Letters Section Editor: Robert M. Golub, MD, Senior Editor.
To the Editor: The results of the study of the NMP22 assay for bladder cancer surveillance by Dr Grossman and colleagues1 do not justify adopting this test for such use. First, the assessment of the sensitivity and specificity of the NMP22 assay was flawed. Diagnostic test evaluation requires the uniform application of an unambiguous reference standard.2 The authors used the reference standard of cystoscopy, but it is unclear how often it was applied. The cross-sectional design of the study implies that all patients underwent a single cystoscopy at the time of NMP22 testing. However, at least 9 patients—and as many as 574 not diagnosed as having cancer at their initial cystoscopy—also underwent repeat cystoscopy. If not all 574 patients underwent a repeat examination, then the reference standard was applied inconsistently, raising questions about whether the NMP22 result prompted the repeat cystoscopy, thus inflating the sensitivity of the NMP22 assay (workup bias).
Wilson CT. NMP22 and Surveillance for Recurrent Bladder Cancer. JAMA. 2006;296(1):44-46. doi:10.1001/jama.296.1.44-b