Evidence-Based Medicine Diagnostic Studies Mindy Smith and Henry Barry
Objectives Formulate precise question Understand test characteristics Sensitivity and specificity Predictive values Likelihood ratios Critically appraise article
Sensitivity & Specificity true positive rate P(T+)|D+ Specificity true negative rate P(T-)|D-
2x2 table Se = a/(a+c) Sp = d/(b+d) Both are constant characteristics of the test
Practice case 500 patients with IBS and 500 normal volunteers have standing stool velocities. 400 IBS patients have abnormal SSV and 150 normal volunteers have abnormal SSV.
Practice 2x2 Se = ? Sp = ?
Practice 2x2 500 500 Se = 0.8 Sp = 0.7
Conundrums Requirement: to calculate sensitivity and specificity, you have to know beforehand who is diseased and disease-free We order diagnostic tests because we DON’T know if the patient is diseased!
What does all that mean? To interpret the results of a test, you have to know how likely the disease is in the first place pre-test probability disease prevalence positive and negative predictive value are dependent upon pre-test probability
Selecting tests Ruling in disease Ruling out disease
Rules of thumb SpPIn SnNOut a very Specific test, when Positive, rules In disease SnNOut a very Sensitive test, when Negative, rules Out disease
Critical Appraisal What are the ONLY things we ever need to know? (hint, three questions) are the results valid? what are the results? can you apply them?
Critical Appraisal: Validity Are the results valid? independent, blind comparison to gold standard spectrum of severity reference used in all patients
Critical Appraisal: Results What are the results? Sensitivity/specificity Likelihood ratios
Critical Appraisal: Applicability Available, affordable, accurate and precise Can you estimate pre-test probability? Will post-test probabilities affect your management? Do the consequences of the test help your patient?
Likelihood ratio LR+ LR- (probability of positive test in diseased subjects)/(probability of positive test in non-diseased subjects) se/(1-sp) or True positives/False Positives LR- (probability of negative test in diseased subjects)/(probability of negative test in non-diseased subjects) (1-se)/sp or False negatives/true negatives
LR 2x2 Table LR+ = ((a/(a+c))/((b/(b+d)) LR- = ((c/(a+c))/((d/(b+d))
Working with Likelihood Ratios Pretest probability = 10% LR+ = 11 What is the post-test probability?
An easier way
An easier way
Summary Post-test probability of disease is a function of pre-test probability Reverend Bayes Sensitivity and specificity have limited clinical utility SpPIn and SnNOut Likelihood ratios have more clinical utility
Teaching Points Very sensitive tests rule out disease when negative (SnNOut) Very specific tests rule in disease when positive (SpPIn)
More Teaching Points Key validity issues: independent and blind comparison to a gold standard, spectrum of illness severity Post-test probability depends upon the pre-test probability of disease Likelihood ratios tell by how much a given test will increase or decrease pre-test probability