Author
Anantharamanan Rajamani1, Henrik Himmelreich1, Florian Simon1, Jasmin Schulz1, Sandra Robrecht1, Can Zhang1, Eugen Tausch2, Christof Schneider2, Matthias Ritgen3, Karl Anton Kreuzer1, Brenda Chyla4, Hyun Yong Jin5, Yanwen Jiang5, John F Seymour6, Arnon Kater7, Stephan Stilgenbauer2, Barbara Eichhorst1, Michael Hallek1,Kirsten Fischer1, Othman Al-Sawaf1
Introduction
Aberrant gene expression impacts disease biology and correlates with patient outcomes in CLL, but expression-based prognostic models have had limited clinical applicability to date. We aimed to develop a composite gene signature-based risk stratification model using transcriptomic data.
Methods
Transcriptomic profiles of 405 CLL patients with available samples from the first-line phase 3 CLL14 trial (Ven-Obi vs. Clb-Obi) were used to train a prognostic model for progression-free survival (PFS). Cox proportional hazards and cross-validated LASSO models were applied subsequently to highly variable genes (HVGs) to identify those associated with PFS. A risk score was derived by applying nonzero model coefficients to normalized expression values. Risk status was assessed using a multivariable Cox model adjusted for genetic markers. The model was subsequently validated in cohorts from the MURANO relapse trial (BR vs. Ven-Rit; n=204) and the Broad (cllmap.org; n=568) dataset.
Results
A total of 107 of 251 HVGs were significantly associated with PFS (p < 0.01). An eight-gene (BCAT1, LHFPL6, C10orf10, ZNF471, PLD1, DNAH14, KLK4, and IL15) derived risk score stratified patients into high-risk (n=251) and low-risk (n=154) classes (A,B). Risk status correlated with PFS (hazard ratio [HR]=3.12 [2.05–4.75], p< 0.001), independent of IGHV mutation and del17p/TP53 alterations (C). Both high-risk IGHV-mutated and unmutated patients had shorter PFS compared to their respective low-risk counterparts (HR=3.68 [2.16–6.29], p< 0.001 for IGHV-mutated; HR=0.403 [0.22–0.73], p=0.002 for IGHV-unmutated, comparing low- vs. high-risk). Among del17p/TP53 subgroups, low-risk patients with or without del17p/TP53 exhibited longer PFS than high-risk patients (HR=0.10 [0.023–0.44], p=0.002 for mutated; HR=3.02 [2.20–4.13], p< 0.001 for wild-type, comparing high- vs. low-risk). C-statistics of 0.62 and 0.63 from the two validation cohorts supported the model’s prognostic performance.
Conclusions
The multi-gene prognostic model reproducibly identifies clinically relevant subgroups with distinct prognoses, facilitating refined risk stratification in both frontline and relapsed CLL.
Keywords : Leukemia, Relapse, Biomarker
Please indicate how this research was funded. :
Please indicate the name of the funding organization.: