Archive-name: powerlines-cancer-FAQ/part4
Last-modified: 1994/7/4
Version: 2.6
Maintainer: jmoulder@its.mcw.edu

FAQs on Power-Frequency Fields and Cancer (Q&A, Part 3 of 3)

21) If exposure to power-frequency magnetic fields does not explain the
residential and occupations studies which show increased cancer incidence,
what other factors could?

There are basically four factors that can result in false associations in
epidemiological studies: inadequate dose assessment (Question 21A),
confounders (Question 21B), inappropriate controls (Question 21C), and
publication bias (Question 21D).

21A) Could problems with dose assessment affect the validity of the
epidemiological studies of power lines and cancer?

If power-frequency fields are associated with cancer, we do not know what
aspect of the field is involved.  At a minimum, risk could be related to
the peak field, the average field, or the rate of change of the field.  The
duration of exposure could also be a factor.  It has even been suggested
that harmonics and/or interactions with the earths static magnetic fields
are involved.  If we do not know who is really exposed, and who is not, we
will usually (but not always) underestimate the true risk [C14].

21B) Are there other cancer risk factors that could be causing a false
association between exposure to power-frequency fields and cancer?

Associations between things are not always evidence for causality.  Power
lines (or electrical occupations) might be associated with a cancer risk
other than magnetic fields.  If such an associated cancer risk were
identified it would be called a "confounder" of the epidemiological studies
of power lines and cancer.  An essential part of epidemiological studies is
to identify and eliminate possible confounders.  Many possible confounders
of the powerline studies have been suggested, including PCBs, herbicides,
ozone and nitrogen oxides, traffic density, and socioeconomic class.
 
- PCBs: Many transformers contain polychlorinated biphenyls (PCBs) and it
has been suggested that PCB contamination of power-line corridors might be
the cause of the excess cancer.  This is unlikely.  First, there is little
evidence for widespread PCB contamination of powerline corridors.  Second,
transformers are found along distribution lines, but not high-voltage
transmission lines, so PCBs could not account for the linkage of childhood
leukemia with transmission corridors [B5]. Three, the evidence that PCB
exposure causes or promotes cancer in people is weak [E9,L1].  Lastly,
PCBs predominantly cause and promote liver cancer in animals; leukemia,
brain and breast cancer have not been reported. 
 
- Herbicides: It has been suggested that herbicides sprayed on the
powerline corridors might be a cause of cancer.  This is also an unlikely
explanation.  First, herbicide spraying would not affect distribution
systems in urban areas (where 3 of 5 positive childhood cancer studies have
been done), and would not explain the reported increase in cancers in
electrical occupations. Second, evidence that herbicides are carcinogens in
humans is weak [L6].  Third, the epidemiology which suggests that the
phenoxy herbicides might be carcinogens indicate that the increased risk is
for lymphomas and soft-tissue sarcomas [L6]; only one study implicates
leukemia [D3], and none implicate brain cancer.

- Ozone and nitrogen oxides:  It has been suggested that ozone and nitrogen
oxides created when high voltage lines arc might be responsible for the
increased cancer along powerline corridors.  This is another unlikely
explanation.  First, while ozone is a cellular genotoxin, there is no
evidence that it causes cancer in humans, and only ambiguous evidence that
it causes lung cancer in rats [L5].  There is essentially no evidence that
the nitrogen oxides are carcinogens.  Second, this potential confounder
would apply only to corridors containing high-voltage lines and would not
explain reports of excess cancer along distribution systems or in
electrical occupations.     
 
- Traffic density: Transmission lines frequently run along major roads, and
the "high current configurations" associated with excess childhood leukemia
in the US studies [C1,C6,C10] are associated with major roads.  It has been
suggested that power lines might be a surrogate for exposure to
cancer-causing substances in traffic exhaust.  This may be a serious
confounder of the residential exposure studies, since traffic density has
been shown to correlate with childhood leukemia incidence [E5].  Note that
this would explain only the reported increase in cancer along power-lines;
it would explain the reports of increased cancer in electrical occupations.

- Socioeconomic class: Socioeconomic class may be an issue in both the
residential and occupational studies, as socioeconomic class is clearly
associated with cancer risk, and "exposed" and "unexposed" groups in many
studies are of different socioeconomic classes [C14].  This is of
particular concern in the US residential exposure studies that are based on
"wirecodes", since the types of wirecodes that are correlated with
childhood cancer are found predominantly in older, poorer neighborhoods,
and/or in neighborhoods with a high proportion of rental housing [C19].

- Other factors: If "other" factors exist that increase the incidence of
cancer they need to be controlled for in studies.  In other words, you have
to make sure that the "exposed" and "unexposed" groups have the same risk
factors.  Every time a new risk factor is discovered, previous studies need
to be reexamined.  Thus the "hot dog factor"!  The same investigators who
reported an increased incidence of childhood leukemia along powerline
corridors in Los Angeles [C10], have recently reported an much higher
leukemia risk factor for hot dog consumption (RR of 9.5) in the same group
of children [E10].   

21C) Could the epidemiological studies of power lines and cancer be biased
by the methods used to select control groups?

An inherent problem with many epidemiological studies is the difficulty of
obtaining a "control" group that is identical to the "exposed" group for
all characteristics related to the disease except the exposure.  This is
very difficult to do for diseases such as leukemia and brain cancer where
the risk factors are poorly known.  An additional complication is that
often people must consent to be included in the control arm of a study, and
participation in studies is known to depend on factors (such as
socioeconomic class, race and occupation) that are linked to differences in
cancer rates.  See Jones et al [C19] for an example of how selection bias
could effect a powerline study.

21D) Could analysis of the epidemiological studies of power lines and
cancer be skewed by publication bias?

It is known that positive studies in many fields are more likely to be
published than negative studies.  This can severely bias meta-analysis
studies such as those discussed in Questions 13 and 15.  Such publication
bias will increase apparent risks.  This is a bigger problem for the
occupational studies than the residential ones.  

Several specific examples of publication bias are known in the studies of
electrical occupations and cancer (see Doll et al [B4], page 94).  In their
review Coleman and Beral [B1] report the results of a Canadian study that
found a RR of 2.4 for leukemia in electrical workers.  The British NRPB
review [B4] found that further followup of the Canadian workers showed a
deficiency of leukemia (a RR of 0.6), but that this followup study has
never been published.  This is an anecdotal report, but publication bias,
by its very nature, is usually anecdotal.

It is also a clear problem for laboratory studies -- it is much easier to
publish studies that report effects than studies that report no effects. 
An example of this can be seen in work by Cain and colleagues.  In a 1993
they published a report [G26] that 60-Hz fields were a co-promoter in a
cell transformation system.  Also in 1993 the same authors reported at
meetings that they could not replicate the co-promotion, and that
subsequent experiments showed a decrease in transformation when 60-Hz
magnetic fields were present.  However, the later data is not published, so
only the positive report is currently in the peer-reviewed literature.

There is also the related issue of "reporting bias", which refers both to
situations where multiple studies are done but only some are reported, and
to situations where abstracts and/or press reports emphasize
unrepresentative subsets of the actual study.  The "Swedish" studies
[C13,C18] provide an case example.  The original report used a number of
different definitions of "exposure", and studied both children and adults. 
Of all the comparisons, the most significant correlations were found for
childhood leukemia and calculated fields.  The published Swedish version
[C13] omits details of some of the exposure definitions that showed no
relationships, and the published English-language version [C18] omits the
adult studies.  The abstract of the English-language version emphasizes the
groups, exposure definitions and cancer types for which there were
significant relationships.  The press reports were based largely on that
abstract.  The result is that a handful of significantly-positive
associations are picked for emphasis from a much larger group of
overwhelmingly non-significant associations.

22) What is the strongest evidence for a connection between power-frequency
fields and cancer?

The best evidence for a connection between cancer and power-frequency
fields is probably:
a) The four epidemiological studies that show a correlation between
childhood cancer and proximity to high-current wiring [C1,C6,C10,C18], plus
the meta-analysis of the Scandinavian studies [B5].
b) The epidemiological studies that show a significant correlation between
work in electrical occupations and cancer, particularly leukemia and brain
cancer [B2,B3,B4,D12,D14]. 
c) The lab studies that show that power-frequency fields do produce
bioeffects.  The most interesting of the lab studies are probably the ones
showing increased transcription of oncogenes at fields of 1-5 G (100-500
microT) [H4,H6,L2].
d) The two laboratory studies that provide evidence that power-frequency
magnetic fields can promote chemically-induced breast cancer [G14,G23].

23) What is the strongest evidence against a connection between
power-frequency fields and cancer?

The best evidence that there is not a connection between cancer and
power-frequency fields is probably:
a) Application of the Hill criteria (Question 20) to the entire body of
epidemiological and laboratory studies.
b) The fact that all studies of genotoxicity, and most study of promotion
have been negative (Question 16). 
c) Adairs [F2,F8] biophysical analyses that indicates that "any biological
effects of weak (less than 40 mG, 4 microT) ELF fields on the cellular
level must be found outside of the scope of conventional physics"
d) Jacksons [E8] and Olsens [C16] epidemiological analysis that shows
that childhood and adult leukemia rates have been stable over a period of
time when per capita power consumption has risen dramatically.  This
argument presumes that exposure has risen in parallel with consumption;
there is little relevant historical data, and there are technical reasons
to question the validity of this assumption.

24) What studies are needed to resolve the cancer-EMF issue?

In the epidemiological area, more of the same types of studies are unlikely
to resolve anything.  Studies showing a dose-response relationship between
measured fields and cancer incidence rates would clearly affect thinking,
as would studies identifying confounders in the residential and
occupational studies.  

In the laboratory area, more genotoxicity and promotion studies may not be
very useful.  Exceptions might be in the area of cell transformation, and
promotion of chemically-induced breast cancer. Long-term rodent exposure
studies (the standard test for carcinogenicity) would have a major impact
if they were positive, but if they were negative it would not change very
many minds.  Further studies of some of the known bioeffects would be
useful, but only if they identified mechanisms or if they established the
conditions under which the effects occur (e.g., thresholds, dose-response
relationships, frequency-dependence, optimal wave-forms).

25) Is there any evidence that power-frequency fields cause any human
health hazards?

While this FAQ sheet, and most public concern, has centered around cancer,
there have also been suggestions that there might be a connection between
non-ionizing EM exposure and birth defects.  This concern has focused as
much on video display terminals (VDTs) as on power lines.  Little
epidemiological or laboratory support for a connection between non-ionizing
EM exposure and birth defects has been found. [J1,J2,J4,J5,J6].  Cox et al
[J3] and Chernoff et al [K5] have recently reviewed this field.

26) What are some good overview articles?

There are really no up-to-date reviews of power-frequency fields and human
health.  The reviews by Davis et al [A2] and Doll et al [B4] are good, but
were published before many of the important epidemiological, genotoxicity,
and promotion studies were available.

27) Are there exposure guidelines for power-frequency fields?

Yes, a number of governmental and professional organizations have developed
exposure guidelines.  These guidelines are based on keeping the body
currents induced by power-frequency EM fields to a level below the
naturally-occurring fields (Question 8).  The most generally relevant are:

- National Radiation Protection Board (UK) [M5]:
  50 Hz: 1.6 mT (16 G) and 12 kV/m 
  60 Hz: 1.33 mT (13.3 G) and 10 kV/m 
  This document also contains guidelines for other ELF frequencies.

- American Conference of Governmental Industrial Hygienists [M6]:
  At 60 Hz: 1 mT (10 G); 0.1 mT (1 G) for pacemaker wearers
  This document also contains guidelines for other ELF frequencies

- International Commission on Non-Ionizing Radiation Protection [M7]
  24 hr general public: 0.1 mT (1 G) and 5 kV/m     
  Short-term general public: 1 mT (10 G) and 10 kV/m     
  Continuous occupational: 0.5 mT (5 G) and 10 kV/m     
  Short-term occupational: 5 mT (50 G) and 30 kV/m      

28) What effect do powerlines have on property values?

There is very little hard data on this issue.  There is anecdotal evidence
and on-going litigation (Wall Street Journal, Dec 9, 1993).  There have
been "comparable property" studies, but any studies done prior to about
1991 (when London et al [C10] was published) might be irrelevant.  One
comparable value study has been published recently [L4], and another has
been presented at a meeting [L7]. Neither study shows hard evidence for an
impact of power lines on property values.  However, both studies indicate
that many owners think that there will be an impact, particularly if
concerns about health effects become widespread.

