lnt. J . Cancer: Supplement 5 , 76-84 (1990)
0 1990 Wiley-Liss, Inc.
Publication of the International Union Against Cancer Publication de I'Union lnternationale Contre 1e Cancer
CHAPTER IV REDUCED BREAST-CANCER MORTALITY WITH MAMMOGRAPHY SCREENING-AN ASSESSMENT OF CURRENTLY AVAILABLE DATA LarS E. RUTQVIST',Anthony B. MILLER*,Ingvar ANDERSON3, Matti HAKAMA4, T h o HAKULINEN', Baldur F. SIGF~~SSON~ and LBszlo TABAR' 'Oncologic Center, Radiumhemmet, Karolinska Hospital, S-104 01 Stockholm, Sweden; 2Department of Preventive Medicine and Biostatistics, Faculty of Medicine, University of Toronto, Toronto, Ontario M5S 1A8, Canada; 3Deparment of Diagnostic Radiology, Malmo General Hospital, S-214 01 Malmo, Sweden; 4University of Tampere, Department of Public Health, PL 607,33101 Tampere, Finland; 'The Finnish Cancer Registry, Elisabetsgatan 21 B , 00170 Helsinki, Finland; 6Department of Mammography, Icelandic Cancer Society, Skbgarhlid 8, IS-I01 Rqkjavik, Iceland; and 'Department of Mammography, Falun Central Hospital, S-791 82 Falun, Sweden. Possible measures to decrease breast-cancer mortality include prevention and improved treatment. Although there may be some strategies that can now be adopted for prevention, their effect will probably be delayed in time. Prevention is also difficult, since it may entail changes in life-style that will not be accepted by all women despite promises of a reduced future risk of breast cancer. Adjuvant systemic therapy has recently been shown to reduce case-fatality during the first 5 years by 20-30% (Early Breast Cancer Trialists' Collaborative Group, 1988). Despite these encouraging early results, many patients still die of their disease and the long-term benefit of adjuvant therapy is not well known. Evaluation of other strategies to improve the outcome of treatment is clearly warranted, e . g . , early diagnosis through mass screening. Most breast-cancer deaths are caused by distant metastases. The probability of distant dissemination is related to the clinical stage at presentation. Many patients with small tumors without axillary nodal involvement achieve long-term survival with local treatment alone. The risk of disease recurrence and early death is considerably higher among patients with large, node-positive tumors. Therefore, it appears reasonable to assume that early detection and treatment should improve outcome. Since this assumption is not self-evident, the applicability of screening as a method to reduce breast-cancer mortality should be evaluated in the context of controlled trials. Many factors other than breast-cancer mortality should be considered in a comprehensive evaluation of screening, e.g., the benefit of being able to use breast-preserving surgery in patients with small tumors, a decreased need of systemic treatment, and a reduced fear of breast cancer among women with a negative screening result, the potential disadvantages of undue anxiety caused by the screening and risk of diagnosing biologically benign lesions as cancer, the cost, inconvenience, and harm of negative biopsies, and the overall cost of the screening program. Several end-points are of interest, such as attendance rate, number of detected cases, stage distribution, interval cancers, mortality and others (Prorok et al., 1984). However, this report will focus on the potential benefit of screening in terms of reduced breast-cancer mortality, since this invariably has been the main objective of the controlled trials. Moreover, breast-cancer death is the only end-point of enough importance to justify the resources spent. SUMMARY OF AVAILABLE STUDIES
Table I presents a summarized description of the major screening studies. The table is not comprehensive. It includes only controlled studies for which mortality data have already been published or will become available in the near future. Most of the studies have been randomized, either individually or by clusters ( e . g . , parishes, municipalities). In the British study, on the other hand, the population of certain areas were selected for screening with the population of other selected areas serving as reference (UK Trial of Early Detection of Breast Cancer Group, 1988). The first trial-the HIP study-started in 1963 and follow-up information is now available up to 18 years postentry. The second generation of trials (Malmo, WE, UK, Scottish, Canadian, Stockholm and Gothenburg trials) started during the late 1970s or early 1980s. Follow-up in these studies is thus generally less than 10 years.
HIP study The study of the Health Insurance Plan (HIP) of Greater New York was a controlled trial in which 62,000 women aged 40-64 years were individually randomized into 2 groups: 3 1,000 were invited to 4 annual screening rounds which included physical examination of the breasts and mammography (the study group), and 31,000 controls who received the normal care offered by the HIP which did not include screening (Shapiro er al., 1988). Of the study group, 65% were screened at least once. Screening took place from 1963 through 1970. After 10 years the breast-cancer mortality was about 29% lower in the study group as compared with the control group (Table 11). The benefit emerged about 3 years post-entry and decreased to about 23% at the end of 18 years. The study group was offered only 4 annual screening rounds with no intervention thereafter. There was no decrease
77
EARLY DETECTION TABLE I - SUMMARY OF MAJOR CONTROLLED TRIALS OF BREAST-CANCER SCREENING
Year of Age at entry initiation (years)
Study
Number Invited
1963 HIP (Shapiro et al., 1988)
40-64
31,000
1976
45-69
21,000
1977
40-74
77,000
I979
45-64
1979
45-64
1980
4w9
25,000
Canadian I1 (Miller, ISl88)
1980
50-59
20,000
Stockholm (Frisell, 1986) Gothenburg (Bjurstam et al.. 19871
1981
40-64
40,000
1982
40-59
22.000
Malmo (7) (Anderssonl et al., 1988) WE (Tab61 et af., 1989) UK (UK Trial of Early Detection of Breast Cancer Group, 1988) Scottish (UK Trial of Early Detection 0 1 Breast Cancer Group, 1988) Canadian I (Miller, 19188)
Of
Screening interval (months)
Study design
Control
Method of allocation
31,000 1: Phys. exam. + mammography 2: Control 21,000 1: Mammography 2: Control
12
Individual random.
18-24
Individual random.
56,000 1: Mammography 2: Control
24, 33'
Cluster random.
12, 242
By domicile
12, 242
Cluster random.
12
Individual random.
12
Individual random.
28
Individual random. Individual random.
1: 46,000 3: 127,000 1: Phys. exam. + 2: 64,000 mammography 2: BSE training ( X 1) 3: Reference population 23,000 1: Phys. exam. + 23,0003 mammography 2: Control 25,000 1: Phys. exam. + mammography 2: Phys. exam. at entry (i.e. x I ) 20,000 1: Phys. exam. + mammography 2: Phys. exam. 20,000 1: Mammography 2: Control 30,000 1: Mammography 2: Control
18
BSE: Breast self-examination.-'Respective average for 4 W 9 and 5C74-year age group.-'Respective interval for physical examination and mammography.-'Included also in the mammography group of the UK trial. TABLE I1 - NUMBERS OF DEATHS IN THE HIP STUDY DUE TO BREAST CANCER BY SELECTED TIME lNTERVALS FROM DATE OF ENTRY. THE STATISTICAL SIGNIFICANCE OF THE PERCENTAGE MORTALITY REDUCTIONS ARE INDICATED. FROM SHAPIRO ET AL. (1988)
s
Breast-cancer deaths throu h following year after entry
Interval to breastcancer diagnosis
Number of breast-cancer
(years)
cases
5
307 30 1
39 63 38*
95 133 29*
126 163 23*
39 63 38*
123 174 29**
180 235 23**
39 63 38*
147 192 23*
260 305 15
Within 5: Study Control % reduction Within 7: Study Control % reduction Within 10: Study Control % reduction
-
43 1 448 -
623 632 -
10
18
'The number of detected breast-cancer cases in the study and control groups was about the same after 5-7 years. Breast-cancer cases detected within 10 years after entry includes cases detected after the conclusion of the screening activities in the study group.-*p < 0.05, **p < 0.01.
in mortality among women aged 4 0 4 9 years at entry during the early follow-up period, whereas there was a major difference for women above 50 years. At longer follow-up, a differential in favor of the study group appeared at entry ages 4 0 4 4 years after about 8 years and at ages 4 5 4 9 years after about 5 years. Since the study was not originally designed to evaluate the effectiveness of screening by age, the differences among the young women were based on small numbers of deaths and were not significant (Table 111). The results are less favourable if the calculations are based on age at diagnosis rather than age at entry, but it may be argued that this approach is biased. The early
78
RUTQVIST ET AL.
TABLE III - NUMBERS OF BREAST-CANCER DEATHS IN THE HIP STUDY BY SELECTED TIME INTERVALS FROM DATE OF ENTRY. ONLY BREAST-CANCER CASES DIAGNOSED WITHIN 5 YEARS ARE INCLUDED. FROM SHAPIRO ETAL. (1988) Interval from entry to death (years): Age at e n y
5
(years)
10
% difference (1 - s/c):
18
Study
Control
Study
Control
Study
Control
9 10 8 7 5
11 9 23 10 10
16 23 23 19 14
21 30 33 28 21
18 31 32 25 20
28 37 41 33 24
40-44 45-49
50-54 55-59 60-44
5 yrs
18 - 11
65 30 50
10 yrs
18 yrs
24 23 30 32 33
36 16 22 24 17
mortality decrease demonstrated among the older women was maintained at long-term follow-up although the relative difference between study and control groups became smaller.
