ELECTRICAL WIRING CONFIGURATIONS AND CHILDHOOD CANCER
NANCY WERTHEIMER1 AND ED LEEPER
Wertheimer, N. (Dept. of Preventive Medicine, U. of Colorado MedicalCenter, Box C-24S, Denver, CO 80262), and E. Leeper. Electrical wiring config-urations and childhood cancer. Am J Epidemiol 109:273-284, 1979.
An excess of electrical wiring configurations suggestive of high current-flowwas noted In Colorado In 1976-1977 near the homes of children who de-veloped cancer, as compared to the homes of control children. The findingwas strongest for children who had spent their entire lives at the same ad-dress, and It appeared to be dose-related. It did not seem to be an artifact ofneighborhood, street congestion, social class, or family structure. The reasonfor the correlation is uncertain; possible effects of current in the water pipesor of AC magnetic fields are suggested.
Electrical power came into use manyyears before environmental impactstudies were common, and today ourdomestic power lines are taken forgranted and generally assumed to beharmless. However, this assumption hasnever been adequately tested. Low levelharmful effects could be missed, yet theymight be important for the population asa whole, since electric lines are soubiquitous. In 1976-1977 we did a fieldstudy in the greater Denver area whichsuggested that, in fact, the homes of chil-dren who developed cancer were foundunduly often near electric lines carryinghigh currents.
In our modern power delivery systems,high-tension wires carrying current at
Received for publication May 11, 1978.Abbreviations: AC, alternating current; HCC,
1 Department of Preventive Medicine and Com-munity Health, Box C-245, University of ColoradoMedical Center, 4200 East Ninth Avenue, Denver,CO 80262. (Reprint requests to Dr. Wertheimer.)
The authors thank the Colorado Department ofVital Statistics and Dr. John Cobb of the Universityof Colorado Preventive Medicine Department fortheir facilitation of this research.
voltages up to several hundred kilovolts(kv) deliver power to distribution substa-tions where the voltage is stepped down,resulting in proportionately higher cur-rent in the medium-voltage (usually 13kv, wire-to-wire) primary lines. These lat-ter radiate out from the substation to dis-tribute power through a neighborhood.Then, at the local transformer, the volt-age of the primaries is stepped down oncemore to produce the 240 volt currentwhich is carried along the secondarywires to service drops which bring thepower to the customer's house. The cur-rent flow will always be greatest in thewires directly issuing from the substationor the transformer. At these points thevoltage has been stepped down and"transformed" into current. And it wasparticularly homes close to these trans-forming points that were over-repre-sented among our cancer cases.
Because our findings appeared to relateto high current rather than voltage, welooked into the magnetic fields induced bycurrent flow. Magnetic fields penetratethe human body (and buildings, etc.)readily. They are not easily shielded, but
they can be cancelled by balancing thecurrents that produce them. Such cancel-lation occurs in electric wires, where thereturn current tends to balance the sup-ply current. However, the cancellation isimperfect because the wires are oftenseparated in space and, more importantly,because some of the return current doesnot flow through the wires at all, but re-turns instead through the ground, andparticularly through the plumbing sys-tem to which most urban electrical sys-tems are grounded at each house.
This results in a locally imbalancedcurrent, both in the distribution wiresand in the plumbing. That imbalancedcurrent produces a 60 hertz (Hz) magneticfield which, though small (table 1), isnonetheless orders of magnitude larg-er than the 60 Hz field found in nature(about 10~8 gauss (1)). The ground-currentflows not only in the street plumbing, butalso through the pipes in the house. Cur-rent which enters the plumbing at onehouse can flow through several homes be-fore it returns to the distribution wires,because the plumbing provides a continu-ous, low-resistance path between houses.
The ground-current produces a magnet-ic field within the house (localized nearthe plumbing) which appears to be relatedroughly to the types of wiring configura-tions nearby (see table 1). This relation-ship between wires and plumbing is to beexpected because, other things beingequal, the greatest unbalanced currenttends to occur where the total current inthe wires is greatest, and the unbalancedportion of the current must detourthrough ground paths such as the nearbyearth and plumbing.
