Ordering Lab Test I Flashcards
PSA (prostate specific antigen)
Protease produced by glands of the prostate
Elevated in prostate cancer (correlated with tumor burden in metastatic prostate cancer)
Can be elevated in patients without prostate cancer (benign prostatic hyperplasia)
Assayed with capture immunoassay, using a chemiluminescent reaction to quantitate the PSA serum level
A quantity that describes accuracy is….?
Bias
The difference between the “true” value and the mean of the measured values is the bias
What is the quantity that describes precision?
% cv (coefficient of variation)
What is the %cv?
Standard deviation divided by the mean x 100
Total error?
The percent bias times percent precision
Total allowable error?
Takes into account clinical decision making - e.g. Break points for clinical action
Functional sensitivity
Level at which the %cv does not exceed 20% (high cv is bad - talks about precision)
The lower end of the AMR (analytical measurement range) is defined by….
The functional sensitivity, provided the linearity at that level is acceptable
The upper end of the AMR (analytical measurement range) is defined by…
The limit of linearity
When should the AMR extend down to a really low value….
Monitoring FSH and LH in children with precocious puberty (being treated with GNRH)
Hospital based labs use 2nd generation and national reference labs use 3rd generation - both are more sensitive
Or!! HCG monitoring for females of reproductive age taking isotretinoin for severe acne or lenalidomide for multiple myeloma
CRR (clinical reporting range)
Typically the same as the AMR at the lower end, but can be much higher at the upper end
This is fixed with the use of a validated dilution protocol for the pt samples that exceed the AMR
Exs. Of when the CRR should be higher than the upper range of the AMR?
Tumor markers
Creatine Kinase (CK) - protein released from muscle fibers that are necrotic - pts with duchenne muscular dystrophy
What does analytical specificity ask?
Does the assay actually measure what it purports to measure and nothing else?
- for immunoasssays, you have to be concerned about cross-reactive substances and interferences
Reference Range - population based
Mean +/- two standard deviations of normal subject data
Central 95% of normal subject data
Normally the population is somewhat skewed
Reference range: physiologic
Rarely available!
Ex. PTH - vitamin D level in serum is construed as the lvl at which PTH plateaus
Reference range - based on health
Based not upon the “normal” population but rather on what is considered optimal for health
Ex. American Heart association considers plasma cholesterol should not exceed 200 mg/dl
Arbitrary cut off
Few ideal lab tests exist where the normal and diseased states do not overlap with respect to analyze values
…
“False results” do not refer to flawed assays but rather mismatches between disease status and analyte values
…
Sensitivity formula?
TP/TP+FN
Specificity formula?
TN/TN+FP x 100
Sensitivity refers to?
How many people with the disease have abnormal results
Specificity refers to?
How good is this test? Does this test really tell us that they have it?
Defined by people who don’t have the disease and how many of them are negative for that marker
Screen with a test for which the cut off has what first? Then what second?
Has been designed for high sensitivity first (get all positives) and then confirm the test for which the cut-off has been designed for high specificity (find all the true negatives)
Ex. Borrelia burgdorferi antibody screen with ELISA then with Western Blot/PCR
Receiver operating characteristic (ROC) curve
Test performance is correlated with area under the curve
Cut off value depends upon use:
1) screening (max sensitivity at the expense of specificity)
2) risks associated with false diagnosis
What to be near the top left
Positive predictive value formula?
TP/TP +FP
Negative predictive value formula?
TN/TN+FN
What is on the predictive value table?
Test result (pos/neg)
Disease (present/absent)
The very same test can perform better or worse depending upon the prevalence of the disease that the test is used to diagnose…. for ex?
Hospitalized pts have typically higher prevalence of disease than out patients, thus have PPV higher and NPV is higher for out patients
Disease Probability: Bayes’ Theorem
“Pre-test probability” may be thought of as the prevalence of disease; ideally it is refined with additional info such as hist/PE
Likelihood ratio - positive test
For a pos test, it is the probability of a person with the disease testing positive, divided by the prob of a person without diesese testing positive (i.e. The ratio of true positive rate to false positive rate)
Sensitivity/(1-specificity)
Likelihood ratio - neg test
Prob of a person with the disease testing negative divided by the prob that a person without the disease testing negative - i.e. , the ratio of the false negative ration to the true negative ratio
1-sensitivity/ specificity
Post test probability
Can be derived from a nomogram
Liklihood ratios for a positive result are better for…?
Diagnosing disease the higher they are (over 10)
Likelihood ratios for a negative result are better at….
Ruling out diseases the lower they are
Exs where Bayes’ theorem can be applied without the help of a computer are oversimplified: they typically treat tests as qualitative (pos or neg) and they are focused on one test
Computer prediction models that treat test results as continuous quantitative variables and that take into account multiple test results (multi-variant analysis) are the future tools you will use; there is some clinical use already with these tools
…
Within lab test performance characteristics dictate the analytical range for an assay
…
Between lab differences exist for accuracy, precision, and reference range for tests
…
ROC curves can be used to…
Compare test and also to find optimal decision cut offs
Disease prevalence figures into test utility
…
Pre and post test probabilities shoudl considered when entertaining a diagnosis or a diagnostic plan
….
