Rational Use of Diagnostic Tests (0)

Rational Use of Diagnostic Tests
Screening tests
Diagnostic tests are often used to screen asymptomatic patients and identify risk factors for
occult disease . Screening tests should be generally noninvasive, inexpensive, and of minimal
risk to the patient . Screening tests should have high diagnostic sensitivity , which means few
false negative results would be expected , as the goal of testing is to rule out the presence of
disease. Screening tests should be used to screen for diseases that (1) have serious
consequences if left undetected, (2) are reasonably prevalent within the population, and (3)
have treatment options readily available . Should a positive result be obtained, a more
accurate, confirmatory test should then be performed.
One example of a screening test would be the urine cortisol-to-creatinine ratio (Cort:Crt)u,
which is used to screen symptomatic patients for canine hyperadrenocorticism.[1,2] The
(Cort:Crt)u ratio tests for the presence or absence of urinary cortisol excretion and is
noninvasive and inexpensive. It also has a high diagnostic sensitivity, which means that a
negative result strongly indicates that the patient most likely does not have the disease.[1,2]
Urinary cortisol excretion can be caused by both pathologic and physiologic processes,
however , and the test cannot distinguish between the two. Therefore , the (Crt:Crt)u test is not
a confirmatory test as it has poor specificity, which means there are many false positives. If a
patient tests positive, additional tests are required to definitively diagnose
hyperadrenocorticism.
Diagnostic tests
Diagnostic tests are often performed to establish a diagnosis or differentiate between two
diagnoses in symptomatic patients. Since the goal is to definitively identify a specific disease
state, these tests should have good specificity, which means there should be few false positive
results. Tests with high diagnostic specificity are often used after screening tests to confirm
diagnoses. In addition to confirming a disease state, diagnostic tests may also reveal
additional information as well about the severity of disease. They can yield useful prognostic
information and establish a baseline for treatment monitoring and disease progression.
Often, several diagnostic modalities can be used to seek a diagnosis in a symptomatic patient.
Each test has unique characteristics of which a clinician should be aware . For example, both
thoracic radiographs and computed tomography (CT) scan can be used to evaluate pulmonary
metastases. CT scan is the more sensitive modality due to superior contrast resolution and
diminished anatomic superimposition.[3,4] CT may be less specific though, since it detects
more lesions, many of which may be nonneoplastic. It would make sense to use CT scan as a
screening test. However, it is less readily available, more expensive , and requires general
anesthesia, which increases risk to the patient and cost to the owner . Therefore, radiographs
are still more commonly used. However, situations exist where CT scan should be strongly
considered . For example, prior to thoracotomy for pulmonary mass excision, CT scan should
be recommended to better characterize lymph node size and determine the presence of smaller
lesions, since both findings strongly affect prognosis.[5]
In general, less expensive, less invasive, and less risky tests are offered first , although owners
should be made aware that more expensive, more invasive tests may follow if diagnosis is not
achieved. For example, fine needle aspiration and cytologic evaluation are often
recommended prior to surgical removal and biopsy of a mass. For many tumor types , such as
lymphoma and mast cell tumors, diagnosis is usually easily achieved with cytology, and
knowledge of the diagnosis prior to surgery augments treatment planning. For spindle cell
tumors, such as fibrosarcoma or hemangiopericytoma, diagnostic cytologic samples are not
always obtained, and a biopsy may need to be performed to achieve a definitive diagnosis.
Useful characteristics of diagnostic tests
Accuracy and Precision
Accuracy is the measure of a test result's accordance with results determined by a gold
standard. It is the frequency with which a given test correctly identifies a patient as having a
disease, or put another way, how closely the mean of test data points accords to the mean of
the reference standard. A test with high diagnostic accuracy has relatively few false results in
comparison to true results.
Precision is the measure of a test result's reproducibility. If repeated testing is performed on
an individual patient's sample, the results from each testing should be similar to each other if
the test is, in fact , precise. For instance , if a white blood cell count from an individual patient
is run 2 to 3 times on a particular hematology analyzer , the results would be considered
precise if the results from each testing were in concordance with each other. Results from a
hematology analyzer would be expected to be more precise than results from manual
differentials.[7] However, it must be confirmed that this is the case , particularly when results
from a particular test do not accord with expectations or with a patient's clinical presentation.
Put a different way, precision can be thought of as the standard deviation, or how tightly data
points cluster around a mean. To achieve valid results, tests must be both accurate and
precise.
Positive and Negative Predictive Values
Predictive values differ from sensitivity and specificity in that they take into account the
prevalence of disease in the test population. They help answer the clinically relevant question
that, given a positive test result, what is the probability that a patient actually has a given
disease?
