European Radiology (2021) 31:5880–5893 https://doi.org/10.1007/s00330-021-07873-2 BREAST Preoperative breast magnetic resonance imaging in patients with ductal carcinoma in situ: a systematic review for the European Commission Initiative on Breast Cancer (ECIBC) Carlos Canelo-Aybar 1,2 & Alvaro Taype-Rondan 3 & Jessica Hanae Zafra-Tanaka 4 & David Rigau 2 & Axel Graewingholt 5 & Annette Lebeau 6 & Elsa Pérez Gómez 7 & Paolo Giorgi Rossi 8 & Miranda Langendam 9 & Margarita Posso 2,10 & Elena Parmelli 11 & Zuleika Saz-Parkinson 11 & Pablo Alonso-Coello 1,2 Received: 23 October 2020 / Revised: 18 February 2021 / Accepted: 11 March 2021 / Published online: 30 May 2021 # The Author(s) 2021, corrected publication 2021 Abstract Objective To evaluate the impact of preoperative MRI in the management of Ductal carcinoma in situ (DCIS). Methods We searched the PubMed, EMBASE and Cochrane Library databases to identify randomised clinical trials (RCTs) or cohort studies assessing the impact of preoperative breast MRI in surgical outcomes, treatment change or loco-regional recur- rence. We provided pooled estimates for odds ratios (OR), relative risks (RR) and proportions and assessed the certainty of the evidence using the GRADE approach. Results We included 3 RCTs and 23 observational cohorts, corresponding to 20,415 patients. For initial breast-conserving surgery (BCS), the RCTs showed that MRI may result in little to no difference (RR 0.95, 95% CI 0.90 to 1.00) (low certainty); observational studies showed that MRI may have no difference in the odds of re-operation after BCS (OR 0.96; 95% CI 0.36 to 2.61) (low certainty); and uncertain evidence from RCTs suggests little to no difference with respect to total mastectomy rate (RR 0.91; 95% CI 0.65 to 1.27) (very low certainty). We also found that MRI may change the initial treatment plans in 17% (95% CI 12 to 24%) of cases, but with little to no effect on locoregional recurrence (aHR = 1.18; 95% CI 0.79 to 1.76) (very low certainty). Conclusion We found evidence of low to very low certainty which may suggest there is no improvement of surgical outcomes with pre-operative MRI assessment of women with DCIS lesions. There is a need for large rigorously conducted RCTs to evaluate the role of preoperative MRI in this population. Key Points • Evidence of low to very low certainty may suggest there is no improvement in surgical outcomes with pre-operative MRI. • There is a need for large rigorously conducted RCTs evaluating the role of preoperative MRI to improve treatment planning for DCIS. Keywords Breast cancer . Ductal carcinoma in situ . Magnetic resonance imaging * Carlos Canelo-Aybar 5 Radiologie am Theater, Paderborn, Germany ccanelo@santpau.cat 6 Institute of Pathology, University Medical Center * Elena Parmelli Hamburg-Eppendorf, Hamburg, Germany Elena.Parmelli@ec.europa.eu 7 University Hospital Dr. Josep Trueta, Girona, Spain 1 CIBER de Epidemiología y Salud Pública (CIBERESP), 8 Epidemiology Unit, Azienda USL – IRCCS di Reggio Emilia, Madrid, Spain Reggio Emilia, Italy 2 Iberoamerican Cochrane Centre - Department of Clinical 9 Department of Epidemiology and Data Science, Amsterdam UMC, Epidemiology and Public Health, Biomedical Research Institute Sant University of Amsterdam, Amsterdam Public Health Institute, Pau (IIB Sant Pau), Sant Antonio María Claret 167, Amsterdam, Netherlands 08025 Barcelona, Spain 10 3 Universidad San Ignacio de Loyola, Unidad de Investigación para la Department of Epidemiology and Evaluation, IMIM (Hospital del Generación y Síntesis de Evidencias en Salud, Lima, Peru Mar Medical Research Institute), Barcelona, Spain 4 11 CRONICAS Centre of Excellence in Chronic Diseases, Universidad European Commission, Joint Research Centre (JRC), Via E. Fermi, Peruana Cayetano Heredia, Lima, Peru 2749. TP127, I-21027 Ispra, VA, Italy Eur Radiol (2021) 31:5880–5893 5881 Abbreviations informed the recommendations of preoperative breast MRI aOR Adjusted odds ratios (Prospero register: 42018099453). During the guidelines pro- BCS Breast-conserving-surgery cess [13], the Guidelines Development Group (GDG) made DCIS Ductal carcinoma in situ detailed considerations on the evidence about effects, values ECIBC European Commission Initiative and preferences, equity, acceptability and feasibility to issue on Breast Cancer recommendations. We encourage readers to refer to these con- GDG Guidelines Development Group siderations in the published recommendations on the ECIBC IBC Invasive breast carcinomas website (https://healthcare-quality.jrc.ec.europa.eu/european- MRI Magnetic resonance imaging breast-cancer-guidelines/surgical-planning/MRI) OR Odds ratios PICO Population, Intervention, Comparison and Outcomes RCTs Randomised clinical trials Methods ROBINS-I Risk Of Bias in Non-randomised Studies of Interventions-I Structured question and outcome prioritisation RR Relative risks SoF Summary of findings The clinical question prioritised by the GDG was “Should additional MRI vs no additional MRI be used in women with histologically confirmed DCIS for preoperative planning?”. Introduction Outcomes were prioritised using a 1 to 9 scale as suggested by the Grading of Recommendations Assessment, In 2018, globally, an estimated 2 million new cases of breast Development and Evaluation (GRADE) approach (Box 1). cancer (BC) were reported [1]. Ductal carcinoma in situ Box 1 Structured clinical question (DCIS) of the breast is the most common form of non- invasive BC, and includes a heterogeneous group of atypical Population Intervention Comparison Outcomes cell proliferation confined within the basement membrane of Women with Preoperative No • MRI triggered the ducts [2]. Over the last decades, the detection of DCIS has confirmed breast preopera- treatment change, as increased, likely because of the widespread use of screening DCIS on MRI tive breast the decision to mammography, accounting for 20 to 25% of newly diagnosed preoperative MRI perform a wider BC in screened populations [3]. histology excision, a mastectomy or a Nowadays, breast-conserving surgery (BCS) has been bilateral adopted as a treatment option for patients with small, screen- mastectomy when a detected lesions [4]. The addition of adjuvant radiation and more conservative hormonal therapy after BCS has been shown to reduce the risk approach were originally planned of invasive recurrence [4]. However, complete surgical exci- before MRI results sion is not always possible due to