Article Text

Download PDFPDF

Periprocedural complications of second-generation flow diverter treatment using Pipeline Flex for unruptured intracranial aneurysms: a systematic review and meta-analysis
  1. Kartik Dev Bhatia1,
  2. Hans Kortman1,
  3. Emanuele Orru1,
  4. Jesse M Klostranec1,
  5. Vitor M Pereira1,
  6. Timo Krings1,2
  1. 1 Division of Neuroradiology, Joint Department of Medical Imaging, Toronto Western Hospital, Toronto, Ontario, Canada
  2. 2 Division of Neuroradiology, Department of Medical Imaging and Division of Neurosurgery, Department of Surgery, Toronto Western Hospital, University Health Network, Toronto, Ontario, Canada
  1. Correspondence to Dr Kartik Dev Bhatia, Division of Neuroradiology, Joint Department of Medical Imaging, Toronto Western Hospital, Toronto ON M5T 2S8, Canada; kartik.bhatia{at}


Background Flow diverters are a breakthrough treatment for large and giant intracranial aneurysms but carry a risk of periprocedural death or major stroke. Pipeline Flex is a second-generation device that is thought to have lower complication rates because of improvements in the delivery system as well as increased operator experience. Our objective was to analyze the risk of periprocedural death or major complications using Pipeline Flex for unruptured intracranial aneurysms.

Methods A systematic search of three databases was performed for studies of ≥10 treatments using Pipeline Flex for unruptured intracranial aneurysms (2014–2019) using PRISMA guidelines. Random effects meta-analysis was used to pool the rates of periprocedural (<30 days) death, major ischemic stroke, symptomatic intracranial hemorrhage, and minor stroke/transient ischemic attack.

Results We included eight studies reporting 901 treatments in 879 patients. Periprocedural mortality (<30 days) was 0.8% (5/901; 95% CI 0.4% to 1.5%; I2=0%). Rate of major complications (death, major ischemic stroke, or symptomatic intracranial hemorrhage) was 1.8% (14/901; 95% CI 1.0% to 2.7%; I2=0%). Aneurysm size ≥10 mm was a statistically significant predictor of a major complication (OR 6.4; 95% CI 2.0 to 20.7; p=0.002). Risk of a major complication in aneurysms <10 mm was 0.9% (95% CI 0.3% to 1.7%; I2=0%). The meta-analysis was limited by the predominance of anterior circulation aneurysms.

Conclusion Treatment of unruptured intracranial aneurysms using the Pipeline Flex flow diverter has a low periprocedural risk of death (0.8%) or major complication (1.8%). The risk of a major complication is significantly higher for large/giant aneurysms (4.4%) and is very low for aneurysms <10 mm (0.9%).

  • aneurysm
  • flow diverter
  • hemorrhage
  • stroke
  • complication

Statistics from


Flow diverter devices are increasingly used for the treatment of intracranial aneurysms, particularly for large or wide-necked sidewall aneurysms that are difficult to treat with traditional endovascular techniques. However, initial enthusiasm was followed by some concerns on reports of periprocedural neurological complications including intracerebral hemorrhage, in-stent thrombosis with major ischemic stroke, thromboembolic infarcts, perforator infarcts, and death.1 Meta-analyses on complication rates suggest a risk of death or major stroke of 5–8% in the anterior circulation and 10–15% in the posterior circulation.1 2 For certain aneurysms, these complication rates exceed the risk of death or disability from the natural history of the aneurysm.3 The relatively high complication rate was believed to arise from thrombus formation along the stent wall,1 that may partly relate to suboptimal wall apposition, difficulty in expanding the device, and the inability to resheath the device.4–7

Since 2014, second-generation flow diverter devices have become available that address many of these issues: they are easier to deploy, can be resheathed to allow for optimization of placement and expansion, and are therefore thought to be associated with lower complication rates.4 8–10 An important additional contributor to suspected improving complication rates is the greater collective operator experience in deploying flow diverter devices that has occurred over time.11 The three most commonly used second-generation devices are: the Pipeline Flex (Medtronic Neurovascular, Irvine, California, USA), an update of the Pipeline Classic that received CE mark in 2014 and FDA pre-market approval in January 2015; the SILK+ (Balt, Montmorency, France), an update of the SILK that has not yet received FDA approval; and the Surpass Streamline (Stryker, Kalamazoo, Michigan, USA) that received FDA approval in July 2018. The Pipeline Flex device has recently also become available with a covalently bonded phosphorylcholine surface coating (Pipeline Flex with Shield technology; FDA approval December 2018) that is thought to further reduce thrombogenicity.12

