Follow-up after embolization of ruptured intracranial aneurysms: A prospective comparison of two-dimensional digital subtraction angiography, three-dimensional digital subtraction angiography, and time-of-flight magnetic resonance angiography

The risk of complications [21] and the high cost of DSA [22] raised the need for alternative imaging methods of follow-up after embolization of intracranial aneurysms. In previous studies, the diagnostic value of MRA has been determined versus various DSA techniques. Reference methods included 2D acquisitions in standard or working projections [10, 11], 3D-DSA [13, 14], or undefined combination of 2D and 3D-DSA [12, 23]. Despite numerous false results of 2D-DSA in the follow-up of embolized aneurysms presented by several authors [13‐19], the method was not prospectively validated to our knowledge. In our study, the diagnostic value of 2D-DSA, 3D-DSA and TOF-MRA were tested against a retrospective consensus of 2D-DSA and 3D-DSA. Three-dimensional DSA presented the best diagnostic performance as concerns both the assessment of residual flow in the aneurysm and the decision on retreatment. The 2D-DSA and TOF-MRA had similar diagnostic values. An unexpected observation was the low value of 2D-DSA in the determination of indications for re-embolization.

The consecutive recruitment of patients scheduled for their first follow-up imaging at 3 months after the embolization resulted in collection of a relatively homogenous study population. This might reduce the influence of sample size on the results. The sizes and locations of the aneurysms were similar to those in other series of patients [3, 10]. The only significant difference was the low number of aneurysms in the posterior circulation in our material. This was caused by the system of bleeding aneurysm treatment in our center, where embolization is a method of choice.

In order to quantify the diagnostic value of intra-arterial arteriography we used the retrospective consensus of two observers simultaneously evaluating 2D and 3D-DSA images. In fact, we found no other reasonable method to noninvasively verify the performance of 2D-DSA, which is commonly considered the golden standard for follow-up. Combining multiple test results to construct a reference standard outcome including deterministic predefined rules, consensus procedures and statistical modeling is an accepted approach in cases when the reference test is imperfect [24, 25]. Generally, evaluation of a new modality which would be more sensitive than the contemporary standard usually causes methodological and statistical problems that result in a low specificity of the new method. Good examples of this dilemma are attempts to verify the value of 3D-DSA in follow-up. A recent systematic review revealed only three papers directly comparing 2D-DSA and 3D-DSA [26]. In the article by Kiyosue et al. [16], 3D-DSA presented as virtual endoscopy yielded the sensitivity and specificity of 100 % and 58.3 %, respectively. Buhk et al. [27], who tested 3D-DSA reconstructed as CT-like multiplanar reformations, reported a better diagnostic accuracy but considered this technique only as a supportive tool. The most comprehensive analysis was presented by Zhou et al. [17], who considered 3D-DSA as definitely more accurate and more reproducible than 2D-DSA despite the calculated moderate specificity and sensitivity of 3D-DSA vs. 2D-DSA used as a standard. In the three presented papers, the inappropriateness of classic test characteristics led to application of comparison statistics, including weighted kappa, chi-square test, and Pearson’s correlation coefficient, which are not able to valuate 3D-DSA [16, 17, 27].

According to our results, 3D-DSA should replace 2D-DSA as a reference method for follow-up, particularly in the process of decision-making on retreatment. The test characteristics of 2D-DSA for both the detection of the residual flow and the decision-making on possible retreatment were lower than those of 3D-DSA and comparable with those of TOF-MRA. The observed PPV of 2D-DSA indicates that only in 75 % of cases, the decision on retreatment was confirmed by the reference test. The low diagnostic value of 2D-DSA would have probably improved if working projections had been used. Working projections, which are used during embolization, are useful especially in detection of the aneurysm neck remnant [10]. However, a precise reconstruction of those projections at follow-up may be difficult even in center where embolization took place and may lead to an excessive exposition to radiation and contrast media. Still, when a precise 2D working projection is necessary, i.e., for re-embolization, a solution may be the use of 3D-DSA again. Several angiographic systems offer an easy c-arm positioning according to the 3D image presentation set by the user. This option is especially useful in retreatment of lesions located at the tip of the basilar artery or in the distal middle cerebral artery, where branch arteries may obscure the aneurysm neck.

