IMPORTANT NOTE:
The UKCCS was published in the December, 1999, issue of The Lancet. Surprisingly, the same issue of The Lancet published a Commentary that, in effect, said that the UKCCS was badly designed using outdated methodologies. This criticism has been born out, as the British Cancer study published 10 months later used the UKCCS's own data to find a relationship between EMF and Cancer. Prof N Day, who is the lead researcher for the UKCCS study, was principle researcher of the British Journal study.
|
|
|
|
|
|
UK Childhood Cancer Study Investigators*
*For information on the full list of investigators see end of paper
Correspondence to: Prof N Day, Strangeways Research Laboratory,
University of Cambridge, Cambridge CB1 8RN, UK
Background Previous studies have suggested an association between exposure to power-frequency electromagnetic fields (EMF) and the development of childhood malignant disease, especially leukaemia and tumours of the central nervous system. We investigated the relation between all childhood cancer and exposure to power-frequency magnetic fields.
Methods The UK Childhood Cancer Study was a population case-control study covering the whole of England, Wales, and Scotland. All children with a confirmed malignant disorder were potentially eligible. For each case, we matched two controls on date of birth and sex, randomly chosen from the list of the Family Health Services Authority in England and Wales or Health Board in Scotland. In the main study, 3838 cases and 7629 controls were interviewed. The EMF part of the study included only one control per case, and household EMF measurements and school measurements where relevant were taken on 2226 matched pairs. These measurements, adjusted for historical line load and appliance fields, were used to estimate average exposure in the year before the date of diagnosis, or an equivalent date for controls. Analyses were by conditional logistic regression, incorporating a census-derived deprivation index used as a measure of socioeconomic status.
Findings For children with mean exposures of more than 0·2 µT compared with children with mean exposures of less than 0·1 µT, the adjusted odds ratios were 0·92 (95% CI 0·471·79) for acute lymphoblastic leukaemia, 0·90 (0·491·63) for all leukaemia, 0·46 (0·111·86) for central-nervous-system tumours, 0·97 (0·462·05) for other malignant disease, and 0·87 (0·561·35) for all malignant disease combined. Higher exposures (>0·4 µT) were recorded for only 17 (<0·4%) individuals (eight cases, nine controls).
Interpretation This study provides no evidence that exposure to magnetic fields associated with the electricity supply in the UK increases risks for childhood leukaemia, cancers of the central nervous system, or any other childhood cancer.
Lancet 1999; 354: 192531
The causes of childhood malignant disorders are poorly understood. Ionising radiation, some cancer chemotherapy agents, viruses, and genetic factors are each the cause of a small proportion of cases, but for most cases the cause is uncertain. Several hypotheses have been proposed, including parental cigarette smoking and, for childhood leukaemia, population mixing and abnormal responses to infection.
One hypothesis has related the risk of development of childhood cancer, especially leukaemia and brain tumours, to increased exposure to electromagnetic fields (EMF), specifically the magnetic component associated with the distribution and use of electricity. The first report1 of a link between childhood leukaemia and exposure to magnetic fields at a frequency of 60 Hz was followed by many studies of power-frequency fields (50/60 Hz), with conflicting results.24 An overview of the early studies based on calculated fields suggested that risk increased smoothly with increasing time-weighted mean exposure to more than 0·2 µT, with a relative risk of about 1·8 associated with exposure of 0·6 µT or more.5 For exposures lower than 0·2 µT, the relative risk seemed to be constant. Several subsequent studies measured magnetic-field exposure rather than inferring or calculating exposure from the proximity and load of neighbouring power lines.610 The results are inconclusive in that they are consistent with no increase in risk at exposures higher than 0·2 µT, but they cannot exclude the possibility of a moderate excess risk at high exposure.
We present the results of the EMF part of the UK Childhood Cancer Study (UKCCS), a nationwide case-control study of childhood cancer done across the whole of the UK, except for Northern Ireland.11 The UKCCS was set up in the early 1990s, when sufficient evidence had accrued to suggest that the roles of several exposures required investigation.
In the full study, five main hypotheses were tested: childhood cancer might be caused by in-utero or postnatal exposure of the child to ionising radiation; specific types may be caused by in-utero or postnatal exposure of the child to certain chemicals; childhood cancer might be caused by exposure of parental germ cells to ionising radiation or certain chemicals before conception of the child; specific types, especially leukaemia and cancers of the central nervous system, might be caused by postnatal exposure to extremely low-frequency EMF; leukaemia and lymphomas may arise as rare and abnormal responses to infection. Results on the non-EMF hypotheses will be reported elsewhere.
We measured the 50 Hz magnetic fields (and harmonics <800 Hz)
generated by the distribution and use of electricity in homes and, when
relevant, schools to obtain estimates of individual exposure. Each
measurement consisted of a series of readings, summarised as the
arithmetic mean. We tested the hypothesis that a mean exposure of more
than 0·2 µT in the year before diagnosis would increase risk of childhood
leukaemia, specifically acute lymphoblastic leukaemia, and cancers of the
central nervous system, compared with a mean exposure of less than 0·1 µT
in the year before diagnosis. A further hypothesis was that risk of the
same diagnoses would increase smoothly with increasing mean exposure in
the year preceding diagnosis. Residential electric fields were measured in
a subset of the study, which will be reported elsewhere.
The UKCCS was a population-based collaborative case-control study covering the whole of England, Wales, and Scotland, based on eight regional centres in England and a centre each in Scotland11,12 and Wales.11
In England and Wales, the UKCCS study population was defined as children aged 014 years, registered with one of the Family Health Service Authorities (FHSAs). In Scotland, which has an independent system, the study population was defined as children registered with one of the 13 Health Boards.
The study began in Scotland on Jan 1, 1991, and in England and Wales on April 1, 1992. All children registered with an FHSA or Health Board after these dates who had pathologically confirmed malignant disease (except for some cancers of the central nervous system diagnosed by scan or radiography alone), as defined in the classification scheme devised by Birch and Marsden,13 were potentially eligible for the study. In Scotland, case accrual ended in December, 1994, and in England and Wales it was restricted to children who had leukaemia and non-Hodgkin lymphoma throughout 1995, and leukaemia alone throughout 1996.
For each case two controls, matched for sex and date of birth, were randomly selected, from the list of the same FHSA or Health Board as the case. In England and Wales, computerised lists of children registered with each FHSA on Jan 1 and July 1 each year were obtained and ten potential controls randomly selected from the list on which cases appeared immediately before diagnosis. When the case child was less than I year old at diagnosis, we chose controls from the first FHSA list on which the case appeared. The general practitioners (family physicians) of the first two potential controls were approached and, with their permission, the parents of those children were contacted and asked to participate in the study. If the general practitioner or parents refused permission, we approached the next control in the same way until two control families participated.
Children (cases and controls) were ineligible if they had been born outside the UK study area or had had previous malignant disease. For the purposes of the study, all controls were assigned a pseudo-diagnosis date--the date on which they were exactly the same age as the corresponding case at diagnosis. Children who themselves, or whose parents, were resident outside the study area in the 3 months leading up to diagnosis or pseudo-diagnosis date were ineligible for inclusion. Children in residential local-authority care at diagnosis date (<1% of total childhood population) were also excluded.11
For the EMF component of the study,11 we chose only one control per case from the two included in the main study because of limited resources. All cases who had participated in the full study were eligible for inclusion in the EMF study. Eligibility was based on home address, since exposure was based on measurements in the home. Home addresses were eligible if the child had lived there for at least 12 months before diagnosis or pseudo-diagnosis, or since birth for children younger than 1 year at diagnosis or pseudo-diagnosis, and the family still lived there. Fixed-site caravans were included, but not mobile homes. If a case's family refused to participate in the EMF study or if the case was ineligible we did not approach either of the matched controls. Otherwise, we approached the control with the lower identification number of the pair. If the first control family refused to participate or the control was ineligible, we approached the second control's family.
If a child had changed address since the diagnosis or pseudo-diagnosis date, we did not take measurements in the previous home. Because of delays in starting the EMF study, the length of time between the diagnosis or pseudo-diagnosis date and the first measurements varied. This delay affected the number of case-control pairs for which measurements were available. The mean time between diagnosis or pseudo-diagnosis and the initial measurement was 20·8 months (SD 10·8) for cases and 21·3 months (10·8) for controls. The last measurement made was in December, 1998.
To address the wide-ranging hypotheses under investigation in the UKCCS, data were collected in stages from several different sources. The first component of the study, without which the others could not proceed, was personal interviews with the parents of cases and controls.
Full residential and occupational histories, including specific information about occupational exposures and individual housing characteristics, were recorded for each parent. To improve the quality of these data, a form asking parents to list the places in which they had lived and the jobs they had had was sent out before the interviews. At interviews, mothers and fathers were also asked about their own health, social habits, and any illnesses in their families. Additional sections on pregnancies and the index child's health, schooling, and social history were incorporated into the mothers' questionnaires.
At the end of the interview, interviewees were asked whether they
were willing to be contacted again. Consent was sought for their
participation in the ionising radiation (radon and
The measurement protocols for the EMF study were based on data acquired in a pilot study14 done by the National Radiological Protection Board. The pilot study showed that a restricted set of measurements would classify, with acceptable sensitivity and specificity, an individual into the lowest 90% of exposure. For exposures in the top 10%, more extensive measurements would be required to give more precise exposure estimates. Since we did not have resources available for making extensive measurements in the households of all cases and controls, we used a two-phase approach.11
In the first phase, we gathered information on EMF exposures from five different sources: specified measurements in the child's home (designated the phase I measurement); the proximity and type of overhead powerlines nearby, from an external-sources questionnaire; a questionnaire on electrical appliances in the home (night storage heaters, underfloor heating, and electric blankets); and measurements in schools or other institutions, such as purpose-built nursery schools, attended by the child; and electricity companies' databases of historical load data and other operating characteristics.
A school was eligible if the child had attended for 15 h or more per week during the winter (October to March) immediately before the diagnosis or pseudo-diagnosis date. If the child had attended more than one school in this period, we chose the school at which the most time was spent.
In the second phase, we took further EMF measurements (phase II
residential measurements) for all children indicated by phase I to be in
the top 10% of exposures (taken as
Exposure assessment was based on measurements for all participants, except for a few for whom specific adjustments were made. These adjustments were: an addition to the measured exposure made for individuals with exposures from certain household appliances, and addition to or subtraction from the measured exposure because exposure from external sources had changed between the year before diagnosis or pseudo-diagnosis and the time of measurement, as determined by line-load data and circuit configuration over the respective periods.
We measured resultant magnetic fields with Emdex II magnetic field meters (Enertech Consultants Ltd, Campbell, CA, USA) in the broadband frequency range 40-800 Hz.11 For the shorter measurements, we used sampling intervals of 1·5 s and 3·0 s in the phase I and phase II assessments, respectively. The sampling interval was adjusted to 10 s for 48 h measurements taken in phase II.
To prevent identification of high-exposure and low-exposure households by study technicians, meter readings were not displayed during the assessment. Information on EMF exposure in individual homes and schools was provided to study participants on request, but otherwise remained confidential.
The protocol was designed specifically to estimate the average EMF exposure in the year before diagnosis or pseudo-diagnosis.
