Article Text

Download PDFPDF

Screen for intracranial dural arteriovenous fistulae with carotid duplex sonography
  1. L-K Tsai1,
  2. S-J Yeh1,
  3. Y-C Chen1,
  4. H-M Liu2,
  5. J-S Jeng1
  1. 1
    Department of Neurology and Stroke Centre, National Taiwan University Hospital and National Taiwan University College of Medicine, Taipei, Taiwan
  2. 2
    Department of Radiology, National Taiwan University Hospital and National Taiwan University College of Medicine, Taipei, Taiwan
  1. Correspondence to Professor J-S Jeng, Department of Neurology, National Taiwan University Hospital, No 7 Chung-Shan South Road, Taipei 100, Taiwan; jsjeng{at}


Objectives: Early diagnosis and management of intracranial dural arteriovenous fistulae (DAVF) may prevent the occurrence of stroke. This study aimed to identify the best carotid duplex sonography (CDS) parameters for screening DAVF.

Methods: 63 DAVF patients and 170 non-DAVF patients received both CDS and conventional angiography. The use of seven CDS haemodynamic parameter sets related to the resistance index (RI) of the external carotid artery (ECA) for the diagnosis of DAVF was validated and the applicability of the best CDS parameter set in 20 400 patients was tested.

Results: The CDS parameter set (ECA RI (cut-off point = 0.7) and internal carotid artery (ICA) to ECA RI ratio (cut-off point = 0.9)) had the highest specificity (99%) for diagnosis of DAVF with moderate sensitivity (51%). Location of the DAVF was a significant determinant of sensitivity of detection, which was 70% for non-cavernous DAVF and 0% for cavernous sinus DAVF (p<0.001). The above parameter set detected abnormality in 92 of 20 400 patients. These abnormalities included DAVF (n = 25), carotid stenosis (n = 32), vertebral artery stenosis (n = 7), intracranial arterial stenosis (n = 6), head and neck tumour (n = 3) and unknown aetiology (n = 19).

Conclusion: Combined CDS parameters of ECA RI and ICA to ECA RI ratio can be used as a screening tool for the diagnosis of DAVF.

View Full Text

Statistics from

Intracranial dural arteriovenous fistulae (DAVF) is a subtype of vascular malformation and consists of abnormal shunts between dural branches of extracranial arteries and the intracranial venous sinus or meningeal veins.1 Patients may be asymptomatic or experience symptoms ranging from mild (eg, pulsatile tinnitus and chemosis) to fatal stroke.2 As catheterised embolisation therapy can eliminate DAVF, early diagnosis and management of DAVF may prevent the occurrence of stroke. Cerebral catheter angiography is a standard diagnostic tool for DAVF, and head CT and MRI have a diagnostic sensitivity of 70–81%.3 However, their invasiveness and high cost are disadvantages for DAVF screening. Therefore, a simple and convenient screening method for DAVF is still needed, especially for screening asymptomatic and mildly symptomatic patients who may possibly develop aggressive manifestations in the future.4

Diagnosis of DAVF by ultrasonography depends on haemodynamic changes in the feeding arteries and drainage veins as well as shortened cerebral circulation time.5 6 7 8 9 In one study using carotid duplex sonography (CDS), 74% of DAVF patients had a low resistance index (RI) in the external carotid arteries (ECA).7 As CDS has been widely used to evaluate the severity of atherosclerosis in carotid arteries,10 simultaneous analysis of the ECA RI during routine CDS may detect occult DAVF in some patients. However, using only ECA RI as a parameter to detect DAVF has led to a false positive rate of 21%,7 possibly because some normal subjects have generalised low flow resistance and thus show low RI values in the ECA. Several other haemodynamic parameters of CDS have also been used for diagnosis of DAVF.7 Our hypothesis was that combining different parameters to diagnose DAVF could reduce this false positive rate. In this study, we first validated the use of seven parameter sets for diagnosis of DAVF. We then chose the best parameter set and tested its usefulness as a screening tool in a large group of patients.

