British Mtthail BulUnn (1991) Vol 47, No. 2, pp. 400-415 O The Britijh Council 1991
Screening for breast cancer NEDay Institute of Public Health, University of Cambridge
Randomised trials of screening have clearly demonstrated a substantial reduction of the order of 40% in breast cancer mortality among women aged over 50 years when regularly screened every two to three years by mammography. No other screening modality has been demonstrated to be of benefit, nor has screening of younger women. The major problem of breast screening at present is the need for greater understanding of the heterogenous natural history of the disease. The extent to which earlier diagnosis improves prognosis is poorly understood, but crucial to determining the full potential of screening.
Survival from breast cancer decreases rapidly with increasing stage of disease. Typical population based figures for 5 year relative survival are 86%, 58%, 46% and 12% for stage I, II, III and IV respectively.1 There is thus considerable potential for reducing population mortality from breast cancer by a systematic approach to improving the stage at presentation by early detection. A range of approaches has been considered including breast self-examination, regular physical examination and a variety of methods of imaging early lesions in the breast, particularly mammography. At present, mammography either alone or in conjunction with physical examination is the only early detection method of proven value for screening. It is in fact the only screening modality for any malignancy for which the value has been demonstrated quantitatively by rigorous randomised trials. Table 1 lists the randomised trials that have been undertaken. There have, in addition, been a number of nonrandomised population based screening programmes which have been used for evaluation, notably in the UK where the populations
45-69 40-47 45-64
40-49
1963
1975
1977
1979
1980
1980
1981
1982
HIP (19)
Malmo (13)
WE (15)
Edinburgh (20)
Canadian I (21)
Canadian II (21)
Stockholm (22)
Gothenburg (Personal Communication)
'Respective avenge for 40-49 and 50-74-year age group 2 Reipective interval for phyiical examination and mammognphy
40-59
40-64
50-59
40-64
Year of start
Study (Ref.)
Age at entry (years)
22 000
40 000
20 000
25 000
23 000
77 000
21000
31000
Invited
1: Phys. exam + mammography 2: Phys. exam. 1: Mammography 2: Control 1: Mammography 2: Control
20 000
20 000 30 000
1: Phys. exam + mammography 2: Phys. Exam, at entry
1: Phys. exam + mammography 2: Control
23 000
25 000
1: Mammography 2: Control
1: Mammography 2: Control
21000 56 000
1: Phys. exam + mammography 2: Control
Screening modality
31000
Control
Number of women
Table 1 Summary of major controlled trials of breast cancer screening
18
28
12
12
12, 242
24, 33'
18, 24
12
Screening interval (months)
JO
w
SCREEN ING FOR BREAST CANC
402
BREAST DISEASE
of certain areas were seleaed for screening with other well-defined areas chosen as comparison groups [UK2, Holland (Utrecht3 and Nijmegen4) and Italy (Florence5)]. Five of the trials listed in Table 1 have reported results on mortality, as given in Table 2. The results for the HIP trial are given for 10 years follow up, to be comparable with the other trials. There is considerable agreement among the trials apart from an aberrantly low reduction in mortality seen in the Malmo trial. This low figure resulted in part from a considerable degree of screening of the control group taking place outside the trial (24% of the control group are reported to have been screened at some time). For all the trials, a lower reduction in breast cancer mortality was seen in women under 50 years of age when entered into the study as given in Table 3 (Malmo gives the results for women aged 45-54 at entry). If the results for the under 50s from the 5 trials are combined naively, one obtains an overall estimate of effect of a reduction in breast cancer mortality of 4.7% with a 95% confidence interval for the relative risk of (0.74, 1.22). This combined analysis takes no account of varying lengths of follow up or degrees of compliance, but is nevertheless a clear demonstration that no appreciable mortality reduction has been achieved in those under 50 years of age with screening frequencies and modalities as given in Table 1. For women over 50 years of age at entry into the trial, all the trials show a reduction in breast cancer mortality ranging from 20% to 40%, a figure which relates to the reduction in the group allocated to screening compared to the control group. Combining the trials naively gives a pooled estimate in the reduction in mortality of 29%, with a 95% confidence interval for the relative risk of (0.61,0.83). These estimates include, of course, a varying degree of compliance over the trials. The degree of compliance in a trial, where the message has to be that the procedure is of unknown benefit, is basically irrelevant in public health terms, since once a procedure is introduced, the message should be that considerable benefit can ensue. The response in the two situations can be expected to differ substantially, as seen by initial findings of the national programme in Britain where responses of 80% or so are seen in some districts, compared to the overall figure of 67% in the UK trial. It is, therefore, of interest to estimate the reduction in mortality seen among women who were actually screened. This figure will be calculated assuming that the underlying breast cancer mortality rate in women who refuse screening is the same as
'Breast oncer cases detected withm 5 years 'Relative benefit at 18 years reduced to 23%
HIF (19) WE (15) Malmo (13) Edinburgh (20) Stockholm (Personal Communication)
1
Study (Ref.)
