DIPSS-R: A revised age-agnostic clinical risk model for chronic phase primary myelofibrosis

Abstract

Abstract Background: The International Prognostic Scoring System (IPSS; Cervantes et al. Blood 2009;113) and dynamic IPSS (DIPSS; Passamonti et al. Blood 2010;115:1703) were developed about 15 years ago, based on 5 clinical risk factors: age >65 years, hemoglobin <10 g/dl, white blood cell count (WBC) >25 x 109/L, peripheral blood (PB) blasts ³1%, and constitutional symptoms. DIPSS-independent genetic risk factors have since been incorporated into molecular risk models with higher predictive accuracy (Guglielmelli et al. JCO 2018;36:310; Tefferi et al JCO 2018;36:1769). Because of its accessibility, DIPSS is often preferred for use in patient selection for clinical trials. The objective of the current study was to revise DIPSS by incorporating new and refined clinical risk variables and excluding those that are subjective (e.g., symptoms) or could exaggerate the need for therapy for older adults (e.g., age). Methods: Diagnostic criteria were according to the International Consensus Classification (Arber et al. Blood 2022; 140:1200). Severity of anemia was adjudicated by the Revised International Working Group-European LeukemiaNet criteria (Blood 2024; 144:1813). Receiver operating characteristic curve analysis was utilized to determine the optimal cutoff points for age, WBC, PB blast percentage, absolute monocyte count (AMC), and platelet count. Data were collected retrospectively corresponding to the time of first referral, which, in the majority of cases, coincided within one year of initial diagnosis and before initiation of cytoreductive therapy. Survival analysis was censored for allogenic stem cell transplantation (ASCT). Time-specific AUC was computed at pre-defined landmark time points and used to compare model performance between DIPSS and the revised model (DIPSS-R). Results: The discovery cohort from the Mayo Clinic included 1,073 patients (median age 65 years; 61% males) and the validation cohort from Italy 635 patients (median age 63 years; 60% males). In the discovery cohort, median follow-up (% deaths, AML, ASCTs) was 45 months (49%, 10%, 17%) and in the validation cohort 60 months (42%, 11%, 5%). In univariate analysis, the following correlated with inferior survival, in both the discovery and validation cohorts, with significance confirmed by multivariable analysis (MVA) in the discovery cohort: severe anemia (transfusion-dependent or hemoglobin level <8 g/dL in women or <9 g/dL in men), moderate anemia (hemoglobin level 8 to <10 in women or 9 to <11 in men), marked leukocytosis (WBC ≥25 x 109/L), moderate leukocytosis (WBC 11 to <25 x 109/L), age >65 years, PB blast ≥2%, AMC ≥1 x 109/L, platelet count <150 x 109/L, constitutional symptoms, and male sex. After excluding age, sex, and constitutional symptoms, MVA-derived HRs (95% CI) were 3.5 (2.8-4.4) for severe anemia, 2.0 (1.5-2.5) moderate anemia, 2.4 (1.8-3.3) marked leukocytosis, 1.5 (1.2-1.9) moderate leukocytosis, 1.7 (1.4-2.1) PB blast ≥2%, 1.5 (1.2-2.0) AMC ≥1 x 109/L, and 1.5 (1.2-1.8) platelet count <150 x 109/L. An HR-weighted risk model was subsequently developed (Figure 1) and validated (Figure 2) using the discovery and validation cohorts, respectively: low risk (0 points; N=205; median survival 226 months), intermediate-1 risk (1 point; N=253; 131 months), intermediate-2 risk (2-3 points; N=457; 62 months), and high risk (4-7 points; N=153; 31 months); the corresponding median survival estimates in the validation cohort were “not reached”, 205 months, 96 months, and 44 months (inter-risk category p values <0.01 in all instances). Ten-year predictive accuracy was comparable between DIPSS-R (AUC 0.78) and DIPSS (AUC 0.78), despite the exclusion of age and constitutional symptoms, as risk variables, from the revised model. Furthermore, predictive performance was superior for DIPSS-R vs. DIPSS when analysis was stratified by age: >65 years (AUC 0.77 vs. 0.74) and ≤65 years (AUC 0.71 vs. 0.66). MVA that included high (ASXl1, SRSF2, U2AF1) and low (CALR type 1/like) risk mutations, karyotype, and DIPSS-R variables showed independent prognostic contribution from all except thrombocytopenia (p=0.2; Figure 3). Conclusion: DIPSS-R utilizes disease-related, as opposed to host-related, predictors of survival whose prognostic contribution was independent and complementary to genetic risk factors. The model preserves global applicability and allows for a more accurate assessment of disease severity for practice and research purposes.