Statistical strategies to analyze local control after radiotherapy.

Abstract

Historically, local progression, an outcome of effectiveness in curative radiotherapy, has not always been statistically analyzed with the same method. This article aims to elucidate the validity and relative usefulness of statistical strategies to analyze local control. In an application example, data were analyzed from a phase III trial comparing different radiation schedules for glottic cancer. The superiority of experimental treatment over standard treatment in local control was detected by a competing risks model but not detected in the analysis of local progression-free survival by a standard survival model. The use of a composite outcome, local progression-free survival, is suboptimal in that local progression and death are equally treated as the events of interest. In the analysis of local progression-free rate, by censoring death, the cumulative incidence of local progression is overestimated. We recommend the consistent use of the cumulative incidence of local progression as the endpoint to assess local control after radiotherapy.