Quality Assurance in Breast Pathology: Lessons Learned From a Review of Amended Reports Open Access

All amended pathology reports for breast surgical specimens in the Department of Pathology at Brigham and Women's Hospital (Boston, Massachusetts) from January 1, 2009, to December 31, 2013, were retrospectively identified from the departmental database of cases. The study specimens included core needle biopsies, excisions, reexcisions, and mastectomies and excluded consultation cases. Addendum reports, presumed to include added information only, were not reviewed.

The classification of amended reports was modeled after the taxonomy of defects previously established by Meier et al¹  and employed in the College of American Pathologists Q-Probes study by Volmar et al⁷  with some adaptations. The 4 categories included (1) report defects, (2) specimen defects, (3) misidentifications, and (4) misinterpretations (Table 1). An amendment was classified as a report defect (ie, a typographical error) only after review of the case suggested that it would not be better classified as 1 of the other 3 types of defects. Differing from this established classification system, we divided misinterpretations into major diagnostic changes, minor diagnostic changes, and additional diagnostic information. Major diagnostic changes were defined as those with the potential to alter patient care, whereas minor diagnostic changes included all others. Major diagnostic changes were further characterized as upgraded, downgraded, and changed diagnoses.

A special emphasis was placed on characterizing major diagnostic changes by investigating their mechanism of detection and their potential impact on clinical care. The mechanism of detection was found by review of the explanatory note in the amended report. If not available, the information was obtained from the original pathologist. The potential impact on clinical care was inferred from a review of clinical notes in the electronic medical record.

During the 5-year period, 122 amended reports were identified from 12 228 breast pathology reports, yielding an amendment rate of approximately 10 per 1000 reports (1%). Most (88 cases; 72%) amendments were due to noninterpretative errors. Of the 88 cases, there were 58 report defects (63.8%; 37 typographical errors and 21 errors in nondiagnostic information), 12 misidentifications (13.6%), and 3 specimen defects (3.4%) (Table 1). Misidentifications included incorrectly specified laterality in 10 cases and anatomic site in the breast in 2 cases. All but one case were detected in a timely fashion, most commonly by a radiologist (8 of 12 cases; 66.7%) or radiologic technologist (1 of 12 cases; 8.3%). None of these cases were associated with clinical consequences. Specimen defects included inadequate sampling of an excisional biopsy without correlate for a mass lesion, inadequate sampling of lymph nodes in an axillary dissection, and discrepancy between margins reported and those intended by the surgeon for an excision specimen.

A few (34 cases; 27.9%) amendments were classified as misinterpretations, including 14 major diagnostic changes (11.5% of all amendments) and 13 minor diagnostic changes (10.7% of all amendments) (Table 1). Only 13 (0.17%) of 7647 breast cancer diagnoses rendered during the 5 years were subject to major changes. A single major diagnostic change involved a nonneoplastic diagnosis. Major diagnostic changes included 7 upgraded, 5 downgraded, and 2 changed diagnoses. Upgraded diagnoses consisted of “missed diagnoses,” primarily cases of missed microinvasion and small foci of invasion (up to 0.3 cm) in ductal carcinoma in situ (DCIS), as well as a single case of a missed micrometastasis (0.15 cm) in a sentinel lymph node. Downgraded diagnoses included 3 cases (for 2 patients) of DCIS involving sclerosing adenosis that was mistaken for invasive ductal carcinoma, and 2 cases of atypical ductal hyperplasia that was overcalled as DCIS. Changed diagnoses included a case of amyloidosis originally diagnosed as fat necrosis and a case of diffuse, large B-cell lymphoma originally diagnosed as “triple-negative” carcinoma (negative for estrogen receptor, progesterone receptor, and HER2) (Table 2). The 14 cases with major diagnostic changes were reported by 9 different pathologists. Both subspecialty breast pathologists (8 cases; 57.1%) and general surgical pathologists (6 cases; 42.9%) and both senior pathologists (>10 years in practice) (10 cases; 71.4%) and junior pathologists (<5 years in practice) (4 cases; 28.6%) were responsible for these errors. The average time in practice at the time of the amendment was 16.0 years.

