Benefits and harms of breast cancer screening revisited: a large, retrospective cross-sectional study quantifying treatment intensity in women with screen-detected versus non-screen-detected cancer in Australia and New Zealand
•,,,,,.
...
Abstract
Objectives Non-mortality benefits of breast cancer screening are rarely considered in assessments of benefits versus harms. This study aims to estimate the rate of overdiagnosis in women with screen-detected breast cancer (SDBC) by allocating cases to either possibly overdiagnosed (POD) or not overdiagnosed categories and to compare treatment recommendations for surgery and adjuvant treatments by category, age at diagnosis and cancer stage.
Methods and analysis Retrospective secondary analysis of 10 191 women diagnosed with breast cancer in Australia and New Zealand in 2018. Treatment recommendations for 5226 women with SDBC and 4965 women with non-SDBC (NSDBC) were collated and analysed. Descriptive statistics were used to calculate proportions and risk ratios (RRs).
Results The POD rate was 15.8%. Screening detected 66.3% of stage 0 tumours, 59% of stage 1, 40% of stage 2 and 27.5% of stage 3 tumours. Women with SDBC were less likely than their NSDBC counterparts to receive chemotherapy (RR 0.60 Aus/0.53 NZ), immunotherapy (mostly human epidermal growth factor 2 receptor therapy) (RR 0.58 Aus/0.82 NZ), mastectomy (RR 0.55 Aus/0.63 NZ) and axillary lymph node dissection (RR 0.49 Aus/0.52 NZ), or to require both mastectomy and radiotherapy (RR 0.41 Aus/0.34 NZ). Less than 1% of POD women were recommended chemotherapy, 9.5% radiotherapy, 6.4% endocrine therapy, 2.2% mastectomy and 0.5% axillary lymph node dissection.
Conclusions Women with SDBCs required less intensive treatment; rates of possible overtreatment of SDBCs are relatively low and may be minimised through multidisciplinary discussion and shared decision-making. Reduced treatment intensity should be considered when balancing the potential benefits and harms of screening.
What is already known on this topic
Recent studies of organised national breast cancer screening programmes have focused on estimates of the mortality benefits and overdiagnosis rates.
What this study adds
This study demonstrates that screening detects a significant proportion of high-risk breast cancer phenotypes.
It documents lower treatment intensity in women with screen-detected breast cancer, especially in women who may be considered potentially overdiagnosed (15.8%).
It reveals that concerns about extensive overtreatment of possibly overdiagnosed cancers are likely to be overstated.
How this study might affect research, practice or policy
This study establishes that earlier detection of high-risk cancers not only improves a woman’s prognosis, but often reduces treatment intensity, and is, therefore, likely to improve post-treatment quality of life, compared with later diagnosis.
Background
Much research and commentary over the last 15 years has focused on overdiagnosis as a major harm of breast cancer screening, although a consensus on the magnitude of the problem is lacking.1–9 Estimates range widely and there is no agreement on the best method to measure it.10 In 2013, the Independent UK Panel on Breast Cancer Screening defined overdiagnosis as ‘the detection of cancers on screening, which would not have become clinically apparent in the woman’s lifetime in the absence of screening’.10 The report reviewed the previous literature and estimated a 20% relative risk reduction in mortality for all women invited to screen, offset by a 1% absolute chance that each woman invited to screen may have a cancer diagnosed and treated which would otherwise never have caused her problems.10 The extent of these estimates has since been challenged.11–13 A 2021 review of published estimates of breast cancer overdiagnosis noted ‘that all the very high estimates were from studies with no individual data on screening exposure, whereas studies with such individual data tended to obtain more modest estimates’ (Chaltiel and Hill14, p.1). The authors concluded that many estimates of breast cancer overdiagnosis represent ‘serious overestimations’ (Chaltiel and Hill14, p. 1). A 2023 large case–control study of women screened through England’s National Health Service (NHS) Breast Screening Programme (NHSBSP) supported this argument, with overdiagnosis estimates of 9.5% without adjustment for self-selection, and 3.7% with adjustment.15 The authors concluded their results ‘showed little if any overdiagnosis, and it is reasonable to conclude that NHSBSP is associated with at worst modest overdiagnosis of breast cancer’ (Blyuss et al,15 p. 1886).
