Assessment of Health Effects from Exposure to Power-Line Frequency Electric and Magnetic Fields. "
NIEHS working group report (1998)

Comments by:
Richard Wilson
Department of Physics
Harvard University
Cambridge, MA, 02138
September 24th 1998
wilson5@fas.harvard.edu

Introduction

In these comments I will concentrate upon one issue: whether or not it is appropriate to claim that "ELF EMF" are possibly carcinogenic to humans (Group 2B) according to an IARC classification. I note that the IARC classification scheme was developed for chemicals. For chemicals the classification has by now a well defined meaning and that meaning tends to define action to be taken by governments, industry and society generally. It is unclear that the meaning can be properly transferred to any another agent without considerable work to clarify what the meaning might be. The NIEHS report does not discuss this properly. In 1991 the U.S. Environmental Protection Agency drafted a report(1) which suggested inter alia that electromagnetic fields be considered a class C carcinogen - a recommendation similar to that of the NIEHS committee. But it is noteworthy that a review committee for that report(2) did not agree. A large part of the disagreement was that the IARC classification scheme is inapplicable.

    What does it mean to say that something is possible?

There would be few people who would claim that it is not possible for a particular chemical, even at a low dose to produce cancer. There are no fundamental physical laws that suggest that sodium chloride (common salt) cannot cause cancer. The latter is not impossible in the sense that the fundamental scientific laws do not exclude it, although it may not be true.(3) The disagreement might come about whether it actually does so. But that is not the case for electromagnetic fields. There are a number of distinguished scientists(4) that insist that it is not possible for a human being to detect, let alone have an adverse effect from magnetic fields at the low intensity of the epidemiological studies. Such detection would be at odds with fundamental laws of electromagnetism, including the theory of relativity coupled with thermodynamics unless a detector exists in the human body which is many centimeters in size and of high conductivity (such as a coil of copper wire wound around the stomach)(5). Pathologists assure us that such a large detector does not exist. Thus the claimed effect is impossible unless new physical laws are suggested. I reiterate that to say that it is possible that electromagnetic fields cause cancer has therefore a vastly different meaning than a statement that a particular chemical causes cancer and such a statement must be made with much greater care. Some of the argument is discussed in the working group report. But there is a confusing summary of the basic science in 4.8.4 which (inter alia) implies (incorrectly) that the constraints imposed by Nyquist's theorem can be avoided by a consideration of non-linear processes(6).

The importance of dose

Since the scientific statement that it is impossible for the low intensity fields to cause cancer depends on the fact that they are very low, it is important to consider intensity or dose when considering evidence for carcinogenicity. One problem with the IARC classification even for chemicals is the failure to consider dose explicitly. Thus a chemical might be considered carcinogenic to people because at high doses it has been known to cause cancer even though exposures are now so far below the high figures of the past that the risk is low. In practice many scientists refuse to allow that a particular chemical is in a carcinogenic classification if the evidence comes from data at exposures much greater than environmental exposures. Examples include the failure to call gasoline (petrol) a Class 1 carcinogen although benzene which is a component of gasoline and is therefore always present in small quantity, is a class 1 carcinogen. The same applies to common salt as noted in reference 3 above. But a bad situation is taken a step further by the working group. I know of no chemical which is claimed to be carcinogenic because of data at low doses which does not also, and in greater measure produce cancer at higher doses. Yet that is the claim for electromagnetic fields.

The importance of exposure or dose is even greater for electromagnetic fields. There is no firm evidence at ANY field of a carcinogenic effect. Yet people have been exposed at fields far greater than the residential of the epidemiological studies that claim to show an effect. Of course electric currents are dangerous. They can cause electric shocks and ultimately death. Electrocution by lightning strikes have been known from time immemorial, and as of 1994 a thousand people are accidentally electrocuted every year in the United States, usually by defective electrical wiring, and some are deliberately electrocuted by the state. But exposure to electromagnetic fields even as high as 10,000 Gauss (1 Tesla) has not been proven to cause cancer.

The effect of magnetic fields on cells is expected, on general symmetry principles (parity conservation in electromagnetic interactions) to vary as the square of the field (B2) at low fields. These arise because the magnetic field comes from the motion of electric charges rather than the charges themselves, and it is not sensible to envisage cancer incidence changing sign as the magnetic field changes sign. Detailed models show this explicitly. But one might ask at what level the B2 law breaks down? Some indication is given in an experiment of mine of 23 years ago. Some materials produce "scintillation light" when irradiated by charged particles. When they are simultaneously exposed to a static magnetic field, this increases slightly (varying as B2 at low values of B) but flattening off, or saturating above 10 gauss to give a 1 1/2% effect above 50 Gauss(7). Thus 10 Gauss might well be an appropriate limit for atomic or molecular processes. Many such effects on atomic and molecular systems are discussed in the working group report but all the ones that are determined to be reliable are at magnetic fields above 1 Gauss.(8)

For any postulated effect on people, a similar saturation might be expected at similar fields. But one should still find an effect at 30 Gauss which is 10 to 100 million times the effect at 3 milliGauss. Indeed since magnetic fields are used in laboratories up to 1 MegaGauss (100 Tesla), the range of fields under consideration is a factor of 109 and the range of the square of the field is 1018. To pretend that it makes no difference is absurd.

Electric power lines produce magnetic fields in nearby houses of 3 milliGauss or less. Yet travelling on Amtrak between New York and New Haven produces fields at the passenger averaging 32 milliGauss, which is 10 times greater than exposure at home, with peaks 10 times larger still(9). There have been no studies on the effects of the magnetic fields resulting from such travel, for example on daily commuters, but a study of Norwegian railroad workers (who work on an electric railroad with similar exposures) showed no effect(10). Electric blankets used to give still larger fields (300 milliGauss) before they were made with twisted pair wire, but epidemiological studies of people who use blankets has not shown a very large effect, even though the general argument suggests that the effect should be 10,000 times greater than the effect at 3 milliGauss. The initial epidemiological study on the effects of electric blanket use showed a small effect, but an study with improved methodology found none. Moreover still larger fields are known in laboratories.