Major diagnostic changes were discovered by a pathologist within the department in all 14 cases (100%). Most cases (9 of 14; 64.3%) were detected at the time of immunohistochemical study interpretation for estrogen, progesterone, and HER2 receptors. All missed foci of microinvasion and invasion, both cases of DCIS downgraded to atypical ductal hyperplasia, and a case of DCIS involving sclerosing adenosis on surgical excision were identified in this manner. All amended diagnoses were valid for the original hematoxylin-eosin levels as well as for the hematoxylin-eosin slide cut as a companion to the immunohistochemical studies. For 3 of the 14 cases (21.4%) that were misinterpreted on core biopsy, DCIS involving sclerosing adenosis (2 cases), and lymphoma, the correct diagnosis became apparent with evaluation of the entire lesion at surgical excision. In addition, the diagnosis of amyloidosis was realized through intradepartmental consultation and further workup completed within few days of sign out, whereas the missed micrometastasis was found on cytokeratin immunostain received shortly after sign out on the same day. No cases were discovered by conference review or by second opinion review during the study period. The median time to detection was 4.6 days, ranging from 0.27 to 78.9 days; the longer time intervals reflect misdiagnoses revealed at surgical excision (Table 2).

All major diagnostic changes discovered at the time of receptor studies were revised in a timely fashion and were not associated with clinical consequences. Major diagnostic changes were associated with clinical consequences in 3 cases, all misdiagnosed on core biopsy and discovered at surgical excision. An unnecessary sentinel lymph node biopsy was performed in 2 cases of DCIS involving sclerosing adenosis mistaken for invasive carcinoma. An unnecessary surgical excision and a sentinel lymph node biopsy were performed in the case of lymphoma mistaken for carcinoma. Of note, for these 3 cases, the original pathologist also reported the results of the receptor studies, emphasizing the value of second review in preventing error. Interpretative errors with clinical consequence comprised 3 of 34 interpretative errors (8.8%).

Some of the cases involved in major diagnostic changes were challenging cases that represent potential diagnostic pitfalls for the surgical pathologist. These cases warrant further consideration because they highlight the importance of generating a differential diagnosis in specific circumstances and suggest pathologic features helpful in making the correct diagnosis.

A core needle biopsy of a 1-cm, circumscribed mass in the right breast showed florid sclerosing adenosis with a complex architectural pattern of densely packed glands as well as cords and single cells in few sclerotic areas. The lesional cells contained apocrine cytoplasm and marked cytologic atypia (Figure, A). The lesion involved most of the tissue cores, which precluded appreciation of a lobulocentric configuration. The initial pathologist rendered a diagnosis of invasive carcinoma. Subsequently, the patient underwent wire localized excision and axillary sentinel lymph node biopsy, which revealed a mass with highly complex architecture and marked cytologic atypia. At that time, a second pathologist diagnosed the lesion as invasive carcinoma. However, at the time of immunohistochemical study interpretation for receptors, a third pathologist noted the circumscribed border of the lesion and questioned the original diagnosis. Double immunostains for p63/cytokeratin and myosin/cytokeratin confirmed that the myoepithelial cell layer was present throughout the entire lesion (Figure, B). Diagnoses for both the core biopsy and surgical excision were amended to DCIS involving sclerosing adenosis after an interval of approximately 2 months.

A core needle biopsy of a palpable mass showed nests, cords, and single epithelioid cells with moderate, pale cytoplasm and vesicular nuclei, considered most consistent with invasive carcinoma (Figure, C). The tumor had a triple-negative profile on immunohistochemical studies for estrogen receptor, progesterone receptor, and HER2. A wire localized excision and sentinel lymph node biopsy was performed. The pathologist for this case, who had expertise in hematopathology, noted that the tumor consisted of sheets of dyshesive, polygonal cells with large vesicular nuclei and prominent nucleoli, worrisome for lymphoma (Figure, D). Further workup confirmed this suspicion and revealed the error on the prior core biopsy approximately 3 weeks after the diagnosis of carcinoma.

The patient originally presented with 2 adjacent palpable irregular masses, 1.2 cm and 1.3 cm, in the upper inner quadrant of the left breast. The lesion was diagnosed as hyalinized fibrous tissue with calcifications on core needle biopsy and as hyalinized fibroadenoma and adjacent breast tissue with fat necrosis on the subsequent excision. A few years later, the patient presented once again with a palpable and ill-defined mass in the region of the prior excision. On core needle biopsy, a third pathologist diagnosed fibroadenomatoid change and fat necrosis with associated calcifications. However, that pathologist revisited the case shortly after sign out and noted an amorphous quality to areas initially appearing hyalinized, raising concern for amyloidosis (Figure, E). Multiple colleagues agreed with this suspicion on intradepartmental consultation, and further workup with Congo red and sulfated Alcian blue special stains confirmed the diagnosis (Figure, F). An amendment was issued, and previous specimens were reevaluated, revealing that amyloidosis had been missed in those as well. Fortunately, clinical evaluation revealed localized amyloidosis, light chain type, without associated systemic disease or malignancy, so the interpretative errors were without consequence.

