Map of life expectancy at birth from Global Education Project.

Friday, June 30, 2006

What you don't know probably can't hurt you

At least in some situations, with the benefit of our old friend Bayes Theorem.

My father had surgery a couple of years ago (titanium knees, to be exact) and he had a difficult recovery. The hospitalist (his own primary care doctor was off somewhere doing veterinary medicine, it seems) ordered a scan of the chest to look for pneumonia, among other imaging tests -- which, incidentally, were read by a radiologist in Australia, but that's for another day. He didn't have pneumonia, but Crocodile Dundee saw a lesion on his left lung. The adrenals were just visible at the bottom of the image and he saw lesions there as well. The hospitalist took me aside and told me that they believed my father had metastatic lung cancer.

After my father got out of orthopedic rehab, amid massive anxiety, my parents saw an oncologist who told them it all looked very grim. He ordered a PET scan. My father had scar tissue on his lung and meaningless cysts on his adrenals.

John Stone in this weeks NEJM (you know the story -- off limits) discusses such "incidentalomas." "How many times each day," he asks, "does an incidental finding lead to other tests, additional inconclusive results, and still further testing? What is the effect of incidentalomas on the costs of health care, physician's concerns about being sued, and patient's fears of serious illness." He goes on to tell us that simple (clinically insignificant) renal cysts occur in 12% of people age 50 to 7. And indeed, incidentalomas commonly occur on the adrenals as well. The finding of an adrenal incidentaloma can lead not only to more radiological procedures, as in my father's case, but to biochemical testing and needle biopsies, and even adrenalectomy -- yet only 1 adrenal mass in 4,000 is malignant.

And here's where Bayes Theorem comes in. If you do an imaging test because there are symptoms that lead to concern, the underlying probability that what you are looking for exists is comparatively high. Therefore the fairly non-specific test of a visible lesion on the image is a good indication that you should pursue the matter further. You are really hoping to rule out, rather than rule in. But if there is no prior indication of a problem, a visible lesion that you just happen to see while looking for something else is probably meaningless. But what if you do nothing, and it turns out to be cancer? Major bummer.

This is quite a conundrum, for which I do not have a snap answer. John Iglehart, who shows up here frequently, discusses cost trends and policy developments in medical imaging elsewhere in the Journal. I will get to all that later.

I will be away for a couple of days, but when I get back on Sunday, I'll blog up a storm. In addition to doing Today in Iraq on Monday and Tuesday, we'll discuss: Why Tiger Woods should play more; The vogue for photographing, and even sculpting, pregnant celebrities in the nude; The physics of leaping tall buildings in a single bound - is Superman scientifically credible? (And, why leap anyway if you can already fly?); the economics of converting abandoned factories into office space, as opposed to demolition and new construction.

Or, if we don't discuss any of those topics, we'll do Iglehart and the cost of health care in Massachusetts, and possibly peace, freedom and democracy, being as it's the 4th of July and all that.

No comments: