Technical Note: A Counting Strategy for Estimating Plasma Cell Number in CD138-Stained Bone Marrow Core Biopsy Sections

Transcription

1 Available online at Technical Note: A Counting Strategy for Estimating Plasma Cell Number in CD138-Stained Bone Marrow Core Biopsy Sections Fred B. Smith and Ashraf Elnawawi Department of Pathology, Saint Vincent s Catholic Medical Centers of New York Manhattan, New York, and Department of Pathology, New York Medical College, Valhalla, New York Abstract. Plasma cells are readily identified microscopically after immunostaining for the CD138 antigen, and CD138-stained bone marrow core biopsy sections have proved superior to Giemsa-stained aspirate smears and hematoxylin-eosin (H&E) sections for determining plasma cell percentages in marrow. The CD138-stained plasma cell percentage is generally obtained by visual estimation, after viewing the entire section. We propose an alternative method, in which the microscopist initially views the immunostained section under low- and medium-power magnification, then selects a single high-power field in which the relative concentration of plasma cells appears most representative of the section as a whole, and performs a manual differential count of plasma cells in that field. On 44 CD138-stained core biopsy specimens from patients with plasma cell myeloma, selected area counting provided plasma cell percentage values that were more closely correlated with total section plasma cell area, determined by computerized image cytometry, than were visual estimates. This simple method requires only slightly more time than visual appraisal, does not require extensive training or experience, and appears to provide a more accurate estimate of the plasma cell population within the entire specimen. Keywords: plasma cell, plasma cell myeloma, immunohistochemistry, CD138 antigen Introduction Address correspondence to Fred B. Smith, M.D., Dept. of Pathology, St. Vincent s Catholic Medical Centers of New York, 170 West 12 Street, New York, NY 10011, USA; tel ; fax ; fsmith@svcmcny.org. The relative number of plasma cells in a bone marrow specimen is a clinical parameter important in the diagnosis and management of plasma cell dyscrasia [1]. The plasma cell number, expressed as a percentage of total marrow hematopoietic cells, has traditionally been obtained by manual differential counting of Giemsa-stained marrow aspirate smears (smear counts) or by estimation from H&E-stained core biopsy sections. CD138 antigen is expressed by neoplastic and nonneoplastic plasma cells, but not by other cells normally present in bone marrow, and the recent introduction into clinical practice of immunostaining for CD138 has greatly facilitated the recognition and quantitation of marrow plasma cells [2,3]. Estimated plasma cell percentages based on visual microscopic appraisal of entire CD138- stained marrow core biopsy sections (CD138 overview estimates) have been reported to show greater inter-observer reproducibility than those based on smear counts or H&E-section estimates, and to correlate well with plasma cell percentages obtained by image cytometry [4]. CD138 overview estimates presumably achieve these superior results because the immunostaining, unlike traditional methods, makes it possible to recognize plasma cells under low-magnification microscopy, when their numbers can be better judged in relation to other cells of the marrow. Although CD138 estimates have thus improved diagnostic accuracy, they must be assumed to be subject to particular sources of error generated by /08/ $ by the Association of Clinical Scientists, Inc.

