Background: Although predictive of chemotherapy toxicity, geriatric assessment measures are not systematically collected in clinical practice and may or may not be predictive for immune-related adverse events. Furthermore, hospitalization during immune checkpoint inhibitor (ICI) treatment for advanced lung cancer has variable prognostic significance. This study aimed to evaluate whether age and documented patient characteristics mapped to geriatric assessment domains (frailty markers, FM) predict survival in this setting. Methods: A single-center retrospective cohort of advanced stage lung cancer patients who received >1 dose of an ICI from 6/1/18 to 2/1/20, were later hospitalized, and received ≥ 1 dose of systemic corticosteroids (n=97) was analyzed. Chart review ascertained documentation of any of the following FMs prior to ICI initiation: inability to walk one block, unintentional weight loss, decreased social activities, recent falls, need for assistance with medications, visual or hearing impairments, living alone, and concern regarding social support. Patients were stratified according to age and three FM categories (0 FM [low risk], ≥1 FMs [at risk], and ≥2 FMs [high risk]). Overall survival (OS) analysis was calculated from first dose of ICI to date of death or last follow-up. Cox’s proportional hazards models were used to assess the relationship between FMs and age on OS; hazard ratios (HR) and 95% confidence intervals (CI) were calculated. Results: Analysis of < 75 and ≥ 75 yo revealed a median OS of 15.1 and 5.4 months respectively (HR 2.76, CI 1.62-4.72). Controlled for performance status (PS), older age (≥75 yo) was associated with a higher risk of death (HR 2.39, CI 1.32-4.31). FMs were associated with higher mortality, adjusted for PS and age (at risk patients HR 1.81, CI 1.03-3.16; high risk patients HR 2.02, CI 1.07-3.78). PS prior to starting ICI was not associated with OS. Conclusions: Age ≥ 75 yo is associated with short survival among lung cancer patients hospitalized while receiving ICI. Pre-treatment FMs documented as part of usual care were associated with worse OS, even after controlling for PS and age. This study shows promise for use of machine learning algorithms to stratify risk in hospitalized patients undergoing treatment for lung cancer with ICIs. These data would allow providers to better target serious illness conversations and end-of-life resources.

Baseline characteristics of pre-treatment groups (low risk: 0 FMs, at risk: ≥1 FMs, high risk: ≥2 FMs).