A number of household appliances andpower tools also produce magnetic fields,but in comparing the fields from ap-pliances with those from power lines, itis important to note that most appliancespresent approximately a magnetic dipolesource, with fields falling off roughly asthe inverse cube of the distance, while a
wire with unbalanced current will have a ent wiring configurations (nearness andfield falling off only as the inverse of the size of wires, closeness to origin of cur-distance. For instance: rent, etc.).
Electrical drillElectrical range (4 burners
on high)Wire carrying 15 amperes
In the literature there are listings of60 Hz magnetic fields produced byappliances which appear quite high.These should not be misinterpreted: Theyare apparently due to the use of mea-surements taken "as close as possible" tothe appliance. Our measurements indi-cate that magnetic-field exposure to thewhole body from normal use of householdappliances rarely exceeds .001 to .002gauss for any extended period, while theambient fields in a house due to nearbydistribution wires or plumbing may some-times reach those levels, or more, forhours or days at a time. If magnetic-fieldexposure is responsible for our finding, itmay be that, above some minimumthreshold, duration of continuous expo-sure is more important than strength ofexposure per se. There is some precedentfor such a threshold effect in the litera-ture on direct current (DC) magneticfields (2).
Our field measurements showed that,on the average, those types of wires as-sociated with cancer in our study exhib-ited high magnetic fields (compare ta-bles 1 and 3). However, the readings var-ied considerably over time; and becauseour observations were all made in goodweather and during work-day hours whendomestic current is minimal, becausecurrent-flow had most probably alteredsince the time of our subjects' residency,and because it was rarely feasible togo close to the house to take a measure-ment, no attempt was made to take sys-tematic measurements at our studyhomes. Rather this study is based on thepotential current flow suggested by differ-
Experimental work on physiologic ef-fects of low-level, extremely low fre-quency magnetic fields is limited. It hasbeen recently reviewed (3). Among thepositive reports are decreased mitosis inslime molds (4), decreased growth ofseedlings (5) and chicks (6), decreased invitro growth of embryonic tissue cells (7),and a number of behavioral and phys-iologic changes in rats (8). All theseresults are for fields considerably higher(.5-30 gauss) than the 60 Hz fields gen-erally found near power lines; however,the findings reported often appear to beunrelated to dose over the range studied.Prolonged exposure to the .001-.1 gaussrange most pertinent to wiring effects hasnot been explored experimentally.
Two studies suggest that a relativelystrong AC (alternating current) field mayinterfere with growth of implantedtumors in animals (9, 10) except wherethe tumor tissue is exposed to the fieldbefore implantation. In this latter condi-tion, tumor "takes" were increased (9).
To explore occupational exposure to ACmagnetic fields, we analyzed data from aUSPH publication on occupation by causeof death (11). All those occupationalcategories which seemed likely to includemen frequently exposed to AC magneticfields were grouped together and found tohave, as a group, a cancer rate signifi-cantly higher than the total population.The "exposed" categories included: powerstation operators; stationary engineers;linemen and servicemen, telephone, tele-graph and power; motor-men, street,subway and elevated railway; electri-cians; and welders and flame cutters. The
standard mortality ratio for cancer forthese categories combined was 115, a sig-nificant increase over the ratio of 100 forall occupations (x2 = 24.5,p < .0001). Forother "natural causes" of death this samegroup showed a standard mortality ratioof 102 (x2 = 1.8, not significant). Whilethis crude analysis in itself proves noth-ing, it underlines the fact that the harm-lessness of AC magnetic fields is stillunproven.