Many scientists and other test "guinea pigs," including myself, have deliberately (or inadvertently) exposed themselves to static magnetic fields of tens of kiloGauss (many Tesla) on many occasions without obvious adverse effect. Ten seconds of work in such a field would be equivalent, on the basis above, to a lifetime of exposure to 3 milliGauss fields. The only observed effects were a tasting of the fillings in the teeth and flashes of light in the retina as the head moves in the field. These effects, magnetophosphenes are discussed in 4.6.12 and have been known for a century. They also occur with alternating magnetic fields. But they are not seen when the intensity is below a threshold which for 60 Hz fields is about 100 Gauss - far above the fields where there are claimed effects of cancer. Electric fields can also be perceived. In elementary physics lectures a demonstration is frequently given when the lecturer is charged up to 100,000 volts and his or her hair stands on end. This can also occur in thunderstorms.(11)

These observed effects where people can see and taste electromagnetic fields are at high intensities. Not the low intensities of the epidemiological studies. Two sets of experiments reported in section 4.6.1.1 at fractions of a Gauss (microTesla). They are not of equal quality as can easily be seen by reading the papers. The earlier ones by Tucker and Schmitt(12) originally found an effect with one volunteer subject, later traced to an auditory clue from a generator. This may have occurred in the later study(13). This is reminiscent of the telepathy experiments of Professor Gilbert Murray in 1915 who found strong evidence for telepathy with several subjects until all auditory clues were removed, even those which were below the level of conscious hearing.

Although there are a few theoretical models that suggest that effects of specific environmental pollutants first increase with dose and then decrease, these have so far only been postulated to occur in situations where there is unequivocal evidence of cancer at high doses. The most important conclusion that can be drawn from the mass of data is that there is no exposure level where electromagnetic fields have been demonstrated to cause cancer. There seems, therefore, no reason to make electromagnetic fields a logical exception to the usual rule that "more is worse." The onus of proof on someone who wishes to claim otherwise is great.

This crucial point has been pointed out many times by many people in the last 15 years. There is only a very perfunctory discussion of doses in the working group report. In view of the importance of dose this shows a remarkable lack of understanding of the logic of the problem of assigning an effect to a cause.

The epidemiological data

The working group state (section 5.1.4) that the "evidence in support of the decision to classify ELF EMF into group 2B is driven by results of studies on childhood leukemia in residential settings and on CLL in adults in occupational settings". The childhood leukemia results are far more limited than studies of adult leukemia. A summary of the data on adult leukemia and on adult brain cancer(14)

(ignored by the committee) listed 86 studies, and performed two types of meta analysis - one using a simple weighted average and another using the DeSimonian and Laird method. They showed that the distribution of the results among studies was NOT Gaussian, but had tails indicating errors other than those of statistical sampling.

Although some individual studies had Risk Ratios greater than 2, this was with large statistical error. The simple weighted average Risk Ratio for CLL for occupational studies was 1.27 and for residential studies 0.97. As often emphasized by epidemiologists, Risk Ratios this low are not considered seriously by epidemiologists unless there is other supporting evidence - which in this case there is not.
The working group seems to have relied upon a meta-analysis by one of its members(15) that is not easily available and upon which I cannot therefore comment directly. But there is in section 4.3.4 a discussion of the various published studies including those from the National Cancer Institute.(16)(17) These NCI authors note that the effect on childhood leukemia vanishes when they exclude the wire code data of the subjects who refused to participate suggesting a selection bias that may well have been operative in the other studies also.(18) This view is explicitly contrary to the statement in the first sentence of the third paragraph of page 186 of the NIEHS report.

There is another type of bias that is common in observational studies. I call it the "Feynman trap" honoring the Nobel laureate who described it very eloquently. If any graph or method of analysis is chosen after the observations are made, the ordinary statistical rules are invalid(19). Even excellent scientists have fallen into the trap at various times. Epidemiologists often refer to the first study in a field as a "hypothesis generating study" to emphasize that it by itself should not be considered as proof. It is clear that the study of Wertheimer and Leeper(20) generated a hypothesis. But however well done it cannot be considered as proof in itself(21). The results should not be averaged in with the rest. Since the other studies described by the working group report were not exact repetitions, it is likely that the Feynman trap was operating to some extent in them also(22)

. This would not render the studies useless but would necessitate increasing the uncertainty band to account for such a problem. The procedure for doing so by a Monte Carlo simulation has been discussed by Shlyakhter et al.(23) This, or something like it should have been discussed somewhere in the NIEHS report.

In view of the discussion of resonance theories later in the report, it is important to note that if studies of exposure at 50 Hz and exposure at 60 Hz are averaged, "supporting evidence" cannot come from any resonance theory with a narrow resonance as would be needed to overcome the theoretical problem of thermal fluctuations. I find no such discussion in the working group report. On another minor detail, the lack of understanding of the basic properties of electromagnetic interactions is displayed in some of the epidemiology studies listed here when the Risk Ratio is plotted agains the root mean square field intensity rather than the mean square field intensity. The data look very different (and in some cases show a very peculiar dose response relationship) when plotted against the correct variable.
 
<>Proceeding from a statistical association (correlation) to an assignment of causality.

