Issues possibly associated with misinterpreting survival data: A method study

Abstract

Objective: Proper interpretation of survival data of clinical cancer studies may be difficult and pitfalls related to the nature of Kaplan-Meier analyses might end up in mistaken inferences. The purpose of the present work is to raise awareness of those pitfalls and to prevent errors in future studies.

Study design and setting: While evaluating a randomized controlled trial, we came across some issues possibly associated with misinterpreting survival data. We thoroughly reviewed the reporting of survival analyses, statistical approaches, baseline characteristics, and choice of primary end point. The reported data were derived from people with high-risk neuroblastoma. Thus, the trial focused on survival. We reenacted survival functions by deducing the data of various treatment groups from pictured survival functions to estimate the concerning hazard ratios.

Results: Opposed to the reporting of the trial, we did not identify a significant difference between treatment groups with respect to overall survival. We were not able to appreciate an effective crossing of survival curves. With respect to event-free survival, we focused on comparable treatment groups and we did not identify a significant difference between treatment groups, thereby again opposing the reporting of the trial.

Conclusions: The present work exemplifies statistical issues that were apparently difficult to detect and that are possibly associated with misinterpreting survival functions. These issues include assumed crossing of survival curves, statistical approach changed in follow-up, different pretreatment between groups, and event-free survival used as primary outcome. Careful handling might prevent similar potential misinterpretation in future studies.

Keywords: Kaplan-Meier analysis; crossing of surival functions; neuroblastoma; randomized controlled trial.