CancerCausesand Control,3, 527- 532
Childhood cancer among Swedish twins
Ylva Rodvall, Zdenek Hrubec, Giiran Pershagen, Anders Ahlbom, Arne Bjurman, and John D. Boice, Jr (Received 25June 1992;accepted29July 1992) The risk of childhood cancer was evaluated among 35,582 twins born in Sweden between 1952 and 1967. Cancers were identified through linkage with national cancer and mortality registries. Overall, 59 childhood cancers and 41 deaths from cancer occurred before the age of 16. For both sexes combined, the cancer incidence was similar to that in the general population of primarily single-born children (number observed/ number expected [O/E] -- 1.0, 95 percent confidence interval [CI] = 0.7-1.2). For males under age five, cancer incidence was reduced significantly (O/E = 0.3, CI = 0.1-0.7). There was a substantial increase in all-cause mortality among twins (O/E = 3.7), attributable to a high mortality during the first year of life. Excluding this first year, the O/E for death (all causes) was 1.1 (CI = 1.0-1.3). For cancer mortality of both sexes, the O/E was 0.9 (CI = 0.6-1.2), with no significant reduction of risk in any sex-age group. We conclude that with the possible exception of males aged 0-4 years, the childhood cancer risk of twins appears similar to that of singletons.
Key words:Cancer incidence, childhood cancer, mortality, twins, Sweden.
Introduction Twins, because of their relatively low weight at birth, die at a much higher rate during the first year of life than single-born children. This high mortality rate has made it difficult to evaluate accurately the risk of childhood cancer in twin populations. Although a number of studies suggest that cancer rates among twins may be low, 1-6the findings are not consistent. Several studies from Scandinavia, for example, report cancer occurrence in twins to be similar to that of the general population. 7-1~ Interestingly, no study has indicated an increase in cancer among twins. The latter observation is noteworthy since, in the past, twins were exposed more frequently to prenatal X-ray than singletons.
Such X-ray exposure has been linked to increases in childhood cancer, both among singletons 12-15 and a m o n g t w i n s Y 6d7
To provide information on the risk of childhood cancer among twins, we examined cancer incidence and mortality in a register of all twin births in Sweden from 1952 through 1967.
Materials and Methods The Swedish Twin Register (STR) is comprised of all twins born in Sweden during 1926-67. is A roster of 35,582 twins born alive between 1952-67 was linked to
Drs Rodvall, Pershagen,Ahlbom, and Bjurman are with the Department of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden. Drs Hrubec and Boice are with the Radiation Epidemiology Branch, Division of Cancer Etiology, National Cancer Institute, Bethesda, Maryland, USA. Dr Pershagenis also at the Department of Environmental Health and Infectious Diseases Control, Karolinska Hospital, Stockholm, Sweden. Dr Ahlbom is also at the Department of Occupational Health, Epidemiology Unit, Karolinska Hospital, Stockholm, Sweden. Address correspondenceto Dr Rodvall, Institute of Environmental Medicine, Box 60208, S-10401 Stockholm, Sweden. This study was supported in part by the National Cancer Institute contract NO1-CP-51033 and in part by the MacArthur Foundation. © 1992 Rapid Communications of Oxford Ltd
Cancer Causes and Control. Vol 3. 1992
527
Y. Rodvall et al the Swedish Cause-of-Death Register, which has been computerized since 1952, and to the Cancer Register, which started in 1958. Since 1947, every individual in Sweden has been assigned a unique civil-registration number which was used in the linkage procedures. For 1,397 twins who did not have a civil registration number recorded in the STR, records in the Parish Office Registers were searched to determine whether they were stillborn. Surviving twins without civil registration numbers were linked to the cancer and mortality registers based on name and birth date. The twins were followed up through age 15. Diagnostic criteria and coding practices have changed over the observation period. The Swedish Cancer Register used the seventh edition of the International Classification of Diseases (ICD-7) 19throughout, while the Cause-of-Death Register used the ICD-82° from 1969 to 1983. In our study, the ICD-7 was used for the cause-of-death data since ICD-7 codes cannot be translated completely into the ICD-8 classification. Standardized ratios of observed to expected incident cases or deaths (O/E) were employed to compare the cancer experience of the twin cohort with that of a comparable Swedish population estimated from national rates specific for sex, age, and calendar time. Approximate 95 percent confidence intervals (CI) were computed except when the observed number of cancers was 20 or fewer. In these instances, exact confidence intervals were calculated based on the Poisson distribution? 1 Person-years (PY) at risk were computed from the date of birth until the date of death or through age 15. There were no twins with more than one cancer. Twins dying within the first month were assumed to have lived for only four days, the average survival for children dying within their first month of life. Those dying between I and 12 months were assumed to have lived 15 days during the month of death. PY at risk for individuals diagnosed with cancer were counted to the date of death or age 15, whichever came first. Twins born before 1952 could not be followed accurately for survival and were excluded. Our previous case-control study was able to include more calendar years and many more twin births (83,316) because it did not require determining individual survival for the entire twin population. ~7 Expected numbers were calculated by multiplying incidence and mortality rates for the general population by the numbers of PY at risk, stratified by age and calendar year in one-year categories, and by sex. Rates for the general population were based on the mean population for each year and age, including the first year. 528
