This recent JAMA article (subscription required) carefully studies the role of adding ultrasound to mammography in breast cancer screening. They found that the combination picked up more cancers but also increased the false positives. In the parlance of clinical epidemiology, it increased sensitivity but decreased specificity.
Or as Homer Simpson would say: DUH!
It is not uncommon that the two co-vary in opposite directions. You can virtually manipulate the sensitivity and specificity of a test by carefully choosing your cutoffs for indicating the presence of absence of disease. The test ends up performing according to your needs. Our purpose in combining tests is to maximize both sensitivity and specificity. HIV is a case in point, since it is an over-sensitive antibody test (won't miss anyone) followed by an extremely specific confirmatory Western blot (won't wrongly diagnose anyone). Most labs won't report a positive antibody unless the Western blot is also positive.
I wish the results of the breast cancer study would have shown a reduction in false positives by the addition of ultrasound. This has always been my fervent hope whenever I have added an ultrasound to a mammography. The lesson of the report is that ultrasound doesn't help that much and must be used with caution and judgment.
Journal Watch reports a JAMA editorialist pointing out that
despite the high number of false-positives, what high-risk women "probably fear most is a late diagnosis." That, she continues, is "the real threat they want to be protected against, not false-positive diagnoses."Again I must say, DUH!
This study is meant to help guide us through the emotionally charged fear of breast cancer. It packs a huge punch for any woman even suspected of having the disease. On the other hand, we need to know the real (scientific= truth) value of doing a test. Perhaps the editorialist is saying that since it is an emotional matter, neither the data nor the truth matter, as long as we only detect more stuff, cancerous or not.
The original article is very detailed and a great read for those with an interest in how to crunch numbers to guide investigative decisions. The editorial set me off!