The phase I residential measurements comprised (in order): three 3 min spot measurements taken in the centre of the child's bedroom, at the centre of the child's bed, and on the centre of the child's pillow; one 90 min measurement taken in the centre of the main family room; a repeat of the three spot measurements after the 90 min measurement. If the period between measurements in the homes of the case and of the corresponding control was 4 months or more, we tried to repeat the earlier measurements. An interval of less than 4 months was achieved for 98% of case-control pairs. During the phase I household visits, we asked about the time that the child spent in bed and at school and used the information in the calculation of time-weighted average exposures.
We made phase II measurements for the matched case-control pairs as close as possible to each other and within 4 weeks. If the phase I questionnaire had identified possible exposure from a night storage heater or underfloor heating in the bedroom, we took measurements during a period in which the appliance was in use, typically during the winter months. We did all phase II measurements at times agreed with the appropriate electricity company as being typical for operation of the local distribution system. The measurements comprised: four 3 min spot measurements taken at the centre of the family room, at the bedside position to be used for the 48 h measurement, at the centre of the child's bed, and at the centre of the pillow; a 48 h measurement taken by the side of the middle of the child's bed; a repeat of the four spot measurements after the 48 h measurement.
For phase I and phase II all measurements, apart from those made on the bed, were done with the meter held 1 m above the floor in a polypropylene stand, and at least 1 m from any operating appliances. Meters were placed in tamper-proof holders for the 48 h measurements.
We took school measurements when the heating systems were operating normally. In England and Wales, measurements were made from October to March, inclusive. There were two measurement schemes. The first was used when the child spent most of his or her time at school in a single classroom during the relevant winter period, typically in primary schools. For this scheme we made five 2 min spot measurements near the centre and four corners of the room. The second scheme for children who used many classrooms, typically those in secondary schools, consisted of spot measurements in up to five of the rooms most frequently used during the relevant winter period. In each room, one measurement was made near the centre; the measurement time totalled 10 min and the measurements in the different rooms were of equal duration. All measurements were made at a height of 1 m from the floor and at least 1 m from any appliance operating on mains electricity.
An external-sources questionnaire was completed for each EMF study participant to identify important sources of electricity supply, such as power lines, near to homes and schools. The questionnaire was designed in cooperation with the National Grid Company, the regional electricity companies for England and Wales, and ScottishPower and Scottish Hydro-Electric in Scotland. The specific purposes of the questionnaire were: to identify high-voltage lines or underground cables that were capable of producing annual average fields of more than 0·1 µT in the home or school; to obtain load and other circuit information to enable reconstruction of historical exposure; to check that the electricity distribution system was operating typically at the time of measurement; to identify substations and particular types of low-voltage circuits that were near to the location of interest. The questionnaires, masked for case or control status, were assessed by the National Radiological Protection Board.
Entry into phase II, based on the external-sources questionnaire, was determined by the following criteria for England and Wales: a National Grid Company overhead line within 400 m or underground cable located within 100 m of the home; a regional electricity company line of 66 kV or more at various threshold distances of up to 200 m from a home; a regional electricity company line of 1133 kV at up to 80 m from a home; an operating substation or a phase-separated underground cable of 33 kV or more within 20 m of a home or school; a three-phase 415 V distribution circuit within 2 m of the home; atypical conditions of distribution circuits at the time of the phase I measurement. In Scotland, we used equivalent criteria based on line voltages.
The threshold distances used to determine the relevant phase II high-voltage circuits were based on design-rating considerations, and were judged conservative. Typical loads on a regional electricity company line were found to be less than 20% of the circuit rating. Analysis of load data from a sample of National Grid Company circuits during a winter period had shown previously that 50% and 95% of the circuits had average loads of less than 30% and 50%, respectively, of their rating (D C Renew, National Grid Company, personal communication).
External-source questionnaires were also issued for interviewed cases and controls who were either ineligible for EMF measurements (because they had moved house during or since the year of interest) or who were eligible but had declined to participate in this part of the study.
To investigate the possible effects of refusal bias, we issued questionnaires for a random sample of 1000 of the 1582 first-choice controls who had refused to participate in the full study.
To account for potentially large variability in exposure from high-voltage lines and cables, we used load data to reconstruct historical exposure for the year of interest. Line-load data were requested for all overhead lines with voltages of 66 kV or more within threshold distances from the location of interest. Thresholds were used to define magnetic flux density reference levels from annual average load data or circuit rating, together with relative circuit phasing. Similarly, we requested load data for phase-separated underground cables of 66 kV or more located within 20 m of the property. Line-load data for the time of measurement and the year of interest (ie, the year before date of diagnosis or pseudo-diagnosis) were requested. The extent to which data were available varied among electricity companies: 70% of annual load data returned covered all or part of the year of interest; 84% fell within 4 years of diagnosis.
The National Grid Company's EM2D program was used to compute magnetic flux densities generated by overhead lines or underground cables. The program, assessed by the National Radiological Protection Board for the purposes of the study, generated a time-averaged value of magnetic flux density, which was used in the exposure algorithm.
We calculated historical exposure for all cases and controls whose residences and schools met inclusion criteria and for whom line-load data were received. To allow for changes in line loading and circuit configuration between the year of interest and the time of measurement, individuals' exposure measurements were adjusted appropriately.11 Previous studies examined distance from power lines as a potential indicator of magnetic-field exposure. We took distance into account in selection for phase II and in the historical exposure calculations. Specific information on proximity to power lines will be reported separately.
Because we took heating appliances to contribute to exposure only during winter months, average exposure in the year preceding date of diagnosis was estimated according to the following algorithm: average exposure=W1×(bed: winter exposure)+W2×(bed: summer exposure) in year of interest+W× (school exposure)+W4×(home non-bed exposure).
This algorithm provides an estimate of the arithmetic mean of exposure in the year before diagnosis or pseudo-diagnosis.
The weights (W1W4,
In phase I, bed and non-bed home exposures were estimated from measurements in appropriate locations covered by a 2 h period. Bed exposure was estimated from the bedroom spot measurements, and the average of the 90 min family-room measurement was used as an estimate of exposure for time not spent in bed or at school. In phase II, bed exposure was estimated by the 48 h measurement. Non-bed home exposure was estimated by the phase I family-room measurement. School exposure was common to phases I and II.
If necessary, we adjusted all home and school exposures for appropriate historical measures and bed-winter exposure for appliance exposure.
About 20% of case-control pairs had phase II measurements. Average annual exposures based on the phase II measurements were used in the analysis when available for a case and the corresponding control. If a phase II measurement was available for only one of a case-control pair, we used phase I measurements for both. For the remaining 80% of case-control pairs, the average annual exposure based on the phase I measurement was used. The repeatability of the phase I and phase II measurements is described elsewhere.11
The validity of the phase I and phase II estimates, compared with average exposure over a year was assessed in a study done by the National Radiological Protection Board, which will be reported elsewhere.
The primary analysis used the estimated arithmetic mean EMF exposure in the year preceding date of diagnosis, confined to matched pairs on whom measurements had been made. The same type of exposure estimate, (ie, whether or not incorporating the phase II measurement) was always used for the cases and controls of each pair. Exposure was divided into four categories (<0·100 µT, 0·1000·199 µT, 0·2000·399 µT, and >0·400 µT), based on previously reported results,68 with the primary analysis combining the top two categories. The analysis preserved case and control matching through use of conditional logistic regression. Treatment of confounding variables was based on the definition of a confounder, that it should be associated with both disease and exposure. Variables that may be associated with exposure but for which there is no evidence of relation to disease, causally or through differential selection into the study, were not included as confounding variables. Some selection into the full study based on socioeconomic variables differed between cases and controls.11 In addition, a weak relation between socioeconomic status and EMF exposure is seen in the control group. We therefore included a measure of socioeconomic status. We based the measure on a census-derived deprivation index.11 for the census enumeration district containing the child's address at the date of diagnosis or pseudo-diagnosis. The measure is based on unemployment, overcrowding, and car ownership. Danesh and colleagues15 and Townsend and colleagues16 have described the usefulness in the UK of small-area-based measures of socioeconomic status instead of measures based on individuals.
