This Article Abstract Figures Only 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 Google Scholar Google Scholar Articles by Beckett, J R Articles by Michaelson, J S Search for Related Content PubMed PubMed Citation Articles by Beckett, J R Articles by Michaelson, J S

Analysis of benefit:risk ratio and mortality reduction for the UK Breast Screening Programme

¹ Regional Medical Physics Department, Newcastle General Hospital, Newcastle-upon-Tyne NE4 6BE, UK and ² Departments of Pathology and Radiology, Massachusetts General Hospital and Department of Pathology, Harvard Medical School, Boston, MA USA

	Abstract

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
A quantitative analysis has been performed to predict the benefit:risk ratio and associated mortality reduction for the UK National Health Service Breast Screening Programme. The analysis is based on the results of an established biological simulation method coupled with dosimetric information and population statistics applicable to the UK breast screening programme. As well as the general breast screening population, the benefit:risk ratios for specific subgroups of women thought to be at higher risk are estimated. The effects of alterations in screening strategy are also investigated. The results indicate favourable benefit:risk ratios and mortality reductions for all women in the programme, with a breast cancer mortality reduction of approximately 9% over the whole UK female population, equivalent to a breast cancer mortality reduction in the region of 25% for the age range 55–69 years.

	Introduction

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
Screening mammography is considered to be effective in reducing breast cancer mortality in women over 50 years of age. As with every procedure involving exposure to the harmful effects of ionizing radiation, this screening modality carries with it a potential risk. The International Commission on Radiological Protection (ICRP) [1] states that the beneficial effects of such procedures should outweigh the associated detrimental effects. It is therefore important that the benefit:risk ratio (BRR) associated with mammography screening is known.

Following the recommendations of the Forrest report in 1986 [2], the National Health Service Breast Screening Programme (NHSBSP) was implemented throughout the UK. In the NHSBSP, mammography screening is currently offered to all women aged between 50 years and 64 years at three yearly intervals, with screening for older women available on request. Previous research has suggested that such a programme will be effective in reducing breast cancer mortality [3, 4]. It has also been stated that exposure of the female breast to ionizing radiation may cause breast cancer in later life [5, 6]. The ratio between the number of lives saved as a result of mammography screening and the number of deaths caused due to radiation induced breast cancer (the BRR) is used in the justification of the technique, although it has been stated that division of years of life gained by years of life lost would be a better measure of this quantity [7].

A number of studies have been conducted with a view to quantifying the BRR [7–9], all of which indicate that the benefits of screening whole populations of women aged over 50 years outweigh the associated risks by a substantial amount. Comparisons of cancers prevented by mammography with cancers induced have also been presented for different subgroups of women attending screening in the UK [10–13]. The results of this work indicate that screening is justified for the general population. Few quantitative estimates of BRR and mortality reduction for the NHSBSP have been published.

This paper attempts a quantitative analysis of BRR and mortality reduction based on the results of an established biological simulation method coupled with dosimetric information and population statistics applicable to the UK breast screening programme. As well as the general breast screening population, the BRRs relating to specific subgroups of women thought to be at higher risk are estimated. The effects of alterations in screening strategy are also investigated.

	Method

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
Computer simulation of breast cancer screening
The present work is based on a revised version of the simulation described by Michaelson et al [14] for the calculation of the beneficial effects of mammography screening. The simulation is biological in basis and utilizes empirical data regarding the rate of invasive breast cancer growth, spread and detectability, which has been extracted from a large American breast imaging data base.

For a given interval between successive screens, the simulation calculates age of screen and screening interval specific probabilities that breast cancer will have metastasized before screen detection, or detection by palpation, whichever occurs first. The probability of metastasis before detection by palpation alone is also calculated in order to account for the scenario in which screening is not offered. The output of the simulation therefore gives the fraction of tumours which would be expected to undergo metastasis before detection by age both for women screened at given intervals, and for unscreened women.

Estimates of age specific reductions in breast cancer mortality
The information on the fraction of metastatic breast tumours in screening and non-screening scenarios was used in conjunction with survival data relating to patients diagnosed with metastatic and non-metastatic breast cancers for the estimation of screening related reductions in breast cancer mortality. This was achieved by estimating and comparing the number of breast cancer deaths that would occur in a screening and in a non-screening scenario. The survival curves for women with metastatic breast cancer have been taken from the work of Debonis et al [15], and the survival curves corresponding to those diagnosed with non-metastatic breast cancer from Arnesson et al [16]. Both curves appear to plateau at around 10 years post diagnosis.