the suboptimal preoperative • Initial evaluation of the extent of the lesion by standard imaging breast-conserving (mammography, ultrasound). Therefore, re-operation for pos- surgery (BCS), a patient not undergo- itive margins is often required in DCIS, with rates ranging ing mastectomy from 17 to 58% [5]. within the initial Magnetic resonance imaging (MRI) has a higher sensitivity surgical treatmenta for BC diagnosis preferentially detecting more aggressive/ • Re-operation after breast-conserving invasive types [6]. It has been proposed as an additional test after surgery, either a mammography, to improve the assessment of the extent of DCIS wider local excision during the preoperative planning, providing better identification or mastectomy after of candidates for BCS especially in the context of extensive the first surgery • Proportion of microcalcifications [7]. Despite previous reviews suggesting positive margins MRI benefits [8–10], there is still uncertainty as some studies after suggest it may overestimate the extent of disease, leading to an breast-conserving increase of unnecessary mastectomies or wider excisions [11, 12]. surgery, absence of clear margins at the The European Commission Initiative on Breast Cancer pathologic assess- (ECIBC) develops the European Guidelines on Breast ment of the Cancer Screening and Diagnosis [13]. This systematic review 5882 Eur Radiol (2021) 31:5880–5893 (continued) Data analysis specimen after sur- gical resection From RCTs, we extracted crude relative risks (RR), and from • Total mastectomy, observational studies, we obtained adjusted odds ratios (aOR) the last definitive mastectomy, or hazard ratios (aHR) when available. We did not pool the including initial and results obtained from both types of designs. Pooled effect additional sizes were estimated using a random effects model with the mastectomy due to Mantel-Haenzel or inverse variance method. To estimate re-operation • Disease-free survival between-study variance and confidence intervals, we used (inferred from the Paule-Mendel and Q-profile methods. To pool the propor- loco-regional recur- tion of MRI-treatment changes, we implemented a generalised rence) linear mixed random model with a logit transformation, and • Quality of life the Clopper-Pearson method to estimate the confidence inter- a An increase in this outcome is a desirable change as it is a complementary val for individual study results. event to initial mastectomy Heterogeneity between studies was evaluated by visual in- spection of forest plots for all outcomes and complemented Data sources and searches with the assessment of the Q statistic and I2 parameter for relative effects, as they are not recommended for proportions We searched the MEDLINE (via PubMed, April 2018), [16]. The following potential sources of heterogeneity were EMBASE (via Ovid, April 2018) and CENTRAL (via The examined: the extent of microcalcifications, risk of bias, ad- Cochrane Library, March 2018) databases using pre-defined justed or crude effect sizes, prospective or retrospective design algorithms. In addition, we updated our initial search during for observational studies and publication year (post hoc). We the first week of January 2021 (Supplementary Table 1), and performed all analysis in RStudio. GDG members were consulted about potential missing studies. Certainty of the evidence We rated the certainty of the evidence separately for RCTs and Study selection observational studies for each of the prioritised outcomes using the GRADE approach [17]. For the assessment of the We included randomised controlled trials (RCT) and cohort certainty of observational evidence, we started from high cer- studies that compared preoperative MRI with no MRI in tainty, as this is the recommended procedure when the women with histologically confirmed DCIS. We excluded ROBINS-I tool is used [18]. studies that included women with invasive breast carcinomas (IBC), those that did not provide stratified results for women with DCIS, conference abstracts and articles published in lan- Results guages other than English. Initially, two calibrated reviewers (A.T.R. and J.Z.) Search results assessed the eligibility at title and abstract level. In a sec- ond step, the two reviewers independently reviewed the In the initial search until April 2018, we retrieved 5260 unique full text of all the selected references. Discrepancies were citations. Initially, we included a total of 20 studies; this was solved by consensus or with the help of a third reviewer the original evidence synthesis used to develop the ECIBC (CCA). recommendations. Finally, after the update search (January 2021), we included six additional studies and the update of a previously included cohort (Fig. 1). Reasons for exclusion are Data extraction and risk of bias assessment detailed in Supplementary table S2. Three reviewers (A.T.R., J.Z., C.C.A.) independently extract- Study characteristics ed data and assessed the risk of bias using the Cochrane Risk of Bias tool [14] for RCTs and the “Risk Of Bias in Non- Three studies were RCTs [19–21], 18 were comparative co- randomised Studies of Interventions-I” (ROBINS-I) for obser- horts [11, 12, 22–36] and five were single arm cohorts [12, 21, vational studies. Before applying the latter tool, we specified 37–39], with a total of 20,415 patients (260 from the RCTs relevant confounding variables (i.e. age, family history of BC, and 20,155 from the observational studies). Most studies were tumour size) [15]. conducted in the USA or in the Netherlands. The mean ages of Eur Radiol (2021) 31:5880–5893 5883 Records identified through multiple database search [March-April 2018] (n = 6,758) Identification Duplicate records removed (n = 1,498) Records screened (n = 5,260) Screening Records excluded (n = 5,155) Full-text articles assessed for eligibility (n = 105) Full-text articles excluded (n = 85): Updated search in January Different outcomes (n=1) 2021, individual records (n = 1,900) Do not show data separately for DCIS (n=72) Eligibility Do not include patients with DCIS (n=11) Studies included up to April 2018 (n = 20: 2 RCTs and 18 Studies included from references surveyed (n = 6) Included Total studies included in the review (n = 26: 3 RCTs and 23 observational studies) Fig. 1 Flowchart of study selection the recruited patients ranged from 40 to 63 years across all presenting microcalcifications and a mean size of 10 mm. studies. Only one study included