The influence of breast screening on breast cancer incidence in England: observational study based on cancer registries and bulletins of the NHS Breast Screening Programme

Abstract

Background

To assess the amount of breast cancer overdiagnosis associated with the National Health Service Breast Screening Programme (NHSBSP) that started in 1988 in England.

Methods

First, numbers of breast cancers in women eligible for breast screening not attending screening were estimated for the period 1995–2019, which were extrapolated to all women. A second method was based on ratios of incidence rates of breast cancers in women aged 50–69 to women aged 70 years or more in 1971–1985. The ratio was used for estimating expected numbers of cancers in 1988–2019, and 1995–2019.

Results

From 1995 to 2019, 506,607 non-invasive and invasive breast cancers were diagnosed among women aged 50–64 years (1995–2001) and 50–70 years (2002–2019). A first method estimated that 95,297 cancers were in excess to the number of cancers that would be expected had the NHSBSP not existed. 42,567 screen-detected non-invasive and micro-invasive cancers represented 45.8% of the total excess cancer. 18.8% of all cancers diagnosed among women invited to screening, 25.1% of cancers found in women attending screening, and 35.1% of cancers detected by screening would represent overdiagnosis. A second method estimated that, 18.0% of all cancers diagnosed in 1988–2019, and 18.2% of all cancers diagnosed in 1995–2019 among women invited to screening would represent overdiagnosis.

Conclusion

The two independent methods obtained similar estimates of overdiagnosis. The NHS Breast Screening Programme in England is associated with substantial amount of overdiagnosis.

Introduction

Following the Forest Report,¹ the National Health Service Breast Screening Programmes (NHSBSP) was started in England in 1988. An undesirable consequence of cancer screening is overdiagnosis, which is the screen-detection of cancers that would not progress into symptomatic cancers during lifetime.^2–4 Overdiagnosed breast cancers mainly consist of ductal in situ cancer and small, localized invasive cancers. However, the magnitude of breast cancer overdiagnosis is difficult to assess because one cannot distinguish screen-detected cancers that would progress or not progress into symptomatic cancers.

The study objective was to assess the amount of overdiagnosis associated with the NHSBSP in England using two independent methods that used activity and cancer statistics reported by the NHSBSP, as well as cancers statistics from cancer registries.

Methods

Overview

A first method used the annual Bulletins entitled “Breast Screening Programme, England” published from 1995 to 2019 and cancer registries for estimating the incidence rates of cancers in women in age groups invited to screening but who did not attend screening. Then, assuming that these incidence rates would have been observed in all women if the NHSBSP had not existed, we estimated the expected numbers of cancers in age groups invited to screening. The excess number of cancers due to screening was computed as the difference between numbers of cancers reported to cancer registries and numbers of cancers expected in the absence of the NHSBSP.

A second method was based on the observation that from 1971 to 1985, incidence rates of breast cancers in England were increasing in all age groups. We computed the expected numbers of cancers in age groups invited to screening if ratios of incidence rates between age groups in 1971-1985 had remained unchanged in 1988–2019.The excess numbers of cancers due to the NHSBSP were computed as the difference between cancers reported to cancer registries in 1988–2019 and expected numbers of cancers obtained using ratios of incidence rates in 1971–1985.

Because data used for the study were retrospective, anonymous and publicly available, prior approval by an ethics committee was not required.

The NHS breast screening programmes

The NHSBSP in England started in 1988 with inviting eligible women aged 50–64 years.⁵ By 1993, all eligible women had been invited at least once for screening. Until 2000, women received a maximum of 5 invitations to screening. A phased extension of the upper age limit to the 71st birthday took place between 2001 and 2006, which represented two additional invitations to screening. In December 2007, the age at which women are invited to screening was extended again at selected pilot sites, from 47 to 73 years of age, thus incorporating nine invitations to screening. Participation to screening has been relatively stable over time, with 70–73% of invited women attending screening (further information in Supplementary Materials 1).

