The authors reviewed their institution's experience treating patients with mammographically detected ductal carcinoma in situ (DCIS) of the breast with breast-conserving therapy (BCT) to determine 10-year rates of local control and survival and to identify factors associated with local recurrence.
From January 1980 to December 1993, 132 breasts in 130 patients were treated with BCT for mammographically detected DCIS at William Beaumont Hospital, Royal Oak, Michigan. All patients underwent an excisional biopsy, and 64% were reexcised. All patients received postoperative whole-breast irradiation to a median dose of 45.0 Gray (Gy) (range: 43.1–56.0 Gy). One hundred twenty-four cases (94%) received a boost to the tumor bed for a median total dose of 60.4 Gy (range: 45.0–71.8 Gy). All cases underwent complete pathologic review by one pathologist. The median follow-up was 7.0 years.
Of the entire study group, 13 patients developed recurrence within the ipsilateral breast, for 5- and 10-year actuarial rates of 8.9% and 10.3%, respectively. Nine of the 13 recurrences (69%) occurred within or immediately adjacent to the lumpectomy cavity and were designated as true recurrences or marginal misses (TR/MM). Four patients (31%) had recurrence elsewhere in the breast. Ten of the 13 recurrences (77%) were invasive, whereas 3 (23%) were pure DCIS. Only 1 patient died of disease, corresponding to 5- and 10-year actuarial cause specific survival rates of 100% and 99.0%, respectively. Multiple clinical, pathologic, and treatment-related factors were analyzed for association with ipsilateral breast failure or TR/MM. In multivariate analysis, only the absence of pathologic calcifications was significantly associated with ipsilateral breast failure. When specifically analyzed for TR/MM, younger age at diagnosis, number of slides with DCIS, number of DCIS and cancerization of lobules (COL) foci within 5 mm of the margin, and the absence of pathologic calcifications demonstrated a statistically significant association. Close or positive margin status did not significantly predict for either TR/MM (P = 0.14) or ipsilateral breast failure (P = 0.19).
The widespread use of screening mammography to detect breast carcinoma has led to a significant increase in the proportion of patients who are diagnosed with clinically occult ductal carcinoma in situ (DCIS).1 Commensurate with this increase in noninvasive cancers has been the improved utilization of breast-conserving therapy (BCT) at most centers throughout the United States.2 Although many more patients diagnosed with mammographically detected DCIS are now treated with BCT, the factors associated with optimal local control are not well defined.3–5 Despite the extremely high local control and survival rates in the majority of patients, a very small (but real) percentage of patients still suffer local treatment failure and ultimately may die of their malignancy.6 Efforts directed at defining criteria associated with local control theoretically may lead to a reduction in the rate of mortality in these patients. To this end, we retrospectively analyzed our institution's experience treating a large population of patients with mammographically detected DCIS with BCT to determine factors associated with local recurrence. To eliminate the confounding effects of incomplete clinical and pathologic information on results, only mammographically detected tumors with complete histologic review (by one pathologist) were included in this analysis.
MATERIALS AND METHODS
From January of 1980 through December of 1993, 169 breasts in 167 patients were treated for DCIS with lumpectomy followed by radiation therapy (RT) at William Beaumont Hospital in Royal Oak, Michigan. One hundred fifty-one of these cases (89%) were detected by screening mammography. Of those cases that were detected mammographically, 137 (91%) underwent complete pathologic review. Five cases were excluded based on findings at histologic review (two for invasive carcinoma and three for atypical ductal hyperplasia). The study population included 128 patients with unilateral DCIS and 2 patients with bilateral DCIS for a total of 132 breasts with pathologically confirmed, mammographically detected DCIS in 130 patients.
All women had American Joint Committee on Cancer (AJCC) clinical Stage 0 disease (Tis N0 M0) DCIS of the breast at presentation.7 Patients with the following findings were excluded in this analysis: 1) invasive carcinoma of the breast, 2) microinvasive carcinoma of the breast, 3) initial detection by any method other than mammography, or 4) incomplete pathologic review. Six patients (5%) with previous or simultaneous contralateral breast carcinoma (two with DCIS and four with invasive cancer) were included in the study population. However, those four patients with previous or simultaneous contralateral invasive breast carcinoma were excluded from all survival analyses.