all patients with Additionally, we included data from small subgroups of microcalcified lesions, while two studies reported the percent- patients with DCIS from two RCTs, the MONET trial age of microcalcified lesions (range 76 to 99%) (Table 1 and [19] which randomised patients with a non-palpable BI- Supplementary table S3). RADS 3–5 lesion to receive routine medical care (mam- One RCT was purposely designed to evaluate the value mography, ultrasound and lesion sampling) or additional of breast MRI in patients with biopsy-proven DCIS who MRI preceding biopsy and the multicentric COMICE trial were scheduled for BCS [21], with almost all lesions [20] which recruited patients with biopsy-proven primary Table 1 Characteristics of the included studies 5884 Author and year (trial) Country Design No. of patients Tumour characteristics Age (years), mean (range) Intervention (MRI) (only DCIS) (microcalcification, size) Balleyguier France RCT MRI: 176 Microcalcification: MRI: 56 (31-80) 1.5 T systems mainly and (IRCIS) 2019 [21] No MRI: 173 MRI: 98%; No MRI: 99% No MRI: 58 (39–80) 3 T systems in 2 centres Mean size (mm): MRI: 10, no MRI: 13 Peters (MONET) Netherlands RCT MRI: 39 Microcalcification: NR NR 3 T, dedicates phased-array 2011 [19] (subgroup analysis) No MRI: 41 Mean size (mm): NR bilateral breast coil Turnbull (COMICE) UK RCT MRI: 43 Microcalcification: NR NR 1.5 T, dedicated breast-surface 2010 [20] (subgroup analysis) No MRI: 48 Mean size (mm): NR coils for signal reception, with a few scans done at 1.0 T Allen 2010 [22] USA Retrospective cohort MRI: 64 Microcalcification: NR MRI: 60.5 (40–83) 1.5 T using 8 channel No MRI: 35 Mean size (mm): NR No MRI: 64.4 (41–89) breast-surface coil Besharat 2018 [40] Iran One arm retrospective MRI: 5 Microcalcification: NR All patients (DCIS 1.5 T cohort Mean size (mm): NR + invasive): 45.5 Davis 2012 [23] USA Retrospective cohort, MRI: 154 Microcalcification: NR NR 1.5-T scanner with use of a comparing two different No MRI: 64 Mean size (mm): NR dedicated prone eight-channel time periods breast coil Duygulu 2012 [41] Turkey One arm retrospective MRI: 18 Microcalcification: NR NR 1.5 T using a standard breast cohort Mean size (mm): NR coil in the prone position Hajaj 2017 [24] UK Retrospective cohort MRI: 70 Microcalcification: NR MRI: 63 (31–75) 1.5-T scanner and a dedicated from one hospital, No MRI: 52 Size range (mm): MRI: No MRI: (56–82) 8-channel breast coil comparing two 2 to 110; no MRI: 3 different time periods to 180 Hlubocky 2018 [38] USA One arm retrospective MRI: 288 Microcalcification: NR NR Initially with 1.5-T magnets; cohort in two sites Mean size: NR later, all were performed on 3.0 T Itakura 2011 [11] USA Retrospective cohort MRI: 38; Microcalcification: NR MRI: median: 50 (24–71) NR No MRI: 111 Size median (mm): MRI: No MRI: median: 59 (38–86) 16, no MRI: 10 Keymeulen 2019 [33] Netherlands Retrospective cohort MRI: 2382 Microcalcification: NR MRI: 50–74 years (74%) NR (population registries) No MRI: 8033 Mean size: NR No MRI: 50–74 years (88%) Kropcho 2012 [25] USA Retrospective cohort from MRI: 62; Microcalcification: NR MRI: 55 (35–78) 1.5-T magnet using a one site No MRI: 98 Mean size (mm): MRI: No MRI: 62 (38–93) dedicated four-channel in 20.9, no MRI: 27.8 vivo breast coil Lam 2019 [34] USA Retrospective cohort from MRI: 332 Microcalcification: NR All patients: 55.5 NR one hospital No MRI: 41 Mean size: NR Lamb 2020 [36] USA Restrospective cohort from MRI: 236 Microcalcification: all patients MRI: 50.6 ± 8.8 1.5 T or 3 T one hospital No MRI: 727 Mean size: NR No MRI: 60.2 ± 10 Lee 2016 [37], Korea One arm retrospective MRI: 199 NR All patients: 50.1 ± 9.4 1.5-T system with a dedicated 2020 [34] cohort in one site 4-channel breast coil Obdeijn 2013 [26] Netherlands Retrospective cohort, MRI: 11 Microcalcification: NR NR 1.5-T system with a dedicated comparing two different No MRI: 27 Mean size (mm): NR double breast coil Eur Radiol (2021) 31:5880–5893 Table 1 (continued) Author and year (trial) Country Design No. of patients Tumour characteristics Age (years), mean (range) Intervention (MRI) (only DCIS) (microcalcification, size) time periods (subgroup analysis) Onega 2017 [27] USA Retrospective cohort (Breast MRI: 354 Microcalcification: NR NR NR Cancer Surveillance No MRI: 2083 Mean size (mm): NR Consortium (BCSC)) Eur Radiol (2021) 31:5880–5893 Pettit 2009 [39] USA One-arm retrospective cohort MRI: 51 Microcalcification: NR NR Siemens 1.5-T Sonata or Espree (subgroup analysis) Mean size (mm): NR magnetic resonance imaging unit with dedicated breast coil Pilewskie 2013 [12] USA Prospective cohort from the MRI: 217; Microcalcification: MRI: MRI: 53 (26–82) NR Lynn Sage Comprehensive No MRI: 135 75.8 %, no MRI: 93.8 % No MRI: 60 (36–86) Breast Center Mean size cm (range): MRI: 2.1 (0.0, 10.0), no MRI: 1.7 (0.0, 9.0) Pilewskie 2014 [28] USA Retrospective cohort from MRI: 596; Microcalcification: NR MRI: 54.0 (26-73) NR the Memorial Sloan-Kettering No MRI: 1723 Mean size (mm): NR No MRI: 53.5 (25–85) Cancer Center (MSKCC) Shin 2012 [29] Korea Retrospective cohort MRI: 62; Microcalcification: NR NR A 1.5-T imager with dedicated (subgroup analysis) No MRI: 25 Mean size (mm): NR double-breast coil was used So 2018 [30] USA Retrospective cohort from MRI: 97; Microcalcification: NR MRI: 56.4 NR one site No MRI: 79 Mean size ± SD: MRI: No MRI: 63.6 1.5 ± 1.9, no MRI: 1.6 ± 2.6 Solin 2008 [42] USA Retrospective cohort from MRI: 31 Microcalcification: NR NR NR one site No MRI: 105 Mean size (mm): NR Vapiwala 2017 [31] USA Retrospective cohort MRI: 31; Microcalcification: NR Microcalcification: NR NR (subgroup analysis) No MRI: 104 Mean size (mm): NR Mean size (mm): NR Vos 2015 [32] Netherlands Retrospective cohort, MRI: 136 Microcalcification: NR NR Dynamic contrast-enhanced MRI population-based No MRI: 478 Mean size (mm): NR (subgroup analysis— high-grade DCIS) Yoon 2020 [35] Korea Retrospective cohort MRI: 106 Microcalcification: NR All patients: 53.5 ± 10 A 1.5-T or 3-T with dedicated from one hospital No MRI: 106 Mean size (cm): 3.0 ± 2.4 (post propensity matching) double-breast coil was used (post propensity matching) MRI magnetic resonance imaging, NR non-reported, MC microcalcifications, T tesla, DCIS ductal carcinoma in situ 5885 5886 Eur Radiol (2021) 31:5880–5893 BC who had undergone triple assessment, and were sched- The observational studies providing adjusted estimates had uled for a wide local excision. an