WE study The WE study was a population-based, randomized trial including 135,000 women aged 40-74 years in 2 Swedish counties (Tabk et al., 1989). Randomization took place in the period 1977-80 and was performed at the community level rather than at an individual level. For this purpose, the combined population of the 2 counties was divided into 19 blocks selected to give relative socio-economic homogeneity within each block. Screening with single-view mammography alone was provided over 8 years to 77,000 women (study group). The average screening interval was 24 months in women aged 40-49 years and 33 months in women aged 5&74 years. Of the study group, 89% were screened at least once. The 56,000 women in the control group were not invited to screening. In an 8-year report the results showed a continued significant deficit of breast-cancer deaths in the study group corresponding to 3 1% breast-cancer mortality reduction (Table IV). The differential emerged about 4 years post-entry (Tabk et al., 1985). Subset analyses showed that the effect was significant and greatest in the age-groups 50-59 years and 60-69 years at entry: the respective mortality reduction was 40% and 35% (Table V). At ages 40-49 years and 70-74 years the number of breast-cancer deaths was small and the estimated mortality reduction (8% and 25% respectively) was not statistically significant. However, a chi-square test for heterogeneity of effect in the different age-groups was not significant. In the 40-49-year age group the results appeared to have improved in comparison with a previous 6-year report which had suggested a slight excess of breast-cancer deaths in the study group (Tabk et al., 1985). Malmo study This study was a population-based, randomized trial including 42,000 women aged 45-69 years in the city of Malmo. Randomization was performed at individual level. The study group comprised 21,000 women who were invited to screening with mammography alone. In the first 2 screening rounds, 2 views were used. In the following rounds, 2 views or only one was done, depending on the parenchymal pattern. The screening interval was 18 to 24 months. Of the study group, 74% were screened at least once. The 21,000 women in the control group were not invited to screening (Andersson et al., 1988). After 9 years there was no significant difference in breast-cancer mortality between the study group and the control group: breast-cancer mortality was only about 4% lower in the study group (Table IV) (Andersson er al., 1988). In TABLE IV - PERCENTAGE REDUCTION OF BREAST CANCER MORTALITY IN SELECTED SCREENING STUDIES Study HIPI
% reduction of breast-cancer
Number of women
deaths in study group (95% C.I.)
62,000
lo2
29 ( 1 1 ; 44)
133,000
8
31 (14; 45)
42,000
9
173,000
7
20 ( - 1; 36)
60,000
I
24 ( - 16; 50)
(Shapiro et al., 1988) WE (Tab& et al., 1985) Malmo (Andersson et al., 1988) UK3 (UK Trial of Early Detection of Breast Cancer Group, 1988)
Stockholm (Frisell, 1989)
4 (-35; 32)
'Breast cancer cases detected within 5 years.-*Relative benefit at 18 years reduced to 23%.-3Population offered physical examination plus mammography vs. reference population.
79
EARLY DETECTION TABLE V
-
PERCENTAGE REDUCTION OF BREAST CANCER MORTALITY BY AGE IN SELECTED SCREENING STUDIES
Study
Age at entry (yeas)
Population and number of breast cancer deaths Study group
% reduction of breast-cancer
deaths in study group’ (95% C.I.)
Control group
Pon.
Deaths
POD.
Deaths
4 w 9 50-59 60-64
13,682 12,756 3.698
39 42 14
13,804 12,967 3,797
51 61 21
24 ( - 16; 50) 31 (-2; 54) 33 (6; 53)
40-49 50-59 60-69 70-74
19,844 23,485 23,412 10,339
28 45 52 35
15,604 16,805 16,269 7,307
24 54 58 31
8 40 35 23
45-54 55-69
7,981 13,107
28 35
8,082 13,113
22 44
-29 ( - 125; 26) 21 (-24; 49)
4M9 50-64
14,375 24,789
16 23
7,142 12,840
8 22
- 7 (-116; 47) 36 ( - 8; 62)
HIP^ (Shapiro e/ al., 1988)
WE (Tabk et d., 1985)
(-60; 48) (10; 60) (5; 36) ( - 27; 53)
Malmo
(Andersson ef al., 1988) Stockholm
(Frisell, 1!)89)
--
‘A negative value corresponds to an increase in mortality in the study group.-’The results relate to the 10-year follow-up of breast cancer cases detected within 5 years :after entry.
women aged above 55 years at entry, the estimated reduction of breast-cancer mortality was 21%, whereas among those aged 45-54 years there appeared to be 29% higher breast-cancer mortality in the study group (95%confidence interval of relative risk: 0.74-2.25). None of these results was significant (Table V). The differential among the older women appeared about 7 years post-entry.
UK study Between 1979 and 1981 the United Kingdom Trial of Early Detection of Breast Cancer enrolled women aged 45-64 years living in 8 selected locations in the UK. Annual screening by clinical examination of the breasts, with mammography in alternate years, was provided over 7 years to 46,000 women; 64,000 were offered teaching in breast self-examination (BSE) and were provided with a self-referral clinic. A total of 127,000 women-for whom no extra services were provided-formed a comparison population. The acceptance rate to first screening (or invitation to education) was higher in the areas in which women were offered screening with physical examination and mammlography (60-72%), compared with the BSE areas (3&53%). The crude reduction in breast-cancer mortality during the first 6-7 years in the population offered screening with physical examination and mammography was 14%, compared with the reference population (Table IV) (UK Trial of Early Detection of Breast Cancer Group, 1988). After adjustment for the difference in pre-study breast-cancer mortality the estimated reduction was 20%.This result was not statistically significant. The differential in favor of the study population appeared after about 3 years and it was significant during the period 6-7 years after entry. Subset analyses according to age have not been published. In the BSE districts, no benefit in terms of breast-cancer mortality was observed. Stockholm study This study was a population-based, randomized trial including 60,000 women aged 40-64 years in the south part of the city of Stockholm. Randomization was done at individual level. The study group comprised 40,000 women who were invited to screening with single-view mammography alone. The screening interval was about 28 months. Of the study group, 82% were screened at least once. The 20,000 women in the control group did not receive invitations to screening (Frisell et al., 1986). A preliminary analysis with a mean follow-up of about 7 years showed a trend towards fewer breast-cancer deaths among the women allocated to screening (Frisell, 1989). The estimated reduction of breast-cancer mortality in the screening group at ages 40-64 years was 24%(Table IV). This result was not significant. Subset analyses suggested that the effect was restricted to women aged above 50 years at entry with an estimated reduction of 36%, but this result was also non-significant. No reduction of breast-cancer deaths was observed among women aged 40-49 years at entry (Table V). Mortality data from the Scottish, Canadian (Miller, 1988) and Gothenburg studies (Bjurstam et al., 1987) have not been published. Table I dloes not include studies which used a case-control method, such as the Dutch DOM and Nijmegen studies or the Italian study from Florence (Collette et al., 1984; Verbeek et al., 1984; Palli et al., 1986). In such studies,
80
RUTQVIST ET A L .