2 Plasma cell estimation in core biopsy of bone marrow 139 visual overviews. Because plasma cells are larger than the majority of hematopoietic cells, one would expect the observer to be influenced by the relatively larger aggregate stained area of the plasma cells and therefore to tend systematically to overestimate their number. In addition, there is a risk that the accuracy of the estimate could be influenced by the distribution pattern of the plasma cells within the marrow, since this is known to vary widely among patients [5]. Perceptual psychologists have noted that when observers visually estimate the numbers of objects too numerous to count (numerosity estimates), they tend to underestimate numbers if the objects are clustered and to overestimate numbers if the objects are dispersed uniformly against the background [6,7]. One would expect the effect of this error to be magnified in plasma cell estimates by overview, since the percentage is derived as a ratio of two numerosity estimates, that of the total marrow cell population and of the stained plasma cells. In carrying out this study, we asked whether we could improve on the visual overview approach to CD138 plasma cell estimates by including a high-magnification manual counting step in the assessment, using the initial low-magnification overview only to select the counting area, and not attempting to estimate cell numbers on the overview. Assessment of CD138-stained core biopsy sections by selected area counting would thus closely parallel the steps used in performing smear counts. We retrospectively assessed CD138- stained bone marrow core biopsies from our departmental archives in this manner, and compared the results to those obtained previously by overview estimates and to those obtained by computerized image cytometry. Materials and Methods Case material. Archived histologic sections from bone marrow core biopsies on patients with plasma cell myeloma, which were performed in 2006 and stained previously for demonstration of CD138 antigen, were retrieved for review. Forty-four sections were judged technically adequate for purposes of the study (ie, marrow area >1 mm 2, cellularity >10%). The patient group from whom the biopsies had been obtained comprised 25 males and 19 females, and their mean age was 60.9 yr (SD ±10.7 yr). Serum total protein values ranged from 4.0 to 8.7 g/dl, and 28 patients had a detectable serum M-protein (11 IgG-κ, 5 IgA-κ, 4 IgG-λ, 2 IgA-λ, 6 other types), at concentrations of 0.08 to 2.16 g/dl. Twentytwo of the patients had detectable free light chains in the urine (23.9 to mg/l). Plasma cell estimates by visual overview and by smear counts were obtained from the surgical pathology reports on the biopsies. Overview estimates were listed on 42 of the pathology reports and smear counts on 32. The procedure for de-identification of the sections and reports, as well as the protocol for the entire study, was approved by the St. Vincent s Catholic Medical Centers of New York Integrated Scientific and Ethical Review Board. Plasma cell estimation by selected area counting. The sections were initially scanned at low (100x) and intermediate (200x) magnification to assess the distribution and concentration of the darkly stained plasma cells. An area within the section was then selected that was judged to have approximately the same concentration of plasma cells as the section as a whole. At high magnification (400x), a region containing approximately 200 cells was selected for counting, and delimited by use of the field diaphragm on the microscope light source. Plasma cells and other nucleated cells in the delimited region were counted with a manual cell counter, and the result was computed as plasma cell percentage. The method of field selection is illustrated in Fig. 1. Image cytometry. The sections were analyzed with a CAS 200 Image Analysis System, using Quantitative Nuclear Antigen software and 400x magnification. Digital images of all portions of the biopsy sections containing hematopoietic tissue were analyzed, with the nuclear threshold setting lowered so that both cytoplasm and nuclei of hematopoietic cells were included in the nuclear image. The antibody threshold was set so that only stained plasma cells were included in the antibody image, and the computed percent area score thus represented the aggregate area of plasma cells, as a percent of the total hematopoietic cell area, within the section. This method was previously used for quantitation of stained cells within a mixed tissue cell population [8]. Data analysis. Results from image cytometry and selected area counting were recorded in an MS Excel file, along with the corresponding overview estimates and smear counts from the pathology reports, and pairwise comparisons of methods were made with StatistXL software, using simple correlation and computations of Pearson s r and Student s t. Results Descriptive statistics for the plasma cell percentages obtained in the study cases by each of the methods (selected area counting, image cytometry, overview estimate, and smear count) are shown in Table 1. The mean of the aspirate smear counts was appreciably lower than the means obtained by the other methods; this may reflect the plasma cell stripping artifact in aspirate smears [3].

3 140 Table 1. Plasma cell percentage of the study cases, assessed by four methods. CD138 section overview Smear count Selected area count Image cytometry % area number mean ± SD 19.2 ± ± ± ± 10.7 range Fig. 1. Area selection for counting of CD138-positive plasma cells. Left, encircled area selected at intermediate-power magnification (200x) to contain concentration of plasma cells representative of section as a whole. Right, selected area at highpower magnification (400x), delimited using microscope field diaphragm. This area was found to contain 196 nucleated cells, 25 of which (13%) were plasma cells. Fig. 2. Correlation of plasma cell percentages by CD138 section overview, Giemsa-stained smear count, and CD138 section selected area count with plasma cell percent area (PA) score obtained by image cytometry.

4 Plasma cell estimation in core biopsy of bone marrow 141 Correlations between each of the manual method results with those obtained by image cytometry are shown in Fig. 2. Selected area counts were more highly correlated with the image cytometry measurements than were the smear counts or overview estimates. Although smear count percentages were lower than those obtained by overview estimates, they showed better correlation with image cytometry. Discussion In this study, selected area counts better approximated plasma cell numbers on CD138- stained bone marrow core biopsy sections than did visual overview estimates, with image cytometry measurements of plasma cell area as a gold standard for comparison. We believe this reflects, at least in part, a decreased impact of numerosity-related error in arriving at the final estimate of plasma cell percentage by the selected area count method. Although the area selection step is subjective and susceptible to observer bias, the counting step in the procedure is not. In overview estimation, however, the entire procedure is based on a subjective judgment. Smear counts in our study also outperformed overview estimates, based on image cytometry correlation, although they systematically undercounted plasma cells. This probably reflects both the partial objectivity of the method (only selection of the area of the smear for counting is subjective) and the under-representation of plasma cells in marrow aspirate samples, as previously noted [3]. It seems likely that the accuracy of CD138 overview estimates could be substantially increased by training the individuals who perform the estimates. Observers could initially examine a training set of marrow sections calibrated by image cytometry and then be provided with feedback on their estimates. In performing a similar visual estimation task, assessment of land areas from aerial overviews, observers dramatically improved their estimates when trained by feedback from image analysis [9]. In a study by Al-Quran et al [4], the reported CD138 overview estimates of plasma cells were performed by 2 individuals identified as expert hematopathologists, who may also have been experienced in image cytometry. We suspect that training effects may explain the much better correlation between overview estimates and image cytometry they reported, compared to our results. The overview estimates in our study were obtained from pathology reports and were provided by 4 general surgical pathologists, none of whom had previously performed image cytometry on marrow specimens. We believe that their estimates represent what would be expected from any conscientious microscopist not specifically trained for the task or aware of numerosity-related illusions. Judging from our experience, selected area counting appears to be a better means of assessing plasma cell numbers in CD138-stained marrow core biopsy sections of myeloma patients than overview estimation in the usual laboratory setting. We caution that our findings have not been tested in non-myeloma patients, and therefore may not pertain to normal patients or those with reactive plasmacytosis. We note, however, that the group of myeloma patients we studied included a subset in disease remission, with plasma cell numbers in the normal and reactive range. Until this issue has been explored more fully, we suggest that laboratories performing estimates of plasma cell numbers in bone marrow biopsies stained for CD138 antigen should not expect overview estimates to closely approximate the actual cell numbers, unless the individuals performing the estimates have been specially trained for this task. Performance of a selected area count on a CD138 section requires no special training and takes less time than a traditional smear count, and the results are not subject to the undercount bias inherent in assessing aspirate smears. References 1. Ravikumar SV, Fonseca R, Dispenzieri A, Lacy MQ, Lust JA, Witzig TE, Therneau TM, Kyle RA, Greipp PR, Gertz MA. Methods for estimation of bone marrow plasma cell involvement in myeloma: predictive value for response and survival in patients undergoing autologous stem cell transplantation. Am J Hematol 2001;68: Chilosi M, Adami F, Lestani M, Montagna L, Cimarosto L, Semanzato G, Pizzolo G, Menestrina F. CD138/ syndecan-1: a useful immunohistochemical marker of

5 142 normal and neoplastic plasma cells on routine trephine bone marrow biopsies. Mod Pathol 1999;12: Dunphy CH, Nies MK, Gabriel DA. Correlation of plasma cell percentages by CD138 immunohistochemistry, cyclin D1 status, and CD56 expression with clinical parameters and overall survival in plasma cell myeloma. Appl Immunohistochem Mol Morphol 2007; 15: Al-Quran SZ, Yang L, Magill JM, Braylan RC, Douglas- Nikitin VK. Assessment of bone marrow plasma cell infiltrates in multiple myeloma: the added value of CD138 immunohistochemistry. Human Pathol 2007; 38: Bartl R, Frisch B, Fateh-Moghadam A, Kettner G, Jaeger K, Sommerfeld W. Histologic classification and staging of multiple myeloma. A retrospective and prospective study of 674 cases. Am J Clin Pathol 1987;87: Ginsburg N. Numerosity estimation as a function of stimulus organization. Perception 1991;20: Cousins JB, Ginsburg N. Subject correlation and the regular-random numerosity illusion. J Gen Psychol 1983;108: Smith FB, Arias JH, Elmquist TH, Mazzara JT. Microvascular cytomegalovirus endothelialitis of the lung. A possible cause of secondary pulmonary hypertension in a patient with AIDS. Chest 1998;114: Gallegos A. Design and evaluation of a computer aided calibration program for visual estimation of vegetation cover. Sveriges Landtbruksuniversitet Arbetsrapport , accessed 10/5/2007 at se/resana/nils/publikationer/arb_rapp_142.pdf.