Margin width is considered the most important risk factor for local recurrence in ductal carcinoma in situ (DCIS) of the breast. The purpose of this report is to assess the predictive utility of lumpectomy specimen margin assessment for the presence and extent of residual DCIS.
Specimens from 253 DCIS cases with lumpectomy and reexcision were studied to determine to the probability of residual DCIS on reexcision. The probability of residual tumor was evaluated with respect to tumor size, margin status, nuclear grade, presence of necrosis, patient age, and the extent of specimen processing (number of sections/volume tissue). Lesions were grouped by size: less than or equal to 2 mm, greater than 2–15 mm, greater than 15–40 mm, or greater than 40 mm. Margin width was recorded as the distance of DCIS to the closest specimen edge or, for positive margins, scored as: extensive (margin involvement in ≥8 sections or >4 low-power fields [LPFs]), moderate (5–7 sections or 2–4 LPFs), minimal (2–4 sections or 1 LPF), or focal (1 section, single focus). The amount of residual tumor was graded by maximum dimension on a semiquantitative basis.
Initial excision margin significantly predicted for the presence of residual tumor on reexcision. Residual tumor was found on reexcision in 85% of extensively positive, 68% of moderately positive, 46% of minimally positive, 30% of focally positive, 41% of greater than 0–1 mm, 31% of greater than 1–2 mm, and 0% of greater than 2 mm margins (P < 0.0001). On univariate analysis, margin width and lesion size of initial excision specimens significantly predicted for the presence of residual DCIS on reexcision. Age, grade, necrosis, and extent of specimen processing were not significant prognostic factors. On multivariate analysis, both initial margin width (P < 0.0001) and lesion size (P = 0.02) significantly predicted for residual DCIS. As for amount of residual tumor, margin width and initial lesion dimension both significantly predicted for medium to large residuum, whereas age 45 years or younger was of borderline significance on univariate analysis. On multivariate analysis, margin width and lesion size on initial excision both remained significant predictors of larger volume residual tumor.
The management of patients with ductal carcinoma in situ (DCIS) of the breast has become a major clinical dilemma. Although the standard of care for the treatment of DCIS remains wide local excision (WLE) followed by adjuvant radiotherapy, there is increasing interest in the use of WLE alone in the management of these lesions.1 The goal of surgery traditionally is to remove enough of the volume of the DCIS lesion so that adjuvant radiotherapy can sterilize the remaining microscopic disease. When excision alone is used, however, the goal of surgery must be to remove all DCIS, or the long-term risk of recurrence without adjuvant treatment can be significant.2
We showed previously that careful examination of the lumpectomy margin in invasive breast carcinoma treated with breast-conserving therapy can predict for residual tumor with an expected error consistent with its role as a sampling procedure.3 Regarding DCIS, others have proposed various schema to reliably choose patients suitable for treatment with excision alone. The Van Nuys Prognostic Index was proposed by Silverstein et al.4 and considered multiple prognostic factors in an attempt to stratify patients according to local failure risk plus or minus radiotherapy. More recently, this idea of patient stratification has been simplified such that tumor size, grade, and histology are subordinated to margin width as a primary indicator of local recurrence risk in DCIS.5
We propose that the goal of WLE when treating without adjuvant radiotherapy must include a necessary condition that all microscopically definable DCIS in the breast is eliminated. Patients for whom it is suitable to forgo adjuvant radiotherapy are usually cases in which all disease has been surgically removed. We examined clinical and pathologic features of DCIS lesions that underwent an initial WLE followed by a later reexcision to determine whether women with pathologically no residual DCIS can reliably be predicted.
METHODS AND MATERIALS
A total of 253 DCIS cases treated with lumpectomy and reexcision from 1987 to 2000 were reviewed and form the basis for this report. Reexcision for margins less than or equal to 2 mm were performed per institutional policy, although final treatment decisions were at physician and patient discretion. Cases with microinvasion as defined by Silver and Tavassoli6 were excluded from analysis. All tumor excisions were conducted with the objective of complete tumor removal with a normal margin of greater than 5 mm. Surgical specimens were oriented with marking sutures and promptly delivered to the pathologist where they were dried and coated with india ink on their external surfaces. Specimens were measured in three dimensions, then multiple sections were obtained with the intent of evaluating deep, superficial, lateral, medial, superior, and inferior surfaces. Additional sections were obtained where the tumor approached the margin of the specimen. Specimen mammograms were obtained in all cases, and postexcisional mammograms were performed as needed to verify complete removal of the radiographic abnormality.