The screening scenario
Using breast cancer incidence [17] and age structure data [18] relating to the British female population, the numbers of breast cancers which would be expected to occur at each year of age were calculated. For a given year of age within the screening age range of a specified screening regimen, the number of expected cancers was multiplied by the corresponding metastatic proportion calculated by the simulation method described above. This led to estimates of the numbers of cancers diagnosed in that age range which would be metastatic, and those which would be non-metastatic for the screening regimen in question. For each age lying outside the screening age range, the age specific fraction of breast cancers expected to be metastatic calculated in the computer simulated non-screening scenario was used to perform similar calculations. The numbers of women with metastatic and non-metastatic disease were reduced using the relevant survival curves, and the ages at which the corresponding deaths occurred were recorded. This was done for all ages between 20 years and 85 years. In this way, the numbers of breast cancer deaths occurring at each year of age in the screening regimen (and at a sufficient number of ages surrounding it) were estimated.

The non-screening scenario
The numbers of deaths occurring at each year of age in the non-screening regimen were calculated using a similar, though slightly more complicated method, as it was necessary to allow for the lead time associated with screening. It was assumed that the numbers of women who had either non-metastatic or metastatic breast cancer at any age within the screening age range were at first predicted by the relevant proportions estimated by the screening regimen computer simulation. These numbers were then reduced in the usual fashion for the duration of the lead time period. Following this, it was assumed that a number of women who at first had non-metastatic breast cancer would swap groups, their cancers becoming metastatic. The survival of these women would then be governed by the metastatic survival curve. In this way, those women whose cancers can be expected to become metastatic during the lead time are accounted for.

For any given year of age within what would be the screening age range, the age specific metastatic proportion corresponding to the computer simulated screening regimen was used to give estimates of the number of non-metastatic and metastatic breast cancers. Similar estimates were made using the age specific metastatic proportion calculated using the computer simulated screening regimen. The difference in the numbers of metastatic breast cancers calculated using the computer simulated screening and non-screening estimates was then calculated.

The numbers of women with metastatic and non-metastatic breast cancer calculated using the computer simulated screening regimen were then reduced in the usual fashion for a period of 3 years (or 2 years for those women aged under 50 years) (corresponding to the lead times estimated by Moskowitz [19] and Fox et al [20]). Following this, a number of women equal to the difference between the number of metastatic cancers estimated for the screening and non-screening scenarios was subtracted from the non-metastatic group. Their numbers were then reduced in accordance with the metastatic breast cancer survival curve. The remaining non-metastatic numbers continued to decay in the usual fashion, as did the number of women left in the initial metastatic group.

For women outside the screening age range, such a procedure was not necessary. For these women, the age specific fractions of metastatic cancers estimated by the computer simulation for the non-screening regimen were used to estimate the numbers of metastatic and non-metastatic breast cancers. These numbers then decayed in accordance with the relevant survival curves under the assumption that no cancers would swap groups.

Age specific reductions in breast cancer death
The reductions in breast cancer mortality, which could be expected at each year of age as a result of a given screening scenario, were then calculated and modified in accordance with the overall attendance rate of 75.4% reported for the NHSBSP [21]. This was done by comparison of the number of breast cancer deaths which would be expected at each year of age in the screening and non-screening scenarios described above. The method was followed for a number of different screening age ranges and intervals. The resulting data gave the quantitative, age-specific estimates of the reductions in breast cancer mortality which were used in the calculation of the BRRs reported below.

Breast radiation dose and radiation risk estimates used in this work
Radiation risk from mammography is thought to be linearly related to mean glandular dose (MGD) [5, 6]. The radiation dose estimates used for the estimation of mammography related radiation risk for women in the general population have been taken from a data set used in a previous publication [22]. These data relate to a total of 1258 women who attended screening in the northern part of the Northern and Yorkshire health region, England. They were collected during 1998 and 1999, take into account the effects of differences in breast composition, and are concerned with women in the age range 35 years to 79 years. In an attempt to make the trends in radiation dose with age and breast thickness independent of variations in output of the mammography sets used, each patient MGD estimate in the set was first divided by the MGD to the standard breast phantom for the mammography set used to screen the breast in question. Equations representing the trends of interest were then fitted to these data as described below.

Equations relating normalized MGD and patient age (A) for the lateral oblique (obl) views and craniocaudal (cc) views commonly used in the NHSBSP were then found. These relationships are explicitly stated in Equations 1 and 2, respectively.