First, prevalence is defined as the number of patients within a population that have a
particular disease at a specific time, or the probability that a patient in a given population is
afflicted with a specific disease. Predictive values are predicated upon prevalence and are
therefore often called “posttest” probabilities, since they modify the probability of disease in a
given patient using result information obtained from a particular test.
Positive predictive value is the probability that given a positive test result, a patient is afflicted
with the disease in question. A high positive predictive value means a positive test result
reliably indicates the presence of disease in a patient.
Negative predictive value is the probability that given a negative test result, a patient is not
afflicted with the disease in question. A high negative predictive value means that a negative
test result reliably indicates the absence of disease in a patient.
Predictive values are predicated both on prevalence and on the specificity and sensitivity
characteristics of the test in question. If test specificity is low, the positive predictive value
will be poor and the likelihood of a false positive result will be high. This is important ,
because costs associated with false positives may be considerable in terms of owner stress,
inappropriate treatment, and unnecessary testing.
In general, for understanding and applying predictive values, it is important to remember that
1. Decreased prevalence improves the predictive value of a negative test.
2. Increased prevalence improves the predictive value of a positive test.
3. Increased sensitivity improves the predictive value of a negative test.
4. Increased specificity improves the predictive value of a positive test.
Other considerations
A test is only as good as the data that support and validate it. The following are considerations
that should be used to understand and evaluate the diagnostic accuracy of a given test result in
a given patient.
Reference Standards
Ideally, reference standards for a test should be determined by testing a sizeable population
(>60) of animals representative of the patient being tested. Accuracy may differ according to
age of patient, species, sex, diet , weight, time of day, and activity status.
For example, reference ranges for hemoglobin concentration are age- and sex- dependent
As another example, an antigen- based heartworm test is the gold standard, antemortem
diagnostic test for dogs ; however, cats often have low worm burden, unisex worm
infestations, or immature worm infestations that are not readily detected by these tests
Thus, sensitivity is much lower in cats, and this test would not be appropriate as a gold
standard for testing cats.
Established reference ranges and interpretation of results are only valid for the methods and
the laboratory described. Reference ranges generally represent the range of test results from a
subpopulation of healthy patients, and values that apply to 95% of the subjects evaluated are
included within the range. Thus, patients who consistently fall slightly outside of the range
but do not have progressive changes in their results may be, in fact, normal. Test results
should always be interpreted in light of other clinical findings. Additionally, the more tests
performed, the greater the chance that a falsely abnormal result may ensue. For example, if a
healthy patient were subjected to 20 independent tests, there would be a 64% probability of
finding at least one abnormal result.
Interfering Factors
External factors such as feeding schedule , sample processing , or medications can interfere
with test results. A high-fat meal prior to evaluation of a serum chemistry analysis will alter
values associated with lipid metabolism and cholesterol. Diuretics may affect sodium and
potassium values. Cephalosporins may cause spurious serum creatinine levels. Corticosteroids
and most all chemotherapy drugs severely decrease lymphocyte levels. Lipemia may decrease
sodium levels, depending on the testing method used. Corticosteroids may increase alkaline
phosphatase levels. If blood samples are not processed quickly after being drawn, glucose
levels will often be artificially lowered. In general, if test results do not correlate with clinical
findings or expected results, the veterinarian should view the results with suspicion and
consider repeating them for confirmation.
Patient preparation
The preparation of the patient is important for certain tests. For example, a fasting state is
needed for optimal glucose and triglyceride measurements. Controlled conditions are
frequently needed for endocrinology testing. Careful attention must be paid to patient
identification and specimen labeling. Knowing when the specimen was collected may be
important. For instance, phenobarbital levels cannot be interpreted appropriately without
knowing whether the specimen was drawn just before (“trough” level) or after (“ peak level”)
drug administration. Drug levels cannot be interpreted if they are drawn during the drug's
distribution phase (e.g., digoxin levels drawn during the first 6 hours after an oral dose ).
Substances that have a circadian variation (e.g., cortisol) can be properly interpreted only with
knowledge of the time of day the sample was drawn.
During specimen collection, other principles should be remembered. Specimens should not be
drawn above an intravenous line, as this may contaminate the sample with intravenous fluid.
Lysis of cells during collection of a blood specimen will result in spuriously increased serum
levels of substances concentrated in cells (e.g., lactate dehydrogenase and potassium). Certain
test specimens may require special handling or storage (e.g., blood gas specimens). Delay in
delivery of specimens to the laboratory can result in ongoing cellular metabolism and
therefore spurious results for some studies (e.g., low blood glucose).
Conclusion
Obtaining a proper medical history, performing a complete physical examination, and having
a thorough understanding of the accuracy, precision, sensitivity, specificity, and indications of
the diagnostic test will result in optimum patient management .