Lymphoma diagnosis at an academic centre: rate of revision and impact on patient care


  • This work was carried out in the Department of Pathology and Microbiology, University of Nebraska Medical Center, Omaha, NE, USA.


Few studies have examined the value of a mandatory second review of outside pathology material for haematological malignancies. Therefore, we compared diagnoses on biopsies referred to an academic medical centre to determine the rate and therapeutic impact of revised diagnoses resulting from a second review. We reviewed 1010 cases referred for lymphoma during 2009–2010. For each case, referral diagnosis and second review diagnosis were compared. Revised diagnoses were grouped into major and minor discrepancies and all major discrepancies were reviewed by a haematologist to determine the effect the diagnostic change would have on therapy. There was no change in diagnosis in 861 (85·2%) cases. In 149 (14·8%) cases, second review resulted in major diagnostic change, of which 131 (12·9%) would have resulted in a therapeutic change. The highest rates of revision were for follicular, high-grade B-cell, and T-cell lymphomas. We found higher rates of major discrepancy in diagnoses from non-academic centres (15·8%) compared to academic centres (8·5%; P = 0·022), and in excisional biopsies (17·9%) compared to smaller biopsies (9·6%; P = 0·0003). Mandatory review of outside pathology material prior to treatment of patients for lymphoma will identify a significant number of misclassified cases with a major change in therapy.

Lymphoma classification has evolved over the last few decades and, before the advent of the 2001 World Health Organization (WHO) classification (Jaffe et al, 2001), multiple different classification systems were used. The most recent WHO classification, published in 2008, stresses the importance of integrating the morphological, immunophenotypic, molecular, cytogenetic and clinical findings in order to accurately diagnose lymphomas (Swerdlow et al, 2008). This approach requires considerable knowledge, experience and skill to interpret and integrate these findings when making a diagnosis of lymphoma. Furthermore, a close working relationship between the pathologist making the diagnosis and the treating haematologist is paramount. In most pathology practices, especially at small non-academic centres, lymphomas are relatively uncommon and most of the pathologists who diagnose these cases have not subspecialized in haematopathology. Additionally, lack of access to the various technologies needed to render a correct diagnosis is a hindrance at such centres (Wilkins, 2011).

A mandatory second review of outside diagnostic pathology material prior to the treatment of patients for a malignancy is a policy at many institutions. Many lymphoma subtypes are treated according to standard protocols, which makes it essential to render a correct diagnosis prior to the initiation of therapy. Several studies have demonstrated the importance of a second review of pathology material (Kronz et al, 1999; Manion et al, 2008; Swapp et al, 2013), but few have addressed its importance in lymphoma pathology (Lester et al, 2003; LaCasce et al, 2008; Proctor et al, 2011; Matasar et al, 2012). Also, all but one of these latter studies (Proctor et al, 2011) were conducted prior to the advent of the 2008 WHO classification (Swerdlow et al, 2008), which introduced a number of new lymphoma entities, and one study (LaCasce et al, 2008) reported on only the most common lymphoma subtypes.

The purpose of this study was to identify the frequency and types of significant discrepancies in lymphoma diagnosis that may affect clinical management. In the study, we reviewed the diagnoses of 1010 cases referred to the Nebraska Lymphoma Study Group (NLSG) in consultation for lymphoma during a 2-year period (2009–2010). We compared the primary (referral) diagnoses to the second review diagnoses to determine the rate and therapeutic impact of revised diagnoses resulting from the second review.


