Second opinion (external specialist referral) practice of breast pathology: the Nottingham experience

Breast pathology is a challenging field, and discrepancies in diagnoses exist and can affect patient management. This study aims to review a breast referral practice and assess the pattern and frequency of breast lesions sent for an external expert review and evaluate potential impacts on patients’ care.


Introduction
The histological diagnosis of most breast lesions is straightforward. However, classification of breast lesions, which are recognised to show a wide spectrum of differentiation with complex morphology and overlapping features, can be subjective and is often based on pathologists' interpretation of multiple morphological features and judgement in applying the defining diagnostic classification criteria. This has been demonstrated in several previous studies examining interpretive classification of breast lesions, some of which have shown high levels of diagnostic discordance, 1-8 even among experts. [8][9][10] The inaccurate classification of breast lesions can adversely affect patient management decision-making for both surgery and systemic therapy and quality of life, as well as increase the personal and health system cost of inappropriate treatment in addition to its medicolegal implications.
Therefore, it is accepted good practice to refer rare, difficult and challenging cases or when there remains uncertainty to a specialist breast pathologist for a second 'primary' diagnostic opinion. Referral for specialists' opinion may not only improve diagnostic accuracy and the quality of patient care, but also provides educational value. The Royal College of Pathologists of the United Kingdom (UK), through their Good Pathology Practice guidance, recommend that all pathologists should actively participate in some form of referral practice as this is in the best interests of patients, provides continuing professional development (CPD), and is considered to be good practice. 11 Consultation practice includes internal consultation within the department (e.g. difficult case local discussion), informally with colleagues in adjacent hospitals, subspeciality tertiary referrals linked to patient pathways or routinely within cancer networks and formal external referral where a primary diagnostic or secondary specialist opinion is required. 11 However, the latter process remains sporadic, implemented on a case-by-case basis and based mainly on local pathologists' preferences. Pathologists tend to seek second opinions when they need more assurance on their interpretation and find the case challenging or showing borderline features. Several factors influence the referral practice among breast pathologists, including primary pathologist experience, breast workload in the department, the number of pathologists with interest in breast pathology in the department, the nature of the lesions (Table 1) and the availability and access to a specialist opinion practice. Patients can also seek a second medical opinion from a specialist. Recommendations from prior studies vary, 9,10,12 and data on the practice of external specialist opinion remain lacking.
In the UK, like most other countries, specialist opinion practice is variable between centres and is typically based on the pathologist's judgement for selecting the cases to be referred and to choose the external specialist referral centre. Some pathologists refer cases directly to experts they know personally or refer to the department in which well-known experts are based. The Nottingham Breast Pathology Unit, located at the Nottingham City Hospital, is one of the leading second-opinion referral practices with more than 40 years of experience in the field, and receives cases from the UK in addition to some other countries. In this study, we aimed to provide the Nottingham experience of specialist opinion referral practice to provide an overview on the type and frequency of referred breast lesions and the impact of such referral on the classification and patient care. Common diagnostic pitfalls are also highlighted to improve awareness of reporting pathologists to these entities.

Materials and methods
This study is based on 740 cases that were referred to the Nottingham University Hospitals NHS Trust, UK for a second opinion during a 4-year period (2019-22) and had slides and reports available for review. Cases with missing slides were excluded. Comparison of these reports was performed for evaluation of change in diagnosis. Data recorded included the reasons for referral, specimen type, the initial diagnosis, the proffered specialist opinion diagnosis, whether both diagnoses were concordant and the type of reclassification in addition to the country of the referral pathologists. Entities included in the concordance of classification included normal, benign non-atypical lesions, atypical lesion, malignant in situ and malignant invasion. Change in diagnosis and/or management was determined based on initial (referral) institution protocols and biopsy type (core or excisional). The lesions were classified based on the 5th edition (2019) of the World Health Organisation (WHO) Classification of Tumours Editorial Board: Breast Tumours. 13 Changes in diagnosis that led to changes in management were further evaluated, recorded and categorised.

Results
In this study 740 cases were identified and reviewed. Most cases were from the UK (92%; other countries included Peru, United Arab Emirates, New Zealand, United States and various European countries) and were from female patients (99%). Seventy per cent were excisional biopsy and 30% were needle core biopsies. Ten cases were sent from a different referral centre in the UK for further specialist opinion. In general, the lesions received can be classified into nine  (Table 1), although overlaps exist as rare and unusual lesions and those that comprise a continuous spectrum or show complex morphology are challenging. The main lesion types referred to Nottingham are summarised in Table 2.