Data sources

This study used data from publicly available sources. Demographic data and breast cancer incidence data (ICD-10 codes C50 and D05) by year and by 5-year age group were obtained from the Office of National Statistics⁶ and cancer registration statistics.⁷ All cancers diagnosed in England are reported to cancer registries. The quality of cancer registration in England is closely monitored and cancer statistics are complete since 1971.⁸ The Bulletins from 1995 to 2019 were retrieved from NHS Digital.⁹

The Bulletins report data on NHSBSP activities running from 1 April to 31 March of next year. The Bulletins include data on (i) the numbers of women resident in England eligible for screening who had at least one screen in the last three years, (ii) the numbers of women invited to screening, and (iii) the numbers of screen-detected breast cancers. The numbers of invasive, and of non-invasive plus microinvasive cancers, are reported separately. The format and level of detail of data reported in the Bulletins has remained fairly constant over time.

From 1995 to 2001, bulletins displayed data for women aged 45–49, 50–54, 55–59, 60–64, 65–69, and ≥70. Starting in 2002–2003, data were displayed for women aged 45–49, 50–54, 55–59, 60–64, 65–69, 70, 71–74 and ≥75. Hence, 1999–2001 were transition years during which numbers of women eligible for screening raised from 4,373 600 in 2000–2001 to 5,783 000 in 2001–2002. Numbers of screen detected breast cancers among women 50–64 years of age in 1991–1993 were found in Reference 10. Similar data for 1994 were not found.

Breast cancer incidence statistics

Breast cancer incidence rates in women aged 30–49, 45–49, 50–64, 65–69 years, 50 years or more and 70 years or more, were computed for each year using numbers of cancers recorded in cancer registries as numerators and demographic data as denominators. There was no adjustment for age.

Subtracting the numbers of screen-detected cancers reported in Bulletins from cancers reported by cancer registries returned the numbers of symptomatic cancers, which encompass the interval cancers (cancers diagnosed between two screening rounds), and cancers diagnosed in women not attending screening. Bulletins published from 2002 to 2003 onwards reported screen-detected cancers in women aged 65–70 years. In order to accommodate cancer numbers for same age groups, one sixth of screen-detected cancers reported for the group 65–70 years in 2002–2019 was transferred to screen-detected cancers reported for women aged 70 years or more.

Estimation of excess breast cancers

First method based on bulletins

Estimations of overdiagnosis were done for women 50–64 years of age between 1995 and 2001, and for women 50–70 years of age from 2002 to 2019. We did not use Bulletins covering the period April 2019–March 2021 because of the disruption caused by the covid-19 pandemic.

Parameters and computations listed in table 1 were done year by year over the 25-year period. Women eligible for screening (row 1) are women resident in England, after exclusion of women not eligible for screening, e.g. because of previous bilateral mastectomy or women who decided to remove themselves from the screening programme. Women with less than three years since the last screening are women who have been screened according to the NHSBSP criteria (row 2). Subtracting row 2 from row 1 gives the numbers of women who had no record of screening over the last three years (row 3). Unscreened women include women who never attended screening plus women with more than three years since last screening. The numbers of women screened (row 4) include women referred by GPs or self-referred to screening. Numbers of women in rows 1, 2 and 4 look enormous because a same woman could be counted several times in successive Bulletins. The numbers of women actually invited to screening each year are not displayed in table 1 because these women consist of a mix of women with diverse screening history (i.e. women invited for their first screening, women with last screen in the last five years or women with no screening record) that do not represent a suitable denominator for cancers found among unscreened women (row 3).

Table 1.

Parameters and computations for estimating the excess breast cancers in women, England, 1995–2019. In situ and invasive breast cancers in women aged 50–64 years from 1995 to 2001; 50–70 years from 2002 to 2019.