screening is generally offered to all women of certain ages in a defined geographical area. The effect of screening is then estimated by calculating the risk of death due to breast cancer among the women who attended screening relative to the non-attenders. These studies have in general replicated the results of the randomized trials, i.e., they have shown a reduction of breast-cancer mortality associated with screening among women aged above 50 years but not among younger women. The case-control method is based on the assumption that the risk of dying from breast cancer is unrelated to the propensity of a woman to attend breast-cancer screening. The HIP study showed lower breast-cancer mortality among non-attenders in the study group as compared with women in the control group (Beahrs et al., 1978). This observation supports the use of the case-control method as the relative risk would-if anything-be underestimated, so that an apparently beneficial effect of screening could not be due to bias. However, the Swedish and UK studies have shown that the non-attenders were at higher risk of dying from breast cancer than the controls. Some non-attenders may have refused the invitation to screening because they feared that a known breast lump would be diagnosed as cancer. Conversely, health-conscious women with a low risk of dying from breast cancer may be more likely to attend screening spontaneously. These circumstances indicate that there is a potential bias in the case-control technique which may-in some settings-result in overestimation of the effect of screening. DISCUSSION
The effect of screening on breast-cancer mortality Comparison of the results of the screening studies is difficult because of differences in setting, attendance, penetration of the intervention into the control group, study design and number of woman-years of follow-up. The HIP study, for instance, compared annual clinical examination plus mammography (with technology of the 1960s) with no intervention. The Swedish studies compared mammography alone (with technology of the 1970s and 1980s) with no intervention. The screening interval was 18 to 24 months in the Malmo study compared with 24 to 33 months in the WE study. In the Malmo study, 24% of the women in the control group had had a clinical mammography, compared with 13% in the WE study. Because of increased public awareness of breast cancer during recent decades, women with breast cancer probably tended to present earlier-xen without screening4uring the 1970s than the 1960s. Therefore, the effect of introducing screening may have been different in the Swedish studies as compared with the HIP study. Despite the differences mentioned, it seems reasonable to conclude that mass screening for breast cancer can achieve a significant reduction in mortality from the disease. The HIP and WE studies have demonstrated reductions of about 30% among women offered screening (Table IV). This is a conservative estimate of the effect of mammography among the women who actually attended screening, because it also includes those who refused to attend. However, it is a true estimate of the effect of the program. Both single-view mammography alone and physical examination plus more complete mammography have thus been shown to be effective as screening modalities. The benefit from adding physical examination to mammography is unknown, as is the effect of using physical examination as the primary screening modality, compared with mammography plus physical examination, a question being addressed in women aged 50-59 years in the Canada I1 trial. The effect of breast-cancer screening on total mortality There has been some controversy over suggestions that screening for breast cancer may increase all-cause mortality, or that the benefit of screening in reducing breast-cancer mortality may have been offset by increases in mortality due to other causes of death. These comments confuse mortality with number or proportion of deaths. There are bound to be more deaths from causes other than breast cancer in a group with years of life saved from breast cancer because of the greater number of woman-years of observation during which these causes operate in the screened group. Eventually, in all groups, the proportion of deaths due to all causes reaches 100%.Recent analyses of the WE data-controlling for county and age-have shown that the relative risk for all-cause mortality in the study group relative to the controls is 0.99 (95% confidence interval: 0.95-1.03) (Tabk et al., 1989). The 1% difference in favor of screening is entirely as expected from the relative importance of breast cancer to other causes of death in women aged 40-74 years. Screening benefit by age The trials with available mortality results were not designed to evaluate the effect of screening by age. Subset analysis of the HIP study generated the hypothesis that the effect on short-term mortality was restricted to women aged above 50 years. The results of the WE, Malmo and Stockholm studies are consistent with this hypothesis, although the numbers in each separate study are too small to permit meaningful conclusions. Table V summarizes the reported mortality reductions by age in the studies mentioned. Since none of the studies in Table V was designed to evaluate age-effects, it has been suggested that they should not be used to reach conclusions on the issue. On the other hand, it may also be reasonable to consider the HIP result as an hypothesis that can be tested in other studies. On this basis, the observation of no early benefit among those
EARLY DETECTION
81
aged 4 W 9 years at entry has been replicated in other trials. A difference in tumor biology between breast cancer in young and old patients and a lower sensitivity of screening in young women could possibly explain a different effect of screening at different ages. It remains to be seen whether long-term follow-up of the more recent trials will show an emerging mortality difference for women aged 40-49 years as suggested in the HIP study (Chu et al., 1988). Two large population-based studies of age-related differences in breast-cancer survival have shown that the outcome is best for patients in their 40s (Host and Lund, 1986; Adami et a l . , 1986). This observation could possibly explain a long latent period before a screening benefit can be demonstrated in young women. In summary, the available data suggest that there is no short-term benefit-in terms of mortality reduction-when screening women aged 40-49 years. Actually, several studies have suggested increased mortality in this age group during the first years of follow-up. On the other hand, an eventual reduction in mortality--even if small-may result in a greater gain of woman-years than mortality reductions at older ages. Longer follow-up of the younger age group is therefore iimportant. If mortality reduction at long-term follow-up is substantiated for young women, it would be appropriate 1.0 evaluate the degree of benefit-in terms of number of woman-years saved-in the same way that is now possible for women aged above 50. Until such data are available, screening of women under the age of 50 should be done only where its effectiveness can be fully evaluated for mortality from breast cancer, i . e . , in the context of a randomized clinical trial or a population-based program with prospective data collection and monitoring and a suitable comparison group. There has been much comment on an apparent difference in the results of the WE and Malmo studies in Sweden. Reasons for this difference include: the lower compliance with invitation to screening in an urban area as distinct from rural areas, a higher background rate of mammography usage in the control population in Malmo, random error due to small1 sample size and number of woman-years of follow-up, and methodological differences between the studies. It may be concluded that the results from the two trials are comparable in showing a reduction in mortality from breast (cancerin women over the age of 50 (in the WE study) or 55 years (in Malmo) and so far no statistically significant reduction in mortality from breast cancer in younger women. Upper age limit for invitation to screening? The best documented short-term breast-cancer mortality reduction with screening is in the 50-69-year age-group. The reduction among those aged 70-74 years at entry in the WE study was lower (23%)compared with those aged 50-69 years (35-40%), possibly because of a lower attendance rate (79% vs. 91%). The problem of competing causes of death also increases with age: the benefit of screening in terms of number of woman-years saved decreases with age, even if the percentage reduction of breast-cancer mortality is the same in all age-groups. Therefore, from society’s point of view, it seems reasonable to give priority to screening women aged below 70 years. Since screening of women in their 60s probably will affect breast-cancer mortality at least up to 10-15 years after cessation of screening, an upper age limit of 69 years can be expected to reduce breast-cancer mortality also among women aged above 70. Frequency of screening? The interval between screening rounds in different studies has varied from 1 to 3 years. There are no experimental data on the benefit of shorter vs. longer intervals. Future trials testing this question will have to be larger than the currently available trials, which were designed to evaluate the overall effectiveness of screening. Analyses of the distribution of interval cancers in the WE study have indicated that the incidence of such cancers approaches the expected incidence-i.e., in the absence of screeningduring the third year after the initial screening round (Tabir et al., 1987). Because of this observation it has been suggested that the screening interval should not exceed 2 years (Anon, 1986). However, until experimental data are available this conclusion should be considered tentative because interval cancers are only a proxy measure of the effectiveness of screening in terms of reduced breast -cancer mortality. Pending further data on the significance of screening intervals, one may speculate whether more benefit is derived from screening a smaller population with shorter intervals than 2 years, as compared with screening a larger population with 2-year or longer intervals. The .si,gn$cance of compliance A screening program must probably achieve a high degree of compliance in order to be maximally effective. The compliance with the first screening round in the HIP study was 65%. In the Swedish studies the figure has generally been higher 74-89%. In the UK study, compliance with screening including physical examination and mammography was hligher (66%) than with BSE instruction (45%). Collaboration with health-promotion experts should be considered to improve attendance rates, particularly since many studies have indicated that non-attenders constitute a high-risk group for fatal breast cancer (UK Trial of Early Detection of Breast Cancer Group, 1988; Tabk et al., 1989; Anderrsson et a l . , 1988). Double readings and multiple views? Double reading of the mammogram and multiple views probably increase both the sensitivity and specificity of the screening. However, there are no experimental data describing to what extent these features contribute to the
82
RUTQVIST ET AL.
reduction of breast-cancer mortality. One may speculate whether more benefit is derived from screening a larger population with single-view mammography than screening a smaller population using double readings and multiple views.
Quality control There are 2 main components of quality control for mammography: ensuring high quality of the films and high quality of the reading of the films. Modem film-screen mammography requires a well-suited mammography machine, careful control of exposure, properly matched films and screens, either vacuum-packed or in special cassettes, a special system for developing the films used only for mammography, a well-trained technician, exquisite attention to positioning, adequate and consistent compression of the breast, to ensure adequate contrast and density of the image (Miller and Tsechkovski, 1987; Tabar and Dean, 1983, 1987, 1989). The systems that were used in the past for screening had several disadvantages, including higher radiation dosage to the breast, breakdown of the processors, and difficulties in quality control. Monitoring for adequate quality requires attention to the physics of the mammography machine and the quality of the individual films. Maintenance of quality control requires regular checks, both done within a department, and utilizing external independent physics centres. A 2-way process is desirable with test films using special phantoms exposed in the physics centre for development in the department and similar films exposed in the department and developed in the physics center, both sets being independently assessed in the physics centre in comparison with preexisting high-quality images (Yaffe et al., 1983). It is well recognized that the quality of the reading of the films is principally dependent on radiologists who are well trained and experienced in mammography screening. Less attention has been paid to external control of the reading (Baines et al., 1986). Such a system was introduced in the National Breast Screening Study in Canada and is being considered in the development of population-based screening programs in that country. In the Swedish programs, double-reading of all films has generally been used and is considered important to ensure good reading quality (Anon, 1986). Both aspects of quality control-film quality and reading quality-are clearly dependent on an adequately trained staff, radiology technicians, radiologists and physicists. Special consideration to the requirements of training is essential when new screening programs are established. Radiation-inducedbreast cancer There are good observational data on breast-cancer induction over a range of high doses (above 0.5-2 Gy). Most data fit a linear relationship between dose and risk. Estimates of the risk of radiation-induced breast cancer with low LET (linear energy transfer) radiation are approximately 400 induced breast cancers per million woman-years per Gy after a latency period of about ten years for women first exposed at the age of 40. Estimates of risk for women exposed at younger ages are higher, some 1,250 induced breast cancers per million woman-years per Gy after a latency period of about 20 years for women first exposed at the age of 10 (Miller et a!., 1989). With the reduction of absorbed dose to the breasts to around 0.3 cGy from a 2-view examination from modem mammography, much of the concern about radiation has dissipated. This has been facilitated by the mentioned observation of a reduced risk of radiation-induced breast cancer in women first irradiated over the age of 40. Gohagen et al. (1986) calculated the possible effect of the American Cancer Society recommendations, assuming a base-line mammogram at age 35 and annual examinations from the age of 40 (Gohagan et al., 1986). Their calculations suggested a lifetime total of about 150 radiogenic cancers compared to about 93,000 otherwise-incident cancers in a population of 1 million women screened by three-view low-dose film-screen mammography. These calculations extrapolated to a population screened by mammography every 2 years from the age of 50 suggest that the effect can be regarded as acceptable in relation to the benefit. Even if this policy was extended to include women screened every 18 months from the age of 40, the numbers of breast cancers induced by mammography would be extremely low. Moreover, these cancers would probably be detected relatively early, as the women in question are those who are likely to attend screening. Eflcacy of other screening modalities The effectiveness of physical examination and BSE education is not well known. The UK study provided some data on BSE, but the results from that study remain inconclusive because of short follow-up and low compliance with BSE teaching. A WHOlUSSR trial of BSE alone should provide data on that modality. Future trials in South America and other third-world countries may provide data on the efficacy of physical examination alone. In summary, screening using physical examination alone or BSE education should be regarded as experimental. CONCLUSIONS
Screening for breast cancer by mammography alone or mammography plus physical examination can reduce mortality from the disease. The largest and most clearly documented benefit is in the age-group 50-69 years: available data from controlled trials indicate that a 3040% reduction in mortality beginning about 5 years after
EARLY DETECTION
83
initiation of screening can be anticipated. However, adherence to strict methods is required to ensure that this degree of benefit is achieved in routine population screening. The benefit among women aged 40-49 years is less clear. No benefit in terms of mortality has been observed in such younger women during 5-8 years after first screening, but may emerge with longer follow-up, as suggested by the HIP results. Pending further data, screening in the age-group 40-49 years should be performed only in the context of a controlled clinical trial or a population-based program that will permit future evaluation of its effectiveness in reducing breast-cancer mortality. The benefit for women aged above 69 years appears to be lower than among those aged 50-69 years, possibly because of lower attendance rates and competing causes of death. The optimal screening interval is not known. Analyses of interval cancers and results of later screening rounds have suggested that the interval should not exceed 2 years. In young women it may be advisable to use 2-view mammography and a shorter interval (e.g., 18 months). However, there are no experimental data describing to what extent these features contribute to the reduction of breast-cancer mortality. Available: data suggest that the risk of radiation-induced breast cancer from mammography screening in women aged 40 or more using modern low-dose techniques is negligible in relation to the potential benefit.