Tumor size was recorded as the greatest dimension of tumor extent as determined pathologically. For multifocal tumors, this included the entire area of abnormality. Residual tumor identified in reexcision specimens was graded using the following criteria:
1Microscopic, a single microscopic focus present in a single histologic section
2Small, tumor in one LPF and/or two to four sections
3Medium, tumor in two to four LPFs and/or five to seven sections
4Large, tumor in five or more LPFs and/or eight or more sections.
All LPFs were examined using a 2-mm lens. Specimen blocks were cut with a thickness 3.0–4.5 mm, with most blocks having a thickness of 3.5–4.0 mm. Nuclear grade was determined from original pathology reports, and when the grade was reported as intermediate between two values, the higher was chosen for reporting purposes.
Margins were classified by the closest margin to the DCIS specimen edge as measured using a micrometer. For those tumors with a positive margin, the extent of positivity was graded using the following criteria:
1Focal, margin involvement by a single microscopic focus in one histologic section
2Minimal, margin involvement in one LPF and/or limited to involvement in two to four sections at one geographic edge of the specimen
3Moderate, margin involvement in two to four LPFs and/or present in five to seven sections
4Extensive, margin involvement in five or more LPFs and/or eight or more sections.
Volume of resected tissue was calculated as macroscopic specimen length times width times depth. Extent of tissue sectioning was determined by the number of tissue sections per specimen volume and expressed numerically as section/volume (S/V ratio).
Relations among baseline variables were examined using the chi-square test. Univariate and multivariate analyses were performed using logistic regression models. Analyses were performed using SAS software system V8.01 for Windows (SAS, Cary, NC). A probability level of alpha = 0.05 was used to determine statistical significance.
The median age of the patients studied was 53 years (range, 31–90). Women aged younger than 45 years old composed 23% of the cohort (n = 58). The mean volume of tissue excised in the initial excision and reexcision were 59 and 91 cc (excluding mastectomy specimens; 404 cc including mastectomies), respectively. The mean volume of tissue excised in WLE for initial excisions was 40 cc for women younger than 45 years old and 53 cc for women 45 years old and older (P = 0.48). For reexcisions, women younger than 45 years old had a mean volume of tissue excised of 66 cc whereas women 45 years old and older had a mean reexcision of 98 cc (P = 0.12).
Demographic and pathologic characteristics of the cohort as a whole are shown in Table 1. Initial excision margins were positive in 127 (50%) specimens, greater than 0–1 mm in 100 (40%), greater than 1–2 mm in 13 (5%), and greater than 2 mm in 10 (4%). Maximum dimension of DCIS in the initial excision was less than or equal to 2 mm in 17 (7%) specimens, greater than 2–15 mm in 117 (46%) specimens, greater than 15–40 mm in 64 (25%) specimens, and greater than 40 mm in 45 (18%) specimens. Regarding nuclear grade, 46 (18%), 92 (36%), and 71 (28%) patients had specimens with Grades 1, 2, and 3, respectively. Necrosis was present in 115 (45%) specimens. Most patients (81%) underwent a lumpectomy as their reexcision procedure, with the remainder (19%) having a mastectomy as their reexcision.
Table 1. Distribution of Evaluated Cases According to Clinical and Pathologic Variables (n = 253)
No. of patients
RE: reexcision; Pos: positive.
Patient age (yrs)
Tumor dimension (mm)
Initial margin (mm)
Type of RE
On reexcision 122 of 253 (48%) specimens showed no evidence of residual tumor. Twenty-six (10%) specimens showed microscopic, 69 (27%) small, 18 (7%) medium, and 10 (4%) large amounts of residual tumor on reexcision. Of the reexcision specimens, 185 (73%) had final margins 2 mm or greater. Twenty-eight (11%) reexcision specimens had positive margins, whereas 40 (16%) specimens had final reexcision margins of greater than 0–2 mm.
To assess the influence of the relative closeness of the measured margin on the probability of finding residual tumor on reexcision, we distributed the evaluated cases accordingly (Table 2). Initial excision margin was found to significantly predict for residual tumor in reexcision specimens with 63%, 41%, 31%, and 0% of positive, greater than 0–1-mm, greater than 1–2-mm, and greater than 2-mm specimens having residual tumor on reexcision (P < 0.0001, Fig. 1). When positive margins were graded by degree of positivity, 85% of extensively positive, 68% of moderately positive, 46% of minimally positive, and 30% of focally positive initial excision specimens contained residual tumor on reexcision (P < 0.0001, Table 3). Overall, 80% of tumors with a moderate to extensively positive initial margin had residual tumor whereas 38% of tumors with a less than moderately positive or negative initial margin had pathologic residuum (P < 0.0001).
