Measuring Extent of Ductal Carcinoma In Situ in Breast Excision Specimens: A Comparison of 4 Methods

Ductal carcinoma in situ (DCIS) is, by definition, a neoplastic proliferation confined to the ductal/lobular system and limited by the basement membrane. Since the study by Wellings,1 we know that DCIS primarily arises within the terminal duct lobular unit. Studies using 3-dimensional and computerized reconstruction have shown that DCIS is mainly unicentric and extends along a complex, branching ductular system, intertwining with adjacent ducts in a pyramid-like shape toward the nipple.2,3 Small gaps have been seen, mainly separating areas of low-grade DCIS,2,4 and anastomoses between ducts have been described by some authors,5 but have not been identified by others.3,6 

The pathologist is presented with a unique challenge when faced with a breast excision specimen for DCIS. Not only is the ductular system complex but also, with the advent of mammography, most cases are detected by the presence of calcifications and are grossly invisible. With the exception of some high-grade lesions that can be recognized by punctate necrosis and areas of ill-defined firmness, sampling and measuring DCIS must be done with the use of radiologic imaging. Tissue processing methods, involving specimen inking per the surgeon's orientation, followed by an x-ray of serially sectioned slices, have been described to address this problem.7–9 This method allows sampling of radiologically suspicious lesions and calculation of extent through reconstruction using the specimen x-ray.9 

The best possible estimate of the size of DCIS is important because that measure is used to guide clinical management. The size of the DCIS has been shown to correlate with local recurrence,10–12 with the risk of a compromised margin,13 and with residual disease.14–16 As a result, this variable is used, along with other factors, by the clinician in counseling patients on the need for further surgery or radiation therapy.11,12 

Given the importance of the extent of the DCIS in clinical practice, it is imperative that this information be provided in pathology reports in a standardized fashion. However, often, this information is not reported or the extent of DCIS is measured imprecisely. In a series of outside slide review of breast excisional cases, only 21% of cases reported the size of DCIS, and most of those measurements were given as the largest dimension measured on a single slide.17 In only one case was the size derived through submitting the specimen in a sequential manner.

The goal of this study was to compare the size of DCIS obtained using various methods, with the gold standard method being serial sequential sampling (SSS) and radiologic correlation in a series of breast excision specimens with DCIS.

The study was approved by the Mount Sinai Hospital Ethics Review Board. A consecutive group of 157 patients who had been diagnosed with DCIS without microinvasion on excisional biopsy/lumpectomy at Mount Sinai Hospital during the period of January 2000 through July 2007 were identified. Additional inclusion criteria were as follows: (1) all histologic slides were available for review, (2) a sliced specimen radiograph and section diagram were available for review, and (3) cases not submitted in toto had negative sections flanking the limits of DCIS. In addition, all reexcision lumpectomy specimens were excluded. A total of 78 cases met these strict inclusion criteria, of which 63 (80%) were submitted in toto.

All specimens were sampled in a sequential fashion as routinely done in our institution and previously described by others.7–9 Briefly, the breast excision was measured in 3 dimensions. The entire specimen was painted with silver nitrate; the anterosuperior and superior margins were repainted with blue ink, and the anteroinferior and inferior margins were repainted with green ink. Following fixation in 10% formalin, the specimen was serially sectioned into thin slices, and the slices were radiographed in the pathology department, generating either a plain film (before January 2006) or a digital image (after January 2006). The sections submitted for histologic examination were documented directly on the x-ray, which then served as a section map. When a slice was too thick to fit into a cassette, it was bisected, and that fact was indicated on the specimen diagram.

Using the gross description, section map, and sliced specimen radiograph, the slides for each case were reviewed, and the extent of DCIS was calculated using 4 methods.

The gold standard method was the SSS method.9 The extent of DCIS was calculated in 3 dimensions using the radiograph and section map. In most cases, the specimen was sectioned in the mediolateral plane. The extent in this plane was calculated by multiplying the average slice thickness by the number of slices involved. The average slice thickness was calculated by dividing the specimen size in the plane of sectioning by the number of slices (see example in Figure 1). In some cases, the largest extent involved a single slice and extended across several glass slides. In these cases, the glass slides were reconstructed side by side, with reference to the radiograph and section diagram, to measure the largest extent across these slides (Figure 2). When DCIS involved a single block, the largest dimension on the glass slide was measured. Other estimates of the extent of DCIS were obtained as follows: (1) the calcification method involved reviewing the sliced specimen radiograph and calculating the extent based only on a theoretical sampling of areas of calcification; (2) the blocks method involved multiplying the number of blocks involved by DCIS by 0.3 cm, the average tissue thickness reported by some research18,19; and (3) the single-slide method involved measuring the largest extent of DCIS on a single slide. An example of each of these calculations is shown in Figure 1.

Graphical displays were employed as the main method of comparing the alternative methods to the SSS method. The measurements from each alternative method were plotted against the SSS method to present the overall pattern. To show more detail on the lack of agreement and its relationship to DCIS size, the differences between the measurements (alternative minus SSS method) were plotted against the means obtained by averaging all 4 methods. The latter was used as estimate of the true DCIS extent because the SSS observations are not without error in spite of being the gold standard and because using SSS by itself tends to induce an artificial relationship between it and the differences.20 

Because the between-method differences exhibited quite severe nonnormality, we employed a simple nonparametric method21 to characterize agreement with the SSS method: the percentage of cases was tabulated where the discrepancies between the alternative and the SSS methods fell in the intervals of less than or equal to 0.5 cm, greater than 0.5 cm but less than or equal to 1 cm, greater than 1 cm but less than or equal to 2 cm, and greater than 2 cm. Results are presented overall and by DCIS lesion size, grouped at 2 cm or less, greater than 2 cm but less than or equal to 4 cm, and greater than 4 cm.

The blocks method is based on a presumed relationship between the number of blocks involved and extent. We, therefore, plotted the DCIS extent as given by the SSS method against the number of blocks involved to display this relationship. Lines are superimposed showing the number of blocks multiplied by 0.3 cm and multiplied by the average slice thickness at our institution (0.5 cm). Statistical calculations were performed using S-PLUS 6.2 for Windows (Insightful Corp, Seattle, Wash).

In this study, 78 cases of DCIS were assessed. One case was multifocal (2 distinct foci of DCIS separated by a distance of 1.0 cm), and for that case, only the largest focus was included. There were 4 grade 1 (5.1%), 40 grade 2 (51.3%), and 34 grade 3 (43.6%) cases. Comedonecrosis was present in 47 (60.3%) cases. The mean number of blocks submitted per case was 26.7 (range, 8–66), and the average tissue thickness was 0.5 cm.

By the SSS method, the largest dimension was in the mediolateral plane in 56 (71.8%) cases. The largest dimension was measured on a single-glass slide in 13 (17%) cases, was measured across 2 or 3 slides within a single slice in 8 (10%) cases, and was calculated by multiplying the average slice thickness by the number of slices involved in 57 (73%) cases. The extent measured ranged from 0.3 to 9.0 cm, with a mean of 2.7 cm. Figure 3, A through C, shows the correspondence between the extent of DCIS as measured by each alternative method versus the SSS method. It is noticeable that the extent of DCIS as measured by both the calcification and single-slide methods was always less than or equal to the SSS estimate. In Figure 3, D through F, the differences (alternative minus SSS method) for each method are plotted against the mean, which acts as the estimate of true DCIS size. The differences show the number of centimeters that the alternative method underestimated or overestimated the extent compared with the SSS method, with the zero line indicating no difference. The following patterns of disagreement can be discerned. In the calcification method, the degree of underestimation increased until approximately 2 cm, when the spread appeared to be capped. The blocks method differences were small when the extent of DCIS was small, but then it produced greater overestimates and underestimates as the DCIS extent increased. The single-slide method showed small differences when DCIS extent was less than approximately 1 cm; however, as DCIS extent increased, the underestimations greatly increased. This is also shown in Figure 3, C, where all lesions greater than 5 cm were measured as 2.5 cm or less by the single-slide method.

For each method, the percentage of cases that agreed with the SSS method to specific degrees of closeness is shown in the Table. The calcification method had 64% of estimates within 0.5 cm of the SSS method versus 42% for the blocks method and 27% for the single-slide method. Deviations larger than 2 cm were seen in 9%, 8%, and 30% of cases for the calcification, blocks, and single-slide methods, respectively. Because the degree of discrepancy did not remain constant during the range of DCIS extent, the percentages are also given separately by DCIS size category. For example, for lesions of 2 to 4 cm in size, 15% of calcification, 8% of blocks, and 58% of single-slide method discrepancies were greater than 2 cm.

