Abstract 208P
Background
Understanding the causes of the discrepancy between clinical trial and real-world outcomes could potentially improve the design and ultimately the outcomes of future clinical trials.
Methods
Stage IV SCLC patients that received platinum etoposide were analysed from data collected by Karolinska University Hospital (Stockholm, Sweden), and open data (Project Data Sphere) of control arms from three phase III and three phase II RCTs (n=1,224). The covariates were age, sex, brain metastasis, performance status (PS), progression free survival, overall survival, and eventual censoring. For the real-world cohort, the 8th version of the TNM staging was available. The real-world patients were compared with the individual and aggregated RCTs to estimate any differences. Kaplan Meier curves and hazard ratios (HR) were computed. The datasets were adjusted to measure the impact of accounting for study differences. The adjustments included variable stratification, exclusion of censoring, and comparison of TNM stages with the clinical trials. A refined analysis was carried out for the censoring (patient inclusion using propensity score), and IVA patients (oversampling).
Results
The real-world cohort exhibited an age distribution skewed towards older age, more balanced sex ratio, and higher number of patients with PS ≥2. A significant difference was observed in survival outcomes (HR: 0.65 [0.55-0.75]). This gap was leveraged when patients were stratified according to PS. Censoring was mostly applied in trials in the key survival range (0-300 days), thus overestimating the survival of these patients. Propensity score censoring correction adjusted the variable selection and closed this gap (HR: 1.1 [0.87, 1.3]). The clinical trial outcomes matched those of IVA real-world patients (HR: 0.96 [0.81, 1.1]). This may be due to selection bias for RCTs, favouring younger and fitter patients, with probably lower disease burden, which can be only partially captured by the TNM staging.
Conclusions
The results showed that a retrospective real-world study could highlight the discrepancies in outcomes compared to RCTs and can help improve future study designs.
Legal entity responsible for the study
The authors.
Funding
The Swedish Cancer Society, Stockholm Cancer Society.
Disclosure
All authors have declared no conflicts of interest.