The most frequent entities sent for a specialist opinion were papillary lesions (19%) and fibroepithelial lesions (17%). These included benign, atypical and malignant papillary lesions as well as benign, borderline and malignant fibroepithelial lesions. Cases of hamartoma and fibroadenoma were included in the fibroepithelial lesion category. Intraductal epithelial proliferations with atypia were also common (19%), and ranged from flat epithelial atypia, atypical ductal hyperplasia (ADH) and the distinction between ADH and low-grade ductal carcinoma in situ (DCIS), various subtypes of DCIS including neuroendocrine, basal-like and apocrine DCIS, and DCIS involving sclerosing lesions in addition to occasional cases of atypical adenosis, lobular carcinoma in situ (LCIS) of the pleomorphic and florid type, distinction between LCIS and DCIS, microinvasion or distinction between DCIS and invasive carcinomas in cases of DCIS-like ducts lacking expression of myoepithelial cell markers.
Another category included invasive carcinomas (14%) that were sent for confirmation of the invasive diagnosis or subtyping of the invasive tumour. These included 10 cases of low-grade adenosquamous carcinomas. Other cases included neuroendocrine carcinomas, high-grade metaplastic carcinoma that was difficult to differentiate from malignant phyllodes or sarcomas, distinction between ductal and lobular carcinomas with unusual E-cadherin staining patterns and subtyping of tumours following neoadjuvant therapy.
In 61 cases (8%) a diagnosis was not offered by the referring pathologists (n = 10), or an initial diagnosis was only suggested but did not include the final specialised diagnosis (n = 47 cases). The other four cases included two cases that were sent for interpretation, which was difficult secondary to poor fixation, one case was sent because no lesion was detected in the excision specimen despite there being a large lesion at the time of core biopsy, and no neoadjuvant therapy was given, while the last case was diagnosed as a B2 lesion (metaplastic changes) and was sent as an example for a case report, but upon review the final diagnosis was changed to a malignant entity.
Tumours with epithelial-myoepithelial phenotype (adenomyoepithelioma) and those showing salivary gland-like tumours, including pleomorphic adenoma and skin adnexal tumours, comprised 9% of the cases.
Fifty-nine cases (8%) were spindle cell lesions which included mesenchymal (fibroblastic, myofibroblastic, smooth muscle and neural lesions) and epithelial spindle cell lesions (spindle cell metaplastic carcinomas). Lesions of the nipple, including epidermal changes suspicious of Paget's disease, nipple adenoma and syringomatous tumour of the nipple, comprised 3% of the cases. Although complex sclerosing lesion/radial scar comprised 1.6% of the cases, more sclerosing lesions were included, but contained higher risk tumours such as DCIS or invasive carcinomas and these were classified as those higher-risk lesions. The least frequent category was vascular lesions of the breast, including atypical vascular lesions and angiosarcomas (1%). Representative images of some challenging cases that were sent for a specialist opinion are presented in Figures 1-4.

I M P A C T S O F T H E S P E C I A L I S T O P I N I O N O N T H E F I N A L C L A S S I F I C A T I O N O F T H E L E S I O N S
After reviewing the cases by the expert pathologists, the initial diagnosis was confirmed or one of the suggested diagnoses was preferred in 584 cases (79%), including 129 cases (17%) in which the final opinion resulted in some non-significant changes in the management, such as benign phyllodes tumour instead of fibroadenoma or adenomyoepithelioma instead of a papilloma. In atypical intraductal epithelial proliferation borderline between ADH and DCIS where both diagnoses were given, 10 cases were classified as DCIS and five as ADH.
Different classification to that initially offered or suggested was made in 132 cases (18%) which resulted in significant change in the management of the patients. These included 30 cases benign to malignant and atypia to invasive tumours, 22 cases benign to atypia or atypia to benign/hyperplastic lesions and 21 cases of malignant to benign or atypia/hyperplasia ( Table 3). The remaining cases included those which resulted in change of management such as fibroadenoma versus borderline phyllodes tumour, carcinomas versus sarcomas or papillomas versus atypical or malignant adenomyoepithelioma (Tables 4 and 5). All significant changes of classification of the lesions with an impact on the management were subjected to consensus discussion among the breast pathologists in the department. In this study, 20% of the cases were subjected to internal discussion among the Nottingham breast pathologists. These included cases with high levels of subjectivity or showing borderline features, as well as cases that show features significantly different from those displayed by the proffered diagnosis by the sending pathologist, or the change in the final classification can result in significant change in the management decision. Fourteen cases were inconclusive and the final opinion was obtained following further external expert opinions, mainly vascular lesions. In 10 cases, the specialist opinion was different from the submitted diagnosis, but there was no significant change in the management.