Cancer Causes and Control. Vol 3. 1992
T a b l e 1. N u m b e r o f cancers r e p o r t e d to t h e C a n c e r R e g i s t r y and t h e C a u s e - o f - D e a t h Registry, b y sex, for S w e d i s h t w i n s b o r n 1952-67 and f o l l o w e d u p to age 16 years Cancer Registry (1958-83)
Leukemia (ICD-7 # 204) Acute lymphatic leukemia Myeloid leukemia Other and unspecified leukemia Total, leukemia Tumors of the central nervous system (CNS) (ICD-7 # 193.0-193.2, 195.3) Brain Intraspinal, meninges Total, CNS Other cancer Lymphomas and other reticuloendothelial neoplasms Retinoblastoma and nervous system tumors other than CNS Bone tumor Other and unspecified malignancies Total, other cancer Total, all cancers
Cause-of-Death Registry (1952-83)
Male
Female
Male
Female
5 2
2 1
2 1
5 --
2 9
5 8
7 10
4 9
8 -8
11 2 13
5 -5
5 2 7
5
--
3
--
-1
6 2
---
-3
4~ 10
3b 11
27
32
2c 5 20
2a 5 21
ICD-code • Includes: appendix maxillary sinus renal parenchyma site unknown
153.4 160.2 180.0 199.9
b Includes: maxillary sinus renal parenchyma site unknown
160.2 180.0 199.9
c Includes: peritoneum suprarenal gland
158.9 195.0
d Includes: suprarenal gland connective tissue (unspecified)
195.0 197.9
Results Fifty-nine childhood cancer cases and 41 childhood cancer deaths occurred before age 16. Of the 41 deaths, 12 were not recorded in the Cancer Register (seven occurred before 1958, two were classified erroneously as leukemia deaths, and three were missed). The 59 incident cases of cancer included 21 tumors of the central nervous system (CNS) and 17 leukemias (Table 1). There was only one twin-pair concordant for can-
Cancer in twins
Table 2. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site and sex Males
Females
O
E
O/E
CI
O
Total
E
O/E
CI
O
E
O/E
CI
Leukemia Tumors of the central nervous system Other cancers
9
9.7
0.93
0.4-1.8
8
8.4
0.95
0.4-1.9
17
18.1
0.94
0.5-1.5
8 10
8.4 15.1
0.95 0.66
0.4-1.9 0.3-1.2
13 11
7.6 12.5
1.71 0.88
0.9-2.9 0.4-1.6
21 21
16.0 27.6
1.31 0.76
0.8-2.0 0.5-1.2
All cancers
27
33.2
0.81
0.5-1.2
32
28.5
1.12
0.8-1.6
59
61.7
0.96
0.7-1.2
Table 3. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site, sex, and calendar year of diagnosis Calendar year
Males O
1958-70 Leukemia Tumors of the nervous system All cancers
Females
E
O/E
CI
6
7.1
0.85
0.3-1.8
5 16
5.6 22.9
0.89 0.70
3
2.6
3 11
2.8 10.3
O
Total
E
O/E
CI
O
E
O/E
CI
6
6.4
0.94
0.3-2.0
12
13.5
0.89
0.5-1.6
0.3-2.1 0.4-1.1
9 24
5.2 19.8
1.73 1.21
0.8-3.3 0.8-1.8
14 40
10.8 42.7
1.30 0.94
0.7-2.2 0.7-1.3
1.15
0.2-3.4
2
2.0
1.00
0.1-3.7
5
4.6
1.09
0.4-2.6
1.07 1.07
0.2-3.2 0.5-1.9
4 8
2.5 8.7
1.60 0.92
0.4-4.2 0.4-1.8
7 19
5.2 19.1
1.35 0.99
0.5-2.7 0.6-1.6
central
1971-83 Leukemia Tumors of the central nervous system Allcancers
Table 4. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site, sex, and by age at diagnosis Males
0-4 Years Leukemia Tumors of the central nervous system All cancers 5-9 Years Leukemia Tumors of the central nervous system Allcancers 10-15 Years Leukemia Tumors of the central nervous system All cancers
Females
Total
O
E
O/E
CI
O
E
O/E
CI
2
4.3
0.47
0.1-1.7
5
3.7
1.35
0.4-3.2
1 3
2.4 12.2
0.42 0.25
0.1-2.3 0.1-0.7
3 11
2.1 10.5
1.43 1.05
2
2.8
0.7
0.1-2.6
1
2.6
0.38
4 8
2.6 9.0
1.54 0.89
0.4-4.0 0.4-1.8
5 8
5
2.7
1.85
0.6-4.4
3 16
3.4 12.0
0.88 1.33
0.2-2.6 0.8-2.2
cer; o n e t w i n had l e u k e m i a a n d t h e o t h e r h a d b r a i n c a n cer. T h e r e w e r e n o c o n c o r d a n t l e u k e m i a pairs. O v e r a l l , c a n c e r i n c i d e n c e w a s s i m i l a r to t h a t e x p e c t e d in t h e g e n e r a l p o p u l a t i o n ( T a b l e 2), a l t h o u g h a d e f i c i t w a s s u g g e s t e d a m o n g m a l e s ( O / E = 0.8). F o r f e m a l e s , t h e r e was a s l i g h t l y i n c r e a s e d risk f o r t u m o r s o f t h e C N S ( O / E -- 1.7). R i s k o f l e u k e m i a was s l i g h t l y
O
E
O/E
CI
7
7.9
0.89
0.4-1.8
0.3-4.2 0.5-1.9
4 14
4.5 22.7
0.89 0.62
0.2-2.3 0.3-1.0
0.0-2.2
3
5.3
0.57
0.1-1.6
2.2 7.5
2.27 0.7-5.2 1.07 0.5-2.1
9 16
4.8 16.5
1.88 0.97
0.9-3.5 0.6-1.6
2
2.1
0.95
0.1-3.4
7
4.8
1.46
0.6-3.0
5 13
3.3 10.5
1.52 1.24
0.5-3.5 0.7-2.1
8 29
6.7 22.6
1.19 1,28
0.5-2.3 0.9-1.8
d e c r e a s e d ( O / E = 0.9). T h e d e f i c i t o f c a n c e r a m o n g m a l e s was a p p a r e n t o n l y in t h e y e a r s 1958-70 ( T a b l e 3). F o r m a l e t w i n s b e t w e e n 0 and 4 y e a r s at d i a g n o s i s , c a n c e r risk was s i g n i f i c a n t l y l o w ( O / E = 0.3, C I = 0.10.7). A m o n g f e m a l e s in this age g r o u p , t h e O / E was 1.1 ( T a b l e 4). T w i n s w e r e at e x c e p t i o n a l l y h i g h risk o f d e a t h f r o m Cancer Causes and Control. Vol 3. 1992
529
Y. Rodvall et al Table 5. O b s e r v e d (O) and expected (E) all-cause mortality, O / E and 95% confidence interval (CI) on O / E b y site and sex Males
All deaths Deaths (0-1 years) Deaths (1-15 years) All cancers Leukemia Tumors of the central nervous system
Females
Total
O
E
O/E
CI
O
E
O/E
CI
O
E
O/E
CI
1,697 1,546 151
463.0 324.3 138.7
3.67 4.77 1.09
3.5-3.8 4.5-5.0 0.9-1.3
1,229 1,119 110
326.7 232.3 94.1
3.76 4.82 1.17
3.6-4.0 4.5-5.1 1.0-1.4
2,926 2,665 261
789.7 556.8 232.9
3.71 4.79 1.12
3.6-3.8 4.6-5.0 1.0-1.3
20 10
25.1 12.0
0.80 0.83
0.5-1.2 0.4-1.5
21 9
20.5 9.6
1.02 0.94
0.6-1.6 0.4-1.8
41 19
45.6 21.6
0.90 0.88
0.6-1.2 0.5-1.4
5
4.6
1.09
0.4-2.6
7
4.2
1.67
0.7-3.4
12
8.8
1.36
0.7-2.4
Table 6. Observed (O) and expected (E) mortality, O/E and 95% confidenceinterval (CI) on O/E by site, sex, and age at death Males