Research issues 1. Long-term benefit of screening with mammography, either alone or in combination with physical examination and BSE education. 2, Benefit associated with physical examination or BSE education alone. 3. Value of initiating screening in different age groups, particularly among women below 50 years. 4. Benefit associated with different screening intervals, multiple vs. single views, and double readings. 5 . Advantages other than mortality reduction (quality of life), and potential disadvantages such as psychological distress, unnecessary biopsies, risk of overdiagnosis, and possible risk of false reassurance from a “negative” screening result. 6 . Cost-benefit analyses in different age-groups and health-care settings. 7. Evaluation of approaches to estimate benefit operationally in screening programs, e.g., compliance to invitations, proportion of the expected incidence found at first screen, interval cancers, rates of advanced disease, and others. 8. Evaluation of strategies to increase compliance with invitation to screening. REFERENCES ADAMI,H.-O., MALKER,B., HOLMBERG, L., PERSON I. and STONEB., The relation between survival and age at diagnosis in breast cancer. N . Engl. J. Med., 315, 559-563 (1986). ANDERSON.I., ASPEGREN,K., JANZON,L., LANDBERG, T., LINDHOLM, K., LINELL,F., LJUNGBERG, O., RANSTAM, J. and SIGF6SSON, €3.. Effect of mammographic screening on breast cancer mortality in an urban population in Sweden. Results from the randomized Malmo Mammographic Screening Trial (MMST). Brit. med. J., 297, 943-948 (1988). ANON,Mammographic screening for early detection of breast cancer (In Swedish). Allmanna r i d fr6n Socialstyrelsen 1986, 3, Swedish National Board of Health and Welfare, Stockholm (1985). BAINES,C.J., MCFARLANE, D.V. and WALL,C., Audit procedures in the national breast screening study: mammography interpretation. J . Can. Ass. Radiol., 37, 256-260 (1987). BEAHRS,O., SHAPIRO,S. and SMART,C. et al., Report of the working group to review the National Cancer Institute-American Cancer Society Breast Cancer Detection Demonstration Projects. J. nut. Cancer Inst., 62, 64&709 (1978). BJURSTAM, N., CAHLIN,E., ERIKSSON, O., HAFSTR~M, L., RuDENSTAM, C.-M. and SAVE-S~DERBERG, J., Breast Cancer Screening Project., Goteborg, Sweden. An eflicient program for breast screening by mammography. Experiences from two screening rounds. (Abstract). 4th EORTC Breast Cancer Working Conference, Lond(on (1987).
CHU, K.C., SMART,C.R. and TARONE,R.E., Analysis of breast cancer mortality and stage distribution by age for the Health Insurance Plan clinical trial. J. nat. Cancer Inst., 80, 1125-1132 (1988). COLLETTE,H.J.A., DAY,N.E., ROMBACH, J.J. and DE WAARD, F., Evaluation of screening for breast cancer in a non-randomised study (the DOM project) by means of a case-control study. Lancet, I, 1224-1226 (1984). EARLYBREASTCANCERTRIALISTS’COLLABORATIVE GROUP.Effects of adjuvant tamoxifen and of cytotoxic therapy on mortality in early breast cancer. An overview of 61 randomized trials among 28,896 women. New Engl. J. Med., 319, 1681-1692 (1988). FRISELL,J., The Stockholm randomized trial of mammography screening. Paper presented at the Nordic Cancer Union Symposium on Breast cancer, Lidingo, Sweden (1989). FRISELL,J., GLAS,U., HELLSTROM, L. and SOMELL,A., Randomized mammographic screening for breast cancer in Stockholm. Design, first round results and comparisons. Breasr Cancer Res. Treat., 8, 45-54 (1986). GOHAGAN, J.K., DARBY,W.P., SPITZNAGEL, E.L., MONSEES, B.S. and TOME,A.E., Radiogenic breast cancer effects of mammographic screening. J . nut. Cancer Inst., 77, 71-76 (1986). HOST, H. and LUND,E., Age as a prognostic factor in breast cancer. Cancer, 57, 2217-2221 (1986). MILLER,A.B., The Canadian National Breast Cancer Screening Study. In: N.E. Day and A.B. Miller (eds.), Screening for breast
84
RUTQVIST ET AL.
cancer, pp. 51-58, published for UICC by H. Huber, Toronto (1988). MILLER,A.B., HOWE, G.R., SHERMAN, G.J., LINDSAY, J.P., YAFFE,M.J., DINNER,P.J., RISCH,H.A. and PRESTON.D., Mortality from breast cancer after irradiation during fluoroscopic examinations in patients being treated for tuberculosis. N. Engl. J. Med., 321, 1285-1289 (1989). MILLER,A.B. and TSECHKOVSKI, M., Imaging technologies in breast cancer control, summary of a report of a World Health Organization Meeting. Amer. J. Radiol., 148, 1093-1094 (1987). PALL],D., ROSSELLIDEL TURCO,M., BUIATTI,E., CARLI,S . , CIATTO,S., TOSCANI,L. and MALTONI,G., A case-control study of the efficacy of a nonrandomised breast cancer screening program in Florence (Italy). In?. J. Cancer, 38, 501-504 (1986). PROROK, P.C., CHAMBERLAIN, J., DAY,N.E., HAKAMA, M. and MILLER,A.B., UICC workshop on evaluation of screening programmes for cancer. In?. J. Cancer, 34, 1-4 (1984). SHAPIRO,S., VENET,W., STRAX,P. and VENET,L., Current results of the breast cancer screening randomized trial: The Health Insurance Plan (HIP) of Greater New York. In: N.E. Day and A.B. Miller (eds.), Screeningfor breast cancer, pp. 3-15, published for UICC by H. Huber, Toronto (1988). TABAR,L. and DEAN,P.B., Screedfilm mammography, quality control. In: Breast carcinoma. Current diagnosis and treatment. Masson, Paris (1983). TABAR,L. and DEAN,P.B., The control of breast cancer through mammography screening. What is the evidence? Radiol. Clin. N. Amer., 25, 993-1005 (1987). TABAR,L. and DEAN,P.B., Optimum mammography technique.
The annotated cookbook approach. Admin. Radiol., 8, 54-56 (1989).
TABAR,L., FAGERBERG, G . , DAY,N.E. and HOLMBERG. L., What is the optimum interval between mammographic screening examinations? An analysis based on the latest results of the Swedish two-county breast cancer screening trial. Brir. J . Cancer, 55, 547551 (1987).
TABAR,L., FAGERBERG, G., DUFFY,S.W. and DAY,N.E., Recent results from the Swedish two-county trial of mammographic screening for breast cancer. J . Epidem. Comm. Hlth, 43, 107-1 14 ( 1989).
TABAR,L., FAGERBERG, C.J.G., GAD, A., BALDETORP, L., HOLMBERG, L.H., GR~NTOFT, O., LJUNGQVIST, U., LUNDSTROM, B., MANSSON,J.C., EKLUND,G., DAY, N.E. and PETTERSSON, F., Reduction in mortality from breast cancer after mass screening with mammography. Randomised trial from the Breast Cancer Screening Working Group of the Swedish National Board of Health and Welfare. Lancet, 11, 829-832 (1985). UK TRIALOF EARLYDETECTION OF BREAST CANCER GROUP.First results on mortality reduction in the UK trial of early detection of breast cancer. Lancet, 11, 411416 (1988). VERBEEK,A.L.M., HENDRICKS, J.H.C.L., HOLLAND,R., MRAVUNAC, M., STURMANS, F. and DAY,N.E., Reduction in breast cancer mortality through mass screening with modem mammography. (First results of the Nijmegen Project 1975-81). Lancet, I, 1222-1224 (1984).
YAFFE,M.J., MAWDSLEY, G.E. and NISHIKAWA, R.M., Quality assurance in a national breast screening study. Proc. SOC.Photo. Opt. Intstr. Eng., 419, 23-30 (1983).