I consider that the statistical evidence for an association between cancer in children and exposure to electromagnetic fields is weak, and that there are many possible sources of non statistical error that can outweigh this evidence. Nonetheless the evidence does exist. Yet we must recognize that a statistical association between two variables, in this case electromagnetic field intensity (or its surrogate) and cancer is insufficient to prove causation. There is a well known statistical association between the (declining) number of storks in Germany and the number of babies being born. This association (statistical correlation) is far stronger than the association between electromagnetic fields and childhood leukemia. Yet few people, and I venture to suggest not one member of the NIEHS working group would venture to state that it is possible that storks cause babies. Of course there are several reasons. There are other possibilities that are considered more plausible and the idea that storks bring babies is often regarded as impossible (although it violates no fundamental physical laws).

Sir Austin Bradford Hill(24) in his Presidential Address to the Section of Occupational Medicine of the Royal Society of Medicine (UK) suggested a list of "attributes" of the association to be considered. Although some epidemiologists consider this list to be out of date, it is nonetheless a list that has been widely used in discussions such as this, and is a useful framework. Although the working group did mention the problem and referred to Hill's criteria, it was somewhat perfunctory. Nor did the working group provide reference to the places where these have been discussed(25)

. But they did in section 5 follow some of the ideas. I find the logic somewhat difficult to trace and confusing. Accordingly I discuss in a later section whether or an association between a surrogate for magnetic fields and childhood leukemia possesses Bradford Hill's attributes. The (weak) association between electromagnetic fields and childhood leukemia has only one of the attributes that Bradford Hill suggests are desirable.

The Basis for Belief in a Phenomenon

Before addressing these suggestions in detail I emphasize that belief is invariably, and properly, strongly influenced by pre-existing prejudice and biases.(26)  I illustrate this by considering the following three declarative statements:

"I saw a dog running down Fifth Avenue"
"I saw a lion running down Fifth Avenue"
"I saw a stegosaurus running down Fifth Avenue"

Little evidence is needed to be convinced that a dog did, in fact, run down Fifth Avenue: dogs are common in contemporary urban society. We might demand a bit more evidence if the speaker were claiming damages for a car accident caused by the postulated dog.   However, before believing the second statement, most of us would require much more auxiliary evidence. Sufficient auxiliary evidence might be that a circus train had crashed and released its animals or that a lion had escaped from the zoo. The geographical context of belief is apparent; if "down Fifth Avenue" is replaced by "along the Serengeti plain of Tanzania" it is likely that belief would be widespread. Most people, however, would steadfastly insist that seeing a stegosaurus is absolutely impossible. The author of the statement would be considered victim of an illusion, drunk, or a lunatic. I suspect that few members of the working group would admit such evidence without qualification.

Whether or not a claim seems stupid, merely possible, or quite sensible often depends upon the knowledge and experience of the perceiver. On the other hand, what is initially thought to be stupid sometimes turns out to be true(27). In order to avoid the problem of ignoring a new idea, as the establishment did with Galileo, we need a few guiding principles. The criteria that the Nobel Prize-winning chemist Irving Langmuir(28)   proposed to identify pathological, or fraudulent, science can be helpful in identifying claims of hazard that society should ignore:

 1. The maximum effect is observed by a process of barely detectable intensity, and the effect is largely independent of the intensity of the apparent causal agent.
2. The effect remains close to the limit of deductibility.
3. Claims of great measurement accuracy, or of profoundness, persist in the face of mounting evidence to the contrary.
4. Theories are put forward that fail the test of being the simplest explanation for the available information.
5. Criticisms are met by ad hoc explanations: the proponents "always have an answer -- always."

I emphasize that the belief of people who are knowledgeable about electromagnetism or cellular biology, the key disciplines involved in the issue of potential risks of EMF to health, is properly governed by their experience.

Other authorities

There is a growing consensus among scientists that the epidemiological studies do not show an association between 60 Hertz EMF and cancer. We refer to reviews by committees composed of distinguished and competent persons and set up by responsible public bodies and professional associations. Of course NIEHS was examining the issue independently. But it is good scientific practice to examine not only data but other scientists interpretation of the data. Although NIEHS might well disagree it seems mandatory to explain in detail the reasons for the disagreement. I cannot find a listing or a discussion in the NIEHS report. I list some reviews below:
(1) There have been three committees of the National Academy of Sciences(29) to review the issue of effects of electromagnetic fields upon health.(30)
(2) An extensive study by Oak Ridge Associated Universities (ORAU)(31), carried out at the behest of the US Government (Committee on Interagency Research and Policy Coordination or CIRRPC).
(3) A report by the Connecticut Department of Health (CASE)(32)

(4) A report of to the National Radiological Protection Board (NPRB) of Great Britain of a committee chaired by Sir Richard Doll(33)
(5) A committee set up by the Governor of Florida(34)
<>(6) The World Health Organization (WHO)(35).
(7) A draft report by the US Environmental Protection Agency(36). But this draft report should only be read in conjunction with the critical report of the review committee.
(8) Reports from the Netherlands,(37) Denmark,(38) and Ireland(39) ;
(9) Report of the French National Institute of Health and Medical Research(40), and the French Academy of Medicine(41)
<>(10) Reports prepared for the states of California, Texas(42),Illinois(43),Maryland(44) and Colorado(45).
(11) The Council of the American Physical Society issued a statement(46) following a report by their "Program on Public Affairs"(47)

We note that none of the committees in (1) - (11) have concluded that there is an effect of electromagnetic fields against which we must guard. The National Academy Committee stated that:

[T]he current body of evidence does not show that exposure to these fields presents a human-health hazard. Specifically, no conclusive and consistent evidence shows that exposures to residential electric and magnetic fields causes cancer, adverse neurobehavioral effects, or reproductive or developmental effects.
In addition a group of distinguished scientists including 6 Nobel laureates (2 each in physics, chemistry and medicine) and three distinguished epidemiologists (one a Dean, one a Chairman of Department) submitted a brief of amicus curiae to the Supreme Court of the State of California(48) (and similar briefs to a district court in NY and to an appeal court in California). That court accepted that there was no scientific case of which the court should take notice.