It appears possible that the presence of obvious transmission lines or
substations will adversely affect property values if there has been recent
local publicity about health or property value concerns.  It would appear
less likely that the presence of "high current configuration" distribution
lines of the type correlated with childhood cancer in the US studies
[C1,C6,C10] would affect property values, since few people would recognize
their existence. If buyers start requesting magnetic field measurements, no
telling what will happen, particularly since measurements are difficult to
do (Questions 29 & 30), and even more difficult to interpret (Question 14).

29) What equipment do you need to measure power-frequency magnetic fields?

Power-frequency fields are measured with a calibrated gauss meter.  The
meters used by environmental health professionals are too expensive for
"home" use.

A unit suitable for home use should meet the following criteria.  First, it
must have a reasonable degree of accuracy and precision (plus/minus 20%
seems reasonable for home use).  Second, it should have true RMS detection,
otherwise readings might be exaggerated if the waveform is non-sinusoidal.
Third, it should have a tailored frequency response, because if the unit is
too broad-band, higher frequency fields from VDTs, TVs, etc. may confound
the measurements.  Fourth, it should have the correct response to overload;
if the unit is subjected to a very strong field, it should peg, not just
give random readings.  Fifth, the presence of a strong electrical field
should not affect the magnetic field measurement.
Meters meeting these requirements are quite expensive, $600 would probably
be the bare minimum.  These meters are not suitable for the non-technically
trained.

There is an understandable reluctance to recommend any unit with unknown
characteristics to a person whose technical abilities are also unknown, and
no peer-reviewed articles on inexpensive instruments appear to be
available.  A recent issue of Consumers report found that the inexpensive
(under $200) meters available in the USA were unreliable.  The suggestions
that one wind a coil and use headphones or a high impedance multimeter are
misguided.  A clever physicist or engineer can anticipate and correct for
nonlinearities and interferences, but for the average person, even one
technically trained, this is unreasonable.

30) How are power-frequency magnetic fields measured?

Measurements must be done with a calibrated gauss meter (Question 29) in
multiple locations over a substantial period of time, because there are
large variations in fields over space and time.  Fortunately, the magnetic
field is far easier to measure than the electrical field. This is because
the presence of conductive objects (including the measurer's body) distorts
the electrical field and makes meaningful measurements difficult. Not so
for the magnetic field.

It is important for the person who is making the evaluation to understand
the difference between an emission and exposure. This may seem obvious, but
many people, including some very smart physical scientists, stick an
instrument right up to the source and compare that number with an exposure
standard.  Also, if the instrument is not isotropic, measurement technique
must compensate for this.

In the case of power distribution line and transformer fields, the magnetic
fields will probably vary considerably over time, as they are proportional
to the current in the system.  A reasonable characterization needs to be
done over time, with anticipated and actual electricity usage factored in.
It may seem to be as simple as walking in and reading the meter, but it's
not.

31) Do the issues discussed in this FAQ sheet apply to EM fields other than
power-frequency fields?

This FAQ sheet concerns itself primarily with sinusoidal fields at
frequencies of 50 or 60 Hz.  However, the basic principles and data
discussed in the FAQ sheet are generally applicable to ELF sources with
frequencies between 1 Hz and 30,000 Hz (30 kHz).  Below 1 Hz, one should
also consider issues associated with static magnetic fields [M8].  Above 30
kHz, one is moving into the radiofrequency (RF) range, and other
biophysical and biological issues arise that are not within the scope of
this document [M2,M4].  

The major issue encountered when dealing with other ELF EM sources is that
the currents induced by ELF magnetic fields depend on the frequency and the
wave-form as well as intensity.  As the frequency increases, so do the
induced currents.  Thus safety guidelines that are based on induced
currents change with frequency [M5,M6].  For example, the NRPB magnetic
field exposure guideline [M5] which is 1.33 mT (13.3 G) at 60 Hz, rises to
80 mT (800 G) at 1 Hz and falls to 80 microT (800 mG) at 3 kHz.  

Estimating the currents induced by non-sinusoidal ELF wave forms is more
complex, because the magnitude of the induced current depends on the rate
at which the magnetic field changes.  Thus a square wave of the same
frequency and amplitude of a sinusoidal wave will induced a much greater
current.

Copyright (C) by John Moulder
end: powerlines-cancer-FAQ/part4
