Practical demonstration of time bias with administration of adjuvant therapy in lung cancer

Abstract

Purpose/BackgroundImmortal time bias (ITB) can hinder appropriate interpretations of studies administering adjuvant therapies. Given the increase in National Cancer Data Base (NCDB) analyses evaluating postoperative radiation therapy (PORT) as an adjuvant therapy, we sought to practically demonstrate the effects of ITB by performing a series of simulated NCDB analyses. MethodsA simulated NCDB analysis was performed to examine how the reported benefit of PORT in stage III non-small cell lung cancer (NSCLC) may change with adjustment for ITB utilizing sequential land mark analysis (SLMA) and time dependent Cox (TDC) modeling. ResultsOn the simulation analysis of 6440 NSCLC patients, we found that the omission of PORT without ITB adjustment was associated with an increased risk of death (HR 1.17, p < 0.0001). After performing a sequential LMA, the detrmient of omitting PORT continued to decrease until it was no longer significant at 8 months, HR 1.05 (p = 0.09). With the TDC model, although still significant, the relative benefit of PORT decreased, to a HR of 1.07 (p = 0.02). ConclusionsImmortal time bias can alter the results of survival analyses if not carefully accounted for. Adjusting for this bias is essential for accurate data interpretation and to better quantify the impact and effect size of adjuvant therapies such as PORT.