Excess all-cause mortality in the evaluation of a screening trial to account for selective participation

Abstract

Objective In addition to disease-specific mortality, a randomized controlled cancer screening trial may be evaluated in terms of excess mortality, in which case no patient-specific information on causes of death is needed. We studied the effect of not accounting for attendance on the calculated excess mortality in a prostate cancer screening trial. Methods The numerator of the excess mortality rate related to prostate cancer diagnoses in each study arm equals the excess number of deaths observed in the cancer patients. The estimation of the expected number of deaths in the absence of the prostate cancer diagnoses has to account for the self-selection of those participating in the trial, particularly if the proportion of non-participants is substantial. Setting The European prostate cancer screening trial (ERSPC). Results In the screening arm, non-attendees had roughly twice the mortality rate of attendees. Approximately twice as many cancers were detected in the screening arm compared with the control arm, primarily in attendees. Unless attendance is properly accounted for, the expected mortality of prostate cancer patients in the screening arm is overestimated by 0.9–3.6 deaths per 1000 person-years. Conclusions Attendees have a lower all-cause mortality rate (are healthier) and a higher probability of a prostate cancer diagnosis than non-attendees and the men randomized to the control arm. If attendance is not accounted for, the excess mortality and the between-arm excess mortality rate ratio are underestimated and screening is considered more effective than it actually is. These effects may be sizeable, notably if non-attendance is common. Correcting for attendance status is important in the calculation of the excess mortality rate in prostate cancer patients that can be used in conjunction with a disease-specific mortality analysis in a randomized controlled cancer screening trial.