Radiation Risks of Mammography

The risks associated with routine mammographic screening which have received the most attention are those concerned with the possible induction of breast cancer by the low-energy radiation associated with mammography, and these are the risks discussed in detail in this Section. However, it must be remembered that there are other and likely more important costs of mammography including the psychological and physical (due to surgical intervention) effects on women with FP diagnoses (Feig, 2004), and the very substantial resource implications for the health care system of a mammography program (Lindfors and Rosenquist, 1995; 2001; Rosenquist and Lindfors, 1998). These latter costs are very difficult to quantify in a cost-benefit analysis, and are not considered further in this Section.

A number of epidemiologic studies of adult women have contributed knowledge of the long-term risks of ionizing radiation to the female breast (Boice, 2001; Preston et al., 2002a; UNSCEAR, 2000). Among these studies are those of Japanese atomic-bomb survivors (Shimizu et al., 1990; Thompson et al., 1994); female tuberculosis patients in Massachusetts who received multiple chest fluoroscopies in conjunction with artificial pneumothorax (Boice et al., 1991); a similar series of female tuberculosis patients in Canada (Howe and McLaughlin, 1996); women in New York State receiving radiotherapy for postpartum mastitis (Shore et al., 1986); and Swedish women receiving x-ray treatment for fibroadenomatosis and other benign breast conditions (Baral et al., 1977; Mattsson et al., 1993).

7.2.1 Factors Defining Breast Cancer Risk

The key findings from the above studies can be summarized as follows:

• The great majority of studies demonstrated increased incidence or mortality from breast cancer following irradiation.

• A linear dose-response function generally provides a reasonable fit to the data, though for some studies it is not possible to exclude the possibility of a linear-quadratic relationship. Using a linear model to fit the data from high dose studies to predict risk from the low doses employed in mammogra-phy is a conservative approach (if the quadratic term is positive) in that it predicts greater breast cancer risk than with the use of the linear-quadratic model.

• Age at exposure has a substantial moderating effect on risk per unit of dose. Generally, the older a woman is at the time of exposure, the lower the risk per unit of dose.

• There appears to be a minimum latent period between exposure and the time at which risk increases, a period that appears to be at least 5 y. In addition, there appears to be no measurable increase in risk until the age at which natural breast cancer risk increases in the population around the age of 30.

• Fractionation of dose or reduced dose rate does not appear to have a major impact on subsequent risk. Thus, risk estimates based on studies, such as the atomic-bomb survivors in which doses were received in a single exposure, are generally similar to those based on studies such as the fluoros-copy studies in which doses were highly fractionated. Therefore, a conservative assumption would be that frac-tionation does not reduce risk per unit of dose, and this assumption is used in the following risk analyses.

• There is no evidence to date that risk of breast cancer returns to the normal background rate for any of the cohorts under observation, some of which have been followed for 40 or more years.

• The nature of the interaction between radiation and other risk factors in inducing breast cancer has been most studied with respect to the interaction between radiation and age at risk. Two simple probabilistic models have often been used, particularly for risk projections, namely the simple multiplicative and simple additive models. In the former, it is assumed that the RR for breast cancer following an exposure to a certain amount of radiation subsequently remains constant and multiplies the natural background age-specific risk of breast cancer. In the latter model, it is assumed that following irradiation, a constant amount of risk is added to the natural background age-specific breast cancer risk. The simple RR model predicts much larger excesses of breast cancer due to radiation than does the simple additive risk model. Recent analyses (Howe and McLaughlin, 1996; NAS/NRC, 1990, Preston et al., 2002a), have used modified versions of these two models as discussed below. Neither the simple additive nor the simple RR model provide adequate descriptions of either breast cancer incidence or mortality.

Several studies (Boice and Stone, 1978; Goodman et al., 1997; Holmberg et al., 2001; Howe, 1989; Land et al., 1994; Shore et al., 1980), have examined the interaction between radiation and other risk factors for breast cancer, in addition to age. In general, these studies are reasonably consistent with a multiplicative (i.e. constant RR) model, though, because of small sample sizes, it generally is not possible to exclude the possibility of other types of interaction models.

• There is no evidence that the case fatality rate from radiation-induced breast cancer is any different than that for other breast cancers.

• Although a number of women in the various cohort studies had breast tissue doses in excess of 1 Gy, both the atomic-bomb survivors study, and some other cohorts, had a substantial number of women with breast tissue doses below 1 Gy. Therefore, although the models used for risk estimation for women exposed to the very low doses involved in mammography (typically, a mean glandular dose of 4 mGy) inevitably involve extrapolation from higher doses, there is a substantial contribution to these risk estimates from women exposed at doses of <1 Gy.

7.2.2 Quantitative Risk Estimates

The BEIR V Committee (NAS/NRC, 1990), conducted a combined analysis of breast cancer incidence from the atomic-bomb survivors study, the Massachusetts fluoroscopy study, and the New York postpartum mastitis study, and a combined analysis of the breast cancer mortality from the atomic-bomb survivors and the Canadian fluoroscopy series.

The BEIR V Committee's preferred risk model for breast cancer incidence (I) for women age 40 y or more is given by:

where RRI is the relative risk for breast cancer incidence, D is the breast tissue dose in gray, LT15 is log of years since exposure divided by 15 (if time since exposure is <15 y and is zero otherwise), and LT30 is log of years since exposure divided by 30 (if time since exposure is >15 y and is zero otherwise). Thus, the excess relative risk (ERR) for incidence does not continue to decrease with age at exposure once a woman is past 40 y of age. Calculated RRs for breast cancer incidence predicted by this model for a mean glandular dose of 4 mGy, a typical dose from a mammographic screen (Section 5), demonstrates the lack of dependence on age at exposure once the age has reached 40 y.

The preferred mortality risk model selected by the BEIR V Committee (NAS/NRC, 1990) was a RR model which was linear in dose,

Table 7.7—RR of breast cancer incidence following a mean glandular dose of 4 mGy predicted by the BEIR V model (NAS/NRC, 1990).

Age at Exposure

0 0

Post a comment