Feasibility of RECIST-based real-world tumor response assessment in metastatic non-small cell lung cancer patients.

e21121 Background: Treatment response to anti-cancer therapies for advanced lung cancer is usually assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST), which is not generally applied in real-world settings. With real-world evidence increasingly being used to support promising treatment, this study evaluated the feasibility of assessing real-world lung cancer response by RECIST-based measurement of lesions on archived radiologic films and assessed its concordance with treatment response based on oncologist narratives in electronic health record (EHR). Methods: This retrospective study included 30 randomly selected metastatic non-small cell lung cancer (mNSCLC) patients diagnosed between January 2015 and December 2019 from the Syapse Learning Health Network of US community health systems. Eligible patients were ≥18 years, histologically confirmed mNSCLC; no other malignancy within 2 years prior to diagnosis; initiated index systemic treatment subsequent to progression on a platinum or anti-PD(L)-1-containing therapy; and had a baseline scan followed by at least one additional scan during the index period. Patients were followed from the initiation of index treatment until the initiation of a post-index line of treatment, death, or date of last contact, whichever occurred first. Tumor response was established using response documented in the medical oncologists’ narratives and compared to a radiologist’s assessment of archived images using RECIST v1.1 criteria. Best overall response was characterized as complete or partial response (CR/PR), stable disease (SD), progressive disease (PD), and not evaluable (NE). Results: The median age at mNSCLC diagnosis was 62 years; 22/30 (73%) were male; 28/30 (93%) were White and 2/30 (7%) were Black. Of the 22/30 (73%) patients with a documented Eastern Cooperative Oncology Group performance status, 21 (95%) had a status of 0 or 1. Patients had a median of 8 months follow-up with the majority surviving (63%) until the end of study. Table shows good concordance of best overall response between medical oncologist-reported and radiologist re-assessed responses; CR/PR was 67%, SD 79% and PD 80%. Causes of discordance were lack of CR/PR confirmation in medical oncologists’ narratives, absence of narratives, or presence of clinical symptoms. Conclusions: This study is an important step in demonstrating the potential to approximate real-world RECIST-based treatment response using EHR abstraction of oncologists’ documentation. Future studies should validate this approach, and improve upon it, by exploring additional real-world data elements as surrogates for clinically relevant responses to anti-cancer therapies. [Table: see text]