This Article Abstract Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Law, J Articles by Young, K C Search for Related Content PubMed PubMed Citation Articles by Law, J Articles by Young, K C

Risk factors for induction of breast cancer by X-rays and their implications for breast screening

¹ Department of Medical Physics, Edinburgh University, Chancellors Building, Little France Crescent, Edinburgh EH16 4SB, ² Quality Assurance Reference Centre, Unit 9, Kingfisher Way, Wallsend, Tyne and Wear NE28 9ND, ³ Radiation Protection Service, St Luke's Wing, Royal Surrey County Hospital, Egerton Road, Guildford GU2 7XX, UK

Correspondence: K Faulkner, Quality Assurance Reference Centre, Unit 9, Kingfisher Way, Wallsend, Tyne and Wear NE28 9ND, UK

	Abstract

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
In order to discuss the balance of benefit and radiation risk in a breast screening programme, it is necessary to have numerical values for the probability of breast cancer induction by X-rays, stratified by age. Various sets of such values have been used hitherto, mainly in relation to breast screening in the UK, both within the NHS Screening Programme and more generally for younger age groups. Further sets have recently been reported. These different sets of values are described and discussed, together with the effects of using additive or relative risk models, and the effect of using a dose and dose rate modifying factor (DDREF). Possible new radiation risk factors for breast cancer induction by X-rays, drawn from these sets, are identified. These are used to calculate fresh values of cancer detection/induction ratios, as an index of benefit/risk, for screening age women and for younger women with and without a family history of breast cancer.

	Introduction

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
Breast cancer screening programmes need to demonstrate that the benefits which they offer will exceed the associated risks, and in particular the risk of induction of future breast cancers by the X-ray exposures involved. This is commonly referred to as Justification. To quantify that risk it is necessary to know, or estimate, the number of cancers predicted to be induced per unit dose, for a range of ages at exposure, during the remaining lifetimes of the exposed women. It is helpful if these estimates can be made for 5 year age bands as this facilitates comparison with cancer detection rates, which in the UK Breast Screening Programme (NHSBSP) are published in 5 year age bands (50–54 years etc.), but it is also possible to use estimates at point ages by interpolation.

Before the NHSBSP began, and in connection with a trial which preceded it, Law [1] made a very simplified set of estimates of radiation risk factors, based on a small number of papers reporting results of repeated medical exposures some decades previously. Those papers [2–4] had shown an encouraging degree of agreement. In 1994, the National Radiological Protection Board (NRPB) provided a fresh set of risk factors which they regarded as the most appropriate for the UK female population. The numerical values were contained in the software described in an NRPB report [5] but the actual numerical values were communicated on a personal basis and did not appear in the open literature until they came to be used in association with the NHSBSP. Those values were then used in a series of papers which discussed the balance of benefit and radiation risk in that programme [6, 7].

A fresh approach to this topic was provided by Preston et al in 2002 [8], and considerable prominence has been given to that paper in a consultation document from the International Commission on Radiological Protection (ICRP) which was widely circulated in 2005 [9]. Preston et al review eight different cohorts of irradiated individuals, including atom bomb survivors and seven other groups who received medical exposures. They do not themselves provide numerical risk factors but offer equations from which such factors may be calculated for any age or population. The ICRP document discusses proposals for revision of cancer induction estimates for all body organs, and in relation to breast cancer also gives useful discussion of some key underlying assumptions, especially the relative claims of absolute and relative risk models, and the question of the dose and dose rate modifying factor (DDREF) to be used at low doses.

In their review of breast cancer risks, Preston et al [8] conclude that the results of their pooled analysis "support the linearity of the radiation dose response for breast cancer". This is sometimes referred to as the linear no threshold (LNT) hypothesis. Preston et al [8] suggest that a dose and dose rate effectiveness factor of two to three is applied for exposures which are small and occur over an extended period. As with all approaches for the estimation of risk factors using the LNT model, there is always an uncertainty associated with the extrapolation from high dose to protracted low dose exposures. This uncertainty is reflected in the comments on DDREF at the end of the paper by Preston et al [8].