0 1990 Wiley-Liss, Inc.
Publication of the International Union Against Cancer Publication de I'Union lnternationale Contre 1e Cancer
CHAPTER IV REDUCED BREAST-CANCER MORTALITY WITH MAMMOGRAPHY SCREENING-AN ASSESSMENT OF CURRENTLY AVAILABLE DATA LarS E. RUTQVIST',Anthony B. MILLER*,Ingvar ANDERSON3, Matti HAKAMA4, T h o HAKULINEN', Baldur F. SIGF~~SSON~ and LBszlo TABAR' 'Oncologic Center, Radiumhemmet, Karolinska Hospital, S-104 01 Stockholm, Sweden; 2Department of Preventive Medicine and Biostatistics, Faculty of Medicine, University of Toronto, Toronto, Ontario M5S 1A8, Canada; 3Deparment of Diagnostic Radiology, Malmo General Hospital, S-214 01 Malmo, Sweden; 4University of Tampere, Department of Public Health, PL 607,33101 Tampere, Finland; 'The Finnish Cancer Registry, Elisabetsgatan 21 B , 00170 Helsinki, Finland; 6Department of Mammography, Icelandic Cancer Society, Skbgarhlid 8, IS-I01 Rqkjavik, Iceland; and 'Department of Mammography, Falun Central Hospital, S-791 82 Falun, Sweden. Possible measures to decrease breast-cancer mortality include prevention and improved treatment. Although there may be some strategies that can now be adopted for prevention, their effect will probably be delayed in time. Prevention is also difficult, since it may entail changes in life-style that will not be accepted by all women despite promises of a reduced future risk of breast cancer. Adjuvant systemic therapy has recently been shown to reduce case-fatality during the first 5 years by 20-30% (Early Breast Cancer Trialists' Collaborative Group, 1988). Despite these encouraging early results, many patients still die of their disease and the long-term benefit of adjuvant therapy is not well known. Evaluation of other strategies to improve the outcome of treatment is clearly warranted, e . g . , early diagnosis through mass screening. Most breast-cancer deaths are caused by distant metastases. The probability of distant dissemination is related to the clinical stage at presentation. Many patients with small tumors without axillary nodal involvement achieve long-term survival with local treatment alone. The risk of disease recurrence and early death is considerably higher among patients with large, node-positive tumors. Therefore, it appears reasonable to assume that early detection and treatment should improve outcome. Since this assumption is not self-evident, the applicability of screening as a method to reduce breast-cancer mortality should be evaluated in the context of controlled trials. Many factors other than breast-cancer mortality should be considered in a comprehensive evaluation of screening, e.g., the benefit of being able to use breast-preserving surgery in patients with small tumors, a decreased need of systemic treatment, and a reduced fear of breast cancer among women with a negative screening result, the potential disadvantages of undue anxiety caused by the screening and risk of diagnosing biologically benign lesions as cancer, the cost, inconvenience, and harm of negative biopsies, and the overall cost of the screening program. Several end-points are of interest, such as attendance rate, number of detected cases, stage distribution, interval cancers, mortality and others (Prorok et al., 1984). However, this report will focus on the potential benefit of screening in terms of reduced breast-cancer mortality, since this invariably has been the main objective of the controlled trials. Moreover, breast-cancer death is the only end-point of enough importance to justify the resources spent. SUMMARY OF AVAILABLE STUDIES
Table I presents a summarized description of the major screening studies. The table is not comprehensive. It includes only controlled studies for which mortality data have already been published or will become available in the near future. Most of the studies have been randomized, either individually or by clusters ( e . g . , parishes, municipalities). In the British study, on the other hand, the population of certain areas were selected for screening with the population of other selected areas serving as reference (UK Trial of Early Detection of Breast Cancer Group, 1988). The first trial-the HIP study-started in 1963 and follow-up information is now available up to 18 years postentry. The second generation of trials (Malmo, WE, UK, Scottish, Canadian, Stockholm and Gothenburg trials) started during the late 1970s or early 1980s. Follow-up in these studies is thus generally less than 10 years.
HIP study The study of the Health Insurance Plan (HIP) of Greater New York was a controlled trial in which 62,000 women aged 40-64 years were individually randomized into 2 groups: 3 1,000 were invited to 4 annual screening rounds which included physical examination of the breasts and mammography (the study group), and 31,000 controls who received the normal care offered by the HIP which did not include screening (Shapiro er al., 1988). Of the study group, 65% were screened at least once. Screening took place from 1963 through 1970. After 10 years the breast-cancer mortality was about 29% lower in the study group as compared with the control group (Table 11). The benefit emerged about 3 years post-entry and decreased to about 23% at the end of 18 years. The study group was offered only 4 annual screening rounds with no intervention thereafter. There was no decrease
77
EARLY DETECTION TABLE I - SUMMARY OF MAJOR CONTROLLED TRIALS OF BREAST-CANCER SCREENING
Year of Age at entry initiation (years)
Study
Number Invited
1963 HIP (Shapiro et al., 1988)
40-64
31,000
1976
45-69
21,000
1977
40-74
77,000
I979
45-64
1979
45-64
1980
4w9
25,000
Canadian I1 (Miller, ISl88)
1980
50-59
20,000
Stockholm (Frisell, 1986) Gothenburg (Bjurstam et al.. 19871
1981
40-64
40,000
1982
40-59
22.000
Malmo (7) (Anderssonl et al., 1988) WE (Tab61 et af., 1989) UK (UK Trial of Early Detection of Breast Cancer Group, 1988) Scottish (UK Trial of Early Detection 0 1 Breast Cancer Group, 1988) Canadian I (Miller, 19188)
Of
Screening interval (months)
Study design
Control
Method of allocation
31,000 1: Phys. exam. + mammography 2: Control 21,000 1: Mammography 2: Control
12
Individual random.
18-24
Individual random.
56,000 1: Mammography 2: Control
24, 33'
Cluster random.
12, 242
By domicile
12, 242
Cluster random.
12
Individual random.
12
Individual random.
28
Individual random. Individual random.
1: 46,000 3: 127,000 1: Phys. exam. + 2: 64,000 mammography 2: BSE training ( X 1) 3: Reference population 23,000 1: Phys. exam. + 23,0003 mammography 2: Control 25,000 1: Phys. exam. + mammography 2: Phys. exam. at entry (i.e. x I ) 20,000 1: Phys. exam. + mammography 2: Phys. exam. 20,000 1: Mammography 2: Control 30,000 1: Mammography 2: Control
18
BSE: Breast self-examination.-'Respective average for 4 W 9 and 5C74-year age group.-'Respective interval for physical examination and mammography.-'Included also in the mammography group of the UK trial. TABLE I1 - NUMBERS OF DEATHS IN THE HIP STUDY DUE TO BREAST CANCER BY SELECTED TIME lNTERVALS FROM DATE OF ENTRY. THE STATISTICAL SIGNIFICANCE OF THE PERCENTAGE MORTALITY REDUCTIONS ARE INDICATED. FROM SHAPIRO ET AL. (1988)
s
Breast-cancer deaths throu h following year after entry
Interval to breastcancer diagnosis
Number of breast-cancer
(years)
cases
5
307 30 1
39 63 38*
95 133 29*
126 163 23*
39 63 38*
123 174 29**
180 235 23**
39 63 38*
147 192 23*
260 305 15
Within 5: Study Control % reduction Within 7: Study Control % reduction Within 10: Study Control % reduction
-
43 1 448 -
623 632 -
10
18
'The number of detected breast-cancer cases in the study and control groups was about the same after 5-7 years. Breast-cancer cases detected within 10 years after entry includes cases detected after the conclusion of the screening activities in the study group.-*p < 0.05, **p < 0.01.
in mortality among women aged 4 0 4 9 years at entry during the early follow-up period, whereas there was a major difference for women above 50 years. At longer follow-up, a differential in favor of the study group appeared at entry ages 4 0 4 4 years after about 8 years and at ages 4 5 4 9 years after about 5 years. Since the study was not originally designed to evaluate the effectiveness of screening by age, the differences among the young women were based on small numbers of deaths and were not significant (Table 111). The results are less favourable if the calculations are based on age at diagnosis rather than age at entry, but it may be argued that this approach is biased. The early
78
RUTQVIST ET AL.
TABLE III - NUMBERS OF BREAST-CANCER DEATHS IN THE HIP STUDY BY SELECTED TIME INTERVALS FROM DATE OF ENTRY. ONLY BREAST-CANCER CASES DIAGNOSED WITHIN 5 YEARS ARE INCLUDED. FROM SHAPIRO ETAL. (1988) Interval from entry to death (years): Age at e n y
5
(years)
10
% difference (1 - s/c):
18
Study
Control
Study
Control
Study
Control
9 10 8 7 5
11 9 23 10 10
16 23 23 19 14
21 30 33 28 21
18 31 32 25 20
28 37 41 33 24
40-44 45-49
50-54 55-59 60-44
5 yrs
18 - 11
65 30 50
10 yrs
18 yrs
24 23 30 32 33
36 16 22 24 17
mortality decrease demonstrated among the older women was maintained at long-term follow-up although the relative difference between study and control groups became smaller.