The UK Panel report did not take into consideration important non-mortality benefits from breast cancer screening when weighing up the benefits versus harms,10 although this idea has been considered since 2001.16–18 The arguments for alerting women to the potential harms of overdiagnosis, and for describing the mortality and the non-mortality related benefits of early detection, are both based on the premise of harm minimisation. Early-stage breast cancer is less likely to require treatment with mastectomy, axillary lymph node dissection (ALND), chemotherapy and postmastectomy radiotherapy and is more likely to be associated with better long-term quality of life than cancer diagnosed at a later stage.19 The early identification of high-risk cancers also provides the opportunity for higher rates of pathological complete response to neoadjuvant therapy, with excellent clinical outcomes for many women, particularly those with triple negative and human epidermal growth factor receptor (HER) 2 positive phenotypes.20 This translates into a better prognosis and less adjuvant systemic therapy, and its associated morbidity, for these women.
This study compared and measured the types of breast cancer treatment women were recommended to have based on whether their cancer was diagnosed within or outside a formal screening programme using a large population database. This study has two objectives: to estimate the rate of overdiagnosis in women with screen-detected breast cancer (SDBC) by allocating cases to either possibly overdiagnosed (POD) or not overdiagnosed (NOD) based on predetermined criteria; and to compare differences in treatment recommendations for adjuvant treatments and surgery by overdiagnosis category, age at diagnosis, cancer stage and country.
Methods
Study design
Cross-sectional study quantifying treatment intensity in women with screen-detected versus non-screen-detected breast cancer in Australia and New Zealand.
Patient and public involvement
There was no involvement from patients or members of the public in the design, or conduct, or reporting of the research. Patient and public involvement will be sought and encouraged in the dissemination of findings stage. This will involve three main avenues of work: (1) codesign with consumers through the Breast Cancer Network Australia’s consumer network to write and publish consumer-friendly materials on the additional benefits of screening on the BCNA website and hard copy publications; (2) discussions with key breast cancer screening stakeholders such as BreastScreen Australia and the senior managers in the Australian government’s national screening section on updating the breast cancer screening consumer information to include our findings and (3) discussions with UK researchers and NHS Breast Cancer Screening representatives on possible similar updates to their consumer-facing programme information.
Definition of terms
The term SDBC is used to refer to breast cancer detected at the two population-based screening programmes: BreastScreen Australia and BreastScreen Aotearoa (NZ). Data on screening in private centres in these countries are not available. SDBC is assumed to be breast cancer detected prior to the appearance of any signs or symptoms (such as a breast lump, nipple discharge, dimpling or other changes in appearance). ‘Early detection’ is defined as breast cancer detected through screening. The term ‘non-SDBC’ (NSDBC) is used to refer to breast cancer that is detected outside the national screening programmes, most often representing cancers detected when a woman presents with symptoms. Interval cancers diagnosed outside the screening programmes are included in this group but cannot be separately identified from the data set, as are asymptomatic cancers identified via private screening or as incidental findings. This project examines phenotypic differences between the two groups and whether SDBC and NSDBC groups differ in the type and extent of surgical and adjuvant treatment (chemotherapy, radiotherapy, endocrine and targeted immunotherapy—mostly HER 2 therapy) they receive.
To account for possible overdiagnosis arising from screening, the criteria used by Elder et al21 to classify patients were adapted, using our terminology, as POD or NOD. The NOD group comprises phenotypes we believe a consensus opinion of breast cancer experts in Australia and New Zealand would consider required treatment. These are: high-grade ductal carcinoma in situ (DCIS); stage 1, grade 1, 1≥10 mm; stage 1, grades 2 and 3; stage 2 and 3; HER2+ and triple negative invasive cancers. This consensus represents the opinions of very senior, experienced clinicians in the authorship of this paper, supported by the strong opinions expressed in their respective multidisciplinary team (MDT) meeting. These occur throughout Australia and New Zealand. Phenotypes outside this group may or may not be overdiagnosed, depending on patient characteristics including age, comorbidities or frailty. POD phenotypes are: low-grade and intermediate-grade stage 0 cancers (DCIS) (regardless of receptor status); and stage 1 cancers (grade 1, 1 to <10 mm, excluding triple negative and HER2+ receptor status).
Data analysis
For analyses, 95% confidence intervals are reported. Given the large sample size permits detection of trivial associations, our interpretation focuses on magnitude of effect, most often using risk ratios (RRs). P values of significance are not presented due to the likelihood of misinterpretation. Background on the data source and descriptors is provided in online supplemental appendix A.
Primary outcome
The primary outcome was treatment intensity. Definitive treatments were determined: for breast surgery, this was wide local excision (WLE) or mastectomy; for axillary surgery, this was sentinel lymph node biopsy (SLNB) if the number of nodes examined was between 1 and 7, or ALND if the number of nodes examined was greater than 7. In addition, the type of adjuvant treatment each woman had received was collated: chemotherapy, radiotherapy and Herceptin (classified in the Breast Quality Audit (BQA) as immunotherapy).
Variables
The variable of primary interest was whether or not breast cancer was detected within a national screening programme. Our data set was extracted from all 2018 entries in the Breast Surgeons of Australia and New Zealand BQA database,22 which classifies cases according to surgeon referral source. There are four options: (a) referred from Breast Screen Australia; (b) referred from Breast Screen Aotearoa (New Zealand); (c) ‘symptomatic’ (usually referred from general practitioner) and (d) ‘other’ (usually referred from a private screening practice or other specialist). For the present analysis, these four categories were merged into two: (1) Breast Screen Australia/Aotearoa (SDBC) and (2) ‘symptomatic’ and ‘other’ (NSDBC).
Data extraction and cleaning was conducted during 2022. Treatment recommendations for 5226 women with SDBC and 4965 women with NSDBC were collated and analysed. Descriptive statistics were used to calculate proportions and RRs. Analysis was performed on women of target screening age only (50–74 in Australia; 45–69 in NZ). Age was recorded in years at time of diagnosis. Postcode was used as a proxy for socioeconomic status, using the Socio-Economic Status for Areas codes, which map Australian postcodes to deciles indicating socioeconomic advantage, where a higher score indicates less disadvantage23 (equivalent data were unavailable for New Zealand). Cancer stage was coded using the seventh Edition of the American Joint Committee on Cancer’s tumour, node, metastases system.24 Triple negative status and HER2 status were also included as variables.
Results
Demographic and tumour variables
Demographic and tumour variables by screening status are shown in table 1. Australian women comprised 76% (n=7746) and New Zealand women 24% (n=2445) of the study cohort, with a mean age of 61 for both the SDBC and NSDBC groups. Not all demographic or tumour variables were available for all women: DCIS grade was missing in 75 women and cancer stage missing in 35 women.
Table 1
|
Demographic and tumour variables by diagnostic pathway and country (screening age; n=10 191)*
The vast majority identified as non-Indigenous, including 83.2% of Australian and 81% of NZ women; 40.9% of Australian women attended public hospitals, compared with 72.5% of NZ women. Among Australian women, 20.3% resided in low socioeconomic areas, 39.2% in middle and 40.5% in high socioeconomic areas. When comparing the number of SDBC cases as a proportion of the total breast cancers, New Zealand women had a higher rate of SDBC (56.6%) compared with Australian women (49.6%).
Compared with Australian women, NZ women had overall fewer recommendation rates for WLE, and combined WLE and radiotherapy, chemotherapy, radiotherapy and endocrine therapy; and higher recommendation rates for mastectomy alone and mastectomy with radiotherapy. Recommendation rates for other treatments were similar between the two subgroups.
In this study, one-third of all triple negative cancers were screen detected, along with 40% of all HER2 positive tumours, and half of both positive oestrogen receptor (ER) and progesterone receptor (PR) positive cancers. Screening detected two-thirds of DCIS (stage 0) including 63.1% of intermediate-grade and 72.3% of high-grade DCIS tumours, which have the potential to become invasive.25 Furthermore, 59% of stage 1 tumours, 40% of stage 2 tumours and 27.5% of stage 3 tumours were detected through screening.
Overdiagnosis rate
Applying the allocation rules from table 2, 639 (16.7%) of Australian women and 179 (13.9%) of New Zealand women were POD, making a combined estimated POD rate of 818 women (15.8%) within this cohort.
Table 2
|
Allocation of SDBC cases to possibly overdiagnosed (POD) and not overdiagnosed (NOD) categories*
Surgical and adjuvant treatments
Cross-tabulations for individual treatments by diagnostic pathway and country, for the total cohort, are shown in table 3. Women with SDBC were more likely to undergo WLE (RR 1.30 Aus/1.39 NZ) and re-excision (RR 1.35 Aus/1.27 NZ) than women with NSDBC, but only half as likely to undergo more extensive surgical treatments including mastectomy (RR 0.55 Aus/0.63 NZ) and ALND (RR 0.49 Aus/0.52 NZ). In terms of adjuvant treatments, women with SDBC were less likely than the NSDBC group to be recommended for chemotherapy (RR 0.60 Aus/0.53 NZ) and immunotherapy (RR 0.58 Aus/0.82 NZ), as likely to be recommended for endocrine therapy in Australia but less likely in NZ (RR 0.99 Aus/0.86 NZ) and slightly more likely to have radiotherapy recommended (RR 1.09 Aus/1.04 NZ). The SDBC group were also far less likely to have both chemotherapy and radiotherapy recommended (RR 0.54 Aus/0.37 NZ).
Table 3
|
Cross-tabulations for adjuvant treatments and surgery by diagnostic pathway, overdiagnosis category (NOD vs POD) and country
When comparing women with NSDBC, with those in the NOD group, the former is more likely to be recommended for chemotherapy (RR 1.45 Aus/1.68 NZ), immunotherapy (RR 1.50 Aus/1.08 NZ) and chemotherapy plus radiotherapy (RR 1.61 Aus/2.42 NZ), as well as mastectomy (RR 1.69 Aus/1.50 NZ) and ALND (RR 1.80 Aus/1.74 NZ). In contrast, they have approximately the same risk of undergoing endocrine therapy (RR 0.94 Aus/1.10 NZ) and a lower risk of radiotherapy (RR 0.88 Aus/0.95 NZ) and WLE (RR 0.78 Aus/0.73 NZ).
The largest between-group differences were found in women with NSDBC and those screened in the POD group. Surgical treatments were more intensive: women with NSDBC were less likely to be recommended for WLE (RR 0.73 Aus/0.67 NZ), up to three times more likely to require mastectomy (RR 3.21 Aus/2.34 NZ), over 45 times more likely to be recommended for mastectomy plus radiotherapy (RR 48.53 Aus/4.89 NZ) and seven times more likely to have ALND recommended (RR 7.13 Aus/7.79 NZ). No women in the POD group were recommended to have chemotherapy plus radiotherapy, or immunotherapy, and only four Australian women were recommended to have chemotherapy (RR 55.17 Aus). Women with NSDBC were also more likely to require endocrine therapy (RR 1.56 Aus/1.78 NZ), while recommendations for radiotherapy alone were similar in both the NSDBC and POD groups (RR 1.09 Aus/1.10 NZ).