All women were treated with excision followed by breast irradiation. The surgical treatment for all patients included at least an excisional biopsy, defined as an excision of tissue around the tip of the needle localization wire. The initial excision was guided by mammographic needle localization for every patient. Eighty-four patients (64%) underwent a reexcision of the primary tumor site due to close (≤2 mm), positive, or uncertain margins at the discretion of the surgeon or the radiation oncologist. For some patients, postexcisional (preirradiation) mammograms were obtained to exclude residual microcalcifications in the breast.
Pathologic lymph node staging was performed in 68 patients (52%). All excised lymph nodes were free of metastasis. The median number of lymph nodes removed was 14 (range, 1–28 lymph nodes). Since 1990, surgical staging of the axilla has not been routinely performed.
Our radiation technique has been reported previously.5, 8 Briefly, RT was initiated at a median interval of 5.1 weeks after the last surgical procedure (range, 0–32 weeks). The entire breast was irradiated with 4–6 MV photons to a median dose of 45.0 Gray (Gy) (range, 43.1–56.0 Gy). Whole-breast irradiation was followed by a supplemental boost to the tumor bed in 124 patients (94%) for a median total dose of 60.4 Gy (range, 45.0–71.8 Gy) to the tumor bed. The tumor bed was boosted with electrons in 96 patients, with an interstitial implant in 27 patients, or with photons in 1 patient. The 8 patients who did not receive a boost received a median of 50.4 Gy to the entire breast (range, 45.0–56.0 Gy). Regional lymphatics were not treated in any patient, and no adjuvant chemotherapy was administered. Five patients (4%) received adjuvant tamoxifen.
All specimen slides were reviewed for this study by one author (N.S.G.) without knowledge of the clinical outcome. Specimens were processed in a similar manner throughout the study period. All specimens underwent postexcisional radiography, and these were compared with the preoperative mammogram. All of the tissue surrounding the tip of the needle-localization wire was embedded and examined. When the excision was performed for abnormal mammographic calcifications, the presence of microcalcifications was confirmed in the biopsy. If microcalcifications were not found in the initial set of slides, then the blocks were step sectioned, and, if necessary, they were X-rayed to identify the appropriate block to step section.
For complete pathologic review, the following information and pathologic features were recorded for both the initial biopsy specimen and the reexcision specimen: 1) total number of slides examined from each specimen; 2) whether the specimen was embedded partially or completely embedded (using the aforementioned processing technique, 39% of entire initial biopsy specimens were completely embedded, i.e., the entire biopsy specimen embedded, not just the tissue surrounding the tip of the localization wire); 3) maximum specimen dimensions; 4) pattern of DCIS involvement, categorized as tumorous, tumorous with dispersion, or predominantly dispersed (DCIS formed a tumorous lesion when the DCIS ducts were closely grouped, such that a tumor size could be measured; DCIS had a dispersed pattern when one or more DCIS ducts were separated by normal breast parenchyma in a random or discontinuous pattern [sometimes on several slides], such that a tumor size could not be measured); 5) maximum DCIS tumor dimensions, measured from that portion of DCIS that had a tumorous pattern (if a tumorous lesion was contiguous across several slides, then this tumor dimension was calculated using an estimated maximal tumor block thickness of 0.4 cm and multiplying by the number of slides containing the tumor; the dispersed pattern associated with tumorous DCIS was not measured); 6) predominant histologic subtype, categorized as comedo, cribriform, papillary, micropapillary, solid, clinging, or cystic;9 7) minor histologic subtype; 8) proportion of DCIS profiles with central necrosis divided into quartiles (none, 1–25%, 26–50%, 51–75%, or ≥76%); 9) predominant nuclear grade; 10) highest nuclear grade; 11) the Van Nuys prognostic classification,10 based on highest nuclear grade and the presence or absence of necrosis; 12) margin status, classified as unifocal positive (single DCIS duct transected at the margin), multifocal positive (two or more ducts transected at the margin that were separated by >0.5 cm or on two or more slides), close (DCIS located within 0.2 cm of the inked margin edge but not