Optimal Minimal Panels of Immunohistochemistry for Diagnosis of B-Cell Lymphoma for Application in Countries with Limited Resources and for Triaging Cases Before

AJCP /ORIGINAL ARTICLE

Optimal Minimal Panels of Immunohistochemistry for Diagnosis of B-Cell Lymphoma for Application in Countries With Limited Resources and for Triaging Cases Before

Referral to Specialist Centers Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019

Maria Giulia Disanto, MD,1 Maria Raffaella Ambrosio, MD, PhD,2 Bruno Jim Rocca, MD, PhD,2 Hazem A. H. Ibrahim, FRCPath, PhD,1,3 Lorenzo Leoncini, MD, PhD,2 and Kikkeri N. Naresh, MD, FRCPath1

From the 1Department of Histopathology, Imperial College Healthcare NHS Trust & Imperial College, London, United Kingdom; 2Department of Medical Biotechnologies, Section of Pathology, University of Siena, Siena, Italy; and 3Department of Histopathology, Faculty of Medicine, Mansoura University, Mansoura, Egypt.

Key Words: Lymphoma; B-cell lymphoma; Immunohistochemistry; Diagnosis; Classification; Developing countries

Am J Clin Pathol May 2016;145:687-695

DOI: 10.1093/AJCP/AQW060

ABSTRACT Lymphomas are a collection of different malignancies “arising” from lymphoid cells. They include about 49 entities, Objectives: Establish and validate optimal minimal and over 19 provisional entities and subsets.1 About 85% of immunohistochemistry panels for usage in a staged lymphomas are of B-cell origin. Precision in lymphoma diag- algorithmic manner for precise diagnosis of B-cell nosis requires expertise and infrastructure. The entities are lymphomas in countries with limited resources. Suggest defined based on morphology, immunohistochemistry (on short panels of immunostains to be used in referring units some occasions in situ hybridization), cytogenetics/fluores- that refer suspected lymphomas to specialist diagnostic cent in situ hybridization (FISH), molecular genetics and clin- centers in resourceful countries. ical information. Thorough knowledge and experience in Methods: Significant proportion of six B-cell lymphomas morphology of lymphoid lesions and a good understanding of has characteristic morphology requiring a short panel of the current World Health Organization (WHO) classification 1 confirmatory immunostains. The rest would go through five of lymphoid neoplasms are essential. Accurate diagnosis different algorithms. also requires application of a wide panel of immunohistochemical stains, high-quality laboratory infrastructure and Results: 812 cases in which a B-cell lymphoma or an HIV- personnel, and resources. In addition, about 20% of cases associated lymphoma was suspected on morphological need cytogenetic and molecular investigations. grounds were evaluated. This led to arriving at a specific Though the incidence of lymphomas varies widely across diagnosis of 799 B-cell lymphomas. A correct diagnosis was the globe, and though the incidence is higher in countries achievable in 69% cases with the application of three to five with better resources, lymphomas are frequent in countries antibodies; others required additional work-up. with limited resources.2 As many countries with limited re- Conclusions: The panels/algorithms assist pathologists in sources have high population density, and as the numbers of practicing lymphoma diagnostics in countries with limited hematopathologists are unfortunately low, the absolute num- resources and in making lymphoma referrals to specialist bers of lymphomas seen by individual pathologists are con- centers. siderably high. Practicing lymphoma pathology in developing countries is challenging, and two of the authors (K.N.N. and L.L.) have been amply exposed to these challenges.3-7 In many countries within Europe, lymphoma diagnosis is being centralized to specialist-integrated hematologic

