Estimating exposure to extremely low frequency magnetic fields near high-voltage power lines and assessment of possible increased cancer risk among Slovenian children and adolescents

There are about 670 km of 400 kV PL, 330 km of 220 kV PL, 2600 km of 110 kV PL and 25 km of 110 kV underground cables in Slovenia. The technical data necessary to calculate and determine the spatial distribution of the ELF MF in the vicinity of HV PL were obtained from national power grid company ELES and from 5 power distribution companies covering Slovenia. The following data about each PL tower were collected: the coordinates, altitude, height and type. The tower type defines the type of the PL geometry (Barrel, Danube...) as well as the exact geometry of the tower cross arms. To take the sag of each span into account, the data about the exact catenary of each span was also obtained. For most of the HV PL the catenary for each span was available in the form of 3D lines, generated by national power grid company from the results of the measurements with LIDAR system. For a few HV PL where the 3D lines were not available, the values of the sag of each span or cable tension in each span were used to calculate the 3D lines. For 110 kV underground cables their exact course was available in a form of points along the cable course, but detailed information about the geometry and depth of the cables was not available.

Based on these technical data, each PL was further split into sections with identical tower types and identical or very similar geometries (position and length of tower cross arm). If the PL was double circuit (in Slovenia only single or double circuit PL are in use), the load data were checked for each system of the PL. If the loads were mostly the same, then both systems were deemed as one section, otherwise each system was deemed as separate segment. In total the HV distribution system was split into 17 segments for 400 kV, 13 segments for 220 kV and 411 segments for 110 kV.

The load data provided by national power grid company ELES consisted of the data about the working and reactive energy transferred by each HV PL in 15-minute intervals in the period from 1. 1. 2006 to 31. 12. 2017. From these data the values of electric currents were calculated for every 15 minutes, as well as several geometric mean values: daily, weekly (7 day), monthly (30 day) and total mean value.

There is no unified approach among the studies regarding which mean value to use as the surrogate for long term exposure. Some studies have estimated the variations over time from available data, for example, on electricity consumption.^24,25 Several studies used the yearly mean load value for the year of the diagnosis with some adding also the year of birth.^{23,26, 27, 28, 29} In this study, the total mean value covers the two-year period from 1. 1. 2006 to 31. 12. 2017.

The ELF MF level at one location is the result of the contributions of all nearby HV PL. As the ELF MF is a vector quantity, the total value depends not only on the amplitude of each contribution, but also on their direction and the angles between them. Typically, when calculating the ELF MF, a numerical model is created that contains all sources and thus the calculation determines the total ELF MF level for the selected loads of HV PV. With the available programs and computational capacity, such a calculation is not feasible for large areas. In addition, usual approach requires recalculation of the whole model for each load conditions (for example at nominal loads, actual maximum loads, average day or annual loads, conditions in case of one HV PL failure...) or if there are changes in technical characteristics like construction of new or reconstruction of existing HV PL.

To allow largescale ELF MF calculations, a new method was developed and validated, which enables the calculation of the total ELF MF based on the values of the ELF MF of individual sources. The methodology was divided into two steps: the first step was to calculate the ELF MF of one HV PL and the second was to calculate the total ELF MF due to all HV PL for desired loads.

New method to calculate the ELF MF of one HV PL was based on the principle that the distribution of the ELF MF in the vicinity of the PL is not affected by the nearby objects or terrain, and on the fact that as long as the geometry of the PL and its load are the same, the value of the ELF MF in a selected point depends only on the distance between the selected point and the HV PL and on the difference of the altitude of the selected point and the HV PL. Each HV PL was split into sections with identical or quite similar geometries. From the data in the registry, a 3D line was generated for each segment, representing the course of the lowest cable of the HV PL. Additionally, for each HV PL segment the ELF MF distribution in a vertical cross section was calculated with a Narda EFC-400 EP program package at the nominal load. The program package calculated the ELF MF by splitting all conductors into short parts and summing their contributions. The value of the ELF MF of one short part of conductor was calculated by the Biot-Savart law:

where is magnetic flux density contribution from a short part of the conductor at the point given by positon vector and I is electric current in the cable.

The calculation took into account exact geometry of the towers in this segment and extended to the distance where the value of the ELF MF fell below 0.05 μT. The result was stored in a GRID matrix with the step of 1 m and positioned so that the center of the matrix aligned with the 3D line of the segment.

