The Gothenburg Breast Screening Trial

BACKGROUND: Although there is evidence for a reduction in breast carcinoma mortality with mammographic screening, some doubts have been expressed, and there is still uncertainty regarding the age specific effects. METHODS: The authors report on a randomized, controlled trial of mammographic screening for breast carcinoma that was conducted among 51,611 women (21,650 women who were invited to a screening the study group and 29,961 women in a control group) ages 39-59 years in Gothenburg, Sweden. Among women in the study group, the screening interval was 18 months. The screening phase of the trial took place in 1982-1991, and follow-up for breast carcinoma mortality continued until December 31, 1996. Mortality from breast carcinoma was analyzed using a Poisson regression model. Overall and age specific effects of invitation to mammography screening on breast carcinoma mortality were calculated. Three mortality effects were estimated: the effect on deaths from breast tumors diagnosed during the screening phase of the trial, as assessed by an independent Endpoint Committee (the EPC evaluation model); the effect on deaths from breast carcinoma diagnosed during the screening phase of the trial, as determined by data from the National Cancer Registry and the National Cause of Death Register (the SCB evaluation model); and the effect on deaths from all breast carcinomas diagnosed up to December 31, 1996, as determined by the National Cancer Registry and the National Cause of Death Register (the SCB follow-up model). RESULTS: A nonsignificant, 21% reduction in the rate of mortality from breast carcinoma with invitation to screening was observed using the EPC evaluation model (relative risk RR, 0.79; 95% confidence interval 95% CI, 0.58-1.08; P = 0.14); and a borderline significant, 23% rate reduction was observed using the SCB follow-up model (RR, 0.77; 95% CI, 0.60-1.00; P = 0.05). Age specific analyses yielded greater mortality rate reductions for the groups of women ages 39-44 years, 45-49 years, and 55-59 years, but there was no mortality rate reduction in the group of women ages 50-54 years. The effects of invitation to mammographic screening on the incidence of lymph node-positive disease closely paralleled the effects of invitation on breast carcinoma mortality. The effect on breast carcinoma mortality was consistent with the effect on all-cause mortality, suggesting no bias in classification of cause of death. Breast carcinoma incidence in the study group was almost identical to the incidence in the control group after trial by screening had ended in the control group (RR, 0.98; 95% CI, 0.88-1.09; P = 0.7). CONCLUSIONS: The current results support the commonly observed 20-30% reduction in breast carcinoma mortality with invitation to screening. The impression that screening is less effective in women younger than 50 years may be an oversimplification. Age specific effects should be a target for further research.