Sixteen studies were included in the review: 11 level 1 studies and 5 level 2 studies. One further study was included in the evaluation of colour Doppler ultrasonography for the detection of isolated calf venous thrombosis only.
In level 1 studies, ultrasonography (any technique) had an overall sensitivity of 62% (95% CI: 54, 70), a specificity of 97% (95% CI: 96, 98), a positive LR of 18.5 (95% CI: 13.7, 25.7) and a PPV of 66% (95% CI: 58, 74) for detecting proximal thrombi. Statistically significant differences in sensitivity (P=0.008) and specificity (P=0.003) were found between studies within each of the three ultrasound modalities. The summary ROC curve did not indicate a correlation between sensitivity and specificity.
In level 2 studies, the sensitivity was 95% (CI: 87, 99), the specificity was 100% (CI: 99, 100), the positive LR was 609.3 (95% CI: 577.3, 643.1) and the PPV was 100% (CI: 94, 100).
A statistically significant difference (P<0.001) was found in the sensitivity, specificity and PPV between level 1 and level 2 studies. Accuracy values and LRs for level 2 studies were all considerably higher.
When ultrasonography was used to detect isolated calf venous thrombosis, a PPV of 93% (CI: 68, 100) was found with a sensitivity of 48% (CI: 29, 67). These data were derived from two level 1 studies.