OBJECTIVES: Real-world data (RWD) is increasingly being used to support regulatory and payer submissions, particularly for genetically-defined cancer subtypes and other rare populations. An important concern for treatment effects estimated from RWD is whether these are generalizable or externally valid beyond the particular sample used for the analysis. This study seeks to evaluate the transportability of treatment effects estimated using RWD for metastatic non-small cell lung cancer (mNSCLC) between United States and Canada using two large, high-quality real-world datasets.

METHODS: Eligible patient cohorts will be defined using RWD from Flatiron Health Analytics Database (FHAD) in the US and using population-based RWD for the entire Canadian province of Alberta, serving as the sample and target populations of interest. G-computation will be used to estimate the effect of adhering to the specified first- and second-line treatment regimen for mNSCLC on overall survival, with an additional transportability adjustment to account for differences in effect measure modifiers between FHAD and the Canadian province of Alberta. Standardized survival curves and hazard ratios for the target Canadian population will be presented, and discrepancies between the transported and actual estimated effect will be explored using bias analysis.

CONCLUSIONS: This study addresses the utility and feasibility of applying transportability analyses for extending treatment effects across RWD when individual-level patient data is available for the target population. Bias analysis will be coupled with transportability analysis to address uncertainty in findings due to unmeasured variables or data elements that may be differently recorded between RWD.