The Epidemiological Principles

Since it is the epidemiological evidence that has been the primary justification for this conclusion of the working group report, it seems worthwhile reviewing it in the light of scientific principles that are used to evaluate whether a statistical association that is found should be considered to be causal. I therefore address this list:
<> 1. Strength
2. Consistency
3. Specificity
4. Temporality
5. Biological gradient
6. Plausibility
7. Coherence
8. Experiment
9. Analogy

The "strength" of the association was most convincing in Percival Pott's original observation a century and more ago that almost all chimney sweeps developed scrotum cancer. (The Risk Ratio was very large). Little other evidence seemed necessary. But even for cigarette smoking where the risk ratio is over 10, it was many decades before scientists were convinced of the causal connection. The risk ratios discussed by the working group are far less.

Although the "ecological" studies started by Wertheimer and Leeper(49) showed a Risk Ratio of about 3, the statistical significance was marginal. The review by Washburn, et al.(50) found a Risk Ratio of 1.57 for some sort of association with the presence of power lines which is statistically significant. Loh et al are in agreement. But one should note the difference between the standard practice in epidemiology and the standard practice in the physical sciences. Physical scientists routinely discuss non-statistical and systematic errors in great detail, and usually attempt a quantitative description of them. Epidemiologists sometimes discuss the non-statistical errors in the text, but do not make a quantitative estimation or include the qualifying phrases in the abstract of an article(51). Great caution is necessary in any interpretation of these numbers, especially when the effect is small. We note that the reduction Risk Ratio from RR = 3 (Wertheimer and Leeper's claim) to RR = 1.57 is a 3.5 fold reduction in predicted excess cancers since the excess cancers are proportional to RR = 1. This is the type of reduction we would observe were the "Feynman Trap"(52)  in partial operation during the first study.

If the average of the either the ecological studies or the occupational studies were a single study, a Risk Ratio of 1.5 would not normally be considered large enough to be deemed evidence for a causal relationship. Of situations where the measured Risk Ratio is less than 2, only two have been accepted as evidence of harm, and these are special situations. The effects of tobacco smoke on the families of smokers (with an average 19% increase or a Risk Ratio of about 1.19) has been accepted by the Environmental Protection Agency(53), and by many physicians and scientists: this is because tobacco smoke is known to be hazardous to the smoker who has a large dose. However, some scientists still question the evidence because there are problems with the measured amount of smoke and they believe the EPA acceptance to be entirely political. Likewise it is generally accepted that there is an effect of X rays during pregnancy on the probability of childhood leukemia, even though the risk ratio averaged over studies is less than two(54), because at high radiation exposure radiation does clearly cause cancer(55). But there is no intensity or situation where electromagnetic fields are known to cause cancer, so one cannot argue that the existence of an effect in a higher field reduces the standard of proof of causation for lower fields.

Hill's Attribute 2 is whether "the same result has been repeatedly observed by different persons, in different places, circumstances and times."(56) The record is mixed. The initial observation that excess childhood leukemias are observed near power lines has been repeated a few times, and there seems to be a consistent relationship with proximity but not the measured magnetic field itself. The Swedish residential study(57)  is consistent with earlier studies in that no association was found with fields measured contemporaneously, but since no one else calculated fields from wire codes and historical usage, the statistically significant result here cannot be properly said to be completely consistent with earlier data. Consistency is also related to the next criterion, specificity. As noted earlier it is not enough for successive studies to find that cancer is elevated in the presence of electromagnetic fields. The Feychting and Ahlbom study suggested that magnetic fields cause an increase in Acute Lymphocytic Leukemia but not in Chronic Lymphocytic Leukemia, whereas that of Floderus, et al. showed an increase in CLL, but not ALL. The most recent study of Savitz and Loomis(58) shows no increase in any leukemia, but a small increase in brain cancer! Moreover the residential studies show no increases in adults, and only effects in children are claimed. The NIEHS working group discussed these various issues but made no attempt to see whether there was consistency in the sense that if an effect is seen one study it should be seen in all others where the sensitivity is adequate. This was done by Loh et al. (59) who showed that there is considerably larger variation between studies than given by a statistical error only. That is partially taken account by the Der Simonian and Laird method and the uncertainty band increases.

It is also important to be clear what it means to be consistent. The characteristics of electromagnetic fields from power lines is different in Europe and USA. In the former, fields alternate at 50 Hz and in the latter at 60 Hz. If one wants to argue that studies in Europe and USA are consistent one cannot at the same time use a resonance model to provide auxiliary evidence(60).

Attribute 3 is often a strong argument. Hill states that "if the association is limited to specific workers and to particular sites and types of disease, and there is no association between the (postulated cause) and other modes of dying, then clearly there is a strong argument in favor of causation."(61) This attribute must be interpreted with full understanding of the generality of the exposure mechanisms. Unlike chemical carcinogens, which give the dose at well defined parts of the body, (those to which the chemical penetrates) electromagnetic fields might well affect all parts of it. In this respect, and this respect only, the problem might be similar to external gamma radiation, which affects all parts of the body. However, if electromagnetic fields produce several types of cancer in one group of people, they should produce the same types of cancer in similar proportions in all other groups similarly exposed.(62)

Attribute 4 of an association is that the adverse outcome occurs after the postulated cause by whatever delay (latent period) has been seen in other studies, or is reasonable from biological principles. "Which is the cart and which the horse?"(63) In most of the existing epidemiological studies, the cancer incidence has not been associated with a contemporaneously measured electromagnetic field, in spite of searches for an association, nor has it been possible to associate incidence with a field measured at an earlier time, because of an inability to get data. Instead the association is with a field assumed, or calculated, from configurations of high tension transmission and local distribution wires.