© American Society for Clinical Pathology, 2016. All rights reserved. Am J Clin Pathol 2016;145:687-695 687 For permissions, please e-mail: [email protected] DOI: 10.1093/ajcp/aqw060 687 Disanto et al /MINIMAL IMMUNOPANELS FOR B-CELL LYMPHOMAS malignancy diagnostic services (SIHMDS). Within many Extensive immunohistochemical work-up had been SIHMDS networks, pathologists at contributing/referring undertaken on these cases. Where required, in-situ hybrid- units would like to maintain their expertise. Pathologists ization for Epstein-Barr virus-encoded RNA (EBER) and from the referring units would prefer to perform some light chains, FISH analysis for MYC, BCL2, BCL6, and IG immunostains, attempt to arrive at a diagnosis, and then for- genes, and antigen receptor gene rearrangement studies had ward the sample to the SIHMDS for final integrated diagno- been undertaken. Initially, reports were evaluated, and each sis.8 Although this helps in maintaining a certain level of of the reports had concise, but meticulous morphological expertise at referring units, it also results in duplication of description and clear documentation of all the additional Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 work-up and a strain on resources. work-up undertaken. Though lymphoma diagnostics have evolved in re- The diagnostic process was employed by two phases, sourceful countries of Europe and North America, some of phase 1 and phase 2. Based on the morphological descrip- the investigators in these countries (such as two of the au- tion (in cases other than those where HIV status was thors of this manuscript, L.L. and K.N.N.) have a keen inter- known), an attempt was made to slot cases into categories est in improving lymphoma diagnostics in developing under phase 1. Cases that could not be placed into any of the countries and in fostering translational research in lymph- categories under phase 1 were then assigned to diagnostic omas prevalent in developing countries.6 The primary aim algorithms based on the morphological description. of the study was to establish minimal panels of immunohis- Biopsies from HIV-positive patients were placed under tochemistry that can be used in a staged algorithmic manner Algorithm 5. to arrive at a precise diagnosis in most cases of suspected B- cell lymphomas in countries with limited resources. Through this work, we also wanted to suggest short panels Phase 1 Diagnosis of immunostains that can be used in referring units in the We identified six types of the B-cell lymphomas that context of SIHMDS networks (or similar centralized ser- had very characteristic morphology and thus required a short vices for lymphoma diagnostics) before the referral is made. panel of immunostains to confirm the diagnosis Table 1 . The limited panel algorithms are not intended for use in These included the following: SIHMDS, where having short turnaround times (TATs), 1. Suspected chronic lymphocytic leukemia/small achieving highest accuracy and diagnostic precision on a lymphocytic lymphoma (CLL/SLL): characteristic cytolo- large volume of cases, training of personnel within the cen- gical features and presence of proliferation centres. Confirm ters and those who refer cases to the centers, and documen- with CD20 (positive), CD5 (positive), CD23 (positive), tation clinically relevant biomarkers and translational CD10 (negative), and cyclin D1 (negative). If CD23 is nega- research, are of paramount importance. tive, evaluate for SOX11 or proceed to Algorithm 2 in phase Overall, the approach we suggest could help in cutting 2. Please note that if flow cytometry/immunophenotyping costs and improving quality in both resource-poor and re- has documented the characteristic immunophenotype of sourceful countries. CLL, lymph node biopsy and immunohistochemistry would not be essential. Suggested algorithms (both on phase 1 and 2 for CLL) are primarily for those cases where peripheral blood/bone marrow immunophenotyping has either not Materials and Methods been performed or has not been informative. The study was initially performed on 296 cases in 2. Suspected follicular lymphoma, grades 1 to 3a (FL1- which a B-cell lymphoma or an HIV-associated lymphoma 3a): a prominent follicular pattern and with presence of cen- was suspected on initial morphological evaluation during trocytes accompanied by variable numbers of centroblasts. the period from January 2012 to June 2012 at the Confirm with CD20 (positive), CD10 (positive), BCL2 Department of Histopathology of Hammersmith Hospital, (positive), and cyclin D1 (negative). If features are not diag- Imperial College Healthcare NHS Trust, London, UK. nostic, the case would need further evaluation by Algorithm Following this, the study was extended to and validated 1 in phase 2. with an additional 516 cases of suspected B-cell lymphomas 3. Suspected follicular lymphoma, grade 3b (FL3b): a seen during the period between January 2005 and June 2012 prominent follicular pattern and with infiltrate entirely com- at the Department of Medical Biotechnologies of the posed of centroblasts (absence of centrocytes). Confirm University of Siena, Italy. All 812 cases had been fixed in with CD20 (positive), BCL6 (positive), BCL2 (positive), formalin and embedded in paraffin. Cases where the initial IRF4/MUM1 (positive) and cyclin D1 (negative). In follicu- morphological description suggested Hodgkin lymphoma or lar lymphoma with a follicular pattern (irrespective of a T-cell lymphoma were excluded. grade), CD21 immunostain is recommended only when the

688 Am J Clin Pathol 2016;145:687-695 © American Society for Clinical Pathology 688 DOI: 10.1093/ajcp/aqw060 AJCP /ORIGINAL ARTICLE

