Int. J.
Cancer: 46, 362-365 (1990)
0 1990 Wiley-Liss, Inc.
PRENATAL X-RAY EXPOSURE AND CHILDHOOD CANCER IN SWEDISH TWINS
Ylva RODVALL'.4, Goran PERSHAGEN', Zdenek HRUBEC~, Anders AHLBOM',Nancy L. PEDERSEN3 and John D. BOICE~
'Department of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, Stockholm; 2RadiationEpidemiology
Branch, National Cancer Institute, Bethesda, MD, USA; and 'Department of Environmental Hygiene, Karolinska Institute,
Stockholm. Sweden.
The association between X-ray exposure in utero and child-
hood cancer was studied in a case-control study, nested in a
national cohort of Swedish twin births. Records of the Swed-
ish Twin Register for those born I 9 3 6 I967 (n = 83,3 16) were
linked to the Swedish Cancer Register (1958-1983) and the
Swedish Cause-of-Death Register (I 952- 1983). Ninety-five
cases of childhood cancer before the age of 16 were identified.
Two controls from the Twin Register were matched to each
case by sex and year of birth. Maternal X-rays durin preg-
nancy were documented for 41% of the cases and 3 6 2 o f the
controls. The crude relative risk of any childhood cancer as-
sociated with any prenatal X-ray was 1.2 (95% CI; 0.7-2.1).
For the cases 64% of the X-rayed women had had abdominal
X-rays, the numbers for the controls were 57%. Fifty-nine per
cent of the X-rayed women had had abdominal X-rays, which
were associated with a relative risk for all cancers of I .4 (CI;
0.8-2.5), leukemias I .7 (CI; 0.7-4.1) and tumors of the central
nervous system I .5 (CI; 0.5-4.2). There was no apparent con-
founding by mother's age, drug use, obstetric complications,
previous miscarriages, social class or length of pregnancy. The
observed relative risks of cancer following prenatal X-ray ex-
posure are consistent with previous studies, suggesting that
the developing fetus may be more sensitive to the carcino-
genic effects of ionizing radiation than are children irradiated
post-natally.
Childhood cancer following prenatal X-ray is the only risk
consistently linked to low doses of radiation of the order of
10-50 mSv (Stewart et al., 1958; MacMahon, 1962; Monson
and MacMahon, 1984; Harvey et al., 1985; Stjernfeldt et al.,
1987; Gardner et al., 1990). However, the causal nature of this
association has been questioned (MacMahon, 1980, 1985; Tot-
ter and Macpherson, 1981), in large part because a comparable
excess risk of childhood cancer was not evident following in
utero exposure to the atomic bombs during World War I1 in
Japan (Jablon and Kato 1970). Although intensive searches
have failed to reveal any evidence of possible confounding
factors, however, the possibility remains that selecting women
for radiography during pregnancy might also select children
who are at higher risk of childhood cancer because of a variety
of conditions not easily ascertained (MacMahon, 1962; Doll,
1981).
To further study prenatal X-ray and childhood cancer, stud-
ies of twins have been recommended because their mothers
were usually X-rayed to confirm the twin pregnancy or to
determine fetal positions (MacMahon, 1980). One study of
cancer incidence conducted in twins confirmed the relationship
between childhood cancer and prenatal X-ray exposure
(Harvey et al., 1985). However, the small number of child-
hood cancers in that study, only 31, was a limitation. To in-
vestigate further the association between X-ray exposure in
utero and childhood cancer, a case-control study based on all
twin births in Sweden during 19361967 was conducted.
SUBJECTS AND METHODS
The Swedish Twin Registry (STR) comprises 110,136 twin
individuals born in Sweden between 1926 and 1967 (Medlund
et al., 1976). Among 83,316 twins born between 1936 and
1967, 106 childhood cancers occurring before age 16 were
identified through record linkage with the Swedish Cancer
Publication of the International Union Against Cancer
Publication de I'Union internationale Contre le Cancer
Registry (1958-1983) and the Swedish Cause of Death Regis-
try (1952-1983).
To verify the diagnoses, either medical records or death
certificates were reviewed for all cases. Four cases were ex-
cluded because they were erroneously classified as cancer. The
remaining 102 cases were divided into 3 diagnostic groups:
leukemia, tumors of the central nervous system, and other
diagnoses, in general following the classification suggested by
Birch and Marsden (1987). The diagnoses are given in Table I.
