The quantitation of plasma cells in bone marrow (BM) is crucial for diagnosing and classifying plasma cell neoplasms. Various methods, including Romanowsky-stained BM aspirates (BMA), immunohistochemistry, flow cytometry, and radiological imaging, have been explored. However, challenges such as patchy infiltration and sample haemodilution can impact the reliability of BM plasma cell percentage estimates. Bone marrow plasma cell percentage varies across methods, with immunohistochemically stained biopsies consistently yielding higher values than Romanowsky-stained BMA or flow cytometry alone. CD138 or MUM1 immunohistochemistry and artificial intelligence image analysis on whole-slide images are emerging as promising tools for accurate plasma cell identification and quantification. Radiological imaging, particularly with advanced technologies like dual-energy computed tomography and radiomics, shows potential for multiple myeloma diagnosis, although standardisation remains a challenge. Molecular techniques, such as allele-specific oligonucleotide quantitative polymerase chain reaction and next-generation sequencing, offer insights into clonality and measurable residual disease. While no consensus exists on a gold standard method for BM plasma cell quantitation, CD138-stained biopsies are favoured for accurate estimation and play a pivotal role in diagnosing and assessing multiple myeloma treatment responses. Combining multiple methods, such as BMA, BM biopsy, and flow cytometry, enhances accuracy of diagnosis and classification of plasma cell neoplasms. The quest for a gold standard requires ongoing research and collaboration to refine existing methods. Furthermore, the rise of digital pathology is anticipated to reshape laboratory medicine and the role of pathologists in the digital era.
This article adds a comprehensive review and comparison of different methods for plasma cell estimation in the bone marrow, highlighting their strengths and limitations. The goal is to contribute valuable insights that can guide the selection of optimal techniques for accurate plasma cell estimation.
Plasma cell neoplasms encompass a wide variety of disorders that result from the clonal expansion of terminally differentiated B cells (plasma cells), typically secreting a single homogeneous monoclonal immunoglobulin.
Quantitation of PC in BM samples has historically relied on manual methods such as microscopy-based cell counting on BM aspirates and immunohistochemically stained trephine biopsies. The surface marker CD138 is sensitive and specific for PC, thus CD138 immunohistochemistry is commonly used for estimating PC in BM biopsies. This practice is supported by the World Health Organization.
Currently, there is no acceptable reference method for estimating PC in BM, although many methods have been developed and evaluated for accuracy and reproducibility. This method review aims to comprehensively evaluate existing techniques for the quantitation of PC in BM samples. By examining the principles, procedures, advantages, and challenges associated with the various methods, we aim to provide a comprehensive understanding of the options available to clinicians, researchers, pathologists, and haematologists.
In this review the authors examine the factors that affect the reliability of plasma cell estimation and share thoughts on the promising role of new technologies. In addition, the challenges of standardisation, clinical relevance and future directions will stimulate discussion and encourage collaboration in this field. We hope to contribute to the refinement of PC estimation methods and their integration into routine clinical practice.
Based on the recommendations by Levac et al., Colquhoun et al. and the methods of Arksey and O’Malley, we conducted a literature search for a narrative (scoping) review article.
Summary of various methods used for the estimation of bone marrow plasma cell percentages, highlighting their advantages and disadvantages.
| Method | Description | Advantages | Disadvantages |
|---|---|---|---|
| Manual differential count (myelogram) | Microscopic examination of Romanowsky-stained BM slides to count PC as a proportion of all other nucleated cells in a 200–500 cell differential. | Readily available. |
Subjective interpretation. |
| Multiparameter flow cytometry | Uses fluorescently labelled antibodies to identify PC in a BMA sample. | Quantitative (controversial) and qualitative analysis. |
Expensive equipment. |
| Immunohistochemistry | Detection of specific PC antigens (CD138 and MUM1) in tissue sections using chromogenic immunohistochemistry. | Readily available. |
Requires tissue processing and specialised reagents. |
| Molecular techniques (ASO-qPCR and NGS) | Based on amplification of unique VDJ rearrangements of |
High sensitivity and specificity. |
Requires specialised equipment and expertise. |
| Digital imaging analysis | Computer-based and AI analysis of IHC-stained slides using static images or WSI. | Objective and reproducible results. |
Initial setup and validation may be time-consuming. |
| Radiology imaging techniques | PET/CT uses a surrogate measure of intracellular glucose metabolism to identify hypermetabolic PC. |
PET/CT more sensitive for focal lesions. |
18F-FDG PET/CT cannot reliably predict PC infiltration of < 10%. |
PC, plasma cell; BM, bone marrow; BMPC%, bone marrow plasma cell percentage; MGUS, monoclonal gammopathy of undetermined significance; SMM, smouldering multiple myeloma; MM, multiple myeloma; MFC, multiparameter flow cytometry; GBM, gradient-boosting machine; NGF, next-generation flow; MRD, measurable residual disease; BMA, bone marrow aspirates; ASO-qPCR, allele-specific oligonucleotide quantitative polymerase chain reaction; NGS, next-generation sequencing; IHC, immunohistochemistry; VDJ, variable, diversity, joining; IGHV, immunoglobulin heavy chain variable; WSI, whole slide imaging; MRI, magnetic resonance imaging; 18F-FDG, 18F fluorodeoxyglucose; PET/CT, positron emission tomography/computed tomography; AI, artificial intelligence.
The first millilitre of aspirated BM represents pure BM elements. The subsequent sampling becomes significantly blood diluted (haemodilution).
Bone marrow aspirate smears are usually stained with Romanowsky-type stains, which are used either in combination or individually. The staining of BM films may be adversely affected by any of the pre-analytical processes.
The common problem with myeloma trephine samples is inadequate or fragmented cores. The biopsy may be inadequate if the same bone puncture site is used for aspiration and biopsy.
Bone marrow trephine biopsy specimens are usually fixed in formalin, embedded in paraffin, and stained with haematoxylin and eosin.
Sections of clotted BMA may be used in cases where it is difficult to obtain a trephine biopsy. The BMA is allowed to clot in a syringe. The clot is then carefully removed, placed in a fixative such as formalin, and processed in the same way as a trephine biopsy, except that no decalcification is required.
Cellular trails of Romanowsky-stained BM aspirates are used for differential counts under the light microscope, but can be misleading if PC do not ‘fall out’ from the BM particles or are not aspirated due to the patchy nature of MM. Plasma cell are counted as a proportion of all other nucleated haematopoietic cells in a 200–500 cell differential to estimate the percentage of PC in the BM.
In a study conducted in the United States in 2007, Al-Quran et al. evaluated CD138-stained BM specimens and showed poor intraobserver concordance with an intraclass correlation coefficient of 0.55 for one observer and an intraclass correlation coefficient of 0.47 for another between PC quantitation on BM aspirate smears and CD138 immunohistochemistry.
Despite this percentage discrepancy, one study found a significant linear correlation between BMPC% in aspirate enumeration and CD138 immunohistochemistry (
The literature supports the use of MFC in the diagnosis and follow-up of MM, particularly regarding establishing the clonal nature of the PC.
The use of MFC to quantify PC in BM is more controversial. It has been shown that the percentage yield of PC is consistently lower with flow cytometric methods than with morphological assessment of smears from BMA.
In contrast to this, a 2017 study by Matsue et al. in Japan showed that the BMPC% on aspirate smear counts correlated with flow cytometry (
With the concern of the significant discrepancy between morphological assessment and flow cytometry PC percentage estimation, Frebet et al., in a study in France in 2011, developed an MFC protocol that uses specific antibody combinations to include PC (CD138+/CD38+) and haematopoietic precursors (CD45+/CD117+/CD34+), and exclude erythroblasts and debris (CD36+/CD45–), to calculate a plasma cell:precursor ratio.
CD138 (Syndecan-1) is a cell surface, sulphate-rich proteoglycan adhesion molecule which is expressed in the late stages of B-cell differentiation and on PC. CD138 is a highly sensitive and specific marker for BMPCs, making it useful in identification and enumeration of BMPC.
In their 2008 study, Smith and Elnawawi proposed an alternative to overview estimation by performing a manual count on a single selected field of view at high magnification (40x objective). The manual count (mean 23.1; s.d. ± 17.7) correlated better with computer-assisted image cytometry (mean 21.4; ± 10.7 s.d.) than the overview estimates (mean 19.2; s.d. ± 20.4), on 44 CD138-stained core biopsy specimens.
In their 2017 study in Japan, Matsue et al. counted more cells (500–2500) in 2–5 representative microscopic fields at low magnification in CD138-stained BM clot and biopsy sections. They showed significant correlation between the marrow clot and biopsy sections (
In their 2023 publication in South Africa, Gantana et al. evaluated a counting method in which reviewers selected three representative biopsy areas with low, intermediate, and high PC densities, followed by a manual count in each of these areas and averaging the BMPC% for each sample. This study showed superior interobserver concordance with the manual counting method at low PC tumour burden compared to an overview estimation method (intraclass correlation coefficient = 0.105).
Many studies show significantly higher numbers of BMPC% on CD138-stained immunohistochemistry biopsies compared to BMA.
Immunohistochemistry determination of PC percentage is valuable for diagnosis and monitoring treatment response, notably the complete and stringent complete response criteria as defined by the International Myeloma Working Group (IMWG).
Molecular techniques are based on the amplification of unique variable, diversity, joining rearrangements of the immunoglobulin heavy chain variable (
Molecular techniques such as allele-specific oligonucleotide quantitative PCR (ASO-qPCR) and next-generation sequencing (NGS) are used during follow-up for the quantification of clonal plasma cells to evaluate treatment response and detect measurable residual disease (MRD) rather than quantification at diagnosis.
Radiological imaging plays a significant role in the diagnosis and management of MM as approximately 90% of MM patients develop bone disease.
Radiographic examination (skeletal survey) to identify lytic bone lesions has been replaced by more sensitive alternatives such as whole body, low-dose computed tomography (CT), 18F fluorodeoxyglucose (18F-FDG) positron emission tomography/CT (PET/CT) scanning, and magnetic resonance imaging (MRI). In under-resourced countries, skeletal surveys are still widely used due to the lack of availability of superior modalities especially for detection of osteolytic lesions in the spine and pelvis.
A South African study in patients at a tertiary hospital showed good concordance between representative BM biopsy samples and 18F-FDG PET/CT image analysis. However, this did not reliably predict PC infiltration of < 10% or diffuse non-hypermetabolic marrow involvement on imaging.
Digital pathology, with or without incorporation of artificial intelligence (AI), is growing in popularity while presenting both challenges and opportunities for the future. Careful consideration is required when introducing this new area of medicine to patient care. Digitisation of immunostained BM biopsy sections can be done by taking still images with a microscope camera or dynamic images with whole slide imaging (WSI) systems. Images can be analysed manually or using AI. Herein, we review several studies that have explored the use of these innovations in estimating BMPC%.
Computer-assisted image analysis (CIA) for estimating BMPCs in BM biopsies dates to the early 2000s. Computer-assisted image analysis was shown to be superior to simple visual estimation of CD138 immunohistochemistry, due to elimination of subjectivity. In a Swiss study by Went et al. in 2006, CIA was performed using image processing software (KS 300, Carl Zeiss AG, Feldbach, Switzerland) after digitising selected fields with a light microscope and colour video camera.
Technology has improved so that an entire biopsy slide can be digitised with WSI scanners, and more advanced software programmes than those used in CIA and AI are used to analyse the images.
The use of an open-source digital image analysis tool, QuPath, was validated by Baranova et al., in a study published in Canada in 2021, for the automation of PC quantification in BM biopsies from patients with MM.
Whole slide imaging may be used to overcome the considerable variability in cellularity and distribution of plasma cells in BM biopsies by providing images of the full length of the biopsy. Whole slide imaging scanners are not readily available in resource-limited settings, hence alternative means of digitising and analysing biopsies are still required. Using the AI of QuPath, Gantana et al., in their South African study published in 2023, counted CD138-positive PC in BM biopsies and showed that digital analysis is superior to both a manual counting method and a previously reported overview estimation method, regardless of the tumour burden of the sample.
The advantage of digital image analysis is therefore considerable. It may be incorporated into routine use, but slides need to be of equal quality. This requires standardisation of biopsy processing protocols.
The accurate determination of BMPCs plays a central role in the diagnosis, prognosis, the choice of therapy and the monitoring of PC neoplasms and is therefore essential. Artificial intelligence, in combination with methods such as flow cytometry and NGS, can provide information to predict risk status and potentially change the diagnostic criteria and risk stratification for MM.
In their 2022 publication in France, Clichet et al. used an AI-based gradient-boosting machine algorithm to develop a decision tree for determining the risk of MM in a cohort of PC neoplasms. They used MFC (10-colour single tube) parameters including PC %, pathological to normal PC ratio, total PC to haematogone ratio, total PC to progenitor CD117+ ratio, mean fluorescence intensity of CD38 and CD27 from pathological PC, and mean fluorescence intensity of CD27 from normal cells at diagnosis. The AI-based decision tree showed a sensitivity of 94.2%, a specificity of 93.7% and an independent validation success rate of 91.0% without misclassification between monoclonal gammopathy of undetermined significance and smouldering multiple myeloma.
Yenamandra et al., in their study published in the United States in 2021, used data at the time of diagnosis which included variables such as patient age, white blood cell count, percentage of PC in the BM on the myelogram, high-risk cytogenetic alterations, and NGS. These data were used to create neural networks with which they developed a predictor for the risk status of MM, which has the potential to predict high-risk patients.
With picture archiving and communications systems as the standard in many tertiary institutions, radiology has a significant potential for radiomics, the use of pattern recognition for the extraction of quantitative descriptors from imaging data.
There are several strategies for MRD detection in myeloma, including MFC, qPCR, and high-throughput sequencing. The two currently validated MRD methods recommended by the IMWG are next-generation flow cytometry and NGS. Next-generation flow refers to the 8-colour 2-tube flow cytometry method developed by the EuroFlow consortium, which uses specific combinations of antibodies to identify and cross-reference neoplastic PC. It has a sensitivity of 10-6. In contrast, conventional 4–8 colour flow cytometry methods can only detect neoplastic PC with a sensitivity of 10-4.
Multiparameter flow cytometry is a sensitive and rapid approach to assessing treatment efficacy with confirmed prognostic value that predicts progression-free and overall survival independent of categorical response assessment and patient biology.
The threshold used to distinguish between MRD positive and MRD negative may vary from sample to sample, and may change as both technologies and treatments improve. The International Clinical Cytometry Society proposed a consensus guideline for the use of flow cytometry for MRD. The International Clinical Cytometry Society recommends the use of at least five initial gating parameters (CD38, CD138, CD45, forward scatter and side scatter) within the same aliquot to accurately identify the total PC compartment and to specify the limit of detection, with a limit of detection of 0.001% (requiring a total number of 3 × 106 cells for analysis) and ideally a limit of quantification of 0.001% (requiring a total number of 5 × 106 cells for analysis).
