Screening details, relative risks and confidence intervals for included studies by study design.

NR = Not reported.
A. Most studies reported mortality as incidence-based breast cancer mortality. These studies report both incidence and prevalence-based breast cancer mortality. Numbers in cells represent incidence-based mortality.
B. Studies report only prevalence-based breast cancer mortality. Numbers shown reflect prevalence-based mortality numbers and RR calculations.
C. For the Otto,2003 study, mammograms were 2-view in the first round and were 1-view in subsequent rounds. For the No Author, 1999 study, type of mammogram varied by screening center. Women in the screened group also received clinical breast exam during non-screened year.
D. Study authors reported all mortality and population numbers by age categories. We excluded women screened ages 40–49 because of the overlap with the Hellquist paper. We then separated the data for women screened 50–69 and 70–74 in order to include in meta-analyses for each age category.
E. RR and CI reported in this table are only for the comparison of Gavleborg to all of Sweden.
F. RR and CI reported in this table are only for the comparison of Turku to Helsinki. The comparison of Turku to Tampere was not included because of overlap and because Tampere screened women 55–59.
G. RR slightly different as RR reported in original article was adjusted for pre-trial rates, which we did not do.
H. Study only reports aggregate population data. No individual cases are identified or linked from medical records or cancer registries.
I. The specific year that screening began varied by county or region.
J. Screening expanded to women age 70 in 2002.
K. We only analyzed women in cohort 1.