Abstract: Achievement of complete remission signifies a crucial milestone in the therapy of acute myeloid leukemia (AML) while refractory disease is associated with dismal outcomes. Hence, accurately identifying patients at risk is essential to tailor treatment concepts individually to disease biology. We used nine machine learning (ML) models to predict complete remission and 2-year overall survival in a large multicenter cohort of 1,383 AML patients who received intensive induction therapy. Clinical, laboratory, cytogenetic and molecular genetic data were incorporated and our results were validated on an external multicenter cohort. Our ML models autonomously selected predictive features including established markers of favorable or adverse risk as well as identifying markers of so-far controversial relevance. De novo AML, extramedullary AML, double-mutated CEBPA, mutations of CEBPA-bZIP, NPM1, FLT3-ITD, ASXL1, RUNX1, SF3B1, IKZF1, TP53, and U2AF1, t(8;21), inv(16)/t(16;16), del(5)/del(5q), del(17)/del(17p), normal or complex karyotypes, age and hemoglobin concentration at initial diagnosis were statistically significant markers predictive of complete remission, while t(8;21), del(5)/del(5q), inv(16)/t(16;16), del(17)/del(17p), double-mutated CEBPA, CEBPA-bZIP, NPM1, FLT3-ITD, DNMT3A, SF3B1, U2AF1, and TP53 mutations, age, white blood cell count, peripheral blast count, serum lactate dehydrogenase level and hemoglobin concentration at initial diagnosis as well as extramedullary manifestations were predictive for 2-year overall survival. For prediction of complete remission and 2-year overall survival areas under the receiver operating characteristic curves ranged between 0.77–0.86 and between 0.63–0.74, respectively in our test set, and between 0.71–0.80 and 0.65–0.75 in the external validation cohort. We demonstrated the feasibility of ML for risk stratification in AML as a model disease for hematologic neoplasms, using a scalable and reusable ML framework. Our study illustrates the clinical applicability of ML as a decision support system in hematology.

Abstract: Increasingly large and complex biomedical data sets challenge conventional hypothesis-driven analytical approaches, however, data-driven unsupervised learning can detect inherent patterns in such data sets. Methods While unsupervised analysis in the medical literature commonly only utilizes a single clustering algorithm for a given data set, we developed a large-scale model with 605 different combinations of target dimensionalities as well as transformation and clustering algorithms and subsequent meta-clustering of individual results. With this model, we investigated a large cohort of 1383 patients from 59 centers in Germany with newly diagnosed acute myeloid leukemia for whom 212 clinical, laboratory, cytogenetic and molecular genetic parameters were available. Results Unsupervised learning identifies four distinct patient clusters, and statistical analysis shows significant differences in rate of complete remissions, event-free, relapse-free and overall survival between the four clusters. In comparison to the standard-of-care hypothesis-driven European Leukemia Net (ELN2017) risk stratification model, we find all three ELN2017 risk categories being represented in all four clusters in varying proportions indicating unappreciated complexity of AML biology in current established risk stratification models. Further, by using assigned clusters as labels we subsequently train a supervised model to validate cluster assignments on a large external multicenter cohort of 664 intensively treated AML patients. Conclusions Dynamic data-driven models are likely more suitable for risk stratification in the context of increasingly complex medical data than rigid hypothesis-driven models to allow for a more personalized treatment allocation and gain novel insights into disease biology.

Abstract: Artificial intelligence (AI) is influencing our society on many levels and has broad implications for the future practice of hematology and oncology. However, for many medical professionals and researchers, it often remains unclear what AI can and cannot do, and what are promising areas for a sensible application of AI in hematology and oncology. Finally, the limits and perils of using AI in oncology are not obvious to many healthcare professionals. Methods In this article, we provide an expert-based consensus statement by the joint Working Group on “Artificial Intelligence in Hematology and Oncology” by the German Society of Hematology and Oncology (DGHO), the German Association for Medical Informatics, Biometry and Epidemiology (GMDS), and the Special Interest Group Digital Health of the German Informatics Society (GI). We provide a conceptual framework for AI in hematology and oncology. Results First, we propose a technological definition, which we deliberately set in a narrow frame to mainly include the technical developments of the last ten years. Second, we present a taxonomy of clinically relevant AI systems, structured according to the type of clinical data they are used to analyze. Third, we show an overview of potential applications, including clinical, research, and educational environments with a focus on hematology and oncology. Conclusion Thus, this article provides a point of reference for hematologists and oncologists, and at the same time sets forth a framework for the further development and clinical deployment of AI in hematology and oncology in the future.

Abstract: Introduction Recently, progress has been made in the treatment of patients with higher risk myelodysplastic syndromes (HR MDS) and acute myeloid leukemia (AML). Nevertheless, patients failing hypomethylating agents (HMA) have a dismal prognosis and very limited treatment options. Targeting CD123 on leukemic stem cells (LSC) is one promising approach in MDS and AML. Talacotuzumab (TAL, JNJ-56022473) is an IgG1 monoclonal antibody targeting CD123 preferentially via antibody-dependent cellular cytotoxicity (ADCC) mediated by natural killer cells (NKs). Aim The SAMBA trial, a phase II study of the German and French MDS study groups within the EMSCO network assessed the overall hematological response rate after 3 months of single agent TAL treatment in AML or HR MDS patients failing hypomethylating agents (HMAs). Methods TAL was given IV at a dose of 9 mg/kg once every two weeks for a total of 6 infusions, responders received up to 20 additional infusions. After the first 3 months, overall hematological response rate (either CR, PR, marrow-CR, HI, SD) was evaluated by bone marrow biopsy. The study was accompanied by an immune monitoring via flow cytometric analysis to investigate the distribution of T- and NK cells in peripheral blood (PB) and bone marrow (BM) at the time of screening and during therapy in comparison with healthy, age-matched controls. Results 24 patients (19 AML and 5 HR MDS) with a median age of 77 (range 71-90) years, who either failed to achieve complete- (CR), partial response (PR), hematological improvement (HI) or relapsed after HMA therapy were included in the study. After TAL administration, 14 patients could be assessed for response after 4 infusions and 10 patients after 6 infusions. The overall response rate (ORR) was 20.8% including 1 complete remission (CRi), 1 patient with hematologic improvement (HI-E) and additionally 3 patients with disease stabilization. The median duration of response in these patients was 3 months (range 3-14 months). Two patients are still on treatment, one patient despite losing objective response (14 months) and one patient with disease stabilization (13 months). The median overall survival for the entire cohort of patients was 3.2 months (range 0.4-11.2 months). In 10 patients (41.6%), therapy with TAL resulted in grade 3/4 infusion related side effects (pneumonia, n=1; infusion-related reaction, n=8; septic shock, n=1). Before treatment initiation, patients had lower levels of CD56dim NK-cells in PB (82% vs. 89% of NK-cells; p=0.069) expressing significantly more inhibiting NK-cell receptors like KIR2DL2 (8.8% vs. 3.2% of NK-cells; p 〈 0.001) and less activating NK-cells receptors like NKG2D (95% vs. 99% of NK-cells; p 〈 0.01) compared to healthy controls. Moreover, expression of PD-1 on lymphocytes and monocytes as well as their matching ligands PD-L1 and PD-L2 on blasts and monocytes in PB was significantly higher in patients compared to healthy controls (p 〈 0.01), another evidence for an exhausted T-cell immune status in our patients prior to treatment initiation. We could not detect any difference in NK-cell levels in responding patients compared to non-responders. Interestingly, pre-treatment expression (MFI and percentage) of CD123 on immature myeloid derived suppressor cells (iMDSC) was higher in responders than in non-responders (p 〈 0.01). Anti-CD123 targeted therapy with TAL resulted in a decreased CD123+ MFI (4239 vs. 2910; p 〈 0.01) on iMDSCs as well as lower levels of iMDSCs in PB and BM (p 〈 0.05).Responding patients displayed a 10-fold reduction of CD123 MFI after 3 months of treatment (2565 vs. 236; p=0.06), indicating that the CD123 molecule on immature MDSCs is targeted effectively by TAL. Conclusion Single agent TAL has limited efficacy in patients with advanced myeloid malignancies failing HMA. Expression of CD123 on immature MDSCs might serve as a biomarker of response for future anti-CD123 targeted approaches. Disclosures Götze: Celgene: Honoraria, Research Funding; JAZZ Pharmaceuticals: Honoraria; Novartis: Honoraria; Takeda: Honoraria, Other: Travel aid ASH 2017. Krönke:Celgene: Honoraria. Middeke:Roche: Membership on an entity's Board of Directors or advisory committees; Abbvie: Membership on an entity's Board of Directors or advisory committees; Janssen: Membership on an entity's Board of Directors or advisory committees, Research Funding. Fenaux:Celgene: Honoraria, Research Funding; Janssen: Honoraria, Research Funding; Jazz: Honoraria, Research Funding; Otsuka: Honoraria, Research Funding. Schlenk:Pfizer: Research Funding, Speakers Bureau. Ades:JAZZ: Honoraria; Takeda: Membership on an entity's Board of Directors or advisory committees; silent pharma: Consultancy; Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding. Platzbecker:Celgene: Research Funding.

Abstract: Background In newly diagnosed acute myeloid leukemia (AML), the general recommendation is to start treatment immediately after the diagnosis has been made. This paradigm is based both on the observation that untreated acute leukemia has a poor prognosis and on retrospective analyses demonstrating a shorter survival in younger AML patients (pts) in whom treatment was delayed by more than 5 days (Sekeres et al., 2009). A more recent single-center analysis came to a different conclusion, showing no prognostic effect for the time from diagnosis to treatment (TDT; Bertoli et al., 2013). We explored the relationship between TDT and prognosis on a large set of real-world data from the AML registry of the Study Alliance Leukemia (SAL) and compared it to the published cohorts. Methods The SAL runs a transregional AML registry in 46 treatment centers across Germany (NCT03188874). All registered patients with an intensive induction treatment, a minimum follow-up time of 12 months and no acute promyelocytic leukemia were selected (n=2,200). Treatment start was defined by the first day of cytarabine, whereas single agent hydroxyurea (HU) was labeled as pretreatment. We analyzed the influence of TDT on complete remission (CR), early death (ED) and overall survival (OS) in univariable analyses for each day of treatment delay, in groups of 0-5, 6-10, 11-15 and & gt;15 days of TDT, and by using the restricted cubic spline (RCS) method for data modelling. In order to adjust for the influence of established prognostic variables on the outcomes, we used multivariable regression models and propensity score weighting. The influence of HU pretreatment on outcomes was investigated by introducing an interaction term between TDT and the presence of HU pretreatment. Results The median age was 59 years (y) (IQR 50-68), the proportion of pts with favorable, intermediate and adverse genetic risk according to ELN was 27%, 53%, and 20%; & gt;95% of pts received induction treatment with standard 7+3. HU pretreatment was administered in 4% of pts. The median