The purpose of this paper is, first, to use the Preston equations to obtain new breast cancer induction risk factors, second to compare those factors with others, and third to assess their implications for breast screening programmes in the UK and elsewhere.

	Absolute and relative risk models

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
For radiation induction of many forms of cancer, the excess relative risk (ERR) model is currently favoured over the alternative model based on excess absolute risk (EAR). In the relative risk model, the increased risk is taken to be proportional to the natural underlying incidence of the cancer concerned, whereas in the absolute risk model the increase is taken to depend on dose and age at exposure but to be independent of the underlying incidence. For some years the majority view in relation to breast cancer also favoured the ERR model, but opinion now seems to be returning to favour the EAR model in this instance, as discussed below.

Preston et al discuss both models in their paper [8], and attempt to relate each to studies of eight cohorts of irradiated females. An equation is stated based on the ERR model but they go on to give greater prominence to a pair of equations based on the EAR model, with parameters based on all eight cohorts, which cover age ranges above and below 50 years at time of exposure. They thus appear to favour the EAR over the ERR model, though they do not explicitly say this. They conclude by recommending this pair of equations based on their EAR model for risk–benefit calculations. The radiation risk factors derived using either model would vary with both age at exposure and attained age. The ICRP consultation document [9] endorses that view for breast cancer. In this paper we explore both approaches and consider the implications for screening mammography.

One of the cohorts of irradiated women studied by Preston et al [8], and by many others, is that of the Japanese atomic bomb survivors. In a Japanese population the natural incidence of breast cancer is considerably lower than in western populations. The ERR model is based on the concept of a multiplication of natural incidence per unit dose, so that by this model higher risk factors result from data on Japanese women than might be derived from western populations. For this reason observations are thought to fit the EAR model better than they fit the ERR approach. Risk factors derived using the EAR model appear to be constant across populations. One of the major problems in this subject is transfer of risk estimates between different populations.

	Dose and dose rate effectiveness factor (DDREF)

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
Although dose limits and regulations in the field of radiation protection are generally based on LNT extrapolations from high dose to low dose, it has long been accepted practice to make some allowance for the likelihood that low dose or dose rate exposures have less effect per unit dose than those at higher doses or dose rates. This is commonly done by applying a DDREF of 2 to risk factor estimates, i.e. halving their numerical values. The ICRP consultation document [9] discusses this point, and concludes that in the present state of knowledge a DDREF of 2 should continue to be applied for fractionated exposures with acute fractions of a few mSv or less and total exposures below 200 mSv. That higher dose level is not normally reached in any single screening round, even if assessment clinics are taken into account. The expression "screening round" in this context means the routine attendance by women in response to invitation as part of a 3 yearly cycle, together with any associated subsequent visits for assessment arising from abnormalities on films taken at the initial visit. All visits within any single screening round will normally be concluded within a total of about a month. Doses from all exposures within a round may be aggregated for risk assessment, though this may not be appropriate over longer time spans. In the NHSBSP, doses in a single round are typically about 4.5 mSv, and will very rarely if ever exceed 200 mSv [10]. Aggregation of doses is not appropriate for successive rounds at longer intervals, where fresh benefit arises on each round. Even if doses were to be aggregated over all seven screening rounds in the NHSBSP, i.e. a total time span of 18 years, the total dose would still not normally reach the 200 mSv level, below which ICRP consider a DDREF of 2 to be appropriate.

It seems appropriate, therefore, to continue to apply a DDREF of 2 in the calculations which follow. Preston et al do not apply it directly to their equations but state at the end of their paper: "For protracted exposures, we suggest that estimates from this model be reduced by a factor of two or three as is customary in radiation risk estimation". We take this statement to refer to the DDREF as discussed above and will use a value of 2 for that factor. Preston et al indicate that in their view an even higher value might not be unreasonable, but some opinion believes that only a DDREF of 1 [11] (i.e. no modifying factor) should be used, and the point remains somewhat controversial.