Figure 4 displays the relationship between the number of positive blocks and DCIS extent as given by the SSS method. The superimposed line A is the blocks-method estimate that was calculated by multiplying the number of positive blocks by 0.3 cm, the proposed average tissue thickness. Line A tends to underestimate small lesions, whereas very large lesions were overestimated. In our study, the actual average tissue thickness of a single slice was 0.5 cm. Therefore, it is not surprising that the blocks method, using a multiplier of 0.3 cm, underestimated extent in most cases. Line B shows the extent of estimates obtained when the average tissue thickness of 0.5 cm was used as the multiplier. Although 0.5 cm is a less-biased estimate at our institution when the number of blocks involved is less than approximately 8 or 10, it overestimates size beyond that point.

Because the single-slide method performed so poorly, we investigated whether results were distorted by its inclusion in the average used to characterize the true DCIS extent in Figure 3 and in the Table. On omitting it from the mean, the only difference to the graph was that points plotted at 8 cm on the x-axis were plotted at 9.5 cm instead. Changes to the Table were restricted to the part broken down by the mean value, and none of the percentages changed in a material way.

Measuring the extent of DCIS in a breast local excision/ lumpectomy specimen is challenging and can be done many different ways. It is important that this measurement be reported in the pathology report and that it be as accurate as possible because this value has prognostic implications and is used to guide clinical management. The size of DCIS is one of the factors included in the University of Southern California/Van Nuys Prognostic Index and has been shown to predict local recurrence in patients with DCIS treated conservatively.10–12 The size of DCIS is also associated with a higher rate of positive or close margins13 and is an independent predictor of residual disease.13–16,22 Clinicians use this information along with other pathologic and patient characteristics to counsel patients on the need for additional surgery, radiation therapy, or tamoxifen.11,23 

In our study, the SSS method was used as the gold standard. This method is routinely used in numerous breast centers and is the preferred method of tissue processing advocated by the Consensus Conference on the Classification of DCIS.18 Yet, a recent study17 showed that numerous hospitals fail to include the size of DCIS in their report. This lack of standardization may partly be due to the time and resources required for the SSS method. The Consensus Conference noted that “when feasible and practical, the entire specimen should be processed in sequence in separate cassettes.” 18 Submitting a specimen in toto requires little additional time but may require additional resources. In our study, the average number of blocks submitted was 26.6 for specimens submitted in toto. When a specimen cannot be feasibly submitted in toto, additional time is required for tissue processing to selectively sample based on radiologic abnormalities. Some may advocate sampling the area of the wire tip in a wire-localization lumpectomy, but that can be inaccurate because the wire is not always located in the precise location of the abnormality. A sliced specimen x-ray must be done either in the pathology department if the technology is available or in the hospital radiology department to sample all suspicious lesions. A diagram indicating the location of the sections taken is extremely important in both selective and complete sampling because it is the only way that the specimen can be reconstructed to accurately measure the size of DCIS. Although this method is initially more time consuming and requires the use of radiography, it is the only way to ensure all radiologic abnormalities are sampled in a grossly normal specimen.

Some individuals may advocate measuring the largest extent on a single slide as an estimate of the size of DCIS. In the outside slide review study by Apple,17 4 (80%) of 5 cases that reported the size of DCIS documented only the maximum size measured on a single slide. In our study, the single-slide method was a good estimate of size when the extent of DCIS was small but was a very poor estimate for lesions greater than 2.0 cm in size. This method will miss large lesions. The largest slide method should only be used when DCIS involves a single block.

When specimen processing precludes measuring the size of DCIS histologically, the mammographic size is sometimes used. Mammographic size has been shown to correlate with histologic size but is often an underestimate13,24–26 and sometimes an overestimate.13 Holland and Hendriks24 showed that the mammographic-histologic difference was less than 2 cm in 80% to 85% of cases. Dillon et al13 showed that the mammographic size equaled the histologic size in only 11% of cases and that underestimations and overestimations of greater than 1 cm occurred in 25% and 17% of patients, respectively.

Although our study did not directly correlate mammographic size with histologic size, we did compare size with theoretical sampling of areas of calcification only with more extensive sampling (calcification vs SSS method). Similar to the mammographic-histologic correlation studies, our study showed that the size of DCIS based on sampling only areas of calcification was either equal to the SSS method or an underestimate. In 91% of our cases, differences were 2 cm or less; in 19% of cases, differences were greater than 1 cm.

