Article Text

Download PDFPDF

Effect of definition and methods on estimates of prevalence of large vessel occlusion in acute ischemic stroke: a systematic review and meta-analysis
  1. Muhammad Waqas1,2,
  2. Ansaar T Rai3,
  3. Kunal Vakharia1,2,
  4. Felix Chin2,
  5. Adnan H Siddiqui3,4
  1. 1 Neurosurgery, University at Buffalo Jacobs School of Medicine and Biomedical Sciences, Buffalo, New York, USA
  2. 2 Neurosurgery, Gates Vascular Institute, Buffalo, New York, USA
  3. 3 Department of Neurointerventional Radiology, West Virginia University, Morgantown, West Virginia, USA
  4. 4 Departments of Neurosurgery and Radiology and Toshiba Stroke Research Center, University at Buffalo, State University of New York, Buffalo, New York, USA
  1. Correspondence to Dr. Adnan H Siddiqui, Departments of Neurosurgery and Radiology and Toshiba Stroke Research Center, University at Buffalo, State University of New York, Buffalo, NY 14260, USA ; asiddiqui{at}, adnan.h.siddiqui{at}


Introduction Accurate estimation of the incidence of large vessel occlusion (LVO) is critical for planning stroke systems of care and approximating workforce requirements. This systematic review aimed to estimate the prevalence of LVO among patients with acute ischemic stroke (AIS), with emphasis on definitions and methods used by different studies.

Methods A systematic literature review was performed to search for articles on the prevalence of LVO and AIS. All articles describing the frequency of LVO frequency among AIS patients were included. Studies without consecutive recruitment or confirmation of LVO with CT angiography or MR angiography were excluded. Heterogeneity of the studies was assessed; meta-regression was performed to estimate the effect of LVO definition and study methods on LVO prevalence.

Results 18 articles met the inclusion criteria: 5 studies presented population based estimates; 13 provided single hospital experiences (5 prospective, 8 retrospective). The AIS denominator (number of all AIS) from which LVO rates were generated was variable. Nine different definitions were used, based on occlusion site. Significant heterogeneity existed among the studies (I2=99%, P<0.001). The prevalence of LVO among patients with suspected AIS ranged from 13% to 52%. Overall prevalence was 30.0% (95% CI 25.0% to 35.0%). Pooled prevalence of LVO among suspected AIS patients was 21% (95% CI 19% to 30%). Based on meta-regression, the method of AIS denominator determination significantly influenced heterogeneity (P=0.018).

Conclusion The heterogeneity of LVO estimates was remarkably high. The method of AIS denominator determination was the most significant predictor of LVO estimates. Studies with a standardized LVO definition and methods of AIS estimation are necessary to estimate the true prevalence of LVO among patients with AIS.

  • large vessel occlusion
  • acute ischemic stroke
  • population-based studies
  • large vessel occlusion definition
  • systematic review

Statistics from

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.


According to the WHO, 15 million people suffer from a stroke each year, leaving 5 million dead and 5 million severely disabled. Large vessel occlusion (LVO) is associated with unfavorable outcomes at 3 and 6 months in patients with acute ischemic stroke (AIS).1 An accurate estimation of the incidence of LVO is important in determining the resources and workforce required for stroke systems of care. A key variable in these estimations is the denominator of all AIS patients from which the prevalence of LVO is derived. The oft quoted incidence of approximately 8 00 000 new or recurrent strokes annually in the USA is derived from studies utilizing specific International Classification of Diseases (ICD) discharge codes.2 A study seeking to estimate the LVO rate among all AIS patients should ideally use the same AIS ICD codes2 for establishing the denominator.

Studies estimating the prevalence of LVO have used different methodologies that vary from population based estimates to consecutive patient series, and from single stroke centers or data provided by emergency medical services (EMS). According to a systematic review by Lakomkin et al, the mean prevalence of LVO among patients with AIS was 31.1%.3 However, these authors did not clearly distinguish among the methods used to determine AIS in the studies they reviewed and how those methods influence estimates of LVO. Further, some key studies were not included and a pooled analysis was not carried out. Estimations of the prevalence of LVO are also complicated by variability in LVO definitions. Lakomkin et al found that 16 of the studies included in their systematic review used nine different definitions of LVO (different combinations of locations of arterial occlusions).3 However, these authors did not assess study quality or stratify results according to the different definitions or methods of estimating LVO. Moreover, estimates of LVO were based on AIS diagnosis and were not studied for patients with suspected AIS.