0-4 Years All deaths All cancers Leukemia Tumors of the central nervous system 5-9 Years All deaths All cancers Leukemia Tumors of the central nervous system 10-15 Years All deaths All cancers Leukemia Tumors of the central nervous system
O
E
O/E
1,624 7 4
385.6 10.5 5.0
3
1.8
1.67
36 7 3
39.5 7.3 4.1
2
Females CI
O
E
O/E
CI
O
E
1,173 7 4
276.1 8.9 3.8
4.25 0.79 1.05
4.0-4.5 0.3-1.6 0.3-2.7
2,797 14 8
661.7 19.4 8.8
0.3-4.8
2
1.6
1.25
0.1-4.5
5
3.4
1.47
0.5-3.4
0.91 0.96 0.73
0.6-1.3 0.4-2.0 0.2-2.1
27 8 4
26.0 6.7 3.7
1.04 1.19 1.08
0.7-1.5 0.5~2.4 0.3-2.8
63 15 7
65.5 14.0 7.8
0.96 1.07 0.90
0.7-1.2 0.6-1.8 0.4-1.9
1.3
1.54
0.2-5.5
2
1.2
1.67
0.2-5.9
4
2.5
1.60
0.4-4.0
37 6 3
37.9 7.2 2.9
0.98 0.83 1.03
0.7-1.3 0.3-1.8 0.2-3.0
29 6 1
24.6 5.0 2.1
1.18 1.20 0.48
0.8-1.7 0.4-2.6 0.0-2.7
66
62.5
12
12.2
4
5.0
1.06 0.8-1.3 0.98 0.5-1.7 0 . 8 0 0.2-2.1
0
1.4
0.00
0.0-2.6
3
1.4
2.14
0.4-6.3
3
2.8
1.07
4.21 4.0-4.4 0.67 0.3-1.4 0 . 8 0 0.2-2.0
all causes (O/E = 3.7), attributable almost entirely to a high mortality in the first year of life. Over 91 percent of the 2,926 deaths occurred prior to age one year. There were no differences by sex, overall or in the first year. The data also were analyzed excluding deaths and PY in the first year of life. The O/Es for cancer deaths were the same as in the general population (Table 5). There were no significant differences in the O/E value for cancer by age at death, although they were somewhat lower for twins under age five than at older ages (Table 6).
Discussion Compared with the general population, twins born in Sweden during 1952-67 had a nearly fivefold risk of 530
Total
CancerCausesand Control Vol 3. 1992
O/E
CI
4 . 2 3 4.1-4.4 0.72 0.4-1.2 0.91 0.4-1.8
0.2-3.1
dying within the first year of life. The reasons for this high mortality are likely to be factors associated with premature birth and low birthweight common among twin pregnancies. When followed to age 16, those surviving the first year experienced a small (12 percent) increased risk of dying. Overall, the risk of cancer (O/ E = 0.96) was similar among twins compared with the general population of primarily single-born children. Both leukemia and other neoplasms occurred slightly below expectation. The only significant finding was a deficit of cancer incidence among males under age five. Most, but not all, studies of twins find a lower rate of childhood cancer compared with singletons.T M Inconsistencies reflect, in all likelihood, the play of chance due to small numbers, differences in study design, and differences in computing the expected number of events. For example, in the report from the Oxford
Cancer in twins
Survey of Childhood Cancers, 2 most often cited as showing reduced cancer risks for twins, very complex adjustments were made in computing the expected numbers. PY of survival were not available for all of the twins, and it was necessary to make assumptions about the percent of twins in the study population, the infant mortality rates of twins and singletons, the percent of twins and singletons discharged alive from hospital who were X-rayed in utero, and the risk associated with this X-ray exposure. The extent to which the estimation of these many parameters influenced study results is unclear. Particularly sensitive is the assumption regarding cancer risks associated with the additional X-ray exposures received by the twins. Applying different radiation risk estimates would affect strongly the estimates of the expected number of childhood cancers. 22 Another study based on twin registry data from California 3,4also reported a 10 percent deficit of childhood cancer based on 100 observed deaths. The O/E for all-cancers at ages 0-4 was 0.9 for both males and females, but no findings were statistically significant. Because special case-finding methods were required to locate cancers in children who had moved outside of the state, incomplete ascertainment of cases could have influenced the study results. Even population-based cancer-registry studies of twins are not without problems. A recent study from Norway found a relative risk for childhood cancers among twins of 0.9 (CI = 0.5-1.5) but the finding was based on only 14 incident cancers. There was a statistically significant excess of renal cancer, but the small numbers preclude meaningful interpretation? Another recent population-based study in Connecticut 6 showed a cancer deficit, which was significant among males, but vital status was not determined for individual twins. Mortality was estimated from United States population rates, and emigration from the state was assumed to be 10 percent. Despite these limitations, it is notable that the deficit of cancer incidence in the Connecticut study was particularly pronounced (O/ E = 0.2) among males under age five years, as in our series. Our study addresses directly most of the deficiencies of previous investigations. Cancer incident-cases and deaths among twins were identified through the same mechanisms in the same population as provided the incidence and mortality rates used to compute the expected numbers. Cancer registration in Sweden is nearly complete and the record-linkage procedure based on personal identification numbers makes it unlikely that many childhood cancers would have been missed. This work differs from two earlier evaluations of cancer in Swedish twins 1°,11in that (i) it is based on
PY at risk, (ii) it is restricted to cancer occurring up to and including age 15, and (iii) it includes cancer incidence. If childhood cancer in twins occurs less frequently than among singletons, the risk difference appears to be small. Although methodologic artifacts might account for some, if not all, of the deficits seen in previous studies, it is interesting that most studies, including ours, find lower than expected rates of childhood cancer and not one finds an overall excess. If the deficits are real, the reasons why twins might be at slightly lower risk for cancer development are not known. They would likely reflect a combination of factors, including low birthweight, competing causes of death or fetal mortality, and sensitivity to environmental carcinogens (see Inskip et al6 for more detailed discussion), although the findings of investigations addressing these factors are not consistent. For example, high birthweight has been associated with increased risk of childhood cancer in some studies23but not in others. 24 It is of some interest, then, that despite the increased frequency of prenatal X-ray exposure among twins, no excess of cancer has been reported in any cohort study of twins. Increased cancer risks would not be expected in the oldest cohorts, born before the time of diagnostic X-rays, such as studied by Jarvik and Falek, 1nor in the youngest cohorts, born after improved equipment lowered exposures and ultrasonic imaging came into use. Some of the cohort studies, however, include subjects born at a time when more than a third of twin pregnancies underwent diagnostic radiography involving appreciable doses, and a slight excess risk might be expected among them. In fact, three casecontrol studies 16,17,25of childhood cancer in twins for whom exposure was ascertained have related prenatal X-rays with increased risks. Because of the relatively low exposure frequency among twins, some 'selection' could occur, but, to date, no selection factors have been identified that are associated both with prenatal X-ray and the risk of childhood cancer. The overall X-ray exposure frequency of 36 percent among Swedish twins, 17 significantly limited our ability to detect a radiation effect, assuming that one existed. Based on prenatal X-ray studies of single births, a 36 percent exposure proportion (not 100 percent) would result in eight radiogenetic childhood cancers in our population of 35,000 twins. This number of excess cancers would lead to an increase of only 14 percent (RR = 1.14). Clearly, this value is too low to detect even in a study of the size we conducted. Thus, cohort studies of twins may not be able to address meaningfully the question of prenatal X-ray carcinogenicity because of the relatively low exposure proportions and associated low statistical power. Cancer Causes and Control. Vol 3. 1992
531
Y. Rodvall et al A c k n o w l e d g e m e n t s - - W e thank Ms A. C. Wistedt for help with the data collection and D r N. L. Pedersen for valuable help with the Twin Register.
References 1. Jarvik LF, Falek A. Cancer rates in aging twins. Am J Hum Genet 1961; 13: 413-22. 2. Hewitt D, LashofJC, Stewart AM. Childhood cancer in twins. Cancer 1966; 19: 157-61. 3. Jackson EW, Norris FD, Klauber MR. Childhood leukemia in California-born twins. Cancer 1969; 23:913-9. 4. Norris FD, Jackson EW. Childhood cancer deaths in California-born twins. A further report on types of cancer found. Cancer 1970; 25: 212-8. 5. Kaprio J, Teppo L, Koskenvuo M, Pukkala E. Cancer in adult same-sexed twins: A historical cohort study. Prog Clin Biol Res 1981; 69: 217-23. 6. Inskip PD, Harvey EB, Boice JD Jr, et aL Incidence of childhood cancer in twins. Cancer Causes Control 1991; 2: 315-24. 7. Harvald B, Hauge M. Heredity of cancer elucidated by a studyof unselected twins. JAMA 1963; 186: 749-53. 8. Iversen T. Leukaemia in children. Acta Pediatr Stockh (Suppl) 1965; 159: 161-2. 9. Windham GC, Bjerkedal T, Langmark F. A populationbased study of cancer incidence in twins and in children with congenital malformation or low birth weight, Norway, 1967-1980. AmJ Epidemio11985; 121: 49-56. 10. Cederl6f R, Floderus B, Friberg L. Cancer in MZ and DZ twins. Acta Genet Med Gemello11970; 19: 69-74. 11. Cederl6f R, Floderus-Myrhed B. Cancer mortality and morbidity among 23,000 unselected twin pairs: In: Gelboin HV, MacMahon B, Matsushima T, Sugimura T, Takayama S, Takebe H, eds. Geneticand Environmental Factors in Experimental and Human Cancer. Tokyo: Japan Scientific Societies Press, 1980; 151-60.
532
CancerCausesand Control.Vol3. 1992
12. Stewart A, Kneale GW. Changes in the cancer risk associated with obstetric radiography. Lancet 1968; 1: 104-7. 13. MacMahon B. Prenatal X-ray exposure and childhood cancer. JNCI 1962; 128: 1173-91. 14. Graham S, Levin ML, Lilienfeld AM, Schuman LM, Gibson R, Dowd JE, Hempelmann L. Preconception, Intrauterine, and Postnatal Irradiation as Related to Leukemia. NCI Monogr 1966; 19: 347-71. 15. Monson RR, MacMahon B. Prenatal X-ray exposure and cancer in children. In: JD Boice Jr, JF Fraumeni Jr, eds.
Radiation Carcinogenesis. Epidemiology and Biological Significance. New York: Raven Press, 1984; 97-105. 16. Harvey EB, Boice JD Jr, Honeyman M, Flannery JT. Prenatal X-ray exposure and childhood cancer in twins. N EngIJ Ailed 1985; 312: 541-5. 17. Rodvall Y, Pershagen G, Hrubec Z, Ahlbom A, Pedersen NL, Boice JD. Prenatal X-ray exposure and childhood cancer in Swedish twins. IntJ Cancer 1990; 46: 362-5. 18. Medlund P, Cederl6f R, Floderus-Myrhed B, Friberg L, S6rensen S. A new Swedish twin registry. Acta Med Scand Suppl. 600, 1976; 1-109. 19. World Health Organization. International Classification of Diseases, Seventh Revision. Geneva: WHO, 1957. 20. World Health Organization. International Classification of Diseases, Eighth Revision. Geneva: WHO, 1967. 21. Rothman KJ, Boice JD. Epidemiologic Analysis with a Programmable Calculator. Boston, MA: Epidemiology Resources Inc., 1982. 22. Mole RH. Childhood cancer after prenatal exposure to diagnostic X-ray examinations in Britain. Br J Cancer 1990; 62; 152-68. 23. Greenberg RS, Shuster JL Jr. Epidemiology of cancer in children. Epidemiol Rev 1985; 7: 22-48. 24. Forsberg J-G, K~illenB. Pregnancy and delivery characteristics of women whose infants develop child cancer. APMIS 1990; 98: 37-42. 25. Mole RH. Antenatal irradiation and childhood cancer. Causation or coincidence. BrJ Cancer 1974; 30:199-208.
Childhood cancer among Swedish twins
Ylva Rodvall, Zdenek Hrubec, Giiran Pershagen, Anders Ahlbom, Arne Bjurman, and John D. Boice, Jr (Received 25June 1992;accepted29July 1992) The risk of childhood cancer was evaluated among 35,582 twins born in Sweden between 1952 and 1967. Cancers were identified through linkage with national cancer and mortality registries. Overall, 59 childhood cancers and 41 deaths from cancer occurred before the age of 16. For both sexes combined, the cancer incidence was similar to that in the general population of primarily single-born children (number observed/ number expected [O/E] -- 1.0, 95 percent confidence interval [CI] = 0.7-1.2). For males under age five, cancer incidence was reduced significantly (O/E = 0.3, CI = 0.1-0.7). There was a substantial increase in all-cause mortality among twins (O/E = 3.7), attributable to a high mortality during the first year of life. Excluding this first year, the O/E for death (all causes) was 1.1 (CI = 1.0-1.3). For cancer mortality of both sexes, the O/E was 0.9 (CI = 0.6-1.2), with no significant reduction of risk in any sex-age group. We conclude that with the possible exception of males aged 0-4 years, the childhood cancer risk of twins appears similar to that of singletons.
Key words:Cancer incidence, childhood cancer, mortality, twins, Sweden.
Introduction Twins, because of their relatively low weight at birth, die at a much higher rate during the first year of life than single-born children. This high mortality rate has made it difficult to evaluate accurately the risk of childhood cancer in twin populations. Although a number of studies suggest that cancer rates among twins may be low, 1-6the findings are not consistent. Several studies from Scandinavia, for example, report cancer occurrence in twins to be similar to that of the general population. 7-1~ Interestingly, no study has indicated an increase in cancer among twins. The latter observation is noteworthy since, in the past, twins were exposed more frequently to prenatal X-ray than singletons.
Such X-ray exposure has been linked to increases in childhood cancer, both among singletons 12-15 and a m o n g t w i n s Y 6d7
To provide information on the risk of childhood cancer among twins, we examined cancer incidence and mortality in a register of all twin births in Sweden from 1952 through 1967.
Materials and Methods The Swedish Twin Register (STR) is comprised of all twins born in Sweden during 1926-67. is A roster of 35,582 twins born alive between 1952-67 was linked to
Drs Rodvall, Pershagen,Ahlbom, and Bjurman are with the Department of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden. Drs Hrubec and Boice are with the Radiation Epidemiology Branch, Division of Cancer Etiology, National Cancer Institute, Bethesda, Maryland, USA. Dr Pershagenis also at the Department of Environmental Health and Infectious Diseases Control, Karolinska Hospital, Stockholm, Sweden. Dr Ahlbom is also at the Department of Occupational Health, Epidemiology Unit, Karolinska Hospital, Stockholm, Sweden. Address correspondenceto Dr Rodvall, Institute of Environmental Medicine, Box 60208, S-10401 Stockholm, Sweden. This study was supported in part by the National Cancer Institute contract NO1-CP-51033 and in part by the MacArthur Foundation. © 1992 Rapid Communications of Oxford Ltd
Cancer Causes and Control. Vol 3. 1992
527
Y. Rodvall et al the Swedish Cause-of-Death Register, which has been computerized since 1952, and to the Cancer Register, which started in 1958. Since 1947, every individual in Sweden has been assigned a unique civil-registration number which was used in the linkage procedures. For 1,397 twins who did not have a civil registration number recorded in the STR, records in the Parish Office Registers were searched to determine whether they were stillborn. Surviving twins without civil registration numbers were linked to the cancer and mortality registers based on name and birth date. The twins were followed up through age 15. Diagnostic criteria and coding practices have changed over the observation period. The Swedish Cancer Register used the seventh edition of the International Classification of Diseases (ICD-7) 19throughout, while the Cause-of-Death Register used the ICD-82° from 1969 to 1983. In our study, the ICD-7 was used for the cause-of-death data since ICD-7 codes cannot be translated completely into the ICD-8 classification. Standardized ratios of observed to expected incident cases or deaths (O/E) were employed to compare the cancer experience of the twin cohort with that of a comparable Swedish population estimated from national rates specific for sex, age, and calendar time. Approximate 95 percent confidence intervals (CI) were computed except when the observed number of cancers was 20 or fewer. In these instances, exact confidence intervals were calculated based on the Poisson distribution? 1 Person-years (PY) at risk were computed from the date of birth until the date of death or through age 15. There were no twins with more than one cancer. Twins dying within the first month were assumed to have lived for only four days, the average survival for children dying within their first month of life. Those dying between I and 12 months were assumed to have lived 15 days during the month of death. PY at risk for individuals diagnosed with cancer were counted to the date of death or age 15, whichever came first. Twins born before 1952 could not be followed accurately for survival and were excluded. Our previous case-control study was able to include more calendar years and many more twin births (83,316) because it did not require determining individual survival for the entire twin population. ~7 Expected numbers were calculated by multiplying incidence and mortality rates for the general population by the numbers of PY at risk, stratified by age and calendar year in one-year categories, and by sex. Rates for the general population were based on the mean population for each year and age, including the first year. 528
