Article type
Abstract
Background: Overdiagnosis is recognised as a common problem in cancer screening, but estimates of its frequency depend on reliable estimates of lead time. Studies of screening mammography that do not allow for lead time may be biased and overestimate overdiagnosis. The methodology for dealing with lead time is diverse, complex and challenging, especially for non-randomised studies. Currently, there is no agreed systematic method to assess the risk of lead-time bias in studies of overdiagnosis.
Objectives: We aim to describe criteria to assess the risk of lead-time bias in studies that estimate overdiagnosis due to screening mammography for breast cancer.
Methods: Evaluating lead-time bias requires consideration of 3 key elements: estimated mean lead time, the shape of the lead-time distributions and follow-up time after screening stops. We searched for literature around these issues and developed criteria to assess the risk of lead-time bias in studies that estimate overdiagnosis due to screening mammography.
Results: There are 2 study design characteristics that underpin our classification for risk of lead-time bias (Table 1). Where observed data were used to adjust for lead time, the risk of bias is judged on whether there was adequate follow-up to capture the tail of the distributions. Where there was a statistical adjustment, bias is judged on whether there was a sufficient mean lead time (based on directly observed data), or whether the model estimate of lead time allows for both progressive and non-progressive preclinical cancers and competing mortality (if directly observed data were not used). We judged lead time adjustment based on a comparison of cumulative incidence in a screened and unscreened population with 5+ years of follow-up to be at low or moderate risk of bias. Statistical correction or insufficient follow-up was judged to be at serious or critical risk of bias.
Conclusions: We have developed clear, transparent and mutually exclusive criteria for judging the risk of lead-time bias. These criteria may be applicable to other systematic reviews of cancer screening programmes.
Objectives: We aim to describe criteria to assess the risk of lead-time bias in studies that estimate overdiagnosis due to screening mammography for breast cancer.
Methods: Evaluating lead-time bias requires consideration of 3 key elements: estimated mean lead time, the shape of the lead-time distributions and follow-up time after screening stops. We searched for literature around these issues and developed criteria to assess the risk of lead-time bias in studies that estimate overdiagnosis due to screening mammography.
Results: There are 2 study design characteristics that underpin our classification for risk of lead-time bias (Table 1). Where observed data were used to adjust for lead time, the risk of bias is judged on whether there was adequate follow-up to capture the tail of the distributions. Where there was a statistical adjustment, bias is judged on whether there was a sufficient mean lead time (based on directly observed data), or whether the model estimate of lead time allows for both progressive and non-progressive preclinical cancers and competing mortality (if directly observed data were not used). We judged lead time adjustment based on a comparison of cumulative incidence in a screened and unscreened population with 5+ years of follow-up to be at low or moderate risk of bias. Statistical correction or insufficient follow-up was judged to be at serious or critical risk of bias.
Conclusions: We have developed clear, transparent and mutually exclusive criteria for judging the risk of lead-time bias. These criteria may be applicable to other systematic reviews of cancer screening programmes.