transected), negative (no DCIS within 0.2 cm of the inked margin), or uncertain (specimen was not inked or was fragmented, such that the specimen margin could not be determined); 13) number of terminal duct lobular units (TDLUs) or interlobular duct profiles involved by cancerization of lobules (COL) that were within 0.5 cm (half of a low-power field with a ×4 objective and a ×10 ocular) of the margin (COL was defined using the criteria of Azzopardi11 and Fechner12); 14) number of TDLUs involved by atypical ductal hyperplasia (ADH) that were within 0.5 cm of the margin ADH was defined using the criteria of Lennington et al.13 as “ADH associated with DCIS”); 15) number of slides with any COL or ADH within 0.5 cm of the margin; 16) number of DCIS ducts within 0.5 cm of the margin; 17) number of slides with DCIS within 0.5 cm of the margin; 18) number of slides with any DCIS; 19) presence or absence of any focus of DCIS with three or more mitoses in one high-power field ×40 ocular); 20) presence or absence of any microcalcifications within DCIS ducts; 21) desmoplastic-type stromal fibrosis around DCIS ducts, categorized as absent-slight or moderate-marked; and 22) lymphocytic infiltration around DCIS ducts, categorized as absent-slight or moderate-marked.
Mammographic findings were recorded from a retrospective review of reports on patient charts. Results were categorized as follows: 1) mass alone (with no calcifications), 2) calcifications alone, and 3) both a mass and calcifications. The vast majority of patients underwent standard two-view film screening mammography (craniocaudal and mediolateral oblique) with magnification views of suspicious calcifications or masses.
After completion of breast irradiation, patients were evaluated every 3 months for the initial 2 years of follow-up and at 6-month intervals thereafter. Patients frequently alternated follow-up visits between their surgeon and radiation oncologist. Mammograms were performed 6 months after completion of treatment and annually thereafter, unless a given mammographic finding warranted earlier follow-up.
An ipsilateral failure was defined as the reappearance of disease in the treated breast before or at the time of metastases. Ipsilateral failures were classified by clinical location in relation to the initial boost volume according to the criteria described by Recht et al.14 A true recurrence/marginal miss (TR/MM) was defined as a recurrence within or immediately adjacent to the boost volume (or the primary tumor site in patients not boosted). An elsewhere failure was defined as an ipsilateral breast recurrence several centimeters from the primary site and was generally felt to be a new primary lesion or multicentric disease. Contralateral breast failure was defined as the subsequent development of breast carcinoma in the opposite, untreated breast. Overall survival reflects all deaths, disease-related or otherwise. Cause-specific survival was based on deaths attributed to breast carcinoma. The four patients with previous or simultaneous contralateral invasive breast carcinoma were excluded from all contralateral breast failure and survival analyses. However, these patients were included in all ipsilateral breast failure, TR/MM, and elsewhere failure analyses. Those two patients with previous or simultaneous contralateral DCIS also were excluded from contralateral breast failure analyses, but they were included in survival analyses.
Actuarial results for ipsilateral breast failure, contralateral breast failure, disease free survival, overall survival, and cause specific survival were calculated using the Kaplan–Meier method.15 The association of clinical, pathologic, and treatment-related variables with any given event was analyzed using the Fisher exact test (two-tailed) for categoric variables and logistic regression for continuous variables. The statistical significance of differences between actuarial curves was calculated with the log-rank test.16 Multivariate analysis was performed using the Cox proportional hazards model.17 A P value < 0.05 was considered statistically significant. Because ADH and COL are mutually exclusive entities, only lower grade cases were considered when analyzing the influence of ADH, and only higher grade cases were considered when analyzing the influence of COL. For calculations of survival, the number of patients was used instead of the number of cases. All time intervals were calculated from the date of diagnosis. Statistical analysis was performed on SAS software (version 6.12; SAS Institute, Inc., Cary, NC).
