Validation of novel prognostic models in symptomatic Waldenström's macroglobulinemia based on real-world data

Abstract Introduction Waldenström's macroglobulinemia (WM) is a rare, indolent, IgM-producing lymphoplasmacytic lymphoma with diverse clinical outcomes. Prognostic models have evolved alongside treatment shifts from chemotherapy to chemoimmunotherapy and targeted therapies. In addition to the original International Prognostic Scoring System for WM (IPSSWM), newer models such as the revised IPSSWM (rIPSSWM) and the Modified Staging System for WM (MSSWM) have been developed, incorporating markers such as lactate dehydrogenase and albumin. However, their performance has not yet been thoroughly validated. Aim: To evaluate the performance of prognostic models in a large, real-world patient cohort and to examine treatment patterns and survival outcomes of symptomatic WM (sWM) in the Czech Republic over the past ten years. Methods sWM patients who started first line (L1) treatment between January 2014, and January 2024 were included. The primary data source was the Czech Myeloma Registry and Czech Lymphoma Study Group Project (NCT03199066). The primary endpoint was overall survival (OS) (calculated from L1 initiation), secondary endpoints were preferred regimen, treatment responses and progression-free survival (PFS). Patients with complete data for scoring systems were included in multivariable analysis and validation of existing prognostic models. The Kaplan-Meier method was used for survival analysis, and groups were compared using pairwise log-rank tests with Benjamini-Hochberg correction. Results In total, 315 patients were included in the analysis (median age 69 years at L1 initiation, 58% male). Molecular genetic testing data for MYD88 and CXCR4 were available for 143 (45%) and 27 (11%) patients, with 137 (96%) and 11 (41%) positively tested for MYD88 and CXCR4 mutations, respectively. Cytopenia, IgM-associated complications, and B symptoms were the most frequent indications for treatment. In the front-line setting, 96% of patients received chemoimmunotherapy (76% cyclophosphamide, 22% bendamustine, and 2% bortezomib-based combinations), 2% rituximab monotherapy and 2% other combinations. With a median follow-up (FU) of 4.8 y (IQR 2.8; 6.8) the overall response rate (ORR) was 86% (257/296 patients with known response: CR 12%, VGPR 17%, PR 67%). Median PFS was 5.3 years (4.5-7.1 years, 95% CI). The 3-year and 5-year probability of OS were 87% (84–91%, 95% CI) and 80% (75–86%, 95% CI), respectively. Among 76 patient deaths 26 (34%) was directly related to WM (progression, histologic transformation, treatment related complications, secondary malignancies). We were able to evaluate the IPSSWM risk score in 304 (97%) patients (low 16%, intermediate 36%, high 48%), rIPSSWM in 291 (92%) patients (very low 15%, low 23%, intermediate 29%, high 21%, very high 12%) and MSSWM in 302 (96%) patients (low 19%, low-intermediate 29%, intermediate 31%, high 21%). The 5-year OS according to the IPSSWM was 95%, 89% and 71% for low, intermediate and high risk respectively, for rIPSSWM was 95%, 95%, 75%, 70% and 64% for very low, low, intermediate, high and very high risk respectively, and for MSSWM 96%, 83% 74% and 70% for low, low-intermediate, intermediate and high risk respectively. All models significantly differentiated between high-risk and low-risk groups, while intermediate groups provided poor resolution. Additionally, in a direct comparison using a subset of 291 patients with all risk system available, the respective C-index and AIC values were: 0.72 and 455 for rIPSSWM; 0.68 and 464 for MSSWM; and 0.66 and 461 for IPSSWM. Interestingly, Beta-2microglobulin (B2M) was a significant prognostic factor, both as a continuous variable and as a binary variable (≥ 4 mg/L). In the binary form, B2M was significant in both univariable (HR: 2.79; p < 0.001) and multivariable Cox regression analyses adjusted for age, LDH, and albumin (HR: 2.2; p = 0.004). Conclusion Over the past decade, rituximab-based immunochemotherapy has been the most commonly used treatment for WM in the Czech Republic. In our real-world cohort, rIPSSWM was the best-performing model, achieving the highest concordance index and the lowest AIC. However, the MSSWM provided comparable results using fewer predictors, potentially simplifying its clinical application.