WE study The WE study was a population-based, randomized trial including 135,000 women aged 40-74 years in 2 Swedish counties (Tabk et al., 1989). Randomization took place in the period 1977-80 and was performed at the community level rather than at an individual level. For this purpose, the combined population of the 2 counties was divided into 19 blocks selected to give relative socio-economic homogeneity within each block. Screening with single-view mammography alone was provided over 8 years to 77,000 women (study group). The average screening interval was 24 months in women aged 40-49 years and 33 months in women aged 5&74 years. Of the study group, 89% were screened at least once. The 56,000 women in the control group were not invited to screening. In an 8-year report the results showed a continued significant deficit of breast-cancer deaths in the study group corresponding to 3 1% breast-cancer mortality reduction (Table IV). The differential emerged about 4 years post-entry (Tabk et al., 1985). Subset analyses showed that the effect was significant and greatest in the age-groups 50-59 years and 60-69 years at entry: the respective mortality reduction was 40% and 35% (Table V). At ages 40-49 years and 70-74 years the number of breast-cancer deaths was small and the estimated mortality reduction (8% and 25% respectively) was not statistically significant. However, a chi-square test for heterogeneity of effect in the different age-groups was not significant. In the 40-49-year age group the results appeared to have improved in comparison with a previous 6-year report which had suggested a slight excess of breast-cancer deaths in the study group (Tabk et al., 1985). Malmo study This study was a population-based, randomized trial including 42,000 women aged 45-69 years in the city of Malmo. Randomization was performed at individual level. The study group comprised 21,000 women who were invited to screening with mammography alone. In the first 2 screening rounds, 2 views were used. In the following rounds, 2 views or only one was done, depending on the parenchymal pattern. The screening interval was 18 to 24 months. Of the study group, 74% were screened at least once. The 21,000 women in the control group were not invited to screening (Andersson et al., 1988). After 9 years there was no significant difference in breast-cancer mortality between the study group and the control group: breast-cancer mortality was only about 4% lower in the study group (Table IV) (Andersson er al., 1988). In TABLE IV - PERCENTAGE REDUCTION OF BREAST CANCER MORTALITY IN SELECTED SCREENING STUDIES Study HIPI
% reduction of breast-cancer
Number of women
deaths in study group (95% C.I.)
62,000
lo2
29 ( 1 1 ; 44)
133,000
8
31 (14; 45)
42,000
9
173,000
7
20 ( - 1; 36)
60,000
I
24 ( - 16; 50)
(Shapiro et al., 1988) WE (Tab& et al., 1985) Malmo (Andersson et al., 1988) UK3 (UK Trial of Early Detection of Breast Cancer Group, 1988)
Stockholm (Frisell, 1989)
4 (-35; 32)
'Breast cancer cases detected within 5 years.-*Relative benefit at 18 years reduced to 23%.-3Population offered physical examination plus mammography vs. reference population.
79
EARLY DETECTION TABLE V
-
PERCENTAGE REDUCTION OF BREAST CANCER MORTALITY BY AGE IN SELECTED SCREENING STUDIES
Study
Age at entry (yeas)
Population and number of breast cancer deaths Study group
% reduction of breast-cancer
deaths in study group’ (95% C.I.)
Control group
Pon.
Deaths
POD.
Deaths
4 w 9 50-59 60-64
13,682 12,756 3.698
39 42 14
13,804 12,967 3,797
51 61 21
24 ( - 16; 50) 31 (-2; 54) 33 (6; 53)
40-49 50-59 60-69 70-74
19,844 23,485 23,412 10,339
28 45 52 35
15,604 16,805 16,269 7,307
24 54 58 31
8 40 35 23
45-54 55-69
7,981 13,107
28 35
8,082 13,113
22 44
-29 ( - 125; 26) 21 (-24; 49)
4M9 50-64
14,375 24,789
16 23
7,142 12,840
8 22
- 7 (-116; 47) 36 ( - 8; 62)
HIP^ (Shapiro e/ al., 1988)
WE (Tabk et d., 1985)
(-60; 48) (10; 60) (5; 36) ( - 27; 53)
Malmo
(Andersson ef al., 1988) Stockholm
(Frisell, 1!)89)
--
‘A negative value corresponds to an increase in mortality in the study group.-’The results relate to the 10-year follow-up of breast cancer cases detected within 5 years :after entry.
women aged above 55 years at entry, the estimated reduction of breast-cancer mortality was 21%, whereas among those aged 45-54 years there appeared to be 29% higher breast-cancer mortality in the study group (95%confidence interval of relative risk: 0.74-2.25). None of these results was significant (Table V). The differential among the older women appeared about 7 years post-entry.
UK study Between 1979 and 1981 the United Kingdom Trial of Early Detection of Breast Cancer enrolled women aged 45-64 years living in 8 selected locations in the UK. Annual screening by clinical examination of the breasts, with mammography in alternate years, was provided over 7 years to 46,000 women; 64,000 were offered teaching in breast self-examination (BSE) and were provided with a self-referral clinic. A total of 127,000 women-for whom no extra services were provided-formed a comparison population. The acceptance rate to first screening (or invitation to education) was higher in the areas in which women were offered screening with physical examination and mammlography (60-72%), compared with the BSE areas (3&53%). The crude reduction in breast-cancer mortality during the first 6-7 years in the population offered screening with physical examination and mammography was 14%, compared with the reference population (Table IV) (UK Trial of Early Detection of Breast Cancer Group, 1988). After adjustment for the difference in pre-study breast-cancer mortality the estimated reduction was 20%.This result was not statistically significant. The differential in favor of the study population appeared after about 3 years and it was significant during the period 6-7 years after entry. Subset analyses according to age have not been published. In the BSE districts, no benefit in terms of breast-cancer mortality was observed. Stockholm study This study was a population-based, randomized trial including 60,000 women aged 40-64 years in the south part of the city of Stockholm. Randomization was done at individual level. The study group comprised 40,000 women who were invited to screening with single-view mammography alone. The screening interval was about 28 months. Of the study group, 82% were screened at least once. The 20,000 women in the control group did not receive invitations to screening (Frisell et al., 1986). A preliminary analysis with a mean follow-up of about 7 years showed a trend towards fewer breast-cancer deaths among the women allocated to screening (Frisell, 1989). The estimated reduction of breast-cancer mortality in the screening group at ages 40-64 years was 24%(Table IV). This result was not significant. Subset analyses suggested that the effect was restricted to women aged above 50 years at entry with an estimated reduction of 36%, but this result was also non-significant. No reduction of breast-cancer deaths was observed among women aged 40-49 years at entry (Table V). Mortality data from the Scottish, Canadian (Miller, 1988) and Gothenburg studies (Bjurstam et al., 1987) have not been published. Table I dloes not include studies which used a case-control method, such as the Dutch DOM and Nijmegen studies or the Italian study from Florence (Collette et al., 1984; Verbeek et al., 1984; Palli et al., 1986). In such studies,
80
RUTQVIST ET A L .