In more than 90% of the cases, comments on further management were included and, in most cases, explanation of the histological findings and their contribution to the final diagnosis and distinction from differential diagnoses was provided. Additional comments on receptor status, advice on quality assurance, information on the relevant immunohistochemical markers and further prognostic information were provided in some cases. At least 25% of the cases were discussed internally with breast pathology, soft tissue and lymphoreticular pathology or dermatopathology colleagues. These were challenging cases and cases with suspected primary breast-specific, soft tissue, lymphoreticular or cutaneous lesions, respectively. It was also found that most cases are completed within a week, except those cases that needed further immunohistochemistry (IHC) or needed specialised non-breast pathologist opinion (on average, not more than 2 weeks). Final reports of the cases sent for further external expert opinion were delayed up to 1 month, but such delays were infrequent.

Discussion
Assessment of certain morphological characteristics is known to be influenced by an element of subjectivity resulting in lower than ideal levels of diagnostic agreement, particularly in conditions showing overlapping features or forming part of a spectrum of entities. Adherence to published guideline recommendations, use of diagnostic criteria and use of immunohistochemistry can help to improve the consistency of reporting. 3,14-16 However, there remains a proportion of lesions that require a specialist opinion. These include rare lesions that a pathologist may rarely encounter in their routine practice, lesions known to have high levels of diagnostic discordance and lesions that show unusual morphology, immunoprofile or where the morphology and immunoprofile do not match, in addition to cases requiring advice on further management. Therefore, discussion with colleagues or seeking a specialist expert opinion in challenging cases for secondary or primary diagnosis can be considered as best practice 11,17 not only for patient care, but also for educational value and to avoid medicolegal litigation risk.
Double-reporting and second-opinion practice are variable, with different strategies and policies adopted by laboratories; in some laboratories double reporting is applied to all breast biopsy specimens or a targeted approach is adopted; for example, if initial interpretation is atypia versus malignant, or DCIS versus invasive carcinoma, or if initial interpretation is invasive carcinoma only. [18][19][20] In a previous study it was reported that laboratory policies mandating internal double reporting varied by diagnosis as follows: invasive cancer 65%; DCIS 56%; atypical ductal hyperplasia 36%; and other benign cases 33%; 81% obtained second opinions in the absence of policies. 12 In a survey of > 250 pathologists, participants believed that second opinion improved diagnostic accuracy (96%), protects from malpractice suits (83%), is easy to obtain, did not take too much time and did not compromise their professional confidence. 12 Other quality assurance methods to ensure accurate diagnosis of difficult breast lesions include consensus meetings, a process in which a difficult case is discussed by multiple pathologists to reach a consensus over a diagnostic opinion before a final report is issued, and which should be approved by all the pathologists contributing to the case. Challenging cases are sent for a specialist pathologist opinion. Most of the laboratories in the UK seek an external specialist opinion if the initial interpretation of the breast lesions is difficult as determined by the reporting pathologists, or if there is lack of agreement on the classification among local pathologists. Medical negligence regarding the management of breast-related cases has been observed to be one major source of medical litigation. [21][22][23] Therefore, pathologists should be encouraged to seek a specialist opinion in difficult cases to avoid such risk.
In this study we reviewed the Nottingham experience of breast pathology referrals to evaluate the pattern of referral and the impacts of the offered opinion. It is important that, in Nottingham, not only are there expert breast pathologists with long-term experience in the field; it is also a training centre with a culture of sharing challenging or borderline cases among members of the breast pathology team and, when deemed appropriate, with other specialist pathologists based on the differential diagnosis of the lesion. Also, external expert opinion is also sought in some cases that are considered inconclusive or exceptionally rare. Although the current study provided data on the distribution of lesion types and reasons of referral, these may be biased by the selective nature of the cases referred to Nottingham for specialist opinion, which is typically based on the referring pathologists' preferences and personal patterns of referral, the perceived type of experience of the Nottingham breast pathologist team in certain lesions and the lack of systematic or mandatory requirement of a referral practice in the UK.
