The research question was clear and inclusion criteria were stated. Both published and unpublished sources were sought and papers in all languages were eligible for inclusion, minimising the risk of publication and language bias. Two reviewers extracted data, reducing the possibility of reviewer error and bias, but only one reviewer selected studies.
It did not appear that that the quality of the primary data was systematically assessed, though aspects, such as attendance at screening programmes, were discussed in the text. Also, there was no information reported about the sources of the data, such as whether data were from national statistics or whether any sampling within surveys was undertaken. The reliability of the primary data sources is unclear.
Rate ratios of the observed and expected cases were based on regression models, so they were calculated from two sets of estimated data. The authors did not state whether the assumptions underlying linear regression modelling were met, or how reliable the results of these models were. Given the nature of the data, other models such as time series analysis may have been more appropriate.
The lack of information about the primary sources of data, or consideration of their quality, and the fact that the increases in cancer were all calculated from estimates from statistical modelling, warrant cautious interpretation of the conclusions. The conclusion that the increase in incidence of breast cancer followed screening was supported by the data presented, but it is unclear whether the conclusion about one in three cancers being overdiagnosed is reliable.