The incidence of ductal carcinoma in situ (DCIS) has increased dramatically since the advent of screening mammography in the 1980s. The age-adjusted DCIS incidence rates increased 17.5% annually from 1983 to 1992. The percentage of patients with DCIS treated with mastectomy has decreased from 71% in 1983 to 44% in 1992. The percentage of patients with DCIS undergoing lumpectomy and radiation in 1992 was 23.3% and lumpectomy only was 30.2%.
The clinical significance of DCIS detected by mammography is still uncertain. It is unclear which patients will ultimately develop invasive disease. Many investigators are concerned about the overdiagnosis and subsequent overtreatment of these cases. However, this concern must be balanced against the finding in one study of an actuarial local recurrence rate of 19% after excision and radiation therapy, with 53% of the recurrences presenting as invasive disease. Therefore, it is certain that a significant percentage of DCIS lesions are obligate precursors to invasive disease and could potentially affect the survival of the patient.
Different Classifications for DCIS
Drs. Silverstein and Lagios, in their thoughtful summary and discussion, present a tool to help predict local recurrence in patients with DCIS and therefore recommend local treatment. It has long been recognized that DCIS is not one homogeneous disease. Therefore, there have been various suggestions for new classifications of DCIS.
Poller et al suggested a simplified classification for DCIS based on cellular proliferation and c-erbB-2 protein expression. This paper is the basis for the current Van Nuys Pathologic Classification. The three categories proposed by Poller et al were comedo DCIS, DCIS with necrosis (non-pure comedo), and DCIS without necrosis. The cases of comedo DCIS and DCIS with necrosis were associated with a higher S-phase and the presence of c-erb-B2 protein. The current Van Nuys classification identifies three factors that are felt to be important in predicting local recurrence: tumor size, pathologic margins, and a pathologic classification using nuclear grade and the presence or absence of necrosis.
Holland et al based their classification primarily on cytonuclear differentiation and secondarily on architectural differentiation. These investigators defined three categories. The first category consists of poorly differentiated tumors with absent or minimal architectural differentiation and usually with necrosis present. The most important characteristic feature is the marked pleomorphic nuclei. The second category is well-differentiated DCIS, usually with a cribriform, micropapillary, changing, and rarely solid pattern. Necrosis is absent and mitoses are uncommon. The third group is intermediately differentiated DCIS.
The strength of this classification is that cytonuclear features are more often consistent in DCIS than is architectural pattern or necrosis. However, as yet there is no correlation between this classification and the biologic activity of DCIS.
In a study by Bobrow et al, two independent observers used the Holland classification to reclassify 105 cases of DCIS. The poorly differentiated group was associated with c-erb-B2 protein staining, a high proliferation rate, p53 protein, and the absence of progesterone(Drug information on progesterone) receptors. Well-differentiated lesions had no c-erb-B2 or p53 present, proliferation was low in all cases, and progesterone receptors were present in 95%. These findings may help determine the biology of DCIS and may be used in the future to determine appropriate treatments.
How to Define a Free Margin?
The specific prognostic factors used in the Van Nuys Prognostic Index (VNPI) include the measurement of tumor margins. When is a free margin free? The answer is not so simple. Determination of negative pathologic margins differs among groups reporting DCIS data. Solin et al defined negative margins as tumor more than 2 mm from the inked margin. Margins were determined by a retrospective review of patient records and were available in only 47%. The NSABP defined negative margins as no transection of the tumor. Of the 790 NSABP cases, 40% either were unavailable for central pathologic review or had unknown margins.
In the Van Nuys series, all cases underwent central pathology review by one pathologist. In their initial publications, Silverstein et al did not rigorously define free margins.[7,8] However, in a subsequent publication, the scoring system used an exact measurement of tumor-free areas, giving each one a different score. This includes, as a best-case scenario, a 10-mm or more margin.
The use of a 10-mm or more margin seems very logical based on the work of Holland's group. They reported on the results of a three-dimensional imaging study of DCIS in 60 mastectomy specimens. They found that continuous and multifocal growth patterns are usual, and that multicentric tumors (defined as a gap of 4 cm or more) are rare, occurring in only one case. Poorly differentiated tumors more often had continuous growth, whereas well-differentiated lesions had multifocal distribution. Therefore, reliability of margin assessment varies with type. In all the cases reviewed, only 8% had a multifocal distribution with gaps more than 10 mm and 7% were in well-differentiated tumors.
The conclusion that can be drawn from the work of Holland et al is that 90% of all DCIS can be removed surgically if a 1-cm margin is attainable. Clearly, it is not always feasible to do this and obtain a good cosmetic result.
Which Factors Predict Local Recurrence?
The next question is, if all DCIS is removed, which prognostic factors will predict local recurrence? The Van Nuys group suggests that it is patients with high-nuclear-grade lesions (grade 3) who are most likely to develop a recurrence, followed by those with grade 1 or 2 lesions with necrosis. Their conclusion is based on a retrospective classification of patients, which, as the authors point out, are a highly selected group. The patients presented in the Van Nuys series received one of two different nonrandomized treatments, either excision or excision followed by radiation therapy. The multivariate analysis for prognostic factors predictive of recurrence was significant for nuclear grade, tumor size, and margin width, and showed a trend toward significance for the presence of necrosis.
The NSABP B-17 study reported that the presence of moderate or marked comedo necrosis predicted for second ipsilateral breast tumors. Solin et al found that no pathologic parameters were predictive of local failure after radiation therapy. Ottesen et al found that tumor size 10 mm, or more large nuclei, and comedo necrosis predicted for local recurrence after excision of DCIS. These data would suggest that necrosis, indicative of a high proliferative rate, and large nuclei with pleormorphism and prominent nucleoli are associated with more aggressive local disease.
In conclusion, DCIS is a very heterogeneous disease. This is no doubt the reason for such controversy, disagreement, and uncertainty over its treatment. The Van Nuys classification attempts to simplify treatment decisions. Intuitively, this classification and the treatment recommendations based on it make good common sense. First, it is ideal to obtain clear margins by 1 cm. If that is impossible, residual disease is likely, and thus, mastectomy is the only treatment available to remove all the disease. This still does not take into consideration the possibility that lesions may be biologically "benign" and not likely to progress to invasive disease. If the tumor is large, a good cosmetic result may not be likely with wide excision, and therefore, mastectomy may be appropriate. If the tumor is small with clear margins, excision alone may be adequate, except in high-grade cases.
There is still a question as to whether the local recurrence rate is higher with high-grade cases compared to good nuclear grade. Treatment options for high-nuclear grade lesions include radiation, with mastectomy reserved for the treatment of the local recurrence (since disease will not recur in a significant number of women), or mastectomy as primary local treatment. In any case, as Drs. Silverstein and Lagios state, patients should be informed of all the pros and cons of each treatment option.
The Van Nuys group appropriately points out that their data are retrospective. Patients were selected for a certain local treatment based on the factors felt to be important at that moment in time. Ideally, it is important to test various local treatments in a prospective randomized trial with attention to such details as margins, nuclear grade, and necrosis.
There are many ongoing European trials evaluating radiation after segmental mastectomy in a large number of women with DCIS. Central pathologic review of these studies could help clarify some of the unanswered questions. Future trial results, such as those of the NSABP and the United Kingdom trial evaluating the use of tamoxifen(Drug information on tamoxifen) (Nolvadex) in DCIS, are eagerly awaited.
The elucidation of the biology of different subsets of DCIS will go a long way toward clarifying treatment decisions. Also, prospective treatment based on biologic markers of proliferation and aggressiveness will help determine the appropriate treatments.