Patients and methods


This study had two parts. Part I was a validation study. We included 63 consecutive adult patients (older than 20 years) with intracranial DAVF who were evaluated by CDS between June 1995 and December 2007. A diagnosis of DAVF was confirmed by a neuroradiologist using results of cerebral catheter angiography in each patient. The types of DAVF were defined according to Cognard’s classification.11 The veins draining the fistulae did not include cortical veins in types I and IIa (relatively benign forms of DAVF) but did include cortical veins in types IIb, IIa+b, III, IV and V (considered to be pernicious forms of DAVF).11 12 The timing of CDS studies was within 2 months before or after the angiographic studies in all except two cases (which were 3 months earlier than angiography). In addition, during the same period, 170 adult patients (96 men; mean age 53 (16) years) without DAVF, confirmed by a complete cerebral catheter angiography study, were enrolled. These patients also underwent CDS 3 months before or after angiography. Their angiographic diagnosis included normal findings (n = 39), internal carotid artery (ICA) stenosis (n = 41), intracranial aneurysm (n = 20), stenosis of the middle cerebral artery (n = 20), brain tumour (n = 5), stenosis of the vertebral artery (n = 30), direct carotid–cavernous shunt (n = 3), subclavian steal phenomenon (n = 2), intracerebral arteriovenous malformation (n = 4), cerebral sinus thrombosis (n = 3), extracranial arteriovenous fistulae (n = 2) and high jugular bulb (n = 1). Medical records were reviewed in detail retrospectively to obtain demographic data and clinical manifestations. On the basis of clinical manifestations, DAVF was separated into “aggressive” and “non-aggressive” forms, which is similar to the classification in some previous reports.3 11 12 The aggressive group included patients with intracranial haemorrhage, venous infarction, seizure, altered mental status or intracranial hypertension. The non-aggressive group included patients who had tinnitus, orbital symptoms not related to intracranial hypertension and cranial neuropathy. CDS and angiography studies were performed and interpreted by different colleagues who were blinded to the clinical information of the patients.

Part II was a study of the applicability of the best CDS parameter set identified in part I. This parameter set was tested retrospectively in 20 400 consecutive cases from March 2003 to March 2007 who had received the CDS study in our hospital. When the selected CDS parameter set identified an abnormality in certain cases, the diagnosis was retrospectively obtained from the medical records and by head CT/MRI and/or angiography. The aim of this part of the study was not to validate the parameter set but to try to find some possible diagnosis leading to false positives. The studies were approved by the ethics committee of the National Taiwan University Hospital (IRB No 200812095R).

CDS study

The CDS study was performed using the HP 4500 series system (with a 3–11 MHz real time B mode imaging transducer and a 3.6 MHz pulsed Doppler transducer), the Diasonic VST Master series system (with a 10 MHz real time B mode imaging transducer and a 6 MHz pulsed Doppler transducer) or the Aloka SSD-3000 series system (with a 7.5 MHz real time B mode imaging transducer and a 5 MHz pulsed Doppler transducer). Peak systolic velocity (PSV), end diastolic velocity (EDV) and RI of the extracranial ECA (proximal to the first arterial branch) and ICA were measured. RI was defined as (PSV − EDV)/PSV and represented flow resistance.13 We analysed seven ECA RI related CDS parameter sets because RI of ECA is the best CDS parameter for predicting DAVF.7 14 The abnormal cut-off points were based on reference control data obtained from 180 adult subjects, as previously described.7 The seven parameter sets and their norms were: (1) a, ECA RI >0.7; (2) b, ICA to ECA ratio of RI <0.9; (3) c, ratio of bilateral ECA RI between 0.9 and 1.1; (4) a and b; (5) a and c; (6) b and c; and (7) a and b and c. We evaluated the haemodynamics in the bilateral carotid arteries. Abnormality on either side was considered abnormal.

Statistical analysis

A Fisher’s exact test was used to determine differences in categorical data between the groups. The t test was used to determine differences in age. Two sided p values of less than 0.05 were considered to indicate a statistically significant difference. Sensitivity, specificity, positive predictive value and negative predictive value were determined in each CDS parameter set. STATA software (V.8.0; Stata, Texas, USA) was used for the statistical analyses.


The 63 DAVF patients who underwent CDS (23 men, 40 women) had a mean age of 54 (15) years (range 23–82) at diagnosis. The locations of DAVF were cavernous sinus and non-cavernous sinus in 17 (27%) and 46 (73%) patients, respectively.

Table 1 shows validation of the method using different CDS parameter sets to diagnose DAVF. The parameter ECA RI was the most sensitive (68%) for predicting DAVF. However, the combination of ECA RI and ICA to ECA ratio of RI had the highest specificity (99%) and positive predictive value (97%) but moderate sensitivity (51%). Our purpose was to determine which CDS parameter was optimal as a screening tool for DAVF. To prevent overdiagnosis and to achieve the highest positive predictive accuracy, we chose the parameter set ECA RI and ICA to ECA ratio of RI for further analysis.