In our series, TOF-MRA presented as an efficient follow-up method. Despite the lower values of test characteristics than that of 3D-DSA, the difference in the diagnostic value was not statistically significant when measured with AUC. Another important observation was only one false-negative result of TOF-MRA in the determination of indications for retreatment, which resulted in high NPV. This reflects very low probability of omitting clinically significant aneurysm remnant when using TOF-MRA. Our results confirm previous reports. In the meta-analysis by Kwee and Kwee [18], the pooled sensitivity and specificity of TOF-MRA in detecting residual flow was 83.3 % and 86.8 %, respectively. In a recent large prospective study by Schaafsma et al. [10], AUC of TOF-MRA was 0.86 (95 % CI, 0.81–0.91).

A significant drawback of TOF-MRA shown in our material was its inability to visualize coil loops protruding into the parent vessel, which was related to the use of 1.5-T field strength for MRA. Such a complication can be visualized with TOF-MRA at 3 T, as presented by Yoneoka et al. [28]. Another observed limitation was susceptibility to artifacts caused by implanted intracranial stents. Similar artifacts had already been observed by several authors [27, 29, 30]. In our patients, these artifacts resulted in false stent lumen narrowing on MR angiograms that might suggest in-stent thrombosis or proliferation of neointima [31]. Stent-related artifacts were most evident on volume-rendered images. In case of a true in-stent stenosis, one may expect a decrease of opacity or a filling defect on 2D-DSA images and the filling defect on 3D-DSA images. More advanced cases of stenosis may present as a difference in contrast medium flow proximally and distally to the stent in 2D-DSA, which reflects a hemodynamically significant blood flow disturbance. In neither of our patients, the stent lumen narrowing was confirmed by the reference test. Therefore, TOF angiograms should always be analyzed including source images and MIP reconstructions. None of the aneurysms treated with stent-assisted coiling presented residual flow. Thus, we may not discuss the possible influence of stent-related artifacts on determination of residual flow. This problem requires further investigation.

Variation in the assessment of aneurysms was reflected by the intraobserver reproducibility and interobserver agreement, which were significant. The 2D-DSA method presented the best agreement, which may be explained by its high resolution and limited possibilities of image adjustment. Since TOF-MRA was analyzed using unreconstructed images, as well as MIP and VR reconstructions, the variability of results was more evident. The observer agreements were at least comparable to previous reports [10, 12, 15, 27].

Study limitations included the mentioned above homogeneity of the sample, which was a result of the applied method of enrollment. We included consecutive patients who were coming for their first scheduled follow-up imaging visit and no active recruitment was carried out. Therefore, patients who died in the early postoperative period and those with severe neurological complications were lost from follow-up. Theoretically, those subjects might have presented a different rate of aneurysm remnants and altered cerebral hemodynamics that might have changed results of the study. Moreover, the method of enrollment limited the study population to patients with ruptured aneurysms only. However, we considered the study population appropriate because test characteristics of imaging modalities appear not to depend on the general patient condition and the preoperative aneurysm status [10]. As mentioned above, the reference test used in the study may be questionable as well. Nevertheless, in our opinion it was the only possible method to test the diagnostic value of 2D-DSA and 3D-DSA. Finally, we performed follow-up DSA examinations by administrating contrast medium into the CCA. At our institution, selective ICA angiograms are not routinely used in diagnostic procedures because we attempt to minimize the patient’s risk as much as possible. Selective catheterization of the ICA may increase the risk of vasospasm in young patients and may result in atherosclerotic plaque fragmentation in the older ones. On the other hand, cerebral vessel opacity might be increased with a selective ICA angiogram. However, in our experience, the used parameters of contrast media flow provide adequate depiction of arteries and of the residual aneurysm flow.

In conclusion, our results indicate that 3D-DSA is the contemporary golden standard for follow-up after aneurysm embolization. On the other hand, the relatively high diagnostic value of TOF-MRA favors this non-invasive method instead of intra-arterial DSA for the routine outpatient follow-up. The use of 3D-DSA should be limited to selected cases, which include patients with contraindications to MR, and those with uncertain MRA results after stent-assisted coiling or with coil protrusion. The 3D-DSA method should be also used to confirm indications for retreatment.