Less common variants of DCIS are also recognized. Some of these variants are defined by the cytologic characteristics of the cells comprising the lesion. For example, certain DCIS lesions are composed of cells that have apocrine features (apocrine DCIS).Others are comprised of cells that have a signet-ring configuration (intraductal signet-ring cell carcinoma). Still other variants of DCIS are composed of cells that exhibit neuroendocrine differentiation (endocrine DCIS and solid papillary carcinoma). Some of the less frequent DCIS types are defined by their architectural features. Cystic hypersecretory carcinoma is characterized by neoplastic cells that line dilatated spaces filled with eosinophilic secretion, resembling the colloid seen in thyroid follicles. The term clinging carcinoma has been used to describe two different types of lesions. One is characterized by a single layer of cytologically malignant cells lining the involved spaces; general agreement exists that such lesions should be categorized as DCIS. In the other variant, ductal-lobular spaces are lined by a layer of cells with low-grade nuclear features. Whereas some authors consider such lesions to represent variants of DCIS, others regard them as atypical ductal hyperplasia.

Some authors believe it useful to subdivide DCIS into two subgroups: the comedo type and the noncomedo type (which encompasses the other variants). This subdivision is based on the observations that the comedo type usually appears more malignant cytologically and is more often associated with invasion than are the other DCIS types. Another difference reported for these two groups is in regard to the relationship between the extent of microcalcifications on the mammogram and the histologic extent of the lesion. When standard two-view mammography is used, the histologic extent of comedo-type DCIS is highly correlated with the extent of the calcifications on the mammogram in most cases. In contrast, the extent of the calcifications associated with noncomedo DCIS on mammography frequently underestimates the pathologic extent of the lesion. A subsequent study, however, has indicated that when standard mammographic images are supplemented with magnification views, a much better correlation is found between the extent of mammographic calcifications and the histologic extent of the lesion, even for DCIS lesions of the noncomedo type.

These classification systems based primarily on architecture have a number of important limitations. First, DCIS lesions may be difficult to classify using architecture alone, because many display a mixture of patterns, particularly when the lesion is large.For example, in one study of 100 consecutive cases of DCIS reviewed in a consultation practice, 23 of 76 noncomedo lesions (30%) showed a mixture of histologic patterns, the most common being cribriform and micropapillary. Among 24 comedo lesions, 10 (42%) also contained areas with noncomedo patterns.32 In another study of 121 cases of DCIS, mixtures of architectural patterns were identified in 62% of cases. Second, although DCIS lesions with a comedo pattern are most often composed of malignant-appearing cells with high-grade nuclei, whereas noncomedo lesions are usually composed of cells with low-grade to intermediate-grade nuclei, the correlation between architecture and nuclear grade is far from absolute. For example, some DCIS lesions with small, uniform, low-grade nuclei display central comedo-type necrosis in the involved spaces  whereas others with cribriform, micropapillary, or papillary architectural patterns may be composed of cytologically malignant cells exhibiting large, high-grade nuclei. In addition, interobserver reproducibility in the categorization of DCIS lesions by architectural pattern is poor, even if a simplified, dichotomous comedo/noncomedo classification scheme is used.Finally, although the architectural classification scheme was perfectly acceptable in an era in which all cases of DCIS were treated by mastectomy, a pressing clinical need now exists to develop a classification system that has prognostic significance for patients considered for treatment with breast conservation.

A number of alternative classification schemes for DCIS have been proposed in an attempt to overcome the limitations of the traditional, primarily architectural classification system. These newer classification systems stratify DCIS lesions largely on the basis of nuclear grade or necrosis, with architectural pattern given secondary or no consideration . Lagios et al. were the first to propose a system based primarily on nuclear grade and necrosis, rather than on architecture. A modification of this system recognizes three major categories of DCIS: high, intermediate, and low grade. Investigators in Nottingham, England, have developed a classification system based primarily on the presence or absence of necrosis.This group divides DCIS into three categories: pure comedo (lesions in which involved spaces show centrally necrotic debris surrounded by large, pleomorphic tumor cells in solid masses); DCIS with necrosis, also called nonpure comedo (lesions with necrotic neoplastic cells but with a cribriform or micropapillary pattern); and DCIS without necrosis (lesions with a cribriform, papillary, micropapillary, or solid pattern and no necrosis). The classification scheme proposed by Silverstein  is essentially a modification of the Nottingham system in which DCIS lesions are classified based on nuclear grade as either high-grade or non–high-grade. The non–high-grade lesions are further stratified by the presence or absence of comedo-type necrosis. Thus, this is a three-tiered system in which DCIS is classified as either high-grade, non–high-grade with necrosis, or non–high-grade without necrosis. A group of European pathologists have proposed classifying DCIS as well differentiated, intermediately differentiated, or poorly differentiated, based primarily on cytonuclear differentiation and cell polarization.Pathologists associated with the United Kingdom National Breast Cancer Screening Program use a classification scheme for DCIS based solely on nuclear grade, recognizing high-grade, intermediate-grade, and low-grade types. Other classification systems have been proposed by other authors as well.

To attain widespread clinical use, a classification system not only must be clinically relevant but also must be able to be applied reliably by different observers. The clinical importance of classifying DCIS using these various approaches is discussed below. Although some authors have claimed that their systems are or should be easy to use, few studies have been conducted to assess interobserver agreement in the classification of DCIS using these newer categorization schemes. In one study, two pathologists classified the DCIS component associated with 180 invasive cancers using six different classification systems. These investigators found that the highest level of agreement was obtained using the Silverstein system.Another study found a 94% level of agreement among six observers using the Lagios classification system.More sobering results were reported by the European Commission Working Group on Breast Screening Pathology. In that study, 33 cases of DCIS were categorized by 23 pathologists using five classification systems. The level of interobserver agreement, as defined by kappa statistics, was only fair to moderate for each of the classification schemes evaluated.

