Running Title: MRI Versus Conventional Preoperative Assessment in DCIS of the Breast

SYSTEMATIC REVIEW AND META-ANALYSIS OF THE EFFECT OF PREOPERATIVE BREAST MRI ON THE SURGICAL MANAGEMENT OF DUCTAL CARCINOMA IN SITU.

Running title: MRI versus conventional preoperative assessment in DCIS of the breast Authors: Alessandro Fancellu1, Robin M. Turner2, J. Michael Dixon3, Antonio Pinna1, Pietrina Cottu1, Nehmat Houssami4 1 Dept. of Clinical and Experimental Medicine, Unit of General Surgery 2, Clinica Chirurgica. University of Sassari, Sassari, Italy. 2School of Public Health and Community Medicine, The University of New South Wales, New South Wales, Australia. 3Breakthrough Breast Cancer Research Unit, Institute of Genetics and Molecular Medicine, University of Edinburgh, UK. 4 Screening and Test Evaluation Program, School of Public Health, Sydney Medical School, University of Sydney, Sydney, Australia.

Corresponding Author: Dr. Alessandro Fancellu, MD, PhD, Dept. of Clinical and Experimental Medicine, Unit of General Surgery 2, Clinica Chirurgica. University of Sassari. V.le San Pietro, 43, 07100 Sassari, Italy Tel: +39 079 228432 Fax: +39 079 223494 e-mail: [email protected]

Category: Systematic review

Funding: N. Houssami receives research support via a National Breast Cancer Foundation (NBCF Australia) Practitioner Fellowship. This paper is not based on a previous communication to a society or meeting.

1 ABSTRACT

Background: Magnetic resonance imaging (MRI) has been used increasingly in the diagnosis and management of women with breast cancer. However, its usefulness in the preoperative assessment of DCIS remains questionable. A meta-analysis was conducted to examine the effects of MRI on surgical treatment of DCIS by analyzing studies comparing preoperative MRI and conventional preoperative assessment.

Methods: Using random-effects modeling, the proportion of women with various outcomes in the

MRI versus no-MRI groups was estimated, and the odds ratio (OR) and adjusted OR (adjusted for study-level median age) for each model were calculated.

Results: Nine eligible studies were identified that included 1 077 women with DCIS who had preoperative MRI and 2 175 who did not. MRI significantly increased the odds of having initial mastectomy [OR, 1.72 (P = 0.012); adjusted OR, 1.76, P = 0.010)]. There were no significant differences in the odds of having positive margins after breast conserving surgery (BCS) between the MRI and no-MRI groups [OR , 0.80 (P = 0.059); adjusted OR, 1.10 (P = 0.716)], nor in the odds of having reoperation for positive margins after BCS [OR 1.06 (P = 0.759); adjusted OR 1.04

(P = 0.844)]. Overall mastectomy rates did not significantly differ according to whether or not MRI was performed [OR 1.23, (P = 0.340); adjusted OR 0.97, (P = 0.881)].

Conclusions: Preoperative MRI in patients with DCIS is not associated with improvement in surgical outcomes. Further studies may identify subsets of DCIS patients who might possibly benefit from preoperative MRI.

INTRODUCTION Ductal carcinoma in situ (DCIS) is the most common form of noninvasive cancer in the breast, and constitutes a spectrum of proliferating malignant cells that are confined within the basement membrane of the ductal epithelium1-3. As a consequence of widespread mammographic screening, DCIS today accounts for 20-25 per cent of newly diagnosed breast cancers, with more than 63 000 cases diagnosed annually in the United States4-7. In patients with DCIS, surgical planning traditionally follows “triple assessment” with clinical examination, mammography, and a tissue diagnosis. However, in recent years, magnetic resonance imaging (MRI) has been used preoperatively in patients with DCIS with the aim of better defining surgical treatment selection and reducing the need for additional surgery after

2 breast conserving surgery (BCS)8-10. In fact, the use of additional MRI imaging after standard mammograpghy may theoretically help in better defining the extent of DCIS 8,11,12.

Whether the use of preoperative MRI in patients with DCIS translates into oncological advantages in terms of local control and survival is far from clear 12-17. In this regard, two recent studies reported no advantages for local control of DCIS in patients who underwent preoperative MRI and BCS12,18. Furthermore, preoperative MRI has other disadvantages in that it increases the numbers of unnecessary surgical biopsies9,10,13,19. Thus, the effectiveness of preoperative MRI in patients with DCIS remains uncertain as recent studies have suggested that MRI does not lead to improved rates of complete excision and may increase rates of mastectomy9,16,20,21. Most studies of MRI in patients with DCIS have included relatively small numbers of patients and individually studies to date may have limited statistical power. To examine the effect of preoperative MRI in the management of biopsy-proven DCIS, a systematic review and meta-analysis were conducted of studies comparing patients who had preoperative MRI (MRI group) with those who had triple assessment only (no-MRI group). The aim was to determine the effect of preoperative MRI on surgical outcomes, specifically the rates of reoperation for positive margins and the rate of mastectomy. A secondary endpoint was to estimate the proportion of patients who had an MRI-triggered change of treatment.

