Review J Med Screen 2021, Vol. 28(4) 389–404 Benefits and harms of breast cancer ! The Author(s) 2021 mammography screening for women Article reuse guidelines: sagepub.com/journals-permissions at average risk of breast cancer: DOI: 10.1177/0969141321993866 journals.sagepub.com/home/msc A systematic review for the European Commission Initiative on Breast Cancer Carlos Canelo-Aybar1,2, Diogenes S Ferreira2,3, M onica Ballesteros2, Margarita Posso2,4, Nadia Montero2, Ivan Solà2, Zuleika Saz-Parkinson5 , Donata Lerda5, Paolo G Rossi6 , Stephen W Duffy7 , Markus Follmann8, Axel Gr€ awingholt9 and 1,2 Pablo Alonso-Coello Abstract Objectives: Mammography screening is generally accepted in women aged 50–69, but the balance between benefits and harms remains controversial in other age groups. This study systematically reviews these effects to inform the European Breast Cancer Guidelines. Methods: We searched PubMed, EMBASE and Cochrane Library for randomised clinical trials (RCTs) or systematic reviews of observational studies in the absence of RCTs comparing invitation to mammography screening to no invitation in women at average breast cancer (BC) risk. We extracted data for mortality, BC stage, mastectomy rate, chemotherapy provision, over- diagnosis and false-positive-related adverse effects. We performed a pooled analysis of relative risks, applying an inverse- variance random-effects model for three age groups (<50, 50–69 and 70–74). GRADE (Grading of Recommendations Assessment, Development and Evaluation) was used to assess the certainty of evidence. Results: We identified 10 RCTs including 616,641 women aged 38–75. Mammography reduced BC mortality in women aged 50–69 (relative risk (RR) 0.77, 95%CI (confidence interval) 0.66–0.90, high certainty) and 70–74 (RR 0.77, 95%CI 0.54–1.09, high certainty), with smaller reductions in under 50s (RR 0.88, 95%CI 0.76–1.02, moderate certainty). Mammography reduced stage IIAþ in women 50–69 (RR 0.80, 95%CI 0.64–1.00, very low certainty) but resulted in an overdiagnosis probability of 23% (95%CI 18–27%) and 17% (95%CI 15–20%) in under 50s and 50–69, respectively (moderate certainty). Mammography was associated with 2.9% increased risk of invasive procedures with benign outcomes (low certainty). Conclusions: For women 50–69, high certainty evidence that mammography screening reduces BC mortality risk would support policymakers formulating strong recommendations. In other age groups, where the net balance of effects is less clear, conditional recommendations will be more likely, together with shared decision-making. Keywords Guidelines, breast cancer, mass screening, mammography Date received: 8 July 2020; accepted: 31 December 2020 1 CIBER de Epidemiologıa y Salud Publica (CIBERESP), Madrid, Spain 8 German Cancer Society, Berlin, Germany 2 Department of Clinical Epidemiology and Public Health, Iberoamerican 9 Radiologie am Theater, Paderborn, Germany Cochrane Centre, Biomedical Research Institute Sant Pau (IIB Sant Pau), Barcelona, Spain Corresponding authors: 3 Department of Epidemiology and Preventive Medicine, School of Public Carlos Canelo-Aybar, Iberoamerican Cochrane Centre, Biomedical Research Health and Preventive Medicine, Monash University, Melbourne, Australia Institute Sant Pau (IIB Sant Pau – CIBERESP), Sant Antonio Marıa Claret 167, 4 Department of Epidemiology and Evaluation, IMIM (Hospital del Mar Medical 08025 Barcelona, Spain. Research Institute), Barcelona, Spain Email: ccanelo@santpau.cat 5 European Commission, Joint Research Centre (JRC), Ispra, VA, Italy 6 Epidemiology Unit, AUSL – IRCCS di Reggio Emilia, RE, Italy Zuleika Saz-Parkinson, European Commission, Joint Research Centre (JRC), 7 Centre for Cancer Prevention, Wolfson Institute of Preventive Medicine, Via E. Fermi, 2749. TP127, I-21027 Ispra (VA), Italy. Queen Mary University of London, London, UK Email: zuleika.saz-parkinson@ec.europa.eu 390 Journal of Medical Screening 28(4) Introduction equity, acceptability and feasibility. Readers are welcome to refer to these considerations in the published recom- Breast cancer (BC) is the second most common malignan- mendations and on the ECIBC website (https://health cy in the world.1 In the European Union, 404,920 women care-quality.jrc.ec.europa.eu/european-breast-cancer- were diagnosed with BC and 98,755 died during 2018.2 guidelines/screening-ages-and-frequencies).12,13 Over the last 20 years, BC mortality has decreased due to improvements in treatment, services delivery and imple- mentation of population screening. However, the role of Methods population screening has been under debate over the last three decades due to conflicting systematic reviews and Structured question and outcome prioritisation recommendations.3,4 The clinical question prioritised by the GDG, ‘Which is Randomised clinical trials (RCTs) carried out during the optimal age range in which to carry out screening for the 1970s and 1980s showed that mammography screening breast cancer?’, followed the Population, Intervention, is associated with a reduction in BC mortality.5 However, Comparison and Outcomes format (Box 1). Three sub- screening a healthy population is also associated with populations were pre-defined: women under 50, 50–69 undesirable effects such as recalling women with a false- and 70–74 years old at the moment of invitation to positive result for additional imaging.6 Overdiagnosis (BC screening. cases that would not have clinically surfaced in the absence During the development of the recommendations,11 the of screening) is another downside of screening, and its GDG decided to split the sub-group of women under 50 magnitude is controversial.7 into two sub-groups: 40–44 and 45–49. The outcomes were Numerous organisations have issued screening recom- prioritised by the GDG using a 1–9 scale as suggested by mendations. The WHO recommended in favour of screen- the GRADE (Grading of Recommendations Assessment, ing starting at 40 years of age in well-resourced settings.8 Development and Evaluation) approach.14 The Canadian Task Force recommended screening only in women over 50, due to the risk of overdiagnosis and unnecessary biopsies in younger women.9 The American Data sources and searches Cancer Society, including evidence from RCTs and obser- We searched MEDLINE (April 2016), EMBASE (April vational and modelling studies, recommended initiating 2016) and CENTRAL (March 2016) databases using annual screening at 45.10 pre-defined algorithms for both systematic reviews and In 2015, the (ECIBC) was launched to develop the individual studies. We adapted the search terms to each European Guidelines on Breast Cancer Screening and database (see Supplemental material 1). We also reviewed Diagnosis. This systematic review informed the recom- lists of references of the included studies, and members of mendations about mammography screening for early the GDG were consulted about potentially missing studies. detection of BC in asymptomatic women at average risk. During June 2018, we performed a new search in During the guideline development,11 the Guidelines MEDLINE as part of the ECIBC’s guideline updating Development Group (GDG) made detailed considerations process. The results were assessed by the GDG and, as on the evidence of effects as well as values and preferences, no relevant studies were identified that could potentially Box 1 Structured clinical question. Population Intervention Comparison Outcomes Women who are at Invitation to mam- No invitation to Critical average risk of mography mammography  Breast cancer mortality breast cancer: screening screening  Overdiagnosis  Under 50 y Important  50–69 y  Stage of breast cancer  70–74 y  Other cause mortality  Rate of mastectomies  Provision of chemotherapy: provision of either adjuvant, neoadjuvant or both  Psychological effects: include anxiety caused or relieved by screening, anxiety caused by assessment of suspicious screening findings, longer length of life qualified by longer periods of life spent with a diagnosis of breast cancer, treatment side effects including psychosocial effects of body image following surgery  False positive related adverse effects: psychological distress  False positive related adverse effects: biopsies and surgeries Canelo-Aybar et al. 391 change the recommendations, they decided to not update To estimate risk differences, we used the baseline risk the review. In November 2019, the GDG met and consid- from the control arms of the RCTs; we also provide esti- ered that, to the best of their knowledge, there were no mates using baseline risks proposed by the GDG members relevant new publications. considering European population surveillance data. All results are expressed as by 100,000 women invited to mam- Study selection mography screening. We included RCTs of women at average risk of BC (with- out family history of BC or inherited changes of BRCA1 Certainty of the evidence and BRCA2 genes), comparing invitation to mammogra- We rated the certainty of the evidence for each outcome phy screening versus no invitation. If no RCTs were iden- taking into consideration the standard GRADE tified, we included systematic reviews of observational domains,19,20 described in the evidence profiles (see studies. For overdiagnosis, we included only trials in Supplemental material 4). which, after completing the study phase, neither women in the control nor the intervention group were offered mammography screening. We excluded studies conducted Results outside the context of screening programmes or not pub- Search results lished in English. Initially, at the title and abstract level, two reviewers From 2393 unique citations, we selected 57 to be appraised after calibration assessed the eligibility of the references as full text. At this stage, we excluded seven reviews, and retrieved. Two reviewers independently reviewed the full 13 observational studies of mammography screening text of the selected references. Discrepancies were solved reporting outcomes available from RCTs. Additionally, either by consensus or with the help of a third reviewer. we excluded the Edinburgh trial because of important baseline differences between the screening and control Data extraction and risk of bias assessment groups, suggesting suboptimal randomisation (see Supplemental material 5).21,22 Details of the study design, population, follow-up and We included 30 publications from nine RCTs: the Health results were extracted by one reviewer and confirmed by Insurance Plan (HIP) of Greater New York trial,5,23–26 the a second reviewer. If needed, we requested additional Canadian Breast Cancer Screening Study (CNBSS-1 and data from authors of the included studies. We assessed the risk of bias (RoB) of RCTs using the Cochrane RoB CNBSS-2),27–32 the United Kingdom Age trial,33–35 the Assessment tool15, and systematic reviews with the Stockholm trial,36–38 the Malm€ o Mammographic Screening AMSTAR (A Measurement Tool to Assess Systematic Trial (MMST I and MMST II),39–41 the G€ oteborg trial,42,43 the Swedish Two-County trial (Osterg€ € otland and Reviews) checklist (see Supplemental material 2).16 44–49 Kopparberg counties), one publication that reported results for the five Swedish mammography trials,18 and Data analysis updated results of the UK Age Trial50 and the G€ oteborg To estimate the effect of mammography screening on BC Trial (Table 1).51 We also obtained additional age-stratified mortality, we used two methods. The ‘short case accrual’ results for BC mortality from the authors of the CNBSS trial. method includes only BC deaths among BC cases diag- Four systematic reviews of observational studies fulfilled nosed during the screening intervention phase.17,18 The the eligibility criteria (Figure 1).52–55 Brett et al.53 assessed ‘long case accrual’ method considers all BC deaths irre- the adverse psychological impact of mammography screen- spective of the date of diagnosis, accrual time being equiv- ing in the general population. Salz et al.55 examined the alent to the follow-up of the study.17,18 effects of false-positive mammogram results. Bond et al.52 We estimated overdiagnosis as the difference in the evaluated the psychological effects of false-positive screen- cumulative number of BC in the groups invited and not ing mammograms in the UK, and one review assessed the invited to screening, expressed: (a) as a percentage of the cumulative risk of false-positive results leading to an inva- number of cancers in the screening group (population per- sive procedure (needle biopsy or surgery).54 spective) or (b) as a percentage of the cancers diagnosed during the screening phase of the trial in the invited to BC mortality screening group (individual perspective). We pooled data as relative risks (RR) using a random-effects model Eight trials included 348,478 women less than (Review Manager v5.3). We assessed the presence of 50,24,26,28,29,36,37,39,40,43,50,51 six trials 249,930 women aged heterogeneity among studies using the Cochrane chi- 50–69,18,23,24,26,27,29,36,37,43,47,48,51 and two trials 18,233 square test and the I2 statistic. Additionally, we provided women aged 70–74.18,39,40,47,48 The trial time ranged subgroup analysis based on the risk of bias assessment from 3.5 to 18.8 years, the median short case accrual and a post-hoc sensitivity analysis excluding RCTs with follow-up time from 9.1 to 24 years, and the median long a substantial concern for breaking the concealment at case accrual follow-up time from 13 to 21.9 years, depend- randomisation. ing on the age strata (Table 1). 392 Table 1. Overview of included randomised clinical trials. Swedish two-county HIP Malm€ o I/Malm€ o II (Ostergotland/Kopparberg) CNBSS-1/CNBSS-2 Stockholm G€ oteborg UK age Start date 1963 1976 1977 1980 1981 1982 1991 Randomisation (if cluster, type) Individual Individual Cluster (geographical Individuala Cluster (day Individual, cluster Individual area) of birth) (day of birth) Number of womenb 60,490 42,283 and 17,793 133,065 50,430/39,405 60,117 52,222 160,836 Age at entry (years) 40–64 45–70/43–49 38–75 40–49/50–59 39–65 39–59 39–41 Intervention Number of views 2 2 then 1 or 2/2c 1 2 1 2 (round 1) 2 then 1 then 1 or 2d Number of readers 2 2 1 1 1 1 (round 1-3), ND and 2 (round 4-5) Screening interval (months) 12 18-24 24,33e 12 24–28 18 12 Number of screening rounds 4 6–8/17 2–4 5/4–5 2 4–5f 4-6 Trial time (y)g 3.5 5.8/18.8 7 5 4.4 6.7 9 Last follow-up (years)h 18 19.2/9.1 20 21.9 17.4 24 17 Attendance, first round (%) 67 74/75–80 89 86/87 82 84 68 Controls invited to screening at ND No/yesj Yes (after 7 y) No Yes (after 4 y) Yes (after 7 y) Yes (after end of interventioni 10 y) ND: not described. a There was a clinical breast examination before randomization, and the personnel in charge of allocation had access to this information. b Numbers of participants included in the analysis, for some trials these numbers vary in different publications. c Depending on parenchymal pattern. d Depending on the density of breast. e Women 38–49: 24 months; Women 50–74: 33 months. f Women born 1923–1932: 4 rounds; women born 1933–1944: 5 rounds. g Number represents median time of women spent in the active intervention phase of the study; otherwise, number of years from start of randomization to last screening in the intervention arm. h Follow-up might differ for specific outcomes and accrual methods (i.e. short vs. long case). Number represents median time when available otherwise reported time from entry. i Mean years after randomization. j Applies to the cohort 1908–1922 (55–69 years at entry). Journal of Medical Screening 28(4) Canelo-Aybar et al. 393 2393 records idenfied from databases 2336 records excluded aer tle/abstract screening 57 records assessed for eligibility 23 full-text arcles excluded Reasons: 7 Reviews of mammography including outcomes available from RCTs 2 studies covering different 30 publicaons from 9 individual RCTs populaon or imaging & modalies 4 systemac reviews of observaonal studies 1 RCT with subopmal randomisaon g 13 observaonal studies Figure 1. PRISMA flowchart. In women under 50, invitation to mammography invited to screening (95%CI 251 fewer to 91 fewer) using screening probably reduces the risk of BC mortality (RR long accrual time (Table 2). Sensitivity analysis including 0.88; 95% CI (confidence interval) 0.76–1.02; I2 ¼ 20%; only RCTs at low RoB yielded similar results. short case accrual) (moderate certainty).18,24,26,28,50,51 