Should major surgery be offered to elderly patients with early-stage lung cancer? An emulated trial using observational data in England
Session type: Proffered paper sessions
Background: Older cancer patients often have suboptimal cancer treatment and poorer cancer outcomes than younger patients: early-stage non-small cell lung cancer (NSCLC) patients diagnosed after 70 years of age in England show up to 60% reduced probabilities of receiving major surgery compared to younger patients. As older cancer patients are generally excluded from clinical trials, the evidence supporting aggressive cancer management is scarce and has to rely on non-randomised studies that are more likely to be prone to bias. For instance, when using observational data to estimate the causal effect of surgery on survival, immortal-time bias is an issue because of the waiting-time between diagnosis and surgery.
Method: To measure the causal effect of receiving a major surgery for lung cancer patients aged over 70 years at diagnosis on 1-year survival probability, we emulated a randomised trial using population-based cancer registry data in England, linked to Hospital Episode Statistics and Lung Cancer Audit Data.
Results: The inclusion criteria led to the selection of 2309 patients. The intervention arm was surgery within six months following diagnosis compared to no surgery within six months. At diagnosis, patients were cloned and entered both arms. Observations were censored when they deviated from the protocol, and this dependent censoring was accounted for using inverse-probability-of-censoring weights when estimating survival probabilities. There were 83% (95% CI: 81-85) and 71% (95% CI: 68-74) of treated and non-treated patients who survived the first year after diagnosis, respectively. This corresponded to 14 days of life saved in the first year after diagnosis.
Conclusion: Our results suggest a strong benefit on survival of surgery among older patients with early-stage NSCLC. We used an innovative study design, enabling us to control for confounding and immortal-time bias, and allowing a transparent reporting of the methods.