In their study of the diagnostic utility of impedance plethysmography (IPG), Anderson and colleagues [1] ignored the first guideline for evaluating a diagnostic test, namely, the need for "an independent, ‘blind’ comparison with a gold standard of diagnosis" [2]. Although the "gold standard" of venography was sensibly used in this study, it was supplemented by compression ultrasonography in a handful of patients, for reasons that are not described. Even so, neither test was done in 242 of 386 patients because venography was done only if the IPG result was positive (or if a "high clinical suspicion" was present). In addition, whether the assessment was "blinded" is not specified in the Methods section. The retrospective nature of the clinically derived database, in which the results of IPG determined referral for venography, suggests that blinding of referring doctors did not occur.

We agree with the authors that these biases tend to overestimate the sensitivity; thus, their primary conclusion, that the sensitivity of IPG is lower than had been thought, is supported by these findings. However, as they note, they "cannot accurately determine specificity of IPG or prevalence of DVT (deep-vein thrombosis) in this cohort of patients." Without sensitivity, specificity, and prevalence, predictive values cannot be calculated, and sufficient data do not exist to extrapolate results to other populations. Finally, the authors' conclusion that "we no longer rely on the safety of a normal IPG result ..." should be based on the negative predictive value, not the positive predictive value; with the "missing cell" of the 2 x 2 table (Figure 2 of reference 1), one cannot even guess the negative predictive value.