METHODS Study Selection A systematic literature search was performed using the PubMed and Cochrane databases, for studies reporting surgical outcomes of preoperative MRI in patients diagnosed with DCIS, using the following keywords: “ductal carcinoma in situ”, “DCIS”, “intraductal”, “MRI”, “Magnetic Resonance Imaging” and “breast”. The “related articles” function was used to broaden the search and all abstracts, studies, and citations scanned were reviewed as were the references of relevant articles. The latest date for this search was March 31, 2014. No language restrictions were made.

Inclusion Criteria For the purposes of this systematic review and meta-analysis, studies reporting on patients with DCIS on preoperative histology were considered. To be included in this meta-analysis, studies had to: compare outcomes of patients who had preoperative MRI (MRI group) with patients who did not have an MRI (no-MRI group) who underwent surgery for DCIS using a controlled study design (either randomized controlled trial or comparative cohort design); and to report on at least one of the defined surgical outcomes (described below).

Exclusion Criteria Studies were excluded according to the following criteria: studies that only compared MRI with no-MRI in patients with invasive breast carcinomas; studies that compared MRI with no-MRI in cohorts that included both invasive breast carcinoma and DCIS where it was not possible to extract data separately for DCIS patients; studies in which the outcomes of interest (specified below) were not reported or were not clearly reported, or if data could not be obtained or calculated separately for each of the MRI and no-MRI groups.

Data Extraction & Quality Assessment One author (AF) reviewed abstracts from the literature search, and identified potentially relevant articles. Two authors (AF, NH) extracted the following data: publication details, study population and time-frame, number of subjects who had surgical treatment with and without preoperative MRI, total number of DCIS cases, preoperative histology, tumour pathologic size, initial surgical intervention (BCS or mastectomy), and final surgical outcomes. To assess the quality of studies, 3 the following information was also extracted: study design (whether randomisation was used; and whether prospective or retrospective); whether consecutive subjects were included; and if a non- randomised design was used whether the comparison group was appropriate.

Data extraction was consistent between the two reviewers. Where data on surgical outcomes were reported but clarification was needed, study authors were contacted to confirm the data or to obtain data clarification; this was required for 2 studies (Itakura et al, Shin et al)9,20. To avoid overlapping patient groups, where studies reported on the same patients in more than one publication, the most informative and/or most recent article was included.

Outcomes of Interest Surgical outcomes were defined as follows: initial surgery (number of patients who had mastectomy, number who had BCS); number of patients with positive margins (or incomplete excision due to non-negative/positive margins) after breast conservation as initial surgery; number of patients who had reoperation after initial BCS (re-excision or conversion to mastectomy); overall number who received mastectomy (mastectomy as initial surgery plus mastectomy after initial BCS); MRI-triggered change of treatment (patients in which initial surgical planning was reported to have been modified because of MRI findings). This included both change from breast conservation to mastectomy and change from unilateral to bilateral surgery.

Statistical Analysis Descriptive statistics (median and inter-quartile range [IQR]), were used to summarise across studies study-level age, DCIS size, and time-frame. Random effects logistic meta-regression was used to model the proportion of subjects with the surgical outcomes in separate models defined under ‘methods’, with odds ratios (OR) estimated for the groups that had MRI versus those that did not have MRI (referent group: no-MRI). Random study effects were included in all models to allow for anticipated heterogeneity between studies beyond what would arise from within study sampling error alone; taking account of both within- and between-study variability provides valid standard errors, confidence intervals, and P-values. Study-specific and pooled proportions (from the meta-regressions), and 95 per cent confidence intervals (CI), were estimated (and displayed in forest plots). Statistical significance was set at P < 0.05. Because an earlier meta-analysis of preoperative MRI22 showed correlation between receipt of MRI and median age, and based on the summary descriptive data for the median study-level age between the groups being compared, a simple adjustment for age was performed. This was done by fitting age as a binary variable, whereby the study-level median age (or mean age if median was not reported) was categorized as younger than the median value of the study-level median ages or older than that median value, to obtain an adjusted OR for each model 22. Because a limited number of studies reported age an ‘unknown’ category was also included so that all studies could be retained in the analysis when adjusted for study-level age information. It should be noted that this represented a minimal study-level adjustment, and that more complex adjustments (or adjustments for more variables) were avoided due to the modest number of studies providing data on covariates. A sensitivity analysis, that excluded one study using a historical comparison group (identified through the quality assessment of study methods), was performed to assess whether excluding that study altered estimates of surgical outcomes. All analyses were conducted in Stata version 13.1 (StataCorp 2009; College Station, TX).

RESULTS The flow diagram summarizing the search and study identification for this systematic review is presented in Fig. 1. Nine studies were eligible for this meta-analysis 9,12,20,23-29: 7 of these were studies that compared cohorts defined by whether or not MRI was performed 9,12,20,25,27,28,29, and 2