Row	Parameters	Source or computation	Cumulative numbers, or %
			1995–2019	1995–2006	2007–2019
1	Number of women eligible for screeningb	Bulletins	140 879 801	60 283 600	80 596 201
2	Number of women screened in the last three yearsb	Bulletins	97 654 718	39 553,300	58 101,418
3	Number of unscreened womena,b	= Row1 − Row2	43 225 083	20 730 300	22 494 783
4	Number of women screened including self/GP referralb	Bulletins	35 982 948	14 854 243	21 128 705
5	Number of cancers diagnosed in all women invited to screening	Cancer registries	506 607	203 780	302 827
6	Number of screen-detected cancers	Bulletins	271 128	101 182	169 946
7	Number of interval cancers	Estimation (see text)	109 280	44 666	64 614
8	Number of cancers diagnosed in screened women	= Row 6 + Row 7	380 408	145 848	234 560
9	Number of cancers diagnosed in unscreened women	= Row 5 − Row 8	126 199	57 932	68 267
10	Number of cancers in women invited to screening in the absence of screening programme	= Row 9 × (Row 1/Row 3)c	411 310	168 466	244 593
11	Excess number of cancers among all women invited to screening	= Row 5 − Row 10	95 297	35 314	58 234
12	Excess proportion of screen-detected cancers (%)	= Row 11 × 100/Row 6	35.1%	34.9%	34.3%
13	Excess proportion of cancers in screened women (%)	= Row 11 × 100/Row 8	25.1%	24.2%	24.8%
14	Excess proportion of cancers among all cancers diagnosed in women invited to screening (%)	= Row 11 × 100/Row 5	18.8%	17.3%	19.2%

^aUnscreened women are women without a record of at least one screening mammography in the last three years.

^bCumulative numbers consist of women who are often included several times in statistics reported in the Bulletins.

^cNumbers in 1995–2006 and 2007–2019 do not add to 411 310 because of the overlap between periods for cancer reporting to cancer registries (January to December) and data reporting in Bulletins (April to March).

The total numbers of breast cancers diagnosed in women resident in England are known from cancer registries (row 5). Because after 2001, women 70 years of age were invited to screening, one fifth of the numbers of women aged 70–74 diagnosed with a breast cancer was added to numbers of women aged 65–69 years diagnosed with a breast cancer.

Numbers of screen-detected cancers, including cancers found following GP or self-referral to screening were extracted from the Bulletins (row 6).

Numbers of interval breast cancers (IBC) for years between 1997 and 2019 were estimated using IBC rates reported by two large cohort studies conducted within the NHSBSP in 1997–2003 and in 2005–2008^11,12(row 7). IBC rates per 1,000 screened women were 0.52 for months 1–11 after screening, 1.20 for months 12–23 after screening and 1.37 for months 24–35 after screening. Because numbers of screened women before 1995 were unknown, the numbers of interval breast cancers in 1995 and 1996 were estimated using the IBC rate of 3.1 per 1,000 screened women.¹³

Adding numbers of screen-detected and of interval breast cancers yielded total numbers of cancers diagnosed among screened women (row 8).

The difference with all breast cancers reported to cancer registries corresponded to breast cancers diagnosed among women who were not screened, regardless of whether they were invited or not invited to screening (row 9).

The extrapolation of numbers of breast cancers diagnosed among unscreened women to the whole population of women eligible for screening was done by multiplying the number of cancers found among unscreened women by the total number of women eligible for screening, divided by the number of unscreened women (row 10).

Subtracting the numbers of cancers expected in the absence of screening (row 10) from all cancer registered in England (row 5) provided an estimate of the number of breast cancers that would not have been diagnosed in the absence of the NHSBSP (row 11).

In a sensitivity analysis, we assessed the influence of women who were self-referred or referred by GPs to screening because they had breast symptoms. The numbers of breast cancers detected in self or GP referred women were counted from Bulletins. Because several Bulletins did not report in the same way age groups for screen-detected cancers and for cancers detected among self or GP-referred women, these calculations were restricted to the period 2007–2019. Interval breast cancers were estimated using the ratio of 0.38 between numbers of interval and of screen-detected cancers during the period 2007–2019.

Second method

Descriptive analyses in Supplementary figures S1–S4 showed that from 1971 to 1985, the incidence of breast cancer was increasing in all age groups, but increases were more pronounced in women aged 50 years or more than in younger women. During that 15-year period, the ratios of incidence rates in the 50–69 age groups to the incidence rates group aged 70 years or more remained fairly constant at around 0.73. The ratios of incidence rates in the 50–69 age group to the 40–49 age group, and to the less than 50 years age group remained at around 1.41 (=1/0.71) and 6.03, respectively.

