AUTHORS
Akram Abushamma, MD, Hussein Awada2, Aashray Mandala, BS3, Serhan Unlu, MD4, Mark David Orland, MD, PhD5, Zachary Brady, BS6, Kartik Lakhotiya, MS7, Carmelo Gurnari, MD,PhD8, Valeria Visconte, PhD9, Arda Durmaz, PhD10, Jaroslaw P. Maciejewski, MD, PhD, FACP11
AFFILIATIONS
1 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
2 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
3 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
4 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
5 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
6 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
7 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
8 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
9 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
10 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
11 Cleveland Clinic - Taussig Cancer Center, Cleveland, Ohio, United States
ABSTRACT
Introduction. The presence of paroxysmal nocturnal hemoglobinuria (PNH) clones is highly specific to acquired idiopathic anemia (AA) in which GPI-deficient clones are believed to be a marker of immune–mediated disease. Currently, the pathogenetic context determining expansion and progression to frank PNH remains unclear. Indeed, some patients bypass the AA stage and present with primary hemolytic PNH (phPNH), retaining some characteristics of immune-mediated AA. PhPNH may result from either failed immune surveillance or an overactive immune reaction leading to early elimination of normal hematopoiesis. The general definition of PNH versus AA/PNH syndrome has been arbitrary, but establishing a cutoff of clonal threshold is needed to analyze the speed of clonal progression from AA to PNH. Following AA, the diverse clonal trajectory may reflect ongoing immune selection versus tolerance or successful therapy, while the implications for the pathogenesis of PNH are diametrically opposite. Nevertheless, the first step is to define the types of evolution using and cutoff value in terms of PNH clone size and PFS as classifier.
Aim. The primary objective is to characterize AA/PNH into immunogenetic and phenotypic subgroups with distinct features and to create a predictive model for their corresponding trajectories.
Methods. We analyzed a large cohort of patients who underwent a flow cytometry (FCM) for PNH clone detection (with a sensitivity threshold of 0.01%). We assigned the cohort into non-progressors (G1; granulocyte clone size remained <10%), progressors (G2; granulocyte clone size expanded to ≥10%), and phPNH (G3, baseline granulocyte clone size of ≥10%). G2 were further stratified into either slow or fast progressors according to the median progression time (G2a: progressed after the median progression time; G2b: progressed at or below the median progression time, respectively).
Results. Of 627 patients who underwent FCM, 449 were included in the subgroup analysis (71.5% G1, 9.1% G2, 19.4% G3), while 181 patients were censored due to follow-up time less than the median time for progression. The cohort had a median age of 54 years (IQR 33–68), 53% were female, and the median follow-up time was 85.5 months (IQR 38.6–175.6). The cohort had a median of 2 FCM measurements (IQR 1–4), and a baseline PNH granulocyte clone size of <1% in 279 patients (62.1%), 1%–<5% in 50 (11.1%), 5%–<10% in 33 (7.3%), and ≥10% in 87 (19.4%). Overall, the median time to progression was 29.5 months (IQR 8.9–68.3); by definition of the median split, median progression was 68.6 months (IQR 35.7–86.1) in G2a and 8.9 months (IQR 5.6–14.0) in G2b. Using baseline characteristics, we compared the subgroups across 30 variables, of which 17 showed significant differences across the 4 groups. However, there were no significant differences between G2a/b. Therefore, we applied machine learning strategy. Both random survival forest and Cox proportional hazards models showed a good discrimination of PNH progression with a C-index of 0.870 vs 0.882, respectively, and a time dependent AUCof 0.892 vs 0.928, respectively, at 30-months. The most important predictors were, as expected, baseline clone size, PIGA mosaicism, and AA severity. Upon using the censored sample for prognostic performance, the model predicted lower 24-months cumulative progression (median 1.1% vs 2.5%, P=.003). Given the trained model, we projected the censored cases and stratified them into predefined groups based on the likelihood of progression at 24 months.
Conclusions. Characterization progression trajectories have clinical implications and may also contribute to understanding PNH pathogenesis. Here, we defined types of progression using multiparameter machine-learning methods and used them to identify parameters that characterize progression groups and their clinical and immunogenetic features.
DISCLOSURES
Akram Abushamma, MD: No relevant conflicts of interest to declare.