Thirty studies were included (number of participants unclear). It appears that six of these studies were later excluded as they did not specify the type of arthrography conducted.
None of the studies met all the quality criteria. Only 2 studies used a prospective design; 8 studies enrolled consecutive patients. All included studies used a valid reference standard but in 10 studies not all patients received this reference standard. None of the included studies measured the reference standard independently of the index test.
Summary ROC plots suggested considerable heterogeneity across the studies in both sensitivity and specificity. There was no association between methodological quality and diagnostic performance.
Acetabular components.
Sensitivity was significantly higher (p=0.0078) for subtraction arthrography (10 studies) than for contrast arthrography (8 studies). The pooled sensitivity was 70% (95% confidence interval, CI: 52, 84) for contrast arthrography and 89% (95% CI: 84, 93) for subtraction arthrography. There was no difference in specificity between the techniques (p=0.84): the pooled specificity was 74% (95% CI: 53, 87) for contrast arthrography and 76% (95% CI: 68, 82) for subtraction arthrography. Sensitivity and specificity did not differ according to whether manual (7 studies) or digital subtraction (4 studies) was used (p=0.65).
Femoral components.
Sensitivity was significantly higher (p=0.003) for subtraction arthrography (9 studies) than for contrast arthrography (13 studies). The pooled sensitivity was 63% (95% CI: 53, 72) for contrast arthrography and 86% (95% CI: 74, 93) for subtraction arthrography. There was no difference in specificity between the techniques (p=0.23): the pooled specificity was 78% (95% CI: 68, 86) for contrast arthrography and 85% (95% CI: 77, 91) for subtraction arthrography. Since the studies only assessed digital subtraction a comparison with manual subtraction was not possible.