Hemorrhage Risk of Brain Arteriovenous Malformations During Pregnancy and Puerperium in a North American Cohort
Background and Purpose—Conclusions reached in existing literature about risk of arteriovenous malformation (AVM) hemorrhage during pregnancy are controversial. This study compares the risk of hemorrhage in pregnant and nonpregnant female patients with AVM in a North American cohort.
Methods—We retrospectively reviewed female patients with AVM evaluated from 1990 to 2015. Exposure period for pregnancy and puerperium was calculated as 40 and 6 weeks, respectively, for each full-term pregnancy and 6 weeks for each abortion. Hemorrhage events and patient-years were calculated during an exposure period (pregnancy and puerperium), and a nonexposure period defined as either the interval from birth until AVM obliteration or until last follow-up after subtracting exposure period. Poisson rate ratio test was used to compare rate of hemorrhage between exposure and nonexposure periods.
Results—For 270 female patients with AVM, mean age was 35.0±19.6 years, and race distribution was white (n=165, 61.1%), black (n=59, 21.9%), Hispanic (n=9, 3.3%), Asian (n=6, 2.2%), and other (n=31, 11.5%). From 191 total pregnancies occurring before AVM obliteration, there were 175 (91.6%) term deliveries and 16 (8.4%) abortions. Overall annual hemorrhage rate for 149 total hemorrhages during an average of 11 097 patient-years was 1.34%. There were 140 hemorrhages in nonexposed women and 9 hemorrhages in pregnant women, translating to an annual hemorrhage rate of 1.3% in nonpregnant women versus 5.7% in pregnant women (P<0.001). Identical analysis for reproductive age patients (15–50) demonstrated a rate of 1.3% versus 7.0% (P<0.001).
Conclusions—Our results conflict with those from a recent study describing no increased rate of rupture during pregnancy. This difference may reflect unique population attributes influencing brain AVM hemorrhage during pregnancy.
Between 5% and 12% of all maternal deaths during pregnancy and puerperium are attributable to hemorrhagic stroke, which harbors a maternal mortality of 35% to 83%.1 Existing reports suggest that pregnancy and the puerperal periods are associated with increased hemorrhagic stroke risk,2–4 specifically from rupture of cerebral aneurysms and brain arteriovenous malformations (AVMs).5–7 Whether the hemorrhagic risk of brain AVMs increases during pregnancy and puerperium remains controversial. Recently, a retrospective study from China describing 393 patients with AVM rupture between ages 18 and 40 years reported no increased risk of intracranial AVM hemorrhage during pregnancy and the puerperal period.8 Bateman et al9 and Horton et al10 describe similar results. Conversely, Gross et al11 have reported an annual hemorrhage rate from brain AVMs during pregnancy of 8.1%, which is higher than that for nonpregnant female patients; however, their study is limited by sample size.
Accurate calculation of hemorrhage risk from brain AVMs is challenging because most studies use population-based statistics to determine at-risk person-years; therefore, factors including whether patients with a known diagnosis of AVM avoid pregnancy cannot be ascertained. Liu et al’s8 description of brain AVMs during pregnancy used a case-crossover analysis design to determine whether the pregnancy exposure period for an individual patient is associated with increased intracranial hemorrhage. Such analysis is made possible by literature describing AVMs as a genetic disease present from birth.12–15 Whether the discrepancies between studies describing pregnancy AVM hemorrhage rate are attributable to calculation methods or other underlying causes, such as population characteristics, is unknown; therefore, the extent to which Liu et al’s8 results can be generalized to other patient populations remains unclear. Provided the rarity of AVM hemorrhage during the relatively short time window of pregnancy, a larger sample size compared with previous North American studies is warranted to better delineate the risk of hemorrhage during pregnancy. The aim of this study is therefore to compare the risk of intracranial hemorrhage in pregnant and nonpregnant female patients with AVM in a large East Coast North American cohort.
Materials and Methods
This study has been approved by our institutional review board. We performed a retrospective review of patients with brain AVM evaluated at our institution from 1990 to 2015. Female patients with pregnancies before AVM obliteration were included. Nine patients with 9 AVM ruptures occurring between the ages of 15 and 50 years were used for secondary analysis to capture patients exclusively within the reproductive age group defined by the US Census Bureau.16 Patients with incomplete clinical and angiographic data, those with diagnosis of hemorrhagic hemolytic telangiectasia, and those lost to follow-up were excluded from the study.
Definition of Variables
We collected patient baseline demographic and clinically related variables. Age was defined as maternal age at time of AVM rupture. Race was categorized as white, black, Hispanic, Asian, and other. AVM rupture was defined as intracerebral hemorrhage or bleeding attributable to brain AVM at presentation. Angiographic features of AVMs were obtained and defined as previously reported.17,18 For patients with AVM rupture during pregnancy or puerperium, we obtained obstetric parameters, including parity, gestational age at rupture, delivery modality, and measures of maternal outcome. Maternal functional outcome was assessed using the modified Rankin Scale. Treatment modality was defined as surgery±embolization, radiosurgery±embolization, embolization only, surgery+radiosurgery±embolization, and medical management.