Where MGD(A)_obl is the normalized MGD relating to a lateral oblique view, MGD(A)_cc is the normalized MGD relating to a cc view. (Table 1 lists the symbols and abbreviations used in this and later sections.) These equations were used to obtain normalized MGD estimates for cc and/or lateral oblique views relating to women in the age range 35 years to 79 years. For example, the normalized MGDs for a 65-year-old woman are estimated to be 1.447 and 1.421 for the oblique and cc views, respectively. Normalized dose estimates made using these equations were then multiplied by 1.34 mGy, the mean standard breast dose in the UK [23] to give average MGD estimates. These were then used for the estimation of the radiation risk incurred by an average woman from a mammographic examination at a given age, or the cumulative radiation risk incurred following a series of examinations.

where (D) is the relative risk estimate, ₀ is the baseline breast cancer mortality, D is the breast absorbed dose in sieverts, T is the time since exposure and E is the age at exposure.

This relative risk model was chosen for this work following a review of sources of radiation risk data and their supporting evidence. It was felt to be prudent to use a relative risk model incorporating a linear dose response relationship for the estimation of radiation induced breast cancer mortality. This conclusion has been drawn because relative risk models when compared with absolute risk models lead to higher overall risk estimates following extrapolation beyond the duration of follow-up of the survival studies available. Similarly, the use of a linear dose response relationship leads to higher low-dose risk estimates than the linear quadratic model. Omission of a dose and dose-rate effectiveness factor also leads to conservative radiation risk estimates as its use leads to a reduction in the assumed effects of exposure. Other authors have made different assumptions. Law [10–13] uses the risk estimates based on those produced by the National Radiological Protection Board (NRPB) [25], which do not include atomic bomb survival data. Comparing projected lifetime risk of fatal breast cancer at high dose and dose rate, the risk estimate used here (0.52 10^-2 Sv^-1) is approximately half that of the equivalent value in the NRPB model (1.1 10^-2 Sv^-1). The equivalent value from the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) 2000 report [26] is 15% higher than that used here at 0.6 10^-2 Sv^-1.

Age-specific MGD estimates for women with breasts of a given lateral oblique thickness were made. The average normalized MGD (D_N,x,t) for each view type (either lateral oblique or cc) received by women in the data set with breasts in each 10 mm lateral oblique breast thickness range from 30 mm to 90 mm who were also in the age range 50 years to 54 years were calculated. These data are shown in Table 2.

Here, D_N,x,t denotes the normalized MGD for a breast of lateral oblique thickness t undergoing a mammographic examination with view type x, and D_N(52.5)_x denotes the normalized MGD estimate at age 52.5 years for view type x calculated using either Equation 1 or 2. (NB. Normalized MGD estimates are labelled by the view type in question (either obl or cc), but by lateral oblique thickness only, as the lateral oblique and cc thicknesses of a given breast differ significantly [23]).

The F factor is simply the ratio of the view type dependent, normalized MGD received by a woman aged between 50 years and 54 years whose breast thickness lies within a given 10 mm interval to that for an average woman at age 52.5 years undergoing a similar mammographic view. Multiplication of either Equation 1 (for oblique views) or Equation 2 (for cc views) by the relevant view type and breast thickness dependent F factor therefore results in an estimate of the age dependency of the normalized MGD estimates relating to breasts of the lateral oblique compressed thickness in question following a mammographic examination of a given view type. F factors relating to breasts in each 10 mm lateral oblique thickness interval in the range 30 mm to 90 mm, and for each view type mentioned are shown in Table 2.

By multiplication of Equations 1 and 2 by the factors described in Equation 4, and by the average standard breast dose of 1.34 mGy [23], MGD estimates for breasts in each 10 mm thickness interval from 30 mm to 90 mm, and in the age range 35 years to 79 years old were made. Using these, radiation risk estimates and their variation with age for women whose breast thicknesses at age 50–54 years are known could be estimated and were used in subsequent breast thickness dependent benefit:risk analyses.

	Method used for all benefit:risk analyses

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
The procedure used to calculate all BRRs was that of Chiacchierini et al [9] which is in turn, an extension of the life table Equations of Chiang [27]. A summary of this method is given below. For a more thorough treatment, the reader is referred to these publications. In what follows, it was necessary again to calculate the number of breast cancer deaths in a screening and in a non-screening scenario. This method, however, was completely different and separate from that used above to estimate age specific mortality reductions, and in fact uses its results.

