Transcript How to practice and teach EBM Chapter 3

How to practice and teach EBM
Chapter 3
May 3, 2006
Vanderbilt Sports Medicine
Diagnosis and Screening
3 questions to consider about tests
– Is the evidence about accuracy of a test
valid
– Does evidence show test can distinguish
patients who do/don’t have disorder
– How can I apply test to a specific
patient
If the evidence is valid, is it
important
– Ie: clinically worthwhile
Validity of evidence
Measurement (most important)
– Was there an independent, blind
comparison with a reference gold
standard
Representative
– Appropriate spectrum of patients
Ascertainment
– Was reference standard ascertained
regardless of diagnostic test result
Importance
Does evidence demonstrate ability of
test to distinguish patients w and
w/o disorder
Sensitivity
– pts with disorder who have positive test
Specificity
– Pts without disorder who have negative
test
Importance
Likelihood ratio - LR(+)
Prob of positive test in presence of disorder
Prob of positive test in absence of disorder
– LR+ = sensitivity/(1-specificity)
– Positive test is more likely in pt w disorder
LR(-) = (1-sensitivity)/specificity
– Negative test more likely in absence of
disord
Importance
SnNout
– Test with high sensitivity (Sn)
– Negative result (N)
– Rules OUT diagnosis
SpPin
– Test with high specificity (Sp)
– Positive result (P)
– Rules IN diagnosis
Applying test to a patient
Is test available, affordable,
accurate, and precise in our setting
– Some tests have higher LR in florid
disease and lower LR in mild
– As patients are referred to specialists,
some tests loose power because
patients displaying symptoms will
include more false-positives
Applying test to a patient
Can we generate an estimate of pre-test
probability
– From experience, prevalence studies, practice
databases, pre-test probability studies
Does post-test probability affect mgmt
and help patient
– Does result cross threshold that causes us to
stop testing
(-) test that r/o likelihood of diagnosis
(+) test that r/in diagnosis
– If not, perform additional tests to pursue
diagnosis
Test treatment thresholds
May not cross threshold until several
tests are performed
If tests are independent, can chain
LR together for each test to increase
overall likelihood ratio
– Ex: LR test 1 (0.06) x LR test 2 (13) x
LR test 3 (11) = 9.13
– Cvt LR to prob: 9.13/10.13 = 90% prob
More than any single test alone
Multilevel LR
Expressing LR for two levels (+/-) is
helpful, but giving LR for multiple
levels allows more distinction
– Ex: 5 levels from extremely negative –
mod negative – neutral – mod pos –
extr pos
– Extremely positive more definitive than
if only had two levels (pos/neg), so
result is more useful
Multiple Tests
Multiple tests grouped together
– Considered a “cluster”
– Combination of results helps make
diagnosis
– When validated in second independent
group
“Clinical prediction guideline”
Critically Appraised Topic (CAT)
Standardized, one page summary of
evidence on one topic
– Shows bottom line result
– Show scenario and clinical question
– Show studies and gold standard
– Presents evidence and results
– References
Screening/Case Finding
Screening
– Making early diagnosis of presymptomatic disease among well
general public
Case Finding
– Making early diagnosis of presymptomatic disease in patients who
presented for an unrelated disorder
Screening/Case Finding
Implicitly recruit pts with promise that
they will live longer or better if they are
tested
Requires that evidence shows pts are
better off in long run with early diagnosis
– Labels all pts (may be harmful)
– False + test is only harmful (no benefit)
– Test must not only be accurate, but treatment
must be efficacious
Screening/Case Finding
Does RCT evidence show early diagnosis
leads to improved survival/quality of life
– RCT type 1 – screen vs no screen
– RCT type 2 – (+) screen test – treat vs no
treat
Will early diag pts cooper with treatment
– If not, no benefit
How do benefits/harms compare
Does frequency/severity of disorder
warrant the effort
Screening/Case Finding
Bias
– Early detection always appears to improve
survival (may in fact not) - “pt sicker longer”
– Pts who volunteer for testing are more cooper
with health advice
Have better outcomes anyway
– Early tests ID pts with slower/more benign dz
– Is follow-up adequate
Early detection is not always a good thing!