The primary outcome measure of this study was the annual hemorrhage rate during the pregnancy and nonpregnancy period. Hemorrhages were counted as AVM-related hemorrhage before AVM obliteration. Exposure period for pregnancy and puerperium was calculated as 40 and 6 weeks, respectively, for each full-term pregnancy and 6 weeks for each abortion. Exposure period calculations were performed using all pregnancies, puerperal periods, and abortions that occurred before documented AVM obliteration. Incidence of hemorrhage can therefore be calculated using hemorrhage counts and patient-years during an exposure and nonexposure period, with the latter defined as the interval from birth until obliteration of AVM or last follow-up after subtracting the exposure period.12–15
We compared characteristics between the exposure and nonexposure groups using ANOVA. Annual hemorrhage rate was obtained by determining the ratio of hemorrhage counts to patient-years of follow-up during either the exposure or nonexposure period. We performed case-crossover analysis of patients with ruptured AVM, with each patient serving as their own control, to obtain an odds ratio of hemorrhage frequency during the exposure and nonexposure periods (Figure 1). Poisson rate ratio test was then used to compare the rate of hemorrhage during the exposure and nonexposure periods. We performed statistical analysis twice: once for all patients and once for patients aged 15 to 50 years. Statistical significance was defined as P<0.05, and all P values were reported as 2 sided. Statistical analyses were performed using R Statistical Software (version 3.2.3, Vienna, Austria).
Overall Patient Characteristics and Treatment Outcomes
From our database of 763 sporadic AVMs, we identified 417 female patients. After exclusion criteria screening, the final cohort included 270 patients. Detailed patient selection is depicted in Figure 2. Mean patient age was 35.0±19.6 years, with most being white (n=165, 61.1%). Of all patients, 31.9% presented with AVM rupture. Average modified Rankin Scale at presentation was 2.23 (Table 1). Mean AVM size was 3.18±1.96 cm (range, 0.05–14.0 cm), and Spetzler–Martin grade distribution was grade 1 (n=27, 10%), grade 2 (n=86, 31.7%), grade 3 (n=96, 35.4%), grade 4 (n=42, 15.5%), and grade 5 (n=19, 7.0%; Table 2).
AVM management modalities included surgery±embolization (n=61, 22.6%), radiosurgery±embolization (n=132, 48.9%), embolization only (n=14, 5.2%), surgery+radiosurgery±embolization (n=7, 2.6%), and medical management (n=56, 20.7%). Obliteration of AVM was achieved in 44.1% of all patients, 88.3% of surgery±embolization patients, and 40.7% of radiosurgery±embolization patients. During an average follow-up of 5.1±5.0 years, 58.9% of patients experienced satisfactory functional outcome with modified Rankin Scale ≤1.
Characteristics of Patients With Hemorrhage During Pregnancy or Puerperium
Nine (3.33%) female patients experienced AVM rupture during pregnancy or puerperium. Mean age at AVM hemorrhage was 29.7 years. Mean AVM size in exposed group patients was 3.2 cm (range, 1.0–7.0 cm) and most were Spetzler–Martin grade 4 (n=5, 55.6%). Five of 9 patients were nulligravid before the hemorrhagic event (n=5, 55.6%). Across all patients, 191 total pregnancies occurred before AVM obliteration, 175 (91.6%) of which were term or pre-term deliveries and 16 (8.4%) were abortions. Detailed descriptions of patient characteristics are included in Table 3.
Overall Hemorrhage Risk and Outcomes
From 149 total hemorrhages >11 097 patient-years, the overall annual hemorrhage rate was 1.34%. Among the 149 hemorrhages, 140 occurred during the nonexposure period (10 937.8 patient-years) and 9 in the exposure period (158.9 patient-years), with 1 hemorrhage during puerperium. All intrapregnancy AVM ruptures occurred in the second (n=5, 62.5%) or third trimester (n=3, 37.5%). Hemorrhage and pregnancy characteristics are summarized in Table 4. We also compared demographic and angiographic factors between the exposure and nonexposure periods (Table I in the online-only Data Supplement) and found no significant differences, suggesting that variables remain reasonably balanced between the nonexposure and exposure periods. It should be noted that age was not included in this comparison because it could not be accurately defined using a per-period analysis.