Cancer Causes and Control. Vol 3. 1992
T a b l e 1. N u m b e r o f cancers r e p o r t e d to t h e C a n c e r R e g i s t r y and t h e C a u s e - o f - D e a t h Registry, b y sex, for S w e d i s h t w i n s b o r n 1952-67 and f o l l o w e d u p to age 16 years Cancer Registry (1958-83)
Leukemia (ICD-7 # 204) Acute lymphatic leukemia Myeloid leukemia Other and unspecified leukemia Total, leukemia Tumors of the central nervous system (CNS) (ICD-7 # 193.0-193.2, 195.3) Brain Intraspinal, meninges Total, CNS Other cancer Lymphomas and other reticuloendothelial neoplasms Retinoblastoma and nervous system tumors other than CNS Bone tumor Other and unspecified malignancies Total, other cancer Total, all cancers
Cause-of-Death Registry (1952-83)
Male
Female
Male
Female
5 2
2 1
2 1
5 --
2 9
5 8
7 10
4 9
8 -8
11 2 13
5 -5
5 2 7
5
--
3
--
-1
6 2
---
-3
4~ 10
3b 11
27
32
2c 5 20
2a 5 21
ICD-code • Includes: appendix maxillary sinus renal parenchyma site unknown
153.4 160.2 180.0 199.9
b Includes: maxillary sinus renal parenchyma site unknown
160.2 180.0 199.9
c Includes: peritoneum suprarenal gland
158.9 195.0
d Includes: suprarenal gland connective tissue (unspecified)
195.0 197.9
Results Fifty-nine childhood cancer cases and 41 childhood cancer deaths occurred before age 16. Of the 41 deaths, 12 were not recorded in the Cancer Register (seven occurred before 1958, two were classified erroneously as leukemia deaths, and three were missed). The 59 incident cases of cancer included 21 tumors of the central nervous system (CNS) and 17 leukemias (Table 1). There was only one twin-pair concordant for can-
Cancer in twins
Table 2. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site and sex Males
Females
O
E
O/E
CI
O
Total
E
O/E
CI
O
E
O/E
CI
Leukemia Tumors of the central nervous system Other cancers
9
9.7
0.93
0.4-1.8
8
8.4
0.95
0.4-1.9
17
18.1
0.94
0.5-1.5
8 10
8.4 15.1
0.95 0.66
0.4-1.9 0.3-1.2
13 11
7.6 12.5
1.71 0.88
0.9-2.9 0.4-1.6
21 21
16.0 27.6
1.31 0.76
0.8-2.0 0.5-1.2
All cancers
27
33.2
0.81
0.5-1.2
32
28.5
1.12
0.8-1.6
59
61.7
0.96
0.7-1.2
Table 3. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site, sex, and calendar year of diagnosis Calendar year
Males O
1958-70 Leukemia Tumors of the nervous system All cancers
Females
E
O/E
CI
6
7.1
0.85
0.3-1.8
5 16
5.6 22.9
0.89 0.70
3
2.6
3 11
2.8 10.3
O
Total
E
O/E
CI
O
E
O/E
CI
6
6.4
0.94
0.3-2.0
12
13.5
0.89
0.5-1.6
0.3-2.1 0.4-1.1
9 24
5.2 19.8
1.73 1.21
0.8-3.3 0.8-1.8
14 40
10.8 42.7
1.30 0.94
0.7-2.2 0.7-1.3
1.15
0.2-3.4
2
2.0
1.00
0.1-3.7
5
4.6
1.09
0.4-2.6
1.07 1.07
0.2-3.2 0.5-1.9
4 8
2.5 8.7
1.60 0.92
0.4-4.2 0.4-1.8
7 19
5.2 19.1
1.35 0.99
0.5-2.7 0.6-1.6
central
1971-83 Leukemia Tumors of the central nervous system Allcancers
Table 4. Observed (O) and expected (E) incident cancers, O / E and 95% confidence interval (CI) on O / E by site, sex, and by age at diagnosis Males
0-4 Years Leukemia Tumors of the central nervous system All cancers 5-9 Years Leukemia Tumors of the central nervous system Allcancers 10-15 Years Leukemia Tumors of the central nervous system All cancers
Females
Total
O
E
O/E
CI
O
E
O/E
CI
2
4.3
0.47
0.1-1.7
5
3.7
1.35
0.4-3.2
1 3
2.4 12.2
0.42 0.25
0.1-2.3 0.1-0.7
3 11
2.1 10.5
1.43 1.05
2
2.8
0.7
0.1-2.6
1
2.6
0.38
4 8
2.6 9.0
1.54 0.89
0.4-4.0 0.4-1.8
5 8
5
2.7
1.85
0.6-4.4
3 16
3.4 12.0
0.88 1.33
0.2-2.6 0.8-2.2
cer; o n e t w i n had l e u k e m i a a n d t h e o t h e r h a d b r a i n c a n cer. T h e r e w e r e n o c o n c o r d a n t l e u k e m i a pairs. O v e r a l l , c a n c e r i n c i d e n c e w a s s i m i l a r to t h a t e x p e c t e d in t h e g e n e r a l p o p u l a t i o n ( T a b l e 2), a l t h o u g h a d e f i c i t w a s s u g g e s t e d a m o n g m a l e s ( O / E = 0.8). F o r f e m a l e s , t h e r e was a s l i g h t l y i n c r e a s e d risk f o r t u m o r s o f t h e C N S ( O / E -- 1.7). R i s k o f l e u k e m i a was s l i g h t l y
O
E
O/E
CI
7
7.9
0.89
0.4-1.8
0.3-4.2 0.5-1.9
4 14
4.5 22.7
0.89 0.62
0.2-2.3 0.3-1.0
0.0-2.2
3
5.3
0.57
0.1-1.6
2.2 7.5
2.27 0.7-5.2 1.07 0.5-2.1
9 16
4.8 16.5
1.88 0.97
0.9-3.5 0.6-1.6
2
2.1
0.95
0.1-3.4
7
4.8
1.46
0.6-3.0
5 13
3.3 10.5
1.52 1.24
0.5-3.5 0.7-2.1
8 29
6.7 22.6
1.19 1,28
0.5-2.3 0.9-1.8
d e c r e a s e d ( O / E = 0.9). T h e d e f i c i t o f c a n c e r a m o n g m a l e s was a p p a r e n t o n l y in t h e y e a r s 1958-70 ( T a b l e 3). F o r m a l e t w i n s b e t w e e n 0 and 4 y e a r s at d i a g n o s i s , c a n c e r risk was s i g n i f i c a n t l y l o w ( O / E = 0.3, C I = 0.10.7). A m o n g f e m a l e s in this age g r o u p , t h e O / E was 1.1 ( T a b l e 4). T w i n s w e r e at e x c e p t i o n a l l y h i g h risk o f d e a t h f r o m Cancer Causes and Control. Vol 3. 1992
529
Y. Rodvall et al Table 5. O b s e r v e d (O) and expected (E) all-cause mortality, O / E and 95% confidence interval (CI) on O / E b y site and sex Males
All deaths Deaths (0-1 years) Deaths (1-15 years) All cancers Leukemia Tumors of the central nervous system
Females
Total
O
E
O/E
CI
O
E
O/E
CI
O
E
O/E
CI
1,697 1,546 151
463.0 324.3 138.7
3.67 4.77 1.09
3.5-3.8 4.5-5.0 0.9-1.3
1,229 1,119 110
326.7 232.3 94.1
3.76 4.82 1.17
3.6-4.0 4.5-5.1 1.0-1.4
2,926 2,665 261
789.7 556.8 232.9
3.71 4.79 1.12
3.6-3.8 4.6-5.0 1.0-1.3
20 10
25.1 12.0
0.80 0.83
0.5-1.2 0.4-1.5
21 9
20.5 9.6
1.02 0.94
0.6-1.6 0.4-1.8
41 19
45.6 21.6
0.90 0.88
0.6-1.2 0.5-1.4
5
4.6
1.09
0.4-2.6
7
4.2
1.67
0.7-3.4
12
8.8
1.36
0.7-2.4
Table 6. Observed (O) and expected (E) mortality, O/E and 95% confidenceinterval (CI) on O/E by site, sex, and age at death Males
