Diagnostic accuracy of unenhanced and gadolinium-enhanced magnetic resonance imaging for acute pulmonary embolism diagnosis: Results of the “IRM-EP” study

Context: Magnetic resonance imaging (MRI) has not been yet fully evaluated for diagnosing pulmonary embolism (PE). Aim: To evaluate MRI performance for PE diagnosis by reference to 64-detector CT angiography (CTA) in patients with clinical suspicion of PE. Intervention: MRI including, unenhanced ECG-gated, perfusion and angiographic sequences was performed within 24 hours from CTA and were interpreted by two independent readers blinded to CTA results. Measures: Sensitivity, specificity evaluated globally and for each sequence. Inter reader agreement evaluated with the kappa statistics. Results: Among 300 patients included, 274, with a conclusive CTA, completed the whole MRI protocol, of which 103 had PE on CTA (prevalence: 37.5%). 76 of the 274 MRI examinations (28%) were judged inconclusive by reader 1 and 83 (30%) by reader 2. Sensitivity and specificity of conclusive MRI on global readings were 84.5% (95% CI, 74.9% to 91.4%) and 99.1% (95% CI, 95.1% to 100.0%) for reader 1 and 78.7% (95% CI, 68.2% to 87.1%) and 100% (95% CI, 96.7% to 100.0%) for reader 2, respectively. Sensitivity was better for proximal (97.7 to 100%) than for segmental (68 to 91%) and sub segmental PE (21 to 33%). Angiographic sequences showed the highest performance and agreement (Kappa: 0.77). Unenhanced sequences, although less sensitive, showed high specificity (96.1%) and good agreement (Kappa: 0.62), whereas perfusion sequences showed lower sensitivity, specificity and agreement. Conclusion: MRI demonstrates high specificity, even for unenhanced sequence and acceptable sensitivity for PE diagnosis at the cost of a 28% to 30% rate of inconclusive result.