The following is reprinted, with permission, from Microwave News. This article/commentary originally appeared in the September/October 2000 issue.

Leading Epidemiologists See Childhood Leukemia Risk at 4 mG

A pooled analysis of raw data from nine different electromagnetic field (EMF) studies has found that children exposed to 4mG or more were twice as likely to develop leukemia. There was no excess risk at lower exposures.

“The level of [statistical] significance that we see for the excess risk at high exposure makes chance an unlikely explanation,” an international team of leading epidemiologists writes in the September issue of the British Journal of Cancer (83, pp.692-698, 2000).

Led by Dr. Anders Ahlbom of the Karolinska Institute in Stockholm, the study team includes Drs. Nicholas Day of the U.K., Maria Feychting of Sweden, Martha Linet of the U.S., Mary McBride of Canada, Jorg Michaelis of Germany, Jørgen Olsen of Denmark, Tore Tynes of Norway and Pia Verkasalo of Finland, each of whom has led an important study in his or her own country.

The new findings are similar to those announced last year by Dr. Sander Greenland of the University of California, Los Angeles (see MWN, S/O 99), in an analysis that combined data from many of the same studies. “It’s a pretty consistent picture. It’s more consistent than one would have any right to expect, given the differences in how these studies were carried out,” Greenland told Microwave News this September. “The main point is, you don’t see anything until you get into the higher categories.” Greenland’s findings will be published in the November issue of Epidemiology.

The Ahlbom and Greenland papers reflect a growing consensus on the apparent association between magnetic fields and childhood leukemia. But there is no consensus on what those data mean. The question of causation—that is, whether EMF exposure is actually responsible for the observed increase—is still very much at issue. Ahlbom and colleagues note evidence of selection bias in some of the studies, and conclude only that, “The explanation for the elevated risk estimate is unknown.”

“We rule out chance and say that selection bias could explain some of the excess,” Ahlbom said in an interview with Microwave News (see p.12). “That leaves us with the options of causation or other possible kinds of bias... .We may not be able to resolve this in the near future.”

The question of causation has important public health implications. While 4 mG is considered unusually high for a residential exposure, present international limits allow for exposures up to 250 times higher. According to a 1998 study by EPRI, between 6 and 12 million Americans live with average exposures above 4 mG (see MWN, M/J98).

Day, of Cambridge University, told Microwave News that, “Interpretation of the finding of an excess risk above 4 mG has to be cautious.” Day headed up the EMF component of last year’s U.K. Childhood Cancer Study (UKCCS; see MWN, N1D99 and JIFOO). “The U.S. study makes a large contribution to this excess, so a lot depends on how one interprets that study,” he said.

Linet, who led the U.S. study, does not think the new combined analysis strengthens the case for an EMF-cancer connection. “All of this increase may be due to bias,” she said in an interview. “We’ve shown in our own study that some of it has to do with who participates and who does not.”*  Linet is at the National Cancer Institute (NCI) in Bethesda, MD, which sponsored her study (see MWN, J/A97).

Does bias play this big a role? “You’ve had this relatively consistent pattern, across different countries and across studies that were conducted in different ways,” Greenland said. “It’s hard for me to see how selection bias would operate in the same way in all these cases.”

Day stands somewhere in between. “Bias and confounding may be the main factors in operation,” he said. On the other hand, Day said, “an excess was seen which is very unlikely to be due to chance, and is consistent with a causal role for EMFs at these high levels. It cannot just be dismissed.”

The new paper notes that leukemia in children is not well understood, so “one cannot exclude” a potential confounding effect from some as-yet-unidentified risk factor. It does conclude that “mobility, traffic exhaust, type of dwelling and urban/rural residency are not important confounders” in these studies—but Greenland thinks this may be premature.

“Many of these things are not well-measured,” Greenland observed. “Traffic exhaust is one of the most controversial. The data on potential confounders is pretty weak, overall.”

“It is frustrating that, with all these studies, you can’t say all that much yet,” commented Greenland. “It’s true that bias or confounding might explain the excess risk. But it’s also true that it could be a real effect.”

In Michaelis’s view, “The main problem is that there is so far no plausible explanation of the observed association.” In the absence of a specific mechanism, he told Microwave News, we cannot conclude that EMF exposure leads to childhood leukemia—even above 4mG. Michaelis is at the University of Mainz.

Ahlbom sounded a similar theme. “If we had supporting experimental data,” he said, “the epidemiology would have been strong enough for a causal interpretation quite some time ago.” But without that, “the requirements on epidemiological data become rather strong.”

