AVM Management Equipoise Survey: physician opinions regarding the management of brain arteriovenous malformations Kevin M Cockroft,1,2,3 Ki-Eun Chang,4 Erik B Lehman,3 Robert E Harbaugh1 1
Department of Neurosurgery, Penn State Hershey Medical Center, Hershey, Pennsylvania, USA 2 Department of Radiology, Penn State Hershey Medical Center, Hershey, Pennsylvania, USA 3 Department of Public Health Sciences, Penn State Hershey Medical Center, Hershey, Pennsylvania, USA 4 Penn State University College of Medicine, Hershey, Pennsylvania, USA Correspondence to Dr K M Cockroft, Department of Neurosurgery-EC110, PO Box 859, Penn State Hershey Medical Center, Hershey, PA 17033, USA; [email protected]
Received 1 November 2013 Revised 18 November 2013 Accepted 19 November 2013 Published Online First 6 December 2013
ABSTRACT Introduction Management of unruptured brain arteriovenous malformations (AVMs) is controversial, and conducting clinical trials in this area has been a challenge. We sought to determine which, if any, patient/AVM characteristics were most likely to inﬂuence clinicians’ decisions regarding management and clinical trial enrollment. Methods We performed an online survey (Survey Monkey) of members of the American Association of Neurological Surgeons (AANS) (n=4836) and Society of NeuroInterventional Surgeons (SNIS) (n=512). Physicians were asked to rate the likelihood that various patient/ AVM characteristics would inﬂuence their treatment decisions and recommendations for clinical trial enrollment. 10 hypothetical case vignettes were also provided, and respondents were asked to select a management recommendation. Results 277 (5.2%) responses were received. Characteristics that the majority (>50%) of respondents felt should lead to treatment included size 50%) of respondents felt should lead to clinical trial enrollment included size > 6cm, deep or eloquent location, and age 51–70 years. The only characteristic that the majority felt should not lead to enrollment was the presence of an intranidal aneurysm. For the 10 vignettes, the majority of respondents (>50%) favored a speciﬁc course of management in only ﬁve cases. Conclusions Clinicians’ opinions on which patient/ AVM characteristics are important for decisions regarding treatment or trial enrollment vary widely. These results suggest that there is no uniform opinion of clinical equipoise with regard to the management of brain AVMs.
To cite: Cockroft KM, Chang K-E, Lehman EB, et al. J NeuroIntervent Surg 2014;6:748–753. 748
Management of unruptured brain arteriovenous malformations (AVMs) is controversial.1 2 While many believe certain unruptured AVMs should be treated, others have gone so far as to say that the treatment of unruptured AVMs is experimental therapy.3 These concerns led to the creation of the ARUBA trial (A Randomized trial of Unruptured
Brain Arteriovenous Malformations), a prospective, unblinded, randomized, clinical trial comparing observational management to interventional therapy for unruptured brain AVMs.4 Enrollment in the trial was slower than expected, particularly in the USA. We hypothesized that one reason for the slow trial enrollment was a lack of consistent clinical equipoise among physicians who evaluate and treat patients with brain AVMs. In an effort to better understand this, we surveyed members of the American Association of Neurological Surgeons (AANS) and the Society of NeuroInterventional Surgeons (SNIS) with regard to which, if any, patient/AVM characteristics were most likely to inﬂuence physicians’ decisions regarding AVM management and clinical trial enrollment.
METHODS Survey After obtaining local institutional review board approval, we performed an online survey (Survey Monkey) of AANS (n=4836) and SNIS (n=512) members. Physicians were invited by email to participate in the survey, and survey participation was linked to the respondent’s computer to prevent duplicate survey completions. Responses were conﬁdential. In the initial portion of the survey, background physician practice data were collected. Physicians were then asked to rate, using a modiﬁed Likert Scale (table 1), the likelihood that various patient/AVM characteristics would inﬂuence their treatment decisions. Next, using a slightly different modiﬁed Likert Scale (table 1), physicians were asked to rate the degree to which various AVM/ patient characteristics would inﬂuence their recommendations for clinical trial enrollment. Finally, 10 hypothetical case vignettes were provided, and respondents were asked to select a management recommendation.
Analysis We determined the number and percentage of respondents agreeing to a given rating of likelihood. If over 50% of respondents agreed to the same likelihood rating, this was considered a ‘majority’ opinion. If 30–50% of respondents agreed to the same likelihood rating, this was considered a ‘signiﬁcant minority’ opinion. We also examined the relationship between certain respondent characteristics and the likelihood of the respondents recommending either treatment, or enrollment in a clinical trial. Speciﬁcally, we ﬁrst
Cockroft KM, et al. J NeuroIntervent Surg 2014;6:748–753. doi:10.1136/neurintsurg-2013-011030
Hemorrhagic stroke Table 1
Modified Likert Scale
Decision for treatment Very unlikely to Less likely Makes no treat=1 to treat=2 difference=3 Decision for clinical trial enrollment Very unlikely to Less likely Not sure or not enroll=1 to enroll=2 relevant=3
No of responses (%) More likely to treat=4
Very likely to treat=5
More likely to enroll=4
Very likely to enroll=5
Table 2 Background of respondents No of responses (%) Specialty Neurosurgery Neurology Neuroradiology Other Subspecialty interest Cerebrovascular disease Neuroendovascular therapy Neurovascular microsurgery Other Years in practice In-training 1–3 In-training >3 Practice 10 Board certified Yes No Type of practice In-training Academic Solo private Single specialty group private Multispecialty group private Military Other Society membership—all that apply AAN AANS AHA/ASA ASNR CNS CV section SIR SNIS SVIN Other Neuroendovascular training Yes No Vascular neurology fellowship Yes No Cerebrovascular surgery fellowship training Yes No
443 (86) 18 (3.5) 45 (8.7) 9 (1.7) 239 133 151 187
(51.4) (28.6) (32.5) (40.2)
31 (6.0) 46 (8.9) 52 (10.1) 79 (15.4) 306 (59.5) 384 (75.6) 124 (24.4)
Gamma knife radiosurgery training Yes 197 (40.0) No 296 (60.0) No of AVMs evaluated per year 20 113 (22.2) No of AVMs treated with or sent for surgical resection per year 20 36 (7.1) No of AVMs treated with or sent for endovascular embolization per year 20 45 (8.9) No of AVMs treated with or sent for stereotactic radiosurgery per year 20 30 (5.9) Participation in a clinical trial as an investigator or sub-investigator Yes 287 (56.8) No 218 (43.2) Likely to enroll a patient in a clinical trial studying brain AVMs without an angiogram Yes 154 (30.4) No 353 (69.6) AAN, American Academy of Neurology; AANS, American Association of Neurological Surgeons; AHA/ASA, American Heart Association/American Stroke Association; AVM, arteriovenous malformation; CNS, Cognitive Neuroscience Society; CV, cerebrovascular; SIR, Society of Interventional Radiology; SNIS, Society of NeuroInterventional Surgeons; SVIN, Society of Vascular and Interventional Neurology.
66 (12.8) 206 (40.1) 34 (6.6) 94 (18.3) 73 (14.2) 5 (1.0) 36 (7.0)
examined the relationship of respondent experience, practice type, and treatment volume with likelihood of recommending treatment, and separately with the likelihood of recommending clinical trial enrollment. For the purposes of this analysis, physician ratings of ‘more likely’ and ‘very likely’ (4 and 5 on the modiﬁed Likert Scale, respectively) were combined as indicating an overall category of ‘likely’ to recommend treatment or clinical trial enrollment. Due to the multiple comparisons involved, a p value of 50%) agreement while lightly shaded regions represent areas where a significant minority (30–50%) agreed. AVM, arteriovenous malformation.
Table 4 Characteristics influencing clinical trial enrollment decisions No of respondents (%) Patient/AVM characteristic
Likely NOT to enroll
Not relevant to enrollment
Likely to enroll
Small size (>3 cm) Intermediate size (3–6 cm) Large size (>6 cm) Deep location Superficial location Posterior fossa location Eloquent location Non-eloquent location Superficial venous drainage Deep venous drainage Single draining veins Multiple draining veins Age 70 years Venous aneurysm Seizures (not specified) Seizures (well controlled) Seizures (poorly controlled) Intranidal aneurysm
74 (33.5) 44 (19.9) 55 (24.8) 40 (18.1) 91 (40.9) 51 (23.1) 34 (15.3) 82 (37.3) 60 (27.1) 39 (17.7) 65 (29.3) 27 (12.1) 89 (40.1) 65 (29.3) 38 (17.2) 75 (34.6) 63 (10.7) 34 (15.3) 43 (19.6) 82 (37.3) 104 (47.5)
55 69 47 61 50 85 61 61 77 84 82 98 52 66 65 41 87 107 80 63 52
90 (41) 108 (48.8) 120 (54) 120 (54.3) 82 (36.8) 85 (38.5) 128 (57.4) 77 (35.0) 84 (38.0) 97 (44.1) 75 (33.8) 98 (43.9) 81 (36.5) 91 (41.0) 118 (53.4) 101 (46.6) 69 (31.5) 81 (36.5) 97 (44.0) 75 (34.1) 63 (28.8)
(24.9) (31.2) (21.2) (27.6) (22.4) (38.5) (27.4) (27.7) (34.8) (38.2) (36.9) (43.9) (23.4) (29.7) (29.4) (18.9) (39.7) (48.2) (36.4) (28.6) (23.7)
Darkly shaded regions represent areas of majority (>50%) agreement while lightly shaded regions represent areas where a significant minority (30–50%) agreed. AVM, arteriovenous malformation.
clinical vignettes are reported in table 5. We also examined the relationship of respondent experience, practice type, and treatment volume with the likelihood of recommending treatment and with the likelihood of recommending clinical trial enrollment (tables 6 and 7).
DISCUSSION Brain AVMs have long been thought of as complex clinical entities. A wide variety of natural history data have been reported, and numerous treatment options are available.5–12 Unfortunately, treatment morbidity and mortality rates also vary widely and are considered by many to be unreliable as the data come mainly from self-reported, often retrospective, case series. These and other issues have led some to believe that a prospective randomized clinical trial is needed assess the ‘true’ role of treatment for brain AVMs. This has been felt to be particularly important for those patients with unruptured AVMs where the margin between risk and beneﬁt for treatment over natural history may be at its smallest. The ARUBA trial is a prospective, randomized, clinical trial dealing with the management of unruptured brain AVMs.4 Any patient with an unruptured brain AVM is eligible for enrollment, so long as the enrolling physician feels that treatment to eventual eradication of the AVM is possible. An angiogram is not required for enrollment, and patients are randomized to either observational follow-up or some form of treatment. Treatments may include any one or a combination of open microsurgery, stereotactic radiosurgery, or endovascular embolization. Many concerns regarding the design of this trial have been raised.2 13–16 In particular, selection bias whereby patients with higher risk lesions are treated outside of the trial, thus resulting in a low risk sample, has often been cited as a potential problem. We suspect that at least part of the reason for this may be a lack of uniform clinical equipoise among physicians evaluating or treating patients with brain AVMs. Such varying opinions regarding optimal management may make it difﬁcult to obtain patients for a clinical trial, especially one dealing with a relatively rare disease such as brain AVMs, and may lead to a cohort of patients that is not representative of the overall disease population. In the case of brain AVMs, strong opinions regarding AVM hemorrhage risk may lead some practitioners to selectively enroll only those patients that they feel will be at low risk for future hemorrhage. These issues may threaten the external validity (ie, overall generalizability) of the trial. The results of this survey shed some light on the opinions of physicians who care for patients with brain AVMs. Although the ARUBA trial did not require a diagnostic angiogram for enrollment, the majority of survey participants responded that they would be unlikely to enroll a patient in a clinical trial without such a study. The ﬁndings also demonstrate considerable variation in the way physicians view AVM risk and the way in which AVM and patient characteristics inﬂuence management decisions. Of the 21 AVM/patient characteristics evaluated, 16 (76%) showed a majority (>50%) agreement with regard to a management recommendation. However, for ﬁve of these characteristics, a signiﬁcant majority of respondents felt completely differently. In addition, of the other ﬁve characteristics where there was no majority opinion, four showed signiﬁcant minorities of respondents (30–50%) taking opposing opinions. With regard to a recommendation for actual clinical trial enrollment, there was considerably less agreement among respondents. Of the 21 AVM/patient characteristics surveyed, majority agreement on clinical trial enrollment was found in only six (28%). Respondents showed pronounced disagreement (a signiﬁcant
Cockroft KM, et al. J NeuroIntervent Surg 2014;6:748–753. doi:10.1136/neurintsurg-2013-011030
Hemorrhagic stroke Table 5
Hypothetical clinical case vignettes Recommend treatment (n (%))
Case 1—36-year-old with an asymptomatic right frontal AVM, measuring 2.5 cm in maximal dimension. Angiogram shows superficial venous drainage and no associated aneurysms Case 2—36-year-old with a history of seizures and a right frontal AVM, measuring 4.5 cm in maximal dimension. Angiogram shows superficial venous drainage and a single intranidal aneurysm Case 3—46-year-old with an asymptomatic left thalamic AVM, measuring 2.5 cm in maximal dimension. Angiogram shows deep venous drainage and a single intranidal aneurysm Case 4—46-year-old with a history of seizures and a right frontal AVM, measuring 6.5 cm in maximal dimension. Angiogram shows superficial and deep venous drainage and multiple intranidal aneurysms. Case 5—56-year-old with an asymptomatic left thalamic AVM, measuring 3.5 cm in maximal dimension. Angiogram shows deep venous drainage and no associated aneurysms Case 6—56-year-old with worsening chronic headaches and left occipital AVM, measuring 3.5 cm in maximal dimension. Angiogram shows superficial and deep venous drainage and no associated aneurysms Case 7—66-year-old with an asymptomatic left thalamic AVM, measuring 2.5 cm in maximal dimension. Angiogram shows superficial venous drainage and no associated aneurysms Case 8—66-year-old with an asymptomatic right frontal AVM, measuring 3.5 cm in maximal dimension. Angiogram shows superficial venous drainage and a single intranidal aneurysm. Case 9—76-year-old with a left thalamic AVM, measuring 2.5 cm in maximal dimension. Angiogram shows deep venous drainage and no associated aneurysms Case 10—76-year-old with an asymptomatic left frontal AVM, measuring 4.5 cm in maximal dimension. Angiogram shows superficial and deep venous drainage with no associated aneurysms
Recommend observation (n (%))
Recommend clinical trial enrollment and randomization (n (%))
Not sure or do not have enough information (n (%))
Darkly shaded regions represent cases of majority (>50%) agreement while lightly shaded regions represent cases where a significant minority (30–50%) agreed. AVM, arteriovenous malformation.
minority for two different responses, or a signiﬁcant minority for one response and a majority for another) with regard to 15 of the 21 AVM/patient characteristics. Of the 10 hypothetical case vignettes, in only one case did a majority favor a speciﬁc mode of treatment without a signiﬁcant minority favoring a completely different management strategy. Interestingly, with the exception of a few AVM/patient characteristics (seizures, large AVM size, deep venous drainage, and deep location), practitioner’s years of experience, practice type, and yearly treatment volume had little impact on which characteristics
respondents felt were important for either management decisions or clinical trial enrollment. Although the results of the management recommendations portion of the survey may suggest a modicum of consistency among practitioners regarding AVM treatment recommendations, the clinical trial enrollment and case vignette portions suggest that this does not translate into a uniform standard for AVM trial enrollment. This variation probably contributed to the poor enrollment seen from US sites in ARUBA. The pattern of responses to the enrollment questions and case vignettes also
Relationship of experience, practice type, and treatment volume with likelihood of recommending treatment
Experience History of seizures Seizures (well controlled) Practice type Large size (>6 cm) Treatment volume Large size (>6 cm)
No (%) likely to recommend treatment 0–10 years 81 (70.4) 47 (40.8) Academic 47 (25.7) 10 years* 141 (52.2) 60 (22.1) Other* 92 (43.0) >11 52 (26.9)
OR (95% CI)
2.18 (1.37 to 3.47) 2.44 (1.53 to 3.01)
0.46 (0.30 to 0.70)
2.16 (1.44 to 3.23)
For the purposes of this analysis, physician ratings of ‘more likely to treat’ and ‘very likely to treat’ (4 and 5 on the modified Likert Scale, respectively) were combined as indicating an overall category of ‘likely to recommend treatment’. Due to the multiple comparisons, a p value of