In one pair, both twins had a cancer diagnosis-one a central
nervous system tumor and the other a leukemia. These were
handled as two cases.
Control twins were selected from the subset of the STR in
which neither member of a pair was a case. Two controls were
matched to each case on the basis of year and month of birth,
sex and sex of co-twin. When both twins in a pair fulfilled the
matching criteria, one of them was selected randomly. Only
controls who were alive at the time of diagnosis of the case
were eligible.
The parish offices were then contacted to determine the
names of the mothers and the place where the twin deliveries
had taken place. To ascertain X-ray exposure, hospital records
were obtained from obstetric and X-ray departments. Seven
cases (Table I) and 14 controls had to be excluded (7%) be-
cause place of birth could not be traced or no record was
available. When no information was found for a control, a new
control was selected according to the original criteria. The final
numbers of cases and controls thus included in the analyses
were 95 and 190, respectively.
Coding of X-ray exposure and other medical record infor-
mation about pregnancy was done blindly with respect to case-
control status. Most of the information about X-rays was found
in the records of the obstetric clinics, whereas radiology de-
partment records seldom contributed any additional informa-
tion. The classification of the subjects as X-rayed or not was
carried out independently by one of the authors (Y.R.) and a
midwife, with complete agreement between them. Among
those having undergone in utero X-rays, one child was diag-
nosed only 2 days after the X-ray. This child was classified as
unexposed in the estimates. For the others, the interval be-
tween date of X-ray and date of cancer diagnosis ranged from
2 months to 15 years; for 45% it was less than 6 years.
As most X-ray films for this time period have been de-
stroyed, it was difficult to estimate the radiation doses. How-
ever, in a study in Stockholm, Larsson (1958) evaluated the
doses from different sources to the ovaries. For pelvimetries
done before 1958, the doses were estimated at 5 4 5 milliSiev-
ert (mSv), with a mean dose of 10 mSv. For abdominal ex-
aminations, the doses were estimated at 5-10 mSv. The doses
in the present study may vary to a large extent because many
hospitals were involved. The pelvimetry technique was
4 T whom
~ reprint requests should be sent, at the Institute of Environ-
mental Medicine, Box 60208, s-104 01 Stockholm, Sweden.
Received: February 21, 1990.
 X-RAY EXPOSURE A N D CHILDHOOD CANCER 363
TABLE I - DISTRIHUTION OF VERIFIED CANCEK DIAGNOSES BEFORE AGE TABLE II - RELATIVE RISK (MATCHED PAIR ANALYSES) OF CHILDHOOD
16 IN SWEDISH TWIN KEGISTRY BY AVAILABILITY OF INFORMATION ON CANCER AND CONFIDENCE INTERVAL ASSOCIATED WITH PRENATAL
X-RAY EXPOSURE X-RAY EXPOSURE
Exposure No exposure EXD UnexD. RR 95% CI
information information
available available All diagnoses n = 95
Leukemia 29 2 All X-ravs Cases 39 56 1.2 0.7-2.1
Acute lymphocytic or 15 1
Controls 69 121
Abdominal X-ravs Cases 25 70 1.4 0.8-2.5
non-lymphocytic Controls 39 151
Undifferentiated and 14 1
Leukemia n = 29
unspecified All X-rays Cases 12 17 1.0 0.4-2.6
Central nervous system 32 4
I1 2 Controls 24 34
Astrocytoma Abdominal X-rays Cases 10 19 1.7 0 . 7 4 1
Medullolblastoma 5 1
Controls 13 45
Other ghorna 6 1
Other inpacranial and 10 - Central nervous system tumors n = 32
All X-rays Cases 13 19 1.1 0.42.6
intraspinal neoplasm Controls 25 39
Other cancer 34 1
Abdominal X-rays Cases 8 24 1.5 0 . 5 4 . 2
Lymphomas and other 6 -
reticulo-endothelial Controls 12 52
neoplasms Other cancer n = 34
Retinoblastoma and 9 - All X-ravs Cases 14 20 1.7 0 . 7 4 . 2
nervous system tumors Controls 20 48
other than CNS Abdominal X-raw Cases 7 27 1.0 0.3-2.9
Malignant bone tumors 5 1 Controls 14 54
Other mid unspecified 14 -
malignant neoplasms
change in radiologic techniques during the 1960’s, the analyses
Total 95 7 were performed separately for X-ray examinations during
1936-1959 and during 1960-1967 (Table ID). For the earlier
period, the relative risk for matched pairs was 1.6 (CI; 0.8-
changed in Sweden about 1960, resulting in a substantial de- 3.4) and for the later 1.1 (CI; 0.4-2.7).
