Abstract
Duffy et al. 1 provide estimates of overdiagnosis (i) and lives saved (ii) based on data from the Two-County Trial. 2 Both might be debated.
Their model distinguishes the prevalence of overdiagnosed cancer (PO) from the prevalence of true cancer (PT). One equation of the model gives the total prevalence when screening starts in the intervention group: PO + PT = 0.0068 [A]. Another equation gives the total prevalence of diagnosed cancer in the control group at the end of the trial: PO + 1.35 PT = 0.0085 [B]. Factor 1.35 takes into account time trend and ageing from the beginning to the end of the trial.
Equations [A] and [B] are satisfied for PO amounting to 0.0019, that is 29% of the total prevalence at the beginning of the trial in the intervention group. This figure is apparently consistent with results from previous publications about overdiagnosis in Norway and in Sweden.3,4 Actually it is not: previously published estimates are restricted to the increase of the proportion of cancer overdiagnosis due to screening; overdiagnosed cancer that also occurs without screening is not included in these estimates.
Equation B is debatable. Without giving to PO another weighting factor than 1, the authors imply that PO is constant over age and over time. Suppose now that prevalence of overdiagnosed cancer (PO) is not constant over age neither over time as is, and for example that it is increased by a factor 1.2. Here is the corresponding equation: 1.2 PO + 1.35 PT = 0.0085 [C]. From equations [A] and [C], PO would amount to 0.0045, that is 67% of the total prevalence at the baseline screening examination.
The number of lives saved based on the Two-County Trial has also been debated. 2 Both points raised here might contribute to get a number of overdiagnosed cancer per life saved by screening closer to previously published estimates. 5