From: Steve Harris <firstname.lastname@example.org>
Subject: Re: How precise is medical science?
Date: 22 Oct 2005 14:06:40 -0700
Murray Grossan wrote:
> On 10/21/05 8:33 AM, in article
> email@example.com, "Jason"
> <firstname.lastname@example.org> wrote:
> > The question comes: how reliable are the studies done in the domain of
> >> medical science given its imprecise nature?
> Imprecise what?
> Kidney transplants
> Smallpox and other vaccines
> Hip replacements
> Cardiac catheterization
> MRI and CAT scans?
> What imprecise are you talking about ??
The imprecision in knowing when to do all these fancy procedures, and
what they'll get you for a particular patient (there's always some
study, but how well does the patient in front of you match the patients
who who were studied?)
And what ARE you seeing in all these fancy tests? There is functional
MRI data we didn't have 10 years ago. There are UBO (unexplained bright
objects) in MRIs we could not even see 20 years ago. And 20 years ago
there was stuff we couldn't see in the CTs of 10 years before that.
What does it all mean? Bayes' theorem bites you in the butt all the
What do bright spots along the VII and VIII cranial nerves mean on a
highfield MRI of a patient with new-onset facial numbness, tics,
vertigo, and tinnitus? Now, suppose this is a migraine patients and
the symptoms disappear? Do you go digging through petrous bone? No,
but you're glad the symptoms went away first or it could have been
embarrasing. Now, what if the bright spots disappear, too? Wups. Could
it be that sometimes doctors see what they expect to find?
Suppose now your patient turns up with semi-Cushingoid symptoms which
could be many other things. Standard cortisol levels and low dose dex
tests are borderline. So his endo guy sends him to the fancy dynamic
MRI pituitary scanner. And finds two masses 2 and 4 mm which are
probably microadenomas. Now what do you do? You couldn't even see
detail this fine not long ago. And you know 1 in 10 people has these
with no Cushings. But only 1 person in 100,000 has Cushings. Hmmmm.
Okay, time for a radiologist to do a CRH-stimulated invasive catheter
inferior petrosal vein ACTH sampling! We're going to do it like Dr.
House on TV, with no screwing up. Except there's no gold standard to
check the results of this kind of thing, so you don't really know how
specific and sensitive it is. And you wish you hadn't done the new MRI
scan, because you don't know how to fit that into the other data. Not
really. It's all way too new to have a database of a thousand patients
with a thousand results of a particular chemical nature, who've had
both tests and then been treated with surgery (or not) and followed.
But meanwhile you have a patient who wants something done....