Objective: To evaluate, by meta-analysis, the accuracy of ultrasound screening for deep venous thrombosis in patients after orthopedic surgery.
Data sources: The MEDLINE database from January 1982 to October 1993. Bibliographies of retrieved articles and recent journal publications were searched independently and using Current Contents.
Study selection: All articles evaluating the use of venous ultrasound imaging (B-mode, duplex, and color Doppler) compared with standard contrast venography for detecting deep venous thrombosis. We excluded abstracts, early reports of studies later reported in full, and studies in which venography was not done in all patients. Seventeen of 30 identified studies were eligible.
Data extraction: Eligible articles were reviewed for the presence of three key criteria necessary for evaluating the accuracy of the diagnostic tests: 1) previously established objective criteria for venography and ultrasound, 2) independent blinded comparisons of venography and ultrasound, and 3) prospective evaluations of consecutive patients. Studies including all three key criteria were defined as level 1 (minimized bias) studies; otherwise, they were defined as level 2 studies.
Data synthesis: In level 1 studies, ultrasonography had a sensitivity of 62% (95 of 153; 95% CI, 54% to 70%), a specificity of 97% (CI, 96% to 98%), and a positive predictive value of 66% (95 of 144; CI, 58% to 74%) for detecting proximal thrombi. For level 2 studies, the sensitivity was 95% (CI, 87% to 99%), the specificity was 100% (CI, 99% to 100%), and the positive predictive value was 100% (CI, 94% to 100%). Differences between level 1 and level 2 studies appeared to be related to bias in study design.
Conclusions: Venous ultrasound imaging has only moderate sensitivity and a moderate positive predictive value when used to screen for deep venous thrombosis in patients after orthopedic surgery; thus, ultrasound imaging may have limitations as a screening test.