Subsequent Hemorrhage in Children with Untreated Brain Arteriovenous Malformation: Higher Risk with Unbalanced Inflow and Outflow Angioarchitecture
Online Supplement
Supplemental Results
Interobserver agreement
Interovserver agreement between the two neuroradiologists regarding generalized venous ectasia and arteriovenous shunt flow led to κ coefficients of 0.84 (95%CI, 0.73-0.94) and 0.87 (95%CI, 0.78-0.97), respectively (κ coefficients of other variables in Supplementary Table 3). For the discrepancies of categorical variables, we have the two neuroradiologist review the images again until they reach a consensus. For the continuous variable (bAVM size), the means of values from two observers were used for further analysis as long as the inter-observer deviation of the values was lower than a threshold (5mm). If the inter-observer deviation was greater than the threshold (5mm), the two radiologists were requested to perform the measurement again until the values were eligible for further analysis.
Association between bAVM size and subsequent hemorrhage
The mean bAVM size was not significantly different between children with or without subsequent hemorrhage (Supplementary Table 1), and univariate Cox regression also did not find association between bAVM size and earlier subsequent hemorrhage (Table 3). Then we classified the bAVM size into three category: <30mm, 30~50mm, and >50mm. 28 bAVMs (25.5%) were small (<30mm), 50 (45.4%) were medium (30-50mm), and 32 (29.1%) were large (>50mm). In 9 children with subsequent hemorrhage, 3 of them harbored a small bAVM, accounting for 10.7% of small bAVM and yielding an annual hemorrhage rate of 5.2% during 58.1 patient-years of follow-up; 1 of them harbored a medium bAVM, accounting for 2% of medium bAVM and yielding an annual hemorrhage rate of 1.3% during 76.5 patient-years of follow-up; 5 of them harbored a large bAVM, accounting for 15.6% of large bAVM and yielding an annual hemorrhage rate of 6.2% during 96.7 patient-years of follow-up. Both the small and large bAVM tend to have higher hemorrhage rates than the medium bAVM. However, log-rank test and univariate Cox regression did not verify the difference of subsequent hemorrhage risk between large or small bAVM and medium bAVM.
Association between hemorrhage presentation and subsequent hemorrhage
Hemorrhagic presentation is identified as independent predictors of hemorrhage during follow-up of untreated bAVM[1; 2]. Most patients with a ruptured bAVM will undergo treatment to obliterate the lesion in order to prevent future hemorrhage. However, unruptured bAVM still requires a careful weigh of risk and benefit of treatment[3]. The time period from diagnosis to interventional treatment could be different between children with ruptured and unruptured bAVM, thereby affect hemorrhage risk evaluation during follow-up. The median treatment-free time after diagnosis is only 4.2 month in children with a ruptured bAVM, which is much shorter than that (12 month) in children with an unruptured bAVM. Although the difference is not significant during the follow-up for all patients, the earlier treatment for ruptured bAVM might result in an underestimate of its subsequent hemorrhage risk in natural course (Supplementary Figure 1). Therefore, our data should be interpreted with caution regarding the role of hemorrhagic presentation in predicting future hemorrhage.
Potential confounding effect of bAVM size on hemorrhage risk evaluation in different types of bAVMs
We observed the hemorrhage events and their correlation with bAVM size in different types of bAVM. In Type 2 bAVM, 1 hemorrhage event occurred in medium bAVM and 2 events occurred in large bAVM; In Type 3 bAVM, 1 hemorrhage event occurred in small bAVM and 1event occurred in large bAVM; In Type 4 bAVM, 3 hemorrhage events occurred in small bAVM and 2 events occurred in large bAVM. Therefore, the size of bAVM might not be correlated with the distinct hemorrhage risk of different types of bAVM.
