MRA versus digital subtraction angiography in acute subarachnoid haemorrhage: a blinded multireader study of prospectively recruited patients

We performed a blinded multireader study comparing MR angiography (MRA) wit h digital subtraction angiography (DSA) in 34 prospectively recruited patie nts who presented with acute subarachnoid haemorrhage (SAH). Two observers independently reviewed the MRA and DSA studies some months after clinical p resentation. Presence of an aneurysm was rated on a LC-point confidence sca le. Cases in which the initial interpretation of the observers varied were jointly reviewed to reach a consensus opinion. DSA was deliberately chosen not to represent the reference standard and the clinical course and surgica l findings were used to explain significant differences between the consens us readings of MRA and DSA. Diagnostic confidence and interobserver agreeme nt were, overall, higher on DSA than on MRA studies (kappa(DSA)= 0.64 versu s kappa(MRA)= 0.52 with 95 % CI for Delta = kappa(DSA)-kappa(MRA) [-0.06, 0 .31]). With both methods, discrepancies between observers were due to aneur ysms overlooked rather than false-positive readings by one observer. Diagno stic accuracy therefore improved when the readings of the two observers wer e combined, particularly for MRA. Intermethod agreement was only fair and s imilar for both readers (kappa (reader 1) = 0.37 versus kappa (reader 2) = 0.32 with 95 % CI for Delta = kappa (reader) (-kappa)(1) (reader 2) [-0.02, 0.11]). Both interobserver and intermethod agreements improved when the da ta were analysed on a per-study (positive or negative study) rather than on a per-aneurysm basis. Differences in the consensus reading were due to fiv e aneurysms (four single and one multiple) detected only with MRA and five (two single and three multiple) detected only with DSA. MRA and DSA should be regarded as complementary in the investigation of patients with acute SA H. DSA can no longer be regarded as the reference standard.