Twenty-nine studies met the inclusion criteria (5,153, range 50 to 779 patients; 5,511, range 50 to 779 lesions). Of the 29 studies, all used an appropriate reference standard and avoided incorporation bias. More than 95% of studies accounted for intermediate results and explained withdrawals. Approximately 40% to 45% of studies avoided differential verification bias, recruited a representative patient spectrum and/or blinded interpreters of the index test. Approximately 10% to 15% of studies avoided clinical review bias and progression bias. None of the studies blinded interpreters of the reference standard.
For ultrasound elastography alone, the summary estimate of sensitivity was 79% (95% CrI 74% to 83%), specificity was 88% (95% CrI 83% to 92%), LR+ was 6.71 (95% CrI 4.60 to 10.20) and LR- was 0.24 (95% CrI 0.19 to 0.30). For B-mode ultrasound alone, the summary estimate of sensitivity was 96% (95% CrI 93% to 98%), specificity was 70% (95% CrI 55% to 83%), LR+ was 3.10 (95% CrI 2.12 to 5.14) and LR- was 0.06 (95% CrI 0.04 to 0.10).
When B-mode and ultrasound elastography were combined, the summary estimate of sensitivity was 73% (95% CrI 67% to 79%), specificity was 97% (95% CrI 95% to 98%), LR+ was 26.20 (95% CrI 16.00 to 48.68) and LR- was 0.28 (95% CrI 0.22 to 0.34). When the combination of the two tests used the disjunctive positivity criterion, the summary estimate of sensitivity was 99% (95% CrI 98% to 99%), specificity was 59% (95% CrI 44% to 68%), LR+ was 2.41 (95% CrI 1.79 to 3.05) and LR- was 0.02 (95% CrI 0.01 to 0.2).
There was no evidence of publication bias.