In this retrospective study, the referral materials (glass slides with or without unstained slides or representative tissue blocks for further testing if required) were received via courier in all cases. In all cases, haematoxylin and eosin (H&E)-stained slides representative of the lesion of concern were received with accompanying IHC-stained slides when available. In most cases, one or few representative tumour blocks (or unstained glass slides) were received. As digital pathology (DP) is increasingly used for primary diagnosis in the clinical practice, sharing whole slide images (WSI) can be an alternative and useful tool to hasten turnaround time of expert specialist referrals, as well as facilitate discussions. This study does not address routine diagnostic errors in pathology reporting, as the initial pathologists have already identified the difficulty of classification of the lesion and sought an external specialist opinion. However, when assessing the impact on the management of the patients considering the difference between the proffered initial classification and the final classification of the lesions, minor impacts on patients' management were observed in 17% of the cases, while significant changes in the management as a result of the expert opinion were observed in 18% of the casesrecognising that interpretative errors are not alike, that expert opinion is still personal and different experts in the field can differ in their interpretation of the same lesion. We have observed this phenomenon in the few cases that were shared among a group of breast pathology experts where weighting of the contribution of the various morphological features present can be subjective and can influence their final classification. 24 Our study is in line with a previous report indicating 35% discordance rate. 25 In another study a significant discrepancy, which was defined as a disagreement that affected patient care, was found in 226 cases (11.5%). 26 Some authors 27 have reported major discordance in 8% of cases 10 ; however, some changes in the management were reported in 80% of the cases. In a systematic mandatory review of cases, as part of the patient pathway in a tertiary hospital, change in diagnosis was documented in 17% of cases and change in surgical management in 10%. 9 Although some authors have documented that the most frequently significant change in diagnosis was between DCIS with invasion and DCIS alone, our study indicated that significant change can also be observed in misclassification of rare tumour types in *Fibroepithelial lesions were associated with higher number of cases due to the management differences between malignant and borderline phyllodes tumour and between benign and borderline phyllodes tumour. LCIS, Lobular carcinoma in situ. **Also, the difference in the management between atypical ductal hyperplasia (ADH) and ductal carcinoma in situ (DCIS) despite the overlaps. ***Cases were considered difference classification if the final diagnosis was not included in the referral letter. addition to papillary, spindle cell and vascular lesions. Classification of fibroepithelial lesions appears to comprise a major challenge not only between borderline and malignant phyllodes tumours or between benign and borderline phyllodes tumours, but also between benign phyllodes tumours and cellular fibroadenoma and between hamartoma and phyllodes tumours. Abnormal epithelial proliferation in fibroepithelial lesions is another diagnostic challenge, with overinterpretation of florid epithelial hyperplasia as in-situ carcinomas and missing foci of lobular neoplasia within the lesion. Management of coexisting lesions such as DCIS and florid LCIS within a fibroepithelial lesion was also a reason for referral. This was similar to the overall rate of major changes in diagnosis among all specimen types. From our study and previous experience, common pitfalls that result in difficulty in the accurate classification of breast lesions are summarised in Table 1. Contributory factors to the misclassification of breast lesions include inadequate use or interpretation of IHC markers or difficulty in interpretation of an unusual immunoprofile, such as the following: 1. Using only basal cytokeratins or oestrogen receptor with or without myoepithelial-specific markers to distinguish intraductal epithelial hyperplasia from neoplasia.

Expression of basal cytokeratins in intermediate-
or high-grade DCIS in a pattern mimicking usual type epithelial hyperplasia (morphology is the key in such scenario). 3. Abnormal expression of cytokeratins in melanoma resulting in its misclassification as breast carcinoma. Melanin pigment can also be misinterpreted as haemosiderin granules secondary to previous haemorrhage in tumours. 4. Malignant phyllodes tumour expressing p63 and cytokeratins resulting in misclassification in core biopsy as metaplastic breast carcinoma. 5. Epithelioid myofibroblastoma is ER-positive and has been misclassified as invasive lobular carcinoma in core biopsy, as other clinical and morphological features were not noted and IHC markers such as CD34 and desmin not used. 6. S100 and CD117 are expressed in a proportion of breast cancer and their positivity does not confirm melanoma, secretory carcinoma or adenoid cystic carcinoma, respectively. 7. A proportion of DCIS, particularly those with papillary pattern adjacent to solid or encapsulated papillary carcinoma or recurrent DCIS, may lack peripheral myoepithelial cell markers, and some of these cases were interpreted as invasive carcinoma. In contrast, p63 IHC shows peripheral location of staining in the proliferating malignant cells of adenosquamous carcinoma mimicking native myoepithelial cells at the epithelial stroma interface, and this pattern resulted in difficulty in distinguishing invasive from in-situ tumours.