The 3D line representing the course of the PL and the matrix with the values of the ELF MF in a vertical cross section were used as the input to a customdeveloped program that calculated the value of the ELF MF along the whole segment length. The program first split the segment into spans. For each span, the program generated a 1m grid of points that covered the area of the whole length of span in one dimension and the same distance as covered by the matrix of the values of ELF MF in a vertical cross section. At both ends of the span (at location of the HV PL towers), additional points were added by rotating the last line of points as presented by blue and black dots in Figure 1. We can see that blue and black dots do overlap in a certain area around the tower. For each point in the 1m grid, first the height difference was calculated between the absolute height of the 3D line representing the course of the PL and the point 1 m above the ground. The absolute height of the point 1 m above the ground was determined from the digital elevation model of Slovenia with the resolution of 1 m constructed based on the LIDAR data. Second, the distance of each point in the 1 m grid from the HV PL was calculated. Both values were rounded to the nearest integer and used to pick the corresponding value from the matrix with the calculated values of the ELF MF in a vertical cross section. By this process the proper value of the ELF MF was assigned to each point in the 1 m grid. Finally, a 10 m orthogonal grid was generated (red circles in Figure 1) to which the highest value of nearby points in a 2 m radius was assigned. The output of the program for each segment was a 10m grid of points with the values of the ELF MF.

Figure 1

Finally, if HV PL was split in more than one segment, all 10 m grids of one HV PL were merged together. For those coordinates where values were available in two (or theoretically more) grids, the highest value was used for the final result.

When multiple HV PL either cross or run in parallel, the total value of the ELF MF is a result of the contributions of all nearby HV PL. As ELF MF is a vector quantity, the resulting value is typically significantly lower than a sum of all contributions. Therefore, four different algorithms to determine the total value of the ELF MF in each considered point were analysed and compared:

As, the actual total value of the ELF MF can never exceed the sum of the two values considered, the MAX + MIN algorithm is the most conservative, the MAX algorithm is the least conservative, and the other two are inbetween.

In order to evaluate all four algorithms, various realworld scenarios were analysed where cumulative effects of several HV PL occur, e.g. crossing of two HV PL, or parallel courses of two and three HV PL. Figure 2 presents the analysis of one such case –the intersection of two 110 kV double-circuit HV PL. The colour scale shows the deviation of the estimated ELF MF determined by all four new algorithms from the actual ELF MF determined by usual numerical modelling procedure. Green colour means that there is practically no deviation, red colour marks areas where the algorithm overestimated the values, and blue colour shows areas where the algorithm underestimated the values. Among four proposed algorithms, the MAX algorithm underestimated the values of the ELF MF the most (as the blue areas are the largest) and overestimated them the least (as the red areas are the smallest). The MAX + MIN algorithm never underestimated the values of the ELF MF, and overestimated them the most among the four proposed algorithms. Information regarding the deviation of the estimated value from the actual value is also provided in Figure 2 by black and red contours. The black contours represent the field lines of 0.1, 0.4 and 1.0 μT for the actual value obtained by usual modelling procedure, and the red contours for the estimated value using a novel method. For the MAX + MIN algorithm red contours never lie inside black contours, which means that the MAX + MIN algorithm overestimated the values of the ELF MF. Using the MAX algorithm, red contours often or even predominantly lie within black contours. The analysis showed that the most realistic results for the conditions considered was given by the MAX + 0.6 MIN algorithm, which was thus chosen as the optimal algorithm. Subsequent feasibility analysis showed that calculations could be performed according to all four algorithms, which allowed later sensitivity analysis, i.e. whether the choice of algorithm affects the result of analyses.

Figure 2

A novel program with all four analysed algorithms was developed to determine total value of the ELF MF. The program also featured the functionality to scale the values of each HV PL with a certain factor before calculating the total value of the ELF MF. This functionality enables relatively easy and fast calculation of the total ELF MF not only for the nominal load of HV PL, but for any load conditions or operating conditions of the whole power distribution network. For the later analysis, the total geometric mean value of the load was used that covers the whole period from 1. 1. 2006 to 31. 12. 2017.

We conducted retrospective geographical epidemiological study in population of Slovenian children and adolescents aged 0–19 years diagnosed in a 12-year period 2005–2016.

Address of residence served as a proxy indicator of ELF MF exposure. From Central Population Registry data on permanent address, which was georeferenced to X and Y coordinates, was obtained for all Slovenian children and adolescents in the studied population. Modelled values for ELF MF density near HV PL were used to classify all population of children and adolescents into five categories of ELF MF exposure: B < 0.1 μT; 0.1 μT ≤ B < 0.2 μT; 0.2 μT ≤ B < 0.3 μT; 0.3 μT ≤ B < 0.4 μT and 0.4 μT ≤ B. Georeferenced addresses were allocated to the closest point on the 10m grid by using the k-nearest neighbours (kNN) method on the Quad Tree to avoid calculation of Euclidian distances between each grid point and each address location (building Quad Tree algorithm has running time O(n·log(n))).³⁰

From the Slovenian Cancer Registry incidence (number of all newly diagnosed cancer cases in one calendar year) data on cancer patients were obtained and linked to population data based on Slovenian personal identification number. Patients with leukemias were identified based on codes C91–C95 according to the 10^th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10). In addition, we analysed all cancer cases occurring in childhood (age at diagnosis 0–14 years) and cases with malignant tumour of the central and autonomic nervous system (or brain tumour in short) with codes C70–C72 diagnosed up to 29 years of age.

