Risk Factors for Recurrence and Metastasis After Breast- Conserving Therapy for Ductal Carcinoma-In-Situ: Analysis of European Organization for Research and Treatment of Cancer Trial 10853 By Nina Bijker, Johannes L. Peterse, Luc Duchateau, Jean-Pierre Julien, Ian S. Fentiman, Christian Duval, Silvana Di Palma, Joe ��lle Simony-Lafontaine, Isabelle de Mascarel, and Marc J. van de Vijver Purpose: In view of the increasing number of pa- tients treated with breast-conserving treatment (BCT) for ductal carcinoma-in-situ (DCIS), risk factors for re- currence and metastasis should be identified. Patients and Methods: Clinical and pathologic char- acteristics from patients with DCIS in the European Organization for Research and Treatment of Cancer trial 10853 (excision with or without radiotherapy) were related to the risk of recurrence. Pathologic fea- tures were derived from a central review of 863 of the 1,010 randomized cases (85%). The median follow-up was 5.4 years. Results: Factors associated with an increased risk of local recurrence in the multivariate analysis were young age ( 40 years) (hazard ratio, 2.14 P 5 .02), symptomatic detection of DCIS (hazard ratio, 1.80 P 5 .008), growth pattern (solid and cribriform) (hazard ratios, 2.67 and 2.69, respectively P 5 .012), involved margins (hazard ratio, 2.07 P 5 .0008), and treatment by local excision alone (hazard ratio, 1.74 P 5 .009). The risk of invasive recurrence was not related to the histologic type of DCIS (P 5 .63), but the risk of distant metastasis was significantly higher in poorly differen- tiated DCIS compared with well-differentiated DCIS (hazard ratio, 6.57 P 5 .01). Conclusion: Patients with poorly differentiated DCIS have a high risk of distant metastasis after invasive local recurrence. Margin status is the most important factor in the success of BCT for DCIS additionally, young age and symptomatic detection of DCIS have negative prognostic value. J Clin Oncol 19:2263-2271. �� 2001 by American Society of Clinical Oncology. CONTROVERSY EXISTS concerning breast-conserv- ing treatment (BCT) for ductal carcinoma-in-situ (DCIS). BCT has proven to be an equivalent alternative to mastectomy in invasive breast cancer. However, whereas ablative treatment for DCIS can almost invariably achieve cure, BCT for DCIS implies a risk of residual disease that might progress into invasive carcinoma, thereby increasing the risk of death from metastatic disease. Two randomized clinical trials have shown that radio- therapy reduces the risk of local recurrence after local excision of DCIS.1-3 Even with radiotherapy, however, recurrence rates of about 10% at 5 years were observed in both trials. The assessment of clinical and pathologic factors that are associated with the risk of recurrence would enable the identification of subgroups of patients for whom ablative treatment should be preferred to BCT, and subgroups that may be treated by local excision without radiotherapy. DCIS forms a heterogeneous group of lesions in which various morphologic subtypes can be recognized. In recent years, several classification systems of DCIS have been proposed.4-7 Nonrandomized studies have suggested that well-differentiated (or low-grade) DCIS is associated with a low risk of recurrence, whereas poorly differentiated (or high-grade) DCIS has a high risk of recurrence and progres- sion to invasion.8-13 In addition to the histologic type, the extent of the lesion and the width of the tumor-free margin play a role in local control after BCT for DCIS.13-18 At present, it is unknown which factors are associated with the risk of distant metastasis and death. The first results of the European Organization for Re- search and Treatment of Cancer (EORTC) 10853 study were published recently.1 In this phase III randomized clinical trial, 1,010 patients with DCIS were randomized between no further treatment and external irradiation (50 Gy in 25 fractions to the whole breast) after microscopic complete excision of the lesion. The trial included a central pathology review to evaluate the diagnosis and classifica- tion of the lesion. Here we report the results of this From the Departments of Radiation Oncology and Pathology, the Netherlands Cancer Institute, Amsterdam, the Netherlands EORTC Data Center, Brussels, Belgium Departments of Surgery and Pathol- ogy, Centre Henri Becquerel, Rouen, Department of Pathology, C.R.L.C. Val D���Aurelle, Montpellier, and Department of Pathology, Institut Bergonie ��, Bordeaux, France Clinical Oncology Unit, Guy���s Hospital, London, United Kingdom and Department of Pathology, Istituto Nazionale dei Tumori, Milano, Italy. Submitted June 30, 2000 accepted January 4, 2001. Address reprint requests to Johannes L. Peterse, MD, Department of Pathology, the Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, the Netherlands. �� 2001 by American Society of Clinical Oncology. 0732-183X/01/1908-2263 2263 Journal of Clinical Oncology, Vol 19, No 8 (April 15), 2001: pp 2263-2271

pathology review and relate clinical and pathologic factors with local recurrence, distant metastasis, and survival. PATIENTS AND METHODS Between January 1986 and July 1996, 1,010 women from 46 institutes were enrolled in the EORTC DCIS study. Detailed informa- tion about study design, eligibility criteria, follow-up procedures, and identification of endpoints have been described elsewhere.1 Patients were selected on the basis of the diagnosis of the local pathologist diagnostic requirements were DCIS without invasion and uninvolved margins. The pathology protocol advised to sample extensively, pref- erably directed by a lamellated specimen mammogram, both from grossly and from mammographically abnormal areas, including all inked margins. The pathology review was performed by one of the authors (J.L.P.) and focused on the diagnosis and classification of the lesion. The centers were either requested to send slides and pathology reports or were site-visited by the reviewer. In case of discordant diagnoses, a third pathologist���s (M.V.) opinion was requested to obtain a final diagnosis. The slides of 889 patients were available for review. These slides were collected through 35 participating institutes, from over 50 pathology laboratories in Europe. In 20 cases, inadequate material was sent and in six cases the quality of the slides did not allow a reliable diagnosis, leaving 863 cases (85%) for analysis. The distribution of the clinical characteristics including age, method of detection, and number of recurrences per treatment arm was similar for reviewed and nonreviewed cases (data not shown). In 57%, all slides were available for review, and for these the median number was 17 (range, five to 66). In the remaining 43%, a selection of slides was sent. Therefore, the extent of the lesion and the width of the tumor-free margin could not reliably be assessed by review of the histologic slides. Thus, the pathology reports were reviewed for the extent of the lesion and the margin status. For the size, only those measurements that were mentioned in the Microscopy section and/or in the Conclusion of the report were used. For margin status, it was recorded whether it was reported whether it was considered free (. 1 mm), close (# 1 mm), or involved and whether an exact measurement of the minimal width of the tumor-free margin was given. When a re-excision was performed and no residual DCIS was found, this was recorded separately. The protocol only allowed the enrollment of patients with DCIS, without evidence of (micro) invasion. In the review, invasion was defined as a focus with the common features of invasive carcinoma of at least 1 mm outside the periductal stromal cuff. For cases of DCIS, the following histologic features were assessed: cytonuclear grade (low, moderate, or high), necrosis (none or moderate/ marked), and the main architectural growth pattern (clinging, micro- papillary, cribriform, or solid/comedo). Eighteen intracystic papillary carcinomas were included in the category of cribriform growth pattern. All cases were classified according to Holland et al5 into well- differentiated, intermediately differentiated, and poorly differentiated subtypes. Initially, necrosis was not scored separately. After publica- tion of the Van Nuys classification,4 necrosis was recorded in the histology review. The Van Nuys classification was applied for the 559 cases in which both nuclear grade and the presence of necrosis were scored. At the time of analysis, the median duration of follow-up was 5.4 years. Univariate analyses stratified for treatment were performed for the age of the patient, method of detection, and all histologic variables in order to assess their prognostic value for local recurrence separately. For these analyses, cases with benign proliferative lesions and lobular carcinoma in situ (LCIS) have been pooled, as well as the invasive lesions with the cases that showed suspicion of invasion. Age was dichotomized (# 40 years v . 40 years), method of detection was divided into symptomatically (nipple discharge, palpable lesion, or both) versus mammographically detected. Cases with a clinging growth pattern were grouped with those with a micropapillary architecture. In addition, the histologic type and architecture of the lesion were related to the risk of DCIS recurrence, invasive recurrence, metastasis, and death. Since nuclear grade and histologic type were highly correlated and the latter was more related to the risk of metastasis and death, the association of nuclear grade with the risk of these events is not presented. The Cox proportional hazards model was fitted for each of these variables with treatment as a stratification factor, and the comparison between the different levels of the variable was made on the basis of the likelihood ratio test. The hazard ratios are presented with the level of the variable considered best as the baseline. Only overall tests over the two treatment arms are performed because too few events are available to test in the treatment subgroups separately or to test for interaction between the treatment and the prognostic factor. A Cox proportional hazards regression model19 was fitted for the multivariate analysis of local recurrence, using variables with signifi- cant P values (P , .05) in the univariate analysis. Since nuclear grade and histologic type were highly correlated, only the histologic type was included in the multivariate analysis. The presence of necrosis and the Van Nuys classification were not used in the multivariate analysis because they were applied for only a subset of patients. For the multivariate analysis, all categories with free margins were grouped, and the close/involved margins were categorized with the nonspecified margins. The likelihood ratio test was used for testing the effect of each of the variables after adjusting for the other variables in the multivariate model. Finally, the association between variables used in the multivar- iate analysis was tested by the likelihood ratio test. Statistical analyses were performed using the SAS software (SAS Institute, Cary, NC). RESULTS Pathology Review The diagnosis of DCIS was confirmed in 775 of 863 lesions (90%). In 45 cases (5%), benign proliferative lesions were diagnosed, consisting of 22 cases with epithelial hyperplasia without atypia and 23 with complex sclerosing lesions. Invasive growth was found in 27 cases (3%), and in another 13 (1.5%) there was suspicion of invasion. In three cases, the lesion was classified as LCIS. The pathology reports described a diameter in 178 cases (25%). Size was reported to be smaller than 1 cm in 121 (68%), between 1 and 2 cm in 41 (23%), and larger than 2 cm in 16 cases (9%). The pathology reports did not specify margin status in 88 cases (12%). Margins were reported ���free��� without further specification in 350 cases (49%), an exact distance was given in 36 cases (5%), no residual DCIS was found at surgical re-excision in 147 (21%), and the margins were reported close/involved in 62 cases (9%). At review, of 775 2264 BIJKER ET AL