The median follow-up for all patients was 7.0 years (range, 1.3–14.2 years). One hundred nine patients (83%) have been followed for a minimum of 5 years, and 29 patients (22%) have been followed for >10 years.
Table 1 lists various clinical, pathologic, and treatment-related characteristics of the entire study population. Of those 76 lesions (58%) that had a measurable tumorous component, 59 (78%) were <1.0 cm in greatest dimension. Forty-two percent of cases had a dispersed pattern, such that tumor size could not be measured accurately. A negative final margin status was confirmed in 67% of cases. Only 2% of specimens were not inked or were fragmented, such that the final margin status could not be determined.
Table 1. Clinical, Pathologic, and Treatment-Related Characteristics of All 132 Cases
No. of cases
DCIS: ductal carcinoma in situ; Gy: Gray; NAC: nipple-areolar complex.
Age at diagnosis (yrs)
Mass with calcifications
Total no. of slides with DCIS
Maximum tumor dimension (cm)
Predominant histologic subtype
Predominant nuclear grade
Final margin status
Close (≤2 mm)
Total tumor bed dose (Gy)
Tumor bed alone
Tumor bed plus NAC
Table 2 lists 5-year and 10-year actuarial outcome data for ipsilateral breast failure, TR/MM, elsewhere failure, contralateral breast carcinoma, disease free survival, overall survival, and cause specific survival for the entire patient population. A total of 13 patients recurred in the ipsilateral breast for 5-year and 10-year actuarial rates of 8.9% and 10.3%, respectively. In nine cases (69%), the recurrence was classified as a TR/MM, and, in four cases (31%), the recurrence was noted elsewhere in the breast (i.e., a new primary lesion). The 5-year and 10-year actuarial rates of ipsilateral failure defined as a TR/MM were 6.7% and 8.1%, respectively. The mean interval from the time of diagnosis to TR/MM was 3.8 years (range, 2.1–6.8 years; median, 3.3 years). For recurrences that occurred elsewhere in the breast, the 5-year and 10-year actuarial rate was 2.4%. The mean interval from diagnosis to elsewhere failure was 4.8 years (range, 2.5–10.3 years; median, 3.2 years). The mean interval for all ipsilateral breast failures was 4.1 years (median: 3.3 years). Eleven ipsilateral breast failures (85%) occurred within 5 years. One TR/MM occurred at 6.8 years, and 1 elsewhere failure occurred at 10.3 years.
Table 2. Five-Year and 10-Year Actuarial Results for All 132 Cases
At 5 years (%)
At 10 years (%)
TR/MM: true recurrences or marginal misses.
Ipsilateral breast failure
Disease free survival
Cause specific survival
Contralateral breast failure
Of the 13 ipsilateral recurrences, the histology at the time of recurrence was invasive carcinoma (with or without associated DCIS) in ten cases (77%), and pure DCIS in three cases (23%). Of the nine recurrences classified as TR/MM, six were invasive, and three were pure DCIS. All four elsewhere failures were invasive. For the ten invasive recurrences, the mean interval to recurrence was 4.2 years (range, 2.1–10.3 years; median, 3.2 years). The three noninvasive recurrences appeared 3.3, 3.9, and 4.4 years after diagnosis. Eleven recurrences (85%) were detected by mammography. One was detected by self-examination, and one was detected by physician examination.
Salvage treatment at the time of ipsilateral breast failure consisted of mastectomy in all 13 cases. Nine of these patients underwent an axillary lymph node dissection. In all nine patients, all lymph nodes were free of metastases. No patient with an invasive recurrence was treated with chemotherapy, and no patient developed a chest wall recurrence. Twelve patients are without evidence of disease at a median of 3.3 years (range, 0.2–7.1 years) after salvage treatment. The 5-year actuarial disease free survival rate after an ipsilateral breast recurrence was 89%. One patient failed distantly 3.0 years after salvage mastectomy and eventually died of disease. For those patients with ipsilateral breast failure, the survival status at last follow-up was alive with no evidence of disease in 11 patients (85%), dead with no evidence of disease in 1 patient (8%), and dead of disease in 1 patient (8%).
