Sample size considerations for the evaluation of prognostic factors in survival analysis.

摘要: When the role of a new prognostic factor is investigated, careful planning an appropriate study required. This includes assessment power in terms sample sizes. An adequate analysis independent effect has to be adjusted for existing standard factors. With survival time as endpoint this will usually done with Cox proportional hazards model. Sample size and formulae have been developed by Schoenfeld randomized treatment comparisons. In factors covariates included are expected correlated primary interest. situation, not valid may applied. paper, Schoenfeld's formula first extended situation where analysis. The validity resulting approximate asymptotic investigated its behaviour numerical integration finite simulation. Second, provided detect interaction between interesting second factor. extends effects. Finally, approach illustrated example on impact DNA ploidy other advanced ovarian cancer.