Although it is common practice for pathology departments to perform reviews of amended pathology reports as part of quality assurance and improvement activities, few studies of amended reports have been published in the literature to date. To our knowledge, this is the first study to investigate amended pathology reports specific to breast specimens at a single institution and, as such, to compare quality measures with those from multi-institutional survey data.^7,9  Our study provides further insight into 2 types of error that warrant special consideration in breast pathology: specimen misidentifications and misinterpretations. More specifically, we found that laterality reporting errors involving core needle biopsy specimens were most often discovered by the radiology team soon after release of the final pathology report and were not associated with clinical consequences, highlighting the importance of radiology-pathology correlation and interdisciplinary communication. Moreover, we found that major interpretive errors were detected intradepartmentally by pathologists, most commonly at interpretation of immunohistochemical studies for estrogen receptor, progesterone receptor, and HER2.

In our practice setting, breast specimens are handled by both general surgical and breast pathologists, whereas biomarker studies are read and reported as an addendum by the breast pathology subspecialty service, frequently by a pathologist other than the one who issued the original diagnosis. In this study, all of the major interpretative errors discovered at the time of biomarker studies were detected in a timely manner and were not associated with clinical consequences. However, major interpretative errors were associated with clinical consequences in 3 cases, and for each of these cases, the original pathologist also reported the results of biomarker studies. These cases were regarded by members of our group as particularly challenging cases that represent known diagnostic pitfalls to the surgical pathologist and that may have been recognized as such if reviewed by a second pathologist. Specialist and generalist pathologists alike, both senior and junior, were vulnerable to interpretative errors. Therefore, any time that a case is revisited for ancillary studies or other reasons, we suggest that pathologists remain mindful of the opportunity to reconsider one's own diagnosis or to provide a second opinion on another's diagnosis. Within our own division of breast pathology, the way in which we approach the interpretation of these studies has changed to reflect this philosophy, although we have not implemented a formal policy of double review because it is not entirely compatible with our workflow. Nevertheless, we believe that double review may be a strategy for error reduction in some practice settings.

Our study builds on recent studies of amended reports that established standardized methodology for data collection and benchmarks for quality measures. A recent College of American Pathologists Q-Probes study of surgical pathology report defects at 73 institutions found a median amendment rate of 5.7 per 1000 (10th to 90th percentile range, 13.5–0.9) with higher rates at institutions with training programs and with policies for review of a set percentage of cases after sign out. Report defects comprised the greatest fraction of all defects. Misinterpretations and specimen defects were most often discovered by pathologists, whereas misidentifications were most often discovered by clinicians. Defects were detected through a number of mechanisms, including clinician-requested review (11.4%), ancillary studies (9.7%), additional clinical information (5.5%), tumor board or case conference review (3.6%), by chance (3.3%), and through extradepartmental second opinion (2.2%), among others. Similar to our study, it appears that review of ancillary studies after sign out was one of the more common mechanisms of error detection, although further information was not provided regarding the types of errors, specimens involved, or ancillary studies performed. As previously mentioned, breast was one of the most common organ sites with total amendments, misinterpretations, and misidentifications, second only to skin.⁷ 

Meier et al^1,8  reported a single institution experience in the development and validation of amended report taxonomy and implementation of a real-time amended report-monitoring system and lean process-improvement program. Amendment rates increased from 2.6 to 4.8 per 1000 during the validation period to 10.1 per 1000 during the initial phases of monitoring and subsequently decreased to 5.6 per 1000 once lean processes took hold. One of the specific interventions introduced was focused double review of breast and prostate cases because these cases together comprised 2% of accessions but 20% to 40% of misinterpretations. This intervention was associated with a marked reduction in misinterpretations, such that the fraction of amendments for misinterpretations fell from 18% to 3%. Although focused double review was an effective pre–sign-out intervention for interpretative error reduction, the authors also emphasized the importance of conference review as a post–sign-out mechanism of detection of interpretative errors (80%) and detection of all types of error (10%–20%).⁸  In comparison to these 2 prior studies, we found a higher amendment rate, which likely reflects exclusive focus on breast specimens, an academic practice setting, and frequent second review of cases at the time of ancillary studies.

