Marrow Assessment for Hemophagocytic Lymphohistiocytosis Demonstrates Poor Correlation With Disease Probability

Abstract

Objectives:

To evaluate the amount of hemophagocytosis in 64 marrow core biopsy specimens and aspirates from 58 patients with clinical suspicion for secondary hemophagocytic lymphohistiocytosis (HLH) or reported findings of hemophagocytosis.

Methods:

A review of medical records assigned patients to a low-risk (45 patients) or high-risk (13 patients) HLH group, and association with histologic findings was examined using the Fisher exact test.

Results:

The amount of hemophagocytosis in aspirate or the core biopsy specimen did not correlate with disease probability (P = .17 and P = .63, respectively). Of the clinical/laboratory criteria assessed, the most significant correlations with HLH were highly elevated ferritin (P = .01), cytopenias (P = .02), and fever (P = .009).

Conclusions:

Our findings indicated that marrow histologic findings alone do not reliably predict the probability of HLH, and an isolated finding of hemophagocytosis, even when present in a high amount, lacks specificity for HLH.

Hemophagocytic lymphohistiocytosis (HLH) is an often fatal syndrome of exaggerated but ineffective inflammatory responses, characterized by excessive macrophage and T-cell activation as well as impairment of the ability of natural killer (NK) and cytotoxic T cells to kill target cells.^1–4 HLH is a group of disorders that include familial and acquired forms of the syndrome and macrophage activation syndrome that is associated with certain autoimmune diseases.^1–3,5 The acquired form of HLH is associated with infections, especially with Epstein-Barr virus, and malignancies, particularly peripheral T/NK-cell or anaplastic large cell lymphomas, and certain medications used for conditions such as systemic lupus erythematosus.^1–3,5,6

According to HLH-2004, the Histiocyte Society’s updated diagnostic and therapeutic guidelines for HLH, a diagnosis of HLH can be established either by a molecular diagnosis with specific gene mutations associated with HLH or by meeting five of eight clinical and laboratory diagnostic criteria for nonfamilial HLH. These criteria include fever; splenomegaly; peripheral blood cytopenias affecting at least two of three cell lineages; hypertriglyceridemia or hypofibrinogenemia; microscopic evidence of hemophagocytosis in bone marrow, spleen, or lymph nodes; low or absent NK-cell activity; elevated ferritin; and elevated soluble CD25 (sCD25; ie, interleukin 2 [IL-2] receptor).⁷ For several reasons, in practice, bone marrow biopsies are often performed to check for evidence of hemophagocytosis when there is suspicion for HLH. Bone marrow biopsies can help rule out other diseases with similar clinical presentations as HLH. Several other diagnostic criteria for HLH, such as fever, cytopenias, and splenomegaly, are not very specific findings. Conventional wisdom suggests that finding evidence of hemophagocytosis can increase clinicians’ confidence in making a diagnosis of HLH. Furthermore, genetic mutation analyses, NK-cell activity, and sCD25 levels are usually send-out tests done at specialized laboratories, which may not be as helpful in acute settings when prompt treatment decisions are crucial. In HLH, as a result of exaggerated immune activation, macrophages nonselectively phagocytize hematopoietic elements, presumably leading to the microscopic finding of hemophagocytosis. However, histologic evidence of hemophagocytosis is not specific to HLH and can be seen in other conditions as well, such as after blood transfusion, chemotherapy administration, sepsis, and major operations,^8–11 but the expected amount of hemophagocytic cells (HPCs) seen in these conditions has not been well defined. At the same time, although it has been suggested that a positive finding in marrow for HPCs requires careful examination of at least three smears, each revealing at least two HPCs,¹⁰ there is so far no accepted interpretative threshold for positive findings or standardized reporting guidelines when such findings are present.

Very few studies to date have systematically looked at the quantitation of hemophagocytosis in the marrow and its correlation with eventual diagnosis of HLH,¹² especially in the adult population. Furthermore, the sensitivity of hemophagocytosis for HLH has been questionable. A pediatric study found evidence of bone marrow hemophagocytosis in only 58% of patients with eventual diagnoses of HLH.¹³ There was no control group in this study to assess for the specificity of hemophagocytosis. A small case-control study found bone marrow quantitation of hemophagocytosis to be higher in patients with HLH, and hemophagocytosis had a sensitivity of 83% and a specificity of only 60% in diagnosing HLH.⁸ However, only six marrow aspirates from three patients with HLH were included in that study, and the control participants were randomly selected and did not necessarily have clinical suspicion for HLH. In both of the above studies, only bone marrow aspirates, but not biopsy specimens, were evaluated.

In this retrospective study, we performed blinded evaluation of 64 bone marrow aspirates and core biopsy specimens from adult patients with signs and symptoms suspicious for acquired HLH, as well as those with reported findings of hemophagocytosis from all other causes. We quantitated the amount of hemophagocytic histiocytes in the marrow aspirates and core biopsy specimens, with the latter aided by CD68 immunostaining for histiocytes. Separately, patients’ medical records were reviewed and the cases assigned to either low- or high-risk HLH after case discussion. We then examined the correlation between the amount of HPCs and the probability of eventual HLH diagnosis in these patients.

Materials and Methods

Participants

Study approval was obtained from the Yale University Human Investigation Committee. A natural language search was conducted in the surgical pathology database of the Yale–New Haven Hospital (YNHH), New Haven, CT. The search included all bone marrow aspirate and core biopsy specimens with accession dates between January 1980 and May 2012, including specimens obtained at YNHH and outreach specimens sent from surrounding health care facilities. The natural language search encompassed the final diagnosis fields as well as the clinical history fields, allowing extraction of cases with clinical suspicion for HLH and all cases with reported marrow findings of hemophagocytosis due to HLH and all other etiologies. The search identified both cases with evidence of hemophagocytosis and cases in which no evidence of hemophagocytosis was seen and documented in the original pathology reports. All adult patients 18 years and older were included in the study.

Bone Marrow Aspirate and Core Biopsy Specimen Review

All cases identified by the search above with aspirates and/or core biopsy specimens available were evaluated blindly, without knowledge of clinical information or pathology diagnoses, regardless of whether hemophagocytosis was identified in the original pathology reports. The bone marrow aspirates available for review were all stained with Wright-Giemsa, and two aspirate slides from each case were evaluated when available. The aspirates were scored independently by a hematopathologist (M.L.X.) and a pathologist-in-training (C.H.). Aspirate cellularity was graded as decreased, adequate, or increased. The number of HPCs was counted over the entire slide and averaged over two slides. Only histiocytes showing phagocytosis of one or more intact erythrocytes and precursors, neutrophils and granulocyte precursors, plasma cells, and lymphocytes were counted as HPCs.

The bone marrow core biopsy specimens available for review were all received in B Plus Fix Ultra solution (BBC Biochemical, Mount Vernon, WA) or formalin, which were briefly decalcified with Decal Stat solution (Decal Chemical Corporation, Tallman, NY) for approximately 30 to 60 minutes before embedding in paraffin blocks. Sections 3 to 4 mm thick were prepared and stained with H&E and the CD68 immunohistochemical stain (clone PG-M1; DAKO, Carpinteria, CA). The core biopsy specimens were also scored independently for cellularity, and marrow cellularity was graded on the H&E slides as either less than 50% or 50% or more. CD68 immunohistochemical stains were used to highlight the histiocytes, and the percentage of HPCs out of total cellularity was estimated with the aid of CD68 staining.

Medical Record Review

Concurrently but independently, the patients’ medical records were retrospectively reviewed by two clinical hematologists with experience in the diagnoses and treatments of HLH (L.T. and N.P.). Patients’ clinical presentations and symptoms, laboratory results, impressions and assessments of the treating clinicians, treatment decisions, and disease courses were all taken into consideration, and data for each diagnostic criterion in the HLH-2004 guidelines were recorded. On the basis of the overall clinical picture and not merely the number of HLH-2004 criteria met, each patient was assigned to either the low-risk HLH group (HLH-L) or the high-risk HLH group (HLH-H) after case discussion.

Statistical Analysis

Analysis was performed for HLH-L and HLH-H patients. Additional analysis was also done to compare patients meeting five HLH-2004 criteria with those who did not. Patient characteristics were presented using descriptive statistics: median and range were calculated for continuous variables, and frequency and percentage were presented for categorical variables. The Fisher exact test was used to investigate the association between risk of HLH and different categorical clinical criteria. The Wilcoxon rank sum test was performed to compare ferritin measurements. The κ coefficient was used to test the agreement of cellularity/cell counts between two pathologists. For the four patients with multiple marrow aspirates and biopsies performed, clinical and laboratory data temporally associated with each patient’s earliest specimen were used in the analyses and shown in the tables. Separate analyses were done with data temporally associated with each patient’s latest specimen; similar results were seen with no significant change in the P values and therefore are not presented. SAS 9.2 (SAS Institute, Cary, NC) was used for analysis, and the statistical significance level was two-sided .05.

Results

Patient Characteristics

The natural language search identified 98 cases from 88 unique patients in which some evidence of HPCs was mentioned in the original pathology reports. At the same time, there were 11 cases from 11 unique patients in which no evidence of HPCs was seen and documented in the original pathology reports. This latter group included patients whose signs and symptoms, like the former cohort, raised suspicion for HLH, but hemophagocytosis was not seen by the original pathologists. Of the 98 cases with HPCs identified, marrow aspirates were available for examination in 66 cases, including four aspirate-only cases without concurrent marrow core biopsy specimens. Core biopsy specimens were available for review in 60 cases. Fifty-six patients’ medical records were available for evaluation of HLH risk. Of the 11 cases with no HPCs identified, nine marrow aspirates and core biopsy specimens were available for examination, and eight patients’ medical records were available for review. A total of 64 cases from 58 unique patients had bone marrow aspirates and/or core biopsy specimens as well as medical records available for review; these were included in this study’s analyses.

Table 1 summarizes the patients’ characteristics, including demographics and underlying medical conditions. Of the 58 unique patients, 24 were female (41.4%) and 34 were male (58.6%). The median age at biopsy was 54.0 years, with an age range of 18 to 87 years. Two patients had two bone marrow biopsy specimens, and another two patients had three specimens obtained at different times. The time interval between consecutive biopsies ranged from three weeks to five months. The remaining 54 patients each had one bone marrow biopsy specimen only.

Table 1

Patient Demographics and Underlying Medical Conditions (n = 58)

Clinical Features

On the basis of the review of the patients’ medical records, 45 (77.6%) patients were classified in the HLH-L group, and 13 (22.4%) were classified in the HLH-H group. Figure 1 summarizes the assignment of patients to different groups for analyses. Table 2 summarizes the time from marrow biopsy/aspirate to the initiation of treatment, along with a list of treatments received for the 13 HLH-H patients. It is worth noting that one patient with diffuse large B-cell lymphoma met five HLH-2004 criteria (fever, splenomegaly, cytopenias affecting two lineages, hypertriglyceridemia/hypofibrinogenemia, and elevated ferritin level), but the overall clinical picture was not consistent with HLH; hence, the patient was classified into the HLH-L group instead. A low amount of hemophagocytosis was seen in this patient’s marrow specimens (1–5 HPCs/slide on marrow aspirate and 10% HPCs of total marrow cellularity on marrow biopsy specimen). Two of 13 patients in the HLH-H group had data for less than five HLH-2004 criteria available, excluding hemophagocytosis, either due to the relevant clinical findings not documented or laboratory tests not ordered. Five patients in the HLH-H group met three or four HLH-2004 criteria, and seven met at least five HLH-2004 criteria excluding hemophagocytosis. Cases meeting less than five HLH-2004 criteria represented patients in whom suspicion for HLH was high enough and treatments were started without the fulfillment of sufficient criteria. Of all the criteria, NK-cell activity was the least frequently documented and was checked in only one patient included in this study.

Figure 1

Algorithm for classification of patients. ^aHLH-H, high-risk hemophagocytic lymphohistiocytosis; HLH-L, low-risk hemophagocytic lymphohistiocytosis. ^bExcluding histologic evidence of hemophagocytosis.

Table 2

Time From Marrow Biopsy/Aspirate to Initiation of Treatment and List of Treatments Received for the 13 HLH-H Patients

Table 3 summarizes the association between six HLH-2004 diagnostic criteria and the probability of HLH diagnosis. Female patients were more likely to be in the HLH-H group than male patients (76.9% vs 23.1%; P = .004). The HLH-H group tended to be younger than the HLH-L group, but the difference was not statistically significant (median age, 48 vs 55 years; P = .06). The HLH-H group had a higher chance of having fever and cytopenias (P = .009 and .02, respectively). Although the HLH-H group had a higher chance of having hypertriglyceridemia and/or hypofibrinogenemia, these two laboratory tests were not checked in more than half of the cases in the HLH-L group. Therefore, their predictive values in determining HLH risk were uncertain. sCD25 (IL-2 receptor) was infrequently checked among patients in this study, and the predictive value was again difficult to be evaluated.

Table 3

Association Between Six of the HLH-2004 Diagnostic Criteria and the Probability of HLH Diagnosis^a

Ferritin level was elevated (≥500 μg/L) in all 13 patients in the HLH-H group. Within the HLH-L group, ferritin level was available for less than half of the patients; for the 17 patients with ferritin level available, 14 had a ferritin level of 500 μg/L or higher as well. However, the HLH-H group had a significantly higher median ferritin level compared with the HLH-L group (16,700 vs 1,870 μg/L; P = .01, Wilcoxon rank sum test). The results are summarized in Figure 2. Most patients in the HLH-H group had a highly elevated ferritin level of 10,000 μg/L or higher, while most in the HLH-L group had a ferritin level between 500 and 10,000 μg/L. The results are summarized in Figure 3.

Figure 2

Ferritin levels (μg/L) between the low-risk hemophagocytic lymphohistiocytosis (HLH) group (n = 13) and the high-risk HLH group (n = 17). Dashed lines represent the median in each group. Wilcoxon rank sum test showed a P value of .01.

Figure 3

Comparison of ferritin levels between the high-and low-risk hemophagocytic lymphohistiocytosis (HLH) groups. P = .02.

Additional analyses were done comparing patients in the HLH-H group who met at least five HLH-2004 criteria (excluding histologic evidence of hemophagocytosis) (n = 7) with all other patients (n = 51). The two groups showed a similar median age (53 vs 55 years; P = .44). However, similar to the analyses mentioned above, the HLH-H group was more likely to be female (P = .001) and to have fever (P = .01), cytopenias (P = .02), and a higher elevation of ferritin (P = .009). Unlike the previous analyses, the HLH-H group was now also more likely to have splenomegaly (P = .02).

Bone Marrow Aspirates and Core Biopsy Specimen Reviews

Bone marrow aspirates from 58 patients were reviewed blindly for cellularity and number of HPCs. Image 1 shows examples of HPCs in selected cases. Table 4 summarizes the association between marrow aspirate cellularity and number of HPCs per slide with the probability of HLH diagnoses. Data from one hematopathologist (M.L.X.) are shown. There was no significant difference in the number of HPCs between the HLH-H and HLH-L groups (P = .17), and there was no optimal cutoff point that would reliably separate the two groups. Most patients in the HLH-H group (61.5%) had a low amount of hemophagocytosis (1–5 cells/slide), and 23.1% of patients in the same group had no evidence of hemophagocytosis. On the other hand, 22.2% of patients in the HLH-L group had a high amount of hemophagocytosis (>10 cells/slide). It is worth noting that the aspirate cellularity in the HLH-H group was more likely to be decreased compared with the HLH-L group (P = .05), which might have resulted in a lower average amount of HPCs present. This analysis was repeated after excluding all cases with no evidence of hemophagocytosis for a comparison of only cases with any degree of hemophagocytosis. The results are summarized in Table 5. There remained no significant difference in the amount of hemophagocytosis between the HLH-H and HLH-L groups (P = .26). Between the two pathologists, κ agreement was 0.50 for aspirate cellularity and 0.51 for the number of HPCs.

Image 1

A–D, Hemophagocytic cells in bone marrow aspirates of selected cases (Wright-Giemsa; ×100). Only histiocytes phagocytizing one or more intact erythrocytes and precursors, neutrophils and granulocyte precursors, plasma cells, and lymphocytes were counted.

Table 4

Association Between Marrow Aspirate Cellularity and Number of Hemophagocytic Cells With the Probability of HLH Diagnosis^a

Table 5

Association Between Marrow Aspirate Cellularity and Number of Hemophagocytic Cells With the Probability of HLH Diagnosis, Excluding Cases With No Evidence of Hemophagocytosis^a

Bone marrow biopsy specimens from 53 patients were also reviewed blindly for cellularity and percentage of HPCs out of total cellularity. Image 2 shows examples of hemophagocytic histiocytes highlighted by CD68 immunohistochemical stains in selected cases. Table 6 summarizes the association between marrow biopsy cellularity and amount of HPCs with the probability of HLH diagnoses. Data from one hematopathologist (M.LX.) are shown. There was no significant difference in marrow biopsy cellularity (P = .14) and percentage of marrow cells with hemophagocytosis (P = .63) between the HLH-H and HLH-L groups, and there was again no optimal cutoff point that would reliably separate the two groups. In fact, most patients in the HLH-H group (58.3%) had a low amount of hemophagocytosis (1%–10%). This analysis was also repeated after excluding all cases with no evidence of hemophagocytosis for a comparison of only cases with any degree of hemophagocytosis. Those results are summarized in Table 7. There remained no significant difference in the percentage of marrow cells with hemophagocytosis between the HLH-H and HLH-L groups (P = .52). Between the two pathologists, κ agreement was 0.58 for biopsy marrow cellularity and 0.26 for the percentage of marrow cells with hemophagocytosis.

Image 2

A and B, CD68 immunohistochemical stain highlights hemophagocytic histiocytes in bone marrow biopsy specimens of selected cases (×100). Due to background and quality of staining of CD68, hemophagocytic cells are not always easily identified (arrows).

Table 6

Association Between Marrow Biopsy Cellularity and Amount of Hemophagocytic Cells With the Probability of HLH^a

Table 7

Association Between Marrow Biopsy Cellularity and Amount of Hemophagocytic Cells With the Probability of HLH, Excluding Cases With No Evidence of Hemophagocytosis^a

Additional analyses were done comparing the patients in the HLH-H group who met at least five HLH-2004 criteria, excluding histologic evidence of hemophagocytosis (n = 7), with all other patients (n = 51). The HLH-H and HLH-L groups again showed no significant differences in the number of HPCs in the aspirates or the percentage of marrow cells with hemophagocytosis in the marrow biopsy specimens. There was again no optimal cutoff point that would reliably separate the two groups.

Discussion

HLH is a syndromic disorder that can lead to life-threatening symptoms in a short interval. While the HLH-2004 guidelines established clinical and laboratory criteria for the diagnosis of HLH, several are very difficult to obtain in a timely manner. Morphologic evaluation for hemophagocytosis, which can be obtained quickly through bone marrow biopsy, is included in this set of criteria, but there is no standard for its assessment by pathologists, and it is not a specific finding for HLH. This retrospective study showed that the amount of hemophagocytic histiocytes in both marrow aspirates and core biopsy specimens does not correlate well with clinical disease probability.