	Calculations from equations derived by Preston et al [8]

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
Preston et al provide the following pair of equations based on the excess absolute risk model: (The typographic errors in Preston's original paper have been corrected in an erratum [12] and the modified version has been used here.)

For attained age 50

For attained age > 50

where agex means age at exposure. These equations give excess cancers per Gy.

No results are provided by these authors for calculations using these equations, and readers are left to make such calculations for themselves. Likewise no precise instructions are given in their paper as to how calculations should proceed, but they do say that they assume a 10 year delay post exposure before any consequent risk arises.

Our calculations have proceeded as follows. For X-ray exposure to unit dose, at a given age, the excess risk was calculated for 1 year at attained age = agex+10. For each following year this was repeated with attained age increased by 1 year each time. For example, for exposure to 1 mGy at age 55 years (agex), Equation (2) was evaluated for age 65 years, which represents the risk for 1 year at that age. It was then evaluated for ages 66, 67, 68 etc. and the risks for each year were summed. This was continued to attained age 85 years, in line with Berrington de Gonzalez and Reeves [11]. For convenience, the results are summed in 5 year age bands to yield total lifetime risks, starting with the 6 years of the 80–85 year band. Each younger 5 year band can then be included to provide lifetime risks at progressively younger ages.

A further correction is applied to allow for the effect of other causes of death, using UK data from conventional life tables [13]. This correction is not mentioned by Preston et al but is made by Berrington de Gonzalez and Reeves. Its effect becomes appreciable after age 75 years, and its application at these ages affects lifetime risks for all younger ages.

Preston et al [8] give an equation for the ERR model but do not evaluate it, presumably because they consider the EAR model to be more appropriate for breast cancer. However, they do provide sufficient information in their paper, mainly in their table 9, to allow others to make calculations on this model if they wish. Berrington de Gonzalez and Reeves [11] suggest that Equation (3) (below) be used with values taken from Table 9 of Preston et al for three of the eight cohorts (see below) cited in that paper [Berrington de Gonzalez, personal communication].

where is the underlying breast cancer incidence rate in the population of interest at a given attained age. In Equation (3), the factor 0.74 and the index –2 are the values found when the ERR model developed by Preston was fitted to the Massachusetts TB fluoroscopy cohort, its extension cohort and the Rochester infant thymic cohort [14, 15]. However, Berrington de Gonzalez and Reeves favour a DDREF of 1 [Berrington de Gonzalez, personal communication] in contrast to Preston et al and to ICRP [9].

Calculations using Equation (3) have been made using the same year-by-year procedure as described earlier for the EAR model, with the same life table corrections to allow for other causes of death. The results are given with those for the EAR model in Table 1.

	Discussion

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

Columns b, c and d show three sets of values prepared by the NRPB. The first set (column b) was used by Law in a series of papers from 1995 [6], and by Law and Faulkner in a further series from 2001 [7]. It derives from SPIDER software [5] and initially was given in 10 year age band averages; these were then modified to 5 year bands by Law after discussion with NRPB. The second set of values produced by the NRPB derived from ASQRAD software [16] and is given at point ages, as are the other values in Table 1 except for column b. Both these NRPB sets (in columns b and c) are based on a relative risk model, using a single value for all ages above 20 years but use slightly different methods of calculation. The decrease in risk with age of exposure results from decreasing life expectancy and competing causes of death. These two NRPB sets differ slightly due to small differences in the life tables used. A DDREF of 2 has been included in both sets. The ASQRAD set was used in NHSBSP Report 54 [17].

The third set from the NRPB (column d) for all attained ages has been derived more recently (Muirhead and Haylock, pers. comm. 2003). It is also based on the ERR model with a DDREF of 2. The calculations were performed by the NRPB using 1998 data for breast cancer incidence in England and Wales (Haylock, pers. comm.). Columns e and f of Table 1 show the results of calculations from the equations of Preston et al as described earlier, also with a DDREF of 2. There is a clear difference in the dependence on age at exposure between the Preston et al EAR values and the NRPB and Preston et al ERR values. Both excess relative risk ERR models show a more gradual decline with increasing age at exposure.