Table 1

Diagnostic performance of the duplex sonographic parameters compared with catheter angiography

Table 2 shows the characteristics of patients in two DAVF groups divided on the basis of abnormal and normal CDS results (ie, by analysis of ECA RI and the ICA to ECA RI ratio). Location of the DAVF had a significant impact on diagnostic sensitivity (0% in cases of cavernous sinus DAVF and 70% in cases of non-cavernous sinus DAVF; p<0.001). Gender had a similar impact (with diagnostic sensitivity higher in females (63%) than in males (30%); p = 0.02). On the other hand, age, clinical manifestations and type of DAVF had no influence.

Table 2

Characteristics of patients and dural arteriovenous fistulae according to duplex sonographic results

Figure 1 shows the results of the application study. Abnormal results were detected in 92 (0.5%) of the 20 400 cases using the ECA RI and ICA to ECA RI ratio parameter set. Among them, 46 (50%) also had other abnormal CDS findings, including severe stenosis or occlusion of the common carotid artery (n = 23), ICA stenosis (n = 9), vertebral artery stenosis (n = 7), low flow velocity and high RI in the ICA indicating distal ICA or middle cerebral artery stenosis (n = 6)15 and mass lesion (n = 1). In the 46 patients without other CDS abnormalities, the diagnoses were DAVF (n = 25), head and neck tumour (n = 2) and unknown aetiology (n = 19). The diagnoses of distal ICA or middle cerebral artery stenosis, head and neck tumour, and DAVF were confirmed by head MRI, CT or angiography.

Figure 1

Diagnosis of patients categorised by results of carotid duplex sonography. CCA, common carotid artery; CDS, carotid duplex sonography; DAVF, dural arteriovenous fistulae; ECA, external carotid artery; ICA, internal carotid artery; MCA, middle cerebral artery; RI, resistance index; VA, vertebral artery.


Patients with DAVF may develop intracranial haemorrhage, cerebral sinus thrombosis or venous infarct as their first presentation,11 while DAVF should reasonably exist for a long time. Early diagnosis of DAVF by screening may thus prevent these aggressive manifestations, especially in asymptomatic and initial mildly symptomatic patients. Here we showed the very high specificity with moderate sensitivity of the CDS parameter set (ECA RI plus ICA to ECA RI ratio) for diagnosis of DAVF. Patients planning to undergo routine CDS studies (which are widely used to evaluate the severity of atherosclerosis in carotid arteries10) should also undergo haemodynamic studies of the bilateral ECAs and ICAs to detect and treat occult DAVF as soon as possible.

Among CDS haemodynamic parameters, ECA RI had the highest diagnostic sensitivity for DAVF in a previous study (74%) and the present study (68%).7 Therefore, if the diagnosis of DAVF is suspected in patients with equivocal symptoms such as tinnitus or chemosis, we can use ECA RI to enhance our confidence to arrange more invasive and expensive studies for diagnosis of DAVF. However, its positive predictive accuracy in the previous study (79%) and in the present study (78%) was not high,7 meaning that DAVF would be misdiagnosed in about one-fifth of patients who would receive an unnecessary extensive workup. To solve this problem, we combined two ECA RI related parameters and increased the positive predictive value to 97%. Using the ICA to ECA RI ratio can exclude patients with low RI in both ECAs and ICAs, as some normal subjects may have generalised low flow resistance.16

Location of DAVF was a major determinant of diagnosis by CDS. Because of the low diagnostic sensitivity (0%) for cavernous sinus DAVF, we suggest using other methods to screen this subgroup of patients, such as transcranial colour coded duplex sonography and head MRI.17 18 Furthermore, as most cases of cavernous sinus DAVF are benign,19 we concentrated on screening for non-cavernous sinus DAVF. After excluding cases of cavernous sinus DAVF from our analysis, we achieved a diagnostic sensitivity of 70% for the method that combined ECA RI with ICA to ECA RI ratio. Patient age, the presence or absence of aggressive disease, and types of DAVF had no influence on predictive value. Although the diagnostic sensitivity of cavernous sinus DAVF is quite low (0%) and most patients (94%) with cavernous sinus DAVF presented with non-aggressive symptoms, the overall diagnostic sensitivity for the non-aggressive group was still 50%. This may be because a high percentage (79%) of large sinus DAVF (eg, transverse sinus and sigmoid sinus) also belonged to the non-aggressive group. Large sinus DAVF usually yields high CDS diagnostic sensitivity (89% in this study).7 For a still unknown reason, diagnostic sensitivity was significantly higher in females than males (63% vs 30%).

