Sixty-seven diagnostic accuracy studies (n=4,711) were included.
Forty studies (61%) were blinded, 33 (59%) used consecutive recruitment, 61 (92%) used some type of reference standard in all patients, while 20 (30%) used the same reference standard test in all patients. Study quality was also reported separately for studies used in the assessment of obstruction, bile duct stones and malignancy.
All imaging end points: MRCP had a high adjusted sensitivity of 95% (95% CI: 75, 99) and specificity of 94% (95% CI: 86, 99).
MRCP had a high sensitivity (97%, 95% CI: 91, 99) and specificity (98%, 95% CI: 91, 99) for the presence of obstruction (30 studies, 1,954 patients); the LR of a positive test result was 49 (95% CI: 25, 62). Similarly high values were also found for the level of obstruction: sensitivity 98% (95% CI: 94, 99), specificity 98% (95% CI: 94, 100), and LR of a positive test result 49 (95% CI: 25, 135). The sensitivity and specificity for the diagnosis of stones were 92% (95% CI: 80, 97) and 97% (95% CI: 90, 99) respectively, (46 studies, 3,592 patients), and the LR of a positive test result was 29 (95% CI: 23, 49). Whilst the sensitivity and specificity for differentiating benign from malignant biliary obstruction were 88% (95% CI: 70, 96) and 95% (95% CI: 82, 99), respectively (22 studies, 12,943 patients), the LR of a positive test result was 16 (95% CI: 10, 30).
Area under the ROC curve: the area was 0.99 (95% CI: 0.93, 1.0) for the presence of obstruction, 0.99 (95% CI: 0.90, 1.0) for the level of obstruction, 0.98 (95% CI: 0.97, 0.99) for bile duct stones, and 0.97 (95% CI: 0.85, 1.0) for malignancy.
The diagnostic accuracy was higher for larger studies, for studies that did not use consecutive enrolment, and for studies that did not use the 'gold' standard for all patients. The results were reported in the paper.
The funnel plots did not show any evidence for publication bias.