Comparable results were obtained using a long case accru- Other cause mortality al follow-up time (RR 0.92; 95%CI 0.83–1.02; In women under 50, mammography screening may make I2 ¼ 6%)18,24,28,43,45,50,51 (Figure 2). In women aged no difference to other-cause mortality, but the evidence is between 50 and 69, mammography screening reduced the uncertain (RR 1.04; 95%CI 0.95 to 1.15; I2 ¼ 62%) (very risk both for short (RR 0.77; 95%CI 0.66–0.90; low certainty).18,28,42,44,50 Two trials were included in the I2 ¼ 49%)18,24,26,27,47,51 and long case accrual 50–69 group27,44 and one trial in the 70–74 group;44 in time18,24,27,43,45,51 (high certainty) (Figure 3). For women these age strata, mammography screening may also aged 70–74, the Malm€ o I reported short case accrual result in no difference (Table 2) (low certainty) follow-up and the Swedish Two-County reported long (Supplemental material 3: Figures S5, S8 and S9). accrual follow-up time; mammography screening reduced the risk of BC mortality (RR 0.77; 95%CI 0.54–1.09; I2 ¼ 0%)18,47 (high certainty) (Figure 4). Advanced BC The risk difference in BC mortality for women aged 50– We defined advanced stage as either stage II or greater, 69 was 138 fewer deaths per 100,000 women invited to tumour size 520 mm or 51 positive lymph node, which is screening (95%CI 204 fewer to 60 fewer) using short accrual consistent with stage IIA disease or higher. Additionally, follow time, and 175 fewer deaths per 100,000 women we used a second definition of advanced disease as 394 Journal of Medical Screening 28(4) Figure 2. Effect on breast cancer mortality of mammography screening (women under 50 years of age): (a) short case accrual, mean follow up across studies 16.8 years; (b) longest case accrual, mean follow-up across studies 15.2 years. Risk of bias legend: (A) Random sequence generation, (B) allocation concealment, (C) blinding of participants and personnel, (D) blinding of outcome assessment, (E) incomplete outcome data and (F) selective reporting, (G) other bias. regional or metastatic or tumour size 5 40 mm, equivalent fewer cases (95%CI 482 fewer to 289 fewer) (Table 2) to stage III or higher. (low certainty) (Supplemental material 3: Figures S1 and Using the stage IIA or higher definition, in women S6). under 50, mammography screening may reduce the risk Using the stage III or higher definition, in women under of advanced disease, but the evidence is uncertain (RR 50, screening may make little difference to the risk of 0.88; 95%CI 0.78 to 0.99; I2 ¼ 0%) (very low advanced disease (RR 0.98; 95%CI: 0.74–1.29; I2 ¼ 0%) certainty).23,24,28,34,43,45,50,51 In women aged 50–69, the (low certainty).23,24,28,34,45,50 In women aged between 50 effect size was similar (RR 0.80; 95%CI 0.64–1.00, and 69, mammography screening may reduce the risk of I2 ¼ 70%) (very low certainty).23,24,27,43,45,51 One trial advanced disease (RR 0.62, 95%CI 0.48–0.80; including older women (aged 50–74) showed that mam- I2 ¼ 0%),23,24,27,45 which is equivalent to 65 fewer cases mography screening may reduce the risk of advanced dis- of advanced BC (low certainty). In women aged 50–74, ease (RR 0.64; 95%CI 0.55–0.73),45 equivalent to 385 mammography screening may reduce the risk of advanced Canelo-Aybar et al. 395 Figure 3. Effect on breast cancer mortality of mammography screening (women aged 50–69): (a) short case accrual, mean follow-up across studies 17.6 years; (b) longest case accrual, mean follow-up across studies 15.5 years. Risk of bias legend: (A) random sequence generation, (B) allocation concealment, (C) blinding of participants and personnel, (D) blinding of outcome assessment, (E) incomplete outcome data, (F) selective reporting and (G) other bias. disease (RR 0.63, 95%CI 0.45–0.89),45 equivalent to 63 I2 ¼ 10%) from an individual perspective (Table 3) (mod- fewer cases (95%CI 94 fewer to 19 fewer) (Table 2) (low erate certainty).27,29,40 certainty) (Supplemental material 3: Figures S2 and S7). Overdiagnosis Rate of mastectomies We identified three trials, the CNBSS-1, the CNBSS-2 and Across all age groups, women invited to screening may a subgroup of women aged 55–69 from the Malmo-I trial undergo more mastectomies (RR 1.20, 95%CI 1.11–1.30; (women aged 45–54 received screening at the end of the I2 ¼ 0%, 180 more in absolute terms) (low certainty) study). In women aged 40–49, the estimates of overdiag- (Supplemental material 3: Figure S3).18,31,39,46 nosis were 12.4% (95%CI 9.9–14.9) from a population perspective and 22.7% (95%CI: 18.4–27.0) from an indi- vidual perspective (moderate certainty).28,29 In women Provision of chemotherapy aged between 50 and 69, we estimated a pooled overdiag- Across all age groups, the evidence was uncertain with an nosis of 10.1% (95%CI: 8.6–11.6; I2 ¼ 0%) from a popu- RR of 0.86 (95%CI 0.53–1.40; I2 71%)18,39,46 (Table 3) lation perspective and 17.3% (95%CI: 14.7–20.0; (very low certainty) (Supplemental material 3: Figure S4). 396 Journal of Medical Screening 28(4) Figure 4. Effect on breast cancer mortality of mammography screening (women 70 years of age or older). Malm€ o I and Swedish two-county reported short case accrual estimate, follow-up across studies 9.5 years. Psychological effects surgical intervention with a benign outcome of 3.9%, 1.5% and 0.9%, respectively.56 The largest study, from a Uncertain evidence showed that mammographic screening Spanish screening programme, reported an estimated may not produce anxiety in women given a clear result cumulative risk of 1.8% for undergoing an invasive pro- after a mammogram.53 However, those requiring further cedure with a benign outcome.57 investigations may experience significant anxiety, in the short and long term, depending on the extent of the addi- tional exams (Table 3) (very low certainty).53 RoB and certainty of the evidence Our main concerns for BC mortality were: (1) the use of False-positive-related psychological distress suboptimal random allocation methods, such as the date of birth to allocate women to each study arm in the One review, including 17 studies, suggested an increase in Stockholm and Gothenburg trials;36,42 (2) in the CNBSS the scores of disease-specific BC measures of psychological trials, participants underwent clinical breast examination distress with false-positive results, being largest for anxiety before randomisation, and this information was available about BC (r ¼ 0.22; 95%CI: 0.18–0.27) and smallest for to the personnel in charge of allocation.27,28 However, no fear (r ¼ 0.08; 95%CI: 0.03–0.14) (low certainty).55 In con- single trial drove the overall results in the subgroup anal- trast, when using non-specific measures, the only sug- ysis of low versus high RoB (Tables 2 and 3). This was gested effect was a higher risk of generalised anxiety corroborated in a post-hoc sensitivity analysis, where we (r ¼ 0.03; 95%CI: 0.00–0.07) (low certainty).55 explored the impact of excluding the CNBSS trials, which Another review included four studies evaluating psy- resulted in a non-meaningful change in the point estimates chological impact;52 false-positive mammograms may be for the outcome breast cancer mortality (i.e. for women associated with negative psychological consequences when under 50, RR 0.84; 95%CI 0.73–0.98, for short case accru- assessed using disease-specific measures (e.g. BC anxiety). al) (Supplemental material 3: Figures S10 and S11). Additionally, the risk of negative effects may be greater if For overdiagnosis, in the CNBSS trials, there was a biopsy is required (RR 2.07; 95%CI 1.22–3.52) than if potential bias due to screening after the trial ended, as only further mammography is needed (RR 1.28; 95%CI screening programmes were subsequently implemented 0.82–2.00) (low certainty).52 under different province jurisdictions with an increased likelihood of attendance of women who had been screened False-positive-related procedures during the trials.29,40 Also, the overdiagnosis results were One systematic review included four primary studies and simple cumulative incidence differences with no adjust- an analysis of performance parameters from 20 screening ment for lead time and, therefore, might be overestimated. programmes (low certainty).54 One Norwegian study Other relevant limitations were related to indirectness, reported a cumulative risk of undergoing fine needle aspi- due to difference in quality control of screening and ration cytology, core needle biopsy (CNB) and having a improvement of care over the last two decades, but given Table 2. Summary of available evidence on desirable effects of breast cancer mammography screening by age groups. Summary of findings Quality assessment No of studies Certainty of Relative effect Age group (population) Design Certainty assessment evidence (95%CI) Basal riska Absolute reduction Canelo-Aybar et al. Breast cancer mortality (short case accrual) <50 8 (348,478) Randomised Downgraded due to imprecision ⨁⨁⨁b,c,d RR 0.88 (0.76– 0.4% 48 fewer per 100,000 (from 84 fewer to trials (serious) MODERATE 1.02) 8 more) 0.7% 84 fewer per 100,000 (from 168 fewer to 14 more) 50–69 6 (249,930) Randomised No relevant considerations were ⨁⨁⨁⨁b,c,e RR 0.77 (0.66– 0.6% 138 fewer per 100,000 (from 204 fewer trials found to justify downgrading HIGH 0.90) to 60 fewer) the certainty of the evidence. 