Objectives The prevailing theory is that mammography on average detects tumors 3.3–7 years earlier, based on models and mean lead time (MLT) estimates not accounting for non-progressive tumors and changes in cancer awareness. We estimate MLT and maximum lead time accounting for these factors.

Method Design: Proportions of tumors with lead times over 1 (age 40–49) and 2 (age 50–69) years are calculated by comparing incidence rates in the first and second screening rounds using relative risk (RR).

Setting and participants: Cohort data were collected from the cancer registries in Denmark, Norway and Sweden. Data were collected from the period when public screening was introduced.

Primary and secondary outcome Incidence/detection rates and interval cancer rates of invasive breast cancer and DCIS.

Results First, MLT, detection- and interval cancer rates for 127,064 (Norwegian) and 517,725 (Danish) women aged 50–69 actually attending a first-time public screening were calculated: 11% (95% CI 3–18%)(Norway) tumors and 25% (95% CI 22–28%)(Denmark) had lead times over 2 years. In Norway, MLT=1.02. In Denmark, MLT=1.21. Second, in cohort studies of 375,064 Swedish (age 40–74), 151,601 Norwegian (age 50–69), and 149,266 Danish women (age 50–69) invited to their first mammography screen in 1986–89, 1996–97, and 2008–9, respectively, we calculate RR=1.33 (95% CI 1.21–1.47) for women aged 40–49 years during the first 5 years of annual screening. There was no prevalence peak; thus, maximum lead time is 1 year for these women. For biennially screened women aged 50–69, RR=1.54 (95% CI 1.44–1.65)(Norway), RR=1.52 (95% CI 1.43–1.60)(Sweden) and RR=1.56 (95% CI 1.45–1.65)(Denmark); prevalence peaks were small, meaning few tumors had lead times over 2 years.

Conclusions Constant high RRs means that many screen-detected tumors were non-progressive. Changes in cancer awareness and many non-progressive cancers influence MLT estimates dramatically. Adjusting for long MLT when calculating overdiagnosis is not justified.