Therefore, we decided to conduct a systematic review of the literature for an estimation of the prevalence of LVO among patients with suspected and confirmed AIS, with an emphasis on the definitions and methodologies used by various studies.


Search strategy and selection criteria

We searched the Excerpta Medica dataBASE (EMBASE), Public/Publisher Medline (PubMed), Web of Science, Google Scholar, and the Cochrane Library to identify relevant articles. Medical subject headings (MeSH) terms used for performing the literature searches were (prevalence of large vessel occlusion in stroke), (prevalence of large artery occlusion in stroke), (large vessel occlusion and stroke), (thrombectomy eligible patients and stroke). Different combinations of MeSH terms were used to enhance the searches. The searches were independently performed by two of the authors and a librarian who had extensive experience in conducting literature searches. Initial searches were performed during the first week of October 2018. Repeat searches were performed on November 14, 2018, and again during the last week of November 2018.

All articles reporting the prevalence of LVO based on CT angiography (CTA) or MR angiography (MRA) findings in a population based or hospital based setting that were published between January 1, 2000 and June 30, 2018 were included in our study, irrespective of the method of determining LVO. Studies without consecutive recruitment or without confirmation of LVO with CTA or MRA were excluded. Articles in languages other than English were also excluded.

The titles and abstracts were reviewed to remove duplicates and irrelevant articles. Further screening of full text articles was performed to meet the inclusion criteria. The screening process was conducted independently by two authors. Additionally, we searched the bibliographies of all initially included articles.

Data were extracted independently by two authors after reading the full text articles. Information extracted included study design and setting, study definition of LVO, denominator determination method, and proportion of patients with LVO. Any conflicts in those authors’ observations were resolved by mutual discussion or the opinion of a third author.

We then devised a 7 point quality assessment tool to evaluate each study. Yes or no ratings were included for the attributes of prospective design, consecutive sampling, uniform definition of stroke, specified vessel occlusion sites, defined catchment area, defined catchment population, and use of CTA or MRA to determine the presence of LVO. The PRISMA guidelines were followed for reporting this systematic review.

Meta-analysis and meta-regression

We planned to assess heterogeneity among the studies using an I2 test. If the heterogeneity value was >50%, we planned to perform meta-regression to assess variables that accounted for the heterogeneity. We intended to use a random effects statistical model if I2 was ≥50% and a fixed effects model if I2 was <50%. Prevalence and CIs were calculated for each study based on the denominator and number of patients with LVO. Prevalence of LVO among confirmed AIS as well as suspected AIS patients was calculated using the appropriate effects model based on the amount of heterogeneity. The results from individual studies for the prevalence of LVO and, if reported, the prevalence of thrombectomy eligible patients were pooled together to estimate the mean prevalence of each with 95% CI. LVO prevalence was calculated separately for population based studies.


The initial literature searches provided 70 citation records (articles). After removing duplicates, 50 records remained. Eighteen records met the criteria for inclusion in this systematic review.4–21 The numbers of records at different stages of the literature searches and the screening process are shown on a PRISMA flow chart (see online supplementary figure 1). The quality of each study, based on the number of positive attributes using the scale mentioned in the methods section, is presented in the online supplementary table 1. The scores ranged from 3 to 6. Individual studies and their characteristics are listed in table 1. The AIS denominator and method of its determination in each study is presented in table 2. The overall prevalence of LVO among AIS patients was 30.0% (95% CI 25.0 to 35.0).

Supplementary material 1

Table 1

Characteristics of the studies included in the systematic review

Table 2

Number of eligible patients and percentage of large vessel occlusion in studies included in the systematic review*

Assessment of heterogeneity

The statistical heterogeneity among the included studies (I2) was 99% (P<0.001). Therefore, the random effect model was used to calculate the pooled prevalence of LVO among patients with confirmed and suspected AIS. Meta-regression was carried out to see the effect of LVO definition and study design and denominator on the measured heterogeneity. Together, these variables accounted for 46.8% of the heterogeneity, with study design having the strongest impact. We have shown our meta-regression model and heterogeneity test results in table 3.

Table 3

Meta-regression to assess the effect of different variables on the estimates of large vessel occlusion

Effect of the denominator

The AIS denominator from which the LVO rate was derived varied among studies. The commonly used format was to estimate LVO from a cohort of patients with suspected AIS who arrived at a hospital’s emergency room. The suspicion of AIS was based on neurological presentation and/or imaging, such as MRI. One of the studies used AIS ICD discharge codes to define the denominator.21 The AIS denominator is critical in comparing LVO rates among different studies because all of these studies report LVO prevalence as a percentage of the larger AIS cohort, which is determined differently from study to study (table 2). To determine the impact of a change in the denominator, we applied two separate statistical models for confirmed and suspected AIS. The first model was based on the number of suspected AIS patients reported by 11 studies. The pooled prevalence was 21% (95% CI 19 to 30). The second model was based on ‘confirmed AIS’ and resulted in a prevalence of 30% (95% CI 25.0 to 35.0) The pooled prevalence of LVO among suspected and confirmed AIS patients is shown in the Forest plots in figure 1.

Figure 1

Random effects model. Top: Forest plot showing pooled prevalence of large vessel occlusion (LVO) based on suspected acute ischemic stroke (AIS) as the denominator (21%). Bottom: Forest plot showing pooled prevalence of LVO based on confirmed AIS as the denominator (30%).

Definition of LVO

The definition of LVO varied among studies (table 1). In five studies, LVO was defined as occlusion of any of the following arteries or arterial segments: internal carotid (ICA), M1, M2, A1, vertebral (VA), basilar (BA), P1, or P2.4 5 7 12 18 In one study, it was defined as occlusion of the ICA, M1 segment, or BA.16 In five studies, it was defined as occlusion of the ICA, M1, M2, or BA.13 15 17 20 21 In three studies, it was defined as ICA, M1, or M2 occlusion.9 10 19 In three studies, ICA, M1, M2, BA, and P1 occlusions were included in the definition.6 8 14 In one study, the definition of LVO was not reported.11

Methods for estimating LVO

Several methods were used in these studies. In five studies, estimation of LVO was based on populations in defined and broad catchment areas.6 8 11 20 21 Numbers of ischemic strokes and LVOs per 1 00 000 population per year were described. Eight studies were retrospective, with data derived from hospital databases or registries.4 5 7 9 12 13 15 16 Five studies were prospective investigations of consecutive patients presenting with ischemic stroke.10 14 17–19 In all studies, the diagnosis of LVO was based on CTA or MRA findings. The method of determining the AIS population is presented in table 1.

Prevalence of LVO among AIS patients according to study design

In the population based studies, the prevalence of LVO among suspected AIS patients ranged from 13%20 to 42.6%.8 The pooled prevalence was 27.1% (95% CI 19.46 to 34.74).6 8 11 20 21 The study by McMeekin et al 11 that was based on 1 year data from the national database of the UK reported an LVO rate of 40%. The combined prevalence of LVO obtained from retrospective studies of single center hospital registries and databases was 26.3% (95% CI 18.6 to 34.4). The range was 21.2%9 to 36.1%.5 According to our pooled analysis of prospective hospital based studies, the prevalence of LVO was 34.8% (95% CI 27.2 to 42.5). The prevalence ranged from 33% to 46%.17 18

Prevalence of LVO according to definition of occlusion site

Five studies with a broad definition of LVO (ICA, M1, M2, A1, VA, BA, or P1) had a pooled prevalence of 32.9% (95% CI 25.3 to 40.5) among patients with suspected AIS.4 5 7 12 16 The range was 22.6% to 46%. In three studies, the pooled prevalence of LVO (ICA, M1, M2, BA, or P1) was 31.3% (95% CI 23.7 to 38.9), ranging from 24.1%6 to 31.3%.4 In four studies, LVO prevalence (ICA, M1, M2, or BA) ranged from 13.0%20 to 33.0%.17 Pooled prevalence of LVO was 23.3% (95% CI 15.71 to 31.0).13 15 17 20

Prevalence of LVO in the USA and abroad

Ten studies estimated the prevalence of LVO in the USA. Not accounting for the heterogeneity of the AIS denominator, pooled analysis showed an LVO prevalence among patients with suspected AIS of 28.8% (95% CI 21.2 to 36.5). Eight studies outside the USA showed a pooled prevalence of 29.3% (95% CI 22.0 to 37.0) (table 2).