4 were randomized controlled trials of preoperative MRI 23,26. Four studies included patients with both DCIS and invasive breast cancer20,23,26,29, whereas 5 studies included patients with DCIS only9,12,25,27,28. Four studies included patients undergoing BCS as primary surgery 12,20,23,27, whereas 5 studies evaluated preoperative MRI in patients scheduled for either mastectomy or breast conservation as initial surgery9,25,26,28,29. The histological diagnosis of DCIS was obtained preoperatively in most cases by using core-biopsy. Overall, 3 252 subjects with DCIS were analyzed; of those, 1 077 had preoperative MRI and 2 175 did not. Characteristics of each study including quality assessment are summarized in Table 1; heterogeneity between studies for various patient and tumour characteristics was present. In all the included studies, patients in the MRI and no-MRI groups were from the same time-frame, except for one study29 that used a historical comparison group. The study-level median age for the patients in the MRI and no-MRI groups was 54.2 years (IQR 51.7 to 57.7) and 60.7 years (IQR 59.2 to 63.0), respectively. The study-level median DCIS size for the MRI and no-MRI groups was 2.6 cm (IQR 1.8 to 3.2) and 2.2 cm (IQR 1.6 to 2.5), respectively. The median study time- frame (using mid-point of timeframe) was 2005 (IQR 2 003.75 to 2 007.5). Four studies reported data for MRI-triggered changes of treatment 9,25,28,29: the proportion of patients who had changes to initial surgical treatment (mostly conversion from a planned initial BCS to initial mastectomy) based on preoperative MRI findings was estimated as 15.7 per cent (95 per cent CI 6.1-35.0).

Modeled Estimates The results of the models are summarized in Table 2. The number of subjects and studies contributing to each model are shown: subject numbers vary according to the number of studies reporting the specific surgical outcome and according to whether the analysis applied to all subjects or only those who received initial BCS. In Table 2 are also shown the estimated pooled proportions with each surgical outcome for the MRI vs no-MRI groups, as well as the OR estimated from each model. In the present meta-analysis, there was evidence that preoperative MRI increased the odds of having mastectomy as initial surgery [OR, 1.72 (P = 0.012)]; this finding did not change after adjustment for study-level age [adjusted OR, 1.76, P = 0.010)]. Conversely, the odds of receiving breast conservation were much higher for women not having MRI [OR, 0.55 (P = 0.006); adjusted OR, 0.53 (P = 0.004)]. There were no statistical differences in the odds of having positive margins or incomplete excision after BCS as initial surgery between the MRI and no- MRI groups [OR , 0.80 (P = 0.059); OR from sensitivity analysis, 0.82 (P= 0.082); adjusted OR, 1.10 (P = 0.716)]. There were no statistical differences in the odds of having reoperation for positive margins after BCS [OR 1.06 (P = 0.759); OR from sensitivity analysis, 1.16 (P = 0.475); adjusted OR 1.04 (P = 0.844)]. The overall mastectomy rate (initial mastectomy plus mastectomy for positive margins after BCS), however did not significantly differ according to whether or not MRI was performed [OR 1.23, (P = 0.340); adjusted OR 0.97, (P = 0.881)]. Fig. 2 and Fig. 3 show study-specific and estimated pooled proportions for each surgical outcome, the OR and the age-adjusted OR.

DISCUSSION The use of MRI in breast practice has progressively increased over the last decade. In particular,

MRI has been used increasingly in the preoperative assessment of patients with both invasive and non-invasive breast cancer. However, its value remains controversial given that it has not been shown to be useful in improving surgical outcomes15,16,22,30 or longer term outcomes17. A recent

5 meta-analysis from Houssami et al, indicated an unfavorable harm-benefit ratio for surgical outcomes from the routine use of preoperative MRI with evidence that it increased mastectomy rates; however, meta-analytic studies have focused mainly on invasive breast cancer 22. The application and potential value of MRI in patients with pure DCIS has been less well studied with much fewer published reports. Hence, the aim of this work was to evaluate preoperative MRI for patients with biopsy-proven DCIS through a systematic review and meta-analysis of studies that have compared surgical outcomes for patients who did and did not have preoperative MRI. DCIS differs in many respects from invasive cancer, in biological behavior, mammographic appearance, surgical perspectives, adjuvant treatments, and even MRI imaging patterns 1,11,13,31,32. Some studies have suggested in patients with biopsy-proven DCIS that MRI provides potentially a more accurate assessment than mammography of the presence of multicentricity and may provide better evaluation of disease extent31,33,34. From a surgical perspective, the potential advantages of preoperative MRI in DCIS include fewer re-excisions for positive margins and detection of multifocal or multicentric disease if MRI does indeed provide a better assessment than mammography of disease extent that also leads to fewer reoperations. It is therefore plausible that

MRI could assist in surgical planning of patients with DCIS of the breast. However, data from the current meta-analysis show no evidence that preoperative MRI does improve surgical outcomes

(collective data in Table 2). As in invasive cancer, in practice preoperative MRI in DCIS patients appears to have some potential harms, such as increased mastectomies and increased surgical biopsies13,16,19,20,22. One reason for this is that MRI has limited specificity in patients with DCIS

(ranging in the literature between 58 and 89 per cent) and variable positive predictive value

(ranging between 25 and 84 per cent)35.