The NLSG was founded in 1982 as a collaborative network of haematologists and pathologists at the University of Nebraska Medical Center (UNMC) and community practices and hospitals throughout Nebraska and the Midlands. UNMC is one of two academic medical centres in Nebraska and has a large referral base for lymphoma patients. In this study, we included 2521 pathology specimens that were sent to the NLSG for expert review in a 2-year period (1 January 2009 to 31 December 2010). We excluded cases in the following diagnostic categories from further analysis: myeloid neoplasms, acute lymphoblastic leukaemia, plasma cell myeloma, and staging bone marrows for non-haematological malignancies. Lymph nodes and extranodal tissues that were reactive or benign were included in the study. After these exclusions, 1830 cases were left. Among these latter cases, 820 (45%) were sent without a primary diagnosis and were excluded from the study. Ultimately, 1010 cases (55%) from 959 patients entered the study. Cases included in the study were then divided into two categories: ‘mandatory reviews’ due to patient referral to the NLSG for clinical management, and ‘outside consultations’, i.e. pathology slides and materials sent for a second opinion. In all cases, we also recorded whether the primary diagnosis was made at an academic centre or a non-academic centre. An academic centre was defined as one affiliated with a medical school. All second review diagnoses were classified according to the 2008 WHO classification (Swerdlow et al, 2008). For each case, the primary referral diagnosis and the second review diagnosis were compared. Discrepant diagnoses were divided into major and minor categories. Major discrepancies were defined as a significant change in the diagnosis with a potential impact on patient care. Minor discrepancies were defined as a change or clarification in the diagnosis without a significant impact on patient care. All major discrepancies were then reviewed by an expert haematologist to determine the potential impact on patient care. Additionally, major discrepancies were divided into five categories: inaccurate classification of non-Hodgkin lymphoma (NHL) or Hodgkin lymphoma (HL); inaccurate grading of follicular lymphoma (FL); benign disorder diagnosed as lymphoma or vice versa; non-haematological malignancy diagnosed as lymphoma or vice versa; and an imprecise or unclear diagnosis which was clarified. Furthermore, we analysed the agreement in diagnosis by the following sample characteristics: type of institution (academic versus non-academic), type of consultation (mandatory review versus outside consultation), biopsy type (excision versus other types, i.e. needle core, punch biopsy, shave biopsy etc.), biopsy site (lymph node, bone marrow, gastrointestinal tract, skin, soft tissue, and all others), and the performance of additional ancillary studies by a UNMC haematopathologist (yes versus no). The study was approved by the Institutional Review Board at the UNMC.

Statistical analysis

Analysis was conducted at the sample level. The Chi-square test was used to compare the proportion with agreement in diagnosis by sample characteristics. P-values <0·05 were considered to be statistically significant. SAS software was used for the statistical analysis (SAS Institute Inc., Cary, NC, USA).


Of the 1010 cases in the study, 683 (67·6%) were mandatory reviews due to patient referral to the NLSG for clinical management and 327 (32·4%) were outside consultations. Furthermore, 142 cases (14·0%) were referred from academic medical centres and 868 cases (86·0%) were from non-academic centres. Upon second review, 838 cases (83·0%) did not undergo a diagnostic change. In 172 cases (17·0%), however, the primary diagnosis was changed or modified. Among these 172 discrepant diagnoses, 149 (14·8%) were considered to be major discrepancies, whereas 23 (2·2%) were classified as minor discrepancies and grouped with the agreement cases, thus bringing the total cases in agreement to 861 (85·2%). Of the 149 cases with major discrepancies, 131 (12·9%) were deemed by the haematologist to result in a significant therapeutic change.

All major diagnostic discrepancies are summarized in Table 1. The largest category of discrepant cases was one in which the diagnosis was revised from one type of lymphoma to another (6·5%). Within this category, the most common change was from one type of B-NHL to another B-NHL (4·3%). A second large category of revised diagnoses (3·0%) was grading discrepancies in FL, with most diagnoses being changed from low-grade (grade 1 or 2) to high-grade (grade 3A or 3B) FL (2·5%). The third category of discrepancy, comprising 2·8% of discordant diagnoses, included benign entities originally diagnosed as lymphoma or vice versa. The most common diagnostic change in this category (1·6%) was from a benign entity to B-NHL. Imprecise or unclear diagnoses (e.g. use of an outdated classification system or the diagnosis of ‘atypical lymphoid infiltrate’) were encountered in 2·1% of discordant cases, whereas in 0·4% of the cases the diagnosis of lymphoma was revised to a non-haematological neoplasm or vice versa.

Table 1. Summary of the major diagnostic discrepancies in lymphoma diagnosis by category.
  N %
  1. NHL, non-Hodgkin lymphoma; B-NHL, B-cell non-Hodgkin lymphoma; T-NHL, T-cell non-Hodgkin lymphoma; CHL, classical Hodgkin lymphoma; HL, Hodgkin lymphoma; NLP-HL, nodular lymphocyte predominant Hodgkin lymphoma; FL1/2; follicular lymphoma grade 1 or 2; FL3, follicular lymphoma grade 3.