In 1997, a consensus conference was convened in an attempt to reach agreement on the classification of DCIS.Although the panel did not endorse any one system of classification, agreement was reached that certain features be routinely documented in pathology reports of DCIS lesions. These include nuclear grade (low, intermediate, or high), the presence of necrosis (comedo or punctate), cell polarization, and architectural pattern. In fact, if these individual features are recorded, sufficient information is then available to permit the categorization of a DCIS lesion according to virtually all of the newly proposed classification schemes.

Biological Markers

The study of tumor markers in DCIS lesions to provide a better understanding of the biology of these lesions and to aid in their classification has provoked considerable interest. The results of studies of these markers are somewhat difficult to compare due to differences in the classification used for the DCIS lesions, patient populations, and methodology; however, a number of trends have emerged. These studies have generally shown that lesions demonstrating a comedo pattern or high nuclear grade exhibit a profile of biological markers that has been associated with aggressive clinical behavior in invasive cancers more often than do noncomedo or low-nuclear-grade lesions. For example, comedo or high-grade lesions are more likely than noncomedo or low-grade lesions to lack estrogen and progesterone receptors, to have a high proliferative rate, and to exhibit aneuploidy,53 overexpression of the HER-2/neu (c-erb-b2) oncogene,mutations of the p53 tumor-suppressor gene with accumulation of its protein product, and angiogenesis in the surrounding stroma.

In addition, genetic analyses have indicated that loss of heterozygosity at various chromosomal loci may differ according to DCIS pattern and grade.

Distribution of Tumor in the Breast and Axillary Lymph Node Involvement

The distribution of tumor in the breast, the incidence of unsuspected invasive carcinoma, and the incidence of axillary lymph node metastases are all important considerations in selecting appropriate therapy for patients with DCIS.

The reported incidence of multicentricity in mastectomy specimens from patients with DCIS varies considerably and has ranged from 0% to 47%. A number of factors have contributed to this variability, including differences in the definition of multicentricity and differences in the methods and extent of specimen sampling. Most authors define multicentricity as the presence of DCIS foci in breast quadrants other than the one harboring the index lesion (in contrast to multifocality, which denotes foci of DCIS in the same quadrant as the index lesion). Others define foci as multicentric if they are a specified distance from the index lesion (e.g., 5 cm), regardless of the quadrant. These studies of multicentricity were conducted before the widespread use of screening mammography, and these data probably cannot be extrapolated to the small (often less than 1 cm), mammographically detected lesions commonly seen today. In these studies, the frequency of multicentricity appeared to be related to the size of the index lesion. In one study, multicentricity was much more common in DCIS lesions larger than 2.5 cm (13 of 25 cases, 52%) than in smaller lesions (4 of 29, 14%).In another study, the frequency of multicentricity also correlated with the size of the lesion as determined by the number of involved ducts in the index lesion.28 These investigators also noted a higher frequency of multicentricity in micropapillary lesions (8 of 10 cases, 80%) than in other types of DCIS (16 of 45, 36%). A similar association between micropapillary DCIS and frequent multiple-quadrant involvement has also been recognized by others.40 Because of the sampling methods used, however, it is not possible to determine if the foci of DCIS characterized as multicentric in these studies were truly independent lesions or if they represented tumor that was, in fact, contiguous with the index lesion.

More recent studies suggest that, in most cases, true multicentricity in DCIS is rare. Holland and Hendriks studied 119 mastectomy specimens containing DCIS by a subgross pathologic-mammographic technique.In all but one case, the tumor was confined to a single “segment” of the breast. Clear-cut multicentric distribution (defined in this study as foci of DCIS separated by 4 cm or more of uninvolved breast tissue) was seen in only one patient. Faverly et al., using stereomicroscopic three-dimensional analysis to define the growth pattern of DCIS within the mammary duct system, studied 60 mastectomy specimens containing DCIS. They found that within the segment of breast involved by DCIS, growth was continuous in some cases and discontinuous in others. Overall, 50% of cases showed a continuous growth pattern and 50% showed a discontinuous pattern, characterized by uninvolved breast tissue between foci of DCIS (“gaps”). In most instances, these gaps were small (less than 5 mm in 82% of cases), and the likelihood of finding such gaps was related to the histologic type of the lesion. Whereas 90% of the cases of poorly differentiated DCIS grew in a continuous manner without gaps, only 30% of well-differentiated lesions and 45% of intermediately differentiated lesions were continuous. The findings in these two studies indicate that, in most cases, DCIS involves the breast in a segmental distribution, and truly multicentric disease is uncommon. In some cases, however, the segment involved by DCIS may be quite large. For example, in the study of Holland and Hendriks, although 86% of the DCIS lesions were nonpalpable and were detected mammographically, 46% were larger than 3 cm. One study of clonality in DCIS supports the contention that most DCIS is unifocal, at least with regard to comedo lesions. In that study, clonality was assessed in widely separated sites of comedo-type DCIS in the same breast. Each of these widely separated sites was found to be monoclonal, and each showed inactivation of the same X chromosome–linked phosphoglycerokinase allele, suggesting an origin from the same clone.

The incidence of nipple involvement in patients with DCIS has been evaluated in a few studies and appears to be related to the method of detection of the lesion. For example, Contesso et al. found nipple involvement in 49% of 117 mastectomy specimens from patients in whom DCIS presented primarily with a palpable mass, nipple discharge, or Paget's disease. In contrast, Lagios et al. found involvement of the nipple in 8 of 40 mastectomy specimens (20%) from patients with DCIS, the majority of whom presented with mammographic calcifications or had DCIS as an incidental finding. Of these eight cases, five were Paget's disease and three had lactiferous duct involvement.

The incidence of occult invasion, either near the primary tumor or in other parts of the breast, has also been examined in mastectomy series. The reported incidence of occult invasion ranges from 0% to 26%. However, these series are difficult to interpret for several reasons. The completeness of initial biopsy varies, which affects the likelihood of finding residual cancer with invasion at mastectomy if the initial biopsy shows only “noninvasive” disease. Also, the extent of sampling of the initial biopsy specimen and of the remainder of the breast also differs substantially from series to series. The likelihood of finding occult invasion appears to be related to the size of the index lesion. In one series, patients with lesions larger than 2.5 cm were more likely to have occult invasion (16 of 55, 29%) than were patients with smaller tumors (1 of 60, 2%). However, all four invasive tumors in the 90 patients with lesions 4.5 cm or smaller were found after inadequate initial excision. The frequency of occult invasion is also related to the method of detection of the DCIS. In one series, invasive cancer was identified in the mastectomy specimen in 6 of 54 patients (11%) with DCIS who presented with a palpable mass, nipple discharge, or Paget's disease and in none of 16 patients who presented with mammographic microcalcifications or in whom DCIS was an incidental finding. In another series, 6 of 41 tumors (15%) from patients that presented with a mass showed occult invasion, compared with only 1 of 21 tumors (5%) detected only on mammography.68 The incidence of occult invasion also appears to be correlated with the histologic type of DCIS and is much more common in comedo lesions. For example, Patchefsky et al. noted microinvasion in 12 of 19 cases (63%) of comedo DCIS and in only 4 of 36 (11%) noncomedo lesions.

Incidence rates reported for axillary nodal involvement in patients given the diagnosis of DCIS range from 0% to 7%,with the higher rates noted in studies performed in the premammography era, when most patients with DCIS presented with a palpable mass. In such cases, invasion is undoubtedly present but is either not recognized by the pathologist or is undetected due to sampling error. Axillary lymph node involvement in patients with DCIS detected by mammography is a rare event. In one series of 189 patients with DCIS, most of whose tumors were detected by mammography alone, none showed metastases on axillary dissection.

In a National Cancer Data Base review of 10,946 patients with DCIS who had an axillary dissection between 1985 and 1991, only 406 (3.6%) were found to have axillary metastases.

Differential Diagnosis

In most instances, the pathologic diagnosis of DCIS is straightforward. However, occasional cases present diagnostic difficulties. At one end of the spectrum, distinguishing low-grade (noncomedo) DCIS from atypical ductal hyperplasia is sometimes difficult. Although a number of authors have published criteria useful in making this distinction, some cases are subject to considerable interobserver variability in diagnosis, even with use of standardized criteria.

At the other end of the spectrum, distinguishing examples of pure DCIS from DCIS with focal stromal invasion (microinvasion) may sometimes be difficult. This distinction is discussed in detail below.

DCIS may also be difficult to distinguish from frankly invasive breast cancer in certain instances, because some breast cancers (e.g., invasive cribriform carcinoma) invade the stroma in rounded nests simulating DCIS.Another diagnostic problem occasionally encountered is distinguishing nests of tumor cells in lymphatic or vascular spaces from DCIS.

Finally, although the distinction between DCIS and lobular carcinoma in situ (LCIS) is usually not difficult to make, areas of overlap exist between these two lesions. DCIS may extend into recognizable lobules, LCIS may involve extralobular ducts,78 and some lesions have cytologic features intermediate between the two disorders.Furthermore, DCIS and LCIS may coexist in the same breast and even in the same ductal-lobular unit.80 For example, in the National Surgical Adjuvant Breast and Bowel Project (NSABP) Protocol B-17 study of patients with DCIS treated with either conservative surgery alone or conservative surgery and radiation therapy, approximately 7% of patients had LCIS in addition to DCIS.

Microinvasive Carcinoma

One of the most important goals in the histologic examination of DCIS lesions is the identification of foci of stromal invasion, because in general the therapeutic algorithm for patients with pure DCIS differs from that for patients with DCIS and associated invasive breast cancer. A frequently encountered problem in the examination of such specimens is the identification of the smallest foci of invasive carcinoma, or microinvasion. Although this diagnosis often appears in surgical pathology reports, this term has not been applied in a consistent, standardized manner, and the histologic diagnosis of microinvasion is not straightforward and is often problematic for the pathologist.

In the 1997 edition of the American Joint Committee on Cancer (AJCC) Cancer Staging Manual,82 microinvasion is defined as “the extension of cancer cells beyond the basement membrane into the adjacent tissues with no focus more than 0.1 cm in greatest dimension.” Lesions that fulfill this definition are staged as T1mic, a subset of T1 breast cancer. The staging manual further states that “when there are multiple foci of microinvasion, the size of only the largest focus is used to classify the microinvasion” and that the size of the individual foci should not be added together. This is the first edition of the AJCC staging manual that recognizes a specific T substage for microinvasion. Unfortunately, widely varying definitions of microinvasion have been used in the past, and some of these definitions differ substantially from that offered in the AJCC staging manual. For example, microinvasion has been variously defined as: (a) DCIS with “evidence of stromal invasion”; (b) “DCIS with limited microscopic stromal invasion below the basement membrane, but not invading more than 10% of the surface of the histologic sections examined”; and (c) “breast cancer cells confined to the duct system of the breast with only a microscopic focus of malignant cells invading beyond the basement membrane of the duct as determined by light microscopy.” This lack of a uniform definition for microinvasion has clearly contributed to the confusion regarding this entity.

The identification of microinvasion in a lesion that is primarily DCIS can be difficult for the pathologist, because a variety of patterns in DCIS may be misconstrued as stromal invasion. Lesions that are commonly mistaken for microinvasion include: (a) DCIS involving lobules (“cancerization of lobules”), (b) branching of ducts, (c) distortion or entrapment of involved ducts or acini by fibrosis, (d) inflammation present in association with and obscuring involved ducts or acini, (e) crush artifact, (f) cautery effect, (g) artifactual displacement of DCIS cells into the surrounding stroma or adipose tissue due to tissue manipulation or a prior needling procedure, and (h) DCIS involving benign sclerosing processes, such as radial scars, complete sclerosing lesions, and sclerosing adenosis.