Despite its place in the currently accepted criteria for the diagnosis of HLH, several prior studies have alluded to the difficulty of marrow assessment of hemophagocytosis. A study of 21 pediatric marrow specimens showed that the numbers of HPCs were often initially low or even absent in patients who developed HLH and that sensitivity of these marrow findings was 58%.¹³ Studies conducted prior to the establishment of the 2004 criteria have attempted to look at the specificity of HPCs. One such large investigation showed that the presence of hemophagocytosis was highly associated with blood transfusions and bacterial sepsis and marginally associated with major surgery and viral infections.¹¹ In 230 unselected autopsy cases, 44.3% showed moderate to severe levels of what was then termed histiocytic hyperplasia with hemophagocytosis. Of course, these data, although valuable, cannot be used to correlate with the subsequently established disease criteria.

A recent study that attempted to determine the sensitivity and specificity of bone marrow hemophagocytosis in HLH compared six HLH cases with random controls and concluded that, although HLH cases have a higher mean percentage of HPCs than control cases, overlap of hemophagocytosis counts between the two groups precluded use of these criteria as a reliable indicator.⁸ The authors of this study performed the quantitation by counting nucleated cells and macrophages with hemophagocytosis using a Miller ocular disk. While this enumeration appeared precise for the fields that were evaluated, it may not be practically useful in the clinical setting. The control marrow were randomly selected and included patients with a variety of conditions, which likely differed from a population of patients who demonstrate clinical signs and symptoms suspicious for HLH and undergo bone marrow aspirate/biopsy.

We compared the quantity of HPCs between patients with and without eventual diagnoses of HLH, many of whom initially presented with signs and symptoms worrisome for HLH. Such a strategy may be useful in the clinical setting when patients are suspected of having HLH; a bone marrow examination is one of several tests procured for ruling out an HLH diagnosis. This study showed that the number of HPCs obtained either by aspirate count or by core biopsy with the aid of CD68 immunohistochemical stains was not significantly associated with the probability of disease, and a high amount of hemophagocytosis could be seen in patients with a low risk of HLH. A critical component of this analysis was the clinical judgment required to determine the risk of disease. Two hematologists with experience in diagnosing and treating HLH reviewed the medical records retrospectively and assigned HLH risk to each case based on all the available information, including the courses of disease and treatment outcomes in addition to available HLH-2004 criteria. This additional layer of interpretation added clarification in cases that met only some of the criteria when other criteria were not available. Furthermore, it is worth noting that, although the reviewers were not blinded to the original bone marrow biopsy reports, they were blinded to the quantitation of HPCs by the two pathologists.

Cases were found via a natural language search for HLH through the pathology database that encompassed the final diagnosis field as well as the clinical history field. This search strategy allowed inclusion of cases in which either the presence or absence of hemophagocytosis was reported by the original pathologists, as well as cases in which HLH was mentioned as a differential diagnosis by the clinician on the requisition forms. We acknowledge that there are limitations to this strategy. At the time of bone marrow biopsy, the primary clinicians might not have mentioned HLH as a differential diagnosis on the requisition form, or the possibility might have come up later as more clinical information was available. Our search would not easily pick up all cases in which HLH was suspected; hence, it was more difficult to accurately assess the sensitivity of hemophagocytosis for HLH. However, our study allowed a better assessment of the specificity of hemophagocytosis for HLH since the search included hemophagocytosis cases in patients with or without HLH. When cases showing no HPCs on the aspirate or biopsy specimens were excluded, the findings remained similar (see Tables 4–7).

The HLH-2004 guidelines requires meeting five of eight criteria for diagnosis, and most of the criteria lack specificity in isolation, particularly fever and splenomegaly, which can be seen in a variety of medical conditions. Another HLH-2004 criterion, ferritin elevation, is also seen frequently among hospitalized patients, especially those with chronic inflammatory conditions. However, a highly elevated ferritin level was thought to be fairly specific in the diagnosis of HLH, as concluded in a recent study.¹⁴ Using the HLH-2004 ferritin threshold of 500 μg/L or higher, 330 pediatric patients were identified, of whom only 10 were diagnosed with HLH. When a higher ferritin threshold of 10,000 μg/L or higher was used, elevated ferritin had 90% sensitivity and 96% specificity for HLH. In our study, the HLH-H group had a significantly higher median ferritin level compared with the HLH-L group. Most of the former group had a ferritin level of 10,000 μg/L or higher, while most in the latter group had a ferritin level between 500 and 10,000 μg/L. However, elevation of ferritin was used together with other clinical information in the assignment of patients to the HLH risk groups. Therefore, the association between the HLH-H group and highly elevated ferritin level must be interpreted with caution.

