Background The debate regarding the lower age limit for mammography service screening is old and lively; a product in part of the lower breast cancer risk in younger ages as well as the limited data available for studies of the younger age group. Recently the idea of inviting only high risk groups has gained momentum, however high risk might not be equivalent to greater benefit from screening. Therefore, there is a need for information on effectiveness of screening as it relates to young women and to specific risk groups. To this end, this thesis evaluates mammography screening for the age group – 40 to 49 year old women – in terms of breast cancer mortality reduction in total and in subgroups based on breast cancer risk factors. Overdiagnosis of mammography screening is also evaluated for women 40 to 49 years old. In addition, this thesis presents a statistical method to estimate this effectiveness and to test for differences in effectiveness between subgroups adjusted for non-compliance and contamination. Methods The studies of this thesis are based on data from the Screening of Young Women (SCRY) database. The SCRY database consists of detailed information on diagnosis, death, screening exposure and risk factors for breast cancer cases and population size by year (between 1986 and 2005) and municipality for women in Sweden between 40 and 49 years old. The material was divided into a study group consisting of the counties that invited women in the age group 40-49 years to mammography screening, and a contemporaneous control group consisting of the counties that did not. Effectiveness was estimated in terms of rate ratios for two different exposures (invitation to and participation in screening), and overdiagnosis for subsequent screening was estimated adjusting for lead time bias. Defining a reference period enabled adjustment for possible underlying differences in breast cancer mortality and incidence. A statistical model for adjusting for non-compliance and contamination in randomised controlled trials was further developed to allow for adjustment in cohort studies using a Poisson model with log-linear structure for exposure and background risk. Results During the study period (1986-2005), there were 619 and 1205 breast cancer deaths and 6047 and 7790 breast cancer cases in the study group and the control groups, respectively. For women between 40 and 49 years old, the breast cancer mortality reduction was estimated at 26% [95% CI, 17 to 34%] for invited to screening and 29% [95% CI, 20 to 38%] for attending screening. The RR estimates for the high-risk groups based on the risk factors parity, age at birth of first child, and socio-economic status were equal to or higher than that of the low risk groups. The new statistical method showed that the decrease in effectiveness with parity was not a statistically significant trend. The overdiagnosis from subsequent screening for 40 to 49 year old women was estimated at 1% [95 % CI, -6 to 8 %] (i.e., not statistically significant). Conclusion Subgroup specific effectiveness was also estimated. The relative effectiveness of screening for breast cancer with mammography for women age 40 to 49 years appears to be comparable to that for older women. These findings and the fact that there was no statistically significant overdiagnosis from subsequent screening speak for inviting women 40 to 49 years old to screening. High-risk screening for nulliparous women aged 40 to 49 years, for example, might be an alternative in countries where population-based screening for all women between 40 and 49 years old is not possible. However, the matter of risk factors and the effect of their combinations is complex and risk group screening presents ethical and practical difficulties. The new statistical model is a useful tool for analysing cohorts with exposed and non-exposed populations where non-compliance and contamination is a potential source of bias.