What, then, are the minimum criteria for identifying bona fide stromal invasion in the setting of DCIS? Remarkably, few guidelines have been published in this regard. Page and Anderson require “more than a single collection of cells outside the lobular unit or immediate periductal area.” Fisher indicates that the suspicious focus should be comprised of a “recognized type of invasive cancer.” Elston and Ellis state that “only when unequivocal invasion is seen outside the specialized lobular stroma should microinvasive carcinoma be diagnosed.” The definition offered by Silver and Tavassoli seems to be less restrictive, because they consider invasive tumor cells singly and in clusters in the periductal stroma to represent microinvasion. In the authors' view, for a diagnosis of unequivocal stromal invasion or microinvasion in the setting of DCIS, the worrisome area should be present clearly beyond the immediate periductal and perilobular region and should consist of a recognized type of invasive cancer. Furthermore, the suspicious area should clearly not be in a benign sclerosing lesion. The clinical significance of a few single tumor cells or tumor cell clusters admixed with inflammatory cells in the immediate periductal region is unclear. In such cases, the authors note the presence of such foci, but indicate uncertainty about their clinical importance. Although the presence of stromal desmoplasia and inflammation should heighten the suspicion of invasion, these phenomena are so often present in association with high-grade DCIS without demonstrable invasion that their presence cannot be depended on to make this distinction.

Another potential problem with the pathologic diagnosis of microinvasion relates to tissue sampling. Previous published studies of microinvasion have generally failed to indicate how much of a given specimen was submitted for microscopic evaluation. Thus, some lesions categorized as microinvasive based on limited tissue sampling could in truth represent frankly invasive carcinomas in which the largest area of invasion was not submitted for histologic evaluation or was not represented on the slides because the cancer was deeper in the blocks. Even when an entire specimen is submitted for histologic evaluation, only a fraction of the tissue is ultimately examined microscopically. For example, if a 6-cm excision specimen is sectioned grossly at 3-mm intervals (producing 20 slices), each of these slices is embedded in a separate paraffin block, and one 5-µm section is cut from each block, less than 1% of the entire specimen will be examined microscopically.

Given the problems with the definition and pathologic diagnosis of microinvasion, the controversy surrounding the clinical significance of this lesion should not be surprising. The reported incidence of axillary lymph node involvement in patients given the diagnosis of microinvasion ranges from 0% to 20%

Ideally, the term microinvasion should be used with regard to the breast in the same way it is used with regard to the cervix—that is, to identify those invasive lesions of limited extent that have virtually no risk of metastasis. Unfortunately, the available data are inadequate to permit the reproducible identification of such a subset due to differences among studies with regard to the definition of microinvasion, variations in the extent of tissue sampling, small patient numbers, and limited follow-up. Additional clinicopathologic studies using a standardized definition of microinvasion are clearly needed to address this important question. Although the definition of microinvasion in the current edition of the AJCC Cancer Staging Manual may ultimately be modified, it represents an important step toward standardization, and its use in both clinical research and clinical practice should be encouraged.
Natural History

The major issue in the management of DCIS is the risk of progression to invasive carcinoma. Few clinically relevant data are available to address this question, primarily because DCIS has traditionally been treated with mastectomy. In addition, most DCIS cases treated in the past for which long-term follow-up is available were gross DCIS, a form that may not be equivalent to the mammographic DCIS more commonly seen today.

Long-term follow-up data are available for several small series of women found to have DCIS on review of biopsy specimens that were originally classified as benign. No attempt was made to assess margin status in these studies, lesion size was unknown, and the completeness of excision remains uncertain. Page  identified 25 such cases in a review of 11,760 breast biopsies. Invasive carcinoma developed in seven women (28%) at intervals of 3 to 10 years (mean of 6.1 years) postbiopsy. This incidence of carcinoma represents a relative risk of 11 compared with that of age-matched controls from the Third National Cancer Survey for white women in Atlanta. An update of this series with follow-up extended to 24 years demonstrated that the relative risk of carcinoma remained constant. In a similar study, Rosen  described 30 women with untreated DCIS; complete follow-up was available only for 15. Seven invasive cancers occurred at a mean of 9.7 years after the diagnosis of DCIS—an incidence of 27% if all cases are included or 53% if only patients with complete follow-up are considered. In both the reports, all carcinomas were in the index breast, usually in the vicinity of the biopsy site. In a similar report, Eusebi described 28 cases of DCIS with an 11% incidence of invasive carcinoma at a median follow-up of 16.7 years. Eusebi subsequently reported on 80 cases of DCIS followed for a mean of 17.5 years, only two of which were high grade. Eleven patients developed invasive carcinoma and five had recurrent DCIS, for a total recurrence rate of 20%. The risk of invasive carcinoma was twice that of the general population. In all these studies, most cases included were low-grade, noncomedo lesions, representing one extreme of the histologic spectrum of DCIS.

Further information on the natural history of DCIS can be obtained from autopsy studies. Alpers  assessed a series of 185 randomly selected breasts from 101 women examined by a subgross sampling technique. One or more foci of DCIS were found in 11 cases (6%). This finding was unrelated to age; DCIS was identified in 3 of 56 women (5%) 49 years of age or younger, in 7 of 70 women (10%) between the ages of 50 and 69, and in 1 of 59 women older than 70 years. In a study with similar methodology, Bartow et al.104 examined the breasts of 519 women aged 14 years or older. Only one case of DCIS was identified; five occult invasive carcinomas were found. An autopsy study from Denmark reported by Andersen et al.105 found DCIS in 11 of 86 breasts examined (13%). As a group, these studies indicate that many, but not all, cases of DCIS progress to invasive carcinoma within a woman's lifetime.

Treatment Options

The uncertainty regarding the natural history of DCIS has resulted in a wide range of treatment practices, ranging from excision alone to mastectomy. Making comparisons among reports is difficult because of differences in patient populations, lack of standardization of surgical and radiotherapeutic techniques, and changes in treatment practice over time.

Mastectomy is a curative treatment for approximately 98% of patients with DCIS, whether gross or mammographic.. Patients whose initial biopsies showed DCIS but for whom invasive carcinoma was later identified in the mastectomy specimens were excluded from these reports. This consideration is important when comparing survival after different methods of local therapy.Although many of the initial reports assessing mastectomy for the treatment for DCIS contained small numbers of patients,more recent, larger studies confirm the finding that treatment failure after mastectomy is rare. Recurrences are almost all invasive carcinomas and may present as local failure or distant metastases without evidence of local recurrence. Treatment failure after mastectomy for DCIS may be due to unsampled or unrecognized invasive carcinoma that results in local recurrence or distant metastases, or it may be due to incomplete removal of breast tissue. Residual breast tissue has the potential to develop a new carcinoma that would be manifested as a “local recurrence.” The failure of recurrence rates after mastectomy to increase with longer follow-up intervals, however, suggests that the majority of recurrences are due to undiagnosed invasive carcinoma rather than the malignant transformation of residual breast tissue.

Mastectomy is a highly effective treatment for DCIS, but it is a radical approach to a lesion that may not progress to invasive carcinoma during the patient's lifetime. It seems somewhat paradoxical that a woman with a palpable invasive carcinoma should be able to preserve her breast, whereas the “reward” for screening and early detection of DCIS is a mastectomy. The acceptance of breast-conserving therapy for the treatment of invasive carcinoma has led to its use as a treatment for DCIS. No randomized trial has ever compared the treatment of DCIS by mastectomy to treatment by excision and irradiation. In many cases, the assumption has been made that, because these two treatments result in equivalent survival for patients with invasive carcinoma, the same is true for patients with DCIS. This assumption is flawed, due to the fundamental difference in the risk of metastatic disease for patients with invasive carcinoma and those with DCIS. In patients with invasive carcinoma, the risk of metastatic disease is present at diagnosis and is not altered by local recurrence in the breast. In DCIS, the risk of metastases at diagnosis is negligible, and an invasive local recurrence carries with it the potential risk of breast cancer mortality. The suitability of excision and irradiation as a treatment for DCIS should be determined by the incidence of invasive recurrence and the results of salvage therapy. Solin reported the results for 268 women in whom 271 breasts were treated with excision and irradiation at ten institutions in Europe and the United States. At a median follow-up of 10.3 years (range, 0.9 to 26.8), 46 failures were observed. These included 43 local failures, 1 combined local and regional, 1 local and distant, and 1 distant-only failure. The 15-year actuarial rate of local failure was 19%, and the median time to failure was 5.2 years. Noteworthy is the fact that, although the local failure rate in this study was relatively high, the 15-year cause-specific survival was 96%113 The methods of evaluation and the extent of surgical resection used in this study would probably not be considered adequate today. Gross excision of the tumor was performed in all cases, but only 15% underwent reexcision, and margin status was unknown in 120 cases (46%). The median whole-breast radiation dose delivered was 5,000 cGy, and 164 of 261 cases (63%) received a boost to the primary tumor site. In spite of these caveats, this study is noteworthy for the large number of patients and relatively long duration of follow-up, and the low cause-specific mortality is reassuring. An examination of the subset of patients with mammographically detected lesions from this series (n = 110) did not reveal a significantly lower rate of local failure than that seen in the group as a whole,a finding also reported by Hiramatsu

Because one-half of the local failures seen after breast-conserving therapy for intraductal carcinoma are invasive carcinoma, the outcome of salvage treatment of these recurrences is important. Solin  described 42 cases of local failure in 274 cases of DCIS treated with excision plus irradiation. The median time to local failure was 5.1 years, and the median follow-up after salvage treatment was 3.7 years. Nineteen of the recurrences (45%) were intraductal carcinoma, and 14 of these were detected with mammographic findings alone. All of the women with intraductal recurrences remained free of disease after mastectomy with a median follow-up of 4.7 years. Five patients with invasive recurrence developed distant metastases, either simultaneously with the recurrence (one patient) or subsequently (four patients). Chest wall recurrences were seen in three patients who had salvage mastectomy for an invasive recurrence, and all of these women developed distant metastases. Of the entire group of 42 women with recurrence, 36 patients (86%) were alive and free of disease, 4 patients (10%) died of disease, 1 patient was alive with disease, and 1 patient died of other causes. Similarly high rates of salvage have been reported in other studies; however, the ultimate breast cancer mortality resulting from breast-preserving treatment cannot yet be assessed. The 42 cases of recurrence reported by Solin et al.120 occurred in a study population with a median follow-up of 78 months. Forty percent of the recurrences reported were seen between 5 and 10 years after treatment, and an additional 12% occurred after 10 years, indicating that a significant risk of further local recurrence exists among these women. In addition, the median follow-up after salvage therapy was only 3.7 years, too short an interval to determine the eventual risk of distant metastases.

