A Novel Strategy for Assessing Bone Marrow Plasma Cell Percentage: Development and Internal Validation of a Surrogate Calculation Approach.

Abstract

Introduction: The differentiation and diagnosis of plasma cell (PC) neoplasms (PCNs) such as multiple myeloma (MM) rely on the quantification of clonal PCs in the bone marrow (BM). For monitoring, the International Myeloma Working Group (IMWG) defines stringent response criteria based on the percentage of BM PC. However, BM biopsies are invasive and painful, and often with sampling variability. This study investigates whether routine biomarkers can predict BM trephine (BMT) PC% using multivariate regression.

Methods: A cross-sectional study was conducted at Tygerberg Hospital, South Africa. Data were extracted from the National Health Laboratory Service (NHLS) database. The final dataset included 112 newly diagnosed MM patients with complete biomarker data for training of the partial least squares regression (PLS-R) model. Variables analyzed included SFLC ratio, paraprotein, Hb, calcium, creatinine, and albumin. Statistical methods included correlation analysis, regression modeling, and internal validation.

Results: The cohort had a median age of 61 years and a male-to-female ratio of 1:1.5. PLS-R analysis identified significant predictors of BMT PC%, including SFLC ratio, paraprotein, Hb, and serum albumin. The final model equation showed moderate predictive power (Q ² = 0.410, R ² Y = 0.432). Spearman correlation analysis showed a moderate positive relationship (p = 0.585) between predicted and actual BMT PC%. Linear regression analysis also confirmed that BMA PC% (R ² = 0.489) was a stronger predictor of BMT PC% than flow cytometry PC% (R ² = 0.184).

Conclusion: This study provides proof of concept for the use of biochemical markers to predict BMT PC% and offers a less invasive alternative for PC quantification in PCN. Standardization of sampling and measurement of biomarkers is essential for the refinement of these predictive models. Future multicenter studies should include prospective data collection to improve model accuracy and clinical applicability.