One of the most important findings the present meta-analysis was that preoperative MRI increased the odds that the initial surgery was mastectomy (27.6 per cent in the MRI group vs 18.2 per cent in the no-MRI patients). Conversely, the odds of having breast conservation were significantly higher for women who did not have preoperative MRI (82.0 per cent vs 71.5 in the MRI patients). These findings probably reflect MRI’s capacity for detecting multifocal and multicentric disease and so results in more patients being considered unsuitable for BCS, and

6 based on the MRI assessment of disease these patients are considered as being better treated by mastectomy. In this regard, whether mastectomy is justified in lower risk patients even if there is extensive DCIS is far from clear. Although MRI identifies more disease than conventional mammography and ultrasound there is no evidence either in invasive breast cancer or DCIS that treating this MRI-detected disease with more radical surgery improves outcome. In contrast, it can be argued that MRI does harm because mastectomy is usually advised if occult and multicentric MRI-only detected foci are identified through preoperative MRI 27,30,36,. In our the present study, the proportion of patients who had a change in their surgical treatment based on preoperative MRI findings was estimated to be 15.7 per cent. Interestingly, this proportion is similar to the studies on preoperative MRI in invasive breast cancer. A meta- analysis from Houssami et al estimated that a median of 16 per cent of breast cancer patients had additional disease identified on preoperative MRI leading to a change in surgical treatment 14. Fancellu et al, also found that 16.5 per cent of patients eligible for breast conservation had MRI- triggered change of initial surgical treatment13. Based on very limited information on conversion from initial BCS to mastectomy due to MRI findings with description of correlation with final pathology, it has been assumed that such conversion to mastectomy was appropriate because in most cases (22 from 23 patients for the 3 studies) there was more extensive disease, either larger tumour, or multifocal or multicentric disease9,25,28. Although this would seem a fair assumption, it remains unclear whether for some of these patients breast conservation would have led to the same prognostic outcomes had MR not been performed, as suggested in recent IPD meta- analysis17, and this constitutes part of the ongoing controversy on MRI staging of the breast. Furthermore, although again based on limited data, there was indication that MRI did not correlate well with exact tumour size measurement at pathology 25,27 suggesting that the above- mentioned assumed ‘appropriate’ conversion to mastectomy may represent unnecessary radical surgery for at least some of the cases converted to mastectomy on the basis of MRI. Thus, it should be acknowledged that the issue of what is ‘necessary’ versus ‘unnecessary’ mastectomy in patients with DCIS is open to debate and interpretation. The presumed justified conversion from BCS to mastectomy (i.e. based on pathologically-proven more extensive or multicentric disease) may not necessarily translate into patient benefit. Studies have reported that only 30 to 50 percent of DCIS evolve to invasive disease if left untreated 8,37. Given that much DCIS will not evolve to invasive carcinoma, it could be argued that most of the patients with DCIS treated with mastectomy receive an overtreatment. Besides, patients with higher risk DCIS receive radiation after BCS and patients with estrogen positive DCIS often receive adjuvant endocrine treatment 38. These treatments are effective because they treat occult disease in the breast and so if MRI detects disease that would otherwise have been effectively treated by radiotherapy or endocrine treatment it is unlikely to improve outcomes12,13,39. In this context, it is possible that the role of MRI in DCIS will become better defined when there is better understanding of DCIS biology.

Reoperation for tumor-involved margins or incomplete excision after BCS remains a primary concern for breast surgeons and patients. In fact, although DCIS is a non-obligate precursor of invasive carcinoma, it can locally recur as in situ or invasive disease and positive margins are one of the most important risk factors for local recurrences 8,40,41. In the literature, the rate of reoperation ranges between 10 per cent and 68 per cent and depends on multiple variables, including differences in selection criteria for breast conservation, surgeons’ expertise, definition of positive margins, surgical techniques, and tumor histotype 42,43. The rates of reoperation after BCS are, in general, higher for patients undergoing surgery for DCIS than in patients with invasive carcinoma. In a retrospective study, it has been reported that 29.5 per cent of the women with DCIS in England had at least one reoperation after primary BCS within 3 months42. In the present analysis, the rates of positive margins or incomplete excision after BCS, as well as the reoperation rates, did not statistically differ between the MRI and no-MRI patients, despite the

7 fact that MRI should theoretically have provided a more accurate assessment of the tumor extent compared to conventional imaging. These data may be explained by the possibility of discrepancy between the MRI estimate of DCIS size and pathology size (as known to also occur 44,45 with mammography) as observed in MRI-specific studies studies, and by possible difficulty in translating MRI images into surgical procedures (although there are no data to support the latter possibility).

The present meta-analysis showed that the overall mastectomy rate did not statistically differ between those who had and those who did not have MRI, even though the odds of having initial mastectomy were higher in the MRI patients. The reoperation rates (including conversion to mastectomy) were also not statistically different between the two groups. This partly reflects that different studies contributed to the different models in the statistical analysis. There was the suggestion of slightly higher reoperation rates in the MRI group, but the lack of significant differences probably also reflect the limited power in some of the models that are based on only a few studies. The model for ‘positive margins or incomplete excision’ included the largest dataset in this meta-analysis, and indicated the lack of an effect from preoperative MRI for this important surgical endpoint. It is also possible that at re-operation more of the MRI-group had wider excisions whereas more of the no-MRI group had mastectomy, which would have the effect of rendering similar final overall mastectomy rates for both groups despite the initially higher mastectomy rate for the MRI group. Histological studies have shown that most DCIS is unicentric. Much of the multifocal disease identified by MRI may be in continuity rather than truly a second focus 46,47. Multifocality is not in itself a contraindication to BCS in either DCIS or invasive breast cancer and satisfactory rates of local control are possible with BCS48,49 assuming that the appropriateness of BCS is determined by the ability to excise all of the disease with a good cosmetic result 50.

Some authors discuss a possible role of MRI in patients with particular subsets of DCIS 21,27,51.

