OBJECTIVE: To propose principles of design and measures of effect for cancer screening as a public health policy. MATERIAL: Finnish routine screening programme with mammography. DESIGN: Evaluation of mortality from breast cancer by before-after time trends among Finnish women, by geographical (by municipalities) comparisons, and by intention to screen at an individual level with individual controls who were cluster randomised and matched for age. OUTCOME MEASURES: Standardised mortality ratios (SMRs) from breast cancer (total), SMRs from breast cancers diagnosed during the screening programme (refined), deaths prevented, prolongation of life for each breast cancer detected, for each death prevented, for each compliant woman, for each screen, and for each invitation. Relative prolongation of life--that is, time gained versus time spent. RESULTS: SMRs only at an individual level with deaths from cancers diagnosed during the screening programme and individually selected controls showed 24% protection, whereas the other SMR measures were too crude or biased because of dilution and selection. Prolongation of life varied from 15 years for each death prevented to two days for each woman screened, with approximation for the prolongation relative to time spent of 3 to 1. CONCLUSIONS: A public health policy should be introduced gradually. Those not covered immediately by the policy serve as controls and they should be randomly allocated. The most relevant outcome measure is prolongation of life, and for public health purposes it should be given per unit of intervention, such as screen or invitation, and also related to the time spent for the intervention. Such gains are often small in a Western society, which implies that medicine, including research, should focus more on other aspects of life than length.
OBJECTIVE: To propose principles of design and measures of effect for cancer screening as a public health policy. MATERIAL: Finnish routine screening programme with mammography. DESIGN: Evaluation of mortality from breast cancer by before-after time trends among Finnish women, by geographical (by municipalities) comparisons, and by intention to screen at an individual level with individual controls who were cluster randomised and matched for age. OUTCOME MEASURES: Standardised mortality ratios (SMRs) from breast cancer (total), SMRs from breast cancers diagnosed during the screening programme (refined), deaths prevented, prolongation of life for each breast cancer detected, for each death prevented, for each compliant woman, for each screen, and for each invitation. Relative prolongation of life--that is, time gained versus time spent. RESULTS: SMRs only at an individual level with deaths from cancers diagnosed during the screening programme and individually selected controls showed 24% protection, whereas the other SMR measures were too crude or biased because of dilution and selection. Prolongation of life varied from 15 years for each death prevented to two days for each woman screened, with approximation for the prolongation relative to time spent of 3 to 1. CONCLUSIONS: A public health policy should be introduced gradually. Those not covered immediately by the policy serve as controls and they should be randomly allocated. The most relevant outcome measure is prolongation of life, and for public health purposes it should be given per unit of intervention, such as screen or invitation, and also related to the time spent for the intervention. Such gains are often small in a Western society, which implies that medicine, including research, should focus more on other aspects of life than length.