“Wire-Code Paradox” Rejected

The 1996 EMF report from the National Academy of Sciences—National Research Council concluded that childhood leukemia was linked to proximity to power lines—as measured by a system of “wire codes”—but not to measured magnetic fields (see MWN, N/D96). Greenland and Ahlbom each conclude, however, that this so-called “wire-code paradox” does not exist.

Both pooled analyses show that at higher levels of exposure, measured field levels are in fact linked to increased risks. Ahlbom and colleagues also question whether the wire-code studies are valid at all: They note that the two North American studies they included—by Linet and McBride—”show no evidence of increased risk...in high wire-code categories.” It is “unclear” why previous wire-code studies found the opposite, they write, but there may have been “considerable potential for bias.”

Day takes an even stronger view. “What does emerge clearly,” he said, “is that the results of the early studies, from which interest in the [EMF—childhood leukemia] hypothesis developed, were the result of bias.” He cited both the NCI data and the pooled analysis led by Ahlbom. “If there is any relationship with magnetic fields,” Day argued, “it is not the one put forward on the basis of wire codes.”

Greenland disagreed: “That may be true, but we can’t conclude that yet.” He pointed out that the pooled analysis by Ahlborn, Day and colleagues only included “the two most negative of the wire-code studies.” While Greenland’s own analysis found that wire-code studies were less consistent than studies based on measured fields, the data from the eight wire-code studies he examined do show a significant excess risk for children living near “very high-current configuration” power lines.

The original plan for Ahlbom’s pooled analysis (see MWN, J/F96 and M/J99) was to include all European studies of EMFs and childhood leukemia that used either calculated fields or 24-or 48-hour magnetic field measurements. Three others were added later: the NCI study, Dr. Mary McBride’s research from Canada and a New Zealand study by Dr. John Dockeity of the U.K.’s University of Oxford (see MWN, M/J99 and N/D99).

“We felt that if we could also incorporate new studies from non-European countries, this pooled analysis would be up to date and presumably stay current for several years,” the paper states.

Ahlbom and colleagues analyzed raw data for 3,247 children with leukemia and 10,400 controls. The children’s magnetic field exposures were based on in-home measurements. They used geometric means—”because they are less affected by outliers”—to gauge average exposure in the year prior to diagnosis.

Previous EMF-cancer studies “have sometimes been criticized on the grounds that the findings might be a consequence of so-called data dredging,” Ahlbom’s team writes. To avoid this, “We specified which primary analyses we planned to do and how to do them before we commenced the analysis,” doing so “before the results of several of the individual studies were known.”

There have been other meta-analyses besides Ahlbom’s and Greenland’s, but these have not combined the raw data from different studies (see MWN, S/O94, J/F99 and J/F00). Ahlbom writes that access to the raw data “gave us two substantial advantages.” First, it allowed his team to make the data from different studies “as compatible as possible,” especially with respect to the categories for exposure assessment. Second, by analyzing a larger number of cases together, it became “possible to analyze..higher cutoff points than the commonly used 2 mG” for the high-exposure category.

Ahlbom’s pooled analysis includes 44 cases and 62 controls with average exposures of 4 mG or more, a far greater number than in any of the individual studies.

Greenland’s study does not have data from the UKCCS, and Ahlbom’s does not include several of the 15 studies used by Greenland—from Mexico and others from Norway, the U.S. and elsewhere. “But none of this seems to matter much in terms of the results,” commented Greenland. “We found in our analysis that the results don’t depend much on any one paper.” Thus, he said, “it’s not surprising that the conclusions of these two pooled analyses came out about the same.”

Neither pooled analysis has data from the recent study by Dr. Lois Green of the University of Toronto, which found significant increases in risk, especially among younger children (see MWN, J/A99). Ahlbom’s paper explains that Green’s “exposure information...was not similar enough to justify inclusion,” and Greenland said that Green’s data were not available in time.

Both Linet and Day emphasized that neither combined analysis showed a risk for most children. “There is no evidence of any appreciable risk below 4mG,” stated Day. “These are the exposure levels of relevance to the great majority of people in this country, and so this confirms the findings from the U.K. study.”

“There is no risk for the 99.2% of kids with exposures under 4 mG,” said Linet. For this reason, she does not view further EMF studies as a priority.

*The pooled analysis notes that controls were “generally characterized by higher socioeconomic status than cases.” particularly in the U.S. study. The UKCCS found some connection between lower status and higher levels of magnetic fields, and an analysis of the NCI data by Linet and Dr. Elizabeth Hatch found that people with lower socioeconomic status were more reluctant to participate fully, especially as controls (see MVN, M/A 00). The NCI researchers concluded that selection bias “led to a slight overestimate of effect in our study.”  [For an opposing view that suggests these biases led to an underestimation of the effect, see the Wartenberg study.]