62 000 133 000 42 000 46 000 60 000
Number of women 65 89 74 61 82
Compliance in study group
Table 2 Percentage reduction of breast cancer mortality in randomised trials of screening
102 10 9 7 7
Follow-up (years)
4 (-35; 32) 17 (-18; 42) 24 (-16; 50)
29 (11; 44) 31 (17; 43)
% Reduction of breast cancer deaths in study group (95% C.I.)
O •n o
45-49 50-64
40-49 50-64
Edinburgh (20)
Stockholm (Personal communication)
'Bremit cmncer cwei detected within 5 yean
45-54 55-69
40-49 50-59 60-69 70-74
40-^9 50-59 60-64
Age at (years)
Malmd (13)
(Personal commun.)
WE
(19)
HIP 1
Study group (Ref.) 39 42 14 37 59 69 47 28 35 13 55 13 55
19844 23 485 23 412 10 339 7981 13 107 5913 17 313 14 375 24 789
Entry Deaths
13 682 12 756 3698
Study Pop.
Control group
7142 12 840
5810 16 094
8082 13 113
15604 16 805 16 269 7307
13 804 12 967 3797
Pop.
- 2 9 (-125; 26) 21 (-24; 49)
22 44
8 22
- 7 (-116; 47) 36 ( - 8 ; 62)
- 2 (-64; 42) 20 ( - 1 7 ; 46)
6 ( - 5 3 ; 43) 38 (11; 57) 41 (18; 58) 19 (-26; 48)
30 67 81 38
13 63
24 ( - 1 6 ; 50) 31 ( - 2 ; 54) 33 (6; 53)
Deaths in study group' (95% C.I.)
51 61 21
Deaths
% Reduction of breast cancel
Population and number of breast cancer deaths
Table 3 Percentage reduction of breast cancer mortality by age in selected screening studies
m
o
SCREENING FOR BREAST CANCER
405
in the control group. Then if a is the degree of compliance, RJOT the reduction in the total study group and RCOMP th e reduction among compliers, one has:
Screening for breast cancer NEDay Institute of Public Health, University of Cambridge
Randomised trials of screening have clearly demonstrated a substantial reduction of the order of 40% in breast cancer mortality among women aged over 50 years when regularly screened every two to three years by mammography. No other screening modality has been demonstrated to be of benefit, nor has screening of younger women. The major problem of breast screening at present is the need for greater understanding of the heterogenous natural history of the disease. The extent to which earlier diagnosis improves prognosis is poorly understood, but crucial to determining the full potential of screening.
Survival from breast cancer decreases rapidly with increasing stage of disease. Typical population based figures for 5 year relative survival are 86%, 58%, 46% and 12% for stage I, II, III and IV respectively.1 There is thus considerable potential for reducing population mortality from breast cancer by a systematic approach to improving the stage at presentation by early detection. A range of approaches has been considered including breast self-examination, regular physical examination and a variety of methods of imaging early lesions in the breast, particularly mammography. At present, mammography either alone or in conjunction with physical examination is the only early detection method of proven value for screening. It is in fact the only screening modality for any malignancy for which the value has been demonstrated quantitatively by rigorous randomised trials. Table 1 lists the randomised trials that have been undertaken. There have, in addition, been a number of nonrandomised population based screening programmes which have been used for evaluation, notably in the UK where the populations
45-69 40-47 45-64
40-49
1963
1975
1977
1979
1980
1980
1981
1982
HIP (19)
Malmo (13)
WE (15)
Edinburgh (20)
Canadian I (21)
Canadian II (21)
Stockholm (22)
Gothenburg (Personal Communication)
'Respective avenge for 40-49 and 50-74-year age group 2 Reipective interval for phyiical examination and mammognphy
40-59
40-64
50-59
40-64
Year of start
Study (Ref.)
Age at entry (years)
22 000
40 000
20 000
25 000
23 000
77 000
21000
31000
Invited
1: Phys. exam + mammography 2: Phys. exam. 1: Mammography 2: Control 1: Mammography 2: Control
20 000
20 000 30 000
1: Phys. exam + mammography 2: Phys. Exam, at entry
1: Phys. exam + mammography 2: Control
23 000
25 000
1: Mammography 2: Control
1: Mammography 2: Control
21000 56 000
1: Phys. exam + mammography 2: Control
Screening modality
31000
Control
Number of women
Table 1 Summary of major controlled trials of breast cancer screening
18
28
12
12
12, 242
24, 33'
18, 24
12
Screening interval (months)
JO
w
SCREEN ING FOR BREAST CANC
402
BREAST DISEASE
of certain areas were seleaed for screening with other well-defined areas chosen as comparison groups [UK2, Holland (Utrecht3 and Nijmegen4) and Italy (Florence5)]. Five of the trials listed in Table 1 have reported results on mortality, as given in Table 2. The results for the HIP trial are given for 10 years follow up, to be comparable with the other trials. There is considerable agreement among the trials apart from an aberrantly low reduction in mortality seen in the Malmo trial. This low figure resulted in part from a considerable degree of screening of the control group taking place outside the trial (24% of the control group are reported to have been screened at some time). For all the trials, a lower reduction in breast cancer mortality was seen in women under 50 years of age when entered into the study as given in Table 3 (Malmo gives the results for women aged 45-54 at entry). If the results for the under 50s from the 5 trials are combined naively, one obtains an overall estimate of effect of a reduction in breast cancer mortality of 4.7% with a 95% confidence interval for the relative risk of (0.74, 1.22). This combined analysis takes no account of varying lengths of follow up or degrees of compliance, but is nevertheless a clear demonstration that no appreciable mortality reduction has been achieved in those under 50 years of age with screening frequencies and modalities as given in Table 1. For women over 50 years of age at entry into the trial, all the trials show a reduction in breast cancer mortality ranging from 20% to 40%, a figure which relates to the reduction in the group allocated to screening compared to the control group. Combining the trials naively gives a pooled estimate in the reduction in mortality of 29%, with a 95% confidence interval for the relative risk of (0.61,0.83). These estimates include, of course, a varying degree of compliance over the trials. The degree of compliance in a trial, where the message has to be that the procedure is of unknown benefit, is basically irrelevant in public health terms, since once a procedure is introduced, the message should be that considerable benefit can ensue. The response in the two situations can be expected to differ substantially, as seen by initial findings of the national programme in Britain where responses of 80% or so are seen in some districts, compared to the overall figure of 67% in the UK trial. It is, therefore, of interest to estimate the reduction in mortality seen among women who were actually screened. This figure will be calculated assuming that the underlying breast cancer mortality rate in women who refuse screening is the same as
'Breast oncer cases detected withm 5 years 'Relative benefit at 18 years reduced to 23%
HIF (19) WE (15) Malmo (13) Edinburgh (20) Stockholm (Personal Communication)
1
Study (Ref.)
62 000 133 000 42 000 46 000 60 000
Number of women 65 89 74 61 82
Compliance in study group
Table 2 Percentage reduction of breast cancer mortality in randomised trials of screening
102 10 9 7 7
Follow-up (years)
4 (-35; 32) 17 (-18; 42) 24 (-16; 50)
29 (11; 44) 31 (17; 43)
% Reduction of breast cancer deaths in study group (95% C.I.)
O •n o
45-49 50-64
40-49 50-64
Edinburgh (20)
Stockholm (Personal communication)
'Bremit cmncer cwei detected within 5 yean
45-54 55-69
40-49 50-59 60-69 70-74
40-^9 50-59 60-64
Age at (years)
Malmd (13)
(Personal commun.)
WE
(19)
HIP 1
Study group (Ref.) 39 42 14 37 59 69 47 28 35 13 55 13 55
19844 23 485 23 412 10 339 7981 13 107 5913 17 313 14 375 24 789
Entry Deaths
13 682 12 756 3698
Study Pop.
Control group
7142 12 840
5810 16 094
8082 13 113
15604 16 805 16 269 7307
13 804 12 967 3797
Pop.
- 2 9 (-125; 26) 21 (-24; 49)
22 44
8 22
- 7 (-116; 47) 36 ( - 8 ; 62)
- 2 (-64; 42) 20 ( - 1 7 ; 46)
6 ( - 5 3 ; 43) 38 (11; 57) 41 (18; 58) 19 (-26; 48)
30 67 81 38
13 63
24 ( - 1 6 ; 50) 31 ( - 2 ; 54) 33 (6; 53)
Deaths in study group' (95% C.I.)
51 61 21
Deaths
% Reduction of breast cancel
Population and number of breast cancer deaths
Table 3 Percentage reduction of breast cancer mortality by age in selected screening studies
m
o
SCREENING FOR BREAST CANCER
405
in the control group. Then if a is the degree of compliance, RJOT the reduction in the total study group and RCOMP th e reduction among compliers, one has:
