Abstract
OBJECTIVES:To estimate the effect of the population based service screening programme in Sweden on mortality from breast cancer among women aged 50–69. SETTING:In 1986, population based service screening with mammography started in Sweden, and by 1997 screening had been introduced in all counties. Half of the counties invite women from 40 years of age whereas women 50 and older are invited in the other counties. The upper age limit was either 69 or 74. Women in the age group 50–69 years are thus invited to screening in all counties. METHODS:The counties which started with mammographic screening in 1986–87 constituted the study group and were compared with the counties which started in 1993 or later. In 1987 the mean number of women aged 50–69 was 161 986 and 98 608 in the study and control groups, respectively. Refined excess mortality (smoothed with the Lowess method) from breast cancer and refined cause specific mortality from breast cancer were used as effect measures. To adjust for geographical differences in mortality from breast cancer a reference period was used. Allowance was made for two potential biases: (a) inclusion bias implying the inclusion of cases diagnosed before invitation to screening in the first screening round, and (b) lead time bias. RESULTS:After a mean follow up time of 10.6 years since the start of screening and a mean individual follow up time of 8.4 years, a non-significant reduction in refined excess mortality for breast cancer was estimated as relative risk (RR) 0.84 (95% confidence interval (95% CI) 0.67 to 1.05). After adjustment for inclusion and lead time biases the RR was 0.80 (20% reduction). Only 27% of the deaths from breast cancer in the total mortality for women aged 50–79 at death consisted of women aged 50–69 at diagnosis who were diagnosed after the start of screening. This figure has important implications for judgement of the impact of screening on age specific national breast cancer mortalities. CONCLUSIONS:A non-significant reduction in mortality from breast cancer was found in counties performing service screening with mammography in Sweden. Adjustment for possible biases changed the result towards a larger effect of screening. The results do not contradict the effects found in the Swedish randomised mammography trials.