1.0% 230 fewer per 100,000 (from 340 fewer to 100 fewer) 2.1% 483 fewer per 100,000 (from 714 fewer to 210 fewer) 70 to 74 2 (18,233) randomised No relevant considerations were ⨁⨁⨁⨁b,c,f RR 0.77 (0.54– 0.9% 207 fewer per 100,000 (from 81 more trials found to justify downgrading HIGH 1.09) to 414 fewer) the certainty of the evidence. 3.0% 690 fewer per 100,000 (from 1380 fewer to 270 more) Breast cancer mortality (longest case accrual available) <50 8 (348,478) Randomised Downgraded due to imprecision ⨁⨁⨁b,c,d RR 0.92 (0.83– 0.5% 38 fewer per 100,000 (from 82 fewer to trials (serious) MODERATE 1.02) 10 more) 50–69 6 (249,930) Randomised No relevant considerations were ⨁⨁⨁⨁b,c,e RR 0.77 (0.67– 0.8% 175 fewer per 1,00,000 (from 251 trials found to justify downgrading HIGH 0.88) fewer to 91 fewer) the certainty of the evidence. Other cause mortality 6 (290,600) Randomised Downgraded due to inconsis- ⨁ RR 1.04 (0.95– 2.5% 100 more per 100,000 (from 125 fewer <50 trials tency (serious), indirectness VERY LOW 1.15) to 375 more) (serious) and imprecision (serious). 2 (119,376) Randomised Downgraded due to indirectness ⨁⨁ RR 0.99 (0.95– 6.6% 66 fewer per 100,000 (from 330 fewer 50–69 trials (serious) and imprecision LOW 1.04) to 264 more) (serious). 1 (17,646) Randomised Downgraded due to indirectness ⨁⨁ RR 1.01 (0.91– 27.0% 270 more per 100,000 (from 2430 70–74 trials (serious) and imprecision LOW 1.10) fewer to 2700 more) (serious). Breast cancer stage IIA or higher 5 (300,307) Randomised Downgraded due to risk of bias ⨁d,g,h RR 0.88 (0.78– 0.4% 46 fewer per 100,000 (from 84 fewer to <50 trials (serious), indirectness (seri- VERY LOW 0.99) 4 fewer) ous) and imprecision (serious). 4 (196,141) Randomised Downgraded due to risk of bias ⨁c,d,i,k RR 0.80 (0.64– 0.7% 140 fewer per 100,000 (from 252 fewer 50–69 trials (serious), inconsistency VERY LOW 1.00) to 0 fewer) (continued) 397 398 Table 2. Continued. Summary of findings Quality assessment No of studies Certainty of Relative effect Age group (population) Design Certainty assessment evidence (95%CI) Basal riska Absolute reduction (serious) indirectness (seri- ous) and imprecision (serious). 1 (97,617) Randomised Downgraded due to indirectness ⨁⨁b,j RR 0.64 (0.55– 1.1% 385 fewer per 100,000 (from 482 fewer 50–74 trials (very serious). LOW 0.73) to 289 fewer) Breast cancer stage – stage IIIþ or tumour size 5 40 mmk 4 (274,212) Randomised Downgraded due to indirectness ⨁⨁b,c,d RR 0.98 (0.74– 0.1% 2 fewer per 100,000 (from 23 fewer to <50 trials (serious) and imprecision LOW 1.29) 26 more) (serious). 3 (170,032) Randomised Downgraded due to risk of bias ⨁⨁c,i RR 0.62 (0.48– 0.2% 65 fewer per 100,000 (from 88 fewer to 50–69 trials (serious) and indirectness LOW 0.80) 34 fewer) (serious). 1 (97,617) Randomised Downgraded due to indirectness ⨁⨁b,j RR 0.63 (0.45– 0.2% 63 fewer per 100,000 (from 94 fewer to 50–74 trialsc (very serious). LOW 0.89) 19 fewer) a The GDG considered that baseline risks higher than 0.6% should be considered to evaluate absolute effects of breast cancer mortality (Breast Cancer Screening, IARC Handbook of Cancer Prevention Volume 15). b Some studies used random allocation methods that would not be currently accepted. One study had a non-blinded assessment of ‘cause of death’. The GDG felt that the CNBSS-2 possibly had issues with achieving prognostic balance. The GDG felt that lack of allocation concealment in this set of studies did not lead to high risk of bias. Given that lack of single trials driving the overall results and similarity in effect sizes (the test for subgroup differences – low vs. high risk of bias trials – was non-significant) and overlapping confidence intervals (CIs), the risk of bias was rated as ‘not serious’. c Trials were conducted more than 20 years ago. Currently, women have higher adherence to breast cancer screening, and the quality control of screening and the care of breast cancer have improved. A large non- randomised study (Hellquist B 2011)79 showed a reduced risk for breast cancer deaths in women aged 40–49 years invited to screening, compared with women not invited (RR ¼ 0.74; 95%CI, 0.66–0.83), which is consistent with the results seen in the RCTs. The GDG did not downgrade for indirectness for breast cancer mortality but considered it serious for other outcomes. d The 95% CI limits crosses the decision threshold (as the CI is wide, a different clinical decision regarding the intervention may be taken depending on whether the lower or the higher limit is considered). e Despite concerns about indirectness from the trials, including the fact that the population age range of 40–74 is broader than the age range in this question, after considering evidence from contemporary non- randomised studies (Broeders et al.3) the GDG decided not to downgrade the quality of evidence for indirectness. f For the mortality-related outcomes, the GDG decided not to downgrade for imprecision because the relative effect is consistent with those in other age groups and that lends support that the estimate of the effect is close to what is reported here. This decision is also reinforced by the fact that, if the indirect evidence from the 50–69 age stratum were considered here, the certainty of the evidence for this outcome would also have been rated as ‘moderate’, as a result of downgrading that evidence from ‘high’ to ‘moderate’ by one level for indirectness and using it here. g Some studies were sub-optimally randomised and had non-blinded assessment of stage of disease; when analysis was restricted to low risk of bias trials, the risk estimate was non-significant. h Indirectness same as for women aged 50–69. i Non-blinded assessment of breast cancer stage is a serious concern. GDG members decided to downgrade to ‘serious’ for risk of bias. j Unexplained inconsistency with statistical heterogeneity (I2 ¼ 70%, p ¼ 0.02). While one study shows clear benefit, in three studies, the 95%CI does not exclude important benefit or harm. k. Analysis includes women aged 40–74 years, but only about 13% of women were 5 70 years. l In the group of older than 70 years, it only included tumour size 550 mm. Journal of Medical Screening 28(4) Table 3. Summary of available evidence on undesirable effects of breast cancer mammography screening by age groups. Quality assessment Summary of findings No of studies Certainty of Relative effect Absolute Age group (population) Design Comments evidence (95%CI) Basal risk reduction Canelo-Aybar et al. Overdiagnosis (population perspective) <50 1 (50,430) Randomised Downgraded due to indi- ⨁⨁⨁a –- 12.4% (95% CI 9.9–14.9%) trials rectness (serious). MODERATE 50–69 2 (64,117) Randomised Downgraded due to indi- ⨁⨁⨁ –- 10.1% (95% CI 8.6–11.6%) trials rectness (serious). MODERATE Overdiagnosis (individual perspective) <50 y 1 (50,430) Randomised Downgraded due to indi- ⨁⨁⨁a –- 22.7% (95% CI 18.4–27.0%) trials rectness due to indirect- MODERATE ness (serious). 50–69 2 (64,117) Randomised Downgraded due to indi- ⨁⨁⨁ –- 17.3% (95%CI 14.7–20.0%) trials rectness (serious). MODERATE Rate of mastectomies All groups 5 (249,550) Randomised Downgraded due to indi- ⨁⨁a,b,c RR 1.20 (1.11– 0.9% 180 more per trials rectness (very serious). LOW 1.30) 100,000 (from 99 more to 270 more) Provision of chemotherapy All groups 2 (99,454) Randomised Downgraded due to incon- ⨁d,e,f RR 0.86 0.4% 56 fewer per trials sistency (serious), indirect- VERY LOW (0.53 to 1.40) 100,000 ness (very serious) and (from 188 imprecision (serious). fewer to 160 more) Quality of life (inferred from psychological effects) All groups 1 SR including Observational No relevant considerations ⨁g –– It does not appear to create anxiety in 54 primary studies were found to justify VERY LOW women who are given a clear result studies. downgrading the certainty after a mammogram and subse- of the evidence quently placed on routine recall. Mixed results about anxiety in women recalled for further testing: several studies reported transient or long term (from 6 months to 1 year after recall) anxiety, while other studies reported no differences in anxiety levels. False-positive-related adverse effects (psychological distress) All groups 2 SR including Observational No relevant considerations ⨁⨁ –- Psychological distress in women that 24 primary studies were found to justify LOW needed further mammography: studies. RR ¼ 1.28, 95%CI 0.82-2.00; For (continued) 399 400 Table 3. Continued. Quality assessment Summary of findings No of studies Certainty of Relative effect Absolute Age group (population) Design Comments evidence (95%CI) Basal risk reduction downgrading the certainty women placed in early recall: of the evidence RR ¼ 1.82, 95%CI 1.22-2.72; For women that needed a fine needle puncture aspiration: RR ¼ 1.80, 95% CI 1.17-2.77; For women that needed a biopsy: RR ¼ 2.07, 95%CI 1.22-3.52. No differences in generic measures of general anxiety and depression at six weeks after assessment and three months after screening False-positive-related adverse effects (biopsies and surgeries) All groups One SR includ- Observational No relevant considerations ⨁⨁h –- Overall 2.9% pooled cumulative risk of ing four pri- studies were found to justify LOW an invasive procedure with benign mary downgrading the certainty outcome (range 1.8% to 6.3% based studies and a of the evidence on 2 studies) and 0.9% risk of cross section- undergoing surgical intervention with al data benign outcome (based on 1 study) analysis. (Hofvind 2012).54 Cross-sectional data from the EUNICE Project showed that 2.2% and 1.1% of all screening examinations resulted in needle biopsy among women without breast cancer (initial and subsequent screens).54 a Trials were conducted more than 20 years ago. Currently, women have higher adherence to breast cancer screening and the quality control of screening and the care of breast cancer have improved. A large non- randomised study (Hellquist B 2011) showed a reduced risk for breast cancer deaths in women aged 40–49 years invited to screening, compared with women not invited (RR ¼ 0.74; 95%CI, 0.66–0.83) which is consistent with the results seen in the RCTs. The GDG did not downgrade for indirectness for breast cancer mortality but considered it serious for other outcomes. b Some studies used methods that would not be accepted for random allocation today. One study had non-blinded assessment of ‘cause of death’. The GDG felt that the CNBSS-1 possibly had issues with achieving prognostic balance. The GDG felt that lack of allocation concealment in this set of studies did not lead to high risk of bias. Given the lack of single trials driving the overall results and similarity in effect sizes (the test for subgroup differences – low vs high risk of bias trials – was non-significant) and overlapping confidence intervals (CIs), the risk of bias was rated as ‘not serious’. c Population include women aged 40–74. Therefore, a much broader age range than the age group studied here. Observational studies do not confirm these results; instead, they provide opposite results. d 95% CI probably crosses the clinical decision threshold (as the CI is wide, a different clinical decision regarding the intervention may be taken depending on whether the lower or the higher limit is considered). e Unexplained inconsistency with statistical heterogeneity (I2 ¼ 71%, P ¼ 0.06). f Chemotherapy protocols and indications have significantly changed (e.g. node status was not determined in earlier studies). g Unexplained inconsistency for variability in anxiety in the group of women recalled for further testing. h Studies included women aged 50–69. Estimates for the 45–49 age stratum are likely to be higher. Journal of Medical Screening 28(4) Canelo-Aybar et al. 401 the consistency with more recent observational studies (see ecological studies have yielded conflicting results; for Discussion section), the GDG did not consider downgrad- example, the incidence of BC stages II–IV has been ing the evidence for this reason. reported to remain unchanged since the introduction of screening in the Netherlands.64 To overcome these limita- Discussion tions, a systematic review suggested to use the primary tumour size as the most direct link to radiological detec- Main findings tion;65 their findings from observational studies suggested a reduction in BC advanced stages after the introduction Our review shows that there is high certainty evidence that of screening.65 mammography screening reduces the risk of BC mortality We observed an increased risk of mastectomies, which in women between the ages of 50 and 69, with the number has not been consistently described in population studies. of deaths averted ranging from 138 fewer to 483 fewer per One Canadian study found that mastectomies were less 100,000 women invited to screening, depending on the frequently performed in screening attendees,59 while baseline risk assumed (from 0.6 to 2.1%). For other age women diagnosed in a New Zealand screening programme groups, the evidence is not conclusive. Consistently, were more likely to undergo conservative surgery;66 similar women invited to screening across all age groups showed results were observed in women aged 40–49 from the US.67 a lower risk of advanced stages of BC. One Norwegian study reported that mammography There is moderate certainty that screening is associated screening was associated with an increase in mastectomy with an increase in undesirable effects. Especially important rates, which later declined, likely explained by changes in was overdiagnosis, regardless of the calculation method, recommended surgical approaches.68 It is noteworthy that which was larger from an individual perspective in the youn- a recent systematic review found that adherence to guide- ger age groups compared to older groups.7 Mammography line recommendations on breast-conserving surgery is screening did not appear to produce significant negative psy- highly variable (35–95%).69 Thus, the increase in mastec- chological effects as long as the results were clearly commu- tomies among RCTs may partly be due to lead time bias, nicated, while false-positive results, especially when further the progress in BC care, or variation in clinical practice. assessment is required, increased the number of invasive Our estimates on chemotherapy are limited by changes procedures and psychological distress. in clinical practices, but recent observational studies sug- gest similar results. One study using Italian population Our results in the context of previous research cancer registries, from 2009 to 2013, observed that the Consistent with our analysis, observational studies suggest neo-adjuvant therapy indication was lower in provinces a reduction of BC mortality after screening implementa- where a screening programme had been present for tion. A systematic review of time trend studies estimated a many years.70 Another study that identified women aged BC mortality reduction from 1 to 9% per year and from 28 40–79 with incident BC from the British Columbia Cancer to 36% in studies comparing post- and pre-screening peri- Registry (Canada) found that chemotherapy use was ods.3 A pooled analysis of seven incidence-based mortality lower among regular screening participants after adjust- (IBM) studies from European countries showed a mortal- ment for age.71 ity reduction of 25% among invited women