Revised diagnosis14914·8
Category 1 – one lymphoma subtype to another 666·5
B-NHL to B-NHL434·3
T-NHL to T-NHL50·5
B-NHL to T-NHL10·1
T-NHL to B-NHL10·1
B-NHL to CHL10·2
B-NHL to composite HL and B-NHL20·1
T-NHL to CHL30·3
CHL to B-NHL30·2
CHL to composite HL AND B-NHL10·1
CHL to T-NHL20·2
NLP-HL to B-NHL20·2
HL to HL
CHL to NLP-HL20·2
Category 2 – FL grading 303·0
FL1/2 to FL3252·5
FL3 to FL1/250·5
Category 3 – benign to lymphoma or vice versa 282·8
Benign to lymphoma
Benign to B-NHL161·6
Benign to T-NHL30·3
Benign to CHL50·5
Lymphoma to benign
B-NHL to benign10·1
T-NHL to benign20·2
NLP-HL to benign10·1
Category 4 – imprecise or unclear diagnosis 212·1
Category 5 – lymphoma to non-haematological neoplasm or vice-versa 40·4
Lymphoma to non-haematological neoplasm10·1
Non-haematological neoplasm to lymphoma30·3

The most common second review diagnoses and the corresponding referral diagnoses are shown in Table 2. The highest discordance rates were seen for FL grade 3 (59·3%), angioimmunoblastic T-cell lymphoma (AITL; 40%) and Burkitt lymphoma (BL; 33·3%). Diagnostic categories with at least 90% concordance included diffuse large B-cell lymphoma (DLBCL), chronic lymphocytic leukaemia/small lymphocytic lymphoma (CLL/SLL), mantle cell lymphoma (MCL), marginal zone lymphoma (MZL), classical HL, peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS), anaplastic large cell lymphoma (ALCL), ALK-positive and -negative types, and benign entities.

Table 2. Most common expert lymphoma diagnoses and the corresponding referral diagnoses.
Referral diagnosisExpert diagnosis
  1. BL, Burkitt lymphoma; DLBCL, diffuse large B-cell lymphoma; FL1/2; follicular lymphoma, grade 1 or 2; FL3, follicular lymphoma, grade 3; FL-NOS, follicular lymphoma, not otherwise specified; CLL/SLL, chronic lymphocytic leukaemia/small lymphocytic lymphoma; MCL, mantle cell lymphoma; MZL, marginal zone lymphoma; HL, Hodgkin lymphoma; NLP-HL, nodular lymphocyte predominant Hodgkin lymphoma; PTCL-AI, peripheral T cell lymphoma, angioimmunoblastic type; PTCL-NOS, peripheral T cell lymphoma, not otherwise specified; ALCL-ALK-, anaplastic large cell lymphoma, ALK negative; ALCL-ALK+, anaplastic large cell lymphoma, ALK positive.

Discordant samples51216354247242011

To address the issue of a potential bias in our cases with regard to the usual epidemiology of lymphoma, we compared the relative frequencies of the different lymphoma subtypes in our study with previously published epidemiological data from North America (Anderson et al, 1998). In our study, there were significantly fewer cases of FL and DLBCL (i.e. common and easily diagnosed in the community) and more cases of SLL and peripheral T-cell lymphoma (i.e. less common and more difficult to diagnose in the community), reflecting the consultation nature of a subset of the cases in our study (data not shown).