A very important anchor for epidemiologists is attribute 5 of Hill's list -- the existence of a biological gradient or dose response relationship. In the usual models, "more is worse" and "less is better," and the adverse effect is at least proportional to the exposure if it does not rise faster than proportionality suggests. There is no accepted medical effect of a pollutant where the effect does not increase as the dose increases, at least initially. As noted above, the claimed effect is very far from satisfying this rule.

It is important here that not only should there be a dose-response within a study but also a dose response by comparison of effects between two studies one at high "dose" (my inserting my head into the 16 kiloGauss (1.6 Tesla) cyclotron field for example) and the low "dose" of the residential studies. The absence of a normal dose response was noted above.

Attribute 6 is that the claimed effect be biologically plausible. Hill emphasized that we cannot always demand this because "what is biologically plausible depends upon the biological knowledge of the day."(64)  But one can interpret this broadly, and Hill does this under "Coherence". A mechanism must be postulated that is not at variance with other knowledge. Before cancer was widely known, an attribution of cancer to any particular cause could have been considered implausible. However, such an attribution was not considered to be impossible. There exist models of cancer, unproven and not easily provable, but nonetheless plausible. That is not the situation with electromagnetic fields. At the present time, no mechanism has been successfully proposed by which 3 milliGauss magnetic fields can cause any cancer(65)

The NIEHS working committee were wishy washy on this. They say (correctly) that "all the theories.... are speculative and unproved". But they give a lot of space to a discussion of theories such as that of Lednev without saying that a consensus of scientists is that they cannot be right because of the fundamental laws above. Indeed this is a seriousweakness of the report in many places. It is a listing (without analysis) of a number of claimed effects and references thereto, which though useful in some ways is inadequate as a scientific approach and completely inadequate to justify any particular recommendation. Because of the absence of any such model, any association of low intensity electromagnetic fields with cancer fails to possess Hill's attribute 6. In the language I used earlier, seeing such an effect would be like seeing a stegosaurus on Fifth Avenue. This point is discussed further under "Analogy".

Attribute 7 requires coherence of the data. This attribute is related to the general plausibility mentioned in the previous paragraph. The idea that the association of lung cancer with cigarette smoking is coherent both with the increase in cigarette use, and the increase in lung cancer that followed it by a couple of decades. It is also coherent with the sex difference in both these variables.(66) Hill also mentions, under this heading, coherence with laboratory experiments on animals and in vitro. In the past many experiments on the effects of electromagnetic fields on animals have been quoted as evidence that low intensity magnetic fields cause cancer. I concur with the NIEHS working group that the animal data are not persuasive(67). Many experiments have been performed in vitro. The working group (68) concluded that a limited number of well-performed studies provide "moderate" evidence. It is hard to tell from the working group report which are the studies to which this conclusion refers. In the text(69) the working group refer to the Ion Paramagnetic Resonance theory. The theory in itself makes no sense but the purported experimental verification is wrong. A simple examination of the data shows that they could not have been obtained in the manner described by the authors because they are more consistent than statistical theory allows. Although the oft repeated claim of Adair that the data must have been fabricated may be extreme, it is clear that they must not be used to justify the theory and a failure of the working group to say this clearly puts doubt on this aspect of their report.    Apart from these and other experiments by the same discredited author, it seems that the working group are referring to data at fields above 1 Gauss (100 microTesla). If these are to be used as confirmatory evidence for an effect at low fields an explanation is necessary on how this can be so and the working group fail to provide one. Moreover there is no indication of any of these studies that they should be carcinogenic, anticarcinogenic, or neutral. The working group does not explain why these should be considered support for a claim of cancer at any intensity. Indeed the discussion of such experiments the working group again follow the non-scientific procedure of quoting experiments and theories without analysis and suggesting that legitimate criticisms and illegitimate rebuttals are equal in weight.

Earlier suggestions were that the experiments on calcium efflux from chicken brains substantiate the epidemiological results. Fortunately I do not find such a statement in the NIEHS report. But I would have preferred a more explicit statement that they should not be considered for two reasons. Firstly the results of these efflux experiments have not been closely the same when they have been repeated, so that the ordinary scientific concept of repeatability, which can and should be applied to laboratory experiments is not satisfied. The arguments of some proponents of an effect that the attempted repetitions have not been properly done are here irrelevant because the burden of proof is on the proposer that electromagnetic fields cause cancer. Secondly, (as noted above) even if such laboratory tests are found to produce results contrary to existing scientific understanding, and show that there is a measurable biological effect, they say nothing about cancer. The measurable effect might be benign, or even a positive good.

Attribute 8 demands that the results be consistent with experiment. Hill here considers the possible experiment of taking preventive action by cessation of exposure. "Does it in fact prevent?"(70) No one has dared to propose cessation of exposure to electromagnetic fields, because society likes the enormous benefits that the technology brings. The opposite of cessation has of course occurred. The considerable increase in electricity use in the last century does not seem to have been accompanied by increases in the cancers discussed(71).

Attribute 9 would suggest that an effect could be accepted if it is analogous to another situation where the proof is more substantial. We can compare the limits of sensitivity for detection of electromagnetic fields with the detection limits of various sense organs. The physiological literature describes over 50 sensory modalities for living organisms(72) For example in the retina of the eye,  cells are sensitive to an individual quantum of light - the limit of sensitivity. But in no situation is there a well established physiological effect at an intensity below the theoretical limit of sensitivity(73).

 - No "new" physics has been necessary to understand the limits of performance for sensory systems. "Limits to the deductibility of small systems are set by noise" (fluctuations ).