Table 2. Probability of Finding Residual Tumor in the Reexcision Specimen as a Function of the Closeness of the Initial Excision Margin, Tumor Dimension, Grade, Presence of Necrosis, and Patient Age
Tumor in reexcision (%)
Initial excision margin (mm)
Initial disease extent (mm)
Patient age (yrs)
Table 3. The Incidence and Extent of Tumor Found at Reexcision as Related to the Initial Margin Statusabc
Margin status (n)
Percentage + RE
Extent residual tumor (row %)
Pos: positive; RE: reexcision.
In the analysis of extent of positivity vs. (+) RE, the chi-square table excluded the unknowns and resulted in a P value of < 0.0001.
In the analysis of extent of positivity vs. (+) RE for those specimens with initial positive margins, the chi-square table excluded the unknowns and combined minimal/focal and extensive/moderate with a resulting P value of < 0.0001.
In the analysis of extent of positivity vs. extent of residual tumor, the chi-square tumor excluded the unknowns with a resulting P value of < 0.0001.
> 0–1 mm (94)
> 1–2 mm (13)
> 2 mm (10)
Tumors with more extensively positive margins were significantly more likely to have a larger volume of tumor on reexcision than specimens with less extensively positive margins (Table 3). Specimens with extensively positive margins had medium/large residual tumor in 23% of reexcision specimens whereas those with moderately positive margins had similar residual in 26% of specimens. Less positive or negative margins had a likelihood of medium to large residual tumor on reexcision of 0–9%. When examining amount of residual tumor, initial specimens with any positive margin on reexcision showed no residual tumor in 37% of cases, microscopic residual in 11%, small in 37%, medium in 10%, and large in 5%. Cases with initially negative margins showed no evidence of residual tumor in 63% of reexcisions, with 10%, 19%, 4%, and 3% having microscopic, small, medium, and large residual tumor, respectively.
Section/volume (S/V) ratios were studied to determine whether degree of pathologic specimen processing influenced the likelihood of detecting residual DCIS on reexcision or of finding positive margins on initial excision. Mean S/V ratios were 0.60 for initial excision and 0.30 for reexcision specimens (including mastectomy specimens; 0.37 excluding mastectomies). Table 4 shows the distribution of S/V ratios for the entire cohort. Using S/V ratio as a surrogate marker, degree of specimen processing was not found to influence the likelihood of finding a positive margin or residual DCIS. Specimen processing was conducted in only two laboratories, however; therefore, there is a fairly narrow range over which this variable is studied.
Table 4. Relations of the Section/Volume Ratio of the Initial and Reexcision Specimens to the Probability of Finding Residual Tumor at Reexcision
S/V: section/volume ratio.
Chi-square test of initial excision S/V ratio vs. probability of residual tumor on reexcision P value 0.39.
Chi-square test of reexcision S/V ratio vs. probability of residual tumor on reexcision P value 0.47.
Results of univariate logistic regression analyses for the probability of finding tumor in reexcision specimens are shown in Table 5. Initial excision margin and initial DCIS maximum dimension both were found to be highly significant predictors for the probability of residual DCIS in reexcision. Although nuclear grade of 1 versus 2 versus 3 was not significant (P = 0.12), a nuclear grade of 2 or 3 versus 1 was found to be of borderline significance (P = 0.07). A nuclear grade of 1 or 2 versus 3, young age, S/V ratio, and the presence of tumor necrosis also were not significant variables in our analysis. On multivariate analysis, lesion dimension (P = 0.02) and initial excision margin (P < 0.0001) both were found to significantly predict for residual DCIS on reexcision.
Table 5. Results of Logistic Regression Analyses of the Probability of Residual Tumor at Reexcision (+RE)
Univariate and multivariate logistic regression analyses also were performed for the probability of finding a medium/large dimension of tumor in reexcision specimens. Initial excision margin and initial DCIS disease dimension significantly predicted for medium/large residual DCIS in reexcision specimens whereas age younger than 45 years was of borderline significance (P = 0.07) on univariate analysis (Table 6). Grade, S/V ratio, and the presence of tumor necrosis were not found to significantly predict for the presence of a medium/large residual tumor in our analysis. On multivariate analysis, however, only initial disease dimension (P = 0.02) and initial excision margin (P < 0.0001) remained significant.