crease in doses (Bore11 and Fernstrom, 1960). When analyses were performed by time of X-ray during
The X-ray examinations were classified into 2 different pregnancy, the relative risk for matched pairs for all childhood
groups according to estimated level of exposure: (1) abdominal cancer was 8.0 (CI; 0.9-71.6) for the second trimester, and 1.1
X-rays (pelvimetry, plain abdominal, placenta previa, urogra- (CI; 0.6-2.1) for the third trimester (Table 111). There were no
phy) and (2) other X-rays (chest, pulmonary, cardiac). For 6 clear trends in childhood cancer risk in relation to age at diag-
cases and 7 controls, abdominal X-rays were done twice during nosis or mother’s age.
the pregnancy. The children’s records were also investigated
for X-rays taken within 6 months after birth. However, only 3 TABLE 111 - RELATIVE RISK (MATCHED PAIR ANALYSES) OF CHILDHOOD
cases and 4 controls had documented X-rays during this period. CANCER AND CONFIDENCE INTERVAL ASSOCIATED WITH ABDOMINAL
X-RAY BY CHARACTERISTICSOF CASE
Relative risks for matched pairs and 95% maximum likeli-
Exp. Unexp. RR 95%CI
hood method confidence intervals (CI) were estimated. Multi-
ple conditialnal logistic regression analysis on the matched sets Year of birth’
was also performed to adjust for the effects of potential con- 19361959 Cases 15 50 1.6 0.8-3.4
founding factors (Breslow and Day, 1980). Controls 20 110
Potential confounders included in the analyses were: moth- 1960-1967 Cases 10 20 1.1 0.4-2.7
er’s age, drug use, obstetric complications, previous miscar- Controls 19 41
Trimester of X-ray exposure
riages, social class based on a Swedish standard classification 2nd Cases 4 91 8.0 0.9-7 1.6
(Statistics Sweden, 1984), and length of pregnancy. Controls 1 189
3rd Cases 21 74 1.1 0.62.1
RESULTS Controls 38 152
Age a? cancer diagnosis
X-rays during pregnancy were performed on the mothers of 0-5 years Cases 12 19 1.6 0.7-4.0
41% of the cases and 36% of the controls. An abdominal X-ray Controls 17 45
was documented for 26% of the cases (25 of 95) and for 21% 6 1 0 years Cases 3 21 0.5 0.1-2.2
of the controls (39 of 190). Of the 25 cases and 39 controls Controls 10 38
with abdominal X.-rays 16% and 3%, respectively, were in the 11-15 years Cases 10 30 1.8 0.7-4.6
second trimester. The remaining X-rays were in the third tri- Controls 12 68
Mother’s age
mester. Only 7 of all abdominal X-rays were pelvimetries; one 525 Cases 8 24 1.1 0.4-3.2
was in the second and 6 were in the third trimester. Controls 9 50
For all maternal X-ray exposure during pregnancy, the rel- 26-35 Cases 12 34 1.7 0.7-3.9
ative risk fior matched pairs of any childhood cancer was 1.2 Controls 22 72
(CI; 0.7-2.1) (Table 11). When only abdominal X-rays were 236 Cases 5 12 1.3 0.4-4.1
included, the relative risk for all cancers was 1.4 (CI; 0.8-2.5), Controls 8 29
Sex
for leukemia RR = 1.7 (CI; 0 . 7 4 . l), for tumors of the cen- Boys Cases 13 32 1.7 0.84.0
tral nervous system RR = 1.5, (CI: 0.54.2), and for other Controls 17 73
tumors RR = 1.0, (CI; 0.3-2.9). For the latter group, how- Girls Cases 12 38 1.1 0.5-2.5
ever, there was an increased risk associated with all X-rays, Controls 22 78
RR = 1.7 (CI; 0.7-4.2). ‘Controls were matched to cases for year and month of birth, sex and sex of
In order to assess whether the relative risk differed after the co-twin.
