Accuracy of Intraoperative Frozen-Section Analysis of Breast Cancer Lumpectomy-Bed Margins

Transcription

1 Accuracy of Intraoperative Frozen-Section Analysis of Breast Cancer Lumpectomy-Bed Margins Juan C Cendán, MD, FACS, Dominique Coco, MD, Edward M Copeland III, MD, FACS BACKGROUND: STUDY DESIGN: RESULTS: CONCLUSIONS: My colleagues and I have been using intraoperative frozen-section analysis (FSA) to evaluate lumpectomy margins in an attempt to reduce the number of additional operations that patients with ductal carcinoma in situ or stage I and II breast cancer would have to endure. We review our experience in breast-conservation therapy (BCT) at the University of Florida (Gainesville) to determine the effectiveness of this approach. Operative reports, operative logs, and pathology reports were retrospectively reviewed for patients who had BCT from January 2001 to January Ninety-seven patients (116 operations) were reviewed. Nineteen patients required an additional operation (19.6%). Forty-three patients had positive margins on paraffin-embedded histologic analysis (44.3%). Accuracy of FSA was 84% when evaluated on a per-case basis, and 96% on a per-slide basis. False negatives were identified in 22 patients, affecting the operative pathway of 19 patients (19.6%) and were identified more frequently in cases of ductal carcinoma in situ (p 0.001). There were no false positives. Additional operative time required for FSA was approximately 13 minutes per case. Eighty-four (86.6%) patients had successful BCT and 13 patients (13.4%) required mastectomy. Intraoperative analysis of margins using FSA is effective at minimizing the number of additional operations, with 19 patients benefiting from immediate intervention in this study. The authors believe that the number of second operations prevented and the high BCT rates justify performing FSA. Ductal carcinoma in situ is more difficult to identify in FSA. Preoperative discussions with the patient should reflect these findings. (J Am Coll Surg 2005;201: by the American College of Surgeons) Breast-conservation therapy (BCT) is now a standard therapeutic option for stages I and II invasive carcinoma and ductal carcinoma in situ (DCIS). Numerous followup studies after the original work of Fisher and colleagues 1 and Veronesi and colleagues 2 have correlated the recurrence of invasive carcinoma and DCIS with the presence of tumor at or near a lumpectomy margin. The standard practice of evaluating breast margins involves a histologic review of tissue embedded in paraffin and stained with hematoxylin and eosin ( permanent section ). This method provides an acceptable histologic definition of both glandular and adipose tissue. Unfortunately, time requirements for processing tissue in formalin and embedding in paraffin does not allow Competing Interests Declared: None. Received January 31, 2005; Accepted March 18, From the Department of Surgery (Cendán, Copeland) and Department of (Coco), University of Florida, College of Medicine, Gainesville, FL. Correspondence address: Juan C Cendán, MD, Department of Surgery, University of Florida, Box , JHMHC, Gainesville, FL for its use intraoperatively. This potentially limits the initial operation by requiring patients to endure an additional operation for reexcision if the margins happen to be involved. Freezing tissue in a glycol-based polyethylene-embedding compound (frozen-section analysis [FSA]) is an alternative embedding method that can be performed in a relatively short period of time and allows for its use intraoperatively. FSA has historically been accurate in identifying the pathologic etiology of palpable and nonpalpable breast masses. 3 Currently, there are fe w studies addressing the accuracy of FSA in evaluating breast margins. Adipose tissue is difficult to freeze and pathologists generally agree that frozen breast margins are difficult to interpret because of the additional challenge of identifying a smaller focus of tumor in a bed of adipose tissue. The degree of inaccuracy is largely unknown. The University of Florida in Gainesville has been evaluating breast margins with FSA over the last 10 years under the hypothesis that the accuracy of FSA is sufficient to reduce a sub by the American College of Surgeons ISSN /05/$30.00 Published by Elsevier Inc. 194 doi: /j.jamcollsurg

2 Vol. 201, No. 2, August 2005 Cendán et al Intraoperative Analysis of Lumpectomy Margins 195 Abbreviations and Acronyms BCT breast-conservation therapy DCIS ductal carcinoma in situ FSA frozen-section analysis MRM modified radical mastectomy stantial number of additional operations to justify its use. We retrospectively review our institution s records over the preceding 3 years to evaluate the efficacy of this method. METHODS Identification of patients, demographic data, and conduct of operations Patients with stage I or II invasive breast cancer or DCIS who underwent BCT at the University of Florida in Gainesville between January 1, 2001 and January 2004 were identified through an IRB-approved database search (University of Florida IRB no. 623; 2003). Only female patients were included. Age, date of the procedure, preoperative and postoperative diagnosis, tumor size and location, frozen-section diagnosis, and number of frozen sections performed were obtained by reviewing