Estimating lead time and overdiagnosis in cancer screening programmes: the MOCCI method.

Abstract

Background: Cancer screening enables earlier cancer diagnosis and treatment. The interval between cancer diagnosis by screening and the symptomatic detection in the absence of screening is termed lead time. If a patient's tumour, detected by screening, would never surface clinically, the patient is considered overdiagnosed. Estimating these quantities is crucial for evaluating cancer screening programmes.

Development: We developed MOCCI (Minimizing Observed and Counterfactual Cancer Incidence), a novel parametric method for estimating the lead time distribution. MOCCI compares age- and calendar-stratified cancer incidence between individuals invited to screening and those not invited and estimates the lead time distribution that minimizes the incidence difference between the two groups. The probability of overdiagnosis is then estimated by comparing the model-predicted lead time with the time to death from other causes.

Application: In an application to the Slovenian breast cancer screening programme, we estimated that 30% (95% confidence interval [CI], 20% to 39%) of screen-detected cases were non-progressive and 33% (95% CI, 25% to 43%) were overdiagnosed; assuming an exponential lead time distribution for progressive cancers, the mean lead time was 1.8 years (95% CI, 1.3 to 2.9).

Conclusions: This study proposes a new method for estimating lead time and overdiagnosis. The proposed method (a) can accommodate various lead time distributions, (b) yields a lead time distribution that is aligned with the observed excess incidence arising from screening, and (c) enables the separation of different sources of overdiagnosis. The method provides stable estimates when sample sizes are large and the assumed lead time model is simple.