Am Fam Physician. 2012;85(9):847-852
Original Article: Usefulness of Procalcitonin Measurement in Reducing Antibiotic Use and Identifying Serious Bacterial Illness [AFP Journal Club]
Issue Date: July 15, 2011
Available at: https://www.aafp.org/afp/2011/0715/p177.html
to the editor: I appreciated the thoughtful and critical analysis by the authors of the AFP Journal Club on the impact of procalcitonin measurement in treating children one to 36 months of age presenting with fever without a source. However, in the discussion of the impact of disease prevalence on a diagnostic test's performance characteristics, the authors may have confused two concepts in clinical epidemiology: sensitivity/specificity and predictive value.1
One author stated that the sensitivity and specificity of a test may change depending on the prevalence of disease in the population in which the test is used. Yet, according to the textbook Clinical Epidemiology: The Essentials, “…the sensitivity and specificity of a test are said to be independent of the prevalence of the diseased individuals in the sample in which the test is being evaluated.”2 In theory, the sensitivity and specificity of procalcitonin measurement are independent of the prevalence of the target disease.
The test characteristic concept that the author may have actually been alluding to is the test's predictive value. The prevalence of the target disease in a population would affect the predictive value; positive predictive value decreases as the prevalence of the target disease decreases, and negative predictive value decreases as the prevalence of the target disease increases.
in reply: Dr. McDiarmid is entirely correct. This discussion should have referred to positive and negative predictive value rather than sensitivity and specificity. I apologize for the confusion.
editor's note: As Dr. McDiarmid points out, sensitivity and specificity are characteristics of the test itself, and generally do not vary with disease prevalence, although this is not a hard and fast rule.1 Positive and negative predictive values are inherently tied to disease prevalence. As most clinicians know, a positive test in a population with a very low disease prevalence is likely to be a false positive. Similarly, a negative test in a patient you think does have a certain disease has a good chance of being a false negative. Thus, positive and negative predictive values vary not only with disease prevalence, but, almost magically, with your clinical suspicion that a patient does or does not have the disease for which he or she is being tested.
For more information about the terms and concepts of clinical epidemiology, see our glossary of terms used in evidence-based medicine at https://www.aafp.org/journals/afp/authors/ebm-toolkit/glossary.html. This is one of several features in our Evidence-Based Medicine Toolkit available at https://www.aafp.org/afp/ebmtoolkit.
The online version of this AFP Journal Club discussion has been corrected to reflect the changes mentioned above.