According to Bandolier (August 2001) it seems that "few systematic reviews
(SR) of diagnostic tests are useful, because they just summarise wrong
results", so "doing SR of diagnostic tests is a complete waste of time". On
the other hand, a few SR which can be considered valid (Reid et al, JAMA
1995; 274: 645-51; Ljimer et al. JAMA 1999; 282: 1061-66; ...) demonstrate
that "most diagnostic tests are evaluated using architecture subject to
immense bias". We should do new and free from bias studies, according to
the EBM rules.
This sounds an excellent advise and an immense task for lab people. But, in
my opinion, a serious problem is then the utilization of a test. According
the bayesian rules, clinicians should revise the pre-test odds of a
disorder to obtain, through the likelihood ratios (LR) of a test, the
post-test odds and make a diagnosis.
A discussion on difficulties for a practical use of pre-test odds ended
around April 1999 in this list, with no conclusions except some bitter
remarks, f.i. the necessity of waiting for "at least a generation of
physicians before any fundamental changes can actually take effect"
So, should we struggle for devising totally new, ebm based tests, in the
hope they will be correctly used by our children? Meanwhile, what's the use
of speaking the LR language?
Dr. Giuseppe Giocoli
Gruppo di Lavoro EBM
Associazione Microbiologi Clinici Italiani
Via Sarca, 19
25015 DESENZANO d/G (BS) Italia
[log in to unmask]