Cancer detection/induction ratios
Cancer detection rates for the NHSBSP age range of 50–70 years have been taken from an earlier paper [18] and are based on observed results given in NHSBSP Reports [19] published up to that time. Although the most recent detection rates in the NHSBSP may be slightly higher than these, the results used in that earlier paper have been retained here to facilitate comparison with previous conclusions.

For younger women aged 30–49 years with no family history of breast cancer, cancer detection rates were previously calculated by scaling from detection rates for older women in proportion to the known incidence of breast cancer in the relevant age groups and dividing by 3 to allow for annual screening [6]. The same principle has been followed in the present work but using the most recent mean detection rate for women aged 50–64 years of 5.2/1000 given in reference [18] in place of the older detection rate of 3.8/1000 as in reference [6]. Detection rates in this age range are thus increased by a factor of 5.2/3.8 = 1.37 compared with earlier work, and further increased by 24% for the effect of two-view screening [20]. It is assumed here that this 24% increase, which was derived from screening of women over 50 years of age every 3 years, will also apply to younger women. These estimated detection rates may be slightly too high if detection is more difficult in younger women with denser breasts, but any error from this cause will be counterbalanced by a reduction in interval cancers due to annual screening. For the age bands 40–44 years and 45–49 years, this procedure gives detection rates for single view screening of 1.0 and 1.5 per 1000 women screened, in very close agreement with rates of 1.1 and 1.6 per 1000 women screened in these age groups recently reported by Moss et al from the UK Age Trial [21]. Thus it seems unlikely that detection rates predicted here for the 30–49 years age range, without a family history of breast cancer, will be overestimated.

Detection rates for women aged 30–49 years with a family history of breast cancer are taken from an earlier paper [22] and are based on the work of Houlston et al [23]. They depend on the age range in which the close relative index patient was herself diagnosed with breast cancer.

Induction rates have been calculated on the basis of a mean glandular dose of 4.5 mGy for two-view screening [10] with no variation due to age [24].

Calculated values of cancer detection/induction ratios have been published in a number of previous papers [6, 7]. However, these are not the same as benefit/risk ratios, which are thought to be in the region of one third to two thirds of the detection/induction ratio, at least in the current screening age range of 50–70 years [17, 25]. At younger ages, where screening benefit in terms of mortality reduction may be less, this proportion may be correspondingly less (for example, breast cancers in younger women may be more aggressive or faster growing). Cancer detection/induction ratios as published previously [6, 7] have been based solely on the NRPB Spider values, as given here in column b of Table 1. In Table 2, detection/induction ratios are given for women aged 50–70 years based on those Spider values, on the latest NRPB risk factors and on the Preston et al EAR equations. All these sets of risk factors incorporate a DDREF of 2.

The NHSBSP, age 50–70 years
In this age range, cancer detection/induction ratios at all ages are quite large, well above 5 or 10, regardless of which set of risk factors is chosen. Thus there are no significant new implications in any of these figures for screening in the 50–70 year age range. The benefit appears to exceed the radiation risk by an appreciable margin.

Ages 30–49 years, without a family history
For women in this age range without a family history of breast cancer, cancer detection/induction ratios exceed five on all sets of risk factors down to age 35 years (Table 3a). Below that age, annual two-view screening should not be considered for this population, and even adopting a 2 year screening interval rather than annual would scarcely resolve the radiation risk. This is important for those who may refer women for screening outside the health service, including women who may wish to refer themselves.

Ages 30–49 years, with a family history
In the following discussion, it is assumed that radiation risks for this subgroup of women are substantially the same as for the general population. If the risk of radiation induction of breast cancer in women with a family history of the disease were raised in proportion to the increased breast cancer incidence in this group, then the detection/induction ratio and hence the benefit/risk ratio would be substantially the same as in the no-family history group discussed above. This point has been made in earlier papers [6, 7, 18].