364 RODVALL ET A L .

Adjustments for potential confounders including mother’s pregnancy or to determine the fetal position. However, it was
age, drug use, obstetric complications, previous miscarriages, sometimes difficult to determine the reason for the X-ray, and
social class and length of pregnancy produced relative risk the relatively low X-ray frequency overall presents the possi-
estimates between 1.3-1.5. They were virtually the same as the bility that twins may not be completely free from selection due
crude relative risks and, thus, are not reported here. to medical indications for the exposure.
Separate analyses of other factors related to pregnancy were We were only able to control for such potential confounders
performed to investigate whether there was an association with about which the records provided information (mother’s age,
childhood cancer (Table IV). For those whose mothers had a drug use, obstetric complications, previous miscarriages, so-
history of miscarriage, the relative risk of childhood cancer cial class and length of pregnancy). When the crude and lo-
was 2.0 (CI; 1 . W . 3 ) . For some complications such as tox- gistic regression results were compared it appeared that none of
emia and infections, there was a suggestion of an increased these factors were in fact confounders. It is not clear whether
relative risk. When drugs were given for infections the relative any of the other suggested risk factors for childhood cancer,
risk of childhood cancer was 3.5 (CI; 1.0-12.0). Detailed anal- such as maternal smoking during pregnancy, are likely to be
yses in relation to specific drugs revealed no clear associations confounding factors in this study (UNCSEAR, 1986).
(not shown in Table). Some X-ray examinations may have been performed but not
noted in the medical records. There were, however, other
DISCUSSION sources of information in the records from which one could
Prenatal abdominal X-ray exposure was associated with a deduce whether or not an X-ray had been carried out. For
40% excess risk of developing childhood cancer. Although instance, when the twin pregnancy was not identified as such
consistent with previous investigations (Stewart et al., 1958; until the delivery, late pregnancy X-ray could hardly have been
MacMahon, 1962; Monson and MacMahon, 1984; Harvey et carried out. To the extent that there was a loss of information
al., 1985; and Stjernfelt et al., 1987), our findings are based on about exposure, it was likely to affect cases and controls to the
a limited number of cases and should be interpreted with cau- same extent and thus lead to a dilution and an underestimation
tion. Nevertheless, the risk was highest following abdominal of the excess risk, if any.
X-rays, which would have exposed the fetus directly. There For 7% of all subjects, neither place of birth nor medical
was also a higher relative risk for the years 1936-1959, when records were available. There were no differences between
the exposures were likely to be higher (RR = 1.6) compared cases and controls in this respect. The hospitals which could
to the years 1960-1967 (RR = 1 . 1 ) . The results indicate ele- not provide records for some of the subjects were situated all
vated relative risks primarily for leukemias and tumors of the over Sweden. The replacement of controls for which we could
central nervous system. On the other hand, several circum- not find any information was therefore not likely to have in-
stances of pregnancy, such as previous miscarriages and use of fluenced the results.
drugs for infection, were linked to an elevated risk of child- As pointed out in previous studies, postnatal X-rays could be
hood cancer and suggest other factors that might also be of a risk factor for childhood cancer and therefore it was impor-
importance in causing cancer in these subjects. tant to document these. It was easier to find out about postnatal
Twins were selected for study to minimize the potential for X-rays for the cases since extensive medical records from the
confounding from factors associated with X-ray examination pediatric clinics were available. Therefore, information about
and to take advantage of the high frequency of radiologic ex- X-rays within 6 months after birth may be biased. However,
amination in twin pregnancies. Among all Swedish mothers the records of only 3 cases indicated the presence of postnatal
having twin pregnancies, 2 1 % had abdominal X-rays during X-rays in this period, and this information has not been used
the period 1936-1967, while among all pregnancies at the for any of the analyses in our report.