Unfortunately, only three studies eligible for this meta-analysis reported data on preoperative

MRI according to tumour characteristics9,25,27; hence, there were limited outcomes data suitable for subgroup analyses. Recent literature suggests that DCIS is a heterogeneous disease also from a molecular perspective, including different subtypes with different biological behaviour 52-56.

Identification of subsets of DCIS patients who might benefit from MRI, if any, awaits further definition of the biology associated with progression to invasive disease or recurrence.

8 The costs of MRI are of concerns in the current era of spending accountability, and the increased time usually taken for obtaining MRI images and results is an important aspect when considering routine use of preoperative MRI. Many have raised concerns also about possible delay in surgical treatment due to the time implications of preoperative MRI13,17,57. These issues are particularly relevant when dealing with a disease typically with an excellent life- time prognosis such as DCIS of the breast.

A limitation of this work is the small number of studies that were available for inclusion in this analysis, and that not all outcomes data were available for all studies. Furthermore, as identified in the quality assessment (Table 1), only two studies were randomized trials and most studies used a comparison group rather than randomization, therefore selection bias may have affected the results of the primary studies and also the meta-analysis. The use of a historical comparison group in one study29 was of concern from a methodology perspective to us , hence we its effect on estimated OR’s was explored by including/excluding that study through a sensitivity analysis (as shown in Table 2, its exclusion did not substantially change results). Despite these acknowledged limitations, this work represents the largest comparative analysis of pre-operative MRI in DCIS patients to date, and our the analyses were adjusted for observed imbalances in study-level age between MRI and no-MRI groups, as identified in previous preoperative MRI research 22. When interpreting the age-adjusted estimates, it is important to recognize that a study-level adjustment does not entirely account for the potential effect of age, and further adjustment (either more complex statistically, or including more variables) is not appropriate given that so few studies were available.

In summary, the present meta-analysis has shown that preoperative MRI in patients with DCIS is not associated with an improvement in surgical outcomes. MRI increases the initial rate of mastectomy, although the overall mastectomy rate was not significantly increased from MRI. Importantly, this meta-analysis shows that preoperative MRI does not reduce the odds of having negative margins after BCS nor does it reduce the odds of patients requiring reoperation for positive margins. On the basis of the collective evidence summarized in this meta-analysis, preoperative MRI does not improve the surgical treatment of women with DCIS of the breast. Further studies may be warranted for the identification of subsets of DCIS patients who might potentially benefit from preoperative MRI.

AKNOWLEDGEMENTS

9 N. Houssami receives research support via a National Breast Cancer Foundation (NBCF Australia) Practitioner Fellowship. Disclosure: The authors declare no conflict of interest.

10 REFERENCES 1 Schmale I, Liu S, Rayhanabad J, Russell CA, Sener SF. Ductal carcinoma in situ (DCIS) of the breast: perspectives on biology and controversies in current management. J Surg Oncol 2012; 105: 212-220. 2 Meijnen P, Oldenburg HS, Peterse JL, Bartelink H, Rutgers EJ. Clinical outcome after selective treatment of patients diagnosed with ductal carcinoma in situ of the breast. Ann Surg Oncol 2008; 15 :235-243. 3 Holland R, Peterse JL, Millis RR, Eusebi V, Faverly D, van de Vijver MJ et al. Ductal carcinoma in situ: a proposal for a new classification. Semin Diagn Pathol 1994; 11: 167-180. 4 Siegel R, Naishadham D, Jemal A. Cancer statistics, 2012. CA Cancer J Clin 2012; 62: 10-29. 5 Virnig BA, Wang SY, Tuttle TM. Ductal carcinoma in situ, and the influence of the mode of detection, population characteristics, and other risk factors. Am Soc Clin Oncol Educ Book 2012; 32: 45-48. 6 Estévez LG, Alvarez I, Seguí MÁ, Muñoz M, Margelí M, Miró C et al. Current perspectives of treatment of ductal carcinoma in situ. Cancer Treat Rev 2010; 36: 507-517. 7 Leeper AD, Dixon JM. DCIS of the breast: Are we over-diagnosing it? Are we over- treating it? Maturitas 2011; 68: 295-296. 8 Jansen SA. Ductal carcinoma in situ: detection, diagnosis, and characterization with magnetic resonance imaging. Semin Ultrasound CT MR. 2011; 32 :306-318. 9 Itakura K, Lessing J, Sakata T, Heinzerling A, Vriens E, Wisner D et al. The impact of preoperative magnetic resonance imaging on surgical treatment and outcomes for ductal carcinoma in situ. Clin Breast Cancer 2011; 11: 33-38. 10 Pilewskie M, Kennedy C, Shappell C, Helenowski I, Scholtens D, Hansen N, Bethke K, Jeruss J, Karstaedt P, Khan SA. Effect of MRI on the management of ductal carcinoma in situ of the breast. Ann Surg Oncol 2013; 20: 1522-1529. 11 Yamada T, Mori N, Watanabe M, Kimijima I, Okumoto T, Seiji K et al. Radiologic- pathologic correlation of ductal carcinoma in situ. Radiographics 2010; 30: 1183-1198. 12 Pilewskie M, Olcese C, Eaton A, Patil S, Morris E, Morrow M et al. Perioperative breast MRI is not associated with lower locoregional recurrence rates in DCIS patients treated with or without radiation. Ann Surg Oncol 2014; 21: 1552-1560. 13 Fancellu A, Soro D, Castiglia P, Marras V, Melis M, Cottu P et al. Usefulness of magnetic resonance in patients with invasive cancer eligible for breast conservation: a comparativestudy. Clin Breast Cancer 2014; 14: 114-121. 14 Houssami N, Ciatto S, Macaskill P, Lord SJ, Warren RM, Dixon JM et al. Accuracy and surgical impact of magnetic resonance imaging in breast cancer staging: systematic review and meta-analysis in detection of multifocal and multicentric cancer. J Clin Oncol 2008; 26: 3248-3258 15 Houssami N, Morrow M. Pre-operative breast MRI in women with recently diagnosed breast cancer-where to next? Breast 2010; 19: 1-2. 16 Morrow M, Waters J, Morris E. MRI for breast cancer screening, diagnosis, and treatment. Lancet 2011; 378: 1804-1811. 17 Houssami N, Turner R, Macaskill P, Turnbull LW, McCready DR, Tuttle TM et al. An individual person data meta-analysis of preoperative magnetic resonance imaging and breast cancer recurrence. J Clin Oncol 2014; 32: 392-401. 18 Solin LJ, Orel SG, Hwang WT, Harris EE, Schnall MD. Relationship of breast magnetic resonance imaging to outcome after breast-conservation treatment with radiation for women with early-stage invasive breast carcinoma or ductal carcinoma in situ. J Clin Oncol 2008; 26: 386-391.