Uncertainty about the biological significance of DCIS has led a number of investigators to examine the use of excision alone as a treatment. In general, patients treated with this approach are highly selected and are usually chosen on the basis of low histologic grade or small size of the tumor. The percentage of patients with DCIS in the study population who meet these selection criteria is usually not stated, so that the number of women with DCIS who are candidates for this type of treatment is unclear. Lagios  described 79 women with DCIS lesions 25 mm or smaller (mean size, less than 8 mm) who were treated with wide excision alone. Fifteen local recurrences (19%) were noted at a mean follow-up of 124 months. Eight of the recurrences were invasive carcinoma and seven were DCIS. No breast cancer deaths occurred; ten deaths occurred from other causes. Schwartz  reported on 191 women in whom 194 breasts were treated by excision alone and who were followed for a mean of 55 months. Approximately two-thirds of the cases were detected as mammographic calcifications, and one-third as incidental findings. The crude rate of recurrence was 14.4%, and the 10-year actuarial rate of recurrence was 24.6%. Only 18% of the recurrences were invasive carcinoma, a much lower rate of invasive recurrence than reported in other studies, and no breast cancer deaths have occurred. The patients were accrued between 1978 and 1996, which indicates the highly selected nature of the population. All of the tumors measured less than 2.5 cm in greatest dimension, and reexcision was routinely used. In contrast, a 43% recurrence rate was noted at a mean follow-up of 85 months for 22 patients treated by local excision alone as part of Protocol B-06.These women were initially diagnosed as having invasive carcinoma and were later reclassified as having DCIS. Only one of these cases was nonpalpable cancer, and tumor size averaged 2.2 cm. These wide variations in local failure rates emphasize the importance of patient selection when attempting to treat women with DCIS by excision alone. In general, studies of the management of DCIS with excision alone show that, when local failure occurs, DCIS is present in approximately one-half of the cases and invasive carcinoma in the other one-half. The time course to local failure is prolonged; studies with longer follow-up show higher local failure rates. In the report of Gallagher et al.,126 the median time to local failure was 47 months; four of eight patients monitored for more than 9 years had recurrent disease, a finding that emphasizes the importance of long-term follow-up.

Two studies from the NSABP have described treatment outcomes in women with DCIS who were randomized to receive excision alone or excision plus radiation therapy. Protocol B-06117 (was designed to evaluate the local therapy of invasive carcinoma. On review of pathologic material, 78 patients with DCIS alone were identified. At a follow-up of 83 months, 12 of the 27 patients (7%) treated with irradiation and 9 of the 21 patients (43%) treated with lumpectomy alone had local failures. No local failures occurred in the 28 women treated with mastectomy. The NSABP has also reported the results of a prospective study designed to evaluate the role of radiation therapy in DCIS.In this study, 818 women were randomized to excision alone or excision plus 5,000 cGy of irradiation to the breast. Histologically negative surgical margins, defined as no contact between tumor-filled ducts and an inked surface, were required in both groups. Eighty percent of the women in the study had tumors detected by mammographic screening. The study was first reported at a median follow-up of 43 months, at which time a 58.8% reduction in the annual incidence of ipsilateral breast recurrence was observed in the irradiated group relative to the nonirradiated group. At 90 months of follow-up, the incidence of invasive recurrence was reduced from 13.4% in the nonirradiated group to 3.9% (p = .000005) in the irradiated group.118 The incidence of recurrent DCIS was also significantly reduced, from 13.4% in the group without radiation to 8.2% in the group with radiation. The continued benefit of radiation therapy in reducing the risk of both invasive and noninvasive recurrences over time strongly suggests that its benefit is not due solely to the control of clinically occult invasive carcinoma, as was suggested after the initial publication of this study. The overall survival does not differ between the two groups. Thirteen deaths have occurred in the 814 evaluable patients; overall survival rate is 94% for patients treated by lumpectomy alone and 95% for those receiving radiation therapy.

Prognostic Factors for Treatment Selection

As is apparent from the preceding discussion, identification of women with a high risk of developing invasive carcinoma after breast-conserving therapy for apparently localized DCIS would be extremely helpful. Physicians are presently unable to differentiate tumors that will recur as invasive carcinoma from those that will recur as DCIS. Some studies have suggested, however, that infiltrating carcinoma that develops after high-grade DCIS is more likely to be poorly differentiated and associated with poor prognosis than infiltrating carcinoma that develops after low-grade DCIS.Histologic subtype and grade of DCIS have generally been the most widely studied predictive factors for recurrence. The limitations of histologic subtyping are discussed in detail in the section on Pathology. Despite these limitations, evidence suggests that histologic subtype and nuclear grade, alone or in combination, may be prognostic factors for local failure after treatment with excision alone or excision plus radiation therapy. In the study of Lagios of women treated with wide excision alone, a 33% recurrence rate (12 of 36 cases) for patients with high-grade DCIS containing comedo-type necrosis was noted, compared with a rate of 2% (1 of 43) for patients with intermediate-grade or low-grade DCIS. Similar findings were reported by Schwartz After wide excision alone, 32% of patients with comedo histology experienced local recurrence, whereas only 3% of those with noncomedo histology had recurrent disease. Eusebi observed that local failure occurs earlier with high-grade DCIS than with low-grade DCIS. Long-term follow-up is needed to see whether these differences persist. However, other studies have observed no relationship between grade and failure.

High nuclear grade and comedo-type histology have also been found to be prognostic for local recurrence when radiation is added to excision. Silverstein reported an 11% rate of local recurrence in women with comedo DCIS compared with a 2% failure rate in those with noncomedo DCIS among 96 women studied; median follow-up was 45 months. An update of this study with a longer follow-up (median of 62 months), however, showed no difference in the rate of local failure based on histologic subtype. Pathology slides were available for review for 172 women from the multicenter study reported by Solin. Sixteen local recurrences occurred in the 172 patients studied. A comparison of recurrence rates for women with the comedo and noncomedo subtypes showed no significant differences (14% compared with 6%, respectively). However, recurrence rates were significantly higher for patients with a tumor of the comedo subtype and a nuclear grade of 3 than for any other groups (20% compared with 5%; p = .009). Forty-four of the 172 patients studied had this combination of factors, and one-half of the observed local recurrences occurred in this group. A multivariate analysis, including histologic subtype of the primary tumor, nuclear grade, amount of necrosis, final pathology margin, and the combination of histologic subtype of comedo carcinoma plus nuclear grade 3, found that only the combination of comedo carcinoma plus nuclear grade 3 correlated significantly with local control method. These patients also had a shorter time to treatment failure (median of 38 months) than did patients without the combination of comedo subtype and nuclear grade 3 (median of 78 months). With additional follow-up, however, the combination of comedo subtype and grade 3 no longer identified a group at increased risk for recurrence.This finding, combined with the Silverstein data discussed earlier, suggests that the importance of the comedo subtype in predicting local recurrence may be overemphasized in studies with short-term follow-up. The pathologic predictors studied had no impact on overall survival or freedom from distant metastases. The NSABP has reported the results of two analyses of the pathologic features of 623 of the 824 patients enrolled in Protocol B-17.In the initial report, moderate or marked comedo necrosis and uncertain or involved margins were associated with an increased risk of local failure. Although radiation therapy reduced the risk of failure in all subgroups, the benefits were greatest in those patients at highest risk for recurrence. In their second report, multivariate analysis of nine histologic features, including margins, histologic type, nuclear grade, tumor size, and comedo necrosis, demonstrated that only comedo necrosis significantly predicted an increased risk of ipsilateral breast recurrence after 8 years. A breast recurrence was seen in 23% of patients with absent or slight comedo necrosis who did not receive radiation therapy. The addition of radiation therapy eliminated most of the risk associated with this factor; 13% of those with absent or slight comedo necrosis and 14% of those with moderate or marked comedo necrosis experienced recurrence after radiation therapy.

