity of power lines of different voltages ranging from 66 to 400 kV measured at 1 m above the ground.
The results of the measurements ranged from 0.03 μT for a measurement of a 90 kV double-circuit power line (16) to 3.5 μT observed in the vicinity of a 66 kV power line (19) . Four studies (13,28,33,35) reported exposure measurements outside the homes of participants. In the study of Wunsch-Filho et al. (13) , 14.8% of the measurements at the front door were above 0.3 μT, while 13.3% of the measurements from Wang et al. (29) were above 0.4 μT at the front door. The other two studies (33,35) reported measured values outside the house depending on the distance to power lines. A house located 24 m away from a 400 kV power line yielded the highest measured value of 4 μT (33) . In studies on exposure of the general public in residential areas (23,24,34,38) , the mean exposure ranged from 0.15 μT obtained in a series of 74 outdoor measurements in an urban area in Romania (23) to 2.18 μT observed below a 500 kV power line in a village of a rural area in Brazil (38) . Three studies used the same measurement device (16,33,34) , while the rest of the studies used different devices (Table 2).

Indoor spot measurements
We described the characteristics and major findings of the studies with indoor spot measurements (n = 10) in Table 3. In most of the studies, measurements were performed inside homes (n = 7). Two studies reported on measurements inside schools (25,27) and one study on exposure during housekeeping, most likely an indoor activity (38) . Two studies (13,18) with measurements in homes provided information on the distribution of the measured ELF-MF exposure values: 24-h measurements under children's beds showed that 6.19% of 727 measurements were ≥0.3 μT (13) . In a study with measurements in apartments in buildings under normal power use, 19% were >0.1 μT and 13% between 0.3-0.4 μT (18) . Average values of all studies with indoor measurements in homes ranged from 0.02 (23,32) to 0.4 μT, which was observed in houses of women with unexplained spontaneous abortions in Iran (31) . There were different settings for the exposure assessment between the studies with measurements in homes including measurements in homes near substations (23) , homes under 110 kV, 220 kV and 400 kV power lines (33) or measurements in apartment buildings with built-in transformer stations or indoor power substations (17,18) . The study on apartment buildings with built-in transformer station measured an average exposure of ∼0.28 μT at 1 m above the floor in apartments directly above or next to the transformer with maximum values of 0.65 μT at 0.5 m height (17) . Other studies with maximum exposure measurements exceeding 0.4 μT reported maximum values of 0.45 μT in a residence with an indoor power substation (18) and 3.2 μT for a house under a 400 kV power line (33) . One of the two studies on measurements in schools reported on ELF-MF depending on the distance to substations (27) . Schools close to substations (30-50 m) showed higher average exposure levels compared to schools far away from substations (610-1390 m) ( Table 3). Mean exposure of measurements in 60 classrooms in Bangkok (25) was 0.11 μT, with 21.67% of the classrooms having an exposure level above 0.2 μT. Only two studies used the same measurement device, the EMDEX-II dosemeter (13,17) .
Feizi & Arabi (37) investigated the potential risk of 123, 230 and 400 kV power lines on acute childhood leukemia in Iran. They calculated the exposure based on the mean intensity of the electrical current and additional line characteristics for 119 children of whom 16 lived in a distance of ≤500 m to a power line. In total, 16.8% of all children were exposed ≥0.45 μT and 83.2% were exposed to <0.45 μT. Lunca et al. (22) calculated the ELF-MF from 400 kV overhead power lines with the software tools PowerMag and Pow-erELT. They report various results of calculated magnetic flux density stratified for single-circuit lines and double-circuit lines. They conclude that the typical levels under 400 kV single-circuit power lines at 1 m above the ground level are 5 μT and under doublecircuit power lines, 4.5 μT. Fadel et al. (36) compared 390 children living <50 m close to a power line with a control group of 390 children from another area in Egypt. Although the authors stated that they calculated the exposure, they did not report results for the calculations. One study in Mexico assessed the distance to transformers, high-voltage power lines and electric substations as an approximation for the expo- Tourab & Babouri (20) Exposure environment: in the vicinity of 220 kV power lines (n = 2) in a city. Measurements taken at 0, 1, 1.  (19) Exposure environment: in the vicinity of 66 kV power lines (n = 40 lines). Measurements taken at 1 m above the ground over a 6 min period.

Spectrum Analyzer NF-5035
Max. and min. mean values: 3.5 μT, 0.89 μT (Continued)  (35) Exposure environment: residential areas of 66 cases of childhood leukemia with information on distance to electric transformers and power lines. Rathebe et al. (24) Exposure environment: residential areas near substations. Measurements at 1 m above the ground for 0, 3, 6 and 9 m distances to substations.