screening is generally offered to all women of certain ages in a defined geographical area. The effect of screening is then estimated by calculating the risk of death due to breast cancer among the women who attended screening relative to the non-attenders. These studies have in general replicated the results of the randomized trials, i.e., they have shown a reduction of breast-cancer mortality associated with screening among women aged above 50 years but not among younger women. The case-control method is based on the assumption that the risk of dying from breast cancer is unrelated to the propensity of a woman to attend breast-cancer screening. The HIP study showed lower breast-cancer mortality among non-attenders in the study group as compared with women in the control group (Beahrs et al., 1978). This observation supports the use of the case-control method as the relative risk would-if anything-be underestimated, so that an apparently beneficial effect of screening could not be due to bias. However, the Swedish and UK studies have shown that the non-attenders were at higher risk of dying from breast cancer than the controls. Some non-attenders may have refused the invitation to screening because they feared that a known breast lump would be diagnosed as cancer. Conversely, health-conscious women with a low risk of dying from breast cancer may be more likely to attend screening spontaneously. These circumstances indicate that there is a potential bias in the case-control technique which may-in some settings-result in overestimation of the effect of screening. DISCUSSION
The effect of screening on breast-cancer mortality Comparison of the results of the screening studies is difficult because of differences in setting, attendance, penetration of the intervention into the control group, study design and number of woman-years of follow-up. The HIP study, for instance, compared annual clinical examination plus mammography (with technology of the 1960s) with no intervention. The Swedish studies compared mammography alone (with technology of the 1970s and 1980s) with no intervention. The screening interval was 18 to 24 months in the Malmo study compared with 24 to 33 months in the WE study. In the Malmo study, 24% of the women in the control group had had a clinical mammography, compared with 13% in the WE study. Because of increased public awareness of breast cancer during recent decades, women with breast cancer probably tended to present earlier-xen without screening4uring the 1970s than the 1960s. Therefore, the effect of introducing screening may have been different in the Swedish studies as compared with the HIP study. Despite the differences mentioned, it seems reasonable to conclude that mass screening for breast cancer can achieve a significant reduction in mortality from the disease. The HIP and WE studies have demonstrated reductions of about 30% among women offered screening (Table IV). This is a conservative estimate of the effect of mammography among the women who actually attended screening, because it also includes those who refused to attend. However, it is a true estimate of the effect of the program. Both single-view mammography alone and physical examination plus more complete mammography have thus been shown to be effective as screening modalities. The benefit from adding physical examination to mammography is unknown, as is the effect of using physical examination as the primary screening modality, compared with mammography plus physical examination, a question being addressed in women aged 50-59 years in the Canada I1 trial. The effect of breast-cancer screening on total mortality There has been some controversy over suggestions that screening for breast cancer may increase all-cause mortality, or that the benefit of screening in reducing breast-cancer mortality may have been offset by increases in mortality due to other causes of death. These comments confuse mortality with number or proportion of deaths. There are bound to be more deaths from causes other than breast cancer in a group with years of life saved from breast cancer because of the greater number of woman-years of observation during which these causes operate in the screened group. Eventually, in all groups, the proportion of deaths due to all causes reaches 100%.Recent analyses of the WE data-controlling for county and age-have shown that the relative risk for all-cause mortality in the study group relative to the controls is 0.99 (95% confidence interval: 0.95-1.03) (Tabk et al., 1989). The 1% difference in favor of screening is entirely as expected from the relative importance of breast cancer to other causes of death in women aged 40-74 years. Screening benefit by age The trials with available mortality results were not designed to evaluate the effect of screening by age. Subset analysis of the HIP study generated the hypothesis that the effect on short-term mortality was restricted to women aged above 50 years. The results of the WE, Malmo and Stockholm studies are consistent with this hypothesis, although the numbers in each separate study are too small to permit meaningful conclusions. Table V summarizes the reported mortality reductions by age in the studies mentioned. Since none of the studies in Table V was designed to evaluate age-effects, it has been suggested that they should not be used to reach conclusions on the issue. On the other hand, it may also be reasonable to consider the HIP result as an hypothesis that can be tested in other studies. On this basis, the observation of no early benefit among those
EARLY DETECTION
81
aged 4 W 9 years at entry has been replicated in other trials. A difference in tumor biology between breast cancer in young and old patients and a lower sensitivity of screening in young women could possibly explain a different effect of screening at different ages. It remains to be seen whether long-term follow-up of the more recent trials will show an emerging mortality difference for women aged 40-49 years as suggested in the HIP study (Chu et al., 1988). Two large population-based studies of age-related differences in breast-cancer survival have shown that the outcome is best for patients in their 40s (Host and Lund, 1986; Adami et a l . , 1986). This observation could possibly explain a long latent period before a screening benefit can be demonstrated in young women. In summary, the available data suggest that there is no short-term benefit-in terms of mortality reduction-when screening women aged 40-49 years. Actually, several studies have suggested increased mortality in this age group during the first years of follow-up. On the other hand, an eventual reduction in mortality--even if small-may result in a greater gain of woman-years than mortality reductions at older ages. Longer follow-up of the younger age group is therefore iimportant. If mortality reduction at long-term follow-up is substantiated for young women, it would be appropriate 1.0 evaluate the degree of benefit-in terms of number of woman-years saved-in the same way that is now possible for women aged above 50. Until such data are available, screening of women under the age of 50 should be done only where its effectiveness can be fully evaluated for mortality from breast cancer, i . e . , in the context of a randomized clinical trial or a population-based program with prospective data collection and monitoring and a suitable comparison group. There has been much comment on an apparent difference in the results of the WE and Malmo studies in Sweden. Reasons for this difference include: the lower compliance with invitation to screening in an urban area as distinct from rural areas, a higher background rate of mammography usage in the control population in Malmo, random error due to small1 sample size and number of woman-years of follow-up, and methodological differences between the studies. It may be concluded that the results from the two trials are comparable in showing a reduction in mortality from breast (cancerin women over the age of 50 (in the WE study) or 55 years (in Malmo) and so far no statistically significant reduction in mortality from breast cancer in younger women. Upper age limit for invitation to screening? The best documented short-term breast-cancer mortality reduction with screening is in the 50-69-year age-group. The reduction among those aged 70-74 years at entry in the WE study was lower (23%)compared with those aged 50-69 years (35-40%), possibly because of a lower attendance rate (79% vs. 91%). The problem of competing causes of death also increases with age: the benefit of screening in terms of number of woman-years saved decreases with age, even if the percentage reduction of breast-cancer mortality is the same in all age-groups. Therefore, from society’s point of view, it seems reasonable to give priority to screening women aged below 70 years. Since screening of women in their 60s probably will affect breast-cancer mortality at least up to 10-15 years after cessation of screening, an upper age limit of 69 years can be expected to reduce breast-cancer mortality also among women aged above 70. Frequency of screening? The interval between screening rounds in different studies has varied from 1 to 3 years. There are no experimental data on the benefit of shorter vs. longer intervals. Future trials testing this question will have to be larger than the currently available trials, which were designed to evaluate the overall effectiveness of screening. Analyses of the distribution of interval cancers in the WE study have indicated that the incidence of such cancers approaches the expected incidence-i.e., in the absence of screeningduring the third year after the initial screening round (Tabir et al., 1987). Because of this observation it has been suggested that the screening interval should not exceed 2 years (Anon, 1986). However, until experimental data are available this conclusion should be considered tentative because interval cancers are only a proxy measure of the effectiveness of screening in terms of reduced breast -cancer mortality. Pending further data on the significance of screening intervals, one may speculate whether more benefit is derived from screening a smaller population with shorter intervals than 2 years, as compared with screening a larger population with 2-year or longer intervals. The .si,gn$cance of compliance A screening program must probably achieve a high degree of compliance in order to be maximally effective. The compliance with the first screening round in the HIP study was 65%. In the Swedish studies the figure has generally been higher 74-89%. In the UK study, compliance with screening including physical examination and mammography was hligher (66%) than with BSE instruction (45%). Collaboration with health-promotion experts should be considered to improve attendance rates, particularly since many studies have indicated that non-attenders constitute a high-risk group for fatal breast cancer (UK Trial of Early Detection of Breast Cancer Group, 1988; Tabk et al., 1989; Anderrsson et a l . , 1988). Double readings and multiple views? Double reading of the mammogram and multiple views probably increase both the sensitivity and specificity of the screening. However, there are no experimental data describing to what extent these features contribute to the
82
RUTQVIST ET AL.
reduction of breast-cancer mortality. One may speculate whether more benefit is derived from screening a larger population with single-view mammography than screening a smaller population using double readings and multiple views.
Quality control There are 2 main components of quality control for mammography: ensuring high quality of the films and high quality of the reading of the films. Modem film-screen mammography requires a well-suited mammography machine, careful control of exposure, properly matched films and screens, either vacuum-packed or in special cassettes, a special system for developing the films used only for mammography, a well-trained technician, exquisite attention to positioning, adequate and consistent compression of the breast, to ensure adequate contrast and density of the image (Miller and Tsechkovski, 1987; Tabar and Dean, 1983, 1987, 1989). The systems that were used in the past for screening had several disadvantages, including higher radiation dosage to the breast, breakdown of the processors, and difficulties in quality control. Monitoring for adequate quality requires attention to the physics of the mammography machine and the quality of the individual films. Maintenance of quality control requires regular checks, both done within a department, and utilizing external independent physics centres. A 2-way process is desirable with test films using special phantoms exposed in the physics centre for development in the department and similar films exposed in the department and developed in the physics center, both sets being independently assessed in the physics centre in comparison with preexisting high-quality images (Yaffe et al., 1983). It is well recognized that the quality of the reading of the films is principally dependent on radiologists who are well trained and experienced in mammography screening. Less attention has been paid to external control of the reading (Baines et al., 1986). Such a system was introduced in the National Breast Screening Study in Canada and is being considered in the development of population-based screening programs in that country. In the Swedish programs, double-reading of all films has generally been used and is considered important to ensure good reading quality (Anon, 1986). Both aspects of quality control-film quality and reading quality-are clearly dependent on an adequately trained staff, radiology technicians, radiologists and physicists. Special consideration to the requirements of training is essential when new screening programs are established. Radiation-inducedbreast cancer There are good observational data on breast-cancer induction over a range of high doses (above 0.5-2 Gy). Most data fit a linear relationship between dose and risk. Estimates of the risk of radiation-induced breast cancer with low LET (linear energy transfer) radiation are approximately 400 induced breast cancers per million woman-years per Gy after a latency period of about ten years for women first exposed at the age of 40. Estimates of risk for women exposed at younger ages are higher, some 1,250 induced breast cancers per million woman-years per Gy after a latency period of about 20 years for women first exposed at the age of 10 (Miller et a!., 1989). With the reduction of absorbed dose to the breasts to around 0.3 cGy from a 2-view examination from modem mammography, much of the concern about radiation has dissipated. This has been facilitated by the mentioned observation of a reduced risk of radiation-induced breast cancer in women first irradiated over the age of 40. Gohagen et al. (1986) calculated the possible effect of the American Cancer Society recommendations, assuming a base-line mammogram at age 35 and annual examinations from the age of 40 (Gohagan et al., 1986). Their calculations suggested a lifetime total of about 150 radiogenic cancers compared to about 93,000 otherwise-incident cancers in a population of 1 million women screened by three-view low-dose film-screen mammography. These calculations extrapolated to a population screened by mammography every 2 years from the age of 50 suggest that the effect can be regarded as acceptable in relation to the benefit. Even if this policy was extended to include women screened every 18 months from the age of 40, the numbers of breast cancers induced by mammography would be extremely low. Moreover, these cancers would probably be detected relatively early, as the women in question are those who are likely to attend screening. Eflcacy of other screening modalities The effectiveness of physical examination and BSE education is not well known. The UK study provided some data on BSE, but the results from that study remain inconclusive because of short follow-up and low compliance with BSE teaching. A WHOlUSSR trial of BSE alone should provide data on that modality. Future trials in South America and other third-world countries may provide data on the efficacy of physical examination alone. In summary, screening using physical examination alone or BSE education should be regarded as experimental. CONCLUSIONS
Screening for breast cancer by mammography alone or mammography plus physical examination can reduce mortality from the disease. The largest and most clearly documented benefit is in the age-group 50-69 years: available data from controlled trials indicate that a 3040% reduction in mortality beginning about 5 years after
EARLY DETECTION
83
initiation of screening can be anticipated. However, adherence to strict methods is required to ensure that this degree of benefit is achieved in routine population screening. The benefit among women aged 40-49 years is less clear. No benefit in terms of mortality has been observed in such younger women during 5-8 years after first screening, but may emerge with longer follow-up, as suggested by the HIP results. Pending further data, screening in the age-group 40-49 years should be performed only in the context of a controlled clinical trial or a population-based program that will permit future evaluation of its effectiveness in reducing breast-cancer mortality. The benefit for women aged above 69 years appears to be lower than among those aged 50-69 years, possibly because of lower attendance rates and competing causes of death. The optimal screening interval is not known. Analyses of interval cancers and results of later screening rounds have suggested that the interval should not exceed 2 years. In young women it may be advisable to use 2-view mammography and a shorter interval (e.g., 18 months). However, there are no experimental data describing to what extent these features contribute to the reduction of breast-cancer mortality. Available: data suggest that the risk of radiation-induced breast cancer from mammography screening in women aged 40 or more using modern low-dose techniques is negligible in relation to the potential benefit.
