The aim of this study was to assess transcranial Doppler (TCD) as a screening test for pulmonary arteriovenous malformation (PAVM) in patients with hereditary hemorrhagic telangiectasia (HHT). This retrospective study included suspected patients with HHT who were screened for PAVM with a TCD and a chest computed tomography (CT) study. The results of TCD and CT were compared to evaluate the usefulness of TCD for detecting PAVM. A TCD Spencer grade ≥3 was defined as positive for a significant right-to-left shunt (RLS). The diameter of the pulmonary arteries feeding the PAVM was measured by calipers from the CT study. In 86 subjects from 74 families with HHT, the sensitivity of TCD for identifying a PAVM at rest was 98% and post-Valsalva was 100%. Specificity was 58% and 35%, respectively, presumably due to pulmonary shunts too small to recognize on CT. Of the patients with HHT who were referred for embolization therapy for their PAVMs, all 20 had TCD grade ≥3. In patients who were diagnosed with a PAVM by chest CT, patients with TCD grade ≥5 had a significantly larger sum of artery diameters feeding the PAVMs compared to those with grade ≤4 (5.0 ± 3.2 mm vs 2.6 ± 1.9 mm, p = 0.01). In conclusion, a TCD examination for evaluating RLS is sensitive for identifying PAVM in patients with HHT and is useful in quantitating the degree of RLS flow. The sensitivity of the TCD examination makes it a useful screening test without radiation in HHT subjects to determine which patients need to undergo CT evaluation to identify PAVMs.
Hereditary hemorrhagic telangiectasia (HHT) is a genetic disorder in which control of vascular growth is abnormal and produces arteriovenous communications in the cutaneous, pulmonary, cerebral, gastrointestinal, and hepatic arterial beds. Pulmonary arteriovenous malformations (PAVMs) have been reported to occur in 15% to 33% of patients with HHT. Patients with PAVMs may experience consequences of right-to-left shunt (RLS) such as hypoxemia, transient ischemic attack, migraine, stroke, and cerebral abscess because of the passage of deoxygenated blood, aseptic, or septic emboli into the cerebral circulation. The incidence of stroke has been reported between 2.6% and 25% of patients with PAVMs, and the incidence of cerebral abscess has been reported between 5% and 9% of patients with PAVMs. Because of the high risk for cerebrovascular events, screening for asymptomatic PAVMs is recommended for patients with HHT. There have been several proposed methods for screening patients with HHT and their families for the presence of PAVMs, including chest computed tomography (CT) and transthoracic echocardiograms with agitated saline contrast. The aim of this study was to assess the usefulness of transcranial Doppler (TCD) to screen for PAVMs in patients with HHT. This study includes the largest number to date of patients with HHT for the assessment of TCD as a screening test for detecting PAVM.
Methods
Between July 2010 and November 2015, 312 suspected patients with HHT and family members were evaluated through the HHT Center of Excellence at University of California, Los Angeles (UCLA). There were 134 subjects who required screening for a PAVM and had a TCD examination; 86 of these also had a chest CT study. Medical charts of the patients were retrospectively reviewed for history, examination findings, and laboratory results to confirm the diagnosis of HHT. The diagnosis of possible or definite HHT was made according to the Curacao criteria. The Curacao criteria is a diagnostic tool that assigns a score of 0 to 4 based on the presence of epistaxis, telangiectasia, visceral lesions, and family history of HHT. The HHT diagnosis is considered definite if ≥3 criteria are present and possible if 2 criteria are present. There was no required genetic testing for the diagnosis of HHT, but there were subjects who had genetic testing performed for the diagnosis of asymptomatic family members. The study received prior approval by the UCLA Institutional Review Board for Human Research, and all patients signed informed consent.