For those with a close relative (index patient) diagnosed in the 30–39 years age range, the detection/induction ratio exceeds 10 on all sets of risk factors down to age 30 years (Table 3b), but annual two-view screening should not be considered at any lower age.

For women whose index patient was diagnosed in the 40–49 years age range, the detection/induction ratio values are shown in Table 3b (ii) and are slightly lower than those in Table 3b (i) for index patients aged 30–39 years at diagnosis. All ratios exceed 10 down to age 35 years. Even if a DDREF of 1 were to be used, the ratio would exceed or be very close to 10 down to age 35 years, and would still be close to 5 down to age 30 years, but the uncertainties outlined earlier would make such screening at ages 30–34 years questionable. Again, this applies to annual two-view screening: a 2 year interval would remove this caveat.

Even if a DDREF of 1 were to be used, the ratio would still exceed 5 down to age 35 years on all risk factors and would exceed 10 using the absolute risk model favoured by Preston et al and by ICRP.

This discussion is based on the current mean dose level of 4.5 mGy for two-view screening. That dose level may decrease with the expected gradual increase in digital imaging in mammography, and cancer detection rates may increase, at least in younger women, for the same reason. If both these effects occur, detection/induction ratios and the corresponding benefit/risk ratios could improve to an appreciable extent.

There may be subgroups of the population (e.g. carriers of the ataxia telangiectasia gene) who may be more radiosensitive to radiation. The approach described in this paper might be adapted for the calculation of radiation risk for groups of this kind. If the ERR model were used, then the underlying risk of breast cancer in the high risk subgroup would need to be known.

	Conclusions

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 
Several new sets of risk factors for breast cancer induction by X-rays have been proposed in recent years, based on a range of assumptions. In the NHSBSP, benefit continues to exceed radiation risk regardless of which of these sets is adopted. At younger ages, there is little if any risk of detriment exceeding benefit down to age 40 years. Annual two-view screening should not be considered below age 35 years for women with no family history, and even for those who do have such a history, it should only be considered if the index patient was diagnosed below age 40 years. There is need for further research and discussion concerning the use of a DDREF of 2 for X-ray exposures at doses below 200 mSv. There also remains a need for further consideration of whether women with a family history of breast cancer have the same susceptibility to X-ray induction of breast cancer as the normal population. It is also desirable to establish more reliable cancer detection rates for younger women with family histories of breast cancer than those derived earlier and used again here. This might be best done by pooling results from all UK family history screening clinics.

Received for publication January 31, 2006. Revision received June 26, 2006. Accepted for publication August 15, 2006.

	References

Top Abstract Introduction Absolute and relative risk... Dose and dose rate... Calculations from equations... Discussion Conclusions References

 