Relative risk of cancer was assessed using a standardized incidence ratio (SIR), which is a method for indirect age standardization and is calculated as a ratio between observed and expected incidence for each studied group, i.e. cancer incidence in each of the five categories of ELF MF exposure. Expected incidence is calculated from age distribution in studied group and age-specific incidence rates in the overall population.³¹ A SIR value of 1 indicates that observed incidence in the studied group is the same as expected incidence, less than 1 indicates a lower than expected incidence, and over 1 indicates a higher than expected incidence.

Data management and analysis of population and cancer data was performed in R software (version 4.0.2) and R packages dplyr (version 1.0.2) and SearchTrees (version 0.5.2).

To measure the ELF MF at the participants homes it is necessary to obtain their consent. This can be potential source of selection bias.²⁹ It is suggested ³² that although it is an imperfect measure of magnetic field exposure, wire code is the only method applicable to nonparticipating subjects. But if technical data about the HV PL near the residence of participants are available, it is possible to determine the exposure at the desired time interval based on the numerical calculations. Such combined estimation of ELF MF exposure has been used in several studies.^{8,27, 28, 29} As numerical calculation of ELF MF is very demanding for larger areas, the studies usually limit the area of interest to a pre-selected distance to the HV PL, usually in the range of a few hundreds of meters.

The value of the ELF MF in a selected point near PL depends on the geometry of the PL, the load of the PL, the distance from the HV PL, and the elevation difference between the PL and the selected point. It is not affected by the terrain and nearby objects. This principle was used by Miravet-Garret³³, where the distance and difference of the altitude of the HV PL and of the selected point was determined from the detailed technical data of the HV PL and the ground profile, and then the value of the ELF MF at the selected point was calculated by a simplified analytical formula.

The mean values of the load of HV PL in Slovenia have an increasing tendency, meaning that on average the energy transfer through HV PL is increasing. For 110 kV PL the average value of geometric means of the loads of all PL was 84 A in 2006 and 87 A in 2017, for 220 kV PL it was 185 A in 2006 and 213 A in 2017, and for 400 kV PL it was 282 A in 2006 and 360 A in 2017. The increase is significantly higher for 400 KV PL compared to either 220 or 110 kV PL. However, the mean loads are not high. Among all 400 kV PL, the highest total mean value of the load for one PL for the whole period from 1. 1. 2006 to 31. 12. 2017 was 742 A and it increased to 842 A for the highest yearly mean value (in year 2016). In Figure 3 the daily, monthly and yearly mean values for this HV PL are presented.

Figure 3

Yearly mean loads do not differ significantly from the total mean loads (1. 1. 2006 ‒ 31. 12. 2017). The highest yearly value was achieved in 2016 (842 A), whereas the lowest was achieved in 2007 (374 A). The fluctuations in monthly mean loads are larger, the highest monthly load was achieved in November 2010 (1588 A) and the lowest in August 2010 (1 A). For weekly mean loads, the highest values were achieved at the end of November 2010 (1987 A).

The ELF MF around HV PL was calculated for the entire territory of Slovenia on the 10m grid. On Figure 4 the value of the ELF MF of several HV PL around one transformer stations are presented for nominal load (top) and the total mean geometric load covering the period from 1. 1. 2006 to 31. 12. 2017. The value of the ELF MF was determined by the new algorithm (from chapter Methodology to calculate the total ELF MF of multiple sources). For the total mean geometric load, the value of the ELF MF was higher than 0.1 μT on almost 2 million points. This represents the area of about 191 km², which means that on about 1% of the territory of Slovenia the value of the ELF MF is 0.1 μT or higher, while on more than 99% of the territory of Slovenia the value of the ELF MF is below 0.1 μT. For 0.4 μT the areas are significantly smaller, as this value is exceeded only on 0.36% of Slovenian territory.

Figure 4

The results were validated by comparing calculated values of ELF MF with measurements in the vicinity of two 400 kV PL and two 220 kV PL. The same load conditions (nominal load) were used for comparison. For the 400 kV PL with the highest mean load, the comparison is given in Table 2.