TriField meter model XE 100
Mean, range, and SD at various distances between substation and residential area:  (38) Exposure environment: Amazon village near 500 kV power lines.  sure to ELF-MF to investigate the risk of childhood leukemia (39) .

Brief summary
The goal of our review was to give the first comprehensive literature overview on exposure to ELF-MF in the general population including children of LMICs by applying methods of a systematic review. We identified 25 studies published between 1993 and 2019 in total, consisting of 21 studies from upper middle income countries, four studies from lower middle income countries and not a single study from LIC. Eighteen studies reported extractable results of spot measurements to estimate the exposure to ELF-MF using outdoor spot measurements (n = 8), indoor spot measurements (n = 6) or both (n = 4). The included studies showed a large heterogeneity in their design, exposure environment, exposure assessment and reported summary statistics, which severely hampered their comparability. Single outdoor spot measurements for ELF-MF ranged from 0.03 μT to a maximum of 4 μT. Average exposure from indoor spot measurements in homes ranged from 0.02 to 0.4 μT.

Comparison with results from HICs
For our comparison of studies from LMICs and HIC, we did not take into account studies that limited their exposure measurements to specific power lines with no reference to proximity to residential areas. Also, we described these studies earlier because the values reported in those studies and measured under power lines could give an idea of the potential exposure in the respective countries, when close to residential areas.
In a review, the WHO estimated the exposure to residential ELF-MF in the USA and Europe (40) . In the USA, the geometric mean of the magnetic fields over one day ranged between 0.06 and 0.11 μT in homes. For Europe, consisting of HIC, the geometric mean of the magnetic fields was lower, estimated to be in the range of 0.03 to 0.07 μT. The WHO also reported the proportion of children being exposed above the thresholds of 0.3-0.4 μT, levels which are possibly associated with an increased leukemia risk in epidemiological studies: (7,8) between 1 and 4% of children were estimated to have exposures ≥0.3 μT and 1-2% of children exposures above 0.4 μT in HIC. Another study assessing the potential health impacts of residential exposure to ELF-MF in Europe (41) estimated the distribution of exposure to residential ELF-MF based on a literature overview. They reported that 0.54% of the general population was exposed to a geometric mean exposure of above 0.3 μT. A measurement survey on residential exposure to ELF-MF in Australia (42) consisting of spot measurements in 296 randomly selected homes in Melbourne showed consistency with the estimates for the USA and Europe reported by the WHO. The average fields were 0.05-0.06 μT and exposure in 2% of the homes was above 0.4 μT. A survey in Taiwan (43) in homes occupied by children under 7 years of age showed higher exposures compared to the USA, Europe and Australia. Spot measurements had been performed in 2214 randomly selected households. Mean magnetic fields were 0.121 μT and 7.3% of the included households were exposed to levels above 0.3 μT and 5.4% above 0.4 μT. However, the study from Taiwan reported results of single spot measurements that lasts about 30-40 s, while the studies from the USA, Europe and Australia reported results of 24-h measurements. Mean value in houses of women with unexplained spontaneous abortion: 0.4 μT ± 0.31 μT mean value in houses of pregnant women: 0.14 μT ± 0.15 μT Ursache et al. (23) Exposure environment: three apartments near substations. Measurements at height of 1 m above the ground and at  The WHO concluded that residential exposure to ELF-MF did not vary dramatically across the world (40) in 2007. Our overview suggests that this is not necessarily generalizable to all places around the world, as it revealed exposure levels in LMICs exceeding the levels reported for the USA, Europe, Australia and Taiwan. In our systematic search, we identified studies from Brazil (13) , China (29) , Iran (37) and the West Bank and Gaza (18) in which larger proportions of persons had been exposed to residential ELF-MF above 0.3 μT or 0.4 μT. In the Brazilian study (13) , 14.8% of the spot measurements performed outside the front door resulted in ≥0.3 μT, and 6.19% was ≥ 0.3 μT for 24-h measurements inside the houses. In a Chinese study (29) , 13.3% of the measurements at the front door of homes of pregnant women were above 0.4 μT, with 4.6% exceeding 1 μT. The Iranian study (37) reported that the calculated exposure to ELF-MF in 16.8% of 119 homes with children was above 0.45 μT. In a study of the West-Bank and Gaza (18) , a comparable high proportion of this exposure level was identified by indoor spot measurements (13% exposed to 0.3-0.4 μT) under normal power use.
The comparison of the results from LMICs with studies from HIC is only possible to a limited extent, since only the Brazilian study (13) performed 24h indoor spot measurements comparable to those in the USA, Europe and Australia. Single spot measurements as performed in the studies of West-Bank and Gaza (18) and China (29) , therefore, can only be compared to the results of the study from Taiwan. Also, the Chinese study reported on outdoor spot measurements. Spot measurements are considered the simplest form of measurements as they are not able to capture the temporal variability during the day as well as between-day variability during e.g. a week or seasonal differences (40) . Twenty-four-hour measurements improve the assessment of temporal variability because short-term increases of magnetic fields by devices or wiring do not influence the average field (41) .