inconsistent effect (OR 0.96; 95% CI 0.36 to 2.61) [30, 32, Of the 18 comparative cohorts included [11, 12, 22–34], 16 35] compared to studies with unadjusted estimates (OR 0.66; were retrospective analyses of medical records or population 95% CI 0.45 to 0.99). Therefore, only those providing adjust- registries [11, 12, 22–27, 29–36], three of them comparing dif- ed estimates were included in the evidence profile for obser- ferent time periods (pre- and post-implementation of MRI) [23, vational studies (low certainty). One study reported the mean 24, 26]. Three cohort studies had larger sample sizes, one used number of re-operations among those patients who received the Breast Cancer Surveillance Consortium registry data with lumpectomy and showed no relevant differences (0.42 vs 2437 patients with DCIS from 2010 to 2014 [27], another study 0.58, p value = 0.31) [11]. included data of 2319 patients from a cancer centre in New York (USA) from 1997 to 2010 [28] and the last study used the Netherlands Cancer Registry (NCR) including 10,415 clinical Total mastectomy records [33] (Table 1 and Supplementary table S3). We also included five single MRI-arm cohorts [37–41] and Two RCTs, including 429 patients [19, 21], showed that MRI the MRI arm of three comparative cohorts and one RCT [12, 21, may result in little to no difference in total mastectomy sur- 26], where all surgical treatment changes due to the MRI find- geries (RR 0.91; 95% CI 0.65 to 1.27), equivalent to 21 fewer ings were recorded. We did not find studies directly reporting the total mastectomies (95% CI 82 fewer to 63 more) performed quality of life of patients receiving pre-operative breast MRI. per 1000 patients assessed (very low certainty) (Table 2). Eight observational studies suggested that MRI may increase the odds of total mastectomies (OR 1.58; 95% CI 1.00 to 2.48) Initial BCS (very low certainty) (Fig. 4b) [12, 22, 23, 25, 32–35], with a larger but imprecise effect observed in studies providing ad- Two RCTs, including a total of 429 patients, showed that MRI justed results (aOR 1.74; 95% CI 0.53 to 5.68) [32, 35]. may result in little to no difference in initial BCS (RR 0.95; 95% CI 0.90 to 1.00) (Fig. 2a) (low certainty) [19, 21]. In 11 observa- tional studies [11, 12, 22–24, 27, 32–36], the odds of BCS was Disease-free survival (inferred from locoregional lower but the evidence was very uncertain (OR 0.49; 95% CI recurrence) 0.29 to 0.82) and only two cohort studies reported adjusted OR showing similar results (very low certainty) [32, 35] (Fig. 2b). One study reported a subgroup of 135 patients with DCIS, over a follow-up of 10 years, and found no differences in Proportion of positive margins (after undergoing BCS) disease-free survival between the use or not of breast MRI (4% vs 4%, p value = 0.25) [31]. Another study reported The results from 91 patients in one RCT suggested that MRI similar results at 8 years of follow-up for the rate of any local may increase the risk of positive margins in the excised lesion failure among 136 patients (6% vs 6%, p value = 0.58) [42]. (RR of 1.58; 95% CI 0.88 to 2.92) (very low certainty) [20]. Lamb et al reported similar rates of second BC events whether However, from observational studies, the pooled analysis in- MRI was used or not (12.8% versus 11.5%, p = 0.68) [36]. cluding 9126 patients suggested that MRI pre-operative as- The largest study, which included 2212 patients with a sessment may have little to no difference in this outcome median follow-up of 4.9 years, showed, in an adjusted multi- (OR 0.91; 95% CI 0.85 to 1.21) [11, 22, 25, 26, 28, 32, 33, variate analysis (i.e. age, margin status, number of excisions), 35] (very low certainty). It is noteworthy that the definitions that MRI may increase the risk of local recurrence (aHR = for positive margins across studies were heterogenous (i.e. < 1.18; 95% CI 0.79 to 1.76) (very low certainty) [28]. 1 mm [11, 27] or < 5 mm [29]) or not clearly described [36], even among the included centres of the only RCT reporting MRI triggered treatment change this outcome [20]. Eight cohorts [12, 26, 33, 35, 37–41] and the intervention arm of Re-operation rate (after undergoing BCS) one RCT [21] informed this outcome. Our pooled estimation showed that 17% (95% CI 12 to 24%) of the initial surgical Two RCTs suggested no difference in the risk of having a re- decisions may change to a more extensive resection or mastec- operation after an initial BCS (RR 1.02; 95% CI 0.56 to 1.84), tomy when breast MRI is used (low certainty) (Supplementary but the evidence was uncertain (very low certainty) [19, 21]. Figure 2). There was important heterogeneity, with the five stud- The pooled estimate from 12 observational studies suggested ies with a larger sample size [12, 21, 37–39] reporting between 9 a decrease in the risk of re-operation (OR 0.72; 95% CI 0.50 to and 18% and those with a smaller sample size reporting much 1.04) (Fig. 3b) [11, 12, 22–26, 29, 30, 32–36]. higher proportions (39% [41] to 50% [26]). Eur Radiol (2021) 31:5880–5893 5887 a b Fig. 2 Meta-analyses of initial breast-conserving surgery. a Randomised clinical trial; b cohort studies (prospective and retrospectives) Subgroup and sensitivity analysis treatment decisions might have been influenced by the knowl- edge of the allocation arm. The assessment of surgical out- One RCT and 11 observational studies reported the age of par- comes is less likely to be biased due to unblinded assessment. ticipants which was similar across studies. We were not able to However, the evaluation of positive margins in the surgical perform subgroup analysis on the extent of microcalcifications as excised specimens might be at higher risk of bias as described most studies did not provide detailed data, but we included a post elsewhere [43]. hoc analysis comparing those studies that reported the proportion In addition, two of the three RCTs might be subject to an of patients with microcalcification lesions with those that did not. imbalance of prognosis factors, as we included results for small As described above (see “Re-operation rate (after undergo- subgroups of the originally allocated patients [19, 20]. The prob- ing BCS)”), we found different estimates among observation- ability of important imbalance for a single prognostic factor is al studies reporting adjusted versus crude estimates for re- higher in RCTs with less than 100 participants [44, 45]. operations but with overlapping confidence intervals (test for For observational evidence, our main concern was risk of subgroup differences p value = 0.499). Other subgroup anal- bias, as only three studies reported adjusted estimates [28, 30, yses, including publication year, did not show meaningful 32]. Confounding variables (i.e. age, lesion size on mammog- results (Supplementary table S4). raphy) might be associated to requesting breast MRI and to performing more aggressive surgical treatments. Most obser- Risk of bias and certainty of the evidence vational studies were retrospective and potentially subject to loss of follow-up or misrecorded data. Three studies compared The included RCTs did not blind participants nor the clinical two different time periods which may lead to bias related to personnel leading to a potential risk of performance bias as 5888 Eur Radiol (2021) 31:5880–5893 a b Fig. 3 Meta-analyses of re-operation rate. a Randomised clinical trial; b cohort studies (prospective and retrospectives) different standards of care due to the progress in treatment finding that was consistent also with the results from observa- quality over time [23, 24, 26]. tional studies. For re-operations, there is also uncertainty; the For the assessment of the certainty of evidence, we rated IRCIS trial designed to include only DCIS patients suggested down for risk of bias for all outcomes, for inconsistency in most a reduction but with confidence intervals including the oppo- estimates informed by observational studies, and for imprecision site effect [21], while the MONET trial showed an increase in for locoregional recurrence, positive margins estimated from a small DCIS subgroup [19]. Among observational studies, RCTs, and for re-operation estimates from both RCTs and ob- Yoon et al, using propensity score matching, suggested a re- servational evidence (Table 2 and Supplementary table S5). duction in re-operations [35]. However, another two observa- tional studies providing adjusted estimates did not find a ben- efit with MRI [30, 32]. The included RCTs had several limitations [46]. The Discussion IRCIS trial recruited patients eligible for BCS after mam- mography, ultrasound and percutaneous biopsy which Main findings might have biased mastectomy effect estimates against MRI. This study might also be underpowered as there were Our review suggests that pre-operative breast MRI for DICS meaningful differences in the results between the intention lesions may have no meaningful impact on surgical outcomes to treat and per protocol analysis [21]. The MONET [19] or on the risk of local recurrence. RCTs showed that MRI had and COMICE [20] trials randomised women with non- little to no effect on initial BCS or total mastectomies, a Eur Radiol (2021) 31:5880–5893 5889 Table 2 Summary of findings a Risk of bias. The intervention (preoperative MRI) was not feasible to be blinded which led to high risk of performance bias, potentially influencing surgeons’ treatment plans b Risk of bias. There was also a potential risk of imbalance of prognostic factors, due to the inclusion of results from a very small subgroup of participants in some RCTs c Indirectness. Initial BCS was considered an intermediate outcome, as women could have received re-excision or a mastectomy depending on the presence of positive margins in the excised specimen d Risk of bias. In some cohort studies, the comparison was between arms over different periods of time (secular bias) e Risk of bias. Most observational studies reported unadjusted estimates f Inconsistency. Potentially important and unexplained heterogeneity across included studies g Other considerations. Although there is an observed large effect, there is a very serious risk of bias concern and the estimate is imprecise; thus, we did not upgrade the certainty of evidence h Risk of bias. The definition of positive margins was variable across clinical centres potentially introducing misclassification bias i Imprecision. The anticipated absolute effects associated to the intervention go from potential benefit to potential harm j Imprecision. There is imprecision of the anticipated absolute effects with the intervention but it is likely due to heterogeneity across studies k Only estimates from studies reporting adjusted ORs are included as the results were different from those unadjusted l Risk of bias. Both studies reported adjusted estimates, although one study did not include tumour size as a pre-defined confounding variable in the analysis. Additionally, there was no information about the time the MRI exam was requested m Risk of bias. Only patients who received breast-conserving surgery were included. There was potential selection bias as those with more aggressive treatments after MRI were not included. Potential over adjustment in the multivariate analysis as positive margins and number of excisions may be in the casual pathway to disease recurrence n Indirectness. A proportion of patients had breast MRI performed after lumpectomy or at re-excision stage o Indirectness. Serious indirectness as locoregional recurrence is considered a surrogate of disease-free survival p Imprecision. The absolute effect of the intervention ranged from significant benefit to significant harm q Risk of bias. Decision to request breast MRI (after mammography and ultrasound) might be associated to the decision to change the initial plan, independent of MRI results palpable or BIRADS 3–5 lesions respectively, while wom- Our results in the context of previous research en with DCIS were only a small fraction of them. Moreover, these RCTs reported difficulties in the acquisi- A systematic review including a lower number of studies than tion of MIR images at 3 T before randomisation [19], our review showed that preoperative breast MRI in patients underwent MRI prior to biopsy which is not considered with DCIS was not associated with an improvement in surgi- standard of care [19] and guided biopsies to verify MRI cal outcomes [47]. Similar results were also observed in pa- findings were not available in all cases [20]. tients with invasive cancer [48–51], with a systematic review 5890 Eur Radiol (2021) 31:5880–5893 a b Fig. 4 Meta-analyses of total mastectomy. a Randomised clinical trial; b cohort studies (prospective and retrospectives) showing that MRI increased mastectomy rates [50] and one suppression [55]. Older studies might be prone to individual person-data meta-analysis reporting that MRI does unstandardised procedures as accreditation was not uniformly not reduce the risk of local recurrence or distant recurrence in implemented. Currently, most facilities perform 1.5 T MRI, these patients [51]. but the use of 3-T magnetic field has increased [56]. As some One potential explanation for the lack of benefit on surgical studies suggest, despite some technical limitations, 3 T could outcomes might be a limited specificity of MRI in patients provide higher correlation with DCIS pathology size com- with DCIS, and variable positive predictive value from 25 to pared to 1.5 T, therefore obtaining higher image quality scores 84% [52]. Previous data has shown that MRI, compared to and better differential diagnosis of enhancing lesions [57]. histopathology, tends to overestimate the size of pure DCIS Another study, including 20 DCIS lesions, found that size lesions and has moderate correlation with pathologically mea- correlation between MRI images and pathology was higher sured tumour size (r = 0.74) [53, 54]. This limitation could with 3 T [58]. Most studies in our review used 1.5 T; thus, impact on having more aggressive treatments than needed, but further studies are needed to assess the clinical impact of 3 T. not on the number of re-excisions. Furthermore, it should have To provide an optimal accuracy, breast MRI should ideally only hampered older studies where biopsies were not per- provide high spatial and temporal resolution [59]; however, formed in all new lesions identified by MRI and patients fre- conventional MRI methods cannot deliver both and usually quently went directly to reassessment of the surgical treatment prioritise spatial resolution [59]. To solve this issue, recent plan. technical advancements have focused on accelerating data ac- An important factor is to ensure MRI images were ade- quisition. Morrison et al described a method that provided six- quately acquired. The ACR Breast MRI Accreditation times faster effective temporal resolution and the same high Program began accrediting facilities in 2010 [55], requiring spatial resolution of standard MRI [60]. Goto et al improved adequate magnetic field strength (1 T or higher) and gradient, the temporal resolution with preservation of spatial resolution bilateral breast coil enabling prone positioning and good fat using ultra-fast DCE-MRI to differentiate malignant and Eur Radiol (2021) 31:5880–5893 5891 benign lesions [61]. As new techniques are developed, the procurement of services between the European Commission’s Joint Research Centre and Asociación Colaboración Cochrane Iberoamericana. performance of MRI in the preoperative setting could improve. Declarations Guarantor The scientific guarantor of this publication is Carlos Canelo- Limitations and strengths Aybar. Conflict of interest Elena Parmelli and Zuleika Saz-Parkinson are We included only English language articles although we in- current employees of the Joint Research Centre, European cluded a larger number of studies than previous reviews. We Commission. Carlos Canelo-Aybar, Pablo Alonso-Coello, and could not explore the effect of MRI in some relevant patient David Rigau are employees of the Iberoamerican Cochrane Collaboration. Axel Gräwingholt, Annette Lebeau, Elsa Pérez subgroups as this data was not available. Our results are ham- Gómez and Miranda Langendam are members of the ECIBC pered by the low to very low certainty of the evidence found Guidelines Development Group. Paolo Giorgi Rossi, as principal for the included outcomes. investigator of an independent study, funded by the Italian Our review has several strengths. We included outcomes Ministry of Health, conducted negotiations with Hologic to obtain that were of interest for women, clinicians and policy makers reagents at reduced price or for free; the reagents have not been used by Paolo Giorgi Rossi’s institutions; he is also member of the when facing the decision of implementing or recommending ECIBC Guidelines Development Group. Jessica Hanae Zafra- preoperative MRI, and used rigorous methods including the Tanaka and Alvaro Taype-Rondan were employees of the GRADE approach to rate the certainty of the evidence. Institute for Health Technology Assessment and Research (IETSI) from the Peruvian Social Security at the time the study took place. Implications for practice and research Statistics and biometry No complex statistical methods were necessary for this paper. Patients and clinicians should be aware that although breast Informed consent Written informed consent was not required for this MRI for pre-operative assessment may improve the morpho- study because it is a systematic review. logical description of DCIS lesions, we did not find evidence to suggest an improvement of surgical outcomes. In fact, ob- Ethical approval Institutional Review Board approval was not required servational evidence, although of very low certainty, may sug- because it is a systematic review. gest that preoperative MRI could lead to more aggressive Study subjects or cohorts overlap The studies included in this system- treatments. Also, MRI-guided biopsy to confirm new lesions atic review were all already published elsewhere. Their references and or important expansions of the detected lesions is not always characteristics are reported within the manuscript. feasible or accessible. From a health system perspective, pre- operative MRI implies greater resource use [62, 63], and given Methodology the uncertain potential benefits, the use of MRI has probably • retrospective (systematic review of already published studies) • performed at one institution limited interest [64]. Given the uncertain evidence, there is a need for Open Access This article is licensed under a Creative Commons conducting well-powered RCTs assessing the role of preoper- Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as ative MRI during treatment planning in patients with DCIS long as you give appropriate credit to the original author(s) and the lesions incorporating new advancements in MRI imaging ac- source, provide a link to the Creative Commons licence, and indicate if quisition, securing the availability of experienced imaging changes were made. The images or other third party material in this article readers and biopsy confirmation of new lesions. are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not Supplementary Information The online version contains supplementary permitted by statutory regulation or exceeds the permitted use, you will material available at https://doi.org/10.1007/s00330-021-07873-2. need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Acknowledgements The authors would like to sincerely thank all members of the Guidelines Development Group of the European Commission Initiative on Breast Cancer for their participation in the discussions generated by this systematic review which led to the different recommendations they developed in the European Guidelines on Breast Cancer Screening and diagnosis (https://healthcare-quality.jrc. References ec.europa.eu/european-breast-cancer-guidelines). 