A recent evidence-based guideline by the College of American Pathologists Pathology and Laboratory Quality Center and the Association of Directors of Anatomic and Surgical Pathology emphasized the value of double review in surgical pathology and cytology. A review of the literature revealed numerous studies in which review of cases by a second pathologist detected disagreements and potential errors. The error rates varied widely, ranging from 0.1% to 10%, depending on the method of review and the types of cases.¹⁰  It is recommended that anatomic pathologists formalize procedures for the review of selected cases to detect interpretative errors and that these reviews should be performed in a timely manner to avoid negative impact on patient care. Because reviews may affect turnaround time, workload, and expense, the ideal approach may depend on the practice setting. Prospective reviews (before sign out), such as double reading, intradepartmental or extradepartmental consultation, or consensus conferences, facilitate the correct treatment plan, minimize confusion, and build confidence and trust. Some studies have shown that prospective review of cases is associated with a decrease in amended reports, particularly those for diagnostic edits.^11–14  However, retrospective reviews shortly after sign out, such as those for conferences, are well-established approaches to ensuring the correct diagnosis and may be more practical in many settings. Our study suggests that a second review at the time of predictive factor reporting could be another valuable strategy for retrospective review of all new breast cancer diagnoses in the appropriate practice settings and does not increase the overall workload.

Our study of amended pathology reports also contributes to an understanding of diagnostically challenging cases in breast pathology. Three cases of DCIS involving sclerosing adenosis were misinterpreted as invasive ductal carcinoma, and these cases represent particularly difficult examples of this type of lesion. One case (2 specimens) consisted of florid adenosis involved by high-grade apocrine DCIS, whereas the other case consisted of sclerosing adenosis with an infiltrative growth pattern and partial involvement by intermediate-grade DCIS. A key to the diagnosis of adenosis is recognition of a lobulocentric pattern of growth at low magnification; however, in some cases, it can be difficult to appreciate this pattern in limited biopsy samples. Therefore, it is important to maintain a low threshold for the use of immunohistochemical studies for myoepithelial cell markers (p63, smooth muscle myosin, calponin, among others) to distinguish adenosis, especially that involved by DCIS, from invasive carcinoma. Another major diagnostic change involved misinterpretation of diffuse, large B cell lymphoma as triple-negative breast cancer. It underscores how easy it is to default to a diagnosis of invasive mammary carcinoma in the setting of clinical presentation with a breast mass and malignancy on biopsy. It is important to consider other entities in the differential diagnosis of a triple-negative, poorly differentiated tumor, such as lymphoma, melanoma, or metastatic carcinoma, and to judiciously perform immunohistochemistry for cytokeratin, and if necessary, a broader panel of lineage-specific markers. Finally, a case of amyloidosis of the breast misinterpreted as fat necrosis on multiple specimens highlights a rare diagnosis that, if missed, may have clinical consequences for the patient. Amyloidosis of the breast is the amyloidosis, light chain, type in most patients and is associated with systemic disease in close to one-half of patients and is associated with hematologic disease in more than one-half of patients.^15,16  The differential diagnosis of fat necrosis versus amyloidosis is a difficult one, given the relative rarity of the latter; however, a Congo red special stain is a relatively inexpensive way to avoid a missed diagnosis in the presence of any suspicion.

Review of amended pathology reports has been shown to be a useful tool for the analysis of errors in surgical pathology and can be performed relatively easily and quickly compared with other methods of error analysis. Much larger, more labor intensive and costly studies are required to detect all possible errors. For example, benign diagnoses are seldom subjected to second review or other mechanisms of error detection, so the rate of underdiagnosis of atypical or malignant lesions is largely unknown. In the context of breast specimens, this is interesting because much of the interobserver variability is related to atypical diagnoses, rather than to cancer diagnoses. The study of addendum reports and extradepartmental second-opinion reports could also be included in such studies. Inappropriate amendment and addendum labeling can lead to underestimated amendment rates and, most concerning, underestimated rates of misinterpretation error. Published audits of addendum reports indicate that approximately 5% to 10% of addendum reports should have been amendments.^8,17  In addition, we did not separately review extradepartmental second-opinion reports to confirm whether any differences of opinion would not be better classified as errors requiring amendment.

In summary, a review of amended pathology reports provides valuable insights into surgical pathology processes, not only reminding us of common-sense good practices and diagnostic pitfalls but also suggesting strategies for error detection and reduction. Our study reveals that second review at the time of ancillary studies is a potentially valuable approach to the detection of interpretive errors in a timely fashion and prevention of associated clinical consequences.