Karolinska Hospital in Stockholm during 1940-1957, the same Although this study is based on cancers among all twins born
proportion was 8% (Borell and Fernstrom, 1960). The medical in Sweden during 3 decades, the limited study size remains a
records of this study indicate that the majority of the X-ray problem as can be seen from the wide confidence intervals.
examinations were performed in order to confirm the twin Recently, prenatal X-ray exposures have become uncommon,
doses have been greatly reduced and any real effects would be
TABLE IV - RELATIVE RISK (MATCHED PAIR ANALYSES) OF CHILDHOOD small and difficult to detect. Therefore, extending the time
CANCER AND CONFIDENCE INTERVAL ASSOCIATED WlTH over which cases could be collected probably would not correct
CHARACTERISTICS OF PREGNANCY this deficiency.
With Without X-ray techniques changed in Sweden about 1960 (Borell and
Characteristic character- character- RR 95% CI
irtir id,. Fernstrom, 1960) partly as a result of the presentation of Stew-
art’s data (1 958). The higher relative risk for the years before
Social class’ Cases 7 80 0.8 0.3-2.1 1960 is consistent with the introduction of the new techniques
Controls 14 148 and concomitant lower doses. However, the doses are difficult
Mother’s age Cases 17 78 0.9 0.4-1.8 to confirm as many X-ray departments were involved, and the
236years Controls 37 153
Miscarriage Cases 15 78 2.0 1.0-4.3 changes were not introduced simultaneously in all hospitals.
Controls 16 174 Nevertheless, our results are similar to the large Oxford Sur-
Toxaemia Cases 13 82 1.6 0.7-3.2 vey, in which risks also declined over time, parallel with the
Controls 17 173 decline in dose per film (Bithell and Stiller, 1988).
Infection Cases 6 89 1.6 0.5-4.7 MacMahon (1962) and Monson and MacMahon (1984) re-
Controls 8 182
Drugs for Cases 7 88 3.5 1.0-12.0 ported a higher relative risk for those exposed early in gesta-
infection Controls 4 186 tion, though the numbers were small. In the present study there
Sedatives or Cases 12 83 1.5 0.7-3.1 were no exposures in the first trimester, and exposure in the
sleeping pills Controls 16 174 second trimester was associated with a higher risk. We also
Length of Cases 28 67 1.3 0.7-2.2 evaluated cancer risks associated with other prenatal factors for
pregnancy Controls 48 142 which information could be obtained. There is a suggestion
‘Professionals and other higher non-manual employees and all self-employed ex- that conditions surrounding a previous pregnancy loss could
cept farmers. increase the risk of childhood cancer. An increased risk was
 X-RAY EXPOSURE AND CHILDHOOD CANCER 365

associated with more than 2 miscarriages by Graham et al. the strongest association was seen for drugs taken for infections
(1966) and by Van Steensel-Moll et al. (1985) but not by during pregnancy. It was not possible to evaluate whether the
Monson and Machfahon (1984). In the study by Harvey et al. association with drug use was merely an indication of an un-
(1985), previous pregnancy loss appeared to modify the effect derlying agent, i.e., infection per se.
of prenatal X-ray exposure. In the present study a relative risk We observed relative risks following prenatal exposure that
of 2.0 (CI 1.0-4.3) was found for previous miscarriage. are consistent internally with respect to time of X-ray and the
Stewart eit al. (1958) did not find an association between amount of radiation exposure and with previous studies, sug-
drugs taken during pregnancy and childhood cancer. Van gesting that the developing fetus may be sensitive to the car-
Steensel-Moll et al. (1985) found an increased risk associated cinogenic effects of ionizing radiation.
with sedatives or sleeping pills but not with antibiotic con-
sumption. Ciilman et al. (1989) found a dose response rela-
tionship for both drugs and illnesses, but the conclusion was ACKNOWLEDGEMENTS
that the effects of drugs taken during pregnancy appeared to be
secondary to the effects of the illnesses for which they were We thank Ms. A. C. Wistedt and Ms. B. Aase for help with
taken. In the present study there was a similar indication of a the data collection and coding of records. This work was sup-
higher risk iissociated with sedative and sleeping pill use, but ported in part by a National Cancer Institute contract.
REFER.ENCES
BIRCH,J.M. and MAKSDEN, H.B., A classification scheme for childhood MACMAHON, B., Prenatal X-ray exposure and childhood cancer. J. nut.
cancer. Int. J. Cance,r, 40, 620-624 (1987). Cancerlnst., 28, 1173-1191 (1962).