The unabridged UK female life table was first constructed using all cause mortality data and breast cancer mortality data relating to the period before the NHSBSP was initiated [28]. For each screening regimen considered, the age specific breast cancer mortality reductions described above, and UK life table quantities (life expectancy, probability of death from all causes at each year of age, number of deaths at each year of age, number of breast cancer deaths occurring at each year of age) from age 20 years to age 100 years were calculated and utilized under the approximation that all women die by age 100 years. As breast cancer mortality is negligible at ages under 20 years, it was thought that the choice of such an age range was justified.

The screening scenario
The screening scenario was simulated with a view to calculating the number of deaths from radiogenic and non-radiogenic breast cancer at each year of age, and the corresponding number of years of life lost. First, the number of deaths at the first year of age considered (age 20 years) from all causes was calculated from a knowledge of the number of women aged 20 years and the probability of death from all causes at this age. The proportion of these women who would have died from breast cancer was then calculated using information on the fraction of the year 20–21 survived by women who died in this interval, and on both the all-cause and breast cancer mortality rates in that interval. This was then used to calculate the number of breast cancer deaths in the interval in question.

The probability of radiogenic breast cancer death was then calculated from a knowledge of the relative risk of radiogenic breast cancer mortality [6], the all-cause mortality rate, and the fraction of that year of life lived by women who die in the corresponding age interval. Multiplication of this probability by the number of women alive at the beginning of the interval in question was then used to estimate the number of deaths from radiogenic breast cancer that would be expected in the corresponding year of life.

The "true" number of breast cancer deaths that would occur in the screening scenario, d_s, after the beneficial effects of mammography screening are taken into account was then calculated. This is simply the sum of the numbers of radiogenic and non-radiogenic breast cancer deaths calculated as above, multiplied by the fraction of breast cancer deaths that are not averted as a result of screening (taken from the age specific breast cancer mortality data generated above). Similarly, the number of "true" radiogenic breast cancer deaths, d_s,r, was calculated as the product of the calculated number of radiogenic breast cancer deaths and the fraction of breast cancer deaths not averted by screening.

The next step was to calculate the probability of death from all causes after adjustment for the beneficial effects of screening, and for the detrimental effects of radiation exposure. This was evaluated as the total number of deaths that occurred in the year of life in question normalized by the number of women who were alive at the start of the interval. Using this and a knowledge of the number of women alive at the beginning of the age interval 20–21, the number of women alive at the beginning of the following year of life was calculated. The process was then repeated for single year age intervals up to age 100 years.

The number of person years of life lost to breast cancer in each age interval, y_s, was simply calculated by multiplying the life expectancy, e, at the relevant age and d_s. Similarly, the number of years of life lost to radiogenic breast cancer y_s,r, was calculated as the product of e and d_s,r.

The non-screening scenario
Again, calculation of parameters relevant to the non-screening scenario was fairly straightforward. For the first age interval, the number of women who are expected to die from all causes is first calculated, as in the screening scenario. The proportion of those women who are expected to have died from breast cancer, d_ns, is then calculated using the same method as for the screening scenario. In the non-screening scenario, there is no screening benefit or radiation risk, therefore, the numbers of women surviving to the next year of life is the same as for the UK life table, and the more complex calculations used in the screening scenario are avoided. As in the preceding discussion, the years of life lost from breast cancer deaths at each year of age interval, y_ns, are estimated as the product of d_ns and e.

Calculation of BRRs, reductions in death and years of life lost
The BRR ratio in terms of breast cancer deaths, BRR(d), was calculated using Equation 5:

where i denotes the age intervals considered.

The BRR in terms of years of life lost to breast cancer, BRR(y), was calculated using Equation 6.

The reduction in breast cancer death, R(d), occurring as a result of a given screening programme was calculated using Equation 7.

Finally, the corresponding reduction in the years of life lost to breast cancer, R(y), was calculated using Equation 8.

	Validation of method

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
A number of groups have investigated the reductions in breast cancer mortality, years of life lost to breast cancer and the BRRs associated with screening. As four of these studies have used simulation methods to estimate the quantities of interest [4, 29–31] they have been used for comparison with the present work.

Jansen and Zoetelief [29] have simulated the effects that screening a population of one million Swedish women over the age range 40 years to 81 years has on breast cancer mortality. Their results have been presented in terms of the lifelong reduction in the number of fatal tumours relating to screening in each of 7 age ranges, each of which covers 6 years of life. 0.5, 1, 2, and 3 year screening intervals have been considered, and the data were normalized by the length of each respective age interval.

To allow for comparison with the present work, these data were converted to give the total number of lives saved for screening a population of 100 000 women over the age range 40 years to 81 years. The present method was then used to calculate similar data for a population of 100 000 British women, and the BEIR V relative risk model coupled with an MGD of 2 mGy per screen was assumed in keeping with the work of Jansen and Zoetelief [29]. The results of these comparisons are shown in Table 3.

