I read with interest the article “Influence of Computer-Aided Detection on Performance of Screening Mammography,” published recently in the New England Journal of Medicine.  The short of it would be that using computer-aided detection did NOT increase the accuracy of mammograms.

I’ve actually had conversations with one of the co-authors as well as a former mentor of mine at the National Cancer Institute who did a lot of work studying mammographic standards.

What the study showed was that using computer-aided detection (CAD) resulted in an increase in biopsy rate of 20% but not a significant change in cancer detection rate.  The CAD did pick up more ductal carcinoma in situ. (See Dr. Knopf's past SharePosts for more on DCIS).

One very interesting point that is not always explained to the lay person is from Table 1 in the article which shows the chance of having a breast cancer diagnosis within one year after a positive screening mammogram.

Table 1: Chance of Having a Breast Cancer Diagnosis within One Year Following a Positive Screening Mammogram

    1.3% in women aged 40-44

    2.2% in women aged 45-49

    3.1% in women 50-54

    4.5% in women 55-59

    5.3% in women 60-64

    6.5% in women 65-69

    7.9% in women 70-74

    8.9% in women over 75 

This means that for a woman between 55-59 the chance of a biopsy of a questionable mammogram (mammographic abnormality) being breast cancer is about 1 in 20. Additionally, this means that most mammographic abnormalities do not result in a diagnosis of breast cancer.

While this may not apply to readers of this SharePost, those who were all in the unlucky 1.3 to 8.9% who DID have a diagnosis of breast cancer, it is important to realize that there are many questionable or abnormal mammograms that do not actually mean that the particular woman will be diagnosed with breast cancer. 

This is the price we pay with mammography.  The stakes of missing a breast cancer are incredibly high – ethically, morally, and legally.  Dr Hall, a mammographer at Beth Israel in Boston, writes that many residents avoid going into mammography and that demand for good mammographers is unmet.  Solutions to this problem are complex.

Computer-aided detection will likely not go away – in general I think people would rather not miss any breast cancers rather than undergo biopsies of normal breast tissue.  This dilemma of radiologic areas that aren’t necessarily breast cancer is also going to be with us with the increasing use of breast MRI, which picks up a lot of abnormalities that aren’t always cancer.

Stay tuned – there will be more to come.