- Some sensory systems operate close to the physical limit of sensitivity, but none have been found to violate physical principles. "Perhaps [our] most important [advance] has been the realization that a sensory system that reaches the physical limits to the performance is exceptional"

Thus I find that the claim that electromagnetic fields at low intensities cause cancer barely satisfies even one of Hill's principles. I do not believe that the assignment "possibly carcinogenic to humans" has been used by IARC for any other substance or agent for which so few of Hill's attributes exist.

What was the working group thinking?

Since the statement that electromagnetic fields (at the low intensity of the epidemiological studies) can possibly cause cancer seems contrary to ordinary interpretation of science, it is obligatory to consider what the working group was thinking and why they said what they did. I consider here three basic possibilities.

(1) The working group consider the word "possible" in a much more general sense than I use here or has ever been used before by anyone discussing the IARC classification.
(2) The working group was not sufficiently knowledgeable about the basic science that the impossibility was not clear to them.
(3) The working group really believe that the "possibility" that the electromagnetic fields cause cancer at the low intensities under consideration is large enough to outweigh the presumption caused by a century and a half of knowledge of electromagnetism and thermodynamics.
I consider each in turn.

If possibility (1) is correct it seems mandatory for NIEHS or any other authority discussing the report to make absolutely clear the meaning of the word "possible". I do not find such a statement in the report.

However the make up of the working group suggests that possibility (2) may be the right one. In scientific studies there is a strong, and often correct tendency to assign a phenomenon one does not understand to a cause one does not understand. In this case perhaps to assign the weak association to electromagnetic fields. Some of the greatest scientists have done this in the past. Sir William Crookes in his Presidential address to the British Association for Advancement of Science in 1897 suggested an attribution of telepathy between people to a region of the electromagnetic spectrum that no one understood at the time - the far ultraviolet. Few scientists would accept that now. The way of avoiding this trap is to have a committee which includes experts in the relevant disciplines. As noted above the discussion of non linear thermodynamics in the report is confusing. There is in the text no clear statement that this or that experiment is unreliable. It suggests that the group did not understand the basic physics. Indeed the group included only one member of the American Physical Society. Nor is there a cellular biologist - who might be expected to understand well the way in which electromagnetic fields act upon cells(74).

Since the epidemiological studies seem to be the primary reason for the vote of the group, it is worth noting that only 4 of them could by any reasonable stretch be called epidemiologists(75). One of the 4 did not agree with the rest of the group on this crucial matter.
Several of the group have been working on projects whose continuation depends upon continued public concern about electromagnetic fields. This is far from improper, and may merely indicate interest on the part of the scientists. However it does suggest a selection bias in the make up of the committee which should be taken into account.
If possibility 3 is true, I state unequivocally my belief that the working group would be wrong.

Conclusion

The working group saw a possible stegosaurus: not a possible lion or a possible dog. Under these circumstances an assignment of classification 2A would extend enormously the well established meaning of that classification and should not be accepted.

References

1. "Evaluation of the Potential Carcinogenicity of Electromagnetic Fields" Review draft EPA/600/6-90/005B (October 1990)

2. Letter to Mr William K. Reilly, Administrator, US Environmental Protection Agency, EPA SAB-RHC-92-013 (1991)

3. This is additional to the obvious point that common salt is not only sodium chloride but contains some potassium chloride. One of the isotopes of potassium, potassium 40 is radioactive and radiation is a definite human carcinogen. Therefore common salt is a carcinogen. This is an inescapable conclusion. However most people use common sense in such discussions and decline to take account of low doses and such small effects.

4. Although distinction is not in itself a criterion for competence or the holding of correct beliefs, it is worth noting that none of the members of the working group seem, for example to be members of the National Academy of Sciences.

5. The argument is simple. The noise in a system varies as the square root of the linear dimensions. The signal can vary linearly with the dimension. Thus a larger detector has a greater signal to noise ratio and therefore greater sensitivity.

6. It is noteworthy that the working group seem to be unaware of the seminal and well reviewed paper by R.K. Adair, "Constraints on the Biological Effects of Weak Extremely-Low Frequency Electromagnetic Fields." 43 Phys. Rev. A. 1039-1048 (1991). This discusses the theoretical constraints in some detail. The reference to the "popular" article by Bennett in "Physics Today" is less detailed and an inadequate substitute.

7. Cavaignac, J.F., E. Jeenicke, B. Vignon, and R. Wilson (1975) "Sensitivity of Organic Scintillators to Magnetic Fields" Nuclear Instruments and Methods v. 126, p 459

8. See for example section 4.8 of the NIEHS working group report

9. W.R Bennett, "Cancer and Power Lines" 47 Physics Today 23-29 (1994). See also the ORAU report at II-16 - II-20. Note also that fields at the floor level of Washington Metro trains (above the motor) can be this high according to measurements by the Department of Transportation (Cambridge Laboratory).

10. T. Tynes, et al., "Leukemia and Brain Tumors in Norwegian Railway Workers, A Nested Case Control Study", 137 Amer. J. Epidemiology 645-653 (1994).

11. I here note a puzzling conclusion by the working group in 5.2.2. That "there is strong evidence that electric fields can be perceived". As noted in the text this is generally accepted. But why talk about electric fields when the rest of the discussion is about magnetic fields? and again why no reference to field intensity? Is there supposed to be a deeper meaning?

12. R.D.Tucker and O.H. Schmitt (1978) "Tests for human perception of 60 Hz moderate strength magnetic fields" IEE Transactions on Biomedical engineering 25:509-518

13. C. Graham and H Cohen (1985) " Influence of 60 Hz fields on human behavior, physiology and biochemistry MRI report

14. Y. S. Loh , A. Shlyakhter and R. Wilson "Electromagnetic Fields and the Risk of Leukemia and Brain Cancer: a summary of epidemiological literature" Technology: Journal of the Franklin Institute 334A, 3-21 (1997)

15. D. Wartemberg, F. Dietrich, R. Goldberg, C Poole and D Savitz (1998) "A meta-anlysis of studies of childhood cancer and residential exposure to magnetic fields" Report to NIEHS

16. Linet. M. et al. (1997) "Residential Exposure to magnetic fields and Acute Lymphoblastic Leukemia" NEJM 337: 1-7

17. Hatch E.E. et al. "Association between Childhood Acute Lymphoblastic Leukemia and Use of Electric Appliances during pregnancy and Childhood" Epidemiology 9:234-245

18. Letter to R Wilson from Dr Martha Linet, August 5 1998

19. The famous physicist Richard Feynman explained a statistical "trap" into which many analysts fall. Striding into a lecture room he said that he had just seen a most unusual thing. A car with a license number BJY 453. The probability of seeing this was less than 1 in 18 million. Yet he saw it! The explanation in this case is obvious, but in epidemiology is less obvious. One must not ask for the probability of something for which one already knows the answer. It is likely that in several of the studies this selection bias was operating to some extent.

20. N. Wertheimer and E. Leeper, "Electric Wiring Codes and Childhood cancer" Amer Journ. Epidemiology 109:273 (1979).

21. The working group calls the Wertheimer and Leeper study a "seminal report". I have heard objections to this because it seems to some to be praising it. But the phrase is of course correct because it was the seed that started the present public concerns.

22. R. Wilson and A. Shlyakhter (1985) "Comments on a paper by Feychting and Ahlbom" Amer. Jour. Epidemiology141:378-379

23. Shlyakhter A.I., L. Mirny, A Vlasov and R. Wilson (1996) "Monte Carlo Modelling in Epidemiological Studies" Human and Ecological Risk Assessment 2(4):920-938

24. A.B. Hill, "The Environment and Diseases: Association and Causation", 58 Proc. Royal Soc. Med., Sec. Occup. Med. 295-300

(1965).

25. One such discussion is in: Richard Wilson and Martin Kaufman "Electromagnetic Fields and the Law" in Science and the Law Ed. R. Clegg, National Legal Center in the Public Interest, 8:23-87, Washington, D.C. (1997)

26. Kammen, D. M.; Shlyakhter, A. I.; Wilson, R. (1994) What is the Risk of the Impossible? Journal of the Franklin Institute v.331A, 97-116 (1994).

27. Feinstein, A. R. (1988) "Scientific standards in epidemiologic studies of the menace of daily life." Science 242: 1257-1263

28. C.G. Suits and H.E. Way, eds., The Collected Works of Irving Langmuir (1960).



29. The last is: National Research Council of the National Academy of Sciences, "Possible Health Effects of Exposure to Residential Electric and Magnetic Fields" (1996).

30. We note that the National Academy of Sciences was set up by Abraham Lincoln largely to advise the government in matters such as these

31. Oak Ridge Associated Universities Panel for the Committee on Interagency Radiation Research and Policy Coordination (CIRRPC) "Health effects of Low frequency Electric and Magnetic Fields" ORAU 92/F8 (1992)

32. Interagency Task Force Studying Electric and Magnetic Fields, "Connecticut 1994 Report on Task Force Activities to Evaluate Health Effects from Electric and Magnetic Fields" (1994).

33. "Electromagnetic Fields and the Risk of Cancer," Report of an Advisory Group on Non-ionizing Radiation, National Radiological Protection Board, No. 3 (1992).

34. "Biological Effects of 60-Hz Transmission Lines," A Report of the Florida Electric and Magnetic Fields Science Advisory Commission to the State of Florida Department of Environmental Regulation (1985).

35. There have been several WHO reports, including "Extremely Low Frequency (ELF) Fields (1984) and "Magnetic Fields" (1987)

36. Reference 1 supra

37. Committee of the Health Council of the Netherlands, "Extremely Low-Frequency Electromagnetic Fields and Health," (April 1992).

38. Expert Group on Non-ionizing Radiation, "Rapport on risiko for kraeft has born med boosel eksoowerat for 50 Hz manerfeiter fra hoispaendugsaenaeet," Report to the Danish Ministry of Health ( May 1993).

39. Report to Minister of Energy by Department of Energy of Ireland, "Electromagnetic Fields: a review of Recent Developments in Research and Public Attitudes and the Response of Authorities to these Developments," (July 1992).

40. P. Guenel and J. Lellouch, "Synthesis of the Literature on Health Effects from Very Low Frequency Electric and Magnetic Fields," National Institute of Health and Medical Research (INSERM) (France)(1993).

41. J.C. Roucayrol, "Sur les champs electromagnetiques de tres basse frequence et la sante [Extremely low frequency electromagnetic fields and health]," 177 Bull. Acad. Nat. Med. 1031 (1993).

42. P.A. Buffler, P.E. Burgess, G.L. Smith, R.A. Beauchamp, H.A. Higgins, S.H. Linder, M.E. McLain, P.L. Zweiacker, "Health Effects of Exposure to Powerline-Frequency Electric and Magnetic Fields," Public Utility Commission of Texas (1993)

43. "Possible Health Effects of Extremely Low Frequency Electric and Magnetic Field Exposure: A Review," Report to the Illinois State Legislature by the Illinois Department of Public Health in coordination with the Illinois Environmental Protection Agency (1992).

44. "Status Report on Potential Human Health Effects Associated With Power Frequency Electric and Magnetic Fields," prepared for the Maryland Department of Natural Resources and the Public Service Commission of Maryland, Document PPSE-T-39 (1994).

45. Quoted in U. S. Environmental Protection Agency, "Electric Power lines: Q & A on Research into Health Effects" (1992).

46. 4 American Physical Society News, No. 7 at 2, July 1995. See also Physics Today July 1995 at page 49.