operative and pathology reports. All patients had a tissue diagnosis before their operation. The possibility of proceeding to a modified radical mastectomy was discussed with patients before their operation. Two of the authors (EMC, JC) were involved in all of the operations reviewed. A sentinel lymph node biopsy was included in every operation after the lumpectomy for those with invasive cancer and a select number of those with DCIS. All instrumentation and gloves were changed before reexcising margins, if a reexcision was performed. Between five and six margins were obtained separately from the lumpectomy cavity in an en-face fashion (lateral, medial, inferior, superior, deep, and superficial) and sent to pathology for evaluation. Margins were not obtained from the lumpectomy specimen. En-face margins were embedded in a glycol-based polyethylene-embedding compound and quickly frozen in an isopentane bath at 60 F. Frozen blocks were then sectioned using a Shandon Cryotome, stained with an hematoxylin and eosin, and then evaluated by a University of Florida pathologist assigned to interpret all frozen sections that day. Comparison of FSA and permanent sections After the evaluation and rendering of the diagnosis, the frozen tissue was fixed in formalin and routinely processed to paraffin-embedded blocks. Frozen-section diagnoses were compared with the permanent pathology slides and categorized as true positive, true negative, false positive, or false negative, accordingly. Sensitivity, specificity, positive, and negative predictive values for the FSA diagnoses were then calculated according to all the margins performed, irrespective of clinical outcomes and then with respect to clinical outcomes. For example, if a false-negative diagnosis were given to one of six margins performed in a single case, FSA would be considered inaccurate for the entire case, despite making five correct diagnoses on the other margins. Statistical evaluation Overall incidence of cancer at a margin was used to establish a baseline expected proportion. Presence of cancer at the margins for invasive ductal and lobular cancer and DCIS was compared with the null hypothesis that all tumor types would be found at a margin equally. The hypothesis was tested using a two-tailed chi-square analysis and significance was defined as p Similarly, the FSA results for each individual pathologic diagnosis were tested against the permanent pathology using a two-tailed chi-square analysis to determine if there was any marked difference between the FSA and permanent results, dependent on the pathologic subtype. Where appropriate, descriptive data are presented as mean 2 SD. Statistical analysis was performed using the SISA web-based package (Vitenbroek DG, SISA Binomial 1997, available at: index.htm, accessed on October 9, 2004). RESULTS Demographics Average patient age was 59.4 ( 11.4) years old. Average tumor size was 1.46 ( 1.4) cm. There were 57 (59%) needle-localization guided excisions and 40 (41%) open excisions. Preoperatively identified tumor characteristics are shown in Table 1. Of 97 patients reviewed, 54 (55%) had true-negative margins on FSA and permanent sections on the initial excision and required no additional operative intervention. Forty-three patients had at least one positive margin on permanent pathology (44%). Twenty-five of these 43 patients had at least one true-positive margin on

3 196 Cendán et al Intraoperative Analysis of Lumpectomy Margins J Am Coll Surg Table 1. Preoperatively Identified Tumor Types n Infiltrating ductal 57 Infiltrating lobular 7 Ductal carcinoma in situ 33 FSA, which resulted in additional intervention at the time of initial lumpectomy (25.8%). Of particular importance, 18 patients had a false-negative margin on initial excision, requiring an additional delayed operation (18.6% false negative, Fig. 1). True positives Of the 25 true-positive patients, 2 underwent immediate modified radical mastectomy (MRM) because of multiple positive margins and the remaining 23 underwent reexcision of the positive margins, in an immediate attempt at BCT. True-negative margins were achieved in 14 patients on immediate reexcision and BCT was achieved. In 5 of the patients, reexcision also yielded true-positive margins and in those patients, 2 underwent immediate MRM and 2 underwent definitive, delayed 54 TN BCT achieved 2 MRM 97 Patients 43 Tumor on permanent section on any margin. (TP) 25 TP on FSA 18 FN* SEE 2nd Tier Below 23 procedures (one skin-sparing and one MRM). In 1 patient, BCT was achieved by reresecting a true-positive margin. Four patients had false-negative margins and are repor ted here (Fig. 1). False negatives On the initial excisions, 18 patients had at least a single false-negative margin, prompting consideration for a second, separate operation. An additional 4 patients had a false-negative margin on immediate reexcision. Three patients did not pursue or require additional operations. One patient with DCIS chose not to reoperate because of advanced age and comorbidities. Two patients had a margin involved with lobular carcinoma in situ and did not warrant reexcision. The remaining 19 patients underwent additional operations; 7 had MRM, and 12 achieved BCT using FSA; of this group, 5 patients had true-positive results at the time of repeat lumpectomy, allowing additional margins to be taken until a negative result was obtained. Patient group benefited by FSA Twenty-five patients benefited from FSA with key intraoperative decisions made based on the finding of cancer at a margin. Of these, 19 patients with tumor-bearing margins did not require a second trip to the operating room because FSA allowed for either true-negative margins to be obtained (15 patients), or mastectomy be performed (4 patients), at the initial operative setting. The remaining 6 patients returned to the operating room at a separate time to allow for reconstruction (2 patients) or because of false-negative margins during the reexcision (4 patients). 