1. Should you order this new blood test for Mr. Jones?
2. What is the next step in evaluating Mrs. Smith’s newly elevated PTT?
Doctors spend a significant portion of their time ordering and interpreting various lab tests, but knowing when and which test to order is not always easy. As a general rule with any diagnostic testing, it is important to ask the following questions: “What am I looking for?” and “How will the results change the management of this patient?” The sensitivity, specificity, positive predictive value, and negative predictive values are all critical statistics that help you determine the answers to those questions (Table 13-1).
Table 13-1. Calculations of Predictive Value, Sensitivity, and Specificity |Favorite Table|Download (.pdf)
Table 13-1. Calculations of Predictive Value, Sensitivity, and Specificity
|Test Result||Disease||No Disease|
|Positive||A (true-positive)||B (false-positive)||Positive predictive value (PPV) = A/A + B|
|Negative||C (false-negative)||D (true-negative)||Negative predictive value (NPV) = D/C + D|
|Sensitivity = A/A + C||Specificity = D/B + D|
It should be noted that the positive predictive value and negative predictive values will be affected by the prevalence of the condition. Hence, when ordering tests for something very rare, the likelihood of a false test goes up. This is demonstrated in Table 13-2, where for the same specificity and sensitivity the positive predictive value increases dramatically as the disease prevalence goes from 2% in Table 13-2A to 20% in Table 13-2B. Awareness of such principles will maximize your likelihood of testing your patients appropriately.
Table 13-2. Effect of Disease Prevalence on Positive and Negative Predictive Values |Favorite Table|Download (.pdf)
Table 13-2. Effect of Disease Prevalence on Positive and Negative Predictive Values
|Test Result||Disease||No Disease||Total|
|+||99||49||Positive predictive value (PPV) = 99/148 = 0.66|
|−||1||4851||Negative predictive value (NPV) = 4851/4852 = 0.99|
|Sensitivity = (99/100) × 100 = 99%||Specificity = (4851/4900) × 100 = 99%|
|+||990||40||PPV = 990/1030 = 0.96|
|−||10||3960||NPV = 3960/3970 = 0.99|
|Sensitivity = (990/1000) × 100 = 99%||Specificity = (3960/4000) × 100 = 99%|
It is also important to note that laboratory tests that measure a continuous variable (eg, CBC, chemistries, LFTs) are subject to unavoidable systematic error intrinsic to the method by which the reference range is determined. For these types of tests the thresholds used to distinguish a “normal” result from an “abnormal” result are determined based on large samples of data from healthy individuals, with the middle 95% of the values simply defined as normal. In other words, the reference range is calibrated for a specificity of 95%, which means that 5% of healthy individuals would be expected to have a falsely positive test result (ie, the result is outside the reference range even though they don’t have the disease). This has important clinical implications. If you order 20 tests (the labs included with a CBC with differential and a basic metabolic panel will generally cross this threshold), you should expect on average 1 value to be outside the reference range. Since “abnormal” values often lead to further testing, some of which are invasive, a false-positive lab test can be associated with substantial risk. While it may seem you are being more thorough when you reflexively order every lab available, taking the time to order only a limited number of labs is actually protective and is therefore better patient care.
Before you order any test you should ask yourself:
“What am I looking for?”
“How will the results change the management of this patient?”
In trying to answer these questions, we should first examine the test characteristics. This test has a low sensitivity, which means that it is less likely to be positive in disease. But the test is specific, which means that a positive result is most likely a true-positive. Furthermore, colon cancer has a relatively high prevalence, increasing the positive predictive value.
Based on these statistics it may at first glance appear that a positive test result might be useful—after all, it is highly specific and a positive result therefore most likely reflects the presence of disease. With this information we can answer question A: You could use this test to find what you are looking for. But in examining the second question regarding the impact the test result might have on patient management, things get a little more complicated. Indeed, on further examination it becomes clear that a positive result would not change management because it would require a colonoscopy to localize the tumor. If we examine the opposite outcome where this poorly sensitive test is negative, we cannot make any conclusions about the presence or absence of disease and—you guessed it—a colonoscopy would be needed. So, while this test might be able to help find colon cancer, it would not change the management of the patient. You should therefore explain this to the patient, decline to use the new test, and encourage the patient to undergo a colonoscopy.
