Analytical approach:
A life table approach was used to model two cohorts of healthy 50 year-old women for 35 years. Each cohort comprised 364,500 women, which was the population of 50 year-old women in England and Wales in 2009 who were eligible for screening. The perspective was not stated explicitly but appeared to be UK NHS perspective.
Effectiveness data:
The primary effectiveness data were relative risks. Relative risks parameters in the model included: relative risk of mortality associated with regular screening; relative risk of death from non-breast cancer causes after breast cancer diagnosis; and relative over-diagnosis of breast cancer related to screening. These were derived from published literature and an independent review of the benefits and harms of breast cancer screening (see Other Publications of Related Interest). Baseline age-specific breast cancer incidence was generated via logistic regression.
Three assumptions were made: incidence of breast cancer in screened women was higher; diagnosis occurred five years earlier between age 50 and 69; and after cessation of screening there was a 10% reduction in incidence.
Monetary benefit and utility valuations:
Utility scores were generated by taking general population utility at age 50 and adjusting for aging and disutility due to breast cancer diagnosis.
Measure of benefit:
The primary measure of benefit was quality-adjusted life years (QALYs). Other outcomes evaluated breast cancer deaths, person years of survival and person years of survival after breast cancer diagnosis. Future survival related benefits were discounted 3.5% annually.
Cost data:
The estimated cost of the screening programme was derived from a published estimate from the NHS screening programme. The overall cost figure reported was divided by 20 to represent years of screening. Costs for treating primary and metastatic breast cancer were derived from NHS reference costs. Costs were calculated for treating clinically detected cancer and cancer detected through screening, with an assumption that the treatment of clinically detected cancers was more expensive. Costs for treating over-diagnosed patients in the screened cohort were included.
Analysis of uncertainty:
The authors conducted scenario and probabilistic sensitivity analyses. The scenarios varied in how many years earlier breast cancer was diagnosed and whether the reduction of incidence after screening was 10% or 20%. Each scenario, including the base case (five-year advance screening, 10% reduction in incidence) were subjected to 5,000 probabilistic simulations where variables were subjected to costs using mostly custom distributions to define parameter uncertainty.