Differentiating Benign From Malignant Bone Marrow B-Cell Lymphoid Aggregates A Statistical Analysis of Distinguishing Features

(cid:2) Context. — Lymphoid aggregates are seen in a minority of bone marrow biopsy specimens, and when present, their neoplastic nature is often apparent by morphologic evaluation. However, the distinction between benign and malignant aggregates can be a diagnostic challenge when there are multiple aggregates with no documented history of lymphoma. Objective. — To aid in the distinction between benign and malignant B-cell lymphoid aggregates. described distribution for within To we and between Results. with infiltrative edges (OR, CI, a B-cell pattern (OR, 30.08; 95% CI, paratrabecular (OR, 95% CI, 3.96–26.12), size greater than 600 l m (OR, 6.83: 95% CI, 3.61–12.93), or cytologic atypia correlated with

L ymphoid aggregates are relatively uncommon in bone marrow biopsy specimens, and when present, their distinction from non-Hodgkin lymphoma, particularly B-cell lymphomas, can represent a diagnostic challenge. Criteria have been suggested in a limited number of studies to aid in that distinction. It has been reported that benign lymphoid aggregates tend to have a nonparatrabecular location, distinct borders without interstitial spillage of lymphoid cells, and are typically small (,600 lm). 1,2 Malignant characteristics of lymphoid aggregates include predominance of B cells, infiltrative edges, paratrabecular location, inclusion of fat cells, location surrounding large sinuses, and large aggregates. [1][2][3][4][5] Moreover, the loss of benign aggregates in deeper sections is considered a prominent indicator of a benign process. 6,7 In a previous study, 8 we demonstrated distinct immunomorphologic patterns of B and T cells in benign and malignant aggregates. Associations between these various characteristics, including the distribution patterns of B and T lymphocytes, and the risk of a B-cell lymphoma in patients who have these bone marrow findings, have not been quantified.
The presence of lymphoid aggregates in the bone marrow has been reported to be more frequently associated with certain conditions including aging, autoimmune diseases, inflammatory conditions, and infectious disorders. 2,9 They have also been reported to be commonly identified in patients with myeloproliferative neoplasms, especially primary myelofibrosis, 4 and rarely in association with myelodysplastic syndromes. 10 Moreover, an increased incidence of benign aggregates in patients with lymphoma who have been treated with rituximab has also been reported. 11,12 These aggregates are often found in postchemotherapy bone marrow specimens and can mimic residual lymphoma. 12,13 Lymphoid aggregates have also been rarely reported in association with tobacco use and certain medications. 14,15 The use of ancillary studies, such as immunohistochemistry, flow cytometry, and molecular analysis, to assist in distinguishing between benign and malignant aggregates has been investigated before, although in a relatively limited fashion. [15][16][17][18] Previous studies 13,19,20 have used only a limited antibody panel, or have only investigated the distinction between benign lymphoid aggregates and follicular lymphoma. We have recently reported a detailed description of the distribution patterns of B and T lymphocytes within bone marrow lymphoid aggregates, and the diagnostic utility of recognizing these patterns. 8 In the current study, we extend our observations by examining the different morphologic and immunophenotypic characteristics of a large number of both benign and malignant bone marrow lymphoid aggregates and apply statistical tools to validate previously proposed criteria for the distinction between benign and malignant B-cell aggregates.

MATERIALS AND METHODS
A search of the pathology files at University of California Irvine Medical Center in Orange, California, from 1993 to 2011 identified 206 bone marrow biopsy cases with reported lymphoid aggregates. A subset of these cases was used in a previous study. 8 All specimens were formalin fixed and included the trephine and/or the clot biopsy specimens. To be considered adequate, a minimum of 3 marrow spaces was required. In a few cases, tissue loss in deeper sections used for immunohistochemical staining was encountered. If the loss of tissue involved the lymphoid aggregates of interest, such cases were excluded from the study. Of all cases, 128 contained benign lymphoid aggregates, while 78 contained malignant B-cell lymphoid aggregates. The benign aggregates were defined from the original morphologic and immunophenotypic examination that included immunohistochemistry and/or flow cytometry testing. Flow cytometry was performed on 82 cases, and 28 cases had B-cell gene rearrangement studies performed by polymerase chain reaction (PCR). Furthermore, all specimens were further reviewed by us and additional stains were used for confirmation of diagnosis in a subset of the cases. Cases with malignant aggregates include 23 cases of chronic lymphocytic leukemia, 18 cases of follicular lymphoma, 8 cases of lymphoplasmacytic lymphoma, 7 cases of mantle cell lymphoma, 9 cases of marginal zone lymphoma, 2 cases of a splenic marginal zone lymphoma, and 11 cases of diffuse large B-cell lymphoma with nodular aggregates. Flow cytometry was performed on 64 cases and B-cell gene rearrangement studies by PCR on 3 cases.
A panel of immunostains was performed on all cases, which included CD3 (polyclonal 1:200; Dako, Carpinteria, California) and CD20 (L26, 1:400; Dako). CD79a (JCB117, 1:200; Dako) immunostaining was done on cases with a history of treatment by rituximab. In addition, other markers were used in select cases to further clarify and confirm the nature of the aggregates. These Histologic characteristics of the aggregates were retrospectively evaluated. These features included the presence of infiltrative edges, paratrabecular location, cytologic atypia, and size of lymphoid aggregates. Infiltrative edges were defined as lymphoid aggregates with poorly circumscribed margins and with lymphocytes at the edge extending into the surrounding interstitium. Assessment of infiltrative edges was best appreciated by using CD3/CD20 immunostains. Paratrabecular localization was defined as aggregates intimately opposed to the marrow bony trabeculae, while aggregates in close proximity to the bony trabeculae were excluded. Cytologic atypia included cells that exhibited open chromatin, irregular nuclear contours, prominent nucleoli, or karyorrhexis. Large aggregates included those greater than 600 lm in greatest dimension. Lymphoid aggregates exhibiting a predominance of T cells, haphazard mixtures of B and T cells, or central cores of T cells surrounded by a rim of B cells were characteristic of a T-cell pattern as previously described. 8 Aggregates composed of predominantly B cells or central cores of B cells surrounded by T cells (excluding germinal center formation) were identified as B-cell patterns. 8 Cases with more than 1 pattern were categorized in accordance with the dominant pattern; however, the presence of a B-cell pattern was used to define a case. The cases were recorded as either exhibiting or not exhibiting these characteristics. Variables including age and sex were also examined.
Follow-up information for patients with reactive lymphoid aggregates was available for 104 patients (81%) with a duration that varied from 1 year to 17 years. Thirty patients had 1-year follow-up (29%), 22 patients had 2-year follow-up (21%), 24 patients had 3 to 5 years follow-up (23%), and 28 patients had more than 6 years follow-up (27%). Characteristics of the lymphoid aggregates were compared to the final benign or malignant diagnosis. The odds ratios (ORs) were calculated to assess the strength of the relationship between individual and combined characteristics, and the benign or malignant diagnoses. We further examined combinations of malignant characteristics to determine if multiple features were predictive of malignancy. The statistical difference between benign and malignant diagnoses with regard to each of 5 malignant characteristics was calculated by using the Pearson v 2 test. After evaluating cases based solely on the number of malignant characteristics present, we examined the cases with regard to which malignant characteristics were present. These additional comparisons were evaluated with the Pearson v 2 test. Furthermore, age and sex differences were also assessed. The statistical analyses were performed by using XLSTAT (version 2013.1.01, Addinsoft, Paris, France).

RESULTS
Histologic and immunophenotypic characteristics of the aggregates were evaluated in all cases. The age range among the reactive cases was 12 to 99 years with a mean age of 57 years. This was similar to the age range of the malignant cases, which varied from 31 to 93 years with a mean age of 57 years. Females represented 54% of the reactive cases as opposed to 45% of the malignant cases (Z ¼ 1.26, P ¼ .21). Eighty-five of the benign cases (66%) and 76 malignant cases (97%) had aggregates present in the trephine biopsy sections, which permitted accurate evaluation for the presence of paratrabecular localization. The remaining cases contained aggregates only in the clot biopsy sections. Of the 128 benign cases, 101 cases (80%) displayed a T-cell pattern (34 cases with predominance of T cells, 50 cases with a mixture of T and B cells, and 18 cases with a core of T cells surrounded by B cells) (Figure 1, all panels). No residual germinal centers were identified in cases with a predominant B-cell pattern with bcl-6, CD23, and PD-1 stains. Forty benign cases (31%) contained large aggregates greater than 600 lm and fewer than 15% of these cases had infiltrative edges or paratrabecular localization ( Figure 2). No benign cases were found to have atypical lymphocytes. In contrast, for the malignant cases, B-cell patterns, infiltrative edges, and large size were the predominant characteristics (Table  1). Notably, more than 88% of the malignant cases had lymphoid aggregates with a B-cell pattern, while 85% of the cases exhibited infiltrative edge. The next highest percentage was for large aggregates at 76% of malignant diagnoses.
The ORs and 95% confidence intervals (CIs) for the presence of each characteristic were calculated and are listed in Table 2. The malignant cases had statistically greater odds of having each of the following characteristics: infiltrative edges, B-cell pattern, paratrabecular location, and large size. The highest ORs were found for infiltrative edges (OR, 80.54; 95% CI, 31.76-204.21) and B-cell patterns (OR, 30.08; 95% CI, 13.28-68.10). Malignant cases with T-cell patterns exhibited an OR of 0.03 and 95% CI of 0.01 to 0.08, indicating decreased odds of these patterns. The number of malignant lymphoid aggregate characteristics exhibited in each case was tallied. Seventy-three of the benign cases (57%) lacked malignant characteristics, whereas only 1 malignant case did not exhibit any of the malignant characteristics ( Table 3). The percentage of malignant cases increased with addition of 1 or 2 characteristics (8.97% and 17.95%, respectively), and nearly a quarter of the cases exhibited 3, 4, or 5 characteristics each. The percentage of benign cases decreased with each additional characteristic, from 57.03% with no characteristics, to 25.00% with 1 characteristic, 15.63% with 2 characteristics, and only 2.34% with 3 characteristics. No benign cases exhibited 4 or 5 malignant characteristics (Table 3).
Cases with no or only 1 malignant characteristic were associated with benign diagnoses (v 2 (1,N ¼ 206) ¼ 65.44, P , .001 and v 2 (1,N ¼ 206) ¼ 8.11, P ¼ .004, respectively). There was no significant difference between the number of benign and malignant cases showing 2 malignant characteristics (v 2 (1, N ¼ 206) ¼ 0.08, P ¼ .78). When lymphoid aggregates had 3 or more malignant characteristics, they were associated with malignant diagnoses (v 2 (1, N ¼ 206) ¼ 28.45, P , .001). The additive effect of characteristics was also evaluated on the basis of the strength of association with malignancy to see if it mattered which combination of characteristics was present. As the presence of a B-cell pattern was the characteristic found in the highest percentage of malignant cases (88.46%), we started with this feature, and then tallied the cases with a B-cell pattern and infiltrative edges (the next highest percentage in the malignant cases at 85.90%) and so on. The results are shown in Table 4. Twenty-six benign cases had B-cell patterns but only 2 benign cases had a B-cell pattern and infiltrative edges. One of these 2 cases also contained a large aggregate and comes from a human immunodeficiency virus (HIV)positive individual. Specimens from HIV-positive patients are summarized in Table 5. No benign cases had more than 3 malignant characteristics when evaluated in this order. Greater than 88% of malignant cases had a B-cell pattern, while 77% had both a B-cell pattern and infiltrative edges.  There were 23% of malignant cases with all 5 malignant characteristics. For every category of characteristics identified, v 2 tests detected significant differences between the benign and malignant cases. Broad approximation of the ORs, relative strengths of associations among the characteristics, and statistical significance can be gleaned from these data. However, the 95% CIs for several of our ORs are wide, necessitating future examination of a larger cohort of bone marrow specimens to provide better approximations of the true OR. Some caution must also be used when interpreting our results as there may be a case selection bias. For cases to be included in our study, documentation of lymphoid aggregates had to be described in the report, which may have led to missing cases with lymphoid aggregates that were not documented.
COMMENT Although the identification of lymphoid aggregates in the bone marrow has been reported by some to be detected in up to 13% of older individuals, 15,21 it is overall a relatively uncommon finding in bone marrow biopsies. 1,2 Recently, we described distinct B-and T-lymphocyte distribution patterns in bone marrow lymphoid aggregates, where lymphoid aggregates with a T-cell pattern (predominantly T cells, haphazard mixture of T and B cells, and centrally located T cells) were predictive of a benign process when correlated with the patient's clinical features, diagnosis, and follow-up specimens, while the predominance of B cells or the presence of a central core of B cells surrounded by T cells was worrisome for malignancy. 8 In this study, we analyzed characteristics that have been reported to be useful in distinguishing benign from malignant B-cell lymphoid aggregates. Bone marrow specimens with malignant aggregates were found to have more than 30 times the likelihood of having a B-cell pattern in the lymphoid aggregates than bone marrows with benign aggregates (OR, 30.08; 95% CI, 13.28-68.10). The odds of having infiltrative edges were even greater (OR, 80.54; 95% CI, 31.76-204.21). Paratrabecular location and large aggregates were also significantly higher in the malignant bone marrows ( Figure 2). As expected, a Tcell pattern had an OR below 1 (OR, 0.03; 95% CI, 0.01-0.08) indicating that T-cell patterns were closely associated with a benign diagnosis, rather than a malignant diagnosis.
From the hypothesis that increasing the number of malignant characteristics found in a case would be positively associated with malignancy, we found that cases displaying none of these characteristics were strongly associated with benign diagnoses. Surprisingly, cases with only 1 malignant characteristic were also strongly associated with benign diagnoses. The presence of 2 characteristics was frequently found in both benign and malignant cases and did not favor one diagnosis over the other. On the other hand, when 3 or more malignant characteristics were identified, nearly all the cases were malignant. This suggests that when evaluating lymphoid aggregates in the bone marrow, the presence of 2 characteristics should raise suspicion of malignancy but can be frequently seen in benign aggregates as well. The presence of 3 malignant characteristics is a reasonable threshold for raising a strong suspicion of malignancy. Also, when taking into account the relative strength of the ORs, certain characteristics, such as B-cell patterns and infiltrative edges, had stronger relationships with malignant cases than large size or paratrabecular location. In our data set, 7% of benign cases had infiltrative edges but only 2 benign cases had both a B-cell pattern and infiltrative edges, 1 of which was an HIV case. In contrast, 77% of the malignant cases had both of these characteristics. Therefore, considering which malignant characteristics are identified is critical in attempting to predict malignancy, especially when only 1 or 2 characteristics are present. Although paratrabecular location is one of the features that historically has been strongly suggestive of malignancy even on a stand-alone basis, 6 of 85 reactive cases with trephine biopsy sections in our study (7%) contained aggregates with a paratrabecular    location. Careful review of these cases showed no definitive evidence of malignancy. Table 6 shows detailed description of these 6 cases along with available follow-up information, and PCR and flow cytometry results. Additionally, information is included for 3 cases in our benign series that had nonparatrabecular lymphoid aggregates but in which some of the lymphocytes touched the paratrabecular bone. Touching of the bony by benign lymphoid aggregates without surrounding the entire paratrabecular space has been previously described. 6 In bone marrow biopsies from HIV-positive patients, caution must be used not to overinterpret lymphoid aggregates. As others have noted, many of the previously described features that can be worrisome for malignancy, including increased size, poorly circumscribed margins, variable cytologic atypia, and a lymphohistiocytic pattern with no distinct granuloma formation, can be found in such patients. [22][23][24] Correlation with flow cytometry results can be helpful in such cases to detect a clonal population. Furthermore, correlation with the clinical presentation is extremely important, as these patients usually do not present with an isolated bone marrow process and if they present with a systemic lymphoma, it is usually a high-grade B-cell lymphoma, which should make the distinction in the bone marrow between a benign and malignant process less difficult. Table 5 shows the details of 10 HIV cases in our study with benign aggregates, where ill-defined margins, increased number of interstitial lymphocytes, and increased number of interspersed histiocytes were the notable morphologic features seen. The neoplastic lymphoid aggregates in Hodgkin lymphoma cases may also cause confusion, but usually contain a mixture of B and T lymphocytes or mainly T cells and are usually accompanied by a background of other inflammatory cells including eosinophils, plasma cells, and histiocytes in addition to Reed-Sternberg cells. 5,15 Sometimes granulomas are identified or evidence of increased background fibrosis is noted. In addition, benign lymphoid aggregates are occasionally seen in Hodgkin lymphoma.
In conclusion, assessment of 5 key morphologic and immunophenotypic features (large size, presence of infiltrative edges, paratrabecular localization, presence of cytologic atypia, and a B-cell distribution pattern) can help in the distinction between benign and malignant B-cell lymphoid aggregates in the bone marrow. We recommend careful morphologic examination and evaluation of CD3 and CD20 stains. The identification of more than 2 of the 5 characteristics was strongly suggestive of malignancy and was statistically associated with malignant diagnoses ( Table  3). The presence of 2 malignant characteristics was frequently found in both benign and malignant cases and should still raise suspicion of malignancy, although it statistically did not favor one diagnosis over the other. There was significant association with benign diagnosis when only 1 or none of the characteristics was noted.