These devices have unique material, design, and deployment properties, and despite having similar underlying mechanisms of action, cannot be grouped together in a meta-analysis. In fact, a previous meta-analysis on flow diverter complication rates demonstrated outcome differences between different flow diverting devices.2 The most extensively used and studied second-generation device is the Pipeline Flex, with several studies suggesting lower rates of major complications compared with the Pipeline Classic.4 9 10 13–16

Prior meta-analyses on flow diverter complications were predominantly based on first-generation device studies.1 2 As such, a focused meta-analysis on periprocedural morbidity and mortality for studies using a second-generation device would be useful to better inform patients of the true risks when using more modern devices. Restricting the meta-analysis to the most commonly used device (Pipeline Flex) and to the unruptured aneurysm setting would potentially reduce heterogeneity of the study outcomes.


The primary objective was to determine the periprocedural mortality and morbidity of the Pipeline Flex flow diverter device for the treatment of unruptured intracranial aneurysms. This was assessed by a systematic review and meta-analysis of the published literature according to Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines.17 Our protocol was submitted prospectively to PROSPERO (CRD42019127172).

Study types

Eligible study types included randomized controlled trials, case–control studies, prospective cohort studies, and retrospective cohort studies, with a minimum sample size of 10 patients. This minimum sample size was chosen to minimize the inclusion of underpowered studies based on the finding of a 17% (95% CI 13.6% to 20.5%) overall complication rate in the prior meta-analysis of Zhou et al in 2017.2

Eligibility criteria (PICO criteria)

Eligible studies required adult participants (≥18 years) with unruptured intracranial aneurysms treated by deployment of one or multiple Pipeline Flex flow diverter(s). Studies incorporating cases with adjunctive coiling, prior endovascular treatment, or use of Pipeline Flex with Shield technology were permitted. In the absence of any known randomized data there is no comparator.


The primary outcome was the incidence of periprocedural mortality (within 30 days of the procedure). The secondary outcomes were the incidence of periprocedural major ischemic stroke (defined as an increase in the National Institutes of Health Stroke Scale (NIHSS) score of ≥4 points), symptomatic intracranial hemorrhage (sICH) (associated with a NIHSS score increase of ≥4 points) with subgrouping based on index aneurysm bleed versus distant hemorrhage, a composite of major complications (including death, major ischemic stroke, and sICH), minor stroke (transient or permanent increase in NIHSS score of 1–3), total treatment-related complication rate, successful final device deployment, use of adjunctive coiling, balloon angioplasty, or mandatory platelet function testing, and in-stent thrombosis requiring emergent use of intravenous glycoprotein IIb/IIIa inhibitors.

Search strategy

A systematic search of three online databases (MedLine via PubMed, Embase via Ovid, and Cochrane Library) was performed by two independent reviewers (KB, HK). The following search terms were used: MeSH intracranial aneurysm AND flow diverter OR pipeline flex. A filter was applied to include only studies published in the last 5 years (since the Pipeline Flex device was first released in 2014). The reference lists of two recent review articles were cross-referenced (Zhou et al, 2017; Yan et al, 2018).2 18 Additional references were also identified from correspondence with other authors.

Selection of studies

Using the PRISMA flowchart for guidance (see online supplementary material), duplicates were removed and the remaining records were independently screened using their titles/abstracts to identify articles that fulfilled the inclusion criteria. Relevant articles were selected for full-text assessment by both reviewers, who then assessed eligibility for inclusion in the quantitative meta-analysis. Where there was disagreement, it was resolved by consensus; otherwise it was resolved by a senior reviewer (TK). Eligible studies were required to report the results of the primary outcome and at least the first two secondary outcomes.

Data extraction and management

The outcome data were extracted from the articles by both reviewers. Data were collated using a standardized dedicated online data extraction sheet using AirTable ( The results entered by each reviewer were checked by the other reviewer; disagreement was resolved by consensus or otherwise via a senior reviewer (TK). The collated data points are summarized in the online supplementary material.