and 38% Our overdiagnosis estimates were in the range described among those actually screened.3 Another review which in the literature for European screening programmes, classified studies according to the quality of the methods which have been roughly estimated to range from 0 to used to estimate the expected mortality in absence of 54% using unadjusted data and from 1 to 10% after screening found an IBM risk reduction of 26% in adjustment for BC risk and lead time.72 However, a pro- women invited for screening from studies with robust portion of the excess of incidence from the CNBSS trials approaches.58 occurred years after screening ceased in the intervention Our results suggest that screened women are diagnosed arm24 and should not be considered as overdiagnosis. with less advanced disease. Observational evidence shows Overall, the certainty of the evidence of overdiagnosis earlier BC staging at diagnosis in women who had received was moderate, due to potential RoB, as women in the mammography screening. One Canadian registry-based control group of the CNBSS trial might have received study showed that screening attendees were more likely opportunistic or programmatic mammography screening to have in-situ disease alone, and in those with invasive at the end of the trial period. cancer, a lower proportion of grade III histology.59 There is no consensus about the method to estimate Furthermore, two studies using the SEER-Medicare data- overdiagnosis. Most common approaches assess the differ- base showed that extending mammography screening to ence in cancer incidence in the presence and absence of elderly women decreased advanced stage at diagnosis.60,61 screening or make inferences about the lead time of The interpretation of BC stage results from RCTs is BC.73 One study observed that a long follow-up time is precluded by stage migration bias due to the introduction needed to account for lead time, as the excess of cumula- of sentinel lymph node dissection62 and by modifications tive BC incidence will fall below 10% after a follow-up of in coding and classification practices.63 Consequently, 25 years in a simulated population.74 Another study, 402 Journal of Medical Screening 28(4) applying a micro-simulation model to the Netherlands a new search in June 2018, and after looking at the results, population, found that estimations made in earlier the GDG decided not to update the systematic review. phases of the screening programme may overestimate Our review also has strengths: we used rigorous meth- overdiagnosis by a factor of 4, underlining the relevance ods including the GRADE approach to rate the certainty of allowing an appropriate follow-up time to obtain reli- of the evidence and included the longest follow-up data able estimates.75 available from the RCTs and systematic reviews of obser- There are discrepancies among previous systematic vational studies. In contrast to previous systematic reviews on the assessment of RoB.4,76 In particular, a reviews, the consideration of contextual evidence allowed Cochrane review considered that only the CNBSS, us to rate the certainty of evidence for BC mortality as Malmo and UK Age trials were at low RoB, showing a high for women aged 50–69 and 70–74 and moderate for non-significant effect of mammography screening on BC women aged 45–49. We also provided results stratified by mortality from those trials.4 We considered the CNBSS age groups of interest for women, clinicians and trial as high RoB due to allocation of women by using policymakers. open lists, and the inclusion of a clinical examination before randomisation which could have led to differential Conclusions assignment;27,28 thus, only two RCTs were at low RoB with similar BC mortality effects when compared to the Our findings have different implications depending on the remaining studies. Consequently, in the 50–69 years strata, stakeholder group. Guideline panellists (and policy makers) we did not downgrade our certainty for RoB due to the are more likely to formulate strong recommendations in similarity in effect estimates across studies. women in the 50–69 age group than in other groups. In Some authors have proposed all-cause mortality as a women under 50 or over 69, where the balance is less clear, better estimate of screening impact. This measure would conditional recommendations are more likely. Moreover, be less prone to ascertainment bias of the cause of death, panels may specify further subgroups among women below an issue described in the Swedish trials with higher all- 50, where baseline risk changes rapidly and recommendations cause mortality in the control group. However, authors could vary between the 40–44 and 45–49 age groups, as in the of the Swedish trials4,77 reported no significant increased ECIBC guidelines13 (https://ecibc.jrc.ec.europa.eu/recommen dations/). Although informed decision-making should be rec- rate of death from other causes after appropriate adjust- ommended in all age groups, this will be especially important ments for age distribution and lead time bias were imple- in these age groups where the balance is less clear. mented.18,47 Moreover, all-cause mortality would be an A number of research priorities were identified with inefficient measure given the unfeasibly large sample size input from the GDG experts, which included: assessing required to detect differences between groups.47 We com- the impact of different screening intervals; the identifica- plemented our estimation of mortality impact with the tion of risk factors to stratify women who should start results for other-cause mortality which suggested no dif- screening earlier (or at shorter examination intervals, ference between women invited or not to screening. such as women with dense breast tissue); better assessment Balancing potential benefits and overdiagnosis, we esti- of the magnitude of overdiagnosis with an emphasis on mate for 100,000 women invited to screening from age 50 methods to estimate the actual impact across age groups; to 69, at least 138 BC deaths would be avoided and 3240 and the use of new technologies for screening (i.e. BC would be diagnosed (2.7 per 1000 annual rate- tomosynthesis).  20 years  0.6 mammography adherence) of which 550 (17%, individual perspective) could be overdiagnosed. Availability of data and materials Thus, for each BC death avoided, approximately four All data sources used during this study are described in this published article and its overdiagnosed cases will be managed. This estimate is in additional information files. The datasets analysed are available from the corre- the range of previous systematic assessments of screen- sponding author on reasonable request. ing.78 However, the potential bias in the overdiagnosis Acknowledgements estimates means that this figure remains tentative. The authors would like to sincerely thank all members of the Guidelines Development Group of the European Commission Initiative on Breast Cancer Limitations and strengths for their participation in the discussions generated by this systematic review which led to the different recommendations they developed in the European Our systematic review has some limitations, as no RCTs Guidelines on Breast Cancer Screening and diagnosis (https://healthcare-quality. have sufficient statistical power to assess the benefit of jrc.ec.europa.eu/european-breast-cancer-guidelines). screening on BC mortality according to age subgroups. Authors’ contributions Additionally, we included only English language articles; Carlos Canelo-Aybar, Diogenes Seraphim Ferreira, Monica Ballesteros, Margarita however, the risk of selection bias is probably small Posso, Nadia Montero and Ivan Solá, Pablo Alonso-Coello were responsible for because we screened previous systematic reviews, and the conducting the systematic review. Monica Ballesteros, Margarita Posso, Nadia GDG includes several international experts, making the Montero and Ivan Solá conducted the search and data extraction. Paolo Giorgi Rossi, Stephen Duffy, Markus Follmann, Zuleika Saz-Parkinson, and Axel possibility of missing studies unlikely. Although our orig- Gr€ awingholt contributed to the definition of the research protocol and provided inal search was conducted up to April 2016, we conducted comments to the preliminary results of the systematic review. Carlos Canelo-Aybar Canelo-Aybar et al. 403 and Pablo Alonso-Coello drafted the first version of the article. All authors con- 14. Guyatt GH, Oxman AD, Kunz R, et al. GRADE guidelines: 2. Framing the tributed to the interpretation and reporting of the results and provided comments question and deciding on important outcomes. J Clin Epidemiol 2011; 64: on subsequent versions of the article. All authors read and approved the final 395–400. manuscript prior submission. 