Wald et al [32] have estimated that a screening programme similar to that used in the UK will result in 412 lives saved per 100 000 women entering the programme at age 50 years, when estimated using the present simulation, this number was found to be 423.

In the light of the differences in the approaches used to generate the data discussed above, the level of agreement between the present work and the other studies considered is thought to be acceptable.

	Results

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
NHSBSP at present
Figure 1 shows the manner in which the BRR is expected to alter with increasing breast thickness for the three yearly screening regimen beginning at age 50 years and ending at age 64 years used in the UK. The MGD estimates used in the calculation of these data were made using the methodology described above. It can be seen that the BRR decreases with increasing breast thickness when expressed in terms of either years of life or breast cancer deaths. This is due to the well documented increase in MGD (and therefore relative risk of radiation induced breast cancer) with breast thickness. It is important to note that the benefit of screening is assumed to be constant with compressed breast thickness, as no quantitative information is available regarding the effects of breast size on mammographic benefit. It is also apparent that the BRR is consistently higher when expressed in terms of years of life lost to breast cancer as opposed to breast cancer death. This is a result of the fact that the latent period for radiogenic breast cancer induction is longer than the mammography related lead time after which a reduction in breast cancer mortality first becomes apparent.

View larger version (14K): [in this window] [in a new window]   Figure 1. Benefit:risk ratio (BRR) vs compressed breast thickness for the current NHSBSP regimen. BRR in terms of years of life (YOL) is shown as circles (); BRR in terms of breast cancer mortality is shown as triangles ().

View larger version (16K): [in this window] [in a new window]   Figure 2. Reduction (R) in breast cancer mortality () and years of life lost (YOLL) to breast cancer () vs compressed breast thickness for the current NHSBSP regimen.

Women requiring high dose, multiple film examinations
For the "worst case" of a woman with thick, relatively glandular breasts who required multiple films, the BRR applying to the breast with the highest MGD estimate in the dose survey described by Beckett and Kotre [22] was estimated. The breast in question was 84 mm thick when compressed, and had a glandular content of 39%. The corresponding mean glandular dose estimate was 9.23 mGy per lateral oblique view. The MGD estimate corresponding to the cc view was estimated to be 9.23 mGy x 0.81=7.48 mGy, where the factor 0.81 was estimated from the ratio of mean lateral oblique and cc views reported from the survey of Young and Burch [23]. At the prevalent screen, four films per lateral oblique view and three films per cc view were assumed, in keeping with the maximum number of films per view reported by Young and Burch [23]. Four films per view were assumed at all subsequent screens, resulting in MGD per screen estimates of 59.4 mGy at the prevalent round, and 36.9 mGy at all incident screens. The estimated BRRs in terms of years of life and breast cancer death were 13.9 and 6.8, respectively. The corresponding years of life lost and breast cancer mortality reductions were estimated to be 8.3% and 7.0%. The effect that multiple film examinations may have on BRRs and mortality/years of life lost reductions has not been allowed for in any other section of this work.

Breast implants
The BRR relating to women with augmented breasts attending screening in the NHSBSP was also estimated. MGD estimates were taken from a recent publication [33], and the screening related benefit was assumed to be 44% lower than for normal women, as this is the mean percentage of breast tissue obscured by a prosthesis [34]. In terms of years of life and breast cancer deaths, the BRRs were estimated to be 105 and 54, respectively. The corresponding reductions in breast cancer death and years of life lost to breast cancer were found to be 5.2% and 5.8%.

Effect of alterations in the screening age range
Shown in Figures 3 and 4 are BRRs in terms of deaths and years of life for "average" women, and for a number of different screening age ranges. In each case, the interval between successive screens is 3 years, in keeping with current NHS policy. It can be seen that, of the 24 age ranges considered, the policy with the highest BRR is that of screening throughout the age range 50 years to 70 years. It is important to realise that 70 years is the oldest screening age considered, and that screening to older ages will increase the BRR still further. The graphs also give information on the manner in which the BRR changes with age at first screen. It is apparent that a reduction in the BRR occurs as age at first screen is reduced.

