Feasibility study of adjustment for contamination and non‐compliance in a prostate cancer screening trial

Abstract

The use of PSA as a screening test has become increasingly prevalent in the general population and therefore also in the control arm of the European Randomized study of Screening for Prostate Cancer (ERSPC). We present a feasibility study and impact simulation of a secondary analysis, which imitates a situation where all participants in the study are managed according to their random assignment. The results of the Rotterdam section of the ERSPC were adjusted for contamination and non-compliance according to Cuzick et al. [Stat Med 1997; 16:1017-1029]. Endpoints of this analysis were simulated reductions in prostate cancer mortality. Of the men allocated to the screen arm, 27.1% were non-compliant. In the control arm 30.7% had their PSA-level measured by a general practitioner (GP) (i.e., contamination). For a scenario in which the intention-to-screen analysis was assumed to give a decrease in the mortality in the men randomized to screening of 6.7%, the secondary analysis resulted in a decrease of 16.1% for those actually screened. Although the definition of contamination as "PSA ever tested" gives an indication of the proportion of contamination, it will be important to differentiate the screening use of PSA from its diagnostic use. For the rest, adjustment for non-compliance and contamination was shown to be feasible in this prostate cancer screening trial. It can therefore be used to carry out a secondary analysis on the definitive outcome of the ERSPC and will provide accurate information for those men who are in fact screened.