0-4 Years All deaths All cancers Leukemia Tumors of the central nervous system 5-9 Years All deaths All cancers Leukemia Tumors of the central nervous system 10-15 Years All deaths All cancers Leukemia Tumors of the central nervous system
O
E
O/E
1,624 7 4
385.6 10.5 5.0
3
1.8
1.67
36 7 3
39.5 7.3 4.1
2
Females CI
O
E
O/E
CI
O
E
1,173 7 4
276.1 8.9 3.8
4.25 0.79 1.05
4.0-4.5 0.3-1.6 0.3-2.7
2,797 14 8
661.7 19.4 8.8
0.3-4.8
2
1.6
1.25
0.1-4.5
5
3.4
1.47
0.5-3.4
0.91 0.96 0.73
0.6-1.3 0.4-2.0 0.2-2.1
27 8 4
26.0 6.7 3.7
1.04 1.19 1.08
0.7-1.5 0.5~2.4 0.3-2.8
63 15 7
65.5 14.0 7.8
0.96 1.07 0.90
0.7-1.2 0.6-1.8 0.4-1.9
1.3
1.54
0.2-5.5
2
1.2
1.67
0.2-5.9
4
2.5
1.60
0.4-4.0
37 6 3
37.9 7.2 2.9
0.98 0.83 1.03
0.7-1.3 0.3-1.8 0.2-3.0
29 6 1
24.6 5.0 2.1
1.18 1.20 0.48
0.8-1.7 0.4-2.6 0.0-2.7
66
62.5
12
12.2
4
5.0
1.06 0.8-1.3 0.98 0.5-1.7 0 . 8 0 0.2-2.1
0
1.4
0.00
0.0-2.6
3
1.4
2.14
0.4-6.3
3
2.8
1.07
4.21 4.0-4.4 0.67 0.3-1.4 0 . 8 0 0.2-2.0
all causes (O/E = 3.7), attributable almost entirely to a high mortality in the first year of life. Over 91 percent of the 2,926 deaths occurred prior to age one year. There were no differences by sex, overall or in the first year. The data also were analyzed excluding deaths and PY in the first year of life. The O/Es for cancer deaths were the same as in the general population (Table 5). There were no significant differences in the O/E value for cancer by age at death, although they were somewhat lower for twins under age five than at older ages (Table 6).
Discussion Compared with the general population, twins born in Sweden during 1952-67 had a nearly fivefold risk of 530
Total
CancerCausesand Control Vol 3. 1992
O/E
CI
4 . 2 3 4.1-4.4 0.72 0.4-1.2 0.91 0.4-1.8
0.2-3.1
dying within the first year of life. The reasons for this high mortality are likely to be factors associated with premature birth and low birthweight common among twin pregnancies. When followed to age 16, those surviving the first year experienced a small (12 percent) increased risk of dying. Overall, the risk of cancer (O/ E = 0.96) was similar among twins compared with the general population of primarily single-born children. Both leukemia and other neoplasms occurred slightly below expectation. The only significant finding was a deficit of cancer incidence among males under age five. Most, but not all, studies of twins find a lower rate of childhood cancer compared with singletons.T M Inconsistencies reflect, in all likelihood, the play of chance due to small numbers, differences in study design, and differences in computing the expected number of events. For example, in the report from the Oxford
Cancer in twins
Survey of Childhood Cancers, 2 most often cited as showing reduced cancer risks for twins, very complex adjustments were made in computing the expected numbers. PY of survival were not available for all of the twins, and it was necessary to make assumptions about the percent of twins in the study population, the infant mortality rates of twins and singletons, the percent of twins and singletons discharged alive from hospital who were X-rayed in utero, and the risk associated with this X-ray exposure. The extent to which the estimation of these many parameters influenced study results is unclear. Particularly sensitive is the assumption regarding cancer risks associated with the additional X-ray exposures received by the twins. Applying different radiation risk estimates would affect strongly the estimates of the expected number of childhood cancers. 22 Another study based on twin registry data from California 3,4also reported a 10 percent deficit of childhood cancer based on 100 observed deaths. The O/E for all-cancers at ages 0-4 was 0.9 for both males and females, but no findings were statistically significant. Because special case-finding methods were required to locate cancers in children who had moved outside of the state, incomplete ascertainment of cases could have influenced the study results. Even population-based cancer-registry studies of twins are not without problems. A recent study from Norway found a relative risk for childhood cancers among twins of 0.9 (CI = 0.5-1.5) but the finding was based on only 14 incident cancers. There was a statistically significant excess of renal cancer, but the small numbers preclude meaningful interpretation? Another recent population-based study in Connecticut 6 showed a cancer deficit, which was significant among males, but vital status was not determined for individual twins. Mortality was estimated from United States population rates, and emigration from the state was assumed to be 10 percent. Despite these limitations, it is notable that the deficit of cancer incidence in the Connecticut study was particularly pronounced (O/ E = 0.2) among males under age five years, as in our series. Our study addresses directly most of the deficiencies of previous investigations. Cancer incident-cases and deaths among twins were identified through the same mechanisms in the same population as provided the incidence and mortality rates used to compute the expected numbers. Cancer registration in Sweden is nearly complete and the record-linkage procedure based on personal identification numbers makes it unlikely that many childhood cancers would have been missed. This work differs from two earlier evaluations of cancer in Swedish twins 1°,11in that (i) it is based on