METHODS
Our cases consisted of persons dying ofcancer in Colorado before age 19 in theyears 1950-1973, who also had a Col-orado birth certificate. Only subjects withaddresses occupied from 1946-1973 inthe greater Denver area were used. Con-trols for these cases consisted of nextDenver-area birth certificates, chosenboth from the files organized by birth-month and county (file 1 controls), andfrom the alphabetical search-listings,which list all Colorado births alphabeti-cally within several wide spans of years:1939-1958, 1959-1969, and 1970-1974.These latter were called file 2 controls. Ifthe next birth certificate was that of asibling it was skipped.
Birth addresses were those listed on thebirth certificates. "Death" addresses wereobtained for both cases and controls bysearching for parents in city directories
for the two years just prior to diagnosis ofthe case. For cases who could not betraced, the address on the death certifi-cate was used. For controls, if the file 1control could not be traced, the file 2 con-trol with most similar birth date whocould be traced was used. There were nosignificant differences in the proportion of"high-current configurations" (HCC's asdefined below) shown by the file 1 controlsused (21 per cent HCC), the file 2 controlsused (23 per cent HCC) and the unusedextra controls (25 per cent HCC), so itseems unlikely that our method of select-ing controls biased our findings.
In all, 344 cases met our criteria.Thirty-nine of these were born before1946, and 33 had a birth address whichwas lost because it had been demolishedor was not adequately specified. Onlydeath addresses were analyzed for these72 cases and their respective controls.Similarly, 16 cases had no usable deathaddress, so only birth addresses were usedfor these cases and their controls. Table 2gives a summary of how many personsand how many addresses were availablefor cases and controls.
The procedure was simply to visit thebirth and "death" addresses of each caseand each control, and to draw a small mapof the electrical wires and transformers inthe vicinity. Primary (13 kv) wires werecategorized as either 'large-gauge" (built
TABLB 2
Distribution of persons and addresses available for analysis, for cases and controls, in a study ofelectrical wiring configurations and childhood cancer in Colorado in 1976—1977
Cases Controlsnesiaenuai status
StableMoved, birth and death
addresses availableOnly birth addressOnly death address
Totals
Persons*
109147
1672
344
Addresses*
109294
1672
491
Persons
128128
1672
344
Addresses
128256
1672
472
* Tables 3, 4, and 9 present data on total addresses, tables 5, 6, 7, 8, and 10 present data on total persons.Tables presenting data on persons are generally broken into total persons with an available birth address{N = 272) and totals with an available death address (N = 328).
to carry high currents) or "thin" depend-ing on whether they were clearly largerthan the secondary wires. Distances weremeasured from the part of the housenearest the wires to the wires, with a rol-latape.
Three types of homes, because of theirproximity to high-current wires, wereconsidered to have "high-current config-urations" (HCC's): 1) homes less than 40meters from large-gauge primaries or anarray of six or more thin primaries; 2)homes less than 20 meters from an arrayof 3-5 thin primaries or from high-tension (50-230 kv) wires; and 3) homesless than 15 meters from "first span" sec-ondary (240 volt) wires. First span second-aries were defined as those secondarieswhich issued directly from the trans-former and had not yet lost any currentthrough a service drop occurring beyondthe transformer pole. The span of second-ary wires separated from any trans-former by at least one intervening servicedrop (ignoring those drops directly at-tached to the transformer pole) werecalled second span secondaries. First spanwires will have more current runningthrough them than second span wires be-cause the first span must carry current forall the drops that mark its distal end pluswhatever current the second span re-quires.
All other configurations were consid-ered "low-current configurations"(LCC's). In addition, where first spanwires could be seen to be carrying currentto no more than two single family homes,on the average (on both sides of the block),those wires were called short first spanwires and, because they carried currentfor so few homes, they were always con-sidered LCC's, regardless of distance.Houses situated beyond the pole at theend of a secondary line ("end poles" in ta-bles 3 and 4) were considered the extremeexample of LCC homes, because they hadno distribution wires at all running pastthem.