Contralateral Breast Failure
Ten patients were diagnosed with subsequent contralateral breast carcinoma after conservative treatment for DCIS. Two of these patients were diagnosed with subsequent contralateral DCIS (one of whom underwent mastectomy), and eight were diagnosed with contralateral invasive carcinoma. The 5-year and 10-year actuarial rates for the development of subsequent contralateral breast carcinoma after diagnosis of DCIS were 5.8% and 9.7%, respectively. The median interval between ipsilateral DCIS and subsequent contralateral breast carcinoma was 3.4 years (range, 0.8–10.8 years). None of these ten patients have recurred in either breast.
An analysis was performed for a number of potential clinical, pathologic, and treatment-related factors relative to ipsilateral breast failure, TR/MM, and elsewhere failure. The factors analyzed included patient age at the time of diagnosis, mammographic findings (calcifications, mass, or both), location of the primary tumor (central, outer quadrant, or inner quadrant), reexcision status (no reexcision, negative reexcision, or positive reexcision), and use of postexcisional mammography to verify complete excision of calcifications. Multiple pathologic factors were analyzed after a complete pathologic review of all cases, as discussed above. Additional treatment-related factors analyzed included whole-breast dose, tumor bed dose, interval from diagnosis to start of RT, interval from diagnosis to completion of RT, boost type (no boost, interstitial implant, electrons, or photons), boost electron energy, and boost volume.
Univariate analysis of selected characteristics is summarized in Table 3, and Table 4 presents the 5-year and 10-year actuarial rates of ipsilateral breast failure and TR/MM with the population separated by these characteristics. Patients who were diagnosed at a younger age demonstrated a significantly higher risk of TR/MM when analyzed either as a continuous variable or with a cut-off value of age 45 years. The number of slides containing any DCIS significantly predicted for both TR/MM and ipsilateral breast failure when the population was divided into cases with <5 slides versus ≥6 slides containing DCIS. Those cases without calcifications on pathologic review demonstrated a higher TR/MM and ipsilateral breast failure rate compared with cases in which pathologic calcifications were identified. Although two of the three cases with uncertain margins developed a TR/MM, margin status was not associated significantly with ipsilateral breast failure or TR/MM when either excluding these cases or when combining them with patients who had close or positive margins. However, when ≥5 foci of COL were identified within 5 mm of the margin, a higher rate of TR/MM and ipsilateral breast failure was noted. No factor was associated significantly with recurrence elsewhere within the breast.
Table 3. Univariate Analysis (P values) of Selected Characteristicsa
Ipsilateral breast failure
TR/MM: true recurrences or marginal misses; DCIS: ductal carcinoma in situ; ADH: atypical ductal hyperplasia; COL: cancerization of lobules.
Continuous variables were analyzed with logistic regression; categoric variables were analyzed with the Fisher exact test.
A Cox proportional hazards model was constructed utilizing variables that were significant on univariate analysis as well as selected variables that have demonstrated association with recurrence in other series (Table 5). Only the absence of pathologic calcifications significantly predicted for ipsilateral breast failure. Cases without calcifications were 4.55 times as likely to experience an ipsilateral breast failure compared with cases that had calcifications. When specifically analyzing for TR/MM, younger age at diagnosis, number of slides with DCIS, number of DCIS and COL foci within 5 mm of the margin, and the absence of pathologic calcifications demonstrated a statistically significant association. For each 1-year incremental increase in patient age, the relative risk of TR/MM decreased by 11%. For each additional slide containing DCIS, the relative risk of TR/MM increased by a factor of 1.55. The number of DCIS slides was examined further to determine the correlation of this variable with tumor size (Fig. 1). The number of DCIS slides was associated significantly with maximum tumor dimension (P < 0.001; linear regression) and tumor volume (P < 0.001) in those 76 cases with a tumorous pattern (in which tumor dimensions could be measured accurately). On multivariate analysis, the number of slides with DCIS (P = 0.007) was a better predictor for a TR/MM than the maximum tumor dimension (P = 0.15) and tumor volume (analysis not shown). Close or positive margin status did not predict significantly for either TR/MM (P = 0.14) or ipsilateral breast failure (P = 0.19). However, for each additional DCIS or COL focus (sum of DCIS and COL foci) located within the 5-mm rim of tissue surrounding the margin, the relative risk of TR/MM increased by 8% (P = 0.02).
