Lobectomy versus stereotactic body radiotherapy for stage I non-small cell lung cancer: Post hoc analysis dressed up as level-1 evidence? - Part 3 of 9

This expert opinion editorial is published in JTCVS in response to the article by Chang and colleagues. Meyers and colleagues begin by emphasizing that a large randomized controlled trial (RCT) would be the best way to answer the question of whether SABR or lobectomy is superior treatment for patients with early stage NSCLC. This editorial highlights the limitations of analyzing and publishing data from failed trials and demonstrates how this contributes little to the ongoing debate. First, neither the STARS trial nor the ROSEL trial achieved 4% of the target sample size which leads to inaccurate results especially since the outcome of interest, death, occurred in an exceedingly small number of patients. Second, this cohort of patients differs significantly from the full cohort, had accrual been successful; early planned analyses in RCTs (when accrual is low) are conservative to prevent false positive results (Type I error). Third, the mortality rate in the surgical arm is 2-4 times the expected mortality for patients with stage I NSCLC undergoing resection. Fourth, the low accrual limits the external validity of these data. Fifth, the fact that there was one treatment-related death in the surgical arm, none in the SABR arm, and no difference with regard to recurrence does not seem to add up to the original article's claim that there is a statistically significant difference in survival favoring SABR. Sixth, the degree of surgical risk for these patients was unclear and it is difficult to account for variation in surgical treatment at different centers. Seventh, in Figure 2, A (overall survival) the hazard ratio is 0.14 in favor of SABR with a confidence interval that includes 1.0 (insignificant), but the p-value listed is 0.037 (significant). Lastly, the authors point out that equipoise existed prior to these RCTs and is still very much present as this controversy continues.