A major obstacle encountered in this study was the absence of test results for one or more of the HLH-2004 criteria among a large number of patients, particularly the specialized send-out tests for sCD25/sIL-2 receptor level and NK-cell activity, as well as ferritin, triglyceride, and fibrinogen levels. Some of the missing data also stemmed from the fact that three of these patients were evaluated prior to the 2004 revision in diagnostic criteria. Cases in the HLH-L group were more likely to have missing data, which could reflect the low clinical suspicion and diminished need for extensive workup. However, we could not exclude the possibility that some of the patients in the HLH-L group with missing data were misclassified and would have been considered at high risk for HLH if the disease had been investigated more thoroughly by treating clinicians at the time. Furthermore, although our search in the pathology database included cases since 1980, only three cases included in this study were between 1980 and 2003. The revision of HLH diagnostic criteria in 2004, increasing suspicion of HLH by clinicians, and increasing awareness of pathologists to specifically search for and report hemophagocytosis in marrows at our institution might have contributed to significantly more cases being identified within the past decade.

In this study, instead of counting the number of HPCs in the marrow aspirates out of a set number of nucleated marrow elements, entire slides were examined—a time-consuming process. Even in cases with florid hemophagocytosis, the HPCs were not necessarily evenly distributed. Examination of the entire aspirates reduced sampling error in cases with patchy areas of hemophagocytosis. Also, similar to any visual assessment of aspirates, the counting of HPCs was susceptible to inter- and intraobserver variabilities and showed only moderate agreement between two pathologists. Furthermore, in patients in whom no evidence of HPCs was seen by the original pathologists, a low amount of hemophagocytosis was seen in all by the two pathologists involved in this study (all had 1–5 HPCs/slide on marrow aspirates and 10%–30% HPCs of total marrow cellularity on marrow biopsy specimens). Using numerical categories (1–5 cells/slide, 6–10 cells/slide, etc) instead of reporting the exact number of HPCs was meant to minimize the effects of these variabilities. Estimation of the percentage of HPCs out of total marrow cellularity with the aid of CD68 immunostains on biopsy specimens also did not reveal good interrater reliability, showing poor agreement between two pathologists. One reason for this is that the staining quality of this particular antibody is difficult to interpret due to high background. However, given the significant amount of staining seen in cases with a low risk of HLH, it does not seem to be a reliable method for quantifying HPCs in the marrow. CD163 immunostains were available in selected cases in this study, which have been found by previous studies to be more specific than CD68 for monocyte/macrophage lineage cells, although they were also less sensitive in certain hematopoietic disorders.^15,16 However, at our institution, CD163 immunostains showed high-background nonspecific staining that made interpretation very difficult. In cases with both CD68 and CD163 stains available, CD163 did not offer better visualization of hemophagocytic histiocytes. Therefore, data for CD68 stains were used in the analyses instead.

Overall, our findings suggest that quantitation of HPCs in the aspirate or in core biopsy specimens with the aid of CD68 immunostains does not correlate well with disease probability. The value of a bone marrow biopsy in light of clinical signs or symptoms suspicious for HLH may still be high given the importance of excluding other disease processes that may or may not be related to HLH. However, the utility of bone marrow evidence of hemophagocytosis as a rule-in criterion for HLH deserves reconsideration. Furthermore, the nonspecificity of hemophagocytosis in the marrow, even when present in high amounts, should remind both pathologists and clinicians that an isolated finding of hemophagocytosis lacks specificity and does not necessarily suggest HLH when the clinical presentation and laboratory findings are not compatible with the diagnosis.

This investigation was supported by National Institutes of Health research grant CA-16359 from the National Cancer Institute.