1084 Arch Pathol Lab MedVol 131, July 2007 Grading Follicular Lymphoma: Histology Versus IHCMartinez et al
Grading of Follicular Lymphoma
Comparison of Routine Histology With Immunohistochemistry Antonio E. Martinez, MD; Li Lin, MS; Cherie H. Dunphy, MD Context.Follicular lymphoma (FL) grading is based on the average number of large transformed cells in 10 neo- plastic follicles at 40 high-power eld (1040 high- power eld) examination (grade 1, 05 centroblasts per high-power eld; grade 2, 615 centroblasts per high-pow- er eld; grade 3, 15 centroblasts per high-power eld). Objective.Since there may be signicant interobserver variability, we analyzed the usefulness of immunohisto- chemical stains in grading FLs more reliably. Design.Forty-three FLs initially graded by World Health Organization criteria (grade 1, 12; grade 2, 18; grade 3, 13) were reviewed and stained with CD3, CD20, Ki-67, CD30, CD68, PAX-5, and BCL-6. Retrospective re- view was performed for the average number of large cells, of large lymphoid cells, of large cells staining with CD3, CD20, BCL-6 (40 cases), and PAX-5, and of all cells stain- ing with CD68, Ki-67, and CD30. Results.By histologic review, 8 of 43 FLs had a signif- icant grade change (4 cases upgraded and 4 cases down- graded). CD3 and CD30 stained only 0 to 3 large cells and 0 to 3 cells, respectively, in neoplastic follicles. CD68
cells represented the large nonlymphoid cells. Increasing
FL grades demonstrated increases in Ki-67
cells. The orig-
inal grade showed substantial agreement with CD20 and moderate agreement with PAX-5 and BCL-6. The original histologic grade agreed with immunohistochemical-based grade using 2 or more antibodies in 5 of 8 discordant cases (4 by CD20 or BCL-6 and PAX-5; 1 by CD20, PAX-5, and BCL-6). Conclusions.Interobserver variability of histologic FL grading may be signicant; we showed low-end substan- tial agreement. Immunohistochemical stains (ie, CD20, PAX-5, and BCL-6) may more reliably determine the num- ber of large transformed cells in neoplastic follicles; Ki-67 staining correlates with higher FL grades. Immunohisto- chemical stains may be evaluated in clinical trials of FL patients to determine prognostic signicance. (Arch Pathol Lab Med. 2007;131:10841088) F ollicular lymphoma (FL) comprises up to 70% of the low-grade lymphomas studied in US clinical trials. 1 Histologic grading can predict clinical outcome. The World Health Organization (WHO) classication recom- mends a 3-grade system that is based on the average num- ber of centroblasts or large transformed cells (LTCs) in 10 neoplastic follicles at 40 high-power eld (HPF) ex- amination (ie, grade 1, 05 centroblasts per HPF; grade 2, 615 centroblasts per HPF; and grade 3, 15 centroblasts per HPF). 2 These criteria, established 2 decades ago, re- sulted from studies using the counting method attributed to Mann and Berard, which demonstrated prognostic sub- grouping of FL. 3,4 It is known that this method suffers from lack of reproducibility among pathologists, primarily Accepted for publication December 20, 2006. From the Division of Hematopathology, Department of Pathology and Laboratory Medicine (Drs Martinez and Dunphy) and the Division of Biostatistics (Dr Lin), Lineberger Cancer Center, University of North Carolina, Chapel Hill. Dr Martinez is now with the Department of Pathology and Laboratory Medicine, Mount Sinai Medical Center, Mi- ami, Fla. The authors have no relevant nancial interest in the products or companies described in this article. Presented as a poster at the annual meeting of the United States and Canadian Academy of Pathology, Atlanta, Ga, February 2006. Reprints: Cherie H. Dunphy, MD, Department of Pathology and Lab- oratory Medicine, CB# 7525, University of North Carolina, Chapel Hill, NC 27599-7525 (e-mail: cdunphy@unch.unc.edu). due to the subjective nature in identifying LTCs, which are morphologically heterogeneous. 5 In the current era of molecular technology, microarray assays capable of producing gene expression proles have dened subgroups of FL, which may supplement the mor- phologic classication used to guide therapy. 6,7 However, such technology is not widely available, and clinical trials to validate the assays utility have not been performed. The most widely available ancillary tool to practicing pathologists is immunohistochemistry (IHC), which has been used in diagnostic surgical pathology for more than 2 decades. The literature describing the utility of IHC in the grading of FL is scant, with a single study showing no signicant improvement in the diagnostic accuracy with its use. 5 The current study was undertaken to com- pare histologic versus immunohistochemical grading of FL using a panel of immunohistochemical antibodies. MATERIALS AND METHODS Case Selection and Subsequent Histologic Grading of FL Forty-three cases of FL initially graded according to WHO cri- teria between 2002 and 2005 were retrieved from our les and retrospectively reviewed. Two observers, the original hematopa- thologist and one of us (A.E.M.), graded the cases. The distri- bution of cases by original histologic grade was as follows: grade 1, 12; grade 2, 18; and grade 3, 13. A.E.M. retrospectively graded the cases by WHO criteria in a blinded fashion (ie, without knowledge of the original grade). Retrospective hematoxylin-eo- sin review for the average number of all large cells (LCs) and Arch Pathol Lab MedVol 131, July 2007 Grading Follicular Lymphoma: Histology Versus IHCMartinez et al 1085 Table 1. Summary of Immunohistochemical Stains Used Antibody Source Clone Staining Pattern CD3 Dako Corporation, Carpinteria, Calif Polyclonal rabbit Membranous CD20 Dako L26 Membranous CD30 Dako Ber-H2 Cytoplasmic and membranous CD68 Dako PG-M1 Cytoplasmic PAX-5 Santa Cruz Biotechnology, Santa Cruz, Calif C-20 Nuclear BCL-6 Cell Marque, Hot Springs, Ariz PG-B6p Nuclear Ki-67 Dako MIB-1 Nuclear Table 2. Measurement of Interobserver Agreement for Categorical Data 16 Value Interpretation 0.00.19 Poor agreement 0.200.39 Fair agreement 0.400.59 Moderate agreement 0.600.79 Substantial agreement 0.801.00 Almost perfect agreement Table 3. Weighted Coefcients for Interobserver Agreement Comparison Weighted (95% Condence Interval) Original grade vs retrospective count 0.59 (0.41, 0.78) Original grade vs CD20 0.65 (0.48, 0.82) Original grade vs BCL-6 0.53 (0.34, 0.73) Original grade vs PAX-5 0.56 (0.38, 0.75) Retrospective count vs CD20 0.71 (0.55, 0.87) Retrospective count vs BCL-6 0.57 (0.38, 0.76) Retrospective count vs PAX-5 0.63 (0.45, 0.81) large lymphoid cells, or LTCs, in 10 to 40 HPFs were tab- ulated. The grading of FL was performed by counting the num- ber of LTCs in 10 HPFs (20-mm eld of view) on hematoxylin- eosinstained sections, and obtaining an average number per HPF, per current WHO classication and grading of FL. 2 An LTC was morphologically dened as 2 to 5 times the size of a normal lymphocyte, with usually round to oval, but occasionally indent- ed, nuclei with vesicular chromatin, 1 to 3 peripheral nucleoli, and a narrow rim of cytoplasm. A minimum of 10 neoplastic follicles was required for counting. In a similar fashion, the num- ber of mitotic gures was recorded per HPF. However, 1 case did not have enough elds for counting mitoses; thus, there were only 42 cases evaluated. Immunohistochemistry The formalin-xed, parafn-embedded blocks of the 43 FLs were stained manually according to kit procedures, with the im- munohistochemical stains outlined in Table 1. The interpretation of the immunohistochemical stains was based on counting the number of LCs staining in 10 neoplastic follicles at 40 (HPF) and obtaining an average number of LCs per HPF for CD3, CD20, BCL-6 (40 cases), and PAX-5. The average numbers of all cells staining with CD68, Ki-67, and CD30 were similarly tabulated. The determination of a positive or negative result for IHC stain- ing was based on the pattern of reactivity for the respective an- tibody on control lymphoid tissue (eg, nuclear staining for BCL-6). Positive and negative controls were prepared and re- viewed for each antibody. Statistical Analysis In order to evaluate the agreement between observers, the weighted coefcient was computed, with equal to 1 when there is perfect agreement and equaling 0 when the agreement equals that expected by chance. Table 2 illustrates the measure- ment of interobserver agreement for categorical data used in this study. Jonckheere-Terpstra, which is a nonparametric test, was used to test for the ordered differences in original grade across mitotic scores. It tested the null hypothesis that the distribution of the original grade does not differ among classes of different mitotic score versus alternatives of ordered class differences (original grade increases directly with increasing mitotic count). Logistic regression was used to analyze the association be- tween original grade and Ki-67. Statistical analyses were per- formed with SAS statistical software, version 9.1 (SAS Institute Inc, Cary, NC). RESULTS Comparison of Original Histologic Grade of FL With Subsequent Histologic Grading by Counting of LTCs and Mitoses By the standardized WHO criteria of the histologic counting method, the interobserver agreement (ie, original histologic grade vs retrospective count) was on the low end of substantial agreement, as demonstrated by a val- ue of 0.60 (Table 3). Retrospectively, 8 of 43 cases with IHC-based grading had a signicant histologic grade change (from grade 2 to grade 3 or vice versa) by the histologic counting method (4 cases were upgraded and 4 cases were downgraded). The original histologic grade increased signicantly with an increasing number of mitoses counted (exact P .01, 2-sided). Table 4 shows the observed distribution of the original grade with an increasing number of counted mitoses and expected cell counts under the null hypoth- esis. IHC of 43 Cases of FL CD3. CD3-positive LCs were rare in the neoplastic fol- licles, ranging from 0 to 3 LCs and averaging less than 1 positive LC per HPF. The CD3-positive cells encountered in the follicles and in the interfollicular regions were pre- dominantly small lymphocytes. CD30. CD30-positive cells were also rare in the neo- plastic follicles, ranging from 0 to 3 of the total cells and averaging less than 1 positive cell per HPF. CD68. CD68-positive cells accounted primarily for the LCs not counted as LTCs in the neoplastic follicles. The number of CD68-positive cells ranged from 10 to 31, with an average of 16 positive cells per HPF. This immunohis- tochemical stain highlighted individual macrophages and some paired cells with cytoplasmic projections, possibly representing follicular dendritic cells (Figure 1). CD20. CD20 demonstrated membranous staining of neoplastic cells in the follicles and interfollicular areas. This antibody highlighted the cell size (Figure 2), exclud- 1086 Arch Pathol Lab MedVol 131, July 2007 Grading Follicular Lymphoma: Histology Versus IHCMartinez et al Table 4. Observed Distribution of Original Histologic Grade With Increasing Number of Mitoses and Expected Cell Counts Under Null Hypothesis Mitotic Level Original Grade* 1 2 3 Total 0 7/4.2 7/6.5 2/5.3 16 1 3/2.4 2/3.6 4/3 9 2 0/1.6 3/2.4 3/2 6 3 1/2.1 4/3.2 3/2.7 8 4 0/0.5 0/0.8 2/0.7 2 5 0/0.3 1/0.4 0/0.3 1 Total 11 17 14 42 * Values are observed count/expected count. Figure 1. CD68 immunohistochemistry shows cytoplasmic staining of a pair of possible follicular dendritic cells in the neoplastic follicles, highlighting the cytoplasmic processes (original magnication 600). Figure 2. CD20 immunohistochemistry shows membranous staining of the neoplastic B cells, highlighting the large cells of a grade 3 fol- licular lymphoma (original magnication 600). Figure 3. PAX-5 immunohistochemistry shows nuclear staining of neoplastic B cells, highlighting the large cells of a grade 3 follicular lymphoma (original magnication 600). ed the macrophages and/or dendritic cells by the absence of staining, and highlighted the architectural features of the neoplastic process. This latter nding was important in dening the percentage of the follicular component and was best observed at low power. BCL-6 and PAX-5. Both PAX-5 and BCL-6, which stain neoplastic as well as nonneoplastic B cells, revealed nucle- ar staining of variable intensity. As with CD20, the neo- plastic cells within the follicles and interfollicular areas were highlighted, and the size of such cells was discern- able. The nuclear features, specically the presence of nu- cleoli, in the LCs was appreciable by ne adjustment of microscopic focusing (Figure 3). Ki-67. Ki-67, which stains all cells in cycle, revealed nuclear staining of the neoplastic cells in the follicles and scattered in the interfollicular areas. As supported by the statistical analysis, FLs of higher grades demonstrated a greater number of Ki-67positive cells (Figure 4, A [grade 1 FL] and B [grade 3 FL]). Grade Change by Counting Method and IHC As previously mentioned, retrospectively, 8 cases had a signicant histologic grade change (from grade 2 to grade 3 or vice versa) by the histologic counting method (4 cases were upgraded and 4 cases were downgraded). However, in 5 of these 8 cases the original histologic grade agreed with immunohistochemical-based grade by at least 2 an- tibodies: in 3 cases, the original grade agreed with CD20 and PAX-5 staining; in 1 case, with BCL-6 and PAX-5 staining; and in 1 case, with CD20, PAX-5, and BCL-6 staining. Statistical Analysis The level of agreement between the original histologic grade and immunohistochemical grading improved with CD20 (original grade vs CD20), with a value of 0.65. However, there were slightly decreased values for PAX- 5 (original grade vs PAX-5: value of 0.56) and BCL-6 (original grade vs BCL-6: value of 0.53). These latter values are within moderate agreement (Table 3). Of interest, there was better, but not signicantly better, agreement between the retrospective count and immu- nohistochemical grading with CD20 ( value increased from 0.65 to 0.71), BCL-6 ( value increased from 0.53 to 0.57), and PAX-5 ( value increased from 0.56 to 0.63). For investigating the statistical association between orig- inal histologic grade and Ki-67, we applied the data to 2 logistic regression models: (1) original grade (1, 2, or 3) as response variable and rank of Ki-67 as explanatory vari- able; and (2) collapse original grades 1 and 2 together, Arch Pathol Lab MedVol 131, July 2007 Grading Follicular Lymphoma: Histology Versus IHCMartinez et al 1087 Figure 4. A, Ki-67 immunohistochemistry shows nuclear staining in a grade 1 follicular lymphoma with a relatively low proliferative index (original magnication 400). B, Ki-67 immunohistochemistry shows nuclear staining in a grade 3 follicular lymphoma with a relatively high proliferative index (original magnication 400). then use the new dichotomous grade as response variable, with Ki-67 remaining in raw scale as an explanatory var- iable. Both models indicated signicant association be- tween original grade and Ki-67 (P .001 in model 1; P .04 in model 2). Increased Ki-67 positivity tended to cor- respond to higher histologic grade. COMMENT Follicular lymphoma is the most common lymphoma in the Western hemisphere. 8 The counting method attributed to Mann and Berard is currently used in the histologic grading of FL and is the method recommended by the WHO. 2,3 Along with clinical prognostic indices, histologic grade continues to be a useful component in the diagnosis and treatment of FL. Gene expression proling has shown to be highly ac- curate in distinguishing low-grade from high-grade dis- ease in FL. 6 This method also has shed light into other molecular aspects of FL, showing immune response sig- natures of the nonmalignant cells that were predictive of survival. 7 While such techniques may improve prognosis beyond the histologic grading scheme and clinical param- eters currently used, they are not available to most prac- ticing pathologists, and they have not been validated in clinical trials. Flow cytometric analysis has been employed in the grading of FL by using gating strategies to analyze the large cells and by analyzing the S-phase fraction of the neoplastic cells. 9,10 Mourad et al 9 demonstrated a signi- cant difference in the number of large cells in low-grade (grades 1 and 2) versus high-grade (grade 3) FL by ow cytometry. While ow cytometry may be a useful ancil- lary test, the major limitations are the limited viability of the large cells and the availability of representative tissue for analysis. Immunohistochemistry is currently the most valuable ancillary tool in the practice of diagnostic surgical pa- thology, and it has achieved great importance in hemato- pathology as well. As it relates to FL, one of the earliest immunohistochemical markers to achieve diagnostic im- portance was BCL-2, which is generally used to distin- guish reactive from neoplastic follicular processes. 11 The immunohistochemical markers of germinal center differ- entiation, BCL-6 and CD-10, have demonstrated prognos- tic signicance in FL. 12 In addition to these, a recent im- munohistochemical study using additional markers of B- cell and germinal center differentiation showed a linear decline in expression of markers in higher grades of FL, adding to the prognostic signicance of IHC. 13 The literature describing the utility of IHC in improving the accuracy of grading FL is limited to a single reference, which showed no improvement with its use. 5 In addition, this study did not describe the number or type of anti- bodies used, or the specic details of the immunohisto- chemical evaluation of FL. Therefore, we set out to evaluate a panel of antibodies to assess whether or not their use would result in an improvement in the grading of FL. In our study, by statistics, the interobserver agreement improved with the use of CD20, showing substantial agreement, and slightly decreased with BCL-6 and PAX- 5, which showed moderate agreement (Table 3). In cases that had a signicant grade change by the counting meth- od, IHC staining showed agreement with 5 of 8 cases by CD20, BCL-6, and PAX-5 in various combinations. Although these results do not demonstrate a signicant improvement in the grading of FL, several practical con- clusions could be drawn from the current study: The membranous staining pattern of CD20 can be used to highlight the larger cells while preserving the nuclear morphology, which is readily appreciated with the he- matoxylin counterstain. The lack of staining of macro- phages and/or follicular dendritic cells also is of use, as these can be confused with LTCs by hematoxylin- eosin evaluation alone. The nuclear staining of PAX-5 and BCL-6 highlight the size of the B cells in the neoplastic follicles and, in the case of the latter, prognostic information may be pro- vided by the staining pattern. While nuclear staining obscures the chromatin detail, ne adjustment of the microscopic focus allows for the visualization of the nu- cleoli seen in the positive B cells. The proliferative activity, as evidenced by Ki-67 nuclear staining and mitotic counts, demonstrated a positive re- lationship with the histologic grade and number of Ki- 67positive cells (Figure 4, A and B; Table 4). While the correlation of proliferative index with histo- logic grade was demonstrated in the current study and by 1088 Arch Pathol Lab MedVol 131, July 2007 Grading Follicular Lymphoma: Histology Versus IHCMartinez et al others, a recent report of low-grade FL with a high pro- liferative index showed a clinical behavior similar to a grade 3 FL. 14,15 Whether this will be borne out in larger series remains to be seen. In summary, immunohistochemical markers of B-cell and germinal center differentiation were shown to be use- ful in identifying LTCs in the grading of FL. In conjunc- tion with the proliferation marker Ki-67 and mitotic activ- ity, cases on the fringe of low-grade versus high-grade FL by the standardized WHO criteria of histologic grading alone may be more accurately categorized by the use of IHC. A recommended panel includes CD20, Ki-67, PAX- 5, and BCL-6. Such immunohistochemical stains may be evaluated in well-designed clinical trials of FL patients to determine prognostic signicance and possible therapeu- tic implications. References 1. Glass AG, Karnell LH, Menck HR. The National Cancer Data Base report on non-Hodgkins lymphoma. Cancer. 1997;80:23112320. 2. Jaffe ES, Harris NL, Stein H, Vardiman JW, eds. Pathology and Genetics of Tumours of Haematopoietic and Lymphoid Tissues. Lyon, France: IARC Press; 2001. World Health Organization Classication of Tumours; vol 3. 3. Mann RB, Berard CW. Criteria for the cytologic subclassication of follicular lymphomas: a proposed alternative method. Hematol Oncol. 1983;1:187192. 4. Nathwani BN, Metter GE, Miller TP, et al. What should be the morphologic criteria for the subdivision of follicular lymphomas? Blood. 1986;4:837845. 5. The Non-Hodgkins Lymphoma Classication Project. A clinical evaluation of the international lymphoma study group classication of non-Hodgkins lym- phoma. Blood. 1997;89:39093918. 6. Glas AM, Kersten J, Delahaye LJMJ, et al. Gene expression proling in fol- licular lymphoma to assess clinical aggressiveness and to guide the choice of treatment. Blood. 2005;105:301307. 7. Dave SS, Wright G, Tan B, et al. Prediction of survival in follicular lympho- ma based on molecular features of tumor-inltrating immune cells. N Engl J Med. 2004;351:21592169. 8. Muller AM, Ihorst G, Mertelsmann R, et al. Epidemiology of non-Hodgkins lymphoma (NHL): trends, geographic distribution, and etiology. Ann Hematol. 2005;84:112. 9. Mourad WA, Rawas F, Shoukri M, et al. Grading follicular lymphoma by using ow cytometry. Ann Saudi Med. 2006;26:205210. 10. Macartney JC, Camplejohn RS, Morris R, et al. DNA ow cytometry of follicular non-Hodgkins lymphoma. J Clin Pathol. 1991;44:215218. 11. Dunphy CH. Applications of ow cytometry and immunohistochemistryto diagnostic hematopathology. Arch Pathol Lab Med. 2004;128:10041022. 12. Bilalovic N, Blystad KB, Golouh R, et al. Expression of bcl-6 and CD10 protein is associated with longer overall survival and time to treatment failure in follicular lymphoma. Am J Clin Pathol. 2004;121:3442. 13. Torlakovic EE, Bilalovic N, Golouh R, et al. Prognostic signicance of PU.1 in follicular lymphoma. J Pathol. 2006;209:352359. 14. Martin AR, Wesenburger DD, Chan WC, et al. Prognostic value of cellular proliferation and histologic grade in follicular lymphoma. Blood. 1995;85:3671 3678. 15. Wang SA, Wang L, Hochberg EP, et al. Low histologic grade follicular lymphoma with high proliferation index: morphologic and clinical features. Am J Surg Pathol. 2005;29:14901496. 16. Landis JR, Koch GG. The measurement of observer agreement for cate- gorical data. Biometrics. 1977;33:159174.