47. For background information, see D. Hafemeister, "Background Paper on 'Power Line Fields and Public Health'," Report to the Panel on Public Affairs of the American Physical Society (1995), available from the American Physical Society or on the World Wide Web at the http://www.calpoly.edu/~dhafemei; see also D. Hafemeister, "Biological Effects of Low-Frequency Electromagnetic Fields," Resource Letter BEL FES-11, 64 Am. J. Physics 974 (1996).

48. Brief of amici curiae, Adair, Bloembergen, Bodansky, Cormack, Gilbert, Glashow, Hafemeister, Merritt, Moulder, Park, Pound, Seaborg, Yalow and Wilson, in M. Covalt et al. Vs San Diego Gas and Light before Supreme Court of California (1995)

49. Reference 20 supra.

50. E.P Washburn, et al., "Residential Proximity to Electricity Transmission and Distribution Equipment and risk of Childhood Leukemia, Childhood Lymphoma and Childhood Nervous System Tumors: systematic review , Evaluation, and Meta Analysis Cancer Causes and Control 5:299-309 (1994)

51. Shlyakhter, A.I. "An Improved Framework for Uncertainty Analysis: accounting for Unsuspected Errors" Risk Anlysis 14:441-447 (1994)

52. Reference 19 Supra.

53. J.L.Repace and A.H. Lowrey, (1993) "An Enforceable Indoor Air Quality Standard for Environmental Tobacco Smoke in the Workplace" Risk Analysis v 13(4); in addition many articles in Risk Analysis 15(1) discuss this topic.

<>54. A. Stewart and G.W.Kneale, (1970) "Radiation Dose Effects in Relation to Obstetric X Rays and Childhood Cancer" Lancet pp1185-1188

55. There are numerous reviews on this subject. We refer in particular to the reports (particularly that of 1993) of the United Nations Subcommittee on the Effects of Atomic Radiation (UNSCEAR) to the General Assembly. United Nations, NY, 1993

56. Reference 24 supra at 74.

57. M. Feychting and A. Ahlbom, "Magnetic Fields and Cancer in Persons living close to High Voltage Power lines in Sweden" Amer. J. Epidemiology 138:467-481 (1993)

58. D.A. Savitz and D. Loomis, "Leukemia and Brain Cancer in Electrical Workers" 141 Amer. J. Epidemiology 123-134 (1995).

59. Reference 14 Supra.

60. Of course as Langmuir pointed out one can always have another ad hoc explanation - such as a change in metabolism as one changes longtitude.

61. A. B. Hill, reference 24 supra.

62. Exact agreement about the proportions is not necessary here, because they may be modified by synergistic factors, and by the different age or gender distributions of the populations.

63. A.B. Hill, reference 24 supra.

64. A.B.Hill, reference 24 supra.

65. The most recent general discussion of the attempts to find models can be found in P.A. Valberg, P.A., "Biology and EMF: Biophysical Mechanisms of Interaction" Gradient Corporation report to Electric Power Research Institute (EPRI) (1994). and P.A Valberg, R. Kavet and C.N. Rafferty "Can low-level 50/60 Hz electric and magnetic fields cause biological effects?" Radiation Research 148 2-21 (1997). The seminal reference showing the constraints upon models is R.K. Adair, "Constraints on the Biological Effects of Weak Extremely-Low Frequency Electromagnetic Fields." 43 Phys. Rev. A. 1039-1048 (1991); see also R.K. Adair "Effects of ELF Magnetic Fields on Biological Magnetite," Bioelectromagnetics (1993); T.S. Tenforde, 13 Ann Revs. Publ. Health 173 - 196 (1993) points out that Adair has ignored the "signal amplification" that can be achieved by large arrays of electrically coupled cells in tissue". But any such array must be tightly coupled to avoid the fluctuation problems and be effectively a large single detector such as my proposed coil of wire around the stomach. It is up to physicians, cellular biologists and pathologists to determine whether such large coupled structures (1 cm or more in diameter) actually exist.

66. Men began to smoke before women did, and the lung cancer incidence increased among men before it increased among women.

67. The working group report lists many of these, and correctly concluded that there is inadequate evidence from experimental animals to support the evaluation.

68. Section 5.1.3

69. Section 4.8.3.2 page 374

70. A.B. Hill, reference 24 supra.

71. J.D. Jackson, "Are the Stray 60 Hz Electromagnetic Fields Associated with the Distribution and Uses of Electrical Power a Significant Cause of Cancer?" 89 Proc. Natl.Acad.Sci. 3508-3510 (1992). Jackson notes that per capita generation has increased a factor of 10 since 1940 and per capita residential consumption increased a factor of 20. Yet cancers have not increased. An examination of the Connecticut cancer registry for example, which has been operating throughout this period, shows no increase in childhood leukemia. Believers in a link might argue that the increase in usage has been accompanied by better household wiring, with better cancellation of magnetic fields, and also with improved cancer prevention thereby masking the effect. But it is hard to believe that the effect could be this big. See also C. Boring, et al.,"Cancer Statistics 1994, 44 Cancer Journal for Clinicians 7-26 (1994)

72. Block, S.M. (1992) "Biophysical Principles of Sensory Transduction" Society of General Physiologists Series 47, 1-17

73. W. Bialek, "Physical Limits to Sensation and Perception"" Ann. Revs. Biophys. Chem. 16:455-478 (1987)

74. I assume here that one rules out the idea that the body contains a large sensor such as a coil wrapped around the stomach.

75. There seems to be no precise definition of an epidemiologist. One court tried to exclude from consideration anyone who had not passed an epidemiology course. This excluded from consideration many of the leading epidemiologists who started the field. According to the more generous and loose definition I adopt here, I could be an epidemiologist although I do not so claim. I have been a coauthor in papers in epidemiological journals and I have been invited, as an epidemiologist to visit an overseas country (Inner Mongolia).