View larger version (17K): [in this window] [in a new window]   Figure 3. Benefit:risk ratio in terms of breast cancer mortality (BRR(d)) vs age at first screen for regimens with final screens at ages 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

View larger version (18K): [in this window] [in a new window]   Figure 4. Benefit:risk ratio in terms of years of life lost to breast cancer (BRR(y)) vs age at first screen for regimens employing a final screen at age 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

As mentioned above, such observations of screening regimens must be considered along with a knowledge of the corresponding female population-wide mortality reductions. These data are shown in Figures 5 and 6. The screening programme that covers the largest age range (40 years to 70 years) results in the greatest overall reductions in breast cancer death and years of life lost to breast cancer, whilst the screening programme covering the shortest age range (50 years to 64 years) leads to the lowest reductions. Regardless of the screening age considered, the beneficial effects of mammography outweigh the associated detriment, therefore this observation is to be expected.

View larger version (16K): [in this window] [in a new window]   Figure 5. Reduction in breast cancer mortality (R(d)) vs age at first screen for regimens employing a final screen at age 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

View larger version (16K): [in this window] [in a new window]   Figure 6. Reduction in the number of years of life lost to breast cancer (R(y)) vs age at first screen for regimens employing a final screen at age 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

View larger version (18K): [in this window] [in a new window]   Figure 7. Mean reduction in breast cancer mortality (R(d)) per screen vs age at first screen for regimens employing a final screen at age 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

View larger version (18K): [in this window] [in a new window]   Figure 8. Mean reduction in the number of years of life lost to breast cancer (R(y)) per screen vs age at first screen for regimens employing a final screen at age 64 years (), 66 years (), 68 years () and 70 years (). The interval between successive screens is 3 years in each case.

View larger version (14K): [in this window] [in a new window]   Figure 9. Benefit:risk ratio in terms of breast cancer mortality (BRR(d)) vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

View larger version (14K): [in this window] [in a new window]   Figure 10. Benefit:risk ratio in terms of years of life lost to breast cancer (BRR(y)) vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

The overall reductions in breast cancer mortality and years of life lost to breast cancer which would occur as a result of the screening strategies discussed are shown in Figures 11 and 12. It can be seen that the largest reductions in these quantities occur at the shortest screening interval considered (1 year), and that increasing the age range for routine screening acts to increase these reductions. This is a consequence of the fact that more frequent screening leads to the detection a greater proportion of breast cancers before they become metastatic, and that screening over a wider age range leads to early detection of a greater number of breast cancers.

View larger version (13K): [in this window] [in a new window]   Figure 11. Reduction in breast cancer mortality (R(d)) vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

View larger version (13K): [in this window] [in a new window]   Figure 12. Reduction in the number of years of life lost to breast cancer (R(y)) vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

View larger version (15K): [in this window] [in a new window]   Figure 13. Mean reduction in breast cancer mortality (R(d)) per screen vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

View larger version (13K): [in this window] [in a new window]   Figure 14. Mean reduction in the number of years of life lost to breast cancer (R(y)) per screen vs interval between successive screens for screening in the age ranges 50–64 years () and 50–70 years ().

	Discussion

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

This information would therefore suggest that mammography screening is justified in that the benefits associated with all of the screening regimens considered outweigh the risks by a substantial margin. It is important to realise that the results presented are imprecise due to the large uncertainties associated with estimates of radiation risk and mammography related benefit. Law [13] states that such data should be considered correct to within no more than one order of magnitude. With this in mind however, it is still true to say that risk is considerably smaller than benefit for all but the "worst case" discussed above.

The screening strategy currently employed by the NHSBSP is not associated with the highest BRR of the screening programmes considered. However, the number of breast cancer deaths induced as a result of this programme is small if not negligible when compared with the number of lives it is expected to save. It is also interesting to note that reducing the age at which screening commences, a subject of current debate, will lead to a reduction in the BRR (Figures 3 and 4). Conversely, increasing the age range for screening above 64 years, a change now being implemented, will tend to increase the BRR. Another observation which is interesting in the context of the NHSBSP is that the 3 yearly screening interval employed at present would appear to lead to the greatest BRR (Figures 9 and 10).

Once the existence of a large difference between the benefits and the risks relating to a given screening regimen is established, it is important to consider the associated reduction in breast cancer mortality. It is apparent from observation of Figures 5, 6, 11, and 12, that the greatest mortality reductions would occur for a screening regimen which incorporates a short screening interval, and a large screening age range. For instance, Figure 12 suggests that a female population-wide mortality reduction of around 20% may be achievable if the present screening age range were to be extended to age 70 years, and an annual screen was employed for all participants. This would however, lead to greatly increased costs and, as Figure 13 indicates, the average reduction in mortality per screen resulting from such a regimen is less favourable than that corresponding to the present screening programme.