1. Law J. Cancers induced and cancers detected in a mammography screening programme. Br J Radiol 1987;60:231–4.[Abstract/Free Full Text]
2. Boice JD, Jr, Land CE, Shore RE, Norman JE, Tokunaga M. Risk of breast cancer following low-dose radiation exposure. Radiology 1979;131:589–97.[Abstract]
3. Boice JD, Jr, Monson RR. Breast cancer in women after repeated fluoroscopic examinations of the chest. J Natl Cancer Inst 1977;59:823–32.[Medline]
4. Shore RE, Hempelmann LH, Kowaluk E, Mansur PS, Pastermak BS, Albert RE, et al. Breast neoplasms in women treated with x-rays for acute postpartum mastitis. J Natl Cancer Inst 1977;59:813–22.[Medline]
5. Stokell PJ, Robb JD. SPIDER-I Software for evaluating the detriment associated with radiation exposure. NRPB – SR261. Chilton: NRPB, 1994
6. Law J. Risk and benefit associated with radiation dose in breast screening programmes. Br J Radiol 1995;68:870–6.[Abstract/Free Full Text]
7. Law J, Faulkner K. Cancers detected and induced, and associated risk and benefit, in a breast screening programme. Br J Radiol 2001;74:1121–7.[Abstract/Free Full Text]
8. Preston DL, Mattsson A, Holmberg E, Shore RE, Hildreth NG, Boice JD, Jr. Radiation effects on breast cancer risk: a pooled analysis of eight cohorts. Radiat Res 2002;158:220–35.[CrossRef][Medline]
9. 20/05 Recommendations of the International Commission on Radiation Protection (Available in draft at www.icrp.org)
10. Young KC, Burch A, Oduko, JM. Radiation doses received in the UK Breast Screening Programme in 2001 and 2002. Br J Radiol 2005;78:207–18.[Abstract/Free Full Text]
11. Berrington de Gonzalez, Reeves G. Mammographic screening before age 50 in the UK: comparison of the radiation risks with the mortality benefits. Br J Cancer 2005;93:590–6.[CrossRef][Medline]
12. Preston DL, Mattsson A, Holmberg E, Shore RE, Hildreth NG, Boice JD, Jr. Erratum. Radiat Res 2002;158:666[CrossRef]
13. World Health Organization. Life tables for 191 countries: data, methods and results. Discussion paper series no. 9. WHO, Geneva (2000). http://w3.whosea.org/healthreport/pdf/paper09.pdf
14. Boice JD, Jr, Preston DL, Davis FG, Monson RR. Frequent chest x-ray fluoroscopy and breast cancer incidence among tuberculosis patients in Massachusetts. Radiat Res 1991;125:214–22.[Medline]
15. Hildreth NG, Shore RE, Dvoretsky PM. The risk of breast cancer after irradiation of the thymus in infancy. N Engl J Med 1989;321:1281–4.[Abstract]
16. European Commission. ASQRAD – Assessment system for the quantification of radiation detriment. EUR16644, CEPN-L-95/2. Luxembourg, EC, 1996
17. Review of radiation risk in breast screening. NHSBSP Publication 54, Sheffield, 2003
18. Law J, Faulkner K. Two-view screening and extending the age range: the balance of benefit and risk. Br J Radiol 2002;75:889–94.[Abstract/Free Full Text]
19. NHSBSP Reviews, 1994–5 to 1998–9 inclusive. Statistical bulletin. London: Department of Health
20. Wald NJ, Murphy P, Major P, Parkes C, Townsend J, Frost C. UKCCCR multicentre randomized controlled trial of one and two view mammography in breast cancer screening. BMJ 1995;311:1189–93.[Abstract/Free Full Text]
21. Moss S, Waller M, Andersson TJ, Cuckle H. Randomised controlled trial of mammographic screening in women from age 40: predicted mortality based on surrogate outcome measures. Br J Cancer 2005;in press
22. Law J. Cancers detected and induced in mammographic screening: new screening schedules and younger women with family history. Br J Radiol 1997;70:62–9.[Abstract]
23. Houlston RS, McCarter E, Parbhoo S, Scurr JH, Slack J. Family history and risk of breast cancer. J Med Genet 1992;29:154–7.[Abstract/Free Full Text]
24. Young KC. Radiation doses in the UK trial of breast screening in women aged 40–48 years. Br J Radiol 2002;75:362–70.[Abstract/Free Full Text]
25. Law J, Faulkner K. Concerning the relationship between benefit and radiation risk, and cancers detected and induced, in a breast screening programme. Br J Radiol 2002;75:678–84.[Abstract/Free Full Text]

This article has been cited by other articles:

				K. Faulkner Ethical concerns arising from screening procedures such as mammography and self-referral Radiat Prot Dosimetry, July 1, 2009; 135(2): 90 - 94. [Abstract] [Full Text] [PDF]

				D. A. Kennedy, T. Lee, and D. Seely A Comparative Review of Thermography as a Breast Cancer Screening Technique Integr Cancer Ther, March 1, 2009; 8(1): 9 - 16. [Abstract] [PDF]

				BJR review of the year -- 2007 Br. J. Radiol., April 1, 2008; 81(964): 265 - 269. [Full Text] [PDF]

This Article Abstract Full Text (PDF) Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Law, J Articles by Young, K C Search for Related Content PubMed PubMed Citation Articles by Law, J Articles by Young, K C