P565: real-world data as supplementary controls for the prospective randomized hovon-103 trial in intensively treated elderly acute myeloid leukemia patients

Background: To replace or supplement control patients in prospective phase II/III studies, it has been suggested to use Real-World Data (RWD). However, it is unknown how outcome of RWD patients compares to outcome of trial control patients. Aims: Here, we compared the outcome of control patients of the prospectively randomized phase II HOVON-103 (H103) trial for elderly, intensively treated AML patients with the outcome of similarly treated elderly AML patients registered by the Netherlands Cancer Registry (NCR). Methods: The H103 randomized 679 patients between standard first-line intensive chemotherapy with or without selinexor, tosedostat, or lenalidomide, in 3 consecutive randomizations. We retrospectively identified 2,469 newly diagnosed elderly AML patients between 2014-2018 from the NCR, of whom 322 were selected for comparison with all H103-controls (n=279) based on first-line intensive treatment, WHO performance score ≤2 or unknown WHO status, and time between diagnosis and treatment ≤33 days. We performed 1:1 nearest neighbor propensity score matching (PSM) for age, modified European LeukemiaNet (ELN) 2017 AML risk classification, and white blood cell count. Results: Before PSM, NCR-patients compared well with H103-controls for age (median: 69 vs. 69 years, P=0.51), sex (61.5% vs. 61.6% male, P=1.00), blood parameters, and modified ELN2017 risk classification (16.8% vs. 13.3% favorable, P=0.43). WHO performance status was missing in 53.7% of NCR-patients vs. 1.4% in H103-controls. Comorbidity and toxicity data were absent in the NCR. Patient characteristics were also balanced after matching. After PSM, the 2-year OS was 29±3% for NCR-patients vs. 39±3% for H103-controls (Figure 1). Summary/Conclusion: Selected NCR-patients were comparable with H103-controls for most patient characteristics present in both databases. However, a significantly lower OS was observed for NCR-patients. The lack of recorded performance status and absence of comorbidity and toxicity scores in NCR-patients might explain these differences. Unless RWD AML are supplemented with these important scores, RWD cannot fully replace phase II/III trial control patients.Keywords: Clinical trial, Real world data, Acute myeloid leukemia