Since the Denver area has been grow-ing fast, many new primary wires havebeen installed to accommodate increasedpower demands. Many of these new in-stallations are of a style easily distin-guished from older wires. For addressesoccupied before 1956 (20 years prior to ourfield work) we noted that only 59 per centof the primary wires found near our studyhomes were of the "old fashioned" typeswhich had been in use at the time of oursubjects' occupancy. (Actually 71 per centof the primary wires observed near pre-1956 case addresses were "old fashioned,"but only 49 per cent of the wires nearpre-1956 control addresses were of theolder types that could have been in use inthose early years.) Where the more mod-ern wiring was observed, we could not tellwhether it represented new installationsor replacement wiring, but we did knowthat it could not have been there in itspresent form in the pre-1956 years.Therefore, we decided to treat all primarywires seen near homes occupied before1956 as unreliable, and to code suchhomes strictly according to their more sta-ble secondary-wire configurations.
This adjustment did not critically affectour findings. Proximity to primary wireswas most strongly associated with cancerfor recent addresses, and the association(as expected) was weaker in the olderdata. But the association was still sig-nificant when all years were consideredand no adjustments made: For birth ad-dresses, 31 per cent of the 272 cases and22 per cent of the 272 controls had homesnear (unadjusted) primaries, a differencesignificant beyond the .025 level by Chi-square. For death addresses the figureswere 29 per cent of 328 cases and 19 percent of 328 controls, significant beyondthe .01 level.
RESULTS
General configurations. Table 3 showshow many cancer and control homesexhibited the various wiring configura-
tions. It can be seen that the most strikingdifference between cases and controls wasfound for subjects who had only one ad-dress from birth to death. This might bebecause, for subjects who moved, the ef-fects of configurations at one address werediluted by effects of configurations atother addresses.
Table 4 indicates that the greater theexposure to current expected from a givenwiring configuration, the greater the ex-cess of cancer found in homes where thatconfiguration was observed.
Type of cancer. The breakdown accord-ing to type of cancer (table 5) shows afairly similar excess of HCC's in cancercases for all categories but one, the deathaddresses of cases with "other tumors."
Such a wide association with differenttypes of cancer is not characteristic ofknown carcinogens such as ionizing radi-ation; thus the broad association observedhere suggests that the HCC-cancer rela-tionship may not be a causal one. Themost likely alternatives are that it is dueto some artifact, or that it reflects someeffect of HCC's on the body's general abil-ity to resist cancer.
Onset age. As table 6 shows, the HCC-cancer relationship was observed in bothyoung and older subjects. The fact thatthe relationship held for the birth as wellas the death addresses of older subjectswould seem to suggest that the effects ofHCC exposure can be long delayed. How-ever, a closer look at the data showed that
TABLE 3
Wiring configurations at the homes of cancer cases and controls, Colorado, 1976-1977
1st span secondaries >15 m"Short" first span wiresSecond span secondariesEnd poles
Total LCC's
(% HCC)
Stable
Case
214
01319
48
336
202
61
(44.0)
X5 =
residence:
Control
060
1010
26
43113315
102
(20.3)= 14.4
p < .001
Birth
Case
214
01126
53
539
408
110
(32.5)
X1 =P =
Movedaddress
Control
013
14
11
29
574
4014
115
(20.1)= 5.4= .02
residence:Death address
Case
238
11723
81
66115110
138
(37.0)
X* =P <
Control
017
01120
48
51196616
152
(24.0)7.6.01
t All six cases within 150 m of a substation were also less than 40 m from large primaries.* HCC =• high-current configuration; LCC = low-current configuration.
TABLE 4
Cancer related to the amount of current expected from different wiring configurations, Colorado, 1976-1977
23 (66 per cent) of the 35 older cases bornat HCC's were also living at a HCC (usu-ally the same address) within two years oftheir cancer onset. Only three (20 percent) of the 15 older controls born atHCC's were living at a HCC within twoyears of the "death" date. Thus the HCC-cancer relationship observed in the birthaddresses of older subjects can be largelyattributed to a HCC residence near thetime of cancer onset, and there is no needto posit a long-delayed effect of HCC's.