This method estimated the expected numbers of cancers in women aged 50–69 years after 1987 if the ratio of incidence rates observed in 1971–1985 between these women and women of other age groups had persisted over time.

Using cancer registry data, the expected cancer incidence rate in women aged 50–69 for each year i from 1971 to 2019 was estimated as follows:

$$\text{EIR}i(50-69)=\text{OIR}i\left(\ge 70\right)*0.73,$$

where OIRi (≥70) were the observed cancer incidence rates derived from cancer registries. Then, for each year i between 1971 and 2019, the expected numbers of cancer (ENCi) in women 50–69 years were computed as follows:

$$\text{ENC}i(50-69)=\text{ONC}i(50-69)*\text{EIR}i(50-69)/\text{OIR}i(50-69),$$

where ONCi (50–69) and OIRi (50–69) were the observed numbers and incidence rates of cancers derived from cancer registries, respectively. Observed (ONCi) and expected (ENCi) numbers of cancers in women aged 50–69 years were summed up for the periods 1988–2019 and 1995–2019. The difference ONCi (50–69) minus ENCi (50–69) was the estimated number of excess cancers associated with screening in women aged 50–69 years.

Likewise, taking the ratio of incidence rates of 6.03 between women aged 50–69 years and women aged less than 50 years in 1971–1985:

$$\text{EIR}i(50-69)=\text{OIR}i\left(<50\right)*6.03,$$

Further computations were the same, using OIRi(<50) instead of OIRi (50–69).

Results

 

All breast cancer trends

Incidence rates of breast cancer in England have constantly increased over time (figure 1). The steeper increase starting in 1985 which preceded the start of breast screening programmes in 1988 was probably the consequence of pilot screening programmes.¹ Breast cancer incidence rates among women aged 50 years or more have increased from 164 per 100 000 in 1971 to 240 per 100 000 in 1985 (46% increase), and then to 413 per 100 000 in 2019 (152% increase since 1971). In women 30–49 years of age, incidence rates increased from 72 per 100 000 in 1971 to 129 per 100 000 women in 2019 (79% increase). Before 1985, about 3% of all cancers in women aged 50 years or more were non-invasive (data not shown). After 1988, that proportion steadily increased for attaining 13% in 2015 (data not shown). The sharp drops in incidence in 2020 reflect the impact of the covid-19 pandemic on society and health services.

Figure 1.

Incidence of non-invasive and invasive breast cancers in women aged 30–49 years and 50 years or more in England, 1971–2020.

Screen-detected and symptomatic cancers in England

The contributions of screen-detected and of symptomatic cancers to incidence rates are depicted in figure 2. In each age group, the difference between all cancer incidence rates and symptomatic cancer incidence rates equates to the incidence rates of screen-detected cancers. Because a minority of women of age groups 45–49 and 70 or more was invited to screening, there were less screen-detected cancers in these age groups. Consequently, the differences between incidence rates of all cancers and of symptomatic cancers were smaller than in other age groups.

Figure 2.

Incidence trends of non-invasive and invasive breast cancers in women aged 45–49, 50–64, 65–69 and ≥70 years, England, 1971–2020. Plain traits are age-specific incidence trends of all breast cancers, and dashed traits are age-specific incidence trends of symptomatic breast cancers. The differences between two age-specific curves are incidence trends of screen-detected breast cancers.

From 1971 to 1985, the incidence rates of all cancers in age groups 40–49, 50–64, 65–69, and 70 years or more increased in parallel. Incidence rates were on average 27% lower in women aged 50–64 years than in women aged 70 years or more. Steep increases in rates increased occurred after 1985, and around 2000, rates in the 50–64 group stabilized at rates close to rates in the 70 years or more group. After 2002, incidence rates in women 65–69 surpassed rates in women aged 70 or more and stayed at a higher level until 2019.