14 TN BCT Achieved 5 TP (4 MRM, 1 BCT) 9 Tumor on permanent 2nd Tier of Surgery 4 FN** Histologic evaluation Six hundred twenty-eight margins were evaluated by FSA and permanent section (5.4 margins per operation). Ninety margins were tumor-bearing on permanent pa- 18 FN* + 4 FN** 19 7 MRM 12 BCT 3 dropouts Figure 1. Clinical impact of frozen-section analysis on patients. BCT, breast-conservation therapy; FN, false negatives; FSA, frozensection analysis; NRM, modified radical mastectomy; TN, true negatives; TP, true positives. *FN from first tier of operations. **FN from second tier of operations. Table 2. Number of Slides Revealing Tumor by Cell Type on Frozen-Section Analysis and Permanent FSA Permanent pathology Infiltrating ductal Infiltrating lobular DCIS Lobular carcinoma in situ 2 3 Mucinous carcinoma 2 Total 57 90

4 Vol. 201, No. 2, August 2005 Cendán et al Intraoperative Analysis of Lumpectomy Margins 197 Table 3. Number of Slides Containing Tumor by Type Total no. of slides Tumor present on FSA (n) Tumor present on permanent pathology n % Infiltrating ductal Infiltrating lobular DCIS *Chi-square analysis reflects a difference in the anticipated distribution of these cases, p 0.001, chi-square thology. Table 2 shows the distribution by carcinoma type. Fifty-seven of the 90 tumor-bearing margins (63.3%) were diagnosed correctly by FSA. When separated by carcinoma type, the distribution is shown in Table 3. Of the three main tumor types, invasive lobular cancer was most frequently tumor-bearing at the margins (18 of 45 slides, 40%); DCIS followed in frequency with 51 of 211 slides (24.2%); and invasive ductal cancer was least frequently tumor-bearing (16 of 385 slides, 4.2%). The difference between these groups was considerable (chi-square 83.7, p 0.001). Difference between FSA and permanent sections of the invasive cancers was not substantial, but DCIS was missed more often on FSA when compared with the invasive cancers on permanent sections, p ( Table 4). Accuracy of the procedure Six hundred twenty-eight pathologic slides were evaluated for the correlation between the FSA and permanent histology; accuracy was 96% (95% CI, %) with a sensitivity of 65%, a negative predictive value of 94%, and 100% specificity and positive predictive value ( Table 5). The 97 patients under went 116 separate operations, yielding 134 different sets of slides for FSA; on a per case basis accuracy was 84% (95% CI, %) with sensitivity of 59%, a negative predictive value of 78% and 100% specificity and positive predictive value. DISCUSSION Ninety margins were tumor-bearing on paraffinembedded analysis. Fifty-seven of those 90 were accurately diagnosed on FSA (63.3%). FSA accuracy in diagnosing breast margins has been largely unstudied and has not gained widespread clinical acceptance. Cox and colleagues 4 repor ted an FSA sensitivity of 77% and Table 4. Analysis of Margin Involvement Accuracy Stratified by Tumor Type Observed on FSA Expected (permanent) Chi-square, two-tailed p value Infiltrating ductal Tumor-bearing No tumor Infiltrating lobular Tumor-bearing No tumor DCIS Tumor-bearing No tumor specificity of 100% on 114 breast margins evaluated. Sauter and colleagues 5 repor ted 90% sensitivity and 97% specificity on 359 margins evaluated by FSA and Weber and colleagues 6 repor ted a sensitivity of 91% and specificity of 100% in their study. Our results more closely resemble that of Cox and colleagues 4 a n d we we re unable to achieve the high sensitivity reported by Weber, Sauter and colleagues 5,6 ; overall accuracy in this study was excellent (84% to 96%, depending whether the analysis is on, respectively, a per case or per slide basis). In our study, 24% of patients required immediate reexcision at the time of the initial lumpectomy. This is slightly less than that defined by Pinotti and Car valho 7 who reported a 40.8% immediate reexcision rate; their technique also incorporated simultaneous combination of radiologic and histopathologic evaluation and their initial margins may not have been as wide as in our patients. Our technique does attempt to obtain a 1-cm margin around the tumor, especially when the tumor is palpable. The ill-defined nature of DCIS and invasive lobular cancer are reflected in our finding a tumor at the margins in these pathologic entities more frequently. Table 5. Accuracy of the Frozen-Section Analysis Method Described on a Per-Case and Per-Slide Basis Initial operation Overall based on case Overall based on slides N Sensitivity Specificity PPV NPV Accuracy (%), 95% CI 82, , , NPV, negative predictive value; PPV, positive predictive value.