Table 1 not conform to the characteristic patterns described in phase Immunohistochemistry Panels in Phase 1 1, though morphological features led to a short panel of Histology Panel immunostains; and morphology and clinical information Suspected CLL/SLL CD20, CD5, CD23, that dictated they should go through different algorithms or CD10, and cyclin D1 decision trees. Suspected FL1-3a CD20, CD21, CD10, BCL2, and cyclin D1 Suspected FL3b CD20, CD21, BCL6, Algorithm 1 Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 BCL2, MUM1, and Nodular pattern—cases with easily identifiable nodular cyclin D1 pattern. The nodular pattern includes various follicular pat- Suspected MCL CD20, CD5, and cyclin terns and other “nodular” patterns mimicking a follicular D1 Suspected DLBCL CD20, CD5, CD21, pattern. Infiltrate could be monomorphic or polymorphic and Ki67 and cell type could be variable. The algorithm addresses the Suspected BL CD20, CD10, BCL2, segregation of follicular lymphoma (FL), MCL, and mar- TdT, and cyclin D1 ginal zone lymphoma (MZL). It also identifies rare CLL/ BL, Burkitt lymphoma; CLL/SLL, chronic lymphocytic leukemia/small lymphocytic lymphoma; DLBCL, diffuse large B-cell lymphoma; FL1-3a, follicular lymphoma SLLs with prominent proliferation centers mimicking a grades 1-3a; FL3b, follicular lymphoma grade 3b; MCL, mantle cell lymphoma; nodular pattern Figure 1 . TdT, terminal deoxynucleotidyl transferase. Algorithm 2 nodular pattern is subtle or is only partially identified. In the Diffuse/interfollicular pattern with small to medium- latter case, quantifying diffuse and follicular components sized cells. CD21 immunostain is used to exclude presence would be aided by CD21 stain. If CD21 demonstrates ab- of nodular/follicular pattern. Cases with focal follicular pat- sence of follicular dendritic cell meshworks, such areas tern would be assessed by Algorithm 1. The number of large need to be evaluated for (diffuse large B-cell lymphoma) nucleolated cells should be <20% of the overall population DLBCL. of cells or <15 per high-power field. If the proportion of 4. Suspected mantle cell lymphoma (MCL): character- large cells is bigger, the case should be assessed for istic mantle zone pattern and monomorphic centrocyte cy- DLBCL. If the proportion of large cells is bigger, and CD21 tology. Confirm with CD20 (positive), CD5 (positive), and identifies a nodular pattern, the case should be assessed both cyclin D1 (positive). If cyclin D1 is negative, evaluate for for DLBCL and by Algorithm 1. Algorithm 2 addresses the SOX11 or proceed to Algorithm 2 in phase 2. segregation of CLL/SLL, MCL, MZL, diffuse FL, and lym- 5. Suspected DLBCL: presence of sheets of large cells phoplasmacytic lymphoma (LPL) Figure 2 . with morphology of centroblasts or immunoblasts. Confirm with CD20 (positive), CD5 (negative), CD21 (absence of Algorithm 3 follicular pattern) and Ki67 (high proliferation rate, >50%). Diffuse blastic infiltrates of medium-sized cells. CD5þ cases would need further evaluation for cyclin D1 Diffuse sheets of monomorphic medium-sized cells with expression. If CD21 highlights a follicular pattern, follicular high mitotic rate and/or fine chromatin. Algorithm 3 areas would need evaluation by Algorithm 1 in phase 2. addresses the segregation of B-lymphoblastic lymphoma, 6. Suspected Burkitt lymphoma (BL): diffuse sheets of blastoid MCL, and BL Figure 3 . It also allows exclusion of monomorphic medium-sized cohesive cells with multiple T-lymphoblastic lymphomas and other mimics of lympho- paracentrally located nucleoli (score 3 morphology accord- blastic lymphoma. ing to the algorithm proposed by Naresh et al7), starry sky pattern, and high mitotic and apoptotic rate. Confirm with Algorithm 4 CD20 (positive), CD10 (positive), BCL2 (negative), termi- Cases with features overlapping between BL and nal deoxynucleotidyl transferase (negative), and cyclin D1 DLBCL. This is addressed using the previously published 7 (negative). If morphology or immunophenotype was found BL diagnostic algorithm Figure 4 . unusual for BL, the case would proceed to Algorithm 4 in phase 2. Algorithm 5 Biopsies from HIV-positive patients were directly evaluated by Algorithm 5. Algorithm 5 addresses the segrega- Phase 2 Diagnosis tion of BL, DLBCL, BL/DLBCL, plasmablastic lymphoma, These included B-cell lymphomas that did not have primary effusion lymphoma and human herpesvirus 8-asso- characteristic morphological features to be slotted into one ciated large cell lymphoma in multicentric Castleman dis- of the phase 1 categories; immunochemistry results that did ease Figure 5 .