Finally, among those screened, women in the POD group were far less likely than those in the NOD group to undergo invasive treatments. For surgical procedures, POD women were much less likely to be recommended for ALND (RR 0.25 Aus/0.22 NZ) and mastectomy (RR 0.53 Aus/0.64 NZ), but slightly more likely to receive WLE (RR 1.07 Aus/1.08 NZ). They were also less likely to receive radiotherapy alone (RR 0.81 Aus/0.86 NZ) and endocrine therapy (RR 0.60 Aus/0.62 NZ).
Age distribution and cancer stage by overdiagnosis category and country
Table 4 shows the distribution of SDBC cases per age group, country, cancer stage and overdiagnosis category. Those in the 45–49 age group (NZ only) comprised 231 (16.8%) NZ women, of which 20 (8.7%) cases were categorised as POD. The majority of cases (4063) were diagnosed in the 50–69 age group: 478 Australian women (16.2%) and 154 New Zealand women (13.9%) were categorised as POD. Women in the 70–74 age group (Australia only) comprised 845 cases, of which 156 (18.5%) were classified as POD.
Table 4
|
Diagnosis of SDBC* by age, breast cancer stage, overdiagnosis category and country (n=5139)†
Of the 808 women with SDBC classified as POD, 11.5% had low-grade DCIS, 32.9% had intermediate-grade DCIS and 55.6% had lower risk stage 1 (grade 1, <10 mm, HER2 negative and ER and/or PR positive) breast cancer. Women classified as NOD (n=4331) had 12.2% high-grade DCIS, 51.1% higher risk stage 1 (grade 1, ≥10 mm, with any receptor status), 32.1% stage 2 and 4.6% stage 3 breast cancers.
Age distribution, cancer stage and recommended treatments by overdiagnosis category
Table 5 shows the proportion of SDBC women, divided into the same three age groups as table 4, in POD versus NOD categories who received each type of treatment. More intensive surgical options were less likely for POD women: 2.2% were recommended for mastectomy and only 0.5% for ALND. In terms of adjuvant treatments, less than 1% of women in the POD categories were recommended to have chemotherapy, 9.5% to have radiotherapy and 6.4% to have endocrine therapy. No POD women required chemotherapy and radiotherapy. HER2 positivity was an exclusion criterion for classification as POD, so none of these women required immunotherapy.
Table 5
|
Treatment of SDBC* by age, cancer stage and overdiagnosis category (n=5169)†
In terms of treatments by age, younger women (45–49, NZ only) were recommended to have the highest rates of mastectomy (41.1%) and ALND (15.6%), chemotherapy (28.1%) and immunotherapy (11.3%), and the lowest rates of chemotherapy plus radiotherapy (2.6%), radiotherapy (54.5%), endocrine therapy (52.9%), WLE (59.7%) and SLNB (68.8%). Older women (70–74, Aus only) had the lowest rates of recommendations for mastectomy (21.1%), ALND (10.1%) and immunotherapy (4.6%), and the highest rates for SLNB (76%) recommendations. Compared with the older group, the majority of women (50–69, Aus plus NZ) were recommended for higher rates of WLE (79.3%), chemotherapy only (25.5%) and slightly higher rates for radiotherapy (72%), chemotherapy and radiotherapy (3.9%) and endocrine therapy (62.7%).
Discussion
This analysis of 10 191 women quantified the differences in diagnoses and recommended treatments (a proxy for treatment-related morbidity) for women with SDBC versus NSDBC. Our results demonstrate that breast cancer screening detects both low- and higher-risk cancers before they become symptomatic. Furthermore, earlier detection results in fewer invasive treatments for women with SDBC compared with women with NSDBC.
Overdiagnosis is recognised as a potential harm of breast cancer screening, and therefore, we assessed the proportion of cases that may have been overdiagnosed. The POD rate was calculated to be 15.8% and was shown to vary with age.
