Test sensitivity in the European prostate cancer screening trial: results from Finland, Sweden, and the Netherlands

Abstract

Test sensitivity pertains to the ability of a test to identify subjects with the target disorder. In cancer screening, test sensitivity can be estimated using interval cancer incidence as an indicator of false-negative result. A randomized trial provides the optimal approach for estimating test sensitivity, as the control arm provides the expected rates. We estimated the sensitivity of the prostate-specific antigen test using incidence method, i.e., based on incidence of interval cancer among subjects with negative screening results, compared with that in the control arm. Data from three centers in the European randomized screening trial were used to estimate interval cancer incidence (I(I)) among 39,389 men with negative screening tests. This was compared with incidence among the 79,525 men in the control arm of the trial (I(c)) to estimate test sensitivity (S = 1 - I(I) / I(C)). Confidence intervals were calculated using simulations, assuming that the number of cases follows a Poisson distribution. The estimated test sensitivity following the first screen was 0.87 (0.83-0.92) in Finland, 0.87 (0.62-1.00) in Sweden, and 0.93 (95% confidence interval, 0.90-0.96) in the Netherlands. There was some indication of a higher test sensitivity for aggressive cancers (0.85-0.98 for non-organ-confined cases or Gleason 8-10) and for the second screening round (approximately 0.85-0.95). Test sensitivity varied to some extent between the three centers in the European trial, probably reflecting variation in screening protocols, but was acceptable in the first screening round, and may be better for aggressive cancers and in the second screening round.