The predictive value of a test is often misinterpreted because it is presented as a percent. It is intuitive to assume that low percentages (70% or less) are "bad" and high percentages are "good". A positive predictive value of 20%, for example, was cited as proof that a test should not be used even though the positive likelihood ratio for that same test was 50. A likelihood ratio of 50 means that the post test odds of disease for a positive test result will be 50 times higher than the pretest odds of disease. Now, that is a large increase in the odds. Critics of laboratory medicine fail to recognize that sensitivity and specificity vary with the strength of the signal. Thus, a value well above the cutoff is far more likely to indicate disease than does a value just above the cutoff--even though both are reported as "positive". Tables of likelihood ratios for a wide range of specific test results, or for multiple test results, provide more information than a simple four-by-four predictive value table. Likelihood ratios are also more informative than predictive values or ROC curves. Finally, critics of laboratory medicine fail to take into account the information to be derived from a confirmatory test, a repeat test at a later time, and from other tests.