Sensitivity analyses for diagnostic studies on ultrasound (‘halo’ sign) and MRI in comparison with clinical diagnosis for cranial giant cell arteritis (GCA) as reference standard
| Number of patients (number of studies) | LR+ (95% CI) | LR– (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | |
| Ultrasound | |||||
| Halo versus clinical diagnosis (main analysis)* | 605 (eight studies)17 19 20 23 25 28 30 31 | 19 (4.8 to 75.5) | 0.2 (0.1 to 0.4) | 77 (62 to 87) | 96 (85 to 99) |
| Excluding high RoB studies*† | 255 (four studies)25 28 30 31 | 16 (7.3 to 35.2) | 0.2 (0.1 to 0.3) | 86 (76 to 93) | 95 (89 to 98) |
| Studies without GC* | 156 (four studies)19 20 25 28 | 4.0 (2.6 to 6.2) | 0.3 (0.2 to 0.7) | 73 (49 to 89) | 82 (75 to 87) |
| Studies with high resolution device*‡ | 292 (four studies)19 25 30 31 | 13.8 (3.9 to 48.2) | 0.1 (0.1 to 0.3) | 87 (77 to 93) | 94 (79 to 98) |
| Suspected diagnosis well-defined§ | 175 (three studies)25 28 31 | 11.5 (4.5 to 29.2) | 0.1 (0.1 to 0.3) | 89 (76 to 95) | 92 (83 to 97) |
| Longitudinal studies§ | 244 (four studies)19 25 28 31 | 7.5 (3.4 to 16.8) | 0.1 (0.1 to 0.3) | 88 (77 to 94) | 88 (78 to 94) |
| MRI | |||||
| MRI versus clinical diagnosis (main analysis)* | 509 (six studies)33–35 38–40 | 5.9 (3.4 to 10.3) | 0.3 (0.2 to 0.5) | 73 (57 to 85) | 88 (81 to 92) |
| Excluding high RoB studies*† | 446 (five studies)33 34 38–40 | 6.8 (3.6 to 13) | 0.3 (0.1 to 0.6) | 75 (56 to 88) | 89 (82 to 93) |
| Studies with high resolution device*‡ | 260 (four studies)33 35 39 40 | 3.8 (2 to 7.5) | 0.4 (0.2 to 0.8) | 68 (44 to 85) | 82 (69 to 91) |
| Suspected diagnosis§ | 270 (three studies)33 34 38 | 8.7 (5 to 15.2) | 0.2 (0.1 to 0.3) | 82 (74 to 87) | 91 (84 to 95) |
| Longitudinal studies§ | 411 (three studies)34 38 40 | 7.5 (4.9 to 11.7) | 0.3 (0.2 to 0.4) | 75 (65 to 84) | 90 (85 to 93) |
↵*Bivariate random-effects binomial generalised mixed model.
†High RoB was defined, in the case of concern on ≥5 RoB items or all 3/3 applicability items out of the QUADAS-2 tool.
‡High resolution devices were defined as >12 MHz probes for ultrasound or 3T MRI machines.
§Univariate random-effects models.
longitudinal studies, studies with clinical diagnosis after follow-up as reference standard; LR, likelihood ratio; QUADAS-2, Quality Assessment of Diagnostic Accuracy Studies-2; RoB, risk of bias; suspected diagnosis, studies with detailed definition of suspicion of giant cell arteritis included; without GC, studies without glucocorticoid treatment before performance of ultrasound.