Table 5. Cox Proportional Hazards Model
Ipsilateral breast failure
DCIS: ductal carcinoma in situ; COL: cancerization of lobules.
No pathologic calcifications
Age at diagnosis (yrs)
No. of slides with DCIS
No. of DCIS and COL foci ≤ 5 mm from margin
Close or positive margin status (vs. negative)
Maximum tumor dimension (mm)
Predominant nuclear grade
Comedo (vs. noncomedo)
In the current study, long term treatment outcome was analyzed in a series of 132 consecutive cases from patients with mammographically detected DCIS who were treated with BCT at a single institution. After complete pathologic review of all cases by one pathologist, histologic variables associated with local recurrence included a larger number of slides containing DCIS, the lack of pathologic calcifications, and greater numbers of DCIS and COL foci within 5 mm of the surgical margin. In addition, on multivariate analysis, young patient age was associated independently with recurrence of the index lesion in the breast. These findings suggest that careful attention to the adequacy of tumor extirpation can result in significant improvements in local control. In addition, young patient age appears to be associated with a higher risk of recurrence in the breast and is one additional variable that should be taken into consideration when determining the acceptability of a patient for BCT.
Several recent studies have attempted to define factors associated with local recurrence in patients with DCIS who receive BCT. The impetus for these investigations is the belief that radiographically detected DCIS may have a natural history that is distinct from that of clinically detected carcinomas and, consequently, may need to be treated differently. Unfortunately, most studies addressing mammographically detected cases consist of small, retrospective reports with short term follow-up and incomplete pathologic reviews. In addition, when analyzing factors associated with local recurrence, all ipsilateral failures (including failures that may be unrelated to treatment of the primary lesion) are included. Consequently, no consistent results have been obtained to establish firm guidelines for treatment in these patients. Those studies that have reported data for mammographically detected cases are shown in Table 6. Only two studies have examined exclusively mammographically detected DCIS.3, 4
Table 6. Studies Reporting Mammographically Detected Ductal Carcinoma In Situ Treated with Lumpectomy Followed by Radiotherapy
No. of patients
Median follow-up (yrs)
Ipsilateral breast failure (%)
Factors significantly associated with any breast recurrence
Factors significantly associated with TR/MM
TR/MM: true recurrences or marginal misses; DCIS: ductal carcinoma in situ; COL: cancerization of lobules; NR: not reported; NSABP: National Surgical Adjuvant Breast and Bowel Project; LHMC: Lahey Hitchcock Medical Center; JCRT: Joint Center for Radiation Therapy; MSKCC: Memorial Sloan-Kettering Cancer Center.
Crude failure during follow-up period.
Exclusive analysis of mammographically detected cases for factors associated with recurrence was not reported.