Meta-analysis and statistical testing

StatsDirect statistical software ( version 3.1.22 (Cambridge, UK) was used for all analyses. The dichotomous variables of the primary and secondary outcomes were assessed using frequency counts to determine proportions relative to the sample sizes. Proportions were transformed using the Freeman–Tukey double arcsine method and random effects meta-analysis performed using the methodology of DerSimonian and Laird.19 Pooled results were displayed using forest plots with weighted estimation of the mean of the sample distribution and 95% confidence intervals. Subgroup analyses for death and major complications were undertaken differentiated by aneurysm size (<10 mm, ≥10 mm), distal location, use of mandatory platelet function testing, and using a leave-one-out analysis. Univariate logistic regression was used to assess the predictive impact of large/giant aneurysm size (≥10 mm) on major complications.

Heterogeneity testing and bias risk assessment

Heterogeneity testing was performed by calculating Cochran Q statistics to determine I2 values as part of the DerSimonian and Laird random effects meta-analyses. P values <0.10 for Q or I2 ≥50% would represent statistically significant heterogeneity. Bias risk assessment was via funnel plot analysis with quantitative assessment using Egger and Harbord–Egger statistics (α <0.05).


Search results

The online search of three databases (Pubmed n=380, Embase n=365, Cochrane n=26) yielded 771 records, and after duplicates were removed, the abstracts and titles of 432 records were screened by both reviewers with exclusion of 422 records (see online supplementary material: PRISMA flow diagram). An additional record was identified through correspondence with other authors, which involved a published abstract for the prospective single-arm PREMIER study that had been presented at the International Stroke Conference 2017.20 The manuscripts of the remaining 14 records were assessed by both reviewers, with exclusion of six records and inclusion of eight records for the meta-analysis. The reasons for exclusion were inability to distinguish Pipeline Flex from Pipeline Classic cases from the manuscript (n=4) and multiple publications on the same patient samples (eg, early outcomes and late outcomes; n=2). For the samples in which there were multiple publications (Colby et al, 20189 and 201610; Martinez-Galmadez et al, 201814 and 201715), data points not available from the latter publication were obtained from the former publication where appropriate.

Included studies and baseline demographics

Eight studies were included in the meta-analysis describing 901 treatments for 935 aneurysms in 879 patients.4 8 9 13 14 20–22 Three of the included studies were prospective.8 14 20 Mean age was 56.7 years (95% CI 54.0 to 59.4) with 752 female patients (83.4%). Mean aneurysm diameter was 7.9 mm (95% CI 4.5 to 11.3) with 671 small/medium (<10 mm), 235 large (10–24 mm), and 29 giant (≥25 mm) aneurysms (71.8%, 25.1%, and 3.1%, respectively). One study included only large/giant aneurysms (n=100)22 while two studies included only small/medium aneurysms (n=46 and 141).4 20 Posterior circulation aneurysms (36 of 794 aneurysms; 4.5%) were included in four and reported in only three of the studies.4 14 20 21 Patient and aneurysm characteristics are summarized in table 1.

Table 1

Patient and aneurysm characteristics for the eight included studies assessing Pipeline Flex treatment of unruptured intracranial aneurysms


The meta-analysis results for the major outcomes are presented in table 2. Subgroup analyses are shown in table 3 and antiplatelet regimens are shown in table 4.

Table 2

Meta-analysis of major outcomes (<30 days), heterogeneity testing, and bias risk testing

Table 3

Subgroup analyses for the major outcomes

Table 4

Antiplatelet regimens and platelet function testing across studies

The overall treatment-related rate of complications using random effects meta-analysis was 10.1% (95% CI 5.3% to 16.2%; I2=83.8%). Death occurred in 5 of 901 treatments (4 of 5 in treatments for large anterior circulation aneurysms) with a meta-analysis weighted proportion of 0.8% (95% CI 0.4% to 1.5%; I2=0%). Major ischemic stroke occurred in 4 of 901 treatments (meta-analysis 0.6%; 95% CI 0.2% to 1.2%; I2=0%; all anterior circulation, 3 of 4 in large/giant aneurysms) and sICH in 9 of 901 treatments (meta-analysis 1.2%; 95% CI 0.6% to 2.1%; I2=0%; all anterior circulation, 6 of 9 in large/giant aneurysms). A composite of major complications (including death, major ischemic stroke, or sICH) yielded 14 events in 901 treatments, with a meta-analysis weighted proportion of 1.8% (95% CI 1.0% to 2.7%; I2=0%). Minor stroke/TIA occurred with 17 of 760 treatments (2.5%; 95% CI 1.0% to 4.8%; I2=54.4%) and other complications occurred in 67 of 712 treatments for which this outcome was reported (9.8%; 95% CI 6.3% to 13.9%; I2=44%); data for these outcomes are not yet reported for the PREMIER study.20