Prevalence of LVO in studies using imaging criteria to define the AIS denominator

In five studies, the denominator was ascertained by MRI, rather than by the incidence of all AIS.9 12 13 15 21 Pooled prevalence of LVO in these studies was 21.49% (95% CI 13.8 to 29.1). Data from individual studies is presented in the  online supplementary table 2.

Thrombectomy eligible patients

Only four studies reported the prevalence of thrombectomy eligible patients; the determination of prevalence was based on different eligibility criteria.5 11 20 21 According to the largest sample study among these, which was from the UK, 28.9–32.9% of all the patients with LVO were eligible for mechanical thrombectomy.11 The prevalence of thrombectomy eligible LVOs reported by Demeestere et al 5 was 68.3% (136 of 199); by Rai et al,21 it was 53.7% (174 of 324); and by Chia et al,20 it was 7.4–14.5% (17–33 of 228).


This systematic review highlights the heterogeneity of the existing literature regarding the definition of LVO and the methods used to determine the prevalence of LVO. Importantly, the AIS denominator is heterogeneously defined, with only a single study using specific AIS ICD discharge codes. Our meta-regression highlighted the value of the AIS denominator and method of its estimation as a statistically significant factor in determining LVO estimates (figure 1). The overall prevalence of LVO decreased from 30% to 21% when the number of patients with suspected AIS was used as the denominator instead of the number of patients with confirmed AIS (figure 1). It can be argued that the estimates should be based using suspected AIS as the denominator because when the emergency medical service encounters a patient with possible stroke, they have to manage it as a stroke unless confirmation is obtained with the help of radiologic investigations. The current systematic review is the first to establish the prevalence of LVO among patients with both confirmed and suspected AIS. A previous review by Lakomkin et al only estimated the prevalence of LVO among patients with AIS with no pooled analysis or stratification of results according to the definition and methods of estimation of the AIS denominator.3 In addition, those authors did not report statistical heterogeneity among the studies included in their review. We not only assessed the heterogeneity among studies but also carried out a meta-regression to look for variables responsible for the heterogeneity.

We selected only those studies in which CTA or MRA had been used for the diagnosis of LVO in consecutive patients. However, the differences in the definition of LVO, access of the stroke center to the catchment area and population, and the method of ascertaining the denominator appear to affect the prevalence of LVO.

Although LVO estimates based on hospital data are important, population based estimates are essential for projecting the true burden of LVO and planning stroke systems of care. The overall estimated prevalence of LVO among suspected AIS patients in the five population based studies in our review was 27.10% (95% CI 19.46 to 34.74),6 8 11 20 21 which was comparable with the pooled prevalence in retrospective studies (ie, 26.3%) and lower than in prospective hospital based studies (ie, 34.9%).

The definition of LVO influences its estimates. Studies defining LVO as an ICA, M1, M2, A1, VA, BA, or P1 occlusion (table 1) had the highest prevalence (34.9% (95%CI 27.2 to 42.5)) compared with 11.8% for ICA, M1, or BA occlusion, reported by Rai et al.22 In a subsequent report, those authors determined the prevalence of M2 occlusions in the same cohort of patients.21 The prevalence of M2 occlusion was found to be 4.2%. If M2 occlusion had been included in their earlier definition of LVO, the prevalence would have been 16.1%. Because both their studies involved the same patient cohort, the latter percentage was used for the current systematic review.

Another possible reason for the difference between reported prevalence rates can be the way in which the diagnosis of stroke was established. Rai et al used ICD-9 coding to determine the diagnosis of AIS at discharge.21 Other studies reviewed the MRI scans of patients to establish the presence of infarcts.9 13 15 21 The pooled prevalence of LVO in the five studies in which the AIS denominator was ascertained with neuroimaging (see online supplementary table 2) was 21.49% (95% CI 13.8 to 29.1). The LVO rate seems to be lower when the same ICD codes are used for AIS as those for the denominator published in the AHA’s ‘Heart and stroke statistics.”2

The overall prevalence of LVO among patients with suspected AIS in the USA was comparable with that of the rest of the world, according to data from articles included in this review; that is, 28.8% (95% CI 21.2 to 36.5) versus 29.3% (95% CI 22.0 to 37.0). However, because the studies are not matched in their definition and methods, these numbers may not be reliable.

Although we searched for relevant studies on LVO from 2000 to 2018, none of the articles published before 2008 met our inclusion criteria. The reason for an increased interest in LVO is a direct consequence of the subsequent rise in mechanical thrombectomy as a treatment option for LVO stroke after the publication of the positive results of randomized controlled trials.23 24 Therefore, estimating the prevalence of thrombectomy eligible patients is critical when estimating the prevalence of LVO. However, only four studies attempted to estimate the prevalence of thrombectomy eligible patients; each of these studies used decision trees based on American Heart Association guidelines and findings of previous clinical trials.23 24 Moreover, the variable estimated rates of thrombectomy eligible patients may well prove to be underestimates now that the DAWN (Diffusion weighted imaging or computerized tomograph perfusion Assessment with clinical mismatch in the triage of Wake-up and late presenting strokes undergoing Neurointervention with Trevo) and DEFUSE 3 (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3) trials have extended the time window wherein thrombectomy could be effective.25

Our systematic review has its limitations. Due to the heterogeneity of the LVO definition and study design as well as the missing information, the meta-analysis has limited value; however, the review does highlight important limitations of the available literature. Additionally, we did not include articles in the non-English literature. Nevertheless, to our knowledge, this is the largest systematic review on this topic and the first to assess the influence of definition and methods on the pooled prevalence of LVO.

Future directions

Population based studies with standardized methods, preferably using ICD-10 coding and including patients up to a minimum of 24 hours after symptom onset, must be carried out. Estimates of LVO should also include estimates of thrombectomy eligible patients. Prospective studies could include neuroimaging evidence to establish the denominator for total stroke populations.


The amount of heterogeneity among studies on the estimates of LVO was remarkably high. The reported prevalence of LVO ranged from 13% to 52%. Our meta-regression indicated that the method applied to determine the AIS denominator was the most significant predictor of LVO estimates. Studies with a uniform definition of LVO and methods of estimation of AIS are necessary to estimate the true prevalence of LVO among patients with AIS.


The authors thank Adrienne Doepp, MLS, Circuit Librarian, Hospital Library Services Program, Western New York Library Resources, for assistance performing the literature searches; Paul H Dressel BFA for preparation of the illustration; and W Fawn Dorr BA and Debra J Zimmer for editorial support (University at Buffalo Neurosurgery staff).



  • Contributors Conception and design: ATR and MW. Acquisition of the data: all authors. Analysis and interpretation of the data: all authors. Drafting the manuscript: MW. Critically revising the manuscript: all authors. Reviewed submitted version of manuscript: all authors.

  • 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 ATR: consulting agreement with Stryker Neurovascular and Cerenovus Siddiqui; financial interest/investor/stock options/ownership in Amnis Therapeutics, Apama Medical, Blink TBI Inc, Buffalo Technology Partners Inc, Cardinal Consultants, Cerebrotech Medical Systems Inc, Cognition Medical, Endostream Medical Ltd, Imperative Care, International Medical Distribution Partners, Neurovascular Diagnostics Inc, Q’Apel Medical Inc, Rebound Therapeutics Corp, Rist Neurovascular Inc, Serenity Medical Inc, Silk Road Medical, StimMed, Synchron, Three Rivers Medical Inc, Viseon Spine Inc; consultant/advisory board for Amnis Therapeutics, Boston Scientific, Canon Medical Systems USA Inc, Cerebrotech Medical Systems Inc, Cerenovus, Corindus Inc, Endostream Medical Ltd, Guidepoint Global Consulting, Imperative Care, Integra LifeSciences Corp, Medtronic, MicroVention, Northwest University–DSMB Chair for HEAT Trial, Penumbra, Q’Apel Medical Inc, Rapid Medical, Rebound Therapeutics Corp, Serenity Medical Inc, Silk Road Medical, StimMed, Stryker, Three Rivers Medical Inc, VasSol, WL Gore and Associates; principal investigator/steering comment of the following trials: Cerenovus LARGE and ARISE II, Medtronic SWIFT PRIME and SWIFT DIRECT, MicroVention FRED and CONFIDENCE, MUSC POSITIVE, and Penumbra 3D Separator, COMPASS, and INVEST.

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

  • Data sharing statement All study data are included in the manuscript and supplementary material.

  • Patient consent for publication Not required.