No screening was implicitly used as the comparator for the screening strategy. This appears to have been a valid comparison. Different patient ages at diagnosis were considered.
The clinical data were mainly from a large, multicentre, randomised controlled trial with a relatively long follow-up of a median of nine years. This should ensure high internal validity and accurate data for the impact of screening on mortality and the screening accuracy. More details on this study would have been useful to fully assess its quality. The authors did not discuss screening compliance, which is important in the real world and might have been different in a clinical trial. Other sources of data were US life tables, which were appropriate. Life expectancy was the main outcome measure and this will allow comparisons with other diseases. The authors stated that quality-adjusted life-years would have been more appropriate, but there were possible limitations in the estimation of the quality-of-life weights.
The authors stated that a societal perspective was adopted, but only the direct medical costs appear to have been considered. They acknowledged that the inclusion of productivity losses and other non-medical costs would have been appropriate. A list of cost items was not given and the unit costs were not presented separately from the resource quantities. The data sources were mentioned, but were not described; they appear to have been standard US sources. The price year was not explicitly stated. These issues limit the transparency of the economic analysis.
Analysis and results:
Only the incremental findings were reported, but they were given in detail. The uncertainty was partly investigated in an analysis that focused on selected parameters and their impact on the costs. The model was not fully described. The authors compared their results with those of other published studies and underlined their limitations. The analysis appears to have been USA-specific and not easily transferable to other settings.
The cost-effectiveness framework was conventional, but some aspects of the analysis were not extensively presented. The authors’ conclusions appear to be valid.