of radiation to which pregnant women have been exposed during examinations (Doll and Wakeford 1997). A decrease over time also was reported in the northeastern U.S. study (Monson and MacMahon 1984).
The possible effect of prenatal exposure has been studied in a number of other populations in the United States and Europe. Results of the case-control studies have been combined in meta-analyses by Bithell (1989, 1990). Although dominated by the OSCC, results of these studies show a significant RR of 1.4 for in utero radiation in association with childhood cancer (Doll and Wakeford 1997).
Controversy continues, however, about the existence and size of the risk following prenatal exposure. Boice and Miller (1999) noted that the increases were restricted to case-control studies and were not seen in cohort studies; they also commented on the similarity of relative risks for leukemia and solid cancers, suggesting an underlying bias in the case-control studies. In their review, Doll and Wakeford (1997) discuss these arguments. In regard to cohort studies, they combine the results of cohort studies for which relative risks can be calculated reliably and note that, when the atomic bomb survivors are excluded, an increased risk is obtained that is consistent with the combined results of case-control studies. They note further that the incomplete follow-up of the Japanese atomic bomb survivor cohort in the years after the bombings may be partially responsible for the apparent inconsistency of results concerning the effects of prenatal exposures. The argument that radiation risks for leukemia and solid cancers differ is based on observations of exposure in childhood and later years. Doll and Wakeford (1997) note that the carcinogenic effects of radiation exposure in utero and in childhood are not expected to be the same because the cells that give risk to most of the typical childhood cancers other than leukemia persist and are capable of dividing for only a short time, if at all, after birth. Doll and Wakeford further conclude that the idea of a causal relationship is supported by the increase in RR with increasing number of X-ray examinations conducted in the third trimester of pregnancy and the significant decline in RR with year of birth, paralleling the decline in fetal doses that occurred over the same period (UNSCEAR 1972).
Based on the results of the Oxford survey and other studies of the effects of maternal irradiation, UNSCEAR (1996) reported a statistically significant leukemia risk (up to age 15 years) and estimated a 40% increase in risk of childhood cancers (up to 15 years) at doses of 10–20 mGy (low LET). Risk estimates have been derived since then by a number of authors and committees (UNSCEAR 1996; Doll and Wakeford 1997; Wakeford and Little 2003). In the most recent analyses, Wakeford and Little (2003) derive an ERR for childhood cancer following prenatal exposure of about 50 Gy−1, with an EAR of about 8% Gy−1. They comment, however, that statistical, dosimetric, modeling, and other uncertainties associated with these risk estimates are appreciable. They also note that when these uncertainties and those associated with equivalent risk coefficients from the Japanese atomic bomb survivor cohort exposed in utero are taken into account, the risk estimates for childhood cancer from these two sources of data are compatible and they conclude that “doses to the fetus in utero of the order of 10 mSv discernibly increase the risk of childhood cancer.”
The use of 131I for diagnostic purposes in childhood is rare; hence information on risk is very sparse. In the cohort of 34,104 patients who had received 131I diagnostic exposures for suspected thyroid disorders in Sweden between 1951 and 1969, reported by Hall and colleagues (1996), only 2408 patients were under age 20 at the time of the examination. Among these, a small excess risk was seen (3 cases observed vs. 1.8 expected).
Information on radiation risks following diagnostic radiation exposure in childhood comes from a study of women who received multiple diagnostic X-rays for the evaluation of scoliosis during childhood and adolescence. This study, in which important efforts were made to reconstruct dose to the breast, has provided an estimate of the risk of radiationinduced breast cancer. This estimate is reviewed, and compared with risk estimates derived from other medical exposure studies, in the following section.
Studies of prenatal exposure to diagnostic X-rays have, despite long-standing controversy, provided important information on the existence of a significantly increased risk of leukemia and childhood cancer following diagnostic doses of 10–20 mGy in utero.
Only one study has examined the effects of using 131I for diagnostic purposes in childhood. A small excess of thyroid cancer risk was seen—based on very small numbers—and no risk estimate is provided.
This section reviews radiation risk estimates for five types of malignancies (lung cancer, female breast cancer, thyroid cancer, leukemia, and stomach cancer). The results of analyses of the risk of heart disease following medical radiation exposures are also reviewed.
Lung cancer was chosen because it is the most common malignancy among humans. Breast cancer was selected because breast tissue in young women is responsive to low levels of low-LET radiation, and thyroid cancer was chosen because of the inherent radiosensitivity of the thyroid gland. Leukemia was chosen because bone marrow is sensitive to low levels of low-LET radiation, and stomach cancer was selected because of its high incidence in many parts of the