Research issues 1. Long-term benefit of screening with mammography, either alone or in combination with physical examination and BSE education. 2, Benefit associated with physical examination or BSE education alone. 3. Value of initiating screening in different age groups, particularly among women below 50 years. 4. Benefit associated with different screening intervals, multiple vs. single views, and double readings. 5 . Advantages other than mortality reduction (quality of life), and potential disadvantages such as psychological distress, unnecessary biopsies, risk of overdiagnosis, and possible risk of false reassurance from a “negative” screening result. 6 . Cost-benefit analyses in different age-groups and health-care settings. 7. Evaluation of approaches to estimate benefit operationally in screening programs, e.g., compliance to invitations, proportion of the expected incidence found at first screen, interval cancers, rates of advanced disease, and others. 8. Evaluation of strategies to increase compliance with invitation to screening. REFERENCES ADAMI,H.-O., MALKER,B., HOLMBERG, L., PERSON I. and STONEB., The relation between survival and age at diagnosis in breast cancer. N . Engl. J. Med., 315, 559-563 (1986). ANDERSON.I., ASPEGREN,K., JANZON,L., LANDBERG, T., LINDHOLM, K., LINELL,F., LJUNGBERG, O., RANSTAM, J. and SIGF6SSON, €3.. Effect of mammographic screening on breast cancer mortality in an urban population in Sweden. Results from the randomized Malmo Mammographic Screening Trial (MMST). Brit. med. J., 297, 943-948 (1988). ANON,Mammographic screening for early detection of breast cancer (In Swedish). Allmanna r i d fr6n Socialstyrelsen 1986, 3, Swedish National Board of Health and Welfare, Stockholm (1985). BAINES,C.J., MCFARLANE, D.V. and WALL,C., Audit procedures in the national breast screening study: mammography interpretation. J . Can. Ass. Radiol., 37, 256-260 (1987). BEAHRS,O., SHAPIRO,S. and SMART,C. et al., Report of the working group to review the National Cancer Institute-American Cancer Society Breast Cancer Detection Demonstration Projects. J. nut. Cancer Inst., 62, 64&709 (1978). BJURSTAM, N., CAHLIN,E., ERIKSSON, O., HAFSTR~M, L., RuDENSTAM, C.-M. and SAVE-S~DERBERG, J., Breast Cancer Screening Project., Goteborg, Sweden. An eflicient program for breast screening by mammography. Experiences from two screening rounds. (Abstract). 4th EORTC Breast Cancer Working Conference, Lond(on (1987).
CHU, K.C., SMART,C.R. and TARONE,R.E., Analysis of breast cancer mortality and stage distribution by age for the Health Insurance Plan clinical trial. J. nat. Cancer Inst., 80, 1125-1132 (1988). COLLETTE,H.J.A., DAY,N.E., ROMBACH, J.J. and DE WAARD, F., Evaluation of screening for breast cancer in a non-randomised study (the DOM project) by means of a case-control study. Lancet, I, 1224-1226 (1984). EARLYBREASTCANCERTRIALISTS’COLLABORATIVE GROUP.Effects of adjuvant tamoxifen and of cytotoxic therapy on mortality in early breast cancer. An overview of 61 randomized trials among 28,896 women. New Engl. J. Med., 319, 1681-1692 (1988). FRISELL,J., The Stockholm randomized trial of mammography screening. Paper presented at the Nordic Cancer Union Symposium on Breast cancer, Lidingo, Sweden (1989). FRISELL,J., GLAS,U., HELLSTROM, L. and SOMELL,A., Randomized mammographic screening for breast cancer in Stockholm. Design, first round results and comparisons. Breasr Cancer Res. Treat., 8, 45-54 (1986). GOHAGAN, J.K., DARBY,W.P., SPITZNAGEL, E.L., MONSEES, B.S. and TOME,A.E., Radiogenic breast cancer effects of mammographic screening. J . nut. Cancer Inst., 77, 71-76 (1986). HOST, H. and LUND,E., Age as a prognostic factor in breast cancer. Cancer, 57, 2217-2221 (1986). MILLER,A.B., The Canadian National Breast Cancer Screening Study. In: N.E. Day and A.B. Miller (eds.), Screening for breast
84
RUTQVIST ET AL.
cancer, pp. 51-58, published for UICC by H. Huber, Toronto (1988). MILLER,A.B., HOWE, G.R., SHERMAN, G.J., LINDSAY, J.P., YAFFE,M.J., DINNER,P.J., RISCH,H.A. and PRESTON.D., Mortality from breast cancer after irradiation during fluoroscopic examinations in patients being treated for tuberculosis. N. Engl. J. Med., 321, 1285-1289 (1989). MILLER,A.B. and TSECHKOVSKI, M., Imaging technologies in breast cancer control, summary of a report of a World Health Organization Meeting. Amer. J. Radiol., 148, 1093-1094 (1987). PALL],D., ROSSELLIDEL TURCO,M., BUIATTI,E., CARLI,S . , CIATTO,S., TOSCANI,L. and MALTONI,G., A case-control study of the efficacy of a nonrandomised breast cancer screening program in Florence (Italy). In?. J. Cancer, 38, 501-504 (1986). PROROK, P.C., CHAMBERLAIN, J., DAY,N.E., HAKAMA, M. and MILLER,A.B., UICC workshop on evaluation of screening programmes for cancer. In?. J. Cancer, 34, 1-4 (1984). SHAPIRO,S., VENET,W., STRAX,P. and VENET,L., Current results of the breast cancer screening randomized trial: The Health Insurance Plan (HIP) of Greater New York. In: N.E. Day and A.B. Miller (eds.), Screeningfor breast cancer, pp. 3-15, published for UICC by H. Huber, Toronto (1988). TABAR,L. and DEAN,P.B., Screedfilm mammography, quality control. In: Breast carcinoma. Current diagnosis and treatment. Masson, Paris (1983). TABAR,L. and DEAN,P.B., The control of breast cancer through mammography screening. What is the evidence? Radiol. Clin. N. Amer., 25, 993-1005 (1987). TABAR,L. and DEAN,P.B., Optimum mammography technique.
The annotated cookbook approach. Admin. Radiol., 8, 54-56 (1989).
TABAR,L., FAGERBERG, G . , DAY,N.E. and HOLMBERG. L., What is the optimum interval between mammographic screening examinations? An analysis based on the latest results of the Swedish two-county breast cancer screening trial. Brir. J . Cancer, 55, 547551 (1987).
TABAR,L., FAGERBERG, G., DUFFY,S.W. and DAY,N.E., Recent results from the Swedish two-county trial of mammographic screening for breast cancer. J . Epidem. Comm. Hlth, 43, 107-1 14 ( 1989).
TABAR,L., FAGERBERG, C.J.G., GAD, A., BALDETORP, L., HOLMBERG, L.H., GR~NTOFT, O., LJUNGQVIST, U., LUNDSTROM, B., MANSSON,J.C., EKLUND,G., DAY, N.E. and PETTERSSON, F., Reduction in mortality from breast cancer after mass screening with mammography. Randomised trial from the Breast Cancer Screening Working Group of the Swedish National Board of Health and Welfare. Lancet, 11, 829-832 (1985). UK TRIALOF EARLYDETECTION OF BREAST CANCER GROUP.First results on mortality reduction in the UK trial of early detection of breast cancer. Lancet, 11, 411416 (1988). VERBEEK,A.L.M., HENDRICKS, J.H.C.L., HOLLAND,R., MRAVUNAC, M., STURMANS, F. and DAY,N.E., Reduction in breast cancer mortality through mass screening with modem mammography. (First results of the Nijmegen Project 1975-81). Lancet, I, 1222-1224 (1984).
YAFFE,M.J., MAWDSLEY, G.E. and NISHIKAWA, R.M., Quality assurance in a national breast screening study. Proc. SOC.Photo. Opt. Intstr. Eng., 419, 23-30 (1983).