TCD was performed using a power M-mode Terumo 150 PMD machine (Spencer Technologies, Redmond, Washington). Ultrasonic probes were mounted on a head frame and positioned over insonation windows located on the temporal bone. Individual probes on each side were manipulated to locate the middle cerebral artery. A mixture of 8-ml normal saline solution combined with 0.5 ml of air and 1 ml of blood was agitated between 2 syringes connected by a 3-way stopcock and injected into the brachial vein; embolic tracks were then counted over the middle cerebral arteries. The addition of blood increases the number of microbubbles produced and may increase the effectiveness of microbubble contrast. The degree of RLS was evaluated by TCD at rest and post-Valsalva maneuver. During Valsalva, patients were instructed to exhale at a sustained pressure of 40 mm Hg, aided by visual feedback with a manometer device; bubble injection commenced just before cessation of Valsalva. The Spencer Logarithmic Scale, which assigns a score of 0 to 5, was used to grade the results. Grade 3 or higher (≥31 microbubbles/min) was considered as a positive result representing a significant RLS. Grade 4 was defined as detection of 101 to 300 microbubbles per minute, grade 5 was defined as detection of >300 microbubbles per minute and microbubbles with a shower effect (too numerous to count) was considered as grade >5. In 12 patients who showed a shunt of grade 4 or ≥5 at rest, the injection was not repeated post-Valsalva. In one 11-year-old female, the bubble study was performed only post-Valsalva to avoid repeated bubble injections. Spencer TCD grades at rest and post-Valsalva were compared with the presence of PAVM on CT. The presence or absence of PAVM on CT was used as the standard to determine the sensitivity and specificity of the TCD examination. The highest TCD grade either at rest or post-Valsalva was compared with the size of the feeding arteries of the PAVMs.
Chest CT was used as the reference for the morphologic evaluation of PAVM. There were 30 patients who brought CT studies from their referring institutions. The remaining 56 patients underwent imaging with a CT scanner at UCLA (SOMATOM Definition Flash [n = 9]; Sensation 64 [n = 41]; Sensation 16 [n = 5]; Emotion 6 [n = 1], Siemens Healthcare, Germany). The diagnosis of PAVM was based on the typical appearance of a nodular opacity of variable size with both an afferent and efferent vessel on CT. The number and the diameter of the feeding arteries for the dominant PAVMs were measured using software calipers on the CT images. Measurements were obtained by a physician blinded to the results of TCD. The sum of the diameters of the arteries feeding the PAVMs were calculated and compared with the TCD grades of RLS flow. Patients were classified in groups according to the feeding artery diameter of the main PAVM (<2 mm, 2 to 2.9 mm, and ≥3 mm).
The recommendation for embolization therapy was based on the presence of a PAVM feeding artery diameter of ≥3 mm on CT, whereas <2 mm was considered to be too small for embolization. A PAVM with a feeding artery diameter between ≥2 and ≤2.9 mm received embolization therapy based on a discussion between the patient and the physician performing this therapy regarding the risks and benefits of undergoing the procedure versus watchful waiting.
To assess the utility of the Valsalva maneuver for distinction between PAVM and patent foramen ovale (PFO), the differences in TCD grades were compared between the patients with HHT with documented PAVM and a population of patients with PFO where TCD is routinely performed. The PFO population consisted of 140 patients who underwent a screening evaluation for an RLS with TCD at rest and post-Valsalva maneuver and had a PFO documented by a right heart catheterization.
Continuous variables were expressed as mean ± standard deviation. Statistical significance was determined using unpaired t tests for continuous variables. Pearson’s chi-square test or Fisher’s exact test were used for categorical variables; a value of p <0.05 was considered significant. Data were analyzed using JMP, version 9.0 for Windows (SAS Institute, Inc., Cary, North Carolina).
Results
Baseline patient characteristics are listed in Table 1 . There were 74 (86%) patients who were diagnosed with definite HHT after a complete evaluation, including CT study to identify any PAVM. Of those who were diagnosed with definite HHT, 40 (54%) patients had a PAVM on CT, whereas 3 (30%) patients diagnosed as possible HHT and 0 (0%) patients diagnosed as unlikely for HHT had a PAVM on CT. Visceral AVMs in multiple organs (lung, liver, gastrointestinal, or brain) were present in 15 (17%) patients. Genetic testing for the gene responsible for the clinical syndrome of HHT was performed in 54 (63%) patients, and 52 were found to have a known gene mutation; these consisted of 27 patients with mutation in the endoglin gene, 19 patients with mutation in the activin A receptor type II-like 1 gene, 1 patient with mutation in the SMAD4 gene, and 5 patients who had previous positive genetic testing but the results of the specific gene mutation were not available. A PAVM was documented on CT in 17 (63%) patients with endoglin gene mutations and 4 (21%) patients with activin A receptor type II-like 1 gene mutations.