There was a significantly higher rate of disagreement in diagnoses from non-academic centres (137/868, 15·8%) compared to academic centres (12/142, 8·5%; P = 0·022). However, referral cases (105/683, 15%) and consultation cases (44/327, 13·5%; P = 0·42) had similar rates of discordance. The biopsy type and site (lymph node versus other) were also examined to determine if these factors affected the concordance rate. Among the 1010 specimens, 625 (61·9%) were excisional biopsies and 385 (38·1%) were other types of specimens (needle core, punch biopsy, shave biopsy etc.). Excisional biopsies had a significantly higher rate of disagreement compared to the other biopsy types (17·9% vs. 9·6%; P = 0·0003). The majority of biopsies were lymph nodes (526/1010, 52·1%), followed by bone marrow (144/1010, 14·3%), soft tissue (86/1010, 8·5%), gastrointestinal tract (64/1010, 6·3%) and skin (59/1010, 5·8%). Biopsy site was not a significant factor affecting the disagreement rate (P = 0·20). In 520 cases (51·5%), additional ancillary studies were performed by UNMC haematopathologists. Cases requiring additional studies had a significantly higher rate of revised diagnosis compared to cases that did not require ancillary studies (20·6% vs. 8·6%; P < 0·0001).


The goal of this study, conducted at an academic medical centre with expertise in the diagnosis and treatment of lymphoid malignancies, was to assess the rate and therapeutic impact of revised diagnoses resulting from expert haematopathology review. The study included 1010 cases referred to the NLSG with a diagnosis of lymphoma or a benign process involving lymphoid tissue. Comparison of the primary referral diagnoses to the second review diagnoses rendered by expert haematopathologists revealed 172 cases (17·0%) in which the referral diagnosis was changed or modified. Among these 172 discrepant diagnoses, 149 (14·8%) were considered major discrepancies and 131 (12·9%) of these were considered to have a significant impact on patient care.

Several studies conducted in North America and Europe have examined the importance of a second review of diagnostic material in haematopathology and are listed in Table 3 (Lester et al, 2003; LaCasce et al, 2008; Proctor et al, 2011; Matasar et al, 2012). The rate of diagnostic discrepancy has ranged from 6% to 27% with two studies reporting discordance rates of about 17% (Lester et al, 2003; Matasar et al, 2012), similar to our rate of 17%. La Casce et al (2008) reviewed NHL cases in the National Comprehensive Cancer Network database and reported a discordance of only 6% between the primary and review diagnoses. However, this study included only the five most common NHL subtypes (FL, DLBCL, MCL, SL, and MZL), which probably explains the much lower discordance rate compared to the other published studies. A recent study from the Mayo Clinic (Swapp et al, 2013) reviewed 71 811 surgical pathology cases received in consultation over a 6-year period, and lymph nodes were the second most common site of disagreement after gastrointestinal specimens. Of the 457 surgical pathology cases with major disagreement in that study, 16% were lymph node biopsies.

Table 3. Comparison of previously-published studies to the current study.
AuthorOverall discrepancy rate (%)Most common discrepant diagnoses (%)
  1. NA, not available; FL3, follicular lymphoma, grade 3; SLL, small lymphocytic lymphoma; MCL, mantle cell lymphoma; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; LPL, lymphoplasmacytic lymphoma; FL1/2, follicular lymphoma, grade 1 or 2; BL/Burkitt-like, Burkitt lymphoma or Burkitt-like lymphoma; TCL, T-cell lymphoma; and PTCL-AI, peripheral T-cell lymphoma, angioimmunoblastic type.

Lester et al (2003)125/745 (16·8)NA
LaCasce et al (2008)43/731 (5·8) FL3 (12·5), SLL (10·3), MCL (7·3)
Proctor et al (2011)513/1873 (27·3)LPL (34·1), MCL (34·1), DLBCL (25·8), FL (23·4), SLL (20)
Matasar et al (2012)123/719 (17·1BL/Burkitt-like (62·5), FL1/2 (41·9), TCL (26)
Current study149/1010 (14·8)FL3 (59·3), PTCL-AI (40·0), BL (33·3), FL1/2 (14·5)

An interesting finding in our study was the significant difference in discrepancy rates between excisional biopsies and other biopsy types (17·9% vs. 9·6%; P = 0·0003). In contrast, LaCasce et al (2008) and Matasar et al (2012) found that the type of biopsy specimen was not predictive of a diagnostic change. Excisional biopsy specimens are larger, providing additional diagnostic information and material for immunophenotyping, molecular analysis, and other diagnostic studies. For example, an excisional biopsy may reveal two different tumour grades or different diagnoses (e.g. composite lymphoma), making the diagnosis more complex and therefore more challenging. Other biopsy types are smaller, providing less tissue for comprehensive evaluation, and more often resulting in a vague or inaccurate diagnosis. This finding emphasizes the importance of an excisional biopsy, rather than smaller biopsies, such as a needle biopsy, for the initial diagnosis of lymphoma.

Similar to our study, some reports also evaluated the impact of revised diagnoses on patient management. Lester et al (2003) reported a change in diagnosis in 17% of lymphoma cases and, upon chart review, found a change in management in 8% of the cases. Proctor et al (2011) reviewed almost 1900 cases over a 5-year period (2003–2008) and found an overall discordance rate of 27·3%. Interestingly, the discordance rate in that study decreased significantly over the years from 32% discordance in 2003 to only 13–15% discordance after 2006. Among the discordant cases, 11% would have resulted in a significant change in clinical management and 39% would have had a minor impact on patient care. Furthermore, in the other 50% of discordant cases, the primary diagnosis often provided insufficient or outdated information and, without a second review, would have resulted in either delayed or inappropriate therapy (Proctor et al, 2011).

The majority of discrepant cases in our study were revisions from one type of lymphoma to another type, with the most common change being from one type of B-NHL to another. The second most common diagnostic problem was FL grading, followed by benign entities diagnosed as lymphoma or vice versa. As for the specific diagnoses, the highest discordance rate was seen for FL grade 3 (59·3%), followed by AITL (40%) and BL (33·3%). Discrepancy rates by diagnosis have varied in the published studies with some common diagnostic problems among the different studies (Table 3). Proctor et al (2011) reported a wide range of discordance depending on the specific diagnosis, from 3·6% for plasma cell neoplasms to 34·1% for lymphoplasmacytic lymphoma and MCL. They also found high discrepancy rates for two of the most common lymphomas, FL and DLBCL. Similar to our study, discordance rates were also high for BL, but the number of these cases was small (Proctor et al, 2011). LaCasce et al (2008) reported their highest discrepancy rates for FL grade 3, followed by SLL and MCL. The most common diagnostic changes reported in the study by Matasar et al (2012) included revisions from benign entities to lymphoma, nondiagnostic/ambiguous to diagnostic/definitive diagnoses, highly-aggressive B-cell neoplasm to aggressive B-cell neoplasm, and indolent B-cell neoplasm to aggressive B-cell neoplasm. Among cases with a primary diagnosis of indolent B-NHL revised to aggressive B-NHL, the most frequent revision was from FL grade 1 or 2 to FL grade 3, similar to our study. Even though primary diagnoses of BL and Burkitt-like lymphoma were uncommon, five of eight cases (62·5%) were reclassified, mostly into DLBCL (Matasar et al, 2012). Furthermore, in the study by Matasar et al (2012), submitted diagnoses of T-cell lymphoma (including cutaneous cases) underwent revision in 26% of the cases.

Grading of FL is a well-recognized problem in haematopathology, as demonstrated by our study and several others (Martinez et al, 2007; LaCasce et al, 2008; Matasar et al, 2012). The WHO classification (Swerdlow et al, 2008) recommends a three-grade system based on the average number of centroblasts counted in 10 neoplastic follicles at high (40×) magnification. However, this method is subjective and suffers from a lack of reproducibility among pathologists (Martinez et al, 2007). Gene expression profiling is reported to be accurate in distinguishing low-grade from high-grade FL (Glas et al, 2005), and Dave et al (2004) identified a molecular signature based on the microenvironment that is predictive of survival in patients with FL. Flow cytometry and immunohistochemistry have also been employed in grading of FL, but with little or no success (Martinez et al, 2007). The proliferation index, as assessed by the Ki67 immunohistochemical staining, usually correlates with FL grade, i.e. the proliferation index is higher in FL grade 3 compared to FL grades 1 and 2 (Wang et al, 2005). Ki67 staining can, therefore, be a useful aid in FL grading. However, Wang et al (2005) found that up to 18% of FL classified as grade 1 or 2, based on histological assessment, have a high proliferation index when stained for Ki67. The clinical value of FL grading has also been a controversial topic for many years. While most studies agree that FL grades 1 and 2 are indolent and incurable with conventional therapy, the management of patients with FL grade 3A varies from watch and wait to aggressive treatment. Ganti et al (2006) studied the outcomes of 421 patients with FL treated with anthracycline-based chemotherapy and found FL grade to be a useful predictor of overall survival (OS). Hans et al (2003) examined the different subtypes of FL grade 3 in 190 patients treated with anthracycline-based chemotherapy. Cases of FL grade 3A, FL grade 3B, and follicular large cleaved cell lymphoma did not differ significantly in clinical characteristics, OS, or event-free survival (EFS). Only cases with a predominant diffuse component had a significantly worse OS and EFS (Hans et al, 2003). Grade 3 FL is usually treated like DLBCL in most centres (Wahlin et al, 2012) and, therefore, it is important to accurately grade FL. For lack of a better and more readily-available tool, counting centroblasts in haematoxylin and eosin-stained slides remains the gold standard for FL grading (Swerdlow et al, 2008). However, new studies are needed to find a better and more reproducible method of grading.

In this study, we found a significantly higher rate of disagreement in diagnoses from non-academic centres compared to academic centres (15·8% vs. 8·5%; P = 0·022). In one study, 121 (91·7%) of the 132 surgical pathology cases with major diagnostic disagreement were from non-academic centres (Manion et al, 2008). The majority of bone marrow cases (91%) revised by Naqvi et al (2011) were also from non-academic centres. In haematopathology, the differences between academic and non-academic centres are most probably due to the lack of haematopathology training and experience of the pathologists who diagnose these cases at non-academic centres, as well as a lack of access to the ancillary technologies needed to accurately diagnose haematological malignancies. This is supported by our finding that cases requiring additional diagnostic studies had a significantly higher rate of revised diagnosis compared to those that did not require such studies (20·6% vs. 8·6%; P < 0·0001). Furthermore, in studies published by groups that provide centralized expert pathology review, such as the All Wales Lymphoma Panel (Lester et al, 2003) and the North Central London Lymphoma network (Proctor et al, 2011), the discordance rates of 16·8% and 27·3% respectively, are similar to or even higher than in our study.

The National Institute for Health and Care Excellence (NICE) guidelines (2003), used in the United Kingdom and other countries, recommend that haematopathology services should be organized at a network level and that all tissue samples from patients with a possible or definite diagnosis of a haematological malignancy should be assessed by an expert haematopathologist. This is essentially the practice of the NLSG. The most widely used guidelines for the diagnosis and treatment of haematological malignancies in North America are those of the National Comprehensive Cancer Network (NCCN, 2013). In our opinion, the NCCN should adopt standard-of-care recommendations similar to the NICE guidelines, encouraging clinicians to request expert haematopathology review of all new malignant diagnoses. In the United States, haematopathologists are trained in a 1- or 2-year focused haematopathology fellowship, which is pursued after a 4-year pathology residency. During fellowship training, the wide spectrum of haematopathology practice is covered, including lymph node, bone marrow and peripheral blood pathology, coagulation, as well as diagnostic techniques (i.e. flow cytometry, cytogenetics, molecular pathology, investigation of haemoglobin disorders). After the fellowship, graduates take a Hematopathology Board Examination administered by the American Board of Pathology. Similar training programmes are recommended for countries lacking this type of expertise. Our findings emphasize the need for such specialized training and experience, as well as sophisticated technology, when dealing with lymphoma pathology.

Lymphoma pathology is a challenging field that requires considerable knowledge, experience, skill, and technology to make the correct diagnosis. We conclude that mandatory review of outside diagnostic pathology material by expert haematopathologists prior to the treatment of lymphoma patients will identify a significant number of misclassified cases and result in major changes in patient management. The cost of mandatory pathology review is small when compared to the cost of treating patients ineffectively or using more expensive drugs or aggressive therapies than needed. Future studies should address the cost-effectiveness of mandatory pathology review.


J.M.B and A.M.P. collected the data and wrote the manuscript; L.M.S. performed statistical analysis; J.A.L., K.K., M.B., P.A., K.F., T.C.G., and W.C.C. collected the data; J.M.V. and J.O.A. collected and interpreted the data; and D.D.W. designed the study, interpreted the data and wrote the manuscript. All authors read and approved the final version of the manuscript.