Women with NSDBC were up to three times more likely to require mastectomy, over 45 times more likely to be recommended for mastectomy plus radiotherapy, and seven times more likely to have ALND recommended. This suggests that detection (or overdetection) of low-risk cancers may lead to minimal harm if they are treated appropriate to their stage and size at time of detection. For example, many of the tumours in the POD group may be managed with surveillance rather than surgery,26 27 and/or endocrine therapy,27 28 or radiotherapy.27 29–34 This de-escalation of treatment for some early-stage disease needs to be balanced against growing evidence that intermediate-grade and high-grade DCIS may progress to invasive cancer and should be treated, rather than monitored, to prevent progression.25 35–37
Strengths and weaknesses
This robust study design comprised secondary analysis of data collected at a single time point (2018), on over 10 000 women diagnosed with primary breast cancer, from approximately 320 breast surgeons across Australia and New Zealand. The analysis is based on recommended treatments for women who did and did not have their breast cancer diagnosed through their national screening programmes.
A major strength is the unique methodology, which enabled the calculation of an overdiagnosis rate in women with breast cancer using a fixed, a priori definition of overdiagnosis. This enabled correction for overdiagnosis in our analyses by separating the most favourable SDBC diagnoses (POD) from those more high-risk/advanced tumours that, on medical consensus, would not be considered overdiagnosed (NOD). We believe this approach provides a clearer assessment of treatment-related quality of life than previous epidemiological estimates based on intention to screen data linked to mortality outcomes; such approaches did not measure the additional benefits of screening related to reduced treatment intensity and the positive impact this may have on women’s quality of life.10 Our estimate of 15.8% is at the lower end of most overdiagnosis estimates1–15 and we believe it is a legitimate method.
Importantly, this approach has allowed for direct linkage between diagnostic pathway and treatment recommendations, demonstrating that overdiagnosis does not necessarily lead to overtreatment. However, relatively small numbers of POD women were recommended for adjuvant treatment; this may be explained by either inexperienced clinicians or factors that were not analysed, such as family history of breast cancer or BRCA1 or BRCA2 status. In both Australia and NZ, the vast majority of breast cancer cases are discussed at multidisciplinary meetings where treatment decisions are matched not only to the histopathology, but also to the individual women; factors such as age, comorbidities, suitability for surgery, potential side effects of treatment, family history, mutation carrier status and the women’s preferences are all considered when making treatment recommendations. Footnotes from Table 3 demonstrate that recommendations for treatment for all breast cancers, however detected, are not always accepted. Women may decline treatments and clinicians may opt for less intensive treatments according to the individual’s circumstances.
The cross-sectional design did not track screening episodes over time and data on interval cancers (those that arise between screening rounds) was not available. However, it does provide a clear and accurate snapshot of the impact of diagnostic pathways on diagnosis and recommended treatments. One inherent limitation is the loose boundaries for allocation of women to either screen-detected or non-screen-detected categories, as outlined in online supplemental appendix A. This is a confounding variable beyond our control.
Clinical implications
This study has three major clinical implications:
Breast cancer screening detects both low-risk and higher-risk tumours. Earlier detection of high-risk cancers not only improves a woman’s prognosis but often reduces treatment intensity and is therefore likely to improve post-treatment quality of life, compared with later diagnosis.
Women with NSDBC have substantially higher rates of more serious/advanced types of breast cancer than SDBC women, associated with higher rates of treatment intensity.
Detection of low-risk tumours that are POD does not lead to extensive treatment of these cancers in the vast majority of cases.
Conclusion
A reduction in breast cancer mortality is not the only screening outcome of importance. Reduced intensity of recommended treatment is also a key benefit. Women contemplating breast cancer screening must be informed about both sides of the harm minimisation argument so that they can make an informed, value-based and personalised decision. Any women considering breast cancer screening must be fully informed of both risks (including overdiagnosis) and benefits (including the option of less harmful treatments) before making the decision to screen or not screen.