In occasional cases, the suggested diagnosis was confirmed using extra molecular tests as well as IHC. Examples included the differential diagnosis of spindle cell lesions [mutation analysis of the b-catenin (CTNNB1) gene for fibromatosis, detection of deletion/ monosomy of chromosome 13q and 16q in myofibroblastoma, MYH9::USP6 fusion gene in nodular fasciitis and COL1A1::PDGFB fusion gene in dermatofibrosarcoma protuberans (DFSP)] 28 and tumour subtyping (detection of ETV6::NTRK3 fusion gene in secretory carcinoma, MECT1::MAML2 fusion gene in mucoepidermoid carcinoma and MYB::NFIB fusion gene in adenoid cystic carcinoma, in addition to detection of cmyc gene mutation in radiation-induced angiosarcoma). 29 MDM2 gene amplification was also tested in one case to differentiate liposarcoma from stroma-rich malignant phyllodes tumour with liposarcoma-like stroma, and HRAS Q61 hot-spot mutation was tested in two cases of adenomyoepithelioma. 30 Although the current study indicated that most of the cases either have minor change in the classification, comments on management and differential diagnosis or a higher percentage of misclassification that resulted in significant change in patient management, this study is biased and does not represent routine second opinion in the clinical settings, but is enriched with challenging and rare cases. These results highlight the importance of a specialist opinion of such cases, which is part of our own practice in Nottingham, as a reasonable proportion of cases were sent to external experts for further opinion.
In this study we did not consider change in the histological grade of in-situ or invasive carcinoma, as this is known to be subjective and inter-and intraobserver concordance is not perfect. Similar changes in the classification of invasive tumour type or pattern of DCIS were not considered in the analysis unless this resulted in the change of management, such as low-grade adenosquamous carcinoma, which is an indolent invasive tumour compared to intermediateand high-grade adenosquamous carcinoma, which is an aggressive type of metaplastic carcinoma.
There is also educational value in understanding the expert's approach and how the diagnosis has been reached and differentials excluded, on which morphological features were the emphasis placed, interpretation of immunohistochemistry profiles and comments on further management or prognosis.
Although data on diagnostic discordances stated in this study do not reflect the diagnostic accuracy of the reporting pathologists in routine clinical settings, as the study cohort was based on challenging cases initially identified by the local pathologists as requiring an external expert opinion (unlike systematic review of cases to identify diagnostic errors in breast pathology), this study highlights the importance of specialist opinion as an important and valuable quality method. Therefore, we encourage reporting pathologists to send challenging and rare cases for specialist opinion and national and international professional organisations and authorities to publish guidelines and standard operating procedures to guide and promote such quality assurance practice. Occasional routine cases that are associated with a low level of concordance can also be subjected to external second opinion as part of a quality assurance activity. Previous studies have indicated that the percentage of cases with a major change in pathological opinion is higher when diagnoses are compared between institutions. 10,18 This difference is due in part to differences in thresholds set by groups of pathologists working together regarding classification of breast lesions. Therefore, regional or smaller institutions and general pathologists may have an increased requirement for referral externally to external experts or specialist breast pathologists, particularly in rare and unusual lesions and in those with overlapping features, to ensure best patient care. In the era of precision medicine and patient-tailored therapy, in addition to the increasing use of costly targeted therapy, the cost of external referral can be considered insignificant and such practice should be encouraged. Moreover, delays in the provision of the final opinion of such cases seem insignificant compared with the magnitude of the refined classification on patient care.
In conclusion, this study demonstrates the value of external referral of challenging, rare and difficult-toclassify breast lesions. It also highlights the most common lesions that are likely to be challenging and expert opinion can improve their classification considering their expertise, the number of such cases encountered in their referral practice and the availability of resources for further consultation and additional ancillary testing.
Adedotun Adebayo and Zsolt Hodi reviewed and approved the manuscript.

Data availability statement
The data that support the findings of this study are available from NUH, NHS Trust. Restrictions apply to the availability of these data, which were used under licence. Data are available from the author(s) with the permission of NUH, NHS Trust.