Univariate logistic regression showed that large/giant aneurysm size (≥10 mm) was a statistically significant predictor of death or major complication (OR 6.4; 95% CI 2.0 to 20.7; p=0.002). The meta-analysis risk of major complications in large/giant aneurysm treatments was 4.4% (95% CI 2.3% to 7.3%; I2=0%) including four deaths. For small/medium sized aneurysms (<10 mm) the risk was 0.9% (95% CI 0.3% to 1.7%; I2=0%) with one death. For distal aneurysms the risk was 3.0% (95% CI 0.6% to 7.2%; I2=0%). All major complications (death, major ischemic stroke, or sICH) in this meta-analysis occurred within the four largest studies,9 20–22 consistent with the larger sample sizes in these studies (n=316, 205, 141, and 94 treatments, respectively).

Heterogeneity testing

No statistically significant heterogeneity was identified for the primary outcome of death or the major secondary outcomes of major stroke, sICH, or the composite of major complications. There was statistically significant heterogeneity for minor stroke/TIA (I2=54.4%), other complications (I2=58.8%), and total complications (I2=83.8%) (table 2). Statistically significant heterogeneity across articles was also present for aneurysm size, anterior versus posterior circulation location, prior rupture status, use of adjunct coiling, and use of balloon angioplasty (see table 1).

Bias risk assessment

No statistically significant bias risk was identified for the major outcomes. The leave-one-out analysis did not suggest bias risk based on the variable sample sizes (see table 3). There was statistically significant bias risk for major complications in the subgroup analysis of studies with non-mandatory platelet function testing (see table 4).


Major findings

Our meta-analysis of flow diversion using Pipeline Flex for unruptured intracranial aneurysms suggests that the periprocedural risk of major complications (1.8%) is markedly lower than that reported for first-generation devices (5% morbidity and 4% mortality).1 Consistent with prior studies,2 23 our meta-analysis demonstrated that these risks were higher when treating large/giant aneurysms (4.4%, OR 6.4). Interestingly, the meta-analysis risk of a major complication when treating small/medium (<10 mm) unruptured aneurysms was very low (0.9%) compared with a reported rate of 3% when using Pipeline Classic.24 With published aneurysm occlusion rates of 87–91% after flow diverter treatment for small anterior circulation aneurysms,25–27 this very low periprocedural risk profile when using a second-generation device would significantly alter the risk-benefit ratio when considering treatment for these aneurysms.

The lower risk profile reported in this meta-analysis cannot only be credited to technological advances between device generations. Increased operator experience in device deployment over multiple years with advancement along the learning curve and improving antiplatelet regimens have likely contributed to lower procedural complication rates.11 In addition, the overall rate of treatment-related complications (10.1%) is still considerable and includes minor stroke/TIA (2.5%) and other complications (9.8%), which were variably defined and reported across studies (eg, headache or nausea).

Despite the known higher risks associated with treating posterior circulation aneurysms,1 2 28 all major neurological complications in our meta-analysis occurred in anterior circulation aneurysms. This was likely due to the purposeful exclusion of posterior circulation aneurysms from four of the eight studies (including the largest study),9 13 and the relatively small sampling (4.5%) of these aneurysms in our pooling. Therefore, it would be fair to state that the results of our meta-analysis are most applicable to unruptured anterior circulation aneurysms and should not be generalized to the posterior circulation.

A previous large meta-analysis of treatments using predominantly first-generation devices with inclusion of more heterogeneous and higher-risk patient groups suggested a periprocedural mortality rate of 2.8% and neurological morbidity rate of 4.5%, with rates of 1.8% and 3.1% for death or severe complications in the subgroup treated using Pipeline Classic (n=1570).2 The risk of complications (all) in that analysis was higher in ruptured cases (OR 2.3) and in posterior circulation compared with anterior circulation aneurysms (44.7% vs 23.7%).2 While such data were useful in providing a global view of the risks of flow diverter treatment, it was affected by statistically significant heterogeneity (I2 >50% for major outcomes) between studies with regard to rupture status, aneurysm location, and device used.2 As such, the quoting of such statistics when undertaking informed consent for any individual patient may not be appropriate. This need for more specific data in the unruptured setting using a second-generation device was addressed in our analysis.