Cancer in children and adolescents is a rare disease, as it represents less than 1% of all cancers in Slovenia as well as in Europe.^34,35 In recent years, around 70 cancers are diagnosed annually in Slovenia during childhood and adolescence (from birth to the age of 19), 20% more among boys than among girls.³⁴ Age-standardized incidence rate has been increasing by an average of slightly more than 1% annually since 1961 in Slovenia and also increased by about 1% per annum in Europe in the past three decades.^34,36 The causes for the increasing incidence are unknown; although it is largely attributed to improved diagnostic and registration procedures, the possibility of increased exposure to different harmful factors, especially intrauterine, cannot be excluded.³⁶

The most common groups of malignant cancers in the age group of 0–19 years are leukaemias, lymphomas and tumours of the central nervous system (in short brain tumours). According to the average for the period 1961–2019, leukaemias represented a quarter of all newly diagnosed cancers, followed by brain tumours with around 15%, Hodgkin’s lymphoma with 10% and non-Hodgkin’s lymphoma with 8%.³⁴ Apart from exposure to benzene and ionizing radiation, the risk factors for leukaemias are not well known.^37,38 The so-called hygiene hypothesis is also possible, where, due to the improvement of infection prevention in the modern era, leukaemia may develop in genetically predisposed individuals due to an incorrect response to infection by the immune system of a child who did not come into contact with non-dangerous infectious agents in early childhood at the right time (early enough).³⁹ Some (but not all) studies have shown that the risk may be increased (only) for childhood leukaemias with longterm daily exposure to ELF MF densities greater than 0.3 or 0.4 μT, depending on the study.

The key advantage of our study was that with inclusion of the population data from the Slovenian Cancer Registry, we avoided performing a case-control study and, at the same time, selection and participation bias.⁸ Research indicates that the fraction of children exposed to higher values of ELF MF (>0.4 μT) is likely to be less than 2% of the population.^40,41 For Slovenia, it was estimated in the past that no more than 1% of children were exposed to long term ELF MF greater than 0.4 μT.⁴² In presented study, using an innovative model for evaluating ELF MF on a small spatial network and georeferencing residences and sources of the electricity network in Slovenia, we precisely determined the exposure to ELF MF in children aged up to 14 years in the period 2005–2016 in Slovenia, who live nearby HV PL. We found that as many as 99.5% of all children did not live in the areas with the mean value of ELF MF 0.1 μT or higher. And only 0.09% of all children were exposed for at least one year to the area of potentially carcinogenic ELF MF density, that is greater than 0.4 μT, in the vicinity of HV PL. This population results and results for age groups 0–19 and 0–29 are presented in Table 3 (the results are equal to the first decimal point). These results emphasize the importance of performing exact calculations on real-life population-based data in order to obtain realistic picture of the situation in a specific country.

In Table 3 we also present results according to five categories of ELF MF exposure for observed and expected number of cancer cases, and standardized incidence ratio for all cancers combined, leukaemia and brain tumours. All of the 516 cancer cases in children aged 0–14 years in 2005–2016 were classified in the lowest category of exposure (ELF MF below 0.1 μT) near HV PL. Relative risk cannot be calculated for other categories of ELF MF exposure, since there were no cases.

Among 195 diagnosed cases of leukaemia in children and adolescents aged 0–19 years in 2005–2016 only one adolescent was classified in the category of ELF MF density between 0.1 and 0.2 μT – a person lived in the vicinity of 110 kV PL. This single case of leukaemia represents big variability and thus “shifted” all residents from this group from a relative risk of zero to a relative risk of more than one. Although the value for relative risk in this group appears high (2.4), the confidence interval is far too wide for statistical significance (since it is based on only one case). No leukaemia case was classified in categories with higher ELF MF exposures (i.e., above 0.2 μT).

Among 196 patients aged 0–29 years diagnosed with brain tumour only one adolescent was classified in the category of ELF MF exposure between 0.2 and 0.3 μT, giving standardized incidence ratio 4.6 with wide 95% confidence interval (0.1–25.4). Also, this person lived in the vicinity of 110 kV PL at the time of diagnosis. No brain tumour case was classified in other categories of ELF MF exposure.

We conclude that none of the cases of leukaemia in the studied population can be attributed to exposure to ELF MF – the relative risk of leukaemia in children and adolescents living near HV PL does not differ from the average risk in the general population. Additionally, we could not assess a possible risk trend over exposure categories, because there were also no cases of leukaemias in other categories.

Our findings are in line with other similar studies, but we cannot directly compare the results with other studies due to the different (in our opinion better, more accurate assessment methodology based on actual exposure and availability of exact coordinates for all dwellings on Slovenian population.^41,43 Using place of residence as a surrogate indicator of ELF MF exposure is more reliable in children than in adults, as they spend most of the day in the vicinity of home and the surrounding area. In addition, they migrate less.