Methodology for estimation of progression-free survival in patients with Non-Small Cell Lung Cancer (NSCLC) using Real-World Data from a French cohort

RECIST is the gold standard for evaluating tumor response, but in clinical practice, patient assessments are not always precisely timed or uniformly documented, making it challenging to use real-world data (RWD) to estimate disease progression. This study aims to present the methodology for assessing the feasibility of estimating Progression-Free Survival (PFS) in NSCLC patients using RWD, highlighting its strengths and limitations.

 

This study analyzed data from 271 patients diagnosed with NSCLC (CIAN-Lung Cohort) in France, from 2017 to 2023. Inclusion criteria were patients with ECOG status at baseline, those receiving at least one pharmacotherapy after Stage IV diagnosis, and those with documented treatment type and line, followed for at least three months. The index date was defined as the initiation of the first treatment after Stage IV diagnosis. Data were manually extracted from electronic medical records, inpatient hospitalization records, pharmacy records, Picture Archiving and Communication System (PACS), Radiology Information System (RIS), and Laboratory Information System (LIS). PFS was defined as the time from index date until disease progression or censoring. Progression events included: disease progression relapse, switch to the next line of therapy, or death. Censoring events included: secondary malignancy diagnosis, lost to follow-up, or dataset cutoff. Only variables with > 70% completion rate were retained.

 

Progression events defined by treatment changes were simpler to implement, as they relied on structured data. However, the variability in the timing of patient assessments often led to discrepancies in recording progression events, potentially affecting the perceived efficacy of treatments and the accuracy of PFS estimation. Although manual data extraction offered flexibility in meeting research criteria, it was time-consuming and required highly trained abstractors. The reliance on human interpretation leaves room for potential bias and variability, emphasizing the need for more standardized documentation practices in clinical environments to improve RWD quality.

 

The CIAN-Lung cohort offers valuable RWD for estimating PFS in NSCLC patients. Future work will explore using natural language processing to optimize data extraction from clinical notes and radiology reports, as well as including radiomics features to enhance PFS analysis.