11 19 McLaughlin S, Mittendorf EA, Bleicher RJ, McCready DR, King TA. The 2013 Society of Surgical Oncology Susan G. Komen for the Cure Symposium: MRI in breast cancer: where are we now? Ann Surg Oncol 2014; 21: 28-36. 20 Shin HC, Han W, Moon HG, Yom CK, Ahn SK, You JM et al. Limited value and utility of breast MRI in patients undergoing breast-conserving cancer surgery. Ann Surg Oncol 2012; 19: 2572-2579. 21 Schouten van der Velden AP, Schlooz-Vries MS, Boetes C, Wobbes T. Magnetic resonance imaging of ductal carcinoma in situ: what is its clinical application? A review. Am J Surg 2009; 198 :262-269. 22 Houssami N, Turner R, Morrow M. Preoperative magnetic resonance imaging in breast cancer: meta-analysis of surgical outcomes. Ann Surg 2013; 257: 249-55. 23 Turnbull L, Brown S, Harvey I, Olivier C, Drew P, Napp V et al. Comparative effectiveness of MRI in breast cancer (COMICE) trial: a randomized controlled trial. Lancet. 2010; 375: 563-571. 24 Turnbull LW, Brown SR, Olivier C, Harvey I, Brown J, Drew P et al. Multicentre randomised controlled trial examining the cost-effectiveness of contrast-enhanced high field magnetic resonance imaging in women with primary breast cancer scheduled for wide local excision (COMICE). Health Technol Assess. 2010; 14: 1-182. 25 Allen LR, Lago-Toro CE, Hughes JH, Careaga E, Brown AT, Chernick M et al. Is there a role for MRI in the preoperative assessment of patients with DCIS? Ann Surg Oncol 2010; 17: 2395-2400. 26 Peters NH, van Esser S, van den Bosch MA, Storm RK, Plaisier PW, van Dalen T et al. Preoperative MRI and surgical management in patients with nonpalpable breast cancer: the MONET-randomised controlled trial. Eur J Cancer 2011; 47 :879-886. 27 Kropcho LC, Steen ST, Chung AP, Sim MS, Kirsch DL, Giuliano AE. Preoperative breast MRI in the surgical treatment of ductal carcinoma in situ. Breast J. 2012; 18: 151-156. 28 Davis KL, Barth RJ Jr, Gui J, Dann E, Eisenberg B, Rosenkranz K. Use of MRI in preoperative planning for women with newly diagnosed DCIS: risk or benefit? Ann Surg Oncol 2012; 19: 3270-3274. 29 Obdeijn IM, Tilanus-Linthorst MM, Spronk S, van Deurzen CH, de Monye C, Hunink MG et al. Preoperative breast MRI can reduce the rate of tumor-positive resection margins and reoperations in patients undergoing breast-conserving surgery. AJR Am J Roentgenol 2013; 200: 304-310. 30 Siegmann KC, Baur A, Vogel U, Kraemer B, Hahn M, Claussen CD. Risk-benefit analysis of preoperative breast MRI in patients with primary breast cancer. Clin Radiol 2009; 64: 403- 413. 31 Chung A, Saouaf R, Scharre K, Phillips E. The impact of MRI on the treatment of DCIS. Am Surg 2005; 71: 705-710. 32 Raza S, Vallejo M, Chikarmane SA, Birdwell RL. Pure ductal carcinoma in situ: a range of MRI features. AJR Am J Roentgenol 2008; 191: 689-99. 33 Hwang ES, Kinkel K, Esserman LJ, Lu Y, Weidner N, Hylton NM. Magnetic resonance imaging in patients diagnosed with ductal carcinoma-in-situ: Value in the diagnosis of residual disease, occult invasion, and multicentricity. Ann Surg Oncol 2003; 10: 381-388. 34 Menell JH, Morris EA, Dershaw DD, Abramson AF, Brogi E, Liberman L.. Determination of the presence and extent of pure ductal carcinoma in situ by mammography and magnetic resonance imaging. Breast J 2005; 11: 382-390. 35 Brem RF, Fishman M, Rapelyea JA. Detection of Ductal Carcinoma in Situ with Mammography, Breast Specific Gamma Imaging, and Magnetic Resonance Imaging: A Comparative Study. Acad Radiol 2007; 14 :945–950 36 Kapoor NS, Eaton A, King TA, Patil S, Stempel M, Morris E et al. Should breast density influence patient selection for breast-conserving surgery? Ann Surg Oncol 2013; 20: 600-606.