| Clinical data | Value (N = 86) |
|---|---|
| Age (years) | 40 ± 19 (9-81) |
| Male | 32 (37%) |
| HHT according to number of Curacao diagnostic criteria present | |
| 4 | 41 (48%) |
| 3 | 33 (38%) |
| 2 | 10 (12%) |
| 1 | 2 (2%) |
| Epistaxis | 77 (90%) |
| Telangiectasia | 69 (80%) |
| Family history of HHT | 82 (95%) |
| Visceral arteriovenous malformations | 54 (63%) |
| Lung (PAVM) | 43 (50%) |
| Single vs. Multiple PAVMs | 18 vs. 25 |
| Brain | 9 (10%) |
| Liver | 12 (14%) |
| Gastrointestinal tract | 7 (8%) |
| Spine | 1 (1%) |
| Gene mutations for HHT | 54 (63%) |
| Endoglin gene | 27 (31%) |
| Activin A receptor type II-like 1 gene | 19 (22%) |
| SMAD4 | 1 (1%) |
| Unknown | 5 (6%) |
| No mutation found | 2 (2%) |
| Not performed | 32 (37%) |
| Referred for embolization therapy of PAVMs | 20 (23%) |
A history of ischemic stroke was present in 5 (6%) patients and 3 (3%) patients had a history of cerebral bleeding. There was 1 patient (1%) with a PAVM who had a history of cerebral abscess. A history of migraine was reported in 35 (41%) patients and an additional 5 (6%) patients had a transient neurologic deficit, which may be difficult to discern from a complex migraine. A PAVM was more common in migraineurs than nonmigraineurs (66% vs 39%, p = 0.03), but the Spencer grade of RLS was not statistically different between the migraineurs and nonmigraineurs (3.4 ± 1.8 vs 2.9 ± 1.6, p = 0.23 for rest; 3.5 ± 1.5 vs 3.0 ± 1.2, p = 0.17 for post-Valsalva).
The sensitivity and specificity of the TCD grade compared with CT is shown in Table 2 . Defining TCD grade ≥3 as positive for RLS, the sensitivity at rest was 98% and post-Valsalva was 100%. One patient with a small PAVM with a feeding artery of 2 mm had a TCD grade 1 at rest, which increased to grade 3 post-Valsalva. The effect of using different thresholds, grade ≥4 or ≥5, for the definition of a positive TCD is also presented. As the cutoff for a positive TCD grade increases, the specificity increases at the expense of sensitivity.
| Definition of positive RLS on TCD | |||
|---|---|---|---|
| Grade ≥3 | Grade ≥4 | Grade ≥5 | |
| TCD grades at rest, N = 85 ∗ | |||
| Sensitivity | 98% (41/42) | 79% (33/42) | 52% (22/42) |
| Specificity | 58% (25/43) | 84% (36/43) | 100% (43/43) |
| Positive Predictive Value | 69% (41/59) | 83% (33/40) | 100% (22/22) |
| Negative Predictive Value | 96% (25/26) | 80% (36/45) | 68% (43/63) |
| TCD grades post-Valsalva, N = 74 † | |||
| Sensitivity | 100% (31/31) | 81% (25/31) | 42% (13/31) |
| Specificity | 35% (15/43) | 84% (36/43) | 100% (43/43) |
| Positive Predictive Value | 53% (31/59) | 78% (25/32) | 100% (13/13) |
| Negative Predictive Value | 100% (15/15) | 86% (36/42) | 70% (43/61) |
∗ An 11-year-old patient underwent TCD only after Valsalva to avoid repeat bubble injection.
† There were 12 patients with TCD grade 4 or 5 at rest who declined to repeat after Valsalva.
Table 3 lists the measurement of the diameter of the PAVM feeding arteries by CT versus the result of the TCD studies. In 43 patients who were diagnosed with PAVMs by CT, patients with a TCD grade ≥5 had a significantly larger sum of artery diameters feeding the PAVM compared with those with grade 4 or less (p = 0.01). There were no patients with a grade 3 or less shunt who had a PAVM with a feeding artery diameter of ≥3 mm. Patients with TCD grade ≥5 were more frequently referred for embolization therapy compared with patients with TCD grade ≤4 (p = 0.03). There were 5 patients who were referred for embolization therapy with a TCD grade ≤4; these consisted of 1 patient with a TCD grade 3 with the main feeding artery diameter of 2.1 mm and 4 patients with a TCD grade 4 where the main feeding artery diameters were between 2.0 mm and 4.0 mm.