No pathologic calcifications
No pathologic calcifications, younger age, increased no. of slides with DCIS, increased no. of DCIS/COL foci near margin
The largest study reporting results with mammographically detected DCIS comes from the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-17 trial of 814 patients who received BCT.18 Six hundred fifty-four of these cases (80%) were detected by mammography only. Three hundred thirty of the mammographically detected cases were randomized to treatment with lumpectomy followed by RT, and 324 received lumpectomy alone. With a mean follow-up of 7.5 years, 33 patients (10%) in the arm receiving postoperative RT developed an ipsilateral breast recurrence, with 84% of recurrences located in the same quadrant as the primary lesion. In the entire study population, no mammographic or clinical characteristics were associated with ipsilateral recurrence. In a previous analysis with complete pathologic review of 573 patients (70%) that included cases that were not mammographically detected, moderate or marked comedo necrosis and uncertain or involved surgical margins were the only factors significantly associated with recurrence in the ipsilateral breast on multivariate analysis.19 On univariate analysis, nuclear grade and histologic subtype also were associated significantly with ipsilateral recurrence. An analysis of exclusively mammographically detected cases has not yet been reported.
Additional studies on mammographically detected DCIS treated with BCT have been reported by Fowble et al.3 and Solin et al.4 Each of those studies analyzed 110 patients who underwent lumpectomy followed by RT. No risk factors were associated significantly with recurrence in either study.
The results from our current analysis contribute further data establishing the necessity of excising adequately all DCIS prior to RT and that greater tumor volumes (as represented by the number of slides with DCIS) are associated with higher disease recurrence rates. Similar to NSABP B-17, it appears that the extent of resection is critical. However, as in our previous studies,5, 20 margin status alone did not correlate with local recurrence on multivariate analysis, regardless of how it was defined (eg., ≤2 mm, ≤5 mm). Even when examining only those 16 cases with strictly positive margins, the 10-year ipsilateral breast failure rate was 6.3% versus 8.1% for those with negative margins. A more appropriate surrogate for the adequacy of excision may be the volume of DCIS and COL in close proximity to the margin rather than the mere presence or absence of any amount of carcinoma within a defined, one-dimensional distance. This is evidenced by the fact that, for each additional focus of DCIS or COL within 5 mm of the margin, the relative risk of TR/MM failure increased by 8%. Collectively, these data imply that the quantity of DCIS in close proximity to the margin may be a better predictor of a greater volume of disease remaining in unexcised breast tissue than margin status alone.
Recent additional work from our institution supports the above hypothesis.21 We studied the factors that predicted for significant residual disease remaining in the breast at reexcision after an initial excisional biopsy in 98 patients with DCIS who received BCT. Margin status alone did not correlate with significant quantities of DCIS remaining in the breast. Rather, an increasing number of slides with DCIS in the initial biopsy (even with a negative margin) as well as an increasing amount of DCIS or COL within 5 mm of the margin were more accurate predictors of significant residual DCIS. No other variables predicted for significant residual disease on multivariate analysis, implying that DCIS volume was the most appropriate prognostic factor.
Though tumor size was not found to correlate with local recurrence in the current study, we believe that the number of slides with DCIS is a surrogate for tumor volume and may be a more reliable indicator of tumor size than ruler-based measurements. As others have pointed out, DCIS does not lend itself routinely to being measured by a ruler unless elaborate pathologic handling of tissue specimens is performed.22 Consequently, tumor size is not a practical or reproducible variable for use in guiding therapy. Again, our data support this hypothesis (Fig. 1). The number of DCIS slides was associate significantly with maximum tumor dimension and tumor volume in those 76 cases with a tumorous pattern (in which tumor dimensions could be measured). Additional studies investigating the above hypothesis need to be conducted, however, to establish the reproducibility of this factor in predicting treatment outcome.
In the current analysis, we chose not to use the Van Nuys Prognostic Index (VNPI),23 because tumor size could be measured accurately in only 76 cases (58%). In addition, because this analysis included only mammographically detected cases, only two cases would have received a VNPI of 8–9 due to smaller tumor size, further reducing the predictive power of the VNPI in this subset of patients.