CD20, CD5, cyclin D1

CD20+, CD5+/–, CD20+, CD5+, cyclin D1– CD20+, CD5–, cyclin D1– cyclin D1+ CD38+ CD38– CD10

CD23–: MCL SOX11 CD44 +– Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 +– +

CD23+/–: BCL2 CD20+, CD5–, ? CD5+ MZL; + – cyclin D1–, CD10– CD23+: ? CLL, CD23+: ? CLL ? CD5+ FL

FL Light chains BCL6 restricted + –

BCL2+, CD38 CD10–: FL + –

BCL2+, CD44 CD10/BCL6–: FL +

? MZL

Figure 1 Diagnostic algorithm for cases with a nodular pattern. Nodular pattern: follicular lymphoma (FL) vs mantle cell lymphoma (MCL) vs marginal zone lymphoma (MZL). If nodularity is subtle or seen only partially, the pattern must be confirmed/quan- tified by CD21 and/or CD23. Diffuse areas to be assessed for percentage of large cells and large cells per high-power field. If MZL with follicular colonization is suspected, perform: light chains, CD43, IgM, and IgD. Questions marks indicate the possibil- ity of diagnosing cases of lymphoma with atypical immunophenotype. CLL, chronic lymphocytic leukemia.

CD20, CD5, cyclin D1

CD20+, CD5+/–, CD20+, CD5+, cyclin D1– CD20+, CD5–, cyclin D1– cyclin D1+

CD23 MCL +– CD20+, CD5–, CD10+/–, BCL6+ cyclin D1–, CD10–, BCL6– CLL SOX11 Diffuse FL + – CD44 + ? CLL (CD23–) ? CD5+ LPD ? MZL/LPL

Figure 2 Diagnostic algorithm for cases with a diffuse an interfollicular pattern, and where the infiltrate is composed of small to medium-sized cells. Diffuse/interfollicular pattern with small to medium-sized cells: chronic lymphocytic leukemia (CLL) vs mantle cell lymphoma (MCL) vs mantle zone lymphoma/lymphoplasmacytic lymphoma. Diffuse pattern must be confirmed by absence of CD21/CD23þ follicular dendritic cell meshworks (if a nodular pattern is recognized, assess as for nodular pattern. Large cells should be <20% or <15/high-power field. CD20–, CD5þ , and cyclin D1– cases: evaluation for T-cell lymphomas. In CD10þ and BCL6– cases, perform terminal deoxynucleotidyl transferase.

All results available on the cases were entered into a panel (phase 1) and the applicability of different diagnostic database. Each case was then systematically analyzed for algorithms (phase 2). The minimal number of antibodies/ the adequacy of the confirmatory immunohistochemistry tests that was essential to arrive at a specific diagnosis was

690 Am J Clin Pathol 2016;145:687-695 © American Society for Clinical Pathology 690 DOI: 10.1093/ajcp/aqw060 AJCP /ORIGINAL ARTICLE

CD20, CD3, TdT, cyclin D1

CD20+, CD3–, CD20+, CD3–, CD20+, CD3–, CD20–, CD3–, CD20–, CD3+, CD20–, CD3–, TdT+, cyclin D1– TdT–, cyclin D1+ TdT–, cyclin D1– TdT+, cyclin D1– TdT+, cyclin D1– TdT–, cyclin D1–

Ki-67 CD10+, PAX5+ Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019

BLL Blastoid MCL See Figure 4 BLL TLL Exclude myeloid sarcoma and other non- hematologic malignancies

Figure 3 Diagnostic algorithm for cases with a diffuse infiltrate of blastic cells. Diffuse blastic infiltrates of medium-sized cells: B-lymphoblastic lymphoma (BLL) vs T-lymphoblastic lymphoma (TLL) vs blastoid mantle cell lymphoma (MCL) vs Burkitt lymphoma. TdT, terminal deoxynucleotidyl transferase.

Phase 1 Phase 2 Phase 3

Morphology (0-3) Morphology (0-3) Morphology (0-3) BCL2 (–, 2; weak, 1) BCL2 (–, 2; weak, 1) BCL2 (–, 2; weak, 1) CD10 (+, 1) CD10 (+, 1) CD10 (+, 1) Ki-67 (>95%, 2; 90%-95%, 1) Ki-67 (>95%, 2; 90%-95%, 1) CD38 (+, 1) CD38 (+, 1) CD44 (–, 1) CD44 (–, 1) FISH (MYC-lg+ and BCL2/BCL6–, 2)

Cumulative score: Cumulative score: Cumulative score: 5-6, BL; 3-4, BL not excluded; ≥8, BL; 6-7, BL not excluded; ≥8, BL; 6-7, BL not excludeda <3, not BL <6, not BL

Figure 4 Diagnostic algorithm for cases where morphological features overlap between Burkitt lymphoma (BL) and diffuse large B-cell lymphoma. CD20þ and terminal deoxynucleotidyl transferase–negative tumor with a diffuse infiltrate of medium to large lymphoid cells. Benefit from karyotype, comparative genomic hybridization, gene expression, and assessment of the impact of each of the parameters. FISH, fluorescence in situ hybridization.

CD20, HHV8, LANA1, EBER

CD20+, HHV8–, EBER– CD20+, HHV8–, EBER+ CD20–, HHV8–, EBER+ CD20–, HHV8+, EBER+ CD20–, HHV8+, EBER–

CD138, PAX5, light chains Light chains Light chains

• PBL (CD138+, PAX5–, • PEL (light chain • HHV8-associated DLBCL DLBCL light chain restricted negative) LBCL in MCD (λ • If BL or BL/DLBCL, • Lymphocyte-depleted light chain restricted) see Figure 4 • If BL of BL/DLBCL, classical Hodgkin see Figure 4 lymphoma (PAX5+, • Exclude lymphocyte- light chain negative; depleted classical perform CD15 and Hodgkin lymphoma CD30

Figure 5 Diagnostic algorithm for lymphoma cases in the HIV setting. Aggressive B-cell lymphoma in HIV setting: Burkitt lymphoma (BL), BL/diffuse large B-cell lymphoma (DLBCL), plasmablastic lymphoma (PBL), primary effusion lymphoma (PEL), and human herpesvirus 8–associated large B-cell lymphoma (LBCL) in multicentric Castleman disease (MCD). EBER, Epstein- Barr virus–encoded small RNA.

© American Society for Clinical Pathology Am J Clin Pathol 2016;145:687-695 691 691 DOI: 10.1093/ajcp/aqw060 Disanto et al /MINIMAL IMMUNOPANELS FOR B-CELL LYMPHOMAS documented in each case. Where required, original slides achieved without thorough knowledge and experience in were retrieved and revisited to record any missing morphology of lymphoid lesions, and without a good under- information. standing of the current WHO classification of lymphoid neoplasms. Without a good knowledge base and expertise, the choice of initial panels and the subsequent decisions made via the algorithms can be compromised. While using limited Results panels, there would be little levy for suboptimal technical abilities in tissue fixation, processing or immunohistochemis- Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 Among the 812 cases, 446 (55%) were evaluated ini- try. Hence, these limited panels are likely to be useful only in tially by phase 1; 366 cases could not be placed under any the setting of a large-volume lymphoma practice that would of the categories of phase 1 and went through phase 2 ensure good technical quality and contribute to substantial algorithms. expertise among the reporting hematopathologists. The qual- Phase 1: Overall, in 89% of cases (398/446 cases), cor- ity of H&E-stained slides would also be crucial. rect diagnosis was achievable based on the suggested pan- Using an algorithmic approach poses certain issues: els. This included 83% of 241 suspected DLBCLs, 95% of129 suspected cases of FL1-3a, 100% of 56 suspected 1. It would be difficult to use this approach in cases requir- CLL and others as described. Thirty-two DLBCLs ex- ing a stringent TAT. The very nature of dichotomized pressed CD5, which necessitated cyclin D1 immunohisto- decision making before progressing to the next panel ex- chemistry. Five cases of FL1-3a were CD10 negative, and tends the TAT. three cases of FL1-3a were BCL2 negative. These cases 2. By using short panels, one loses the safety net of using also required further immunostains to arrive at the correct multiple overlapping antibodies to arrive at similar con- diagnosis. The initially suspected diagnosis on morphology clusions. For example, if one used CD20 in isolation in- was incorrect in nine (4%) of 446 cases, and these included stead of CD20, CD79a and PAX5, it is possible that a seven T-cell lymphomas and two blastoid MCLs Table 2 . B-cell lymphoma with weak/absent CD20 expression Phase 2: Overall, of 366 cases analyzed in phase 2, might be missed (or sample from a patient on rituximab 41% required five or fewer, and 93% required eight or fewer therapy, when information is not provided). antibodies to arrive at a specific diagnosis. Among the 159 3. It hampers continued growth of hematopathologists. cases with a nodular pattern, 38% requiredfive or fewer, and Due to these issues, the proposed algorithmic approach 92% required eight or fewer antibodies to arrive at a specific needs to be introduced with caution, and only where neces- diagnosis. Among the 120 cases with a diffuse interfollicu- sary, ie dictated by extraordinary economic constraints. lar pattern, 52% required five or fewer, and 100% required The rationale for incorporating specific antibodies eight or fewer antibodies to arrive at a specific diagnosis. within the algorithm in specific situations is summarized as Among the 14 HIVþ lymphomas, 57% required five or few- follows: er, and 93% required eight or fewer antibodies/tests to arrive 1. In our studies on material from developing countries, at a specific diagnosis Table 3 . Proportion of each of the we have found preservation of nuclear antigens such as different lymphomas among cases with different morpho- BCL6 to be suboptimal. Hence, CD10 has been pre- logic patterns is provided in Table 4 . ferred over BCL6 in most phase 1 algorithms (with ex- Overall, 799 B-cell lymphomas were evaluated. ception of FL3b, where CD10 is frequently not DLBCL and FL1-3a were the most frequent lymphomas expressed). Furthermore, BCL6 is expressed on T-cells Table 5 . A correct diagnosis was achievable in 69% cases including T-cells within follicles. with the application of three to five antibodies. 2. CD20 is required to establish the B-cell nature of the Furthermore, a correct diagnosis was achievable in 90% disease and also to provide confidence to the clinicians cases and 96% cases using seven antibodies and eight anti- to use rituximab if deemed necessary. bodies, respectively Table 6 . 3. BCL2 would be essential in FLs to exclude hyperplastic follicles mimicking FL. 4. CD21 has been used for identification of follicular architecture and for quantification of follicular and dif- Discussion fuse components. Through this study, we document that in the hands of 5. With respect to CLL, appreciation of proliferation well trained hematopathologists, dependable diagnosis of B- centers depends on the quality of H&E sections. cell lymphomas is achievable, even under considerable eco- Prominence of proliferation centers is variable between nomic constraints. It should be stressed that this cannot be cases. Even in cases with identifiable proliferation

692 Am J Clin Pathol 2016;145:687-695 © American Society for Clinical Pathology 692 DOI: 10.1093/ajcp/aqw060 AJCP /ORIGINAL ARTICLE

Table 2 Diagnostic Ability at Phase 1

% (No. of) Cases % (No. of) Cases Where Correct % (No. of) Cases That Requiring Additional Diagnosis Was Not Suspected Category on Reached Correct Work-up to Reach Suspected on Morphology (No.) Diagnosis Correct Diagnosis Morphology CLL (56) 100 (56) 0 0 Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 FL1-3a (129) 95 (122) 5 (7) 0 5 were CD10 negative 3 were BCL2 negative FL3b (4) 100% (4) 0 0 MCL (9) 100% (9) 0 0 DLBCL (241) 83% (200) 13% (32) 4% (9) due to CD5 7 were T cell expression lymphomas 2 were blastoid MCL BL (7) 100% (7) 0 0 Total (446) 89% (398) 9% (39) 2% (9)

BL, Burkitt lymphoma; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; FL1-3a, follicular lymphoma grades 1-3a; FL3b, follicular lymphoma grade 3b; MCL, mantle cell lymphoma.

Table 3 Minimum Number of Antibodies That Would Have Been Required to Arrive at the Correct Diagnosis in Phase 2

Pattern/Algorithm (No.) 3 Antibodies 4–5 Antibodies 6–8 Antibodies 9–10 Antibodies/Testsa >10 Antibodies/Testsa Nodular (159) 8% (13) 30% (47) 54% (86) 4% (6) 4% (7) Diffuse/interfollicular 12% (14) 40% (48) 48% (58) — — pattern (120) Diffuse blastic 77% (20) 23% (6) pattern (26) BL/DLBCL algorithm (47) 72% (34) 19% (9) 9% (4) HIV lymphomas (14) 7% (1) 50% (7) 36% (5) — 7% (1) Total (366) 8% (28) 33% (122) 52% (189) 4% (15) 3% (12)

BL, Burkitt lymphoma; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma aSome tests included fluorescent in situ hybridization analysis for MYC, BCL2, BCL6 and IG gene rearrangements. In some cases, it included insitu hybridization for Epstein-Barr virus (EBV)-encoded small RNA (EBER) (EBV detection).

centers, they can be easily missed in poor-quality prep- into and analysis of patients in locally relevant clinical trials arations, which are not uncommon in resource-poor will be feasible. Without classifying lymphomas correctly, countries. Hence, a panel that would take help arriving treatment outcomes cannot be measured meaningfully. at a specific diagnosis has been designed. Recently, antibodies such as LEF1 have been 6. Without the use of Ki67, there is a chance of overdiag- introduced and appear to be very specific for CLL.9 nosis of DLBCL in suboptimal-quality material in cases Similarly, utility of antibodies such as IRTA1 and MNDA of FL with diffuse areas and MZL with relatively prom- are being investigated in the diagnosis of MZL.10 inent large cells. Here Ki67 acts as a safety net. Diagnostic algorithms may need further modifications with 7. In FL3b, BCL2 could be negative. In such cases, strong introduction of these “new” antibodies to clinical practice. expression of MUM1 is often helpful. Furthermore, As a spin-off from this study, we can also propose a MUM1 expression dichotomizes FLs into FL grades 1 limited panel of antibodies to be used in units referring sus- to 2 and FL grade 3. Cyclin D1 is used in the panel to pected lymphoma cases (units that do not make the final differentiate from blastoid and pleomorphic variants of diagnosis) to specialist referral centers. This pertains to net- MCL with a nodular pattern. works such as those in the United Kingdom, where lymphoma diagnosis is centralized to SIHMDS. In many such Despite the above reservations, this approach will help in networks, a general pathologist initially evaluates the lesion, usage of the WHO classification of lymphoma in large parts performs immunohistochemistry and forwards cases of of the world with limited resources, as well as impact entry lymphoma or those suspicious of lymphoma to the

Table 4 Final Diagnoses as Percentages (No.) Within Individual Patterns Using Different Algorithms in Phase 2

Pattern/Algorithm (No.) BLL CLL MZL and LPL FL1-3a FL3b Diffuse FL1-2 MCL DLBCL Nodular (159) — — 16% (25) 62% (98) <1% (1) — 13% (20) — Diffuse/interfollicular pattern (120) — 53% (64) 18% (22) — — 6% (7) 16% (19) — Diffuse blastic pattern (26) 69% (18) — — — — — 23% (6) — BL/DLBCL algorithm (47) — — — — — — 87% (41) Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 HIV lymphomas (14) — — — — — — 50% (7) Total (366) 5% (18) 17% (64) 13% (47) 26% (98) <1% (1) 2% (7) 12% (45) 13% (48)

BL, Burkitt lymphoma; BLL, B-lymphoblastic lymphoma; cHL, classical Hodgkin lymphoma; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; FL1-2, follicular lymphoma grades 1-2; FL1-3a, follicular lymphoma grades 1-3a; FL3b, follicular lymphoma grade 3b; MCL, mantle cell lymphoma; MZL, marginal zone lymphoma; PBL, plasmablastic lymphoma; PEL, primary effusion lymphoma; TLL, T-lymphoblastic lymphoma.

Table 5 Table 6 B-Cell Lymphomas Evaluated in the Study Number and Proportion of Cases Where a Specific Diagnosis Could Be Reached and the Number of Antibodies/Tests Used to Lymphoma Numbers Reach a Specific Diagnosisa B-lymphoblastic lymphoma 18 Chronic lymphocytic leukemia 120 No. (%) of Cases Where a Specific Lymphoplasmacytic lymphoma 1 No. of Antibodies/Tests Diagnosis Was Reached Extranodal marginal zone lymphoma, mucosa-asso- 21 3 37 (5) ciated lymphoid tissue type 4-5 548 (69) Nodal marginal zone lymphoma 25 6 631 (79) Follicular lymphoma grades 1-3a 234 7 719 (90) Follicular lymphoma grade 3b 5 8 767 (96) Mantle cell lymphoma 56 9-10 775 (97) Diffuse large B-cell lymphoma 280 11-12 783 (98) Diffuse large B-cell lymphoma with follicular 20 a lymphoma Data combines both phases 1 and 2. Diffuse large B-cell lymphoma with marginal zone 1 lymphoma Burkitt lymphoma 10 B-cell lymphoma, unclassifiable with features over- 5 Table 7 lapping between diffuse large B-cell lymphoma Suggested Short Panels on Adequate Biopsies in Preparation with follicular lymphoma and Burkitt lymphoma of Referral to SIHMDS Plasmablastic lymphoma 1 Histology Panel Extracavitary primary effusion lymphoma 1 B-cell lymphoma, unspecified 1 Suspected “small” cell lymphoma CD20, CD5, CD10, BCL2, Total 799 or lymphoma with a follicular CD23 and cyclin D1 pattern Suspected lymphoma with me- CD20, CD3, CD43 and Ki67 dium- to large-sized cells and SIHMDS. At the SIHMDS, original slides are reviewed, and with a diffuse pattern additional immunostains (with some repeat of immunos- Suspected Hodgkin lymphoma CD20, CD3, CD15 and CD30 tains) are undertaken with or without investigations pertain- ing to FISH and molecular diagnostics before a final diagnosis is issued.8 This two-stage reporting can lead to (a) duplication of work in terms of immunohistochemistry; (b) 1. Not to perform immunostains in small or needle core lack of clarity in the extent to which the case should be biopsies; paraffin block should be referred to investigated at the referring center; (c) delay in referral; and SIHMDS on evaluation of the initial H&E section, (d) depletion of tissue in the paraffin block in small biopsies and paraffin block should not be subjected to any before the referral. additional levels. Some of the essential antibodies such as PAX5 (in the 2. Though a limited panel of immunostains may be per- diagnosis of classical Hodgkin lymphoma) have not been formed, referral should be made within 72 hours of the included in the algorithms. Such essential antibodies would receipt of biopsy. be used at the referral center before treatment is initiated. 3. Suggested short panels on adequate biopsies include the Our proposal for referral of lymphoid lesions in the following (especially if one is working under economic context of SIHMDS networks includes the following: constraints) Table 7 :

694 Am J Clin Pathol 2016;145:687-695 © American Society for Clinical Pathology 694 DOI: 10.1093/ajcp/aqw060 AJCP /ORIGINAL ARTICLE

MZL and Not Classified T-Cell FL and DLBCL DLBCL BL DLBCL/BL PBL PEL B-NHL cHL T-LL Lymphomas 9% (14) — — — — — — — — <1% (1) 5% (6) 1% (1) — — — — 1% (1) — — — ——4% (1) — — — — — 4% (1) — ——2% (1) 11% (5) — — — — — — Downloaded from https://academic.oup.com/ajcp/article-abstract/145/5/687/2195691 by World Health Organization user on 09 January 2019 ——7% (1) — 7% (1) 7% (1) — 29% (4) — — 5% (20) <1% (1) 1% (3) 1% (5) <1% (1) <1% (1) <1% (1) 1% (4) <1% (1) <1% (1)

a. Suspected “small” cell lymphoma or lymphoma with 2. International Agency for Research on Cancer. Worldwide a follicular pattern: CD20, CD5, CD10, BCL2, incidence of and mortality from non Hodgkin lymphoma; Globocan 2012. http://globocan.iarc.fr/Pages/ CD23, and cyclin D1 fact_sheets_population.aspx. b. Suspected lymphoma with medium- to large-sized 3. Naresh KN, Srinivas V, Soman CS. Distribution of various cells and with a diffuse pattern: CD20, CD3, CD43, subtypes of non-Hodgkin’s lymphoma in India: a study of and Ki67 2773 lymphomas using R.E.A.L. and WHO Classifications. Ann Oncol. 2000;11(suppl 1):63-67. c. Suspected Hodgkin lymphoma: CD20, CD3, CD15, 4. Naresh KN, Agarwal B, Nathwani BN, et al. Use of the and CD30. World Health Organization (WHO) classification of non- The above panels should be adequate to separate react- Hodgkin’s lymphoma in Mumbai, India: a review of 200 con- ive lymphoid lesions and nonhematopoeitic lesions, and to secutive cases by a panel of five expert hematopathologists. enable targeted referrals to SIHMDS. This would also allow Leuk Lymphoma. 2004;45:1569-1577. maintaining a certain level of lymphoma diagnostic skills in 5. Naresh KN, Advani S, Adde M, et al. Report of an International Network of Cancer Treatment and Research the referring units. workshop on non-Hodgkin’s lymphoma in developing coun- To conclude, the study and the panels suggested intend tries. Blood Cells Mol Dis. 2004;33:330-337. to (1) allow well trained hematopathologists to practice 6. Naresh KN, Raphael M, Ayers L, et al. Lymphomas in lymphoma diagnostics in countries with limited resources; sub-Saharan Africa: what can we learn and how can we help in improving diagnosis, managing patients and fos- and (2) referral pathways to function in the context of tering translational research? Br J Haematol. SIHMDS networks in countries such as the UK. These algo- 2011;154:696-703. rithms are not intended to be part of a standard operating pro- 7. Naresh KN, Ibrahim HA, Lazzi S, et al. Diagnosis of Burkitt cedure in specialized diagnostic services such as SIHMDS. lymphoma using an algorithmic approach — applicable in both resource-poor and resource-rich countries. Br J Haematol. 2011;154:770-776. Corresponding author: Kikkeri Naresh, Hammersmith Hospital, 8. Proctor IE, McNamara C, Rodriguez-Justo M, et al. Du Cane Road, London, W12 0HS, UK; [email protected]. Importance of expert central review in the diagnosis of lymphoid malignancies in a regional cancer network. J Clin Oncol. 2011;29:1431-1435. 9. Menter T, Dirnhofer S, Tzankov A. LEF1: a highly specific marker for the diagnosis of chronic lymphocytic B cell leu- kaemia/small lymphocytic B cell lymphoma. J Clin Pathol. References 2015;68:473-478. 1. Jaffe ES, Harris NL, Stein H, et al. Introduction and over- 10. van den Brand M, Mathijssen JJ, Garcia-Garcia M, et al. view of the classification of the lymphoid neoplasms. In: Immunohistochemical differentiation between follicular Swerdlow SH, Campo E, Harris NL, et al, eds. WHO lymphoma and nodal marginal zone lymphoma - combined Classification of Tumours of Haematopoietic and Lymphoid performance of multiple markers. Haematologica. Tissues. Lyon, France: IARC; 2008:158-166. 2015;100:e358-e360.