BIAS DUE TO OBSERVATION OF DIFFERENT PARTS OF A NONCONSTANT HAZARD CURVE OVER TIME

Objectives, Uncontrolled data often have to be used in clinical decisi on making and in the planning of clinical trials. When such data are u sed as a basis for comparing different treatment strategies, they tend to generate sources of bias such as inconsistent patient selection, m isrepresentation, and measurement errors. A rational usage of uncontro lled data requires identification and handling of different systematic errors when comparing different treatment strategies. Therefore we so ught to define a systematic error that deserves more attention in the clinical literature. Methods. Using hypothetical examples, we illustra te the bias introduced by variations in ''lead time,'' proportions of so-called prevalent/incident cases, and lengths of follow-up between r eference and treatment groups. Results, We found these three situation s conceptually identical. Bias is introduced because different parts o f a nonconstant hazard curve over time are observed. Conclusions, The systematic error described is potentially important in uncontrolled da ta provided the hazard rate over time is nonconstant. Thus the seeming ly worse outcome after external beam radiation therapy as compared to the outcomes of deferred treatment or radical prostatectomy in compile d patient series of localized prostate cancer may in some part be expl ained by bias due to observation of different parts of a nonconstant h azard curve over time.