OBJECTIVE: The purpose of this study was to determine the correlation between transcranial Doppler (TCD) velocities and angiographic vasospasm after aneurysmal subarachnoid hemorrhage. METHODS: In the first part of this study, patients were retrospectively reviewed to correlate middle cerebral artery absolute blood flow velocities with angiographic vasospasm. In the second part of the study, the middle cerebral artery/ipsilateral extracranial internal carotid artery velocity ratio (Lindegaard ratio) was prospectively correlated with angiographic vasospasm. Angiographic vasospasm was independently graded, by observers blinded to the TCD results, as either none, mild (less than one-third artery luminal narrowing), moderate (one-third to one-half narrowing), or severe (more than one-half narrowing). The sensitivity, specificity, likelihood ratios for positive and negative TCD results, positive and negative predictive values, and kappa and P values were calculated. RESULTS: One hundred one patients were analyzed in the first part of the study, and 44 patients were analyzed in the second part. Interobserver agreement regarding angiographic vasospasm was good (kappa = 0.86). Despite significant correlation between mean velocities and the degree of vasospasm, the clinical dependability of TCD velocities (evaluated using predictive values and likelihood ratios) was limited. The positive predictive value of velocities of > or =200 cm/s for moderate/severe angiographic vasospasm was 87% but that of lower velocities was approximately 50%. The negative predictive value of velocities of <120 cm/s was 94% but that of higher velocities was approximately 75%. Only the likelihood ratios for velocities of <120 or > or =200 cm/s were useful (likelihood ratio for negative result = 0.17, likelihood ratio for positive result = 16.39). Overall, 57% of patients exhibited maximum velocities in the indeterminate range between 120 and 199 cm/s. Lindegaard ratios did not improve the predictive value of TCD monitoring. CONCLUSION: For individual patients, only low or very high middle cerebral artery flow velocities (i.e., <120 or > or =200 cm/s) reliably predicted the absence or presence of clinically significant angiographic vasospasm. Intermediate velocities, which were observed for approximately one-half of the patients, were not dependable and should be interpreted with caution.