(303) Impacts of the COVID-19 Pandemic on Assessment of Progression and Surveillance Bias in Advanced Non-Small Cell Lung Cancer

Publication Number: 977

Background: The COVID-19 pandemic disrupted cancer care in the U.S. Delays in assessment of tumor burden in routine clinical care may introduce surveillance bias in real-world studies of patients (pts) with cancer.

Objectives: The primary objective of this study was to explore the extent to which the observed real-world progression free survival (rwPFS) varied among pts with advanced non-small lung cancer (aNSCLC) treated before versus during the pandemic.

Methods: In this retrospective observational study, we utilized the nationwide de-identified, electronic health record-derived Flatiron Health database. The analysis included pts with aNSCLC who initiated at least one line of systemic therapy in the pre-pandemic (Mar 1 to Nov 30, 2018) and/or pandemic (Mar 1 to Nov 30, 2020) eligibility periods. Follow-up ended on Mar 1, 2019 and Mar 1, 2021, respectively, with data cutoff on Sept 30, 2021. The observed rwPFS was analyzed from the time of treatment initiation using Kaplan-Meier methods. Cox proportional hazards models with adjustment for pt demographics (age at diagnosis, race/ethnicity, region, smoking status) and clinical characteristics (practice type, group stage at diagnosis, systemic therapy class, ECOG) were used to estimate hazard ratios (HR) and 95% confidence intervals (CI). All analyses were stratified by line of therapy (first line [1L] and later) with consistent censoring rules for progression and death applied for both study periods.

Results: Overall, 6,440 eligible pts with aNSCLC were included in the pre-pandemic period and 5,796 in the pandemic period. Most pts (>90%) initiated 1L and/or 2L systemic therapy in the study period. The proportion of pts with an observed rwP event in the pre-pandemic period was greater than those initiating one of these lines during the pandemic (1L: 31% vs 27%, respectively, p < 0.05; 2L: 41% vs 35%, respectively, p < 0.05). A lower proportion of deaths was observed among pts initiating 1L treatment during the pre-pandemic period compared to the pandemic period (8% vs 10%, respectively; p = 0.03). Among pts who initiated 1L therapy, treatment in the pre-pandemic period was associated with significantly shorter observed median rwPFS compared to treatment during the pandemic (8.0 vs 9.6 months; HR 1.12; 95% CI 1.03-1.22; p  < 0.01). A stronger association was observed among pts who received 2L (5.5 vs 6.7 months; HR 1.30; 95% CI 1.16-1.45; p  < 0.01).

Conclusions: Our findings suggest that pts with aNSCLC initiating treatment during the COVID-19 pandemic were less likely to have a progression event observed despite having a similar length of potential follow up. Additional research should determine the extent to which surveillance bias may artificially lengthen the observed median rwPFS.