Urban-suburban differences. Sincecancer may show a different incidence inurban and non-urban areas, it seemedimportant to rule out the possibility thata difference in urbanization betweencases and controls was the significant vari-able in this study, and simply carried theHCC differences with it, spuriously. Thisseemed unlikely, intuitively, because thefield work was done one neighborhood ata time, and on none of the 22 days of fieldwork did the individual day's results failto show a preponderance of HCC's in thecase addresses.
A more formal survey shows that, al-though there was a slight excess of subur-ban addresses in the controls, it was notstatistically significant. Furthermore, thecases showed more HCC's than the con-trols independently in three areas: in oldDenver, in the more recently developedDenver areas (as estimated from a plan-ning department publication (12)), and inthe Denver suburbs (see table 7).
Socioeconomic class. The literature re-ports an excess of leukemia in families ofhigher socioeconomic class (13). Ourdata, dealing with all types of childhoodcancer, show only an insignificant trendin this direction. It seemed possible thatour method of choosing controls mighthave biased our control group againstlower-class controls, since only controlswho could be traced in directories wereused. However, a check on the discardedcontrols showed that upper and lower
class controls were discarded equally of- and sibling-order category tested, so weten, while Class HI controls were some- see no clue in these variables as to whywhat disproportionately retained. There the relationship between HCC's andwas no significant difference in the per cancer should exist,cent of discarded and retained controls Traffic congestion. A recent report (15)showing a HCC and, as table 8 shows, the suggests that cancer may occur undulyassociation between HCC's and cancer often near heavy-traffic routes. Our datawas observed within each social-class did show a mild excess of case-addressesgroup. It therefore seems unlikely that near such routes; case-addresses weresome spurious relationship to social class more likely than control-addresses to beexplains our findings. found within 40 meters of streets having a
Family pattern. The literature reports daily traffic count of 5000 vehicles oran excess of first siblings and older more on the 1960 Department of High-mothers among children with leukemia ways traffic map. However, once again, a(14). In our total sample of childhood significant excess of HCC's in cancer casescancer cases, a trend towards both more was found independently for addresses onfirst siblings and older mothers was heavy-traffic routes and for other address-noted, but neither was statistically sig- es. (For heavy-traffic routes, 53 per centnificant. Furthermore, the HCC-cancer of 74 case-addresses showed HCC'srelationship holds to approximately the against 30 per cent of 48 control-same degree within each maternal-age addresses; for other locations, 35 per cent
TABLE 6
Wiring configurations and cancer onset age, Colorado, 1976-1977*
Birth address
Death address
Type of wiringconfiguration!
HCCLCC
(% HCC)
HCCLCC
(% HCC)
Cancer onsetCase
66103
(39.1)
68105
(39.3)
0—6 yearsControl
40129
(23.7)
37136
(21.4)
OnsetCane
3568
(34.0)
6194
(39.4)
6-18 years
Control
1588
(14.6)
37118
(23.9)
* Case-control differences are significant by Chi-square (p < .01) for each category in the table.t HCC = high-current configuration; LCC = low-current configuration.
TABLE 7
Wiring configurations in different neighborhoods of cancer cases and controls in Colorado in 1976—1977*
Birth address
Death address
Type of wiringconfigure tionf
HCCLCC
(% HCC)
HCCLCC
(% HCC)
OldCase
4277
(35.2)
4962
(44.1)
DenverControl
2691
(22.2)
2477
(23.8)
Newer DenverCase
2740
(40.3)
3549
(41.7)
Control
944
(17.0)
1955
(25.7)
SuburbanCase
3254
(37.2)
4588
(33.8)
Control
2082
(19.6)
31122
(20.3)
* Case-control differences are significant by x1 (p < 05 or better) for each category in the table.t HCC = high-current configuration; LCC = low-current configuration.
Father's occupational class* at subject's birth, related to wiring configurations at birth residencesof cancer cases and controls, Colorado, 1976-1977
Type of wiringconfiguration!