You have measured 3 consecutive PTT values in the therapeutic range and now there is suddenly a value way above your goal. It is important to first confirm that the drip rate did not change, as a result of either a miscommunication or a nursing error. If the rate has remained the same, then it is likely that some type of lab error has occurred. A very common instance in the case of PTT testing is that the sample sent to the lab was drawn off of a line through which heparin was infusing. A repeat sample should be sent.
Similar situations can occur with almost any medication or infusion going through an intravenous line. For example, one may see falsely elevated glucose levels in patients receiving TPN when the blood sample was drawn off of the line. In such a case, correlation with finger-stick glucose levels is advised.
Other lab abnormalities that can occur are hemolysis in the sample tube, resulting in falsely elevated potassium levels, falsely elevated lactic acid levels from the sample being drawn from a vein in an arm with a tourniquet on it for an extended time, or pseudohyponatremia in the presence of extremely high glucose levels.
First, let us review urine testing. There are 2 common urine tests—urinalysis and urine culture. A urinalysis looks at the following: appearance, pH, specific gravity, protein, glucose, ketones, red blood cells, leukocyte esterase, nitrites, bile, and urobilinogen. Additional elements of a urinalysis analyzed at the microscopic level are white blood cells, urinary casts, yeast, and epithelial cells.
The interpretation of urinalysis is not as easily defined as for some other tests. Each element of the urinalysis should be integrated into a whole, balancing those elements that suggest infection against those that argue against. Ms. Doe has some elements of her urinalysis that might suggest a urinary tract infection, such as the presence of bacteria and the presence of leukocytes. This is counterbalanced by the presence of squamous epithelial cells, negative leukocyte esterase, and negative nitrate. But how do we integrate these conflicting data?
Bacteria often appear in urine specimens. Normal microbial flora of the vagina in the female and the external urethral meatus in both sexes can rapidly multiply in urine standing at room temperature. Bacteriuria alone is therefore not diagnostic and, in a case of suspected urinary tract infection, requires culture to determine if it represents a pathogen or contamination. A bacterial count of more than 100,000/mL of 1 organism reflects significant bacteriuria and, in the setting of other positive findings on the urinalysis, a likely urinary tract infection. The presence of multiple organisms is usually indicative of contamination.
Leukocytes (white cells) may appear with infection in either the upper or lower urinary tract or with acute glomerulonephritis. White cells may also be contaminants from other, non-urinary tract sources. For example, samples may be contaminated with white cells from the vagina (often present in vaginal and cervical infections), or white cells from the external urethral meatus (in both men and women). Yet the presence of squamous epithelial cells indicates likely contamination from the skin surface or from the outer urethra. In general, 2 or more leukocytes per high-power field in urine that does not include any other markers of contamination suggest that the specimen is abnormal. Higher numbers of leukocytes are even more convincing in this regard.
Nitrates and leukocyte esterase are more specific tests of urinary tract infection. Nitrates are the by-product of some bacteria, especially gram-negative rods such as E coli. Leukocyte esterase is normally contained within white blood cells and its presence indicates likely pyuria. Its absence, in contrast to nitrates, strongly argues against a urinary tract infection, especially in the absence of other compelling markers. As with all urinalysis data both leukocyte esterase and nitrate results should be integrated with other available data prior to drawing a final conclusion.
Lastly, a urinalysis should also be interpreted in the context of a patient’s clinical picture, including whether the patient is having symptoms of a UTI such as fevers, dysuria, or increased urinary frequency, urgency, or hesitancy. If the patient is symptomatic, then a urinalysis may not even be needed, and if performed, may require less criteria to be considered positive.
In our example, Ms. Doe’s urinalysis likely reflects a contaminated specimen. The squamous cells indicate contamination, she has relatively few white cells, leukocyte esterase and nitrate are negative, and the bacterial count is below the standard cutoff for urinary tract infections. Although she has 3 to 5 white cells, this is difficult to interpret in the context of contamination. Lastly, both negative leukocyte esterase and nitrate support the negative laboratory diagnosis.
3. Does Jane Doe have a urinary tract infection?