PY at risk, (ii) it is restricted to cancer occurring up to and including age 15, and (iii) it includes cancer incidence. If childhood cancer in twins occurs less frequently than among singletons, the risk difference appears to be small. Although methodologic artifacts might account for some, if not all, of the deficits seen in previous studies, it is interesting that most studies, including ours, find lower than expected rates of childhood cancer and not one finds an overall excess. If the deficits are real, the reasons why twins might be at slightly lower risk for cancer development are not known. They would likely reflect a combination of factors, including low birthweight, competing causes of death or fetal mortality, and sensitivity to environmental carcinogens (see Inskip et al6 for more detailed discussion), although the findings of investigations addressing these factors are not consistent. For example, high birthweight has been associated with increased risk of childhood cancer in some studies23but not in others. 24 It is of some interest, then, that despite the increased frequency of prenatal X-ray exposure among twins, no excess of cancer has been reported in any cohort study of twins. Increased cancer risks would not be expected in the oldest cohorts, born before the time of diagnostic X-rays, such as studied by Jarvik and Falek, 1nor in the youngest cohorts, born after improved equipment lowered exposures and ultrasonic imaging came into use. Some of the cohort studies, however, include subjects born at a time when more than a third of twin pregnancies underwent diagnostic radiography involving appreciable doses, and a slight excess risk might be expected among them. In fact, three casecontrol studies 16,17,25of childhood cancer in twins for whom exposure was ascertained have related prenatal X-rays with increased risks. Because of the relatively low exposure frequency among twins, some 'selection' could occur, but, to date, no selection factors have been identified that are associated both with prenatal X-ray and the risk of childhood cancer. The overall X-ray exposure frequency of 36 percent among Swedish twins, 17 significantly limited our ability to detect a radiation effect, assuming that one existed. Based on prenatal X-ray studies of single births, a 36 percent exposure proportion (not 100 percent) would result in eight radiogenetic childhood cancers in our population of 35,000 twins. This number of excess cancers would lead to an increase of only 14 percent (RR = 1.14). Clearly, this value is too low to detect even in a study of the size we conducted. Thus, cohort studies of twins may not be able to address meaningfully the question of prenatal X-ray carcinogenicity because of the relatively low exposure proportions and associated low statistical power. Cancer Causes and Control. Vol 3. 1992
531
Y. Rodvall et al A c k n o w l e d g e m e n t s - - W e thank Ms A. C. Wistedt for help with the data collection and D r N. L. Pedersen for valuable help with the Twin Register.
References 1. Jarvik LF, Falek A. Cancer rates in aging twins. Am J Hum Genet 1961; 13: 413-22. 2. Hewitt D, LashofJC, Stewart AM. Childhood cancer in twins. Cancer 1966; 19: 157-61. 3. Jackson EW, Norris FD, Klauber MR. Childhood leukemia in California-born twins. Cancer 1969; 23:913-9. 4. Norris FD, Jackson EW. Childhood cancer deaths in California-born twins. A further report on types of cancer found. Cancer 1970; 25: 212-8. 5. Kaprio J, Teppo L, Koskenvuo M, Pukkala E. Cancer in adult same-sexed twins: A historical cohort study. Prog Clin Biol Res 1981; 69: 217-23. 6. Inskip PD, Harvey EB, Boice JD Jr, et aL Incidence of childhood cancer in twins. Cancer Causes Control 1991; 2: 315-24. 7. Harvald B, Hauge M. Heredity of cancer elucidated by a studyof unselected twins. JAMA 1963; 186: 749-53. 8. Iversen T. Leukaemia in children. Acta Pediatr Stockh (Suppl) 1965; 159: 161-2. 9. Windham GC, Bjerkedal T, Langmark F. A populationbased study of cancer incidence in twins and in children with congenital malformation or low birth weight, Norway, 1967-1980. AmJ Epidemio11985; 121: 49-56. 10. Cederl6f R, Floderus B, Friberg L. Cancer in MZ and DZ twins. Acta Genet Med Gemello11970; 19: 69-74. 11. Cederl6f R, Floderus-Myrhed B. Cancer mortality and morbidity among 23,000 unselected twin pairs: In: Gelboin HV, MacMahon B, Matsushima T, Sugimura T, Takayama S, Takebe H, eds. Geneticand Environmental Factors in Experimental and Human Cancer. Tokyo: Japan Scientific Societies Press, 1980; 151-60.
532
CancerCausesand Control.Vol3. 1992
12. Stewart A, Kneale GW. Changes in the cancer risk associated with obstetric radiography. Lancet 1968; 1: 104-7. 13. MacMahon B. Prenatal X-ray exposure and childhood cancer. JNCI 1962; 128: 1173-91. 14. Graham S, Levin ML, Lilienfeld AM, Schuman LM, Gibson R, Dowd JE, Hempelmann L. Preconception, Intrauterine, and Postnatal Irradiation as Related to Leukemia. NCI Monogr 1966; 19: 347-71. 15. Monson RR, MacMahon B. Prenatal X-ray exposure and cancer in children. In: JD Boice Jr, JF Fraumeni Jr, eds.
Radiation Carcinogenesis. Epidemiology and Biological Significance. New York: Raven Press, 1984; 97-105. 16. Harvey EB, Boice JD Jr, Honeyman M, Flannery JT. Prenatal X-ray exposure and childhood cancer in twins. N EngIJ Ailed 1985; 312: 541-5. 17. Rodvall Y, Pershagen G, Hrubec Z, Ahlbom A, Pedersen NL, Boice JD. Prenatal X-ray exposure and childhood cancer in Swedish twins. IntJ Cancer 1990; 46: 362-5. 18. Medlund P, Cederl6f R, Floderus-Myrhed B, Friberg L, S6rensen S. A new Swedish twin registry. Acta Med Scand Suppl. 600, 1976; 1-109. 19. World Health Organization. International Classification of Diseases, Seventh Revision. Geneva: WHO, 1957. 20. World Health Organization. International Classification of Diseases, Eighth Revision. Geneva: WHO, 1967. 21. Rothman KJ, Boice JD. Epidemiologic Analysis with a Programmable Calculator. Boston, MA: Epidemiology Resources Inc., 1982. 22. Mole RH. Childhood cancer after prenatal exposure to diagnostic X-ray examinations in Britain. Br J Cancer 1990; 62; 152-68. 23. Greenberg RS, Shuster JL Jr. Epidemiology of cancer in children. Epidemiol Rev 1985; 7: 22-48. 24. Forsberg J-G, K~illenB. Pregnancy and delivery characteristics of women whose infants develop child cancer. APMIS 1990; 98: 37-42. 25. Mole RH. Antenatal irradiation and childhood cancer. Causation or coincidence. BrJ Cancer 1974; 30:199-208.