In women 50–64 years, incidence rates of symptomatic cancers in 1990–1992 and after 1994 were similar. After 2005, incidence rates in women aged less than 50 years and 70 or more were gradually pushed up by screen-detected cancers. Hence, increases in cancer incidence rates after 1985 were essentially due to increases in screen-detected cancers.

First method

From 1995 to 2019, 506 607 cancers were diagnosed among women aged 50–64 years (1995–2001) and 50–70 years (2002–2019) resident in England (table 1). Of these, 95 297 cancers were in excess to the number of cancers that were expected had the NHSBSP not existed. Overall, the excess numbers of cancers represented 35.1% of screen-detected cancers, 25.1% of cancers diagnosed in women attending screening, and 18.8% of cancers diagnosed in all women. Overdiagnosis estimates were slightly higher for the period 2007–2019 than for the period 1995–2006.

Non-invasive and microinvasive cancers represented 21.2% of all screen-detected cancers, while they represented 5.5% of symptomatic cancers. Estimates were that 42 567 screen-detected non-invasive and micro-invasive cancers represented 45.8% of the total excess cancer.

In 2007–2019, 479 322 (2.3%) of 21 128 705 women aged 50–70 years were self or referred by GPs to screening. The number of screen-detected breast cancers in these women was 5013 and the number of interval breast cancers was estimated as 1905. The 6918 breast cancers found in self- or GP-referred women represented 2.9% of the 234 561 breast cancers found in all women who attended screening in 2007–2019.

Second method

Computations summarized in table 2 estimated 18.0% excess cancers in women aged 50–69 years for the 1988–2019 period, and 18.2% for the 1995–2019 period. The observed and expected cancer incidence rates in women aged 50–69 years are displayed in Supplementary figure S5. The variability of incidence rates in women aged 70 years or more after 2013 (figure 2) stirred instable expected incidence rates in women aged 50–69. Taking women less than 50 as comparison group, computations estimated 15.5% excess cancers in women aged 50–69 years for the 1988–2019 period, and 15.6% for the 1995–2019 periods.

Table 2.

Observed and expected numbers of non-invasive and invasive breast cancers among women aged 50–69 years residing in England.

Parameters	Numbers and % for comparisons with women aged
	70 years or more	Less than 50 years
1988–2019
Observed numbers of cancers in women aged 50–69 years	633 319	633 319
Expected numbers of cancers in women aged 50–69 yearsa	519  399	535 460
Difference observed minus expected cancers	113  920	97 859
Proportion of excess cancers (%)	18.0%	15.5%
1995–2019
Observed numbers of cancers in women aged 50–69 years	537 908	537  908
Expected numbers of cancers in women aged 50–69 yearsa	440 063	453 908
Difference observed minus expected cancers	97 845	84 064
Proportion of excess cancers (%)	18.2%	15.6%

^aExpected numbers computed year by year using a ratio of 0.73 for incidence rates in women aged 50–69 years to incidence rates in women aged 70 years or more, and a ratio of 0.166 for incidence rates in women aged 50–69 years to incidence rates in women aged less than 50 years.

Discussion

This study has used two different methods for estimating the amount of non-invasive and invasive breast cancer overdiagnosis associated with the NHSBSP in England. The first method based on administrative records of the NHSBSP and cancer registries, has never been used before. Regarding the second method, because incidences rates in women aged 50–69 years and in women aged 70 years or more increased nearly at the same pace in 1971–1985, estimates based on these two age groups are the most credible. In summary, about 18% of all cancers diagnosed in women in women invited to screening, 25% of cancers found in women attending screening, and 35% of cancers detected by screening would represent overdiagnosis. Because data reported in Bulletins for age groups not eligible for screening were less detailed, the number of women with overdiagnosed cancers would in fact be greater if the study had had the possibility to appropriately use data of screen-detected cancers in women aged less than 50 years or 71 years or more. The NHSBSP invites women every three years to screening. Most probably that overdiagnosis would be higher if invitation to screening was every year or every two years.