15. Higgins JP, Altman DG, Gotzsche PC, Cochrane Statistical Methods Group, et al. The cochrane collaboration’s tool for assessing risk of bias in randomised trials. BMJ 2011; 343: d5928. Declaration of conflicting interests 16. Shea BJ, Hamel C, Wells GA, et al. AMSTAR is a reliable and valid measure- The author(s) declared the following potential conflicts of interest with respect to ment tool to assess the methodological quality of systematic reviews. J Clin the research, authorship, and/or publication of this article: Zuleika Saz-Parkinson Epidemiol 2009; 62: 1013–1020. and Donata Lerda are current employees of the Joint Research Centre, European 17. Nystrom L, Larsson LG, Rutqvist LE, et al. Determination of cause of death Commission. Carlos Canelo-Aybar, Diogenes Seraphim Ferreira, Monica among breast cancer cases in the Swedish randomized mammography screening Ballesteros, Margarita Posso, Nadia Montero, Ivan Solà and Pablo Alonso- trials. A comparison between official statistics and validation by an endpoint Coello are employees of the Iberoamerican Cochrane Collaboration. Paolo committee. Acta Oncol 1995; 34: 145–152. Giorgi Rossi, Stephen Duffy, Markus Follmann, and Axel Gr€ awingholt are mem- 18. Nystr€ om L, Andersson I, Bjurstam N, et al. Long-term effects of mammogra- bers of the ECIBC Guidelines Development Group. Diogenes Seraphim Ferreira phy screening: updated overview of the Swedish randomised trials. Lancet was supported by the fellowship MTF 2015–02 from the European Respiratory 2002; 359: 909–919. Society during the conduct of the study. 19. Guyatt GH, Oxman AD, Schunemann HJ, et al. GRADE guidelines: a new series of articles in the journal of clinical epidemiology. J Clin Epidemiol 2011; 64: 380–382. Funding 20. Guyatt GH, Oxman AD, Vist GE, GRADE Working Group, et al. GRADE: The author(s) disclosed receipt of the following financial support for the research, an emerging consensus on rating quality of evidence and strength of recom- authorship, and/or publication of this article: The systematic review was carried out mendations. BMJ 2008; 336: 924–926. by Iberoamerican Cochrane Collaboration under the Framework contract 443094 21. Roberts MM, Alexander FE, Anderson TJ, et al. The Edinburgh randomised for procurement of services between the European Commission’s Joint Research trial of screening for breast cancer: description of method. Br J Cancer 1984; Centre and Asociacion Colaboraci on Cochrane Iberoamericana. 50: 1–6. 22. Alexander FE, Anderson TJ, Brown HK, et al. 14 years of follow-up from the ORCID iDs Edinburgh randomised trial of breast-cancer screening. Lancet 1999; 353: 1903–1908. Zuleika Saz-Parkinson https://orcid.org/0000-0002-3518-5899 23. Chu KC, Smart CR and Tarone RE. Analysis of breast cancer mortality and Paolo G Rossi https://orcid.org/0000-0001-9703-2460 stage distribution by age for the health insurance plan clinical trial. J Natl Stephen W Duffy https://orcid.org/0000-0003-4901-7922 Cancer Inst 1988; 80: 1125–1132. Axel Gr€awingholt https://orcid.org/0000-0003-4778-8731 24. Habbema JD, van Oortmarssen GJ, van Putten DJ, et al. Age-specific reduc- tion in breast cancer mortality by screening: an analysis of the results of the Supplemental material health insurance plan of greater New York study. J Natl Cancer Inst 1986; 77: Supplemental material for this article is available online. 317–320. 25. Shapiro S, Strax P, Venet L, et al. Proceedings: changes in 5-year breast cancer mortality in a breast cancer screening program. Proc Natl Cancer Conf 1972; 7: References 663–678. 1. Ferlay JE, Lam F, Colombet M, et al. Global cancer observatory: cancer today. 26. Shapiro S. Periodic screening for breast cancer: the HIP randomized controlled Lyon, France: International Agency for Research on Cancer, https://gco.iarc. trial. Health insurance plan. J Natl Cancer Inst Monogr 1997; 22: 27–30. 27. Miller AB, To T, Baines CJ, et al. Canadian national breast screening study-2: fr/today (2018, accessed 25 July 2019). 13-year results of a randomized trial in women aged 50-59 years. J Natl Cancer 2. ECIS. European Cancer Information System, https://ecis.jrc.ec.europa.eu Inst 2000; 92: 1490–1499. (2018, accessed 20 July 2019). 28. Miller AB, To T, Baines CJ, et al. The Canadian national breast screening 3. Broeders M, Moss S, Nystrom L, EUROSCREEN Working Group, et al. The study-1: breast cancer mortality after 11 to 16 years of follow-up. A random- impact of mammographic screening on breast cancer mortality in Europe: a ized screening trial of mammography in women age 40 to 49 years. Ann Intern review of observational studies. J Med Screen 2012; 19 (Suppl 1): 14–25. Med 2002; 137: 305–312. 4. Gotzsche PC and Jorgensen KJ. Screening for breast cancer with mammogra- 29. Miller AB, Wall C, Baines CJ, et al. Twenty five year follow-up for breast phy. Cochrane Database Syst Rev 2013; CD001877. cancer incidence and mortality of the Canadian national breast screening 5. Shapiro S. Screening: assessment of current studies. Cancer 1994; 74: 231–238. study: randomised screening trial. BMJ 2014; 348: g366. 6. Hubbard RA, Kerlikowske K, Flowers CI, et al. Cumulative probability of 30. Miller AB, Baines CJ, To T, et al. Canadian national breast screening study. 1. false-positive recall or biopsy recommendation after 10 years of screening mam- Breast cancer detection and death rates among women aged 40 to 49 years. mography: a cohort study. Ann Intern Med 2011; 155: 481–492. Canad Med Assoc J 1992; 147: 1459–1476. 7. Marmot MG, Altman DG, Cameron DA, et al. The benefits and harms of 31. Miller AB. The costs and benefits of breast cancer screening. Am J Prev Med breast cancer screening: an independent review. Br J Cancer 2013; 108: 1993; 9: 175–180. 2205–2240. 32. Miller AB, Baines CJ, To T, et al. Canadian national breast screening study. 2. 8. WHO position paper on mammography screening. Geneva: World health Breast cancer detection and death rates among women aged 50 to 59 years. Organisation, 2014. Canad Med Assoc J 1992; 147: 1477–1488. 9. Tonelli M, Gorber SC, Joffres M, et al. Recommendations on screening for 33. Moss S. A trial to study the effect on breast cancer mortality of annual mam- breast cancer in average-risk women aged 40-74 years. Canad Med Assoc J mographic screening in women starting at age 40. Trial steering group. J Med 2011; 183: 1991–2001. Screen 1999; 6: 144–148. 10. Oeffinger KC, Fontham ET, Etzioni R, American Cancer Society, et al. Breast 34. Moss S, Waller M, Anderson TJ, Trial Management Group, et al. Randomised cancer screening for women at average risk: 2015 guideline update from the controlled trial of mammographic screening in women from age 40: predicted American cancer society. JAMA 2015; 314: 1599–1614. mortality based on surrogate outcome measures. Br J Cancer 2005; 92: 11. Schunemann HJ, Lerda D, Dimitrova N, for the European 955–960. Commission Initiative on Breast Cancer Contributor Group, et al. Methods 35. Moss SM, Cuckle H, Evans A, Trial Management Group, et al. Effect of for development of the European commission initiative on breast cancer guide- mammographic screening from age 40 years on breast cancer mortality at 10 lines: recommendations in the era of guideline transparency. Ann Intern Med years’ follow-up: a randomised controlled trial. Lancet 2006; 368: 2053–2060. 2019; 171: 273. 36. Frisell J, Glas U, Hellstrom L, et al. Randomized mammographic screening for 12. (ECIBC) ECIOBC. Recommendations from European Breast Guidelines, breast cancer in Stockholm. Design, first round results and comparisons. Breast https://ecibc.jrc.ec.europa.eu/recommendations/list/3 (2018, accessed 29 Cancer Res Treat 1986; 8: 45–54. March 2018). 