Potential confounding effect of hemorrhage presentation on hemorrhage risk evaluation in different types of bAVMs
Since there was more hemorrhagic presentation from Type 1 bAVM to Type 4 bAVM, we also analyzed the treatment-free time of children with different type of bAVM (Supplementary Figure 1). The interventional treatment is initiated earlier in Type 4 bAVMs, while their subsequent hemorrhagic events are much more than other types of bAVM. This might suggest that the risk for subsequent hemorrhage in Type 4 bAVMs is not overestimated.
Children without interventional treatment for bAVM
In this pediatric bAVM series, 24 children (24/110, 21.8%) did not undergo treatment for bAVM. 18 of them (18/24, 75%) harbored a bAVM located in eloquent region, 10 of them (10/24, 41.7%) had a large (>5cm) bAVM and 13 of them (13/24, 54.2%) had a Spetzler-Martin Grade IV-V bAVM. There were 10 children with Type 1 bAVM, 8 with Type 2 bAVM, 2 with Type 3 bAVM and 4 with Type 4 bAVM.
References
1Gross BA, Du R (2013) Natural history of cerebral arteriovenous malformations: a meta-analysis. J Neurosurg 118:437-443
2Kim H, Al-Shahi Salman R, McCulloch CE, Stapf C, Young WL, Coinvestigators M (2014) Untreated brain arteriovenous malformation: patient-level meta-analysis of hemorrhage predictors. Neurology 83:590-597
3Lawton MT, Rutledge WC, Kim H et al (2015) Brain arteriovenous malformations. Nat Rev Dis Primers 1:15008
Supplemental Figures
Supplementary Figure 1.The proportion remaining treatment-free in childrenwith (A)hemorrhagic presentationvs.non-hemorrhagic presentationand those with (B) different subtypes of bAVMs.
Supplemental Tables
SupplementaryTable 1. Characteristics of patients according to the occurrence ofarteriovenous malformation ruptureduring follow-up*
Characteristic / No Rupture(n=101) / Rupture
(n=9) / Total
(n=110) / P value
Demographic
Gender / .52
Female / 34(33.7) / 4(44.4) / 38(34.5)
Male / 67(66.3) / 5(55.6) / 72(65.5)
Age at diagnosis(ys) / 11.24±4.41 / 7.00±3.39 / 10.89±4.48 / .006
Clinical
Hemorrhagic Presentation / 1.00
No / 53(52.5) / 5(55.6) / 58(52.7)
Yes / 48(47.5) / 4(44.4) / 52(47.3)
Radiological
Fast A-V shunt / .15†
No / 54(53.5) / 2(22.2) / 56(50.9)
Yes / 47(46.5) / 7(77.8) / 54(49.1)
Generalized venous ectasia / .09†
No / 43(42.6) / 7(77.8) / 50(45.5)
Yes / 58(57.4) / 2(22.2) / 60(54.5)
Venous drainage / .57†
Superficial / 75(74.3) / 8(88.9) / 83(75.5)
Any Deep / 26(25.7) / 1(11.1) / 27(24.5)
Deep location / .99†
No / 61(60.4) / 6(66.7) / 67(60.9)
Yes / 40(39.6) / 3(33.3) / 43(39.1)
Infratentoriallocation / .58‡
No / 95(94.1) / 8(88.9) / 103(93.6)
Yes / 6(5.9) / 1(11.1) / 7(6.4)
Eloquence / .49†
No / 52(51.5) / 3(33.3) / 55(50.0)
Yes / 49(48.5) / 6(66.7) / 55(50.0)
Associated aneurysm / .66‡
No / 94(93.1) / 8(88.9) / 102(92.7)
Yes / 7(6.9) / 1(11.1) / 8(7.3)
Nidus size / .058‡
30mm / 25(24.8) / 3(33.3) / 28(25.5)
30-50mm / 49(48.5) / 1(11.1) / 50(45.5)
>50mm / 27(26.7) / 5(55.6) / 32(29.1)
Maximal nidussize(mm) / 41.56±18.18 / 44.14±22.88 / 41.77±18.50 / .69
Spetzler-Martin Grade / .72§
I-II / 51(50.5) / 4(44.4) / 55(50.0)
III / 31(30.7) / 3(33.3) / 34(30.9)
IV-V / 19(18.8) / 2(22.2) / 21(19.1)
Treatment-free Follow-up
Total person-years / 186.0 / 45.4 / 231.4
Mean (range), years / 1.84(0.1-15.4) / 5.04(1.0-10.3) / 2.10(0.1-15.4)
*Table entries are No. (%)or mean ± SD.P value in boldface indicates statistical significance.