Self-selection

In the absence of screening mammography, the incidence of breast cancer is known to be higher among socially advantaged women than among socially disadvantaged women.¹⁴ Also, socially advantaged women are more inclined to attend screening than socially disadvantaged women (self-selection).¹⁵ In the setting of studies on the effect of screening on cancer-specific mortality, adjusting for self-selection is warranted because the effects of socio-economic status on cancer-specific mortality existed before the start of screening programmes. In the setting of studies on the effect of screening on overdiagnosis, adjusting for self-selection is not warranted because it is impossible to have an overdiagnosed breast cancer without attending screening mammography. Hence, screening attendance is a causal intermediate between socio-economic status and overdiagnosis. Controlling for self-selection in this setting would lead to overadjustment bias.¹⁶

In addition, two Swedish randomized trials reported breast cancer numbers without those cancers found at first screening of the control group. The cumulative incidence of breast cancer in women of the intervention group who did not attend screening and of women in the control group were 10.6 and 10.9 per 1,000, respectively, in the Two-County trial,¹⁷ and 19.5 and 21.1 per 1,000, respectively, in the Malmö trial.¹⁸ Hence, these two trials did not suggest meaningful differences in breast cancer incidences observed among women not attending screening and women not invited to screening.

Strengths and limitations

Main strengths of our study are first that the two methods are based on reliable observational data and do not rely on statistical modelling necessitating unverifiable assumptions. Second, the first method rests on use of the Bulletins that kept the same format for data reporting, which enabled a consistent appraisal of activities and outcomes over the entire 25-year period. Third, the second method is based on high-quality cancer registries.

Regarding limitations, in the first method, cancer registries reported data by calendar year, while the Bulletins reported data from April to March. The overlap in reporting periods had probably a minor influence on estimates because during the 25-year long period of the study, temporal disagreements between data sources compensated each other from one year to the next. The differences in age group definition between cancer registries and the Bulletins required to adjust numbers of cancers in order to make correspond age groups, especially the 50–69 and 50–70 year age groups.

Cancers found in women self or GP-referred to screening represented a small proportion of all cancers found in women attending screening upon invitation. These cancers are unlikely to have markedly influenced estimates in table 1.

The limitation of the second method is the assumption that in the absence of screening, that ratio of incidence rates between age groups would have remained parallel until 2019.

Comparison with the previous estimations

The Independent UK Panel on Breast Cancer Screening estimated that 26% of screen-detected breast cancers represent overdiagnosis.⁴ The Panel estimates were based on the Swedish Malmö randomized trial (women aged 55–69 years) and the two Canadian randomized trials (women aged 40–59 years). The older age of women invited by the NHSBSP, and replacement of film-mammography with digital mammography after 2010 may explain why our study estimated that 35% of screen-detected breast cancers represented overdiagnosis.

Long-term temporal trends

Increases in incidence rates after start of the NHSBSP were greater than expected.¹⁹ However, prospects were that after completion of the prevalence rounds in women aged 50–64 years in 1994, rates would return to levels expected if pre-screening rates had persisted over time.^20–22 But invitation of women aged 65–70 years to screening boosted the burden of excess cancers, and in 2019, there was no sign that incidence rates would return to levels expected from pre-screening levels (figure 2). Overdiagnosis is thus a pervasive side effect of screening, which comes in addition to secular trends in incidence documented for England since the end of the 1950s.²³

Conclusions

The breast screening programme in England is associated with substantial amount of overdiagnosis. There was no sign that the burden of overdiagnosis in women aged 50–71 years would reduce with time, and recent efforts for extending screening to women aged 45–49 and 71–74 years is likely to further expand overdiagnosis.

Supplementary data

Supplementary data are available at EURPUB online.

Funding

This work was supported by the International Prevention Research Institute funded works required by the study. This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflicts of interest: None declared.

Data availability

Data used for this study are publicly available from sources mentioned in the article.

Key points.

• Overdiagnosis is the screen-detection of cancers that would not progress into a symptomatic cancer during lifetime.

• The study is the first to use reports on breast cancers detected by the English NHS Breast Screening Programme for estimating overdiagnosis.

• About 18% of non-invasive and invasive breast cancers found among English women in age groups eligible for screening would represent overdiagnosis.

• For reducing the burden of overdiagnosis, breast screening should be targeted to women at high risk of breast cancer.