Thirty-two studies met the inclusion criteria (n=1,022 patients; range 10 to 76). Study quality was considered to be generally high; the median QUADAS score was 13 (range 11.5 to 14). Two studies avoided clinical review bias. Twenty-eight studies reported recruiting a consecutive sample. The proportion of excluded (uninterpretable) segments from 0 to 13%. There was no evidence of publication bias.
Entire arterial tree (19 studies): Magnetic resonance angiography (MRA) pooled sensitivity was 94.7% (95% CI 92.1 to 96.4); specificity was 95.6% (95% CI 94.0 to 96.8); positive likelihood ratio was 21.56 (95% CI 15.70 to 29.69); and negative likelihood ratio was 0.056 (95% CI 0.037 to 0.083). Heterogeneity was high for both sensitivity (I2=81%) and specificity (I2=89%). There was no evidence of publication bias.
Aortoiliac subregion (19 studies): MRA pooled sensitivity was 93.5% (95% CI 90.51 to 95.64); specificity was 96.3% (95% CI 94.5 to 97.6); positive likelihood ratio was 25.33 (95% CI 16.80 to 38.41); and negative likelihood ratio was 0.067 (95% CI 0.045 to 0.099).
Femoropopliteal subregion (20 studies): MRA pooled sensitivity was 95.3% (95% CI 90.6 to 97.7); specificity was 95.6% (95% CI 93.9 to 96.8); positive likelihood ratio was 21.40 (95% CI 15.26 to 29.81); negative likelihood ratio was 0.050 (95% CI 0.024 to 0.099).
Tibiofibular subregion (25 studies): MRA pooled sensitivity was 92.2% (95% CI 89.5 to 96.4); specificity was 93.3% (95% CI 89.8 to 95.7); positive likelihood ratio was 13.80 (95% CI 8.95 to 21.53); and negative likelihood ratio was 0.083 (95% CI 0.060 to 0.114).
Analysis of the 3x3 data showed 95.3% of arterial segments were correctly staged/classified, 3.1% were overstaged, and 1.6% were understaged.
Results of several subgroup analyses were reported.