37. Frisell J, Lidbrink E, Hellstrom L, et al. Followup after 11 years – update of 13. Schunemann HL, Quinn C, Follmann M, for the European Commission mortality results in the Stockholm mammographic screening trial. Breast Initiative on Breast Cancer (ECIBC) Contributor Group, et al. Breast cancer Cancer Res Treat 1997; 45: 263–270. screening and diagnosis: a synopsis of the European breast guidelines. Ann 38. Frisell J. Mammographic screening for breast cancer. PhD Thesis. Stockholm: Intern Med 2020; 172: 46. S€odersjukhuset, 1989. 404 Journal of Medical Screening 28(4) 39. Andersson I, Aspegren K, Janzon L, et al. Mammographic screening and mor- 60. Randolph WM, Goodwin JS, Mahnken JD, et al. Regular mammography use tality from breast cancer: the malmo mammographic screening trial. BMJ 1988; is associated with elimination of age-related disparities in size and stage of 297: 943–948. breast cancer at diagnosis. Ann Intern Med 2002; 137: 783–790. 40. Zackrisson S, Andersson I, Janzon L, et al. Rate of over-diagnosis of breast 61. McCarthy EP, Burns RB, Freund KM, et al. Mammography use, breast cancer cancer 15 years after end of malmo mammographic screening trial: follow-up stage at diagnosis, and survival among older women. J Am Geriatr Soc 2000; study. BMJ 2006; 332: 689–692. 48: 1226–1233. 41. Andersson I and Janzon L. Reduced breast cancer mortality in women under 62. Maaskant AJ, van de Poll-Franse LV, Voogd AC, et al. Stage migration due to age 50: updated results from the malmo mammographic screening program. J introduction of the sentinel node procedure: a population-based study. Breast Natl Cancer Inst Monogr 1997; 22: 63–67. Cancer Res Treat 2009; 113: 173–179. 42. Bjurstam N, Bj€ orneld L, Duffy SW, et al. The Gothenburg breast screening 63. Larsen IK, Myklebust TA, Johannesen TB, et al. Stage-specific incidence and trial: first results on mortality, incidence, and mode of detection for women survival of breast cancer in Norway: the implications of changes in coding and ages 39-49 years at randomization. Cancer 1997; 80: 2091–2099. classification practice. Breast 2018; 38: 107–113. 43. Bjurstam N, Bjorneld L, Warwick J, et al. The Gothenburg breast screening 64. Autier P, Boniol M, Koechlin A, et al. Effectiveness of and overdiagnosis from trial. Cancer 2003; 97: 2387–2396. mammography screening in the Netherlands: population based study. BMJ 44. Tabar L, Fagerberg G, Duffy SW, et al. The Swedish two county trial of 2017; 359: j5224. mammographic screening for breast cancer: recent results and calculation of 65. Broeders MJM, Allgood P, Duffy SW, et al. The impact of mammography benefit. J Epidemiol Community Health 1989; 43: 107–114. screening programmes on incidence of advanced breast cancer in Europe: a 45. Tabar L, Fagerberg G, Chen HH, et al. Efficacy of breast cancer screening by age. New results from the Swedish Two-County trial. Cancer 1995; 75: literature review. BMC Cancer 2018; 18: 860. 2507–2517. 66. Samnakay N, Tinning J, Ives A, et al. Rates for mastectomy are lower in 46. Tabar L, Chen HH, Duffy SW, et al. Primary and adjuvant therapy, prognostic women attending a breast-screening programme. ANZ J Surg 2005; 75: factors and survival in 1053 breast cancers diagnosed in a trial of mammogra- 936–939. phy screening. Jpn J Clin Oncol 1999; 29: 608–616. 67. James TA, Wade JE and Sprague BL. The impact of mammographic screening 47. Tabar L, Duffy SW, Yen MF, et al. All-cause mortality among breast cancer on the surgical management of breast cancer. J Surg Oncol 2016; 113: 496–500. patients in a screening trial: support for breast cancer mortality as an end point. 68. Suhrke P, Mæhlen J, Schlichting E, et al. Effect of mammography screening on J Med Screen 2002; 9: 159–162. surgical treatment for breast cancer in Norway: comparative analysis of cancer 48. Tabar L, Vitak B, Chen TH, et al. Swedish Two-County trial: impact of mam- registry data. BMJ 2011; 343: d4692. mographic screening on breast cancer mortality during 3 decades. Radiology 69. Ni~ no de Guzmán E, Song Y, Alonso-Coello P, et al. Healthcare providers’ 2011; 260: 658–663. adherence to breast cancer guidelines in Europe: a systematic literature 49. Tabar L, Vitak B, Chen HH, et al. The Swedish Two-County trial twenty years review. Breast Cancer Res Treat 2020; 181: 499–518. later. Updated mortality results and new insights from long-term follow-up. 70. Mangone L, Mancuso P, Tagliabue G, et al. Neoadjuvant therapy for breast Radiol Clin N Am 2000; 38: 625–651. cancer. Tumori 2019; 105: 488–493. 50. Moss SM, Wale C, Smith R, et al. Effect of mammographic screening from age 71. Coldman AJ, Phillips N and Speers C. A retrospective study of the effect of 40 years on breast cancer mortality in the UK age trial at 17 years’ follow-up: a participation in screening mammography on the use of chemotherapy and randomised controlled trial. Lancet Oncol 2015; 16: 1123–1132. breast conserving surgery. Int J Cancer 2007; 120: 2185–2190. 51. Bjurstam NG, Bjorneld LM and Duffy SW. Updated results of the Gothenburg 72. Puliti D, Duffy SW, Miccinesi G, EUROSCREEN Working Group, et al. trial of mammographic screening. Cancer 2016; 122: 1832–1835. Overdiagnosis in mammographic screening for breast cancer in Europe: a lit- 52. Bond M, Pavey T, Welch K, et al. Psychological consequences of false-positive erature review. J Med Screen 2012; 19 (Suppl 1): 42–56. screening mammograms in the UK. Evid Based Med 2013; 18: 54–61. 73. Nelson HD, Pappas M, Cantor A, et al. Harms of breast cancer screening: 53. Brett J, Bankhead C, Henderson B, et al. The psychological impact of mam- systematic review to update the 2009 U.S. Preventive services task force rec- mographic screening. A systematic review. Psychooncology 2005; 14: 917–938. ommendation. Ann Intern Med 2016; 164: 256–267. 54. Hofvind S, Ponti A, Patnick J, EUNICE Project and Euroscreen Working 74. Duffy SW and Parmar D. Overdiagnosis in breast cancer screening: the impor- Groups, et al. False-positive results in mammographic screening for breast tance of length of observation period and lead time. Breast Cancer Res 2013; cancer in Europe: a literature review and survey of service screening pro- 15: R41. grammes. J Med Screen 2012; 19 (Suppl 1): 57–66. 75. de Gelder R, Heijnsdijk EA, van Ravesteyn NT, et al. Interpreting overdiag- 55. Salz T, Richman AR and Brewer NT. Meta-analyses of the effect of false- nosis estimates in population-based mammography screening. Epidemiol Rev positive mammograms on generic and specific psychosocial outcomes. 2011; 33: 111–121. Psychooncology 2010; 19: 1026–1034. 76. Nelson HD, Fu R, Cantor A, et al. Effectiveness of breast cancer screening: 56. Hofvind S, Thoresen S and Tretli S. The cumulative risk of a false-positive recall in the Norwegian breast cancer screening program. Cancer 2004; 101: systematic review and meta-analysis to update the 2009 U.S. preventive services 1501–1507. task force recommendation. Ann Intern Med 2016; 164: 244–255. 57. Salas D, Ibanez J, Roman R, CFPR (Cumulative False Positive Risk) group, 77. Black WC, Haggstrom DA and Welch HG. All-cause mortality in randomized et al. Effect of start age of breast cancer screening mammography on the risk of trials of cancer screening. J Natl Cancer Inst 2002; 94: 167–173. false-positive results. Prev Med 2011; 53: 76–81. 78. The Research Council of Norway: research-based evaluation of the 58. Njor S, Nystrom L, Moss S, Euroscreen Working Group, et al. Breast cancer Norwegian breast cancer screening program. Lysaker: Division for Society mortality in mammographic screening in Europe: a review of incidence-based and Health, 2015. mortality studies. J Med Screen 2012; 19 (Suppl 1): 33–41. 79. Hellquist BN, Duffy SW, Abdsaleh S, et al. Effectiveness of population-based 59. Olivotto IA, Mates D, Kan L, et al. Prognosis, treatment, and recurrence of service screening with mammography for women ages 40 to 49 years: evalua- breast cancer for women attending or not attending the screening mammogra- tion of the Swedish Mammography Screening in Young Women (SCRY) phy program of British Columbia. Breast Cancer Res Treat 1999; 54: 73–81. cohort. Cancer 2011; 117: 714–722.