Artery of Adamkiewicz: a meta-analysis of anatomical characteristics

Purpose The artery of Adamkiewicz (AKA) provides the major blood supply to the anterior thoracolumbar spinal cord and iatrogenic injury or inadequate reconstruction of this vessel during vascular and endovascular surgery can result in postoperative neurological deficit due to spinal cord ischemia. The aim of this study was to provide comprehensive data on the prevalence and anatomical characteristics of the AKA. Methods An extensive search was conducted through the major electronic databases to identify eligible articles. Data extracted included study type, prevalence of the AKA, gender, number of AKA per patient, laterality, origin based on vertebral level, side of origin, morphometric data, and ethnicity subgroups. Results A total of 60 studies (n = 5437 subjects) were included in the meta-analysis. Our main findings revealed that the AKA was present in 84.6% of the population, and patients most frequently had a single AKA (87.4%) on the left side (76.6%) originating between T8 and L1 (89%). Conclusion As an AKA is present in the majority of the population, caution should be taken during vascular and endovascular surgical procedures to avoid injury or ensure proper reconstruction. All surgeons operating in the thoracolumbar spinal cord should have a thorough understanding of the anatomical characteristics and surgical implications of an AKA. Electronic supplementary material The online version of this article (10.1007/s00234-019-02207-y) contains supplementary material, which is available to authorized users.


Introduction
The artery of Adamkiewicz (AKA), also known as the great anterior radiculomedullary artery, is a major artery that joins the anterior spinal artery in the lower one-third of the spinal cord ( Fig. 1) [1]. Because of its large role in feeding the spinal cord, many reports have stressed the importance of reattaching the intercostal or lumbar arteries to the AKA in the event of spinal cord ischemia following vascular and endovascular surgery (Fig. 2). Identification of the AKA preoperatively helps surgeons to determine the appropriate range of aortic lesions that require graft replacement [2]. Therefore, accurate localization and detailed anatomical knowledge of the AKA are important when planning surgical and interventional radiological treatments of thoracoabdominal diseases and spinal lesions in order to help reduce the risk of postoperative ischemic spinal complications and paraplegia.
The AKA is the most dominant anterior radiculomedullary artery and is responsible for the arterial blood supply to the spinal cord from T8 to the conus medullaris [3]. Its origin is highly variable and extends from the mid-thoracic level to the DT and BS contributed equally to this work.
Electronic supplementary material The online version of this article (https://doi.org/10.1007/s00234-019-02207-y) contains supplementary material, which is available to authorized users. lumbar levels, including the bilateral T3-T12 intercostal arteries [4] and L1-L4 lumbar arteries [5]. It typically arises from the T8-L1 neural foramina [6] from the left intercostal or lumbar arteries [7]. The AKA has a diameter of 0.8-1.3 mm, and the distal portion of this artery, together with the anterior spinal artery, forms a characteristic Bhairpin^turn [8] (Fig. 3). Various techniques have been devised to preoperatively identify the location and anatomy of this artery. Such techniques include computed tomography angiography (CTA), magnetic resonance angiography (MRA), and digital subtraction angiography (DSA), with the latter considered the gold standard [9].
The most important cause of injury to the AKA is iatrogenic, and in part, this is a factor of the high degree of variability in the anatomical location of this artery [10]. Preoperative AKA identification and its subsequent reconstruction or preservation may aid in reducing the incidence of postoperative neurological deficits and improving the outcomes of thoracolumbar surgical procedures. To this end, the aim of this study was to provide comprehensive data on the prevalence and anatomical characteristics of the AKA.

Search strategy
A search of all major electronic databases (PubMed, EMBASE, ScienceDirect, China National Knowledge Infrastructure (CNKI), SciELO, BIOSIS, and Web of Science) was performed in order to identify potential articles. The following search terms were employed: artery of Adamkiewicz, arteria radicularis magna (ARM), radicularis magna, great radicular artery of Adamkiewicz, major anterior segmental medullary artