Full text of DOI:10.1002/(SICI)1521-186X(200001)21:1<2::AID-BEM2>3.0.CO;2-0

Bioelectromagnetics 21:2^7 (2000)
d'Arsonval Address: Adequacy of Wire Codes
and Other Proxies Used to Assess Historic
Magnetic Field Exposure
Nancy Wertheimer*
I want to thank you all for this heartwarming and downstream pole was placed, as was then the typical
unexpected honor. It is an honor that really belongs, of practice, just two lot lines away from the transformer
course, to a long list of researchers, since our own poleÐa fact that corresponded rather startlingly to my
work, Ed Leeper's and mine, took root at all only earlier observation of excess cancer risk for homes on
because the ground had already been turned, and its the two lots nearest the transformer on either side. The
value lies primarily in the wide range of research it has pattern was not always so clear in every alley, but it
stimulated.
Often I'm asked why we chose to study power
de®nitely held, on average.
It may be important that what I saw in those
lines and cancer in the ®rst place: The fact is, of course, alleys were largely ®elds from net currents in the
that we didn't. What I originally set out to do was just a secondary linesÐcurrents that were produced only
bit of ``shoeleather'' epidemiologyÐvisiting birth because old-Denver had an all-metal, conductive water
homes of young leukemia cases and controls to see system at the time, which allows some neutral current
what clues to etiology might appear. I wanted to see if to be diverted through houses into the iron water
cases had lived close to each other, which might mainsÐa diversion that cannot occur with more recent
suggest an infectious etiology; if they had lived near Denver plumbing practices. Those neutral currents can
polluting factories or on busy streets; and so on.
also produce ``hot spots'' of very high exposure inside
About halfway through my sample I began to feel a house as they ¯ow through itÐa fact I will return to
that I was seeing too many pole-mounted transformers later.
in the backyards of my leukemia cases. Checking back,
Thus our interest in a link between environmental
and then moving forward through the rest of my magnetic ®elds and cancer arose. That hypothesis led
sample, I found that, indeed, not only the homes to Ed's construction of a state-of-the-art gaussmeter, to
nearest the transformer, but homes on the two lots two years of taking hundreds of measurements, both
nearest the transformer on either side, had especially inside and outside representative homesÐand to
high cancer risk.
scores of consultations with power company engi-
At that point I went to my physicist friend Ed neers, including old timers. Those measurements and
Leeper, saying, ``Look, I have this really crazy ®nding! consultations guided construction and testing of a
What could transformers possibly do to affect a kid's ``wire code'' aimed speci®cally at identifying the
cancer risk?'' Well, he couldn't tell me, and neither homes likely to have unusually high ®elds in our old
could anyone else. Ed did tell me that transformers Denver neighborhoods.
would produce 60 Hz magnetic ®eldsÐbut he assured
Using that code on a new, enlarged sample, we
me that those ®elds would drop off so quickly that they found a two- to three-fold increase of cancer risk
would probably not be discernable even at the base of among the children living at high-code homes in
the pole, much less at the house. Nonetheless, he put Denver. Analyses for confounding by urbanicity,
together a very crude gaussmeter made from a socioeconomic level, mobility, traf®c density, and
degaussing coil and the audio circuit of an old such, did not change the ®nding. so ®nally we
walkie-talkie, and I took this little meter out to explore. published our results. We knew our hypothesis would
I began at the base of a transformer pole and be met with skepticismÐas indeed it should have
started walking down the alley, under the secondary been. We were ourselves still skeptical, and spent the
line fed by the transformer. At ®rst I saw no change at ensuing years refusing interviews with the National
all in the ®eld as I walked away from the transformer.
But then suddenly the ®eld diminished sharply as I
passed the ®rst downstream pole, where four service
drops took current off the secondary line. And that ®rst
ÐÐÐÐÐÐ
*Correspondence to: Nancy Wertheimer, 1330 5th Street, Boulder,
CO 80302.
ß 2000 Wiley-Liss,Inc.

Proxies to Assess Historic MF Exposure
3
Enquirer on the one hand, and trying to encourage And there is good reason to question how well
adequate replication studies on the other. And cer- present-day measurements can re¯ect such historic
tainly nothing would have come out of our work had exposure.
not a few brave souls in the scienti®c community
®nally taken on the dif®cult and essential task of of day, day of the week, and season of the year; with
replication. plumbing changes in the street or changes in nearby
Fields change over time: They change with time
When replications came, they fostered some business uses; or with placing an electric toothbrush
alarming press reports. While those reports did help to holder on the other side of the wall from a child's bed;
stimulate funding for further research, they also, and so on.
unfortunately, produced a premature, politically-driven
In particular, worn dosimetry has special pro-
demand for immediate answers, in a ®eld where we blems, along with its advantages. It records exposures
don't yet know even how to ask some of the essential that depend on exactly where a child spends his time,
questions, much less how to answer them. Today I which can vary over the long term or from day to day at
would like to explore one of those questions: That is, least as much as can his exposure sources. If, on
just what exposures should we be measuring and how measurement days, he spends about the same times in
should we measure them?
the same places (TV chair, bed, schoolroom) as in his
After Savitz et al. reported a replication of our prediagnosis years, his present-day measurements may
wire code results in 1988, the idea that residential be a good proxy for exposure in his critical historic
magnetic ®elds might affect childhood cancer gained period. But what if he is now in a different schoolroom,
momentum until at least the mid 1990s, when the or his TV chair has been moved directly over a
positive Scandinavian studies appeared. Recently, plumbing current ``hot spot?'' Of further concern is
however, two large multi-area studiesÐthe National that, to some degree, any AC ®eld meter will register
Cancer Institute study [Linet et al., 1997] and the ®ve- motion in the earth's ®eld as increased AC ®eld
province Canadian study [McBride et al., 1999]Ðhave exposure. If controls are suf®ciently more active than
reported results that are largely null, except for a few convalescing cases, they may be spuriously under-
modestly elevated odds ratios at the high end of represented in the lowest-®eld (referent) group, biasing
exposure. What changed?
There are many, I think, who believe that what
higher-®eld odds ratios downward.
Thus wire codes, when well designed, may still
changed is simply that, over time, the studies got better be the most reliable way we have yet found to assess
and better; and that therefore the risk ratios obtained historic exposure. They are historically stable, and they
got smaller and smaller because in fact there is no do not require subject cooperation. In fact, much of the
magnetic ®eld effect on cancer. I don't hold that view. evidence for risk thus far has come from studies that
On the contrary, I believe that the increasingly used a well-tailored wire code to assess historic
null results may be largely due to changes made, with exposure. By ``a well tailored code,'' I mean using
the best intentions, in the way exposure has been pertinent information about historic engineering prac-
assessed; I think that the measurement techniques used tices to develop a set of observations (a ``code'') that
recently have often estimated a subject's historic effectively identi®es which homes were likely to have
exposure less well than did the assessments that had high ®elds historicallyÐand then using measure-
produced the earlier, more positive ®ndings. ments at representative homes in the study area to
There is an adage attributed to Eric Sevareid that validate the code.
``the cause of most problems is solutions''; and I think
Fortunately, ours is not the only code that has
that may apply here. In trying to improve on the rough been tailored to ®t the homes it was used on. One of the
exposure categories provided by the earlier wire codes, best-tailored codes to date is the ``wire code,'' if I may
the new studies focused their efforts on present-day call it that, used by Feychting and Ahlbom in their
measurements. But the very speci®city with which 1992 Swedish studyÐthat is, the historic ®eld
modern instruments re¯ect the exposure on a particular calculations they made, which were grounded in good
and present day can make them ill-suited for estimat- historic information about current loads and wire
ing the typical and historic exposure that each study con®gurations. Contemporaneous measurements con-
really must assess.
®rmed that their calculations did an excellent job of
It is important to remember here that what indicating high ®eld homes for the single-family
are often called ``actual measurements'' are not, and residences, where in fact they saw evidence of risk.
never have been, actual measurements of the his- However, the calculations were not tailored to re¯ect
toric exposure of interestÐexposures which occur- ®elds from the ground current loops that commonly
red over some uncertain time period in the past. occur in the internal wires and plumbing of apart-

4
Wertheimer
ments. Therefore, the calculations did only a mediocre
In a well-tailored code, high-®eld homes should
job of indicating high ®eld homes for the apartment be both frequent in the high code group and rare in the
dwellings, where no evidence of risk was seen. low code group, thus producing a high ratio between
Even the positive results from the Savitz study the high-®eld-home percentages in the two groups.
were obtained using a code adjusted to ®t the area and Small ratios such as those seen in Table I for the non-
time period studied: Savitz noted that homes with Denver United States data, and for the apartment
underground wiring, which were common in his dwellings in Sweden, may signal use of a poorly
sample but rare in our older study, characteristically tailored code.
had low ®elds. Accordingly, he introduced a new
Looking next, in Figure 1, at the results from
``underground'' wire code category to re¯ect that. twelve different data sets, we see that use of a wire
Thus he appropriately applied our method, but not our code to assess historic exposure has produced positive
original code, to his sample. Had the code been further evidence of risk in almost every study where the code
adjusted for certain other recent technology changes, was speci®cally tailored to ®t the actual population
the Savitz odds ratio would have increased from 1.5 to studied. On the other hand, use of a code not
1.8 [Leeper et al., 1991]. speci®cally tailored for its study sample has produced
Eventually it became clear that the reason essentially null results, on average.
Savitz's buried-wire homes had low ®elds was that
So if there is in fact a real cancer risk associated
most houses, new enough to have underground wiring, with environmental magnetic ®elds, what have we
also had the non-conductive plumbing elements now learned in the last 20 years about assessing the
required in Denver. They were thus unlikely (in exposure responsible for that risk?
Denver, but not necessarily elsewhere) to have ``hot
Neither for well-tailored wire codes nor for
spots'' of exposure from ground currents ¯owing in present-day measurements have we seen much evi-
their water lines [Wertheimer et al., 1995]. Some such dence for a risk that is dose-related over the full
distinction between homes that do and do not have intensity range. For both, the most consistent evidence
potential plumbing-®eld exposure is probably essential of risk has come from the high-end distinctions;
if a code is to adequately assess North American particularly for homes with average exposures over
homes occupied by children in recent years.
3 mG. At that level the evidence of risk has been
TABLE 1. Tailored and Untailored Applications of Two Codes From Wertheimer [1998]
A. Percent with ꢀ 0.2 mT measurements in each Wertheimer±Leeper code category
Ratios
Standardized Ny
VHCC OHCC OLCC VLCC
264
UG
229
Total
HCC
Total HCC/ VHCC/
LCC
85
193
220
LCC
LCC
Tailored
Applications
Wertheimer±Leeper [1979]
Denver
Savitz [1988] Denver
67.9
60.0
30.6
21.0
4.8
6.0
0.0
0.0
NA
3.0
42.0
32.9
2.2
2.8
19.2
11.7
30.9Ã
21.3Ã
Untailored
Applications
Preston Martin [1996]
Los Angeles
Severson [1988] (Seattle)
35.0
57.1
14.1
9.1
20.0
9.4
13.9
14.5
12.0
2.3
6.3
7.1
0.0
2.7
20.6
23.8
26.1
9.6
5.2
7.7
2.1
4.6
3.4
3.6
11.1
5.2
Linet [1997] Midwest states 40.0
B. Percent with ꢀ 0.2 mT measurements in homes calculated as ꢀ 0.2 mT and < 0.2 mTz
(a) In homes with calculated
(b) In homes with calculated
®elds < 0.2 mT
®elds ꢀ 0.2 mT
90.7
92.8
Ratio a/b
Tailored application
Untailored application
One-family homes
Apartments
2.4
15.6
37.8Ã
5.9
ÃThese data sets showed signi®cant odds ratios at high code homes.
yIn calculating percentages for combined wire code categories (e.g., in weighting ``Total HCCs''), it was necessary to adopt a standardized
distribution for the numbers in each code category, in order to prevent confusing differences due to wire code distribution in different
regions with differences in wire code adequacy in different areas. The standardized numbers used are from Zaffanella's ``1000 house
study'' [Zaffanella, 1993].
zData from Feychting and Ahlbom, 1992, Table 3.8 and 3.10.

Proxies to Assess Historic MF Exposure
5
Fig. 1. Oddsratiosfor thehighestexposure categorypresentedindatafromtwelvestudiesof child-
hoodcancer, dividedintogroupswithtailoredanduntailoredwire codesor fieldcalculations.
Tailored Codes
1. Wertheimerand Leeper [1979],Table 3
2. Savitzand Kaune [1993],Table 4
Untailored Codes
8. Fulton et al. [1980],Table1
9. Thomaset al. [1998] (Wertheimer^Leepercode),Table1
10. Feychtingand Ahlbom [1992] (Appartments),Table 4.21
3. Thomaset al. [1998], (Predictedfields),Table1
4. FeychtingandAhlbom[1992](Single-familyhomes),Table4.20 11. Preston-Martinet al. [1996],Table 6
5. Olsenet al. [1993],Table IV
6. Verkasaloet al. [1993],Table II
7. Tynes and Haldorsen [1997],Table 5
12. Gurneyet al. [1996],Table 3
13. Linet et al. [1997],Table 3
14. McBride et al. [1999],Table 5
remarkably consistent, but the numbers are very whether a potential existsÐand existed historicallyÐ
small.
for having ``hot spots'' of still-higher ®elds elsewhere
There is, however, at least one case where in the house.
somewhat lower and more commonly seen measure-
For instance, a 1.0 mG (0.1mT) ®eld that is essen-
ments have also been a good indicator of risk. In two tially vertical usually signals that the main source of
data sets thus far (data from the 1988 Savitz study and ®eld is a relatively distant power line, or a large current
the 1988 Severson adult-leukemia study), signi®cant loop around the house. And those sources will produce
risk estimates up to four-fold have been seen for a fairly uniform 1.0 mG ®eld throughout the house
analyses that considered the average of both ®eld angle (except near appliances of course).
and ®eld intensity simultaneously [Wertheimer et al.,
1995].
But if instead the 1.0 mG ®eld has a non-vertical
angle (less than about 55 ꢁ from the horizontal), this
Why is the angle important? Probably because, in usually signals that the main source of our ®eld is quite
this country at least, the angle of a non-negligible ®eld, nearbyÐoften net current in plumbing under the ¯oor
measured at waist height, can tell us a lot about of the same room, that may be producing ®elds of

6
Wertheimer
5 mG or more in its immediate vicinity. If the resident
happens to sit, sleep or play near such a ground current
``hot spot'', he can be exposed to 5 mG for long periods
even though the average ®eld in his house may be only
1.0 mG.
laboratory work suggests. If so, then only the
part of a study population having those cofactors
will be susceptible to magnetic ®eld exposure at
any given timeÐstill further nulling evidence of
any real risk.
Remember, such net currents in the grounding
system were responsible for our very ®rst observations
Thus, in interpreting epidemiologic studies, we
in the old-Denver area; and they deserve more must stay aware that, while associations based on some
attention than they have been given. systematic bias are always possible, and must be
I believe that in fact an average intensity guarded against, bias towards a null result is always
measurement over 3 mG in one instance, and a non- likely in all studies. So if the modestly-elevated risk
vertical ®eld over 0.5 mG in another instance, may ratios that have been repeatedly reported cannot be
both be imperfect indicators of the same risk factor. attributed to systematic biasÐand so far they have not
Either could indicate, for instance, a risk that is beenÐthey are likely to be underestimates of a
incurred only with repeated and extended exposure to substantial real risk.
®elds above some quite high intensity threshold.
This whole question matters a great deal because
homes that have ``hot spots'' from plumbing currents
REFERENCES
Feychting M, Ahlbom A. 1993. Magnetic ®elds and cancer in
children residing near Swedish high-voltage power lines.
Amer J Epidemiol 138:467±481.
Feychting M, Ahlbom A. 1992. Magnetic ®elds and cancer in
people residing near Swedish high voltage power lines.
Institutet for Miljomedicin (IMM) Report 6/92, Stockholm,
Sweden.
Fulton JP, Cobb S, Preble L, Leone L, Foreman E. 1980. Electrical
wiring con®gurations and childhood leukemia in Rhode
Island. Amer J Epidemiol 111:292±296.
Gurney, JG, Mueller BA, Davis S, Schwartz SM, Stevens RF,
Kopecky KS. 1996. Childhood brain tumor occurrence in
relation to residential power line con®gurations, electrical
heating sources, and electric appliance use. Amer J
Epidemiol 143:120±128.
are much more common than homes with an average
®eld over 3 mG. Consequently those two types of
measurement imply very different estimates of com-
munity risk.
In the Savitz data, for instance, an average ®eld
over 3 mG shows about a doubled risk and occurs in
less than 3% of the control homes. On the other hand,
non-vertical ®elds over 0.5 mG show a signi®cant four-
fold risk estimate and occur in one-®fth of the homes.
Those last ®gures, obviously, suggest a much more
widespread and serious community risk from environ-
mental magnetic ®elds than do the former.