12 37 Erbas B, Provenzano E, Armes J, Gertig D. The natural history of ductal carcinoma in situ of the breast: a review. Breast Cancer Res Treat. 2006; 97: 135-144. 38 Fentiman IS. Lessons learned in breast cancer surgery. Br J Surg 2014; 101: 145-147. 39 Cuzik J, Sestak I, Pinder S, Ellis IO, Forsyth S, Bundred NJ et al. Effect of tamoxifen and radiotherapy in women with locally excised ductal carcinoma in situ: long term results from the UK/ANZ DCIS trial. Lancet Oncol 2011; 12: 21–29. 40 de Roos MA, de Bock GH, Baas PC, de Munck L, Wiggers T, de Vries J. Compliance with guidelines is related to better local recurrence-free survival in ductal carcinoma in situ. Br J Cancer 2005; 93: 1122-1127. 41 Fancellu A, Cottu P, Feo CF, Bertulu D, Giuliani G, Mulas S et al. Sentinel node biopsy in early breast cancer: lessons learned from more than 1000 cases at a single institution. Tumori 2012; 98: 413-420. 42 Jeevan R, Cromwell DA, Trivella M, Lawrence G, Kearins O, Pereira J et al. Reoperation rates after breast conserving surgery for breast cancer among women in England: retrospective study of hospital episode statistics. BMJ 2012; 345: e4505. 43 Bani MR, Lux MP, Heusinger K, Wenkel E, Magener A, Schulz-Wendtland R et al. Factors correlating with reexcision after breast-conserving therapy. Eur J Surg Oncol 2009; 35: 32-37. 44 Mennella S, Garlaschi A, Paparo F, Perillo M, Celenza M, Massa T et al. Magnetic resonance imaging of breast cancer: factors affecting the accuracy of preoperative lesion sizing. Acta Radiol 2014;. pii: 0284185114524089. [Epub ahead of print] 45 Kumar AS, Chen DF, Au A, Chen YY, Leung J, Garwood ER et al. Biologic significance of false-positive magnetic resonance imaging enhancement in the setting of ductal carcinoma in situ. Am J Surg 2006; 192: 520-524. 46 Holland R, Hendriks JH, Vebeek AL, Mravunac M, Schuurmans Stekhoven JH. Extent, distribution, and mammographic/histological correlations of breast ductal carcinoma in situ. Lancet 1990; 335: 519-522. 47 Faverly DR, Burgers L, Bult P, Holland R. Three dimensional imaging of mammary ductal carcinoma in situ: clinical implications. Semin Diagn Pathol 1994; 11: 193-198. 48 Gentilini, O., Botteri, E., Rotmensz, N. Da Lima L, Caliskan M, Garcia-Etienne CA et al. Conservative surgery in patients with multifocal/multicentric breast cancer. Breast Cancer Res Treat. 2009; 113: 577-583. 49 Lim W, Park EH, Choi SL, Seo JY, Kim HJ, Chang MA et al. Breast conserving surgery for multifocal breast cancer. Ann Surg 2009; 249: 87-90. 50 Clough KB, Ihrai T, Oden S, Kaufman G, Massey E, Nos C. Oncoplastic surgery for breast cancer based on tumour location and a quadrant-per-quadrant atlas. Br J Surg 2012; 99: 1389-1395. 51 Greenwood HI, Heller SL, Kim S, Sigmund EE, Shaylor SD, Moy L. Ductal carcinoma in situ of the breasts: review of MR imaging features. Radiographics. 2013; 33: 1569-1588. 52 Solin LJ, Gray R, Baehner FL, Butler SM, Hughes LL, Yoshizawa C et al. A Multigene Expression Assay to Predict Local Recurrence Risk for Ductal Carcinoma In Situ of the Breast. J Natl Cancer Inst 2013, 105: 701-10. 53 Benson JR, Wishart GC. Predictors of recurrence for ductal carcinoma in situ after breast- conserving surgery. Lancet Oncol 2013; 14: e348-357. 54 Joh JE, Esposito NN, Kiluk JV, Laronga C, Lee MC, Loftus L, et al. The effect of Oncotype DX recurrence score on treatment recommendations for patients with estrogen receptor-positive early stage breast cancer and correlation with estimation of recurrence risk by breast cancer specialists. Oncologist 2011; 16: 1520-1526. 55 Allred DC, Wu Y, Mao S, Nagtegaal ID, Lee S, Perou CM, et al. Ductal carcinoma in situ and the emergence of diversity during breast cancer evolution. Clin Cancer Res. 2008 Jan 15; 14: 370-8.