TDT was 3 days (IQR 2-6). Descriptive statistics after grouping of pts showed the highest median age and the lowest proportion of NPM1 mutated and favorable risk in the TDT group 11-15. Of all pts, 79% achieved a CR/CRi; unadjusted CR rates for the patient groups with TDT of 0-5, 6-10, 11-15 and & gt;15 days were 80%, 77%, 74% and 76%, respectively (p=0.317). In multivariable analysis accounting for the influence of ELN risk, age, WBC, LDH, de novo versus secondary AML and ECOG, the OR for each additional day of TDT was 0.99 (95%-CI, 0.97-1.00; p=0.124). Four percent of pts died within the first 30 days from treatment start. The respective rates in the four TDT categories were 4.0%, 3.8%, 5.1% and 4.1% (p=0.960). In multivariable analysis, the OR for TDT was 1.01 (95%-CI, 0.98-1.05; p=0.549). After a median follow-up of 40 months, the 2-y OS of all pts was 51%. The unadjusted 2-y OS rates stratified by TDT of 0-5, 6-10, 11-15, & gt;15 days were 52, 49, 46, and 51% (see Table 1 and Figure 1). The hazard ratio (HR) for each day of treatment delay was 1.00 (95%-CI; 0.99-1.01; p=0.317). In multivariable Cox regression analysis, the HR for TDT as continuous variable was 1.00 (95%-CI, 0.99-1.01; p=0.689). When OS was analyzed separately stratified for age ≤60 and & gt;60 ys and for high versus lower initial WBC defined by a threshold of 50 x 109/L, no significant differences between TDT groups were observed. Multivariable models using TDT as a grouped variable or with RCS did not provide evidence for a significant influence of TDT on outcomes. Propensity score matching of pts in the four TDT groups did not reveal an influence on outcomes. The use of HU was not associated with CR, ED nor OS. Conclusion Our study on 2,200 newly diagnosed registry pts receiving consistent intensive induction with standard-dose cytarabine plus daunorubicin (7+3) suggests that TDT is not related to response or survival, neither in younger nor in older pts. Despite multivariable analyses, a bias towards longer TDT intervals in pts judged to be clinically stable by the treating physician cannot be excluded entirely. As treatment stratification in intensive first-line treatment of AML evolves, the TDT data suggests that it may be a safe and reasonable approach to wait for genetic and other laboratory test results in order to assign clinically stable pts to the best available treatment option before the start of intensive treatment. Disclosures Krämer: Daiichi-Sankyo: Honoraria, Membership on an entity's Board of Directors or advisory committees; Bayer: Research Funding; BMS: Research Funding; Roche: Consultancy, Honoraria, Membership on an entity's Board of Directors or advisory committees. Hänel:Roche: Honoraria; Amgen: Honoraria; Celgene: Other: advisory board; Novartis: Honoraria; Takeda: Other: advisory board. Jost:Daiichi: Honoraria; Sanofi: Honoraria; Gilead: Other: travel grants; Jazz Pharmaceuticals: Honoraria. Brümmendorf:Merck: Consultancy; Janssen: Consultancy; Novartis: Consultancy, Research Funding; Pfizer: Consultancy, Research Funding; University Hospital of the RWTH Aachen: Employment; Ariad: Consultancy. Krause:Siemens: Research Funding; Takeda: Honoraria; MSD: Honoraria; Gilead: Other: travel; Celgene Corporation: Other: Travel. Scholl:Novartis: Other: Project funding; Pfizer: Other: Advisory boards; Gilead: Other: Project funding; Daiichi Sankyo: Other: Advisory boards; AbbVie: Other: Advisory boards. Hochhaus:Pfizer: Research Funding; Novartis: Research Funding; BMS: Research Funding; Incyte: Research Funding; MSD: Research Funding. Kiani:Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau. Middeke:Sanofi: Research Funding, Speakers Bureau; Roche: Speakers Bureau; AbbVie: Consultancy, Speakers Bureau; Gilead: Consultancy; Janssen: Consultancy, Speakers Bureau; MSD: Consultancy. Thiede:AgenDix GmbH: Employment, Equity Ownership; Novartis: Research Funding, Speakers Bureau; Bayer: Research Funding; Daiichi-Sankyo: Speakers Bureau. Stoelzel:JAZZ Pharmaceuticals: Consultancy; Neovii: Other: Travel funding; Shire: Consultancy, Other: Travel funding. Platzbecker:Celgene: Consultancy, Honoraria, Research Funding; Abbvie: Consultancy, Honoraria; Novartis: Consultancy, Honoraria, Research Funding.