Power lines and childhood leukemia: the pooled studies side by side
Two pooled analyses published in 2000 found that children whose homes averaged more than 0.3 to 0.4 microtesla (3 to 4 milligauss) had about twice the leukemia risk of children in lower fields. Very few children live with fields that high, and the pooled analyses that followed found a weaker association or none. IARC and WHO describe the evidence as limited and say it does not show that the fields cause the disease.
How the studies measured exposure
The first study, by Nancy Wertheimer and Ed Leeper in Denver in 1979, used "wire codes," a rating of the power lines outside each home. Later work measured fields inside homes with spot readings or 24 to 48 hour recordings, calculated historical fields from line loads, or simply recorded the distance from a home to the nearest high-voltage line.
The method changes the answer. The June 2002 NIEHS booklet put the conclusion in two lines: wire codes are not associated with childhood leukemia, while measured fields show an association that is "weak," with no clear cause and effect. WHO's review warns that distance is "a very poor predictor" of magnetic field exposure. Two large measurement studies from the late 1990s found little or nothing: a National Cancer Institute study of more than 1,000 children in 9 states found little evidence of risk, and the UK Childhood Cancer Study of nearly 4,000 children with cancer found no excess, though most of those UK children had fields below 0.2 µT.
The pooled analyses and big studies
Dark bars compare magnetic field levels. Light bars compare distance from a high-voltage line. When the interval in parentheses includes 1, the result is compatible with no difference.
| Study and comparison | Odds ratio (95% CI) |
|---|---|
| Greenland 2000, above 0.3 µT | 1.7 (1.2 to 2.3) |
| Ahlbom 2000, 0.4 µT or more | 2.00 (1.27 to 3.13) |
| Draper 2005, within 200 m | 1.69 (1.13 to 2.53) |
| Kheifets 2010, 0.3 µT or more | 1.44 (0.88 to 2.36) |
| Bunch 2014, within 200 m | 1.12 (0.90 to 1.38) |
| Crespi 2016, within 50 m | 1.4 (0.7 to 2.7) |
| Amoon 2018, within 50 m | 1.33 (0.92 to 1.93) |
| Amoon 2022, 0.4 µT or more | 1.01 (interval not in abstract) |
Source: The study papers listed in the table below. Dashed line: odds ratio of 1, meaning no difference between groups.
| Study | Where and how many | Comparison | Result (95% CI) |
|---|---|---|---|
| Greenland et al., Epidemiology, 2000 | 15 studies, 12 with field data; summary based on 2,656 cases and 7,084 controls | Above 0.3 µT vs 0 to 0.1 µT | 1.7 (1.2 to 2.3) |
| Ahlbom et al., British Journal of Cancer, 2000 | 9 studies; 3,247 cases, 10,400 controls | 0.4 µT or more vs lower | 2.00 (1.27 to 3.13); 44 exposed cases |
| Draper et al., BMJ, 2005 | England and Wales, births 1962 to 1995; 9,700 cases, 9,700 controls | Birth address within 200 m of a line vs over 600 m | 1.69 (1.13 to 2.53); at 200 to 600 m, 1.23 (1.02 to 1.49) |
| Kheifets et al., British Journal of Cancer, 2010 | 7 studies published after 2000; 10,865 cases, 12,853 controls | 0.3 µT or more vs under 0.1 µT | 1.44 (0.88 to 2.36); 26 exposed cases |
| Bunch et al., British Journal of Cancer, 2014 | Britain 1962 to 2008; 53,515 children with cancer and matched controls | Within 200 m vs over 1,000 m, whole period | 1.12 (0.90 to 1.38) |
| Crespi et al., British Journal of Cancer, 2016 | California 1986 to 2008; 5,788 cases, 5,788 controls | Within 50 m of a 200 kV or larger line | 1.4 (0.7 to 2.7); 22 exposed cases |
| Amoon et al., British Journal of Cancer, 2018 | 11 record-based studies; 29,049 cases, 68,231 controls | Within 50 m of a 200 kV or larger line vs 300 m or more | 1.33 (0.92 to 1.93) |
| Amoon et al., Environmental Research, 2022 | 4 recent studies; 24,994 cases, 30,769 controls | 0.4 µT or more vs under 0.1 µT | 1.01 (no interval in the abstract) |
The two analyses from 2000 used different rules for which studies to include and different cutpoints, yet WHO called them "in close agreement." The 2,656 and 7,084 figures for Greenland come from the NIEHS booklet's reprint of its summary table. For Ahlbom, 44 of the 3,247 children with leukemia, about 1.4%, were in the top group, against 62 of 10,400 controls, about 0.6%.