HCCLCC
(% HCC)
Classes I and II
Case
1934
(35.8)
Xs -P <
Control
941
(18.0)3.2.10
Class m
Case
4998
(33.3)
X1
P
Control
30111
(21.3)= 4.7<.O5
Classes IV and V
Case
3339
(45.8)
x1
P
Control
1665
(19.8)= 10.8S.001
* Class categories follow the schema provided in "Mortality by Occupation Level and Cause of Death,"Vital Statistics Special Reports 53, #5, 1963, and are as follows: Class I: Professional. Class II: Technical,Administrative and Managerial. Class HI: Clerical, Sales, and Skilled Workers. Class IV: Semi-skilledWorkers. Class V: Laborers.
t HCC = high-current configuration; LCC = low-current configuration.
TABLE 9
Increase of cancer cases within 40 meters of heavy-traffic routes, as related to the presence or absence ofnearby* primaries, Colorado, 1976-1977
Type of subject
CasesControls(% cases)
NearTraffic routes
329
(78.0)
.05
primary wiresOther locations
8453
(61.3)
< p < .10
Not near primariesTraffic routes Other locations
42 33339 371
(51.9) (47.3)
Not significant
* "Nearby" primaries here means that the primaries were near enough to the house to qualify it as ahigh-current configuration (HCC).
of 417 case-addresses showed HCC's the males; 51 per cent of the 197 maleagainst 21 per cent of 424 control- cases had a HCC at their birth-or death-addresses), address, or both, while 45 per cent of the
In fact, the excess cancer we found on 147 female cases had such an address,heavy-traffic routes seems to be related to This compares with only 28 per cent of thethe frequent presence on such routes of 168 control males and also 28 per cent ofprimary wires carrying especially high the 176 control females,currents. Table 9 shows that the excess of It is interesting that significant malecancer cases on high-traffic routes occur- excess among our cancer cases appearedred to a significant extent only where to be confined to two categories: 1) casesprimary wires were nearby. whose birth address had a lower current
Sex distribution. Many cancers, includ- configuration than the death address, anding leukemia, occur more frequently in 2) cases with stable address who de-males than females. This is reflected in veloped cancer after at least one year ofour data where 57 per cent of our cases postnatal life at a residence situated nearwere males, as compared to 49 per cent of primary wires (table 10).the controls. The excess of HCC's among Because these two categories were cho-cases was significant for both males and sen from a number of ways we might havefemales when the sexes were analyzed categorized the data, they must remainseparately, but the trend was stronger in suspect until a replication confirms or
Sex distribution of cancer cases in a study of electrical wiring configurations and childhood cancerin Colorado in 1976-1977
Type of address
Birth address had lower currentconfiguration than death address
Stable residence at HCCt involvingprimary wires
Other cases with any HCC addressOther cases with no HCC address
Males
28
22
5691
Females
14
4
4881
% male
66.7
84.6
53.852.9
Significance*
X» = 4 .0 ,p<.05
Xs = 11.1, p < .001
X1 = 0.5, not significantX1 = 0.5, not significant
* An expected value of 50 per cent male was used to calculate the chi-squares.t HCC = high-current configuration.
disputes them. However, we chose thesecategories for a reason: We hypothesizedthat males might be excessively suscepti-ble to HCC's at all ages, including pre-natally. (It is of interest here that malerats appear especially susceptible to ex-perimental magnetic fields (8, p. 182)(16), as do embryos (17).) If males aremore susceptible, they might frequentlybe aborted when pregnancy occurs at aHCC, but pregnancy at a LCC wouldallow the most susceptible males to beborn and then to develop cancer laterwhen exposed to a situation with highercurrent nearby. This hypothesis is consis-tent with the male excess in category 1above.
Category 2 is presumed to provide a po-tentially similar situation: Where prima-ry wires are found running near a house(in 1976), it is always possible that thesewires were first installed or were "beefedup" at some time after the subject's birth.Or if they were present all along, the cur-rent they carry may sometimes have in-creased with time. If any of these thingshappens, the postnatal current flow nearthe house will be increased over the pre-natal flow, even without a change in resi-dence. Should this happen, the suscepti-ble male who escaped abortion duringpregnancy might develop cancer, and thiswould explain the male excess in cate-gory 2.
Substations. Power carried at highervoltage is stepped down to produce in-creased current at two points in our elec-trical distribution system: at the distri-bution substation, and again at theneighborhood transformers. As indicated,cancer cases were found in excess close tothe "first span" wires issuing from thetransformers. An even stronger trend wasfound for substations.
None of the 702 control addresses vis-ited (including our unused extra controls)was within 150 meters of a substation.This is to be expected since probably lessthan one home in 1000 in the Denver areais that near a substation. What is surpris-ing is that six of 491 case addresses werefound within 150 meters of a substationand, in each case, less than 40 metersfrom the large primary wires issuing fromthat substation. These six are shown intable 3. Each cancer case had lived at thesubstation address within three years orless of his illness. Although these num-bers are small, they are striking.
Blind studies. It should be noted thatour Denver-area study, being exploratory,was not done blind. This could lead to er-ror, although our observations werereasonably unambiguous. To check justhow reliable our coding was, an assistantobserved and coded 70 case and 70 controladdresses randomly chosen from thosepreviously coded by the principal inves-
tigator. The assistant did not know thecase-control status of the addresses shecoded. Her coding agreed with ours in 128(91 per cent) of the 140 instances. In fiveof the 12 instances of disagreement, theassistant's judgment favored the hy-pothesis of a HCC-cancer correlation,while ours did not. In seven instances, thereverse was true.
Also, a blind study was done (for birthaddresses only) in Colorado Springs andPueblo. This study showed the same cor-relation as the Denver study, similar indegree but less significant due to thesmaller numbers; 32 per cent of the 65cases and 18 per cent of the 65 controlsshowed HCC's. The correlation wasstrongest for cases with onset before sixyears of age, possibly because many of theolder cases had been gone from their birthaddresses for many years before theircancer onset.
DISCUSSION
It is not clear how residence near aHCC might affect the development ofcancer, but several possibilities should beconsidered:
1) Some association of both cancer andHCC's with a third factor may spuriouslyaccount for our correlation. Although wefound no indication of such a third factorin our analyses of social class, neighbor-hood, congested streets, or family make-up, the possibilities have not beenexhausted.
2) The magnetic fields produced by wirecurrents may somehow directly "cause"cancer. There is, however, no independentevidence or theoretical understandingwhich seems to support this possibility.The evidence concerning mutagenic ef-fects of extremely low frequency magneticfields, for instance, is ambiguous, butprobably negative (18).
3) Carcinogenic activity may be as-sociated with some indirect effect of theHCC's. For example, fields around power
lines might change the distribution ofsome ambient environmental carcinogen,such as particles which emit ionizingradiation. (However, the fields neardomestic wires are too weak to make thisseem probable.) Or the increased currentflowing in the plumbing might locally af-fect the drinking water. (There is often asmall amount of lead in copper waterpipes, for instance, due to imperfect sol-dering. And lead in the water supply iscorrelated with cancer, at least geo-graphically (19). However, it is not clearthat AC current in pipes could affect thissmall amount of lead enough to make adifference.)
4) AC magnetic fields might affect thedevelopment of cancer indirectly, throughsome effect on physiologic processes. It isconceivable, for instance, that contact-inhibition of cellular growth, or the basicimmune reaction of recognizing "self"from "not self," involves electrical poten-tials occurring at cell surfaces. Against anelectromagnetic background differentfrom that provided during evolution, anysuch cell mechanism might be altered.
Whatever the basis for our observedcorrelation, it should be emphasized that,although the risk of cancer appears to beincreased for children living near HCC's,it is rarely increased by a factor of morethan two or three. Therefore, if in thegeneral population one child in 1000 islikely to get cancer before age 19, no morethan two or three in 1000 living near aHCC would be expected to get it. Thepractical significance of the correlation, ifany, lies in the high prevalence of HCC's,not in any very high risk posed by mostHCC's.