Some previous studies have shown better mammographic-histologic correlation with the presence of comedonecrosis,25,26 whereas others have shown this is not the case when magnification views are used.24 In our study, the presence of comedonecrosis did not improve size estimation for the calcification method because virtually the same percentage of cases with and without comedonecrosis had differences within 0.5 cm of the SSS method (64% vs 65%). One possible explanation is that DCIS often shows mixed types, and the presence of comedonecrosis is not necessarily uniform throughout the lesion. Often cases showed comedonecrosis centrally and no calcifications at the limits of the lesion. These cases demonstrate that it is important to sample, at a minimum, the mammographically negative breast tissue flanking areas of calcification to be certain the entire extent of the lesion is represented.

In our study, the calcification method did not produce any overestimates because of experimental design. Some areas of calcification were sampled but did not correspond to DCIS but were, rather, associated with benign epithelium, stroma, and blood vessels. For the calcification method, the size was calculated based on areas that corresponded to DCIS only (Figure 1). Various types of calcification have been described and correlated with histologic findings in DCIS, and these can be used to distinguish benign from malignant types of calcifications.24,27 In our study, all calcifications were sampled regardless of appearance because imaging and interpretation of the sliced specimen radiograph is routinely done in the pathology department at our institution.

The blocks method involved multiplying the number of positive blocks by 0.3 cm, a proposed average tissue-section thickness.18,19 As shown in Figure 4, the relationship between the number of positive blocks and size by the SSS method is best when the number of blocks is less than approximately 10. In our study, the average slice thickness was 0.5 cm, which is similar to other reports.7,25,28 As shown in the Table, when this value is used as the multiplier, the size estimates are closer to the SSS method but overestimates size when the number of positive blocks is greater than approximately 8 to 10. If data points with more than 8 positive blocks are eliminated, the best-fit linear relationship is obtained using 0.47 cm as the multiplier. Similarly, if 10 positive blocks are used as the cutoff, the best fit multiplier is 0.45 cm. It is also possible to derive an adjustment to the formula to be applied when the number of blocks involved is more than 8 or 10, to obtain a curved relationship. All such formulas derived from our data are applicable only to our data set and would require validation with other data sets and other institutions to be applicable for general use.

Poor correlation between size by the SSS method and the number of positive blocks when the number of blocks is large is explained by the SSS method measuring the greatest linear extent, whereas the blocks method gives an estimate of the volume when DCIS occupies a large space and is sampled across that space. When DCIS spans in a linear fashion, within the volume of a block, the blocks method is a good estimate, particularly when adjusted for the tissue thickness at one's institution. When correlated with mammographic findings and a diagram indicating the location of the sections taken, a multiplier could be used as an estimate of tissue slice thickness, and the extent could be calculated by multiplying the number of slices involved by this multiplier. Without the knowledge of the location of blocks within a specimen, the blocks method is at risk of false estimations because multiple blocks may be taken to encompass a lesion of large volume, but smaller greatest-linear extents or too few blocks may be sampled from a large lesion.

In our study, 14 cases had more than 10 positive blocks and only 2 of these cases measured less than 4 cm by the SSS method. Therefore, it may be argued that, when sampled carefully, an accurate measurement of cases with greater than 10 positive blocks is not necessary because most of these cases will be of large volume DCIS, not amenable to breast-conserving surgery.

In our study, the SSS method was used as the gold standard because it is the most thorough method of sampling. To determine the true gold standard method, a study with outcomes would be required but is not feasibly ethically. Therefore, although size of DCIS has been shown to predict local recurrence,10–12 is associated with a higher rate of positive or close margins,13 and is an independent predictor of residual disease,13–16,22 the method by which that size is determined has not been correlated with outcomes. Although all size determinations are estimates, it is important that this information be included in the pathology report and reported as accurately as possible.

In conclusion, determining the size of DCIS by any method is an estimation. But given its clinical implications, it is important that this measurement be as precise as possible, particularly for lesions less than 4 cm. We recommend that the single-slide method be used only when DCIS is limited to a single slide. If the blocks method is used, a basic diagram of where the blocks were taken is necessary to avoid overestimating linear extent in large volume DCIS. Both the calcification method and the blocks method frequently give substantial underestimates and/ or overestimates to a degree that could affect clinical decisions. The SSS method is thorough and, although thought by some to be too expensive, is the standard protocol at many institutions. In our experience, once this method is in place, it requires little additional time and greatly enhances one's understanding of disease within the breast excision specimen.

We thank Pathologists' Assistants Alan Wolff, MLT; Sarah James, MLT; Brian Chow, BSc, MLT; and Oliver Pangan, BSc, MLT, in the Department of Pathology and Laboratory Medicine at Mount Sinai Hospital, for their role in specimen processing.