No fetal malformations attributable to intrapartum radiation exposure were reported. There were no fetal or maternal deaths during pregnancy or the puerperal period. The annual hemorrhage rate for the 140 nonexposed and 9 exposed hemorrhages, respectively, were 1.3% and 5.7% (rate ratio=4.43; confidence interval, 1.98–8.65; P<0.001). Identical analysis for reproductive age patients (15–50 years per the US Census Bureau) revealed a more distinct rate of 1.3% versus 7.0% (rate ratio=5.61; confidence interval, 2.48–11.18; P<0.001) for nonexposed and exposed patients, respectively. Management of AVM after pregnancy delivery was through surgery (n=2, 22.2%), radiosurgery (n=6, 66.7%), or embolization (n=1, 11.1%). Obliteration was achieved in 55.6% of these patients.
Risk of Hemorrhage Recurrence During Pregnancy
The consideration of pregnancy termination in the context of a ruptured brain AVM requires knowledge of hemorrhage recurrence. Of the 8 patients who hemorrhaged during pregnancy, 7 delivered after 38 weeks, and the remaining patient delivered at 29 weeks. All 8 patients with hemorrhage during pregnancy delivered via cesarean section. Two of the 8 patients who initially hemorrhaged during pregnancy experienced recurrent bleeding during the same pregnancy or in the puerperal period. After initial intraventricular hemorrhage presentation at 23 weeks of pregnancy, 1 patient (Table 3) experienced 2 recurrent intraventricular hemorrhages at weeks 27 and 29, which prompted pregnancy termination at 29. The second patient had a recurrent intracerebral hemorrhage during puerperium, 10 days after cesarean section.
Summary and Interpretation of Results
The goal of this study was to use a previously described case-crossover analysis in a sizeable, racially diverse North American population of AVM patients with hemorrhage during pregnancy. To this end, we presented a cohort of 270 female AVM patients with complete pregnancy history. To our knowledge, this is one of the largest North American cohorts describing the risk of AVM hemorrhage during pregnancy. The overall baseline characteristics and AVM angiographic features of our patient population are like those previously described. Moreover, our cohort’s racial distribution broadens the generalizability of our results to the US population. We found an overall annual hemorrhagic risk of 1.34%, which was 1.30% during the nonpregnant period and 5.7% during pregnancy and puerperium, suggesting an increased risk during the exposure period.
Natural History of Brain AVMs
The overall annual rate of hemorrhage in our study was 1.34%, which is similar to that reported by Gross et al,11 but lower than the 2% to 4% described by others.8,10,11 This discrepancy may be because of reporting of annual hemorrhage rates in the context of follow-up after initial hemorrhagic presentation, which consequently decreases follow-up time and inflates AVM hemorrhage rates. Like Gross et al,11 we assumed AVM presence since birth, which might explain our lower overall annual rate of hemorrhage. Although Liu et al8 also regarded AVMs as a congenital disease, they reported an annual hemorrhage rate of 2.71%, perhaps reflecting patient ascertainment bias and lack of follow-up once younger patients experienced a hemorrhagic event.
Hemorrhagic Risk During Pregnancy
The risk of brain AVM hemorrhage during pregnancy and puerperium remains contentious because reported rupture rates differ between studies. In the most recent large series of AVMs in pregnancy, Liu et al8 did not find an increased risk of hemorrhage during pregnancy. Using case-crossover analysis, the authors described 12 hemorrhages in 12 patients across 452 pregnancies, yielding an annual hemorrhage rate of 3.32%. Between the remaining 381 patients, the annual hemorrhage rate was 4.14% from 10 627 patient-years of follow-up. Although this study is distinguished by its large cohort size and use of case-crossover analysis, its generalizability to North American patients is limited in that it describes a solely Chinese population.
About North American cohorts, in a prospective, population-based study, Bateman et al9 found no substantially increased risk of intracranial hemorrhage from cerebrovascular malformations during pregnancy (0.50 versus 0.33 per 100 000 person-years for pregnant versus nonpregnant patients, respectively). However, the authors assessed all-causes of intracranial hemorrhage, of which AVMs were only a small portion and therefore its statistical use is limited. Conversely, Horton et al10 argued for no increased hemorrhage risk during pregnancy albeit with a calculated pregnancy time of 1 year to include an assumed full-term pregnancy and a 12-week puerperium. Although their cohort describes a North American population, it is distinct from the general AVM population in that the dominant treatment modality was radiosurgery, and a greater proportion of AVMs were classified as having a deep location.