Table 6. Results of Logistic Regression Analyses of the Probability of Medium/Large Residual Tumor on Reexcision
The role of postoperative breast irradiation in DCIS of the breast has been established by two large, randomized trials. The first such trial was the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-17, which randomized 818 women with DCIS and uninvolved margins to WLE alone or WLE plus adjuvant irradiation.7 At a median follow-up of 90 months, this trial found a decrease in ipsilateral invasive or noninvasive recurrence from 27% to 12% with the addition of breast irradiation.8 The European Organization for Research and Treatment of Cancer (EORTC) trial 10853 randomized 1002 women with DCIS and negative margins to WLE or WLE plus adjuvant irradiation and reports a decrease in 4-year local failure rates from 17% to 11% with the addition of whole breast irradiation to WLE.9
The NSABP B-24 trial served both to confirm the experimental arm of NSABP trial B-17 and to determine the additional beneficial effects of tamoxifen.10 In the NSABP B-24 trial, 1804 women with DCIS, including women with margins involved by tumor, were randomized to WLE with adjuvant radiotherapy plus either placebo or tamoxifen for 5 years. In this trial, the addition of tamoxifen to postoperative radiotherapy reduced both ipsilateral and total breast carcinoma events at 5 years follow-up. Taken as a whole, these 3 randomized trials show 5-year breast carcinoma events with DCIS to be approximately 25% with WLE alone, 13% with WLE and breast irradiation, and 8% with WLE, breast irradiation, and tamoxifen. The overall risk of local recurrence is shown, on an annualized basis, to be 3.4–4.0% with WLE alone, 1.9–2.3% with WLE plus adjuvant irradiation, and 1.2% with WLE, adjuvant irradiation, and tamoxifen.
Despite the aforementioned results, interest in treatment of DCIS of the breast with WLE alone remains intense. There may be a subset of women at such low risk of local recurrence that the relative risk reduction by the addition of breast irradiation to WLE is negligible. Potential clinical and pathologic factors proposed to help define this subset include young age, grade, margin width, tumor dimension, architecture, and the presence of tumor necrosis.4 Various single-institution studies have studied this and often have produced inconsistent results regarding risk factors for local recurrence with and without adjuvant irradiation.11–14 This may be explained in part by the lack of data with the necessary prolonged follow-up.
According to a central review of cases enrolled in the NSABP B-17 trial, comedo architecture with necrosis and involved margins were the only significant pathologic predictors of local recurrence,15 although margin status lost statistical significance on later reanalysis.16 Patients in the NSABP B17 trial with negative margins had an annualized ipsilateral breast tumor recurrence rate of 3.6% and 1.6% with WLE or WLE plus adjuvant irradiation, respectively. Patients in the same study with uncertain or involved margins had annualized ipsilateral breast tumor recurrence rates of 4.8% and 2.3% with WLE or WLE plus adjuvant irradiation, respectively. Notably, no group of patients could be identified in the NSABP analysis that did not have a significantly improved local control rate with the addition of adjuvant irradiation.
The EORTC recently has reviewed clinical and pathologic risk factors for recurrence in DCIS lesions treated on trial 10853.17 On multivariate analysis, age, clinical presentation, margins, architecture, and adjuvant radiotherapy were all significant factors in predicting for local recurrence. Lesion size was not a significant prognostic factor, but only approximately 25% of lesions had a measured dimension and of these, greater than 90% were less than or equal to 2 cm in size. In addition, the trial eligibility did not require exact reporting of margin width; therefore, it was not designed to identify a margin width with which it is safe to treat with WLE alone. Of note, the trial does show a local recurrence rate of 18% for patients treated with WLE alone who underwent a reexcision showing no evidence of residual DCIS. This raises concerns regarding surgical misses of excisional cavities as well as of underappreciation of multifocality of disease. In many series, a reexcision with no residual DCIS is considered for reporting purposes a specimen with greater than or equal to 10-mm margins. In addition, high rates of local failure are seen in high-risk groups, such as those patients with high nuclear grade or positive margins, with or without the use of adjuvant radiotherapy. This suggests that in these high-risk subgroups, the volume of residual DCIS may be routinely underestimated and too large to be sterilized by adjuvant radiation.