The current results also suggest that there may be a potential significant methodological flaw in many of the studies of DCIS relating to the definition of recurrent carcinoma. The majority of locally recurrent carcinomas are neoplasms that arise in or near the bed of the initial DCIS lesion (TR/MM neoplasms), suggesting that “recurrences” are a persistence of the initial DCIS lesion.24–35 “Recurrences” that develop elsewhere in the breast may represent de novo neoplasms. However, for the purpose of analysis, most studies have defined a recurrent carcinoma as a subsequent carcinoma arising anywhere in the breast.19, 36–50 Unfortunately, statistical analyses of pathologic features associated with locally recurrent carcinomas have been based on the combination of both TR/MM neoplasms and recurrent carcinomas located elsewhere within the breast (elsewhere neoplasms). Because recurrences are uncommon, the lumping together of two potentially biologically distinct types of recurrences into one patient group may alter the findings of these studies significantly. In our previous study, all TR/MM neoplasms had the same nuclear grade and morphology as the initial DCIS lesion, but two of the three elsewhere failures had a different nuclear grade or cytology than the initial DCIS lesion, suggesting that they were different neoplasms unrelated to treatment of the index lesion.20 Also, the greatest association of risk factors with outcome was with TR/MM failures. It is interesting that, in studies with a high percentage of recurrences defined as TR/MM, factors associated with adequacy of excision of the initial DCIS lesion (such as margin status) generally are predictive of outcome.51 This is demonstrated clearly by the NSABP B-17 study, in which 87% of recurrences were defined as TR/MM, and margin status was associated with recurrence in the breast.19 Conversely, in studies in which a large percentage of recurrences were defined as elsewhere failures, no variables (particularly treatment-related) correlated with outcome, probably reflecting the dilutional effect of these biologically unrelated carcinomas on the statistical power to identify an association.41 Collectively, these data imply that the location of the recurrent neoplasm must be taken into consideration when analyzing factors associated with recurrence of the index lesion.
One additional factor that was associated independently with local recurrence in our study was the absence of calcifications on histologic review. Although mammographic calcifications have been correlated with treatment outcome in other studies,52 we know of only one other study that has demonstrated an association between recurrence and pathologic calcifications.53 In a multivariate analysis of 144 patients who underwent lumpectomy and postoperative RT at the Institut Curie, the absence of calcifications was associated significantly with ipsilateral recurrence. Those patients without calcifications had a 5-fold increase in the rate of ipsilateral recurrence compared with those with pathologic calcifications (P = 0.003). These results are strikingly similar to our data. Because the presence of pathologic calcifications was associated significantly with calcifications on mammography (P = 0.03) in our series, tumors without these calcifications potentially could be underestimated radiographically. However, unlike the data from Institut Curie, our analysis did not reveal significantly smaller excision specimens for those patients without calcifications (mean, 161 cm3 and 113 cm3 for those with and without calcifications, respectively; P = 0.38). In addition, the absence of calcifications was not associated with the initial excision volume (P = 0.43), reexcision volume (P = 0.36), number of slides examined in the initial excision (P = 0.74), number of slides examined in the reexcision (P = 0.21), number of slides with DCIS in the initial excision (P = 0.55), number of slides with DCIS in the reexcision (P = 0.39), tumor size (P = 0.50), margin status (P = 0.73), or amount of DCIS or COL near the margin (P = 0.35). At this time, the factors responsible for the higher recurrence rate in patients without calcifications remain uncertain.
Finally, the issue of young patient age needs to be taken into consideration when determining the acceptability of a patient with DCIS for BCT. Younger patients had a significantly higher relapse rate at 10 years (23.4% vs. 4.3%) that was associated independently with outcome on multivariate analysis. Although we continue to offer younger patients with DCIS the option of breast conservation, further studies need to be performed to establish the factor(s) responsible for this higher relapse rate and techniques that may be useful in optimizing outcome in these patients.
The findings from this retrospective analysis of mammographically detected DCIS patients who were treated with BCT suggest that adequate excision of all DCIS prior to RT can improve local control in the breast. However, margin status alone may not predict complete tumor extirpation adequately. The volume of DCIS and COL within 5 mm of the margin appears to be a more reliable surrogate for the adequacy of excision. In addition, young patient age and the absence of pathologic calcifications are independent risk factors for the development of local disease recurrence.
The authors thank Ms. Vicky Dykes for her assistance in the preparation of the article.