The technological updates in the Pipeline Flex device compared with the Pipeline Classic include the presence of 2–3 mm length PTFE (polytetrafluoroethylene) sleeves that help protect the distal end of the device and facilitate deployment, the ability to resheath the device with the 0.027 inch microcatheter due to the presence of a proximal silicone resheathing pad and increased pushability due to replacement of the Classic pusher wire with a larger hypotube.29 Optimization of placement, reduced procedural times and complexity, and enhanced wall apposition may all contribute to a reduced risk of in-stent thrombosis.4–7 These advantages may have partly contributed to the low risk of major complications seen in this meta-analysis; the rate of successful final deployment of the device was 99.3% (95% CI 98.5% to 99.8%). However, other significant contributors would include our purposeful assessment of unruptured aneurysms only,2 the predominance of small/medium anterior circulation aneurysms in several of the included studies, as well as improved operator experience and more efficient antiplatelet regimens over time.11 Within the limitations of this meta-analysis, mandatory platelet function testing was not associated with improvement in major complication rates (see table 4).

Use of second-generation devices may offer some additional procedural advantages. Le et al demonstrated significantly reduced fluoroscopy times, procedural times, radiation dosage, and contrast use with Pipeline Flex procedures compared with Pipeline Classic procedures performed by the same group of experienced operators.4 There are also reported disadvantages with the use of Pipeline Flex; Mooney et al prospectively observed failure to recapture the delivery wire of the device in 10 of 18 cases.30

Causes of significant heterogeneity and bias risk

Statistically significant heterogeneity was present in our meta-analysis for aneurysm size, location, and prior remote rupture status. For size, this was likely due to three studies having almost entirely small/medium aneurysms in their samples4 13 20 while, on the other hand, large/giant aneurysms comprised 100%22 and 44%21 of the samples for another two studies. For location, the heterogeneity likely resulted from the purposeful exclusion of posterior circulation aneurysms from four of the eight studies.8 9 13 22 Significant heterogeneity was also present for the outcomes of minor stroke/TIA (partly due to two TIA episodes in a study of 10 patients8), other complications (likely due to variable inclusion and reporting across studies), and the overall rate of treatment-related complications.

No statistically significant heterogeneity was present for the major outcomes of death, major ischemic stroke, sICH, or the composite of major complications. There was a statistically significant bias risk in the subgroup analysis for non-mandatory use of platelet function testing, likely due to the absence of major complications in the smallest two of the five studies in this subgroup; this raises the possibility of a type II error in these studies.


This meta-analysis is limited by the relatively small number of eligible studies (n=8) and the imbalance across studies for sample size (range 10–316 treatments). Long-term aneurysm occlusion outcomes and complication rates could not be accurately addressed as long-term follow-up has not yet been reported for the two largest studies.9 21 An additional significant confounder was the use of an updated device (Pipeline Flex with Shield technology) as the sole device in one prospective study with 50 patients14; this device was designed to reduce complication rates through the use of an anti-thrombogenic coating but is otherwise identical in design and deployment. Five of the eight included studies were retrospective. The findings of this meta-analysis are most applicable to unruptured anterior circulation aneurysms (95% of the pooling) and should not be generalized to the posterior circulation setting.


Treatment of unruptured intracranial aneurysms using the second-generation Pipeline Flex flow diverter has a low periprocedural risk of death (0.8%) or major complications (1.8%). Risk of a major complication is significantly higher for large/giant aneurysms (4.4%) and is very low for aneurysms <10 mm (0.9%). Our findings are most applicable to unruptured anterior circulation aneurysms.



  • Contributors KDB was the primary author of the study protocol, was one of the two reviewers, performed data extraction and collation, undertook data analysis, and drafted and revised the manuscript. He is the guarantor. HK was one of the two reviewers, performed data collation and data quality checking, assisted with data analysis, and edited the manuscript. EO advised upon and edited the study protocol, performed data analysis, and edited the manuscript. JMK advised upon and edited the study protocol, performed data analysis, and edited the manuscript. VMP advised upon the methodology, assisted with revising the draft protocol, checked collated data for accuracy, analyzed the data output for suitability of meta-analysis, and edited and helped revise the manuscript. TK supervised the project, acted as the senior reviewer in cases of non-consensus, advised upon and assisted with editing of the protocol, advised upon the methodology, analyzed the data output for suitability of meta-analysis, and edited and helped revise the manuscript. All authors assisted in the design of the study.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests VMP is a consultant and proctor for Medtronic Neurovascular and Stryker and a consultant for Phenox and Balt. One of the studies included in this meta-analysis (Pereira et al, 2015) was published by members of our research group.

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

  • Patient consent for publication Not required.

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.