What changed after 2000
The newer field-based analyses moved toward no effect. Kheifets and colleagues pooled seven studies from after 2000 and got 1.44 at 0.3 µT or more, with an interval that includes 1. They wrote that the association "is weaker in recent studies and imprecise because of small numbers of highly exposed individuals." Amoon's 2022 pooled analysis of four more recent studies found an odds ratio of 1.01 at 0.4 µT or more. Its authors also combined three pooled analyses in a meta-analysis and got 1.45 (0.95 to 2.20) at 0.4 µT or more.
The distance studies tell a stranger story. Draper's team found raised risk at 200 to 600 meters, where, they note, typical calculated fields from the lines are below 0.1 µT and often below 0.01 µT. That is less than the average field in homes from other sources. When Bunch and colleagues extended the British data to 2008, the risk within 200 meters fell decade by decade, from 4.50 in the 1960s to 0.71 in the 2000s. They concluded that "a risk declining over time is unlikely to arise from any physical effect of the powerlines." Amoon's 2018 analysis of distance found a small, imprecise excess within 50 meters of lines of 200 kV or more that was "not explained by high magnetic fields," with a higher odds ratio in children diagnosed before age 5.
Why agencies call the link limited, not causal
In IARC Monograph 80, published in 2002, IARC rated the human evidence on childhood leukemia "limited" and the animal evidence "inadequate," which led to Group 2B, possibly carcinogenic to humans. ELF electric fields went into Group 3, not classifiable. The panel of 21 experts from 10 countries based the decision on a fairly consistent doubling of risk above 0.4 µT and found no scientific explanation for it.
WHO's 2007 review, Environmental Health Criteria 238, weighs the evidence in both directions. Against a causal reading: case-control studies that measured fields in homes often had low response rates, which can bias the comparison, and nearly all laboratory and mechanistic work fails to support an effect at these levels. In favor of taking it seriously: Nordic studies that calculated fields from line records are free of any known selection bias, though they had few exposed children, and exposure errors that are similar in cases and controls tend to push results toward no effect. WHO also wrote that if the link is causal, the true risk "could also be greater than what is reported." Its verdict: not strong enough to be considered causal, but strong enough to remain a concern.
ICNIRP's 2010 guidelines add that no biophysical mechanism has been found and that there is no adequate animal model of acute lymphoblastic leukemia, the most common type in children. NIEHS, in its 1999 report to Congress, judged the chance that these fields are a real hazard to be "currently small," while saying they could not be called entirely safe.
How many children are in the high-exposure group
| Estimate | Figure | Source |
|---|---|---|
| Children with mean home exposure above 0.3 µT | About 1% to 4% | WHO EHC 238 |
| Children with median exposure above 0.4 µT | About 1% to 2% | WHO EHC 238 |
| National surveys, mean above 0.3 µT | About 4% to 5% (Korea 7.8%) | WHO EHC 238 |
| Children in the pooled studies at the high level | Less than 1% | National Cancer Institute |
| US homes above 3 mG and above 4 mG | 4.5% and 2.5% | NIEHS 2002 |
The National Cancer Institute's fact sheet states the problem plainly: less than 1% of the children in the studies had exposures at the level where the excess appeared. The estimates differ because some count children and others homes, some use means and others medians, and the surveys come from different countries.
If the link were causal, how many cases
WHO worked through the numbers on the assumption of causality, while calling the calculation very imprecise. Its estimate was 100 to 2,400 cases a year worldwide, or 0.2% to 4.9% of the roughly 49,000 childhood leukemia cases worldwide in 2000. It put the attributable share at about 0.5% in a low-exposure country such as the UK and about 5% in a high-exposure country such as the US. Greenland's analysis gave a US figure of 3%, with an interval from minus 2% to 8%. Draper's team estimated about 1% of cases in England and Wales, roughly five a year, and called that estimate imprecise.
Reading the next study
- Count the children in the top exposure group. Ahlbom had 44 cases there, and Kheifets had 26 at 0.3 µT or more.
- Check whether exposure was measured, calculated from line loads, or inferred from distance.
- Look for the participation rate in any study that measured fields in homes.
- Check whether the confidence interval includes 1.
- Note whether the data are new or a re-analysis of children already counted in an earlier pool.
What to do with all this is a personal call. Some families ask their utility for a measurement, and some check a home's distance from a line before renting or buying. Some buy consumer products, for example the EMFCALM 12-pack from a shop that sells EMF stickers for phones and other devices. None of the studies above looked at products like these, and we don't evaluate them.
The figure to watch in the next pooled analysis is the count of exposed cases, more than the odds ratio. Ahlbom's group had 44 in 2000. Kheifets's 2010 pooled analysis had 26 at 0.3 µT or more. With groups that small, a few children more or fewer can move the headline result.