More recently, Gross et al11 reported an increased risk of hemorrhage during pregnancy in a North American cohort. They included 54 female patients with 5 hemorrhages occurring in 62 pregnancies during follow-up, yielding an annual hemorrhagic risk of 8.1% during pregnancy. Although limited by sample size and an alternate methodology for follow-up period calculation, their results are concordant with our study’s finding that pregnancy increases the hemorrhagic risk of AVMs. The annual hemorrhage rate during pregnancy and puerperium in our study was 5.7% for all patients and 7.0% for patients aged 15 to 50 years. In another study limited to patients undergoing radiosurgery, the annual hemorrhage rate during pregnancy was reported as 9.3% for 191 patients as compared with 4.5% during the remaining reproductive years.19 Analysis of 58 429 deliveries by Skidmore et al20 described hemorrhagic stroke in 11 patients, 4 of which were caused by AVM rupture. Other case series and reports also support higher AVM hemorrhage rates during female reproductive years.21–25
Another major concern about hemorrhage during pregnancy is the risk of recurrent hemorrhage. This risk was not captured in the study by Liu et al8 because patients were not followed after initial hemorrhagic presentation. In 1974, Robinson et al26 described 152 women <45 years of age with either brain AVM or aneurysm. Twenty-four of these patients had an AVM and presented within 2 years of pregnancy. Twenty-one of the 24 women presented with hemorrhage, 11 of which occurred during pregnancy. Importantly, the rebleed rate of these patients during the same pregnancy was 27%, and fetal death rate was reported to be 26%. Gross et al11 have also described that 1 in 4 of their patients experienced a rebleed during pregnancy. In our series, the rebleed rate after initial intrapartum hemorrhage (25%) was similar to previously reported figures. We counsel attention to patient variability because 1 patient experienced multiple hemorrhage recurrences necessitating cesarean section at 29 weeks. Although we did not observe patient or fetal mortality during the exposure period, the demonstrated propensity for hemorrhage recurrence—particularly during the second and third trimesters and puerperium—should prompt vigilance from the patient, obstetrics, and neurosurgery.
Management Strategy of Pregnant Patients With AVM
Pregnant patients presenting with signs and symptoms suggestive of intracranial hemorrhage require timely diagnosis. Computed tomography remains the standard for evaluation of potential intracranial hemorrhage and can be performed in pregnancy with lead shielding to the abdomen and pelvis.5 The gold standard method of diagnosing brain AVM remains digital subtraction angiography. Digital subtraction angiography is thought to deliver ≈0.17 to 2.8 mGy of radiation to the fetus, which is well below the accepted limit of 50 mGy.27,28 There were no reported instances of radiation-induced fetal malformations within our cohort or that of Liu et al.8 Once AVM rupture is confirmed, there is minimal evidence to guide management decisions. Early surgical resection has not been reported to have better outcomes in pregnant patients compared with conservative management.1 In part, AVM grade guides treatment as grades V and VI lesions are generally managed conservatively because of predicted morbidity and mortality.29 Emergency AVM resection has been recommended for patients with signs of brain herniation and lesions graded I to IV.8 For patients presenting in late pregnancy, surgery can be deferred until after cesarean delivery.
Accurate interpretation of our study requires acknowledgment of its limitations, including its retrospective design and reflection of a single institution experience, which renders it susceptible to attrition bias. Among a total of 417 female patients with AVM, 401 did not have hereditary hemorrhagic telangiectasia. Ultimately, 270 patients (67%) satisfying our study’s inclusion criteria were included. We mitigated this risk through vigorous record tracking and management. Recall bias is also possible if abortion reporting was not complete. Our study is also limited in that we cannot prospectively study patients with this diagnosis and are therefore relegated to use of case-crossover analysis. Regardless, this study represents one of the largest cohorts describing the natural history of brain AVMs in pregnant women.
This study describes a large North American cohort of female patients with AVM and suggests an ≈5-fold increased risk of hemorrhage during pregnancy. This result conflicts with that reported by a large Chinese cohort but is consistent with findings from another North American cohort. For patients with known AVM who become pregnant, careful monitoring should be pursued throughout pregnancy but most particularly during the second and third trimesters and the puerperal period. Evidence guiding whether to intervene prophylactically during pregnancy remains sparse.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.016828/-/DC1.
- Received January 27, 2017.
- Revision received March 20, 2017.
- Accepted April 5, 2017.
- © 2017 American Heart Association, Inc.
- Sharshar T,
- Lamy C,
- Mas JL
- Davie CA,
- O’Brien P
- Tiel Groenestege AT,
- Rinkel GJ,
- van der Bom JG,
- Algra A,
- Klijn CJ
- Young WL,
- Yang GY
- Hademenos GJ,
- Alberts MJ,
- Awad I,
- Mayberg M,
- Shepard T,
- Jagoda A,
- et al
- Nishida T,
- Faughnan ME,
- Krings T,
- Chakinala M,
- Gossage JR,
- Young WL,
- et al
- Monte LM,
- Ellis RR
- Yang W,
- Porras JL,
- Garzon-Muvdi T,
- Xu R,
- Caplan JM,
- Hung AL,
- et al
- Yang W,
- Porras JL,
- Hung AL,
- Khalid S,
- Garzon-Muvdi T,
- Caplan JM,
- et al
- Crawford PM,
- West CR,
- Chadwick DW,
- Shaw MD