Analysis of the relationship between long-term operative success and a transient or delayed operative side effect.

Abstract

Investigators sometimes correlate the success of an operative procedure with the presence of an operative side effect that is thought to reflect the efficacy of the procedure in some way. When the prevalence of the side effect increases or decreases with time, this type of analysis can introduce a systematic bias. Monte Carlo-simulated patient cohorts with independently and randomly generated operative success and presence of side effects were generated. Kaplan-Meier analyses of operative success were stratified by the presence of side effects at various times after the "operation." When the side effect (such as numbness after an ablative procedure) resolved during the study period in a significant proportion of patients, stratification by presence of the side effect at the end of the study period introduced a serious bias in favor of apparent better pain relief in patients without numbness. When the prevalence of the side effect increased with time (such as the requirement for a subsequent operation for radiation necrosis after brachytherapy or radiosurgical treatment of a glioma), the analysis was biased in favor of apparent longer survival for patients who experienced the side effect (i.e., those who required subsequent operations). Analyses of operative success stratified by the presence of postoperative side effects with time-varying prevalence may be seriously biased and should be interpreted with caution.