BITHELL,J.F. and STILLER,C.A., A new calculation of the carcinogenic MACMAHON,B., Childhood cancer and prenatal irradiation. In: J.H.
risk of obstetric X-raying. Stat. Med., 7, 857-864 (1988). Burchenal and H.F. Oettgen (eds.), Cancer achievements, challenges and
prospects for the 1980s. Vol. 1, pp. 223-228, Grune and Stratton, New
BORELL,U. and FEKNSTROM, I., Radiologic pelvimetry. Acta Radiol., Yolk (1980).
Suppl. 191 (1960).
MACMAHON, B., Prenatal X-ray exposure and twins. New Engl. J. Med.,
BRESLOW, N.E. and DAY,N.E., Sfatistical methods in cancer research. 312, 576 (1985).
Vol 1. The analyses of case-control studies. IARC Scientific Publications
32, IARC, Lyon (1980). MEDLUND,P., CEDERL~F, R . , FLODERUS-MYRHED, B., FRIBERG,L. and
S~RENSEN, S., A new Swedish twin registry. Acta med. scand. Suppl. 600
DOLL,R., Radiation hazards: 25 years of collaborative research. Brit. J . (1976).
Radiol., 54, 179-186 (1981). MONSON,R.R. and MACMAHON, B., Prenatal X-ray exposure and cancer
GARDNER, M.J., SNEE,M.P., HALL,A.J., POWELL,C.H., DOWNES,S. in children. In: J.D. Boice, Jr. and J.F. Fraumeni, Jr., (eds.), Radiation
and TERREL,J.D., Results of case-control study of leukemia and lym- carcinogenesis: Epidemiology and biological significance. pp. 97-105,
phoma among young people near Sellafield nuclear plant in West Cumbria. Raven Press, New York (1984).
Brit. Med. J., 300, 423-429 (1990). STATISTICS SWEDEN,Swedish socioeconomic classification, SCB-tryck.
GILMAN,E.A., WILSON,L.M.K., KNEALE,G.W. and WATERHOUSE, Orebro (1984).
J.A.H., Childhood cancers and their association with pregnancy drugs and STEWART, A., WEBB,J. and HEWITT,D.I., A survey of childhood rna-
illnesses. Pediatr. Perinatal. Epidemiol., 3, 66-94 (1989). lignancies. Brit. med. J., 1, 1495-1508 (1958).
GRAHAM, S., LEVIN,M.L., LILIENFELD, A.M., SCHUMAN, L.M., GIB- STJERNFELDT, M., BERGLUND, K., LINDSTEN,S . and LUDVIGSSON, I.,
SON, R., DovvD, J.E. and HEMPELMANN, L., Preconception, intrauterine Maternal smoking and irradiation during pregnancy as risk factors for
and postnatal irradiation as related to leukemia. Nat. Cancer Inst. childhood leukemia. VTT-Grafiska, Vimmerby (1987).
Monogr., 19, 347-3’71 (1966).
TOTTER,J.R. and MACPHERSON, H.G., Do childhood cancers result from
HARVEY, E B . , BOICE,J.D., JR., HONEYMAN, M. and FLANNERY, J.T., prenatal X-rays? Health Phys., 40, 511-524 (1981).
Prenatal X-ray exposure and childhood cancer in twins. New Engl. J.
Med., 312, 541-545 (1985). UNITEDNATIONSSCIENTIFIC COMMITTEE ON THE EFFECTSOF ATOMIC
RADIATION. Genetic and somatic effects of ionizing radiazjon -UN-Kbl.,
JABLON,S. and KATO,H., Childhood cancer in relation to prenatal expo- E.86.IX.9, UNO, New York (1986).
sure to atomic-bomb radiation. Lancet, 11, 1000-1003 (1970).
VAN STEENSEL-MOLL, H.A., VALKENBURG, H.A., VANDENBROUCKE,
LARSSON, L.-E., Radiation doses to the gonads of patients in Swedish J.P. and VAN ZANEN,G.E., Are maternal fertility problems related to
roentgen diagnostics Acta Radiol., Suppl. 157 (1958). childhood leukemia? Int. J. Epidemiol., 14, 555-559 (1985).