The NHSBSP would appear to be capable of causing a breast cancer mortality reduction of approximately 8% over the UK female population. This reduction however, occurs mainly in the age group 55 years to 69 years. The corresponding breast cancer mortality reduction in this age group is approximately 24%, a result which agrees well with the 24% mortality reduction reported by van den Akker-van Marle et al [4].

It has recently been proposed that two view screening should be utilized in the NHSBSP at all screening rounds. This proposal has been made based on the results of two studies in which the extra benefit associated with two view screening has been estimated [32, 35]. It has been found [32] that 24% more women with breast cancer were detected through two view screening. Such practice would therefore appear to increase the benefit attainable as a result of the present screening programme. Indeed, Wald et al [32] state that the use of two view mammography will increase the mortality reductions associated with a one view screening programme by 24%.

It would appear therefore, that the use of two view screening at each round in the NHSBSP will lead to approximately double the risks associated with current practice, and will be associated with an increase in benefit of approximately 24%. The BRRs associated with the current screening programme will therefore be reduced, but will still be well in excess of unity. On the other hand, the overall mortality and years of life lost reductions associated with current practice will be increased by approximately 24%, as the risks associated with the programme will remain extremely small in comparison with the benefits. This is a simplistic attempt to estimate the additional benefits associated with two view screening. As pointed out by Blanks et al [35], there is an increase in the proportion of small cancers detected when two view screening is employed, therefore the additional benefit is likely to be higher than the 24% estimated by Wald et al [32].

It is important to realise that all the observations and results presented here relate to screening programmes that have been implemented for some time, and have therefore reached their full potential, or a "steady state". As pointed out by Blanks et al [36], the NHSBSP is not yet at this stage, and it may take another decade until this happens. They report that a breast cancer mortality reduction of 6.4% in the age range 55 years to 69 years can at present be attributed to screening.

	Conclusion

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 
Analyses based on a biological model of breast screening have been used to obtain estimates of BRR and mortality reduction for the current UK breast screening programme and possible variations on the present screening regimen. The results indicate favourable BRRs and mortality reductions for all women in the programme, with a breast cancer mortality reduction of approximately 8% over the whole UK female population, equivalent to a breast cancer mortality reduction in the region of 24% for the age range presently invited for screening.

	Footnotes

 
This work was supported in part by a Department of Health research contract (number RPK31).

Received for publication April 8, 2002. Revision received September 23, 2002. Accepted for publication January 23, 2003.

	References

Top Abstract Introduction Method Method used for all... Validation of method Results Discussion Conclusion References

 

