AUTHORS
Hadil Zureigat, MD, Shulan Tian, PhD2, Rong He, MD3, Dragan Jevremovic, MD PhD4, Patricia Greipp, DO5, Antoine N. Saliba, M.D.6, Aasiya Matin, M.B.B.S.7, Mehrdad Hefazi Torghabeh, MD8, Abhishek Avinash Mangaonkar, MBBS9, Naseema Gangat, MBBS10, William Hogan, M.B., B.Ch.11, Hassan B Alkhateeb, M.D.12, Cecilia Arana Yi, M.D.13, Yael Kusne, MD, PhD14, James Foran, FRCPC15, Talha Badar, MD16, Mark R. Litzow, MD17, Mithun Shah, MD, PhD18, Kebede Begna, M.D.19, Mrinal M. Patnaik, MD, MBBS20, Aref Al-Kali, M.D.21, Tariq Kewan, MD22
AFFILIATIONS
1 Mayo Clinic, Rochester, Minnesota, United States
2 Mayo Clinic, Rochester, Minnesota, United States
3 Mayo Clinic, Rochester, Minnesota, United States
4 Mayo Clinic, Rochester, Minnesota, United States
5 Mayo Clinic, Rochester, Minnesota, United States
6 Mayo Clinic, Rochester, Minnesota, United States
7 Mayo Clinic, Rochester, Minnesota, United States
8 Mayo Clinic, Rochester, Minnesota, United States
9 Mayo Clinic, Rochester, Minnesota, United States
10 Mayo Clinic, Rochester, Minnesota, United States
11 Mayo Clinic, Rochester, Minnesota, United States
12 Mayo Clinic, Rochester, Minnesota, United States
13 Mayo Clinic, Scottsdale, Arizona, United States
14 Mayo Clinic, Scottsdale, Arizona, United States
15 Mayo Clinic, Jacksonville, Florida, United States
16 Mayo Clinic, Jacksonville, Florida, United States
17 Mayo Clinic, Rochester, Minnesota, United States
18 Mayo Clinic, Rochester, Minnesota, United States
19 Mayo Clinic, Rochester, Minnesota, United States
20 Mayo Clinic, Rochester, Minnesota, United States
21 Mayo Clinic, Rochester, Minnesota, United States
22 Mayo Clinic, Rochester, Minnesota, United States
ABSTRACT
Introduction: Pathogenic DDX41 (DEAD-box helicase 41; 5q35.3) mutations are associated with a strong predisposition to myeloid neoplasms, characterized by distinct prognostic features not adequately captured by existing risk models, including the molecular international prognostic scoring system (IPSS-M) for myelodysplastic syndrome (MDS). Our aim is to identify key predictors of overall survival (OS) and develop a machine learning-based risk model to improve risk stratification in this population.
Methods: Patients with pathogenic DDX41 mutations detected by next-generation sequencing were identified. Demographics, diagnosis, treatment, and survival data was obtained from electronic health records. OS was calculated from diagnosis until death or last follow up date. Patients undergoing allogeneic hematopoietic stem cell transplantation (AHSCT) were censored at the time of transplant. Variable selection was performed using elastic net–regularized Cox regression (α = 0.2) with 5-fold cross-validation, followed by a final multi-variable Cox proportional hazards (CPH) model. In parallel, a gradient boosting machine (GBM) with Cox loss was trained on the full predictor set to rank variable importance. Model discrimination was assessed by Harrell’s C-index. A risk score derived from the final Cox model stratified patients into low-, intermediate-, and high-risk groups by tertiles.
Results: A total of 143 patients were identified, median age of 69 years (inter-quartile range 62–75), predominantly male 66% (n=94). Close to half 44% (n=63) had MDS, 32% (n=46) had acute myeloid leukemia, and 9% (n=13) were carriers, the remaining 15% (n=21) had other diagnoses including myeloproliferative neoplasm, clonal cytopenia of undetermined significance, bone marrow failure, and clonal hematopoiesis of indeterminate potential. A total of 70% (n=100) had hotspot DDX41 mutations including p.Arg525His, p.M1I and p.Asp140GlyFs*2. Overall, 56% (n=79) had no co-existing mu (CoM). Most common CoMs were ASXL1 (n=16, 11%), TET2 (n=14, 10%), DNMT3A (n=11, 8%), and SRSF2 (n=10, 7%). A total of 43% (n=62/96) received less-intensive regimens (hypomethylating agents +/- Venetoclax) and 9% (n=9/96) received intensive chemotherapy. A total of 23% (n=33) underwent AHSCT. Most patients (n=112, 81%) had normal karyotypes, del20q and loss of Y were the most common cytogenetic abnormalities. About 84% (n=117) had a single DDX41 mu only, while 16% (n=23) had multiple. A total of 63% (n=88) had suspected germline DDX41 mu, 14% (n=20) had somatic DDX41 mu only, and 23% (n=32) had both germline and somatic DDX41 mutations. GBM model ranked age, platelet count, bone marrow (BM) blast, absolute lymphocyte count (ALC), absolute neutrophil count (ANC), hemoglobin (Hb), treatment type, absolute monocyte count (AMC), and DDX41 variant allele frequency as the top predictors (C-index 0.90). Elastic Net model similarly selected age, BM blast, platelets, ALC and Hb. It also additionally selected DNMT3A CoM and abnormal cytogenetics.
Final CPH model included 10 predictors. Independent predictors of worse OS included higher BM blast (HR 1.77, 95% CI 1.27–2.48, p < 0.001), DNMT3A CoM (HR 6.20, 95% CI 2.00–19.23, p = 0.002), older age (HR 1.07 per year, 95% CI 1.02–1.12, p = 0.003), and abnormal cytogenetics (HR 2.10, 95% CI 1.04–4.22, p = 0.038). Higher platelet count (HR 0.38, 95% CI 0.16–0.87, p = 0.023) and higher ALC (HR 0.62, 95% CI 0.39–0.998, p = 0.049) were associated with better OS. Patients with a single somatic mu had similar OS compared to those with a single germline mu (HR: 1.5, p=0.43) or combined somatic /germline mutations (HR:2.5, p=0.11). The final CPH model achieved a C-index of 0.829.
The CPH-derived risk score stratified patients into three groups with significantly different outcomes: low-risk (n = 48; median OS 80.4 months), intermediate-risk (n = 47; median OS 56.3 months), and high-risk (n = 48; median OS 26.5 months). Pairwise log-rank tests confirmed significant separation across all groups (Low vs Intermediate, p = 0.017; Intermediate vs High, p < 0.001; Low vs High, p < 0.001).
Conclusion: Age, platelet count, BM blast percentage, abnormal cytogenetics, ALC, and DNMT3A CoM are key predictors of OS in patients with DDX41 pathogenic mu. A simple Cox-based risk score provides clear prognostic stratification (C-index 0.829), while GBM offers complementary ranking and higher discrimination (C-index 0.900), supporting improved risk assessment.
DISCLOSURES
Hadil Zureigat, MD: No relevant conflicts of interest to declare.