13 56 Polyak K. Molecular markers for the diagnosis and management of ductal carcinoma in situ. J Natl Cancer Inst Monogr. 2010; 41: 210-213. 57 Angarita FA, Acuna SA, Fonseca A, et al. Impact of preoperative breast MRIs on timing of surgery and type of intervention in newly diagnosed breast cancer patients. Ann Surg Oncol 2010; 17: 273-279.

14 FIGURE LEGENDS Figure 1: Systematic search and selection strategy.

Figure 2 : Study-specific and estimated pooled proportions with each surgical outcome for MRI versus no-MRI groups. Study-specific proportions are shown as circles, and pooled proportions are shown as squares; lines represent 95% CI with solid lines for MRI group and dashed lines for no-MRI.

Figure 3 : Estimates of the effect of preoperative MRI on surgical outcomes in DCIS. Number in each model shown in square brackets; adjusted OR is based on study-level age adjustment (see methods)

15 16 17 Table 1: Characteristics of Studies in the Meta-Analysis reporting on Preoperative MRI in DCIS Author/ Eligible year of No. of DCIS patients Characteristics of the study & quality appraisal subjects publication Was randomisation Were consecutive If not a RCT, was Design used to assign to subjects with DCIS comparison group intervention? included or recruited? appropriate?

Patients with biopsy- proven Turnbull Not applicable, a RCT, primary (COMICE) 23, 91 RCT Yes Yes patients randomized within breast 2010 same time-frame (2002-2007) cancer scheduled for wide local excision

Patients with Allen25, 2010 98 DCIS who RCC No Yes Yes, same time-frame (2007) underwent core biopsy

Patients with Yes, same time-frame (2000- Itakura9, 2010 149 DCIS who RCC No Yes 2007) underwent core biopsy

Patients with Peters non-palpable Not applicable, a RCT, (MONET)26, 80 DCIS and IC RCT Yes Yes patients randomized within 2011 who same time-frame (2006-2010) underwent core biopsy

Patients with DCIS Kropcho27, Yes, same time-frame (2002- 158 undergoing RCC No Yes 2011 2009) breast conserving surgery

Patients with DCIS Yes, same time-frame (2003- Shin20, 2012 87 scheduled RCC No Yes 2005) for breast conserving surgery

18 Patients with Yes, same time-frame (2007- Davis28, 2012 218 DCIS who RCC No Unclear 2011) underwent core biopsy

Patients with DCIS and IC Unclear (consecutive for No, a historical comparison: Obdeijn29, 50 suitable for RCC No two different time- MRI group 2007-2010, no-MRI 2013 breast periods) group 2005-2006 conserving surgery

Patients with DCIS who Pilewskie12, Yes, same time-frame (1997- 2321 underwent RCC No Yes 2014 2010) breast conserving surgery

RCC, Retrospective Comparative Cohorts of DCIS patients; RCT, Randomized Controlled Trial; DCIS, Ductal Carcinoma In Situ; IC, Invasive Carcinoma; NR, not reported Table 2 Models Comparing Surgical Outcomes In Subjects with DCIS Who Had Preoperative MRI vs Those Who Did Not Have MRI Model Comparing Outcomes in Those Who Had MRI vs Those Who Did Not OR Adjusted for Age§ Have MRI [No. Had MRI vs No MRI in [Referent: no-MRI] Each Model]* Model OR [Referent: no-MRI] OR 95% CI OR 95% CI

Studies of all subjects with DCIS (9 studies)

Initial surgery: mastectomy [316 vs 279; 18.2%(13.6-24.0) 1.13-2.61 0.012 1.76 1.15-2.69 5 studies]9,25,26,28,29

Initial surgery: BCS [316 vs 279; 82.0(76.1-86.7) 0.36-0.84 0.006 0.53 0.35-0.82 5 studies]9,25,26,28,29

Positive margins or 0.64-1.01 0.059 25.0%(18.6-33.0) 1.10 0.64-1.89 incomplete excision after BCS as initial (0.65-1.03)† (0.082)† surgery [762 vs 1906; 4 studies]23,25,27,29

Reoperation: re-excision or conversion 0.72-1.57 0.759 34.5%(26.1-44.0) 1.04 0.70-1.55 to mastectomy after BCS as initial (0.78-1.72)† (0.475)† surgery [356 vs 276; 5 studies]20,23,25,28,29

Overall mastectomy rate: Initial mastectomies plus mastectomies for 11.5%(2.7-37.4) 0.80-1.9 0.340 0.97 0.61-1.53 positive margins after BCS [378 vs 263; 5 studies]20,25,26,27,28

DCIS, Ductal Carcinoma in Situ; BCS, Breast Conserving Surgery; OR, Odds Ratio; CI, Confidence Interval * Subject numbers vary according to the number of studies reporting the specific surgical outcome and according to whether the analysis applied to all subjects or only those who received initial breast conserving surgery. ¥ P value for model comparing estimates of each surgical outcome in those who had MRI relative to those who did not have MRI (referent group). § OR adjusted for model outcomes using study-level median age (see methods). † For these outcomes (‘Positive margins or incomplete excision after BCS as initial surgery’ and ‘Reoperation: re-excision or conversion to mastectomy after BCS as initial surgery’) the actual OR is reported based on all studies reporting data; in brackets we also report the OR from sensitivity analysis that excluded one study using a historical comparison group.