Evidence of overdiagnosis

Sandblom and colleagues’ paper compared the case fatality rate of prostate cancer between men randomly assigned to screening and those who were not invited for screening.1 A more reliable outcome of such a randomised controlled trial is the difference of mortality as a result of prostate cancer between intervention and control groups. Of 1494 men invited for screening, 30 (2%) died from prostate cancer. Of 7532 men not invited, 130 (1.7%) died from prostate cancer. These proportions are not given in the paper. The outcome in the invited group is not significantly worse than in the control group, but it provides a negative answer to the question about screening efficacy for reducing prostate cancer mortality.

Survival analysis limited to diagnosed cases is biased by overdiagnosis of clinically insignificant prostate cancer.2 In table 1 of Sandblom and colleagues’ paper we found a significant 1.8% (95% confidence interval 0.6% to 3.0%) (χ²=10.23; P=0.0014) absolute increase in prostate cancer as a result of overdiagnosis in the invited group. Figures 2 and 3 show an illusionary improved prognosis among cases in the intervention versus control group because of an excess of 27 overdiagnoses among the 85 cases in the intervention group.1 Both case fatality rate and survival are overdiagnosis dependent: patients who have been overdiagnosed cannot die from prostate cancer. Evidence based medicine must be independent of misleading influence and use well recognised principles, such as analysis by intention to treat.3