1. ICRP. Recommendations of the International Commission on Radiological Protection. Ann ICRP 1991;60:No.1–3
2. Department of Health, Social Security (DHSS). Breast Cancer Screening. A report of a Working Group chaired by Sir Patrick Forrest. London: HMSO, 1986.
3. Tabar L, Fagerberg G, Duffy S, Day NE. The Swedish two county trial of mammographic screening for breast cancer: recent results and calculation of benefit. J Epidemiol Community Health 1989;43:107–14.[Abstract/Free Full Text]
4. van den Akker-van Marle E, de Koning H, Boer R, van der Maas P. Reduction in breast cancer mortality due to the introduction of mass screening in the Netherlands: comparison with the United Kingdom. J Med Screen 1999;6:30–4.[Abstract/Free Full Text]
5. Howe GR, McLaughlin J. Breast cancer mortality between 1950 and 1987 after exposure to fractionated moderate-dose-rate ionizing radiation in the Canadian fluoroscopy cohort study and a comparison with breast cancer mortality in the atomic bomb survivors study. Radiat Res 1996;145:694–707.[Medline]
6. BEIR V. National Research Council Committee on the Biological Effects of Ionising Radiations. Health effects of exposure to low levels of ionizing radiation. Washington DC: NAS/NRC, 1990.
7. Feig SA, Ehrlich SM. Estimation of radiation risk from screening mammography: recent trends and comparison with expected benefits. Radiology 1990;174:638–47.[Abstract]
8. Breemsterboer PMM, Warmerdam PG, Boer R, de Koning HJ. Radiation risk of mammography related to benefit in screening programmes: a favourable balance? J Med Screen 1998;5:81–7.[Abstract/Free Full Text]
9. Chiacchierini RP, Lundin FE. Risk-benefit analysis for reduced dose mammography. In: Logan WW, Muntz EP, editors. Reduced Dose Mammography, Mosby 1979:61–72.
10. Law J. Cancers induced and cancers detected in a mammography screening programme. Br J Radiol 1987;60:231–4.[Abstract/Free Full Text]
11. Law J. Variations in individual radiation dose in a breast screening programme and consequences for the balance between associated risk and benefit. Br J Radiol 1993;66:691–8.[Abstract/Free Full Text]
12. Law J. Risk and benefit associated with radiation dose in breast screening programmes—an update. Br J Radiol 1995;68:870–6.[Abstract/Free Full Text]
13. 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]
14. Michaelson JS, Halpern E, Kopans DB. Breast Cancer: Computer simulation method for estimating optimal intervals for screening. Radiology 1999;212:551–60.[Abstract/Free Full Text]
15. Debonis D, Terz JJ, Eldar S, Hill R. Survival of patients with metastatic breast cancer diagnosed between 1955 and 1980. J Surg Oncol 1991;48:158–63.[Medline]
16. Arnesson LG, Smeds S, Fagerberg F. Recurrence-free survival in patients with small breast cancers. Eur J Surg 1994;160:271–6.[Medline]
17. Office for National Statistics. Cancer statistics registrations. London: The Stationary Office, 1999.
18. Office for National Statistics. Population Trends. London: The Stationary Office, 2000.
19. Moskowitz M. Breast cancer: age-specific growth rates and screening strategies. Radiology 1986;161:37–41.[Abstract/Free Full Text]
20. Fox SH, Moskowitz M, Saenger EL, Kevelakes JG, Milbrath J, Goodman MW. Benefit/risk analysis of aggressive mammographic screening. Radiology 1978;128:359–65.[Abstract]
21. Department of Health. Breast Screening Programme, England: 1997–1998. London: Department of Health, 1999. (http://www.doh.gov.uk/public/bcscreen.htm)
22. Beckett JR, Kotre CJ. Dosimetric implications of age related glandular changes in screening mammography. Phys Med Biol 2000;45:801–13.[CrossRef][Medline]
23. Young KC, Burch A. Radiation doses received in the UK breast screening programme in 1997 and 1998. Br J Radiol 2000;73:278–87.[Abstract]
24. Brenner DJ, Amols HI. Enhanced risk from low-energy screen-film mammography X rays. Br J Radiol 1989;62:910–4.[Abstract/Free Full Text]
25. NRPB. Estimates of Late Radiation Risks to the UK Population. Documents of the NRPB 1993; 4 no. 4.
26. UNSCEAR. United Nations Scientific Committee on the Effects of Atomic Radiation. Sources and Effects of Ionizing Radiation, Vol 2: Effects, 2000.
27. Chiang CL. Introduction to Stochastic Processes in Biostatistics. New York: John Wiley and Sons, 1969.
28. World Health Organisation. World Health Statistics Annual. Geneva: World Health Organisation, 1988.
29. Jansen JT, Zoetelief J. Optimisation of breast cancer screening using a computer simulation model. Eur J Radiol 1997;24:137–44.[CrossRef][Medline]
30. Jansen JT, Zoetelief J. Assessment of lifetime gained as a result of mammographic breast cancer screening using a computer model. Br J Radiol 1997;70:619–28.[Abstract]
31. Beemsterboer PMM, Warmerdam PG, Boer R, de Koning HJ. Radiation risk of mammography related to benefit in screening programmes: a favourable balance? J Med Screen 1998;5:81–7.
32. Wald NJ, Murphy P, Major P, Parkes C, Townsend J, Frost C. UKCCCR multicentre trial of one and two view mammography in breast cancer screening. BMJ 1995;311:1189–93.[Abstract/Free Full Text]
33. Beckett JR, Kotre CJ. Estimation of Mean Glandular Dose for mammography of augmented breasts. Phys Med Biol 2000;45:3241–52.[CrossRef][Medline]
34. Silverstein MJ, Handel N, Gamagami P. The effect of silicone-gel-filled implants on mammography. Cancer 1991;68 (Suppl.):1159–63.[CrossRef][Medline]
35. Blanks RG, Moss SM, Wallis MG. Use of two view mammography compared with one view in the detection of small invasive cancers: further results from the NHSBSP. J Med Screen 1997;4:98–101.[Medline]
36. Blanks RG, Moss SM, McGahan CE, Quinn MJ, Babb PJ. Effect of NHS breast screening programme on mortality from breast cancer in England and Wales, 1990–8: comparison of observed with predicted mortality. BMJ 2000;321:665–9.[Abstract/Free Full Text]

This Article Abstract Figures Only 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 Google Scholar Google Scholar Articles by Beckett, J R Articles by Michaelson, J S Search for Related Content PubMed PubMed Citation Articles by Beckett, J R Articles by Michaelson, J S