Several studies have suggested that age may influence the risk of local recurrence after breast-conserving therapy. Solin  noted a 25% incidence of local failure in patients aged 50 or younger who were treated with excision and irradiation compared with 2% in patients older than age 50, in spite of the fact that nuclear grade, tumor size, and margin status did not differ between groups. The median time to local failure was also shorter in the younger patients (4.9 years versus 8.7 years). Van Zeealso observed higher rates of local failure in women younger than age 40 than in their older counterparts after treatment with excision and irradiation or excision alone. Fourquet et al.133 noted an actuarial 10-year recurrence rate of 30% for women aged 40 and younger treated with excision and irradiation, compared with 14% for those older than age 40. One possible explanation for these results may be the presence of higher circulating levels of estrogen in the younger patients, because estrogen is known to have promotional effects in breast cancer cell lines.

Other studies have suggested that a family history of breast cancer may influence the risk of local failure after excision and irradiation. Hiramatsu observed a 37% failure rate in patients with a family history of breast cancer compared with a 9% failure rate in those without a family history. McCormick et al.115 reported that 40% of patients who experienced local-control failure had a first-degree relative with breast cancer compared with 11.4% of patients in whom local control was maintained.

Use of Tamoxifen

Data from the NSABP Breast Cancer Prevention trial134 demonstrating that tamoxifen (tamoxifen citrate) therapy reduces the risk of both invasive and intraductal carcinoma in women at increased risk for breast cancer development , coupled with data from tamoxifen treatment trials demonstrating a reduction in contralateral breast cancer incidence,135 strongly suggests that tamoxifen therapy would be beneficial in DCIS. The initial results of NSABP Protocol B-24, in which 1,804 patients with DCIS treated by lumpectomy and radiation therapy were randomized to tamoxifen 20 mg daily for 5 years or placebo, have been reported after a mean follow-up of 62 months. The addition of tamoxifen reduced the average annual rate of invasive breast recurrence from 0.90 per 100 patients to 0.50 per 100 patients (relative risk, 0.56; p = .03) and reduced the rate of recurrent DCIS from 1.10 to 0.87 per 100 patients (relative risk, 0.82; p = .43). Overall, the risk of ipsilateral recurrence of any type (invasive or noninvasive) or of new contralateral breast cancers, or distant disease was reduced from 13.4% to 8.2% at 5 years, a highly significant reduction. As discussed in the section Treatment Options, because invasive recurrence has the potential to impact mortality, these results provide a strong rationale for the use of tamoxifen in patients with DCIS treated with a breast-conserving approach. These benefits must be weighed against the potential risks of treatment, which are lowest in patients younger than age 50 years and in those with a prior hysterectomy.

Treatment Selection

The available information on DCIS suggests that, although all patients can be treated with mastectomy, many are candidates for treatment with excision and irradiation, and a smaller group may be appropriately treated with excision alone. In treatment selection, it is useful to consider the risk of breast cancer recurrence, the risk of invasive breast cancer, and the risk of dying of breast cancer associated with breast-conserving treatment.

The available data on breast-conserving treatment combined with radiation therapy generally show recurrence rates of 10% to 15% at 10 years. Approximately one-half of these recurrences are invasive carcinoma, a risk of 5% to 7%. The risk of dying of breast cancer is approximately one-third the risk of developing the disease, so the risk of breast cancer death is 2% to 3% at 10 years. The risk of death 10 years after a mastectomy for DCIS is 1% to 2%. The major force of breast cancer mortality after mastectomy is likely to be evident in the first 10 years after treatment, however, given that death is presumably due to occult invasive disease present at the time of diagnosis. Local recurrences in DCIS continue to occur after 10 years, with additional breast cancer–associated mortality. Comparisons of breast cancer mortality rates 30 years after treatment could show greater differences in survival between women treated with mastectomy and those treated with excision and radiation therapy than the 1% to 2% estimated here. However, the use of tamoxifen significantly reduces the risk of invasive recurrence, eliminating most of the potential survival difference between treatment with excision and radiation therapy, and mastectomy. Whether radiation is necessary for all patients with DCIS treated with a breast-sparing approach remains uncertain. Retrospective data indicate that highly selected patients, usually with small, low-grade DCIS, have a very low local failure rate after excision alone. The only results available from a prospective study are those from NSABP Protocol B-17.118 Although this study indicates that radiation therapy reduces the risk of local recurrence in all subgroups of patients with DCIS, Page and Lagios have questioned whether a more detailed mammographic and pathologic evaluation would allow the identification of a subgroup of patients who will do well without radiation therapy. In addition, the impact of tamoxifen therapy on recurrence when radiation therapy is not given is uncertain.