A retrospective analysis of a patient series from the Memorial Sloan-Kettering Cancer Center has shown age, comedo subtype, grade, margin status, and the use of postoperative radiotherapy all to be significant predictors of local failure on univariate analysis. On multivariate analysis, however, only margin status remained significant.18 Others have largely confirmed these risk factors.11–14, 19, 20 Of potential prognostic factors, margin width has garnered the most attention, both because of the proposal that this is the dominant factor in predicting for local recurrence with WLE alone5 and because of the finding that this is the only potential risk factor that can be influenced by physician intervention.
For invasive breast carcinoma, the status of the specimen margin on initial excision specimen has been shown to predict for residual tumor on reexcision, although this is restricted by the limitations of margin assessment as a sampling procedure.3 Our study shows that this is true for DCIS as well. Margin width on initial excision is the dominant factor in predicting the presence of residual tumor in DCIS reexcisions. In addition, more extensively positive margins are more likely to contain residual tumor on reexcision than those specimens with less extensively positive margins. Patients with minimally or focally positive margins on initial excision have a likelihood of having residual DCIS on reexcision that is similar to patients with close margins, but significantly less than patients with extensively or moderately positive margins. Goldstein et al., in a smaller data set from William Beaumont Hospital, have made similar findings.21 These results are consistent with a theory of escalating residual tumor burden reflected in increasing margin positivity.22
Much interest has focused on the data of Silverstein et al. suggesting a very low rate of local recurrence in women with greater than or equal to 10-mm margins treated with WLE alone, regardless of other risk factors.5 Silverstein studied 469 women with DCIS treated at two hospitals. Although the data are retrospective, local failure rates in patients with margins greater than or equal to 10 mm are only 2–3% in those treated with WLE with or without adjuvant radiation (P = 0.92).5 The number of these lesions with margins directly measured by ocular micrometer as greater than or equal to 10 mm as compared with those placed in this group because of a reexcision without evidence of residual tumor is not stated. This very low local failure rate does not seem to be consistent with EORTC analysis of their data, in which 18% of patients with negative reexcision had local failure at 4.25 years follow-up when treated with surgery alone.9 Our study has a limited number of patients with margins greater than 2 mm who still underwent a surgical reexcision. Despite this, no patients in this cohort had residual DCIS on reexcision. It is possible that in terms of likelihood of harboring residual DCIS, patients with negative margins of greater than 2 mm, as directly measured from tumor edge to specimen margin, are very dissimilar to patients with reexcisions without evidence of residuum. Although our data suggest that if margins are directly measured to be greater than 2 mm then residual DCIS in the breast is unlikely, the EORTC data suggest that this is not true for patients with negative reexcisions. Particular emphasis should be placed, perhaps, on obtaining sufficient margins on initial excision if there are plans to forgo adjuvant radiotherapy.
Lesion dimension also is shown to be a significant prognostic factor in our data for the presence of residual DCIS on reexcision. On univariate and multivariate analyses, dimension significantly predicts for the presence of residual disease. In addition, on both univariate and multivariate analyses larger dimension significantly predicts for larger amounts of residual DCIS. Dimension is often difficult to assess in DCIS because of multifocality, both real and because of specimen processing. Our results show that an educated estimation of lesion dimension is worthwhile and important. Although exact determination of lesion size may be difficult, our results show that pathologists and clinicians should try in all cases to determine lesion size, at least as grouped as microscopic (< 2 mm), small (2–15 mm), medium (16–40 mm), and large (> 40 mm).
Contrary to some previously published reports,18, 23 young age does not predict for the presence of residual tumor on reexcision in our data. The reasons for this are unclear. There is a suggestion in our data that reexcision specimens in younger women tend to be smaller. This may be because of physician concerns regarding cosmetic outcome in younger women requiring reexcisions. If concerns about cosmesis, in turn, have driven smaller reexcisions in younger women, this could cause an underestimation of the true residual tumor burden in these women compared with older women having larger reexcisions.
Central to the controversy regarding the use of WLE alone in DCIS are the relative roles of surgery and radiation therapy. When adjuvant radiotherapy is planned, a low volume of residual DCIS is accepted. When excision alone is planned, however, the goal of surgery should be to eliminate all microscopically detectable DCIS in the breast. Margin assessment remains the most reliable method of determining whether this goal has been attained, according to our data. The theory that there are subsets of women based on age, grade, and tumor architecture in whom low volumes of residual DCIS are unlikely to progress and can be accepted without reexcision or radiotherapy is tempting but has yet to be demonstrated in a controlled prospective manner. This study should serve to help identify those women less likely to have residual tumor on reexcision and therefore are suitable for future study of treatment with WLE alone.