We thank Dr. Forsyth for his interest in our article1 and for addressing the important issue of cost versus benefit of mammography screening. There are a few points in Dr. Forsyth's letter that need to be addressed.
Dr. Forsyth claims, “The authors conclude that the high percentage reduction in deaths is, in and of itself, proof that screening at age 40 years is a rational choice for society.” No such statement was made in the article. The conclusion in our article is that screening for women aged 40 to 49 years is efficient in reducing breast cancer mortality in this age group.
The miscalculation of cost per life saved (640,000,000/604 is 1,059,603 not 1,058,225) makes little difference, but more importantly, the misunderstanding of the number of years the women are screened seems to have been mixed up with the average follow-up for death measured from screening start. Women aged 40 to 49 years are screened for 10 years, not 16 years. Neither was the number of women in the study presented as 1 million, which was used by Dr Forsyth for his calculations.
In our article, the number needed to invite to screening during a 10-year period to save 1 life was estimated at 1252 women. If we multiply 1252 women by 5 mammography examinations and assume a participation rate at 80%, then the number of mammography examinations per life saved is 1252 × 5 × 0.8 = 5008. For a cost of $100 per examination, the cost per life saved will be $500,800.
Cost per life saved is not a perfect measure when comparing screening of young women with older women. Cost per life year gained would be a better alternative. In a complete cost analysis, it is also important to take into consideration that breast cancers are found at an earlier stage with screening, which reduces the cost of treatment.
Unfortunately we had some difficulty understanding what Dr. Forsyth means by the “many more controls than treated patients” causing a “strong possibility of a study bias,” but we assume he refers to the possibility of a difference in baseline mortality between the study group and the control group. In the study, the difference in baseline mortality was estimated to be 6% and nonsignificant. This potential difference most likely decreased over the study period, which is further discussed in our article.