Wednesday, December 21, 2011

Reinterpretation on Negative Findings for CAD in Mammography

I have recently published an article in the Journal of the National Cancer Institute (Oxford University Press) a leading medical journal. It is a letter in reply to a recently published and highly publicized study on the use of computer-aided detection (CAD) technologies in breast cancer screening via mammography. That study, published late last summer was the world's largest study on the effects of computer-aided detection technologies in breast cancer screening (in terms of the number of examinations incorporated into the analysis - 1.6 million screening examinations). The study reported largely negative findings for the performance of breast cancer screening with the help of computer-aided detection technologies.

My letter in reply has indicated that their method for evaluating a screening center's sensitivity is not appropriate for comparing centers with and without CAD technology. I demonstrated that the method used for evaluating how sensitive a screening center is to detecting cancer can easily overestimate the sensitivity of the control centers (with no CAD technology) relative to the experimental group (the ones with CAD technology). This means that the methods chosen by the authors have the potential to result in a misleadingly small improvement in test sensitivity between the groups we are comparing (CAD and no CAD).

I also highlighted that their own results indicated a statistically significant shift in tumour yield towards pre-invasive malignancies. A cancer that is pre-invasive has not yet invaded into neighbouring tissues and typically results in a much better treatment and survival prognosis than invasive cancers. Their own results demonstrate that there was a shift (that should not be attributed to randomness based on their own statistics), towards more pre-invasive tumours among those centers with CAD technology.

You can read my article in the journal through this link, but because of standard academic publication practice, it is behind a paywall. Since a copyright was taken out on the letter, I cannot reprint it here. It will also be appearing in the journal's upcoming print edition. You can also access it through a University who pays the journal for the service of providing their students and faculty with access.

I was pleased to see that the exceptional breast cancer and statistics researchers at the University of Chicago (Bob Nishikawa, Maryellen Giger, Yulei Jiang and Charlie Metz) also published a letter on problems with this recent study. I was fortunate enough to have Bob Nishikawa, the Director of Breast Imaging at the University of Chicago as the external examiner on my PhD thesis years ago.