Anderson and colleagues' [1] retrospective analysis differs fundamentally from previous prospective analyses [2]. In the methods section, Anderson and colleagues report that "deep-vein thrombosis was diagnosed ... by non-filling of a vein or venous segment despite repeated injections with contrast material." There is international agreement [2] that non-filling is not a criterion for DVT. Previous prospective analyses have shown that non-filling is inadequate for interpretation [2] but that IPG is sensitive to obstructive DVT. Non-filling as described by Anderson and colleagues may have been caused by technical problems [2] rather than an established venous thrombosis.

The use of a nonspecific criterion for DVT introduces two potentially serious flaws. First, the positive ultrasound result in the presence of a nonspecific, non-filled segment on the venogram may have been false-positive. Second, a positive ultrasound interpretation in the presence of an inadequate venogram introduces a serious bias; in patients with an inadequate venogram, the positive ultrasound result was assumed to be correct in the presence of normal IPG results. This method of analysis is scientifically inappropriate.

In addition, Anderson and colleagues introduced an incorporation bias by including four additional patients whose results were classified by the ultrasound results alone. These patients should not have been included in the analysis.

It is crucial that a reanalysis be presented according to internationally standardized venography interpretation [2] because Anderson and colleagues' findings may reflect the method of venographic interpretation rather than characteristics of the noninvasive tests evaluated.