That DCIS, rather than being a single entity, represents a spectrum of diseases of differing biological potential is becoming increasingly clear. This clinical observation is supported by studies of biological markers, which indicate that characteristics of the malignant phenotype are more likely to be expressed in high-grade DCIS. Until developments in molecular biology allow more precise prediction of which tumors progress to invasive carcinoma, efforts must be directed toward minimizing local recurrence in women treated with a breast-conserving approach. The initial step in the evaluation of patients with DCIS is the determination of the extent of the lesion. Because most patients with DCIS have nonpalpable mammographic lesions, careful mammographic evaluation before treatment selection is critical. Holland have previously reported that the extent of poorly differentiated DCIS assessed by microscopy correlated well with the extent of the lesion evaluated radiologically, but the mammographic appearance of well-differentiated tumors substantially underestimated the microscopic extent. However, the routine use of magnification views as part of the mammographic evaluation allowed the detection of additional calcifications that reduced the discrepancy between the pathologically and mammographically determined extent of well-differentiated DCIS. Needle localization should be used to guide the biopsy; if the calcifications are extensive, bracketing wires are useful to aid in complete excision. Specimen mammography is essential to confirm the excision of calcifications. In cases in which calcifications are extensive or approach the edge of the surgical specimen, postexcision mammograms are useful to confirm the removal of all suspicious calcifications. Gluck performed postexcision mammograms, including spot compression views, on 43 women who required reexcision due to positive or unknown margins after a diagnosis of breast carcinoma. Twenty-eight patients had DCIS; the positive predictive value of residual calcifications as an indicator of residual tumor was 0.67 and increased to 0.9 when more than five calcifications were present. Even when the margins are negative, postexcision mammography can demonstrate residual calcifications indicative of the need for further resection. Although DCIS lesions are not clinically detectable, they may be quite large. Morrow et al.140 found that contraindications to breast-preservation methods were present in 33% of patients with DCIS compared with only 10% of patients with stage I invasive carcinoma. Extensive disease that could not be encompassed with a cosmetic resection was the major contraindication to breast-conserving therapy in patients with DCIS.

A detailed pathologic evaluation is also needed and should include orientation marking, inking of the specimen, and measurement of both specimen and tumor size before sectioning. Because accurate measurement of microscopic DCIS is often difficult, reporting the number of blocks in which DCIS is present, as well as its largest single extent in any one slide, is often useful. The correlation of microcalcifications with DCIS (i.e., whether DCIS is present only in areas of calcification or in calcification and adjacent breast tissue) as well as the margin status should be noted. If margins are involved, the extent of involvement should be stated; when margins are negative, proximity of the lesion to the margin should be noted.

Attempts have been made to incorporate the size of the lesion, its histologic features, and the extent of the surgical excision into a prognostic index that would direct treatment selection. One such index is the Van Nuys Prognostic Index (VNPI), which assigns scores of 1, 2, or 3 for histologic type, width of the surgical margin, and size of the lesion. Lesions with low VNPI scores are said to be suitable for excision alone; those with intermediate scores (5 to 7) require the addition of radiation therapy; and those with high scores require mastectomy. Although such a simplification of the decision-making process is attractive, this index has a number of limitations. The index was developed using retrospective data on 254 patients and was validated using retrospective data on 79 patients from another institution. The use of the classification system is dependent on the reproducibility of the individual components. Because the histologic classification scheme and method of tumor measurement are not in universal or even routine use, this is a significant issue. The potential problems in duplicating these elements have been discussed in detail by Schnitt  Equally important is the fact that the patients used to develop this index were treated over a large time span from 1972 and 1995. However, treatment with excision alone was used in more recent years, whereas treatment with excision and irradiation was more common in the past. This suggests that the low rate of local recurrence seen after excision alone may be due to improvements in mammographic and pathologic evaluation. Hiramatsu reported that the incidence of local recurrence 6.5 years after excision and irradiation decreased from 12% to 2% when patients treated between 1976 and 1985 were compared with those treated between 1985 and 1995, although radiation technique did not change. Finally, although the VNPI is based on factors that most clinicians would consider important in predicting the behavior of DCIS, whether these are the most important factors in determining outcome is not clear. In a subsequent report, the authors of the VNPI noted that, when DCIS was widely excised to negative margins, nuclear grade was not a predictor of recurrence in patients treated with excision alone or excision and radiation therapy. As noted previously, age and a family history of breast carcinoma have been suggested to influence the risk of local recurrence in retrospective studies. For these reasons, the authors do not believe that the VNPI is an appropriate substitute for an individualized assessment of the risks and benefits of the available treatment options for DCIS.

The lack of a single appropriate treatment option for all patients with DCIS is reflected in national patterns of care. A review of 39,010 patients with DCIS reported to the National Cancer Data Base between 1985 and 1993 demonstrated that the use of breast-preserving techniques increased from 31% to 54% in that 8-year interval. Overall, only 45% of the patients treated with breast-preserving techniques received radiation therapy, although the use of this modality increased from 38% of cases to 54% during the study period. Smaller tumors and low-grade lesions were most likely to be treated with breast-preserving surgery alone.

MANAGEMENT SUMMARY

DCIS represents a heterogeneous group of lesions of varying malignant potential. Total (simple) mastectomy is associated with a cure rate of 98% to 99% for all types of DCIS.

Patients with localized DCIS are candidates for breast-sparing surgery and irradiation. Detailed mammography and careful pathologic evaluation are essential to confirm the localized nature of the lesion and to judge the adequacy of resection. The goals of surgery are to remove all suspicious microcalcifications and to achieve negative margins of resection.

Excision alone may be an appropriate treatment for selected women with small (less than 1-cm to 2-cm) low-grade DCIS lesions with clearly negative margins.

Axillary dissection is not indicated in DCIS. In women with large high-grade lesions undergoing mastectomy, a low axillary sampling obviates the need for reoperation if invasion is identified.

The use of tamoxifen should be considered to reduce the risk of ipsilateral breast tumor recurrence after breast-sparing surgery and to reduce the risk of contralateral breast cancer in all patients.

A detailed discussion of the risks and benefits of the various options must be undertaken to allow each woman with DCIS to make an informed treatment choice.