It is noteworthy that a BM aspirate sample containing 10 x 106 cells is required for consistent disease detection, rapid processing avoids cell degeneration, and the use of therapeutic monoclonal antibodies such as daratumumab can lead to false-negative MRD results. In addition to sample processing problems, immunophenotype in MM has been shown to be unstable, and many patients (41%) change immunophenotype, which can complicate interpretation.
In an attempt to overcome the potential loss of typical gating markers (e.g., CD38 and CD138) after treatment with monoclonal antibodies, the members of the signalling lymphocyte activation molecules family, CD229 and CD319, have been identified as new gating markers for PC in flow cytometric immunophenotyping, particularly in relapsed or resistant cases of MM.
The use of NGS, as with ASO-qPCR, is based on the identification and amplification of the sequence of interest, usually in the variable regions of the IGH gene on chromosome 14. However, for the follow-up of patients using PCR-based methods, a patient-specific primer must be developed from the diagnostic sequence. With NGS, these regions are amplified and sequenced with specific primers that are not patient specific.
The quantitation of BMPC% on Romanowsky-stained BMA alone may misrepresent the accurate count of PC due to the patchy nature of infiltration and sample haemodilution. Bone marrow PC percentage is consistently lower on BMA as compared to immunohistochemistry-stained BM biopsies. Flow cytometry alone yields consistently lower results compared to BMA myelograms and are not readily available in resource-constrained settings. Flow cytometry is, however, useful for determination of PC clonality, MRD, and risk stratification of MM. A combination of BMA, BM biopsy, and flow cytometry assessment of PC yield superior results compared to using each sample on its own.
CD138 or MUM1 immunohistochemistry improves PC identification compared to haematoxylin and eosin and other stains. Manual counting in representative areas of CD138-stained biopsies has been shown to improve reproducibility, compared to overview estimation. Use of AI image analysis on WSI or static images may currently be the best method for BMPC estimation. Guidelines for the standardisation and validation of digital techniques have been published and using static images in combination with open-source software for image analysis will be possible in situations with limited resources.
Radiological imaging is crucial for the diagnosis and treatment of MM. Introduction of new technologies such as double-energy CT and radiomics as standard practice may necessitate changes to IMWG guidelines. These techniques are still in the early stages of development for the assessment of patients with MM and are an exciting and rapidly evolving field. Further development will be required before radiological imaging can replace BM biopsy for BMPC quantitation.
Allele-specific oligonucleotide-qPCR and NGS are commonly used to quantify clonal PC in the follow-up of treatment response and detection of MRD. Allele-specific oligonucleotide-qPCR correlates well with MFC for MRD quantification, but has limitations in clonality detection and sequencing, which affects its applicability. Next-generation sequencing has replaced ASO-qPCR because it is more sensitive and does not require patient-specific primers for MRD monitoring, but using either molecular technique for this purpose can be challenging.
Diagnosis and classification of PC neoplasms depends on accurate quantification of PC in BM biopsy. Although there is no consensus on the optimal method for estimating PC in BM, CD138-stained biopsies are the preferred sample for this estimation and remain invaluable for diagnosis and assessment of response. No sample or method should be used in isolation, as they are better used in combination for this estimation, and laboratories without advanced techniques should continue to use the best methods available to them. In our quest for a gold standard method, this will require continued research and collaboration to expand and improve the currently published methods. Finally, digital pathology is likely to redefine the way we practise laboratory medicine and the role of the pathologist in this new digital age, and limited resources need not necessarily be a limiting factor.
The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.
E.J.G. conceptualised the manuscript, wrote the original draft and participated in the review and editing. Z.C.C. and E.M. were involved in the preparation of subsequent drafts and in the review and editing of the manuscript.
This article followed all ethical standards for research without direct contact with human or animal subjects.
Data sharing is not applicable to this article as no new data were created or analysed in this study.
The views and opinions expressed in this article are those of the authors and are the product of professional research. It does not necessarily reflect the official policy or position of any affiliated institution, funder, agency, or that of the publisher. The authors are responsible for this article’s results, findings, and content.