e18844 Background: Tissue diagnostics for metastasized non-small cell lung cancer (NSCLC) is hampered by the difficulty of obtaining tissue biopsies, and sequential testing. This leads to patient burden, and a time-consuming diagnostic trajectory, sometimes without valid test results. We analyzed the diagnostic process of liquid biopsies (ctDNA) as an addition to tissue diagnostics, for patients with stage IV NSCLC. Methods: Data was retrieved from stage IV patients (N=209) included in the observational, Lung cancer Early Molecular Assessment trial (LEMA; NCT02894853). Within this study, plasma and tissue was collected upfront. Sequencing of ctDNA was performed retrospectively, and compared to results from tissue diagnostics, as performed in the including hospital, including test results, success-rates, and identified targets per patient. Throughput time for tissue, and costs for diagnostics were calculated within a single specialized institute. Because of the retrospective nature of the ctDNA assessment, diagnostics for implementation of ctDNA was modelled (1 run per week), based on estimations for the trajectory and throughput time. A discrete-event simulation model was developed with three scenarios: 1) diagnostics with tissue biopsy alone (comparator); 2) diagnostics with ctDNA first, and tissue biopsy if required; 3) ctDNA if biopsy failed. The process scope ranged from tissue/blood retrieval to outcome in terms of targets. Simulation analysis was performed, calculating mean price per patient (pp), median throughput time pp, and targets found per scenario. Results: In scenario 1, 85% of the patients had a clinically relevant test result, compared to 93% and 92% when ctDNA was added in scenario 2 and 3, respectively. Scenario 2 was the fastest, but also the costliest scenario: median throughput time 15 days pp (IQR 8-30) and mean costs of €2.716 pp (standard deviation [SD] €789) (Table). In this scenario, 9.9% of the biopsies could be cancelled, and in 39.6% a simple biopsy with immunohistochemistry was sufficient. Conclusions: The addition of ctDNA may lead to better informed treatment decisions in patients with stage IV NSCLC. Discrete-event simulation analysis showed that the addition of ctDNA in the diagnostic process increased proportion of patients with a clinically relevant test-result in both scenarios. Throughput time is faster, and additional costs per patient are limited, especially when ctDNA is added after biopsy failure. Clinical trial information: NCT02894853. [Table: see text]
Machine learning for tissue diagnostics in oncology: brave new world