†P values are from the χ2 test (correction for continuity).
‡P values are from the Fisher’s exact test.
§P values are from the Mann-Whitney U test.
SupplementaryTable 2. Characteristics of patients according to the occurrence of hemorrhage at presentation*
Characteristic / Non-hemorrhage(n=58) / Hemorrhage
(n=52) / Total
(n=110) / P value
Demographic
Gender / 0.988
Female / 20(34.5) / 18(34.6) / 38(34.5)
Male / 38(65.5) / 34(65.4) / 72(65.5)
Age at diagnosis(ys) / 11.12±4.61 / 10.63±4.36 / 10.89±4.48 / 0.572
Radiological
Fast A-V shunt / 0.037
No / 35(60.3) / 21(40.4) / 56(50.9)
Yes / 23(39.7) / 31(59.6) / 54(49.1)
Venous ectasia / 0.005
No / 19(32.8) / 31(59.6) / 50(45.5)
Yes / 39(67.2) / 21(40.4) / 60(54.5)
Venous drainage / 0.001
Superficial / 51(87.9) / 32(61.5) / 83(75.5)
Any Deep / 7(12.1) / 20(38.5) / 27(24.5)
Deep location / 0.003
No / 43(74.1) / 24(46.2) / 67(60.9)
Yes / 15(25.9) / 28(53.8) / 43(39.1)
Infratentoriallocation / 0.881†
No / 55(94.8) / 48(92.3) / 103(93.6)
Yes / 3(5.2) / 4(7.7) / 7(6.4)
Eloquence / 0.252
No / 32(55.2) / 23(44.2) / 55(50.0)
Yes / 26(44.8) / 29(55.8) / 55(50.0)
Associated aneurysm / 0.873†
No / 54(93.1) / 48(92.3) / 102(92.7)
Yes / 4(6.9) / 4(7.7) / 8(7.3)
Nidus size / 0.312
< 30 mm / 12(20.7) / 16(30.8) / 28(25.5)
30~50 mm / 26(44.8) / 24(46.2) / 50(45.5)
>50mm / 20(34.5) / 12(23.1) / 32(29.1)
Maximal nidussize(mm) / 43.96±17.65 / 39.32±19.27 / 41.77±18.50 / 0.190
Spetzler-Martin Grade / 0.552§
I-II / 31(53.4) / 24(46.2) / 55(50.0)
III / 16(27.6) / 18(34.6) / 34(30.9)
IV-V / 11(19.0) / 10(19.2) / 21(19.1)
Treatment-free Follow-up
Total person-years / 144.3 / 87.1 / 231.4
Mean (range), years / 2.49(0.1-15.4) / 1.67(0.1-11.0) / 2.10(0.1-15.4)
*Table entries are No. (%)or mean ± SD.P value in boldface indicates statistical significance.
†P values are from the χ2 test (correction for continuity).
‡P values are from the Fisher’s exact test.
§P values are from the Mann-Whitney U test.
Supplementary Table 3
Radiological variables / κappa value / 95%CIDeep location / 0.96 / 0.91-1.01
Infratentorial location / 1.00 / 1.00-1.00
Eloquence / 0.96 / 0.91-1.01
Fast A-V shunt / 0.87 / 0.78-0.97
Generalized venous ectasia / 0.84 / 0.73-0.94
Deep venous drainage / 0.95 / 0.89-1.01
Associated aneurysm / 0.88 / 0.71-1.04