1. Ferlay J, Ervik M, Lam F et al (2018) Global Cancer Observatory: Funding The systematic review was carried out by Iberoamerican cancer today. Lyon, France: International Agency for Research on Cochrane Collaboration under the Framework contract 443094 for 5892 Eur Radiol (2021) 31:5880–5893 Cancer. Available from: https://gco.iarc.fr/today. Accessed 25 07 21. Balleyguier C, Dunant A, Ceugnart L et al (2019) Preoperative 2019 breast magnetic resonance imaging in women with local ductal 2. Schmale I, Liu S, Rayhanabad J, Russell CA, Sener SF (2012) carcinoma in situ to optimize surgical outcomes: results from the Ductal carcinoma in situ (DCIS) of the breast: perspectives on bi- randomized phase III trial IRCIS. J Clin Oncol:18.00595 ology and controversies in current management. J Surg Oncol 105: 22. Allen LR, Lago-Toro CE, Hughes JH et al (2010) Is there a role for 212–220 MRI in the preoperative assessment of patients with DCIS? Ann 3. Kerlikowske K (2010) Epidemiology of ductal carcinoma in situ. J Surg Oncol 17:2395–2400 Natl Cancer Inst Monogr 2010:139–141 23. Davis KL, Barth RJ, Gui J, Dann E, Eisenberg B, Rosenkranz K 4. Barrio AV, Van Zee KJ (2017) Controversies in the treatment of (2012) Use of MRI in preoperative planning for women with newly ductal carcinoma in situ. Annu Rev Med 68:197–211 diagnosed DCIS: risk or benefit? Ann Surg Oncol 19:3270–3274 5. Jansen CHP, Perera D, Wiethoff AJ et al (2017) Contrast-enhanced 24. Hajaj M, Karim A, Pascaline S, Noor L, Patel S, Dakka M (2017) magnetic resonance imaging for the detection of ruptured coronary Impact of MRI on high grade ductal carcinoma insitu (HG DCIS) plaques in patients with acute myocardial infarction. PLoS One 12: management, are we using the full scope of MRI? Eur J Radiol 95: e0188292 271–277 6. Mann RM, Kuhl CK, Moy L (2019) Contrast-enhanced MRI for 25. Kropcho LC, Steen ST, Chung AP, Sim MS, Kirsch DL, Giuliano breast cancer screening. J Magn Reson Imaging 50:377–390 AE (2012) Preoperative breast MRI in the surgical treatment of 7. Houssami N, Ciatto S, Macaskill P et al (2008) Accuracy and sur- ductal carcinoma in situ. Breast J 18:151–156 gical impact of magnetic resonance imaging in breast cancer stag- 26. Obdeijn I-M, Tilanus-Linthorst MM, Spronk S et al (2013) ing: systematic review and meta-analysis in detection of multifocal Preoperative breast MRI can reduce the rate of tumor-positive re- and multicentric cancer. J Clin Oncol 26:3248–3258 section margins and reoperations in patients undergoing breast- 8. Tajima CC, de Sousa LLC, Venys GL, Guatelli CS, Bitencourt conserving surgery. AJR Am J Roentgenol 200:304–310 AGV, Marques EF (2019) Magnetic resonance imaging of the 27. Onega T, Weiss JE, Goodrich ME et al (2017) Relationship be- breast: role in the evaluation of ductal carcinoma in situ. Radiol tween preoperative breast MRI and surgical treatment of non- Bras 52(1):43–47 metastatic breast cancer. J Surg Oncol 116:1008–1015 9. Menezes GL, Knuttel FM, Stehouwer BL, Pijnappel RM, van den 28. Pilewskie M, Olcese C, Eaton A et al (2014) Perioperative breast Bosch MA (2014) Magnetic resonance imaging in breast cancer: a MRI is not associated with lower locoregional recurrence rates in literature review and future perspectives. World J Clin Oncol 5:61 DCIS patients treated with or without radiation. Ann Surg Oncol 21:1552–1560 10. Albert ML, Gao Y, Moy L (2016) Role of breast MRI in patients 29. Shin H-C, Han W, Moon H-G et al (2012) Limited value and utility with newly diagnosed breast cancer. Curr Breast Cancer Rep 8:80– of breast MRI in patients undergoing breast-conserving cancer sur- 89 gery. Ann Surg Oncol 19:2572–2579 11. Itakura K, Lessing J, Sakata T et al (2011) The impact of preoper- 30. So A, De La Cruz LM, Williams AD et al (2018) The impact of ative magnetic resonance imaging on surgical treatment and out- preoperative magnetic resonance imaging and lumpectomy cavity comes for ductal carcinoma in situ. Clin Breast Cancer 11:33–38 shavings on re-excision rate in pure ductal carcinoma in situ—a 12. Pilewskie M, Kennedy C, Shappell C et al (2013) Effect of MRI on single institution’s experience. J Surg Oncol 117:558–566 the management of ductal carcinoma in situ of the breast. Ann Surg 31. Vapiwala N, Hwang WT, Kushner CJ, Schnall MD, Freedman Oncol 20:1522–1529 GM, Solin LJ (2017) No impact of breast magnetic resonance im- 13. Schunemann HJ, Lerda D, Dimitrova N et al (2019) Methods for aging on 15-year outcomes in patients with ductal carcinoma in situ development of the European Commission Initiative on Breast or early-stage invasive breast cancer managed with breast conser- Cancer Guidelines: recommendations in the era of guideline trans- vation therapy. Cancer 123:1324–1332 parency. Ann Intern Med. https://doi.org/10.7326/M18-3445 32. Vos E, Voogd AC, Verhoef C, Siesling S, Obdeijn I-M, Koppert 14. Higgins JP, Altman DG, Gøtzsche PC et al (2011) The Cochrane LB (2015) Benefits of preoperative MRI in breast cancer surgery Collaboration’s tool for assessing risk of bias in randomised trials. studied in a large population-based cancer registry. Br J Surg 102: BMJ 343:d5928 1649–1657 15. Sterne JA, Hernán MA, Reeves BC et al (2016) ROBINS-I: a tool 33. Keymeulen K, Geurts SME, Lobbes MBI et al (2019) Population- for assessing risk of bias in non-randomised studies of interven- based study of the effect of preoperative breast MRI on the surgical tions. BMJ 355:i4919 management of ductal carcinoma in situ. Br J Surg 106:1488–1494 16. Nyaga VN, Arbyn M, Aerts M (2014) Metaprop: a Stata command 34. Lam DL, Smith J, Partridge SC et al (2020) The impact of preop- to perform meta-analysis of binomial data. Arch Public Health 72: erative breast MRI on surgical management of women with newly 39 diagnosed ductal carcinoma in situ. Acad Radiol 27:478–486 17. Balshem H, Helfand M, Schünemann HJ et al (2011) GRADE 