Figure 1.

Expected concordance rate when the same assay is repeated on the same patient cohort. In this example, the test has 90% sensitivity and specificity (overall accuracy of 90%), the measurement error is random, and the test is applied to 160 tumors. Cases in the boxes represent discordant results between two measurements: 9 + 9 + 5 + 5 = 28, corresponding to a 17.5% discordance rate. After the second test, the same number of cases remain false negative or false positive as after the first test, n = 10 (cases underlined), resulting in no overall improvement in diagnostic accuracy. If repeat testing is limited only to the false-negative cases of the first test, the discordance rate would be high (90%) but the overall diagnostic accuracy would improve substantially.

Abbreviation: ER, estrogen receptor.