In a large patient population, the test using ECA RI with the ICA to ECA RI ratio detected abnormal CDS results in 0.5% of patients, half of whom also had other CDS abnormalities. These included:

  1. Stenosis or occlusion of the common carotid artery. Normally, blood flow goes through the common carotid artery to both the ECA and ICA. Therefore, stenosis or occlusion of the common carotid artery decreases flow resistance with low RI at the ECA.20 In some patients with occlusion or high grade stenosis in the common carotid artery, ICA flow may be supplied by retrograde ECA flow, which comes from collateral ECA branches. In these patients, CDS reveals the reverse flow and low RI in the ECA (fig 2A, B).20

  2. ICA stenosis.

  3. Vertebral artery stenosis. Long term cerebral hypoperfusion due to ICA or vertebral artery stenosis may lead to collateral cerebral blood flow from the branches of the ECA. Because the flow resistance in the cerebral circulation system is much lower than the extracranial circulation system,13 the collateral cerebral blood flow from the branches of the ECA will reduce the ECA RI (fig 2C, D).

  4. Low flow velocity and high RI in the ICA. These findings may indicate severe stenosis of the distal ICA or middle cerebral artery that stenosis related high flow resistance distally reduce the flow velocity and increase RI in the proximal ICA.15 High ICA RI increases the ICA to ECA RI ratio and again, the possible ECA collateral flow to intracranial cerebral arteries resulting from arterial stenosis may decrease the ECA RI.

  5. Head and neck tumours. Highly vascular tumours may have abundant blood supply from branches of the ECA, which reduce the ECA RI. Some mass lesions may be detected by neck sonography (fig 2E, F). On the other hand, DAVF was detected in more than half (54%) of patients with abnormal ECA RI, abnormal ICA to ECA RI ratio but no other CDS abnormality. Further extensive studies such as head CT, MRI or catheter angiography were necessary to confirm the diagnosis of DAVF in this group of patients.

Figure 2

Examples of patients with low resistance index (RI) in the external carotid artery (ECA), and high internal carotid artery (ICA) to ECA ratio of RI by carotid duplex sonography (CDS). (A, B) A 20-year-old girl had Takayasu arteritis with severe stenosis in the common carotid artery (CCA) noted by conventional angiography and colour coded duplex sonography (CCDS). CDS showed trickle flow in the CCA and reverse flow with low RI in the ECA. EDV, PSV. (C, D) A 62-year-old man had severe stenosis in the left proximal ICA noted by CCDS. Conventional angiography of the left CCA showed total occlusion of the left ICA at the cervical portion and collateral flow from the ECA branches to the left distal ICA and middle cerebral artery (MCA). (E, F) A 36-year-old man had oropharyngeal cancer. CCDS revealed a hypoechoic mass lesion near the carotid arteries. Head MRI showed a large oropharyngeal tumour with abundant vascular supply.

The diagnostic sensitivity is not high (51%) using the CDS parameter set of ECA RI plus ICA to ECA RI ratio for the diagnosis of DAVF. Therefore, we do recommend this parameter set to confirm a diagnosis of DAVF. Instead, we suggest applying this CDS parameter set on patients planning to undergo routine CDS studies for reasons other than to detect occult DAVF. There were only 25 of 20 400 cases (0.1%) with DAVF in the application study. Because large numbers of patients will undergo a CDS study for various reasons (about 6000 cases per year in our hospital), it is still worth screening for DAVF by CDS. Furthermore, the final diagnosis in some cases in our large population screening was not obvious because further examinations were not performed. These patients may be simply false positive cases or they may have undiagnosed DAVF. However, the probability of DAVF in patients with abnormal ECA RI, abnormal ICA to ECA RI ratio but no other CDS abnormality is already high enough to justify a further extensive workup. In addition, in this retrospective study, most of our patients with DAVF were evaluated by CDS after they experienced DAVF related symptoms. However, it is possible that CDS results could be normal in asymptomatic or mildly symptomatic DAVF patients before aggressive symptoms appear. We validated the diagnostic performances after summing up the data from the different patient groups with disease or non-disease, which may lead to some bias. Future prospective studies to confirm the usefulness of the CDS haemodynamic parameters for screening of DAVF are mandatory.

In summary, patients planning to undergo routine CDS for any reason should also undergo haemodynamic studies in the bilateral ECAs and ICAs. ECA RI less than 0.7, ICA to ECA RI ratio more than 0.9 and no other CDS abnormality (such as carotid/vertebral arterial stenosis) may indicate DAVF is present in at least half of the patients. Physicians should thus be alert to mild symptoms of DAVF (eg, pulsatile tinnitus and chemosis). Head contrast CT or MRI should also be performed. If the diagnosis of DAVF is highly suspected by the above clinical presentations or imaging findings, then catheter cerebral angiography should be carried out.


View Abstract


  • Competing interests None.

  • Ethics approval The study was approved by the ethics committee of the National Taiwan University Hospital (IRB No 200812095R).

  • Provenance and Peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.