REFERENCES
1. Polk C: Sources, propagation, amplitude andtemporal variation of extremely low frequency(0-100 Hz) electromagnetic fields./n Biologicaland Clinical Effects of Low Frequency Magneticand Electric Fields. Edited by JG Llaurado, ASances Jr. Springfield, IL, Charles C Thomas,1974, pp 21-48
2. Cook ES, Fardon JC, Nutini LG: Effects ofmagnetic fields on cellular respiration. InBiological Effects of Magnetic Fields. Edited byMF Barnothy. New York, Plenum Press, 1969,pp 67-78
3. Sheppard AR, Eisenbud M: Biological Effects ofElectric and Magnetic Fields of Extremely LowFrequency. New York, New York UniversityPress, 1977
4. Goodman EM, Greenebaum B, Marron MT: Ef-fects of extremely low frequency fields ongrowth and differentiation of physarumpolycepbalum. Naval Electronic StystemsCommand, Technical Report Phase I, April1975. Available from NT1S as ADA 010187
5. Coate WB: Seed germination and early growthstudy. In Project Sanguine Biological EffectsTest Program Pilot Studies, Naval ElectronicsSystems Command, Final Report, November1970, pp H-l to H-10. Available from NT1S asADA 717409
6. Krueger WF, Giarola AJ, Bradley JW, et al: In-fluence of low level electric and magnetic fieldson the growth of young chicks. Biomed Sci In-stnun 9:183-186, 1972
7. Yates VJ, Chang PW, Chen HJ, et al: Influenceof ELF magnetic and electric fields upon the irevitro growth rate of chicken embryo cells. InNaval Electronics Systems Command, Techni-cal Report Phase I, March 1975, pp 60-99 .Available from NTIS as ADA 007578
8. Persinger MA, Lafreniere GF, Ossenkoff K: Be-havioral, physiological and histological changesin rats exposed during various developmentalstages to ELF magnetic fields. In ELF and VLFElectromagnetic Field Effects. Edited by MAPersinger. New York, Plenum Press, 1974, pp177-225
9. LenziM: A report of a few recent experiments onthe biologic effects of magnetic fields. Radiology35:307-314, 1940
10. Batkin S, Tabrah FL: Effects of alternatingmagnetic field (12 gauss) on transplantedneuroblastoma. Res Commun Chem Pharmacol16:351-362, 1977
11. Guralnick L: Mortality by occupation and causeof death among men 20 to 64 years of age:United States, 1950. Vital Statistical Special Re-ports 53, #3, 1963
12. Denver Regional Council of Governments, Of-fice of Data Services: Profile of the Denver Re-gion, 1960-1970. May, 1973
13. Fasal E, Jackson EW, Klauber MR: Birthcharacteristics and leukemia in childhood. JNatl Cancer Inst 47:501-509, 1971
14. MacMahon B, Newill VA: Birth characteristicsof children dying of malignant neoplasms. JNatl Cancer Inst 28:231-244, 1962
15. Blumer M, Blumer W, Reich T: Polycyclicaromatic hydrocarbons in soils of a moun-tain valley: Correlation with highway trafficand cancer incidence. Environ Sci Tech11:1082-1086, 1977
16. Barnothy JM: Development of young mice. InBiological Effects of Magnetic Fields. Edited byMF Barnothy. New York, Plenum Press, 1964,pp 93-99
17. Reno VR, Nutini LG: Tissue respiration. InBiological Effects of Magnetic Fields. Edited byMF Barnothy. New York, Plenum Press, 1964,pp 211-217
18. Naval Electronics Systems Command: Navy-sponsored extremely low frequency biologicaland ecological research summary (update). May1976, pp 44-53. Available from NTIS as ADA02761
19. Berg JW, Burbank F: Correlations betweencarcinogenic trace metals in the water supplyand cancer mortality. Ann NY Acad Sci199:249-264, 1972