Considerations for studies on ELF-MF in LMICs
We identified several methodological limitations in studies carried out in LMICs with respect to the conduct and the reporting of the studies. Applying standardised measurement routines and strengthening standardisation in reporting would be beneficial for future studies in LMICs and would allow for a better comparability with studies from HIC.

Conduct
Studies in LMICs could have been prompted by high exposure situations, providing a biased view of exposure levels. Nevertheless, some studies showed in fact that high exposure situations exist, but they have to be set in context, e.g. by showing how frequently these situations occur. Results should therefore be interpreted with caution and they cannot be generaliesd. In LMICs, permanent and reliable access to electricity is not always available for the whole population. In parallel to improvements of the electricity distribution networks, conducting measurement surveys drawn from a random sample must be a future goal. Therefore, studies should not solely include homes near substations, power lines and apartment buildings with transformer stations, but include randomly selected homes to obtain population-representative exposure. By conducting a pilot study, first, the feasibility of such an approach needs to be assessed in the local context. Results of the pilot study should also allow to perform power calculations (42) . Predefined measurement protocols for at least 24-h measurements should be developed and used. These protocols should include information about the measurement device and the calibrations for this device. The protocol should also give clear instructions on how to perform the measurement on-site (e.g. in the apartment). Besides measurements, authors should try to collect additional information that could have impacted the exposure, e.g. type of household, status of electric devices (on or off during measurements), temperature, measured distance to and voltage of power lines, presence and type of substations.

Reporting
A strengthening of standardisation of reporting could lead to a better comparison of the studies. Especially reporting on key elements of the exposure assessment is important, including but not limited to the measurement device, time and season of measurement and measurement techniques. The authors should, therefore, state if they used uniaxial or triaxial measuring probes, as well as which field values they report (total field, maximum along a single axis). They should also indicate whether the measurement device was calibrated according to manufacturer's specifications. It is also important that results are reported in a comparable way, in tables. This should include reporting on average measured exposure of magnetic fields (at least arithmetic and geometric mean, standard deviation, and median and other percentiles) and distribution of exposure levels, with emphasis on the exposure levels higher than 0.3 μT and 0.4 μT.

Strengths and limitations of the study
One strength of our study is that the process of literature search, screening and study selection was done systematically and in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA) workflow (44) . We included two different databases, performed the study selection by applying a priori defined eligibility criteria and extracted data in a predefined database.
Additional strengths of our overview are that we made extensive efforts to identify relevant publications on this topic by checking two electronic databases (PubMed and WoS) and complementing this by additional efforts, including the examination of the issues of the last 2 years of the journal that published the largest number of included studies, checking the reference lists of included articles for other relevant articles and conducting an informal survey among expert in the field of electromagnetic fields via an existing email list including experts who are collaborators of the WHO. These efforts confirmed that our search had been exhaustive, since we only identified two additional articles: both articles had been missed because of the rare situation that the official country name 'West Bank and Gaza' had not been used in the article which referred to the study location as 'Palestine'.
Despite the strength, our study has some limitations. The objectivity in terms of selection and data extraction process could have been further improved by adding a second independent reviewer. Furthermore, we cannot rule out the possibility that we miss potential relevant articles by using search terms only in the English language.

CONCLUSION
Data on exposure to ELF-MF in the general population of LMICs were sparse and even non-existent for LIC. Studies were heterogeneous and quality of reporting was limited. Direct comparison with systematic monitoring surveys from HIC is very limited with the data available. Some studies showed measured ELF-MF levels higher compared to studies from HIC, indicating a need for further analysis. However, the generalizability of the currently available evidence is unknown. There is an urgent need for future systematic studies with randomly drawn samples, sound measurement methods based on predefined protocols and standardised reporting, before any firm conclusions on ELF-MF exposure in the general population in LMICs can be drawn.

SUPPLEMENTARY DATA
Supplementary materials are available at Clinical Infectious Diseases online.