Intended for healthcare professionals


Retrospective study of reasons for improved survival in patients with breast cancer in East Anglia: earlier diagnosis or better treatment?

BMJ 1997; 314 doi: (Published 15 February 1997) Cite this as: BMJ 1997;314:472

This article has a correction. Please see:

  1. Diane Stockton, university research associatea,
  2. Tom Davies, directora,
  3. Nicholas Day, professor and directorb,
  4. Jenny McCann, university research associatea
  1. a East Anglian Cancer Intelligence Unit Department of Community Medicine Institute of Public Health University of Cambridge Cambridge CB2 2SR
  2. b Department of Community Medicine Institute of Public Health University of Cambridge Cambridge CB2 2SR
  1. Correspondence to: Miss Stockton
  • Accepted 29 November 1996


Objectives: To investigate the recent fall in mortality from breast cancer in England and Wales, and to determine the relative contributions of improvements in treatment and earlier detection of tumours.

Design: Retrospective study of all women with breast cancer registered by the East Anglian cancer registry and diagnosed between 1982 and 1989.

Subjects: 3965 patients diagnosed 1982-5 compared with 4665 patients diagnosed 1986-9, in three age groups 0-49, 50-64, ≥65 years, with information on stage at diagnosis and survival.

Main outcome measures: Three year relative survival rates by time period, age group, and stage; relative hazard ratios for each time period and age group derived from Cox's proportional hazards model, adjusted for single year of age and stage.

Results: Survival improved in the later time period, although there was little stage specific improvement. The proportion of early stage tumours increased especially in the 50-64 year age group, and adjustment for stage accounted for over half of the improvement in survival in women aged under 65 years.

Conclusion: Over half of the drop in mortality in women aged under 65 years seems to be attributable to earlier detection of tumours, which has been observed since the mid-1980s. This could have resulted from an increase in breast awareness predating the start of the breast screening programme.

Key messages

  • There has been a similar reduction in mortality from breast cancer in East Anglia and in England and Wales

  • Survival has improved in later years

  • There has been no significant improvement in survival for specific stages of disease

  • There has been a shift to earlier diagnosis of tumours since the mid-1980s; this explains over half of the improvement in survival

  • Reduction in mortality does not seem to be due to improved treatment


In East Anglia in 1993 breast cancer accounted for almost 30% of all malignancies in women (excluding non-melanoma skin cancer) and was attributed as the cause of 591 deaths in a population of just over one million women. The cumulative risk to age 85 years of developing breast cancer is 11.2%–that is, about one in nine women will develop breast cancer by this age.

There has recently been much publicity regarding a reported drop in mortality from breast cancer in England and Wales in the 1990s.1 2 Much of this recent fall may be due to improvements in treatment1 2–for example, more younger women with breast cancer being treated with tamoxifen and chemotherapy.

An alternative hypothesis is that tumours are being detected earlier, thus allowing more effective intervention. In 1986 there was much publicity about breast cancer after the publication of the Forrest report,3 which listed guidelines for a national screening programme. The improvement in public awareness caused by the media coverage at this time could have led to the earlier detection of tumours.

Considering these two hypotheses, if improvements in treatment are the primary cause of the fall in mortality from breast cancer we would expect to see a corresponding fall in stage specific mortality. As the results of the overview of the trials on treatment of breast cancer show the main effect to be on early stage breast cancer,4 one would expect, in particular, to see an overall improvement in survival from stage I and II breast cancers. If, alternatively, changing stage at diagnosis has been the main contribution to the fall in mortality we would expect to see little change in stage specific survival but a shift towards earlier stage at diagnosis. We investigated the relative contributions of improved treatment and earlier diagnosis on the observed reduction in mortality.

The reported drop in mortality began in the years 1989-90. We have, therefore, attempted to relate changes in mortality between the years 1985-8 and 1989-92 to characteristics of patients' tumours at diagnosis and also to the three year survival of women with cancers diagnosed on average three years earlier (that is, between the two time periods of four years–1982-5 and 1986-9). During the later time period there was considerable publicity surrounding breast screening, but the programme itself did not start in East Anglia until April 1989 and then began with the 60-64 year age group. Only 1.5% of cancers diagnosed in the later period had been detected through screening. Thus any changes in stage at presentation between the two time periods are unlikely to be the result of screening.

Subjects and methods

The ascertainment and quality of the data in the East Anglian cancer registry are high, and it is one of the few registries that collects stage information for breast cancer and performs active follow up of patients. All invasive breast cancers diagnosed between 1982 and 1989 and registered by the registry were identified. Active follow up of patients is carried out by the registry three years after diagnosis and then every five years until death, so vital status three years after diagnosis was known for almost all (99.4%) of the identified patients. All the cancers were staged by a clinical oncologist working in the registry.

In the period 1982-5 there were 3965 registered cases of invasive breast cancer, and of these, 14 (0.4%) patients had fewer than three years' follow up. In 1986-9 there were 4665 registered cases, and 36 (0.8%) patients had fewer than three years' follow up. Patients lost to follow up were entered into the analysis as censored cases. Some patients, especially elderly women, did not have enough information in the records for staging. These patients are included in the analysis as unstaged.

Mortality was directly standardised for age with the European standard population. The breast cancer mortality and population data used were obtained from the Office of Population Censuses and Surveys (OPCS). Relative survival estimates, with expected death rates obtained from OPCS life tables, were used to compare the two time periods by age group and stage.5 An analysis of survival with adjustment for age and stage was carried out for each age group to compare the two time periods by using Cox's proportional hazards model.6

Treatment details were available for patients aged 50-69 years. These had previously been extracted from clinical notes and from the cancer registration details as part of the evaluation of the breast screening programme.


Table 1) shows the change in mortality seen in East Anglia between 1985-8 and 1989-92. Corresponding rates for the whole of England and Wales are also given. The fall in East Anglia is comparable with that seen nationally and was confined to women under 65 years of age.

Table 1

Mortality from breast cancer (deaths per 100 000 woman years) by time period and age group, standardised to European standard population

View this table:

Table 2) and figure 1) show the distribution by age and stage of breast cancer cases diagnosed in the two time periods. In the age groups under 65 years there was a considerable increase in early (stage I and II) tumours. The odds ratio for being diagnosed as stage I or II in the later time period compared with the earlier was 1.5 in these younger women (see statistical appendix), representing a considerable stage shift. There was a relatively large proportion of unstaged tumours in women aged 65 years or over at diagnosis, so we could not determine whether the stage distribution had changed in this age group between the two time periods, and analysis of data relating to these women was difficult.

Table 2

Numbers (%) of patients diagnosed with breast cancer by time period, stage, and age group

View this table:
Fig 1
Fig 1

Percentage of patients diagnosed with breast cancer in first (1982-5) and second (1986-9) period by age group and stage

The percentage increases in the three age groups <50, 50-64, and ≥65 years were 18.7%, 18.0%, and 17.0%, respectively, suggesting that the effect of screening in the 50-64 year age group in the second time period has been minimal, despite a 28% increase in the number of cancers diagnosed in women aged 60-64 years (some of whom would have been invited for screening in 1989).

Estimates for relative survival at three years (that is, adjusted for all cause mortality; table 3)) showed an overall improvement in survival of women diagnosed in 1986-9 compared with 1982-5. This improvement was largely confined to women under the age of 65 years; for women aged 65 years or over at diagnosis the change in survival was minimal, reflecting the absence of an effect on mortality in this age group. As expected, there was high survival in those with early stage tumours and poorer survival in those with stage III and IV tumours. There was little change in stage specific survival, however, particularly for stages I and II for which we would expect treatment to be most effective.

Table 3

Percentages of patients (95% confidence intervals) surviving three years from diagnosis (relative survival estimates)

View this table:

Cox's proportional hazards ratios comparing the risk of death within three years after diagnosis of breast cancer in 1986-9 with 1982-5 were calculated overall and for each age group (table 4)). There was a significant (P<0.01) improvement in survival in the later time period when we examined it overall (that is, not taking age or stage into account). This improvement remained after adjustment for single year of age, but when adjustments for stage were also made almost all the improvement disappeared. Thus, adjustment for stage accounted for almost all of the improvement in survival. When we made comparisons specific for age group the difference in survival between time periods was again considerably smaller in each age group after adjustment for stage and was not significant. For the two younger age groups, in which there were few unstaged tumours, adjustment for stage accounted for more than half the improvement in survival.

Table 4

Cox's proportion hazards ratios6 (95% confidence intervals) comparing patients diagnosed in 1986-9 with those diagnosed in 1982-5

View this table:


In East Anglia mortality from breast cancer has fallen in much the same way as it has done nationally. This was not caused by a fall in incidence; indeed this has been rising. The most likely explanations are that treatment has improved and that women are presenting at an earlier stage.

The King's Fund guidelines published in 1986 suggested various treatment protocols that could reduce mortality.7 In particular, they recommended use of tamoxifen; possibly the increased use of tamoxifen may be the major contributor in the improvement of treatment.1 Figure 2 shows the proportion of women in East Anglia aged 50-69 years who received tamoxifen as part of their primary treatment. A clear rise in use in women with early stage cancers is seen. The proportion of patients receiving tamoxifen treatment in East Anglia is low in comparison with that reported in south east England,8 Edinburgh,9 and Yorkshire.10 Although we believe there may have been some under-reporting of use in East Anglia, however, we see no reason why under-reporting should not be consistent over the study period and do not believe it could account for the total difference in reported use between East Anglia and other areas of the United Kingdom.

Fig 2
Fig 2

Percentage of patients aged 50-69 diagnosed with breast cancer who were receiving hormone treatment

In stage I and II tumours the 10 year relative hazard of death for patients with breast cancer who are using tamoxifen is 0.7.4 In the period covered by the data, tamoxifen use has increased in these women from 17% to 30%. Given that the improvement in survival is 30% in those being treated with tamoxifen, however, the overall improvement in survival in this age group would be only 4% and probably less over three years. There has been no significant improvement in stage specific survival in the women aged 50-64 years for early stage cancers (see tables 3 and 4), and it therefore seems unlikely that tamoxifen has improved survival significantly. As the number of patients being treated with tamoxifen in East Anglia increases in line with other regions it is hoped there will be more of a treatment effect. Similar results, showing no improvement in prognosis with increased use of treatment modalities, were found recently in the south eastern Netherlands.11

The incidence of breast cancer has been rising in East Anglia, with the annual rate of rise increasing from 1986 and then again from 1989. All of the increase in incidence is in stage I and II tumours. Reasons for this increase in incidence of early stage tumours from 1986 onwards are unclear. Screening cannot be the cause because it did not start until 1989. The increase could, however, be due to a general increase in breast awareness (with the publication of the Forrest report) causing a reduction in tumour size at diagnosis and hence a decline in mortality from breast cancer.12

In conclusion, earlier diagnosis probably accounts for more than half the improvement in survival in women aged under 65 years. The causes of the rest of the improvement are not clear but probably involve improved and more extensively used effective treatment. Together with the activity of the screening programme these should result in further reductions in mortality from breast cancer in the future.


We acknowledge the invaluable work of the staff of the East Anglian cancer registry and thank them for all their time and effort.

Funding: JM was funded by a public health and operational research grant from the East Anglian Regional Health Authority.

Conflict of interest: None.

Table A1

Calculation of odds ratio

View this table:


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
View Abstract