35. Yoon GY, Choi WJ, Kim HH, Cha JH, Shin HJ, Chae EY (2020) guidelines: 3. Rating the quality of evidence. J Clin Epidemiol 64: Surgical outcomes for ductal carcinoma in situ: impact of preoper- 401–406 ative MRI. Radiology 295:296–303 18. Schunemann HJ, Cuello C, Akl EA et al (2019) GRADE guide- 36. Lamb LR, Oseni TO, Lehman CD, Bahl M (2020) Pre-operative lines: 18. How ROBINS-I and other tools to assess risk of bias in MRI in patients with ductal carcinoma in situ: is MRI useful for nonrandomized studies should be used to rate the certainty of a identifying additional disease? Eur J Radiol 129:109130 body of evidence. J Clin Epidemiol 111:105–114 37. Lee J, Jung JH, Kim WW et al (2016) The role of preoperative 19. Peters N, Van Esser S, Van Den Bosch M et al (2011) Preoperative breast magnetic resonance (MR) imaging for surgical decision in MRI and surgical management in patients with nonpalpable breast patients with triple-negative breast cancer. J Surg Oncol 113:12–16 cancer: the MONET–randomised controlled trial. Eur J Cancer 47: 38. Hlubocky J, Bhavnagri S, Swinford A, Mitri C, Rebner M, Pai V 879–886 (2018) Does the use of pretreatment MRI change the management 20. Turnbull LW, Brown SR, Olivier C et al (2010) Multicentre of patients with newly diagnosed breast cancer? Breast J 24:309– randomised controlled trial examining the cost-effectiveness of 313 contrast-enhanced high field magnetic resonance imaging in wom- 39. Pettit K, Swatske ME, Gao F et al (2009) The impact of breast MRI en with primary breast cancer scheduled for wide local excision on surgical decision-making: are patients at risk for mastectomy? J (COMICE). Health Technol Assess 14:1–182 Surg Oncol 100:553–558 Eur Radiol (2021) 31:5880–5893 5893 40. Besharat Sara KM, Azadeh J, Nasser M (2018) Influence of preop- 53. Baur A, Bahrs SD, Speck S et al (2013) Breast MRI of pure ductal erative magnetic resonance imaging in surgical planning for breast carcinoma in situ: sensitivity of diagnosis and influence of lesion cancer. Int J Cancer Manag 11:e81459 characteristics. Eur J Radiol 82:1731–1737 41. Duygulu G, Oktay A, Bilgen IG, Kapkaç M, Zekioglu O (2012) 54. Schouten van der Velden AP, Schlooz-Vries MS, Boetes C, The role of breast MRI in planning the surgical treatment of breast Wobbes T (2009) Magnetic resonance imaging of ductal carcinoma cancer. Diagn Interv Radiol 18:460 in situ: what is its clinical application? A review. Am J Surg 198: 42. Solin LJ, Orel SG, Hwang WT, Harris EE, Schnall MD (2008) 262–269 Relationship of breast magnetic resonance imaging to outcome after 55. Hendrick RE (2014) High-quality breast MRI. Radiol Clin North breast-conservation treatment with radiation for women with early- Am 52:547–562 stage invasive breast carcinoma or ductal carcinoma in situ. J Clin 56. Soher BJ, Dale BM, Merkle EM (2007) A review of MR physics: 3 Oncol 26:386–391 T versus 1.5 T. Magn Reson Imaging Clin N Am 15:277–290 43. Skov BG, Braendstrup O, Hirsch FR, Lauritzen AF, Nielsen HW, 57. Kuhl CK, Jost P, Morakkabati N, Zivanovic O, Schild HH, Gieseke Skov T (1994) Are pathologists biased by clinical information?: a J (2006) Contrast-enhanced MR imaging of the breast at 3.0 and 1.5 blinded cross-over study of the histopathological diagnosis of me- T in the same patients: initial experience. Radiology 239:666–676 sothelial tumours versus pulmonary adenocarcinoma. Lung Cancer 58. Rahbar H, DeMartini WB, Lee AY, Partridge SC, Peacock S, 11:365–372 Lehman CD (2015) Accuracy of 3 T versus 1.5 T breast MRI for 44. Cui L, Hung HM, Wang SJ, Tsong Y (2002) Issues related to pre-operative assessment of extent of disease in newly diagnosed subgroup analysis in clinical trials. J Biopharm Stat 12:347–358 DCIS. Eur J Radiol 84:611–616 45. Chu R, Walter SD, Guyatt G et al (2012) Assessment and implica- tion of prognostic imbalance in randomized controlled trials with a 59. Kuhl C (2007) The current status of breast MR imaging. Part I. binary outcome–a simulation study. PLoS One 7:e36677 Choice of technique, image interpretation, diagnostic accuracy, 46. Pinker K (2020) Preoperative MRI improves surgical planning and and transfer to clinical practice. Radiology 244:356–378 outcomes for ductal carcinoma in situ. Radiology 295:304–306 60. Morrison CK, Henze Bancroft LC, DeMartini WB et al (2017) 47. Fancellu A, Turner RM, Dixon JM, Pinna A, Cottu P, Houssami N Novel high spatiotemporal resolution versus standard-of-care dy- (2015) Meta-analysis of the effect of preoperative breast MRI on namic contrast-enhanced breast MRI: Comparison of Image the surgical management of ductal carcinoma in situ. Br J Surg 102: Quality. Invest Radiol 52:198–205 883–893 61. Goto M, Sakai K, Yokota H et al (2019) Diagnostic performance of 48. Parsyan A, Alqahtani A, Mesurolle B, Meterissian S (2013) Impact initial enhancement analysis using ultra-fast dynamic contrast- of preoperative breast MRI on surgical decision making and clinical enhanced MRI for breast lesions. Eur Radiol 29:1164–1174 outcomes: a systematic review. World J Surg 37:2134–2139 62. Pataky RE, Baliski CR (2016) Reoperation costs in attempted 49. Houssami N, Turner R, Morrow M (2013) Preoperative magnetic breast-conserving surgery: a decision analysis. Curr Oncol 23: resonance imaging in breast cancer: meta-analysis of surgical out- 314–321 comes. Ann Surg 257:249–255 63. Onega T, Tosteson AN, Weiss J et al (2016) Costs of diagnostic and 50. Houssami N, Turner RM, Morrow M (2017) Meta-analysis of pre- preoperative workup with and without breast MRI in older women operative magnetic resonance imaging (MRI) and surgical treat- with a breast cancer diagnosis. BMC Health Serv Res 16:76 ment for breast cancer. Breast Cancer Res Treat 165:273–283 64. Kandel M, Dunant A, Balleyguier C, Bonastre J (2020) Cost- 51. Houssami N, Turner R, Macaskill P et al (2014) An individual effectiveness of preoperative magnetic resonance imaging to opti- person data meta-analysis of preoperative magnetic resonance im- mize surgery in ductal carcinoma in situ of the breast. Eur J Radiol aging and breast cancer recurrence. J Clin Oncol 32:392–401 129:109058 52. Brem RF, Fishman M, Rapelyea JA (2007) Detection of ductal carcinoma in situ with mammography, breast specific gamma im- Publisher’s note Springer Nature remains neutral with regard to jurisdic- aging, and magnetic resonance imaging: a comparative study. Acad tional claims in published maps and institutional affiliations. Radiol 14:945–950