5 198 Cendán et al Intraoperative Analysis of Lumpectomy Margins J Am Coll Surg Obtaining 100% specificity supports the belief that FSA is a safe method for interpreting margins by avoiding potential situations where additional breast tissue is removed unnecessarily. For this reason, our institution has been reluctant to try touch preparation cytology, given a number of studies that have had false-positive diagnoses. Another strength of this study was use of multiple general pathologists. We believe that this reflects a more common support environment for a practicing surgeon and will make the data more relevant outside of specialized breast care centers. From a clinical perspective, 22 of the 43 patients with positive margins (51.2%) had at lease one false-negative margin, either at the first excision or at reexcision. Two patients chose not to return to the operating room and 1 had lobular carcinoma in situ involving a margin. Nineteen patients had a second operation. Although an argument questioning use of FSA could arise from the fact that 51.2% of the patients with positive margins were incorrectly diagnosed; in our view, the more important fact is that critical intraoperative decisions were made based on the FSA in 25 patients, and 19 patients did not require a second operation all of whom would have required another operation without FSA. When FSA of margins is used, the patient must understand that the pathologic status of her margins will not be known until the permanent sections are available. On the other hand, this is true for all patients in whom FSA margins are not used. Those patients with DCIS and lobular invasive cancer are at the highest risk for needing an additional procedure because of a false-negative FSA. Likewise, the technique can be valuable when attempting BCT on large lesions in relation to breast size. If negative FSA can be obtained with a large lesion and the cosmetic result is acceptable, then BCT is achieved. If margins are positive or cosmetic result is poor, then the patient should be prepared for mastectomy. Many patients will take comfort in knowing that an attempt at BCT was tried and then will be more accepting of a mastectomy. A breast cancer that is so large as to preclude BCT and require mastectomy will, in our treatment algorithm, require radiation therapy, and precludes use of immediate reconstruction. FSA required 13 minutes, the time required to close the wound. If immediate reexploration was required because of a positive margin, the wound was reopened to do so. Important information was available intraoperatively in 25 of 43 patients; and, at a minimum, a second operation was avoided in 19 of 43 patients with tumor at the margins (44.2%), with no practical increase in operative time for those patients with negative margins on FSA. REFERENCES 1. Fisher B, Anderson S, Bryant J, et al. Twenty-year follow-up of a randomized trial comparing total mastectomy, lumpectomy, and lumpectomy plus radiation for the management of invasive breast cancer. N Engl J Med 2002;347: Veronesi U, Cascinelli N, Mariani L, et al. Twenty-year follow-up of a randomized study comparing breast-conserving surgery with radical mastectomy for early breast cancer. N Engl J Med 2002; 347: Ferreiro JA, Gisvold JJ, Bostwick DG. Accuracy of frozen-section diagnosis of mammographically directed breast biopsies: results of 1,490 consecutive cases. Am J Surg Pathol 1995;19: Cox C, Ku NN, Reintgen DS, et al. Touch prep cytology of breast lumpectomy margins with histologic correlation. Arch Surg 1991;126: Sauter ER, Hoffman JP, Ottery FD, et al. Is frozen section analysis of reexcision lumpectomy margins worthwhile? Margin analysis in breast reexcisions. Cancer 1994;73: Weber S, Storm FK, Stitt J, Mahvi DM. The role of frozen section analysis of margins during breast conservation surgery. Cancer J Sci Am 1997;3: Pinotti JA, Carvalho FM. Intraoperative pathological monitorization of surgical margins: a method to reduce recurrences after conservative treatment for breast cancer. Eur J Gynaecol Oncol 2002;23:11 16.