Evaluation of an innovative new method for quantitation of plasma cells on CD138 immunohistochemistry

Abstract

AimTo compare the frequently used CD138 immunohistochemistry-based method of plasma cell quantitation, to a proposed new method, using interobserver and intraobserver concordance parameters.MethodsArchival CD138 immunohistochemically stained slides made from paraffin-embedded bone marrow biopsies of 33 patients with a confirmed diagnosis of multiple myeloma were used. Light microscopic examination was performed using low magnification lenses (10×) for both the overview estimation method (method A) and the new method (method B), and high magnification lenses (50×), for method B only. For method B, reviewers selected three areas with low, intermediate and high plasma cell densities using 10× lenses. Using a well-defined technique, the 50× lens was then used to count plasma cells as a percentage of all nucleated cells. After blinded relabelling of all the slides, the nine reviewers repeated the plasma cell quantitation using both methods. The plasma cell counts were obtained, and the review times were recorded.ResultsOverall intraobserver concordance was comparable for method A (concordance correlation coefficient (CCC)=0.840) and method B (CCC=0.733). Interobserver concordance for method A (intraclass correlation coefficient (ICC)=0.793 and 0.713) and method B (ICC=0.657 and 0.658) indicated high similarity between reviewers. Method A showed poor interobserver concordance (ICC=0.105) at low plasma cell densities.ConclusionsThe new method is comparable to the frequently used overview estimation method in terms of intraobserver and interobserver concordance, and cost. The new method has superior interobserver concordance at low plasma cell densities. The new method appears more amenable to digital scanning and analysis.

Evaluation of a new technique using artificial intelligence for quantification of plasma cells on CD138 immunohistochemistry

“…In our previous study, method B had lower mean plasma cell percentages than method A. 13 The mean plasma cell percentages for method C are also statistically significantly lower than method A in this study. This could be due to method A overestimating plasma cell tumour burden or method B and method C underestimating plasma cell tumour burden.…”

“…In our previous study, we had found that method A1 (overview estimation method) showed poor interobserver concordance at low plasma cell tumour burdens (ICC = 0.105 [95% CI À0.009 to 0.445]). 13 The mean plasma cell percentages and standard deviations for method C are comparable to those of method B. In our previous study, method B had lower mean plasma cell percentages than method A.…”

“…The latter method also had poor interobserver concordance compared to the new manual counting method, especially when the plasma cell tumour burden was low. 13 It has been suggested that this new method of selecting three areas representative of the bone marrow may be more suitable for digital scanning and analysis with the appropriate software. Baranova et al 14 validated the use of an open-source digital image analysis tool, QuPath, 15 to automate the quantification of plasma cells in bone marrow biopsies from patients with multiple myeloma.…”

“…The nine pathologists (reviewers) included two senior pathologists with 4-29 years postqualification experience (category 1), four junior pathologists with 0-3 years post-qualification experience (category 2) and three pathologists in specialist training (category 3). 13 After selecting the three areas most representative on the tissue biopsy and image capture (digitisation) of the areas, digital analysis of these areas was performed using the recently validated open-source machine learning software, QuPath, for quantification of plasma cells in bone marrow. 15 Each reviewer received written instructions on selecting representative areas and capturing images (digitising samples) for digital analysis using the open source software (online Supplemental Appendix S1).…”

Comparison of laboratory methods for the detection of neoplastic plasma cells in plasma cell dyscrasias

“…A more objective computational method could potentially avoid this bias, although such a method was not employed in this study. Other authors have recently found that measuring concordances of plasma cell quantification between IHC with light microscopy at different magnifications have yielded comparable results 21…”