We obtained information on proximity to powerlines and the
associated line-load information for cases and controls included in the
main study for whom no EMF measurements were available, and for some
first-choice controls not included in the main study. We analysed this
information to find out whether our results, based on measurement, could
have been affected by selection bias.
87% of all eligible cases diagnosed in the UKCCS in England, Scotland, and Wales in the defined periods were included, with at least one of the parents interviewed. The corresponding response rate among controls was 64% (figure), with evidence of under-representation of those living in the most deprived census areas. 2226 case-control pairs were eligible for analysis (58% of interviewed case-control sets, 50% of all eligible cases). The main reason for non-inclusion in the EMF part of the study was change of residence in the time between date of diagnosis or pseudo-diagnosis and measurements being taken. The most deprived category, from census-based small-area deprivation indices, was strikingly under-represented, compared with the full set of first-choice controls. Distribution of deprivation classification differed little, however, between the cases and controls with EMF measurements, which showed only slight relative under-representation among controls in the most deprived categories (table 1). Table 2 gives the age and deprivation index distribution by diagnostic category and table 3 the distribution of exposure by age and deprivation index. Exposure seemed not to be age-related, but was moderately associated with deprivation (table 3).
Adjustment for deprivation index had only a small effect on odds ratios for acute lymphoblastic leukaemia, leukaemia, tumours of the central nervous system, other malignant disease, and all malignant disease (table 4). Some participation bias not captured by this index is possible, but that it would have concealed a substantial positive association seems implausible. We did trend tests for the odds ratios for each sub-table; p values ranged from 0·33 to 0·80. There was no evidence, for any category of malignant disease, supporting either of the hypotheses of the EMF study. Compared with the baseline group, who had exposures of less than 0·1 µT, there was no excess among children with exposures of more than 0·2 µT, nor was there any evidence of increasing risk with increasing dose.
The data were examined for children aged 5 years or younger and children aged 6 or older for total leukaemia and for all other malignant disease. Risk did not differ by age.
We made line-load data adjustments for 48 children (80% of requests for load data), and only on one occasion (for a control) was an exposure estimate increased sufficiently, based on line-load data, to cause an upward change in exposure category (table 5). Only eight of 83 children (four cases, four controls) with substantial external-source exposure had exposures of more than 0·2 µT. More cases than controls were included through their proximity to high-voltage power lines (p=0·04, table 5). The entire excess, however, is in the exposure category of less than 0·1 µT. If this difference is true, it does not seem to be related to average exposure.
Of the 170 individuals who had an appliance of interest in the home, only two (one case and one control) changed exposure category when the estimated field from the appliance (electric blanket) was included in the assessment (from lowest to second lowest category). Only three individuals (all controls) were in one of the two highest exposure categories because of exposure at school. All three moved from the lowest category to the second highest.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
) and the
non-ionising radiation (EMF) components of the study. Signed agreement was
requested for blood samples to be taken at a later date and for medical
and other records to be accessed.
0·1 µT) and for
the relevant matched case or control. We also took further measurements
for individuals who were exposed to specified appliances described in the
phase I questionnaire, and those living within specified distances from
high-voltage overhead power lines and underground cables.
Wi=1)
were individually calculated for each child to reflect the time spent in
bed and in school, as recorded in the study questionnaire. If any of the
information needed to calculate the weights was missing, we used average
age-related values. There was no evidence of differential recall between
cases and controls; the average value of the total bed exposure weight
(W1+W2) was 0·46 for cases and controls, and the
average weight for school exposure (W3) was 0·093 for cases and
0·095 for controls.