To sum up: In studying environmental magnetic
®eld exposures, the exposure assessment problems are
formidable. Too often we forget this, and proceed to
think of whatever metric a study chooses as if that
metric were indeed an accurate re¯ection of the true
historic exposure of interest. That confusion of intent
with fact has led some to conclude that risks from
magnetic ®eld exposure, ``if real, are very small.'' But
this should be understood as a leap of faith.
Kaune WL, Zaffanella LE. 1994. Assessing historical exposures of
children to power±frequency magnetic ®elds. J Exposure
Analysis Environ Epidemiol 4:149±169.
Leeper E, Wertheimer N, Savitz D, Barnes F, Wachtel H.
1991. Modi®cation of the 1979 ``Denver wire code'' for
different wire or plumbing types. Bioelectromagnetics 12:
315±318.
Linet MS, Hatch EE, Kleinerman RA, Leslie L, Robeson LL,
Kaune WT, Friedman DR, Severson RK, Haines CM,
Hartsock CT, Shelley N, Wacholder S, Tarone RE.
1997. Residential exposure to magnetic ®elds and acute
lymphoblastic leukemia in children. New England J Med
337:1±7.
We need to remember:
*
*
First, that we don't know exactly what to
measure;
Â
McBride ML, Gallagher RP, Therialt G, Armstrong BG, Tamaro S,
Spinelli JJ, Deadman JE, Fincham S, Robson D, Choi W.
1999. Power±frequency electric and magnetic ®elds and
risk of childhood leukemia in Canada. Amer J Epidemiol
149:831±842.
Second, that even at best our exposure assess-
ment probably includes considerable random
misassignment of exposure, which will bias our
estimates of any true risk that exists towards the
null;
Olsen JH, Nielsen A, Schulgen G. 1993. Residence near high
voltage facilities and risk of cancer in children. Brit Med J
307:891±895.
Preston-Martin S, Navidi W, Thomas D, Pey-Jivan L, Bowman J,
Pagoda J. 1996. Los Angeles study of residential magnetic
®elds and childhood brain tumors. Am J Epidemiol 143:
105±119.
Savitz DA, Wachtel H, Barnes FA, John EM, Tvrdik JG. 1988.
Case±control study of childhood cancer and exposure to
60-Hz magnetic ®elds. Am J Epidemiol 128:21±38.
*
*
Third, that our hypothetically ``unexposed''
referent group is always exposed to a number
of unexamined sources of magnetic ®eld, which
will further null the evidence of real risks; and
Fourth, that magnetic ®elds may act only in
conjunction with speci®c cofactors, as some

Proxies to Assess Historic MF Exposure
7
Savitz DA, Kaune WT, 1993. Childhood cancer in relation to a
modi®ed residential wire code. Environ Health Perspectives
101:76±80.
Severson RK, Stevens RG, Kaune WT, Thomas DB, Heuser L,
Davis S, Sever LE. 1988. Acute nonlymphocytic leukemia
Finnish children living close to power lines. Br Med J
307:895±899.
Wertheimer N, Leeper E. 1979. Electrical wiring con®gura-
tions and childhood cancer. Amer J Epidemiol 109:
273±283.
and residential exposure to power frequency magnetic ®elds. Wertheimer N, Savitz DA, Leeper E. 1995. Childhood cancer in
Am J Epidemiol 128:10±20.
Thomas DC, Bowman JD, Liangzhong J, Feng J, Peters JM. 1998.
relation to indicators of magnetic ®elds from ground current
sources. Bioelectromagnetics 16:86±96.
Residential magnetic ®elds predicted from wiring con®g- Wertheimer N. 1998. Childhood cancer Appendix one, NIEHS
urations: II. Relationships to childhood leukemia. Bioelec- EMF Science Symposium: Breakout group reports for
tromagnetics 20:414±422. epidemiological research ®ndings, ppl5±22.
Tynes T, Haldorsen T. 1997. Electromagnetic ®elds and cancer in Zaffanella LE. 1993. Survey of residential magnetic ®eld sources.
children residing near Norwegian high-voltage power lines.
Am J Epidemiol 145:219±226.
Volume 1: Goals, results, and conclusions. Report TR-
102759-VI, Electrical Power Research Institute, Palo Alto,
CA 94304.
È
Verkasalo PK, Pukkala E, Hongisto MY, Valjus JE, Jarvinen PJ,
Heikkila KV, Koskenvuo M. 1993. Risk of cancer in
È