To the Editor.
—The article by Dr Johnson1 summarizes a special case of pretest probability (high pretest probability) as a major factor in determining utility of a test. I believe two points in the article deserve comment.First, Fig 1 in Johnson's article fails to account for false-negative results. When the pretest probability of disease is high, the test results may actually detract from diagnostic accuracy. For example, it has been stated in The Journal that "the ultrasound scan actually detracts from diagnostic accuracy in patients with a high (pretest) probability of extrahepatic obstruction."2 As the probability of disease approaches 1, any "negative" (normal) value must be a false-negative. Therefore, the information provided by the test is not zero, but less than zero—a negative number, indicating the test's potential to mislead, reduce diagnostic accuracy, and worsen outcome.Second, Johnson states that "the information provided by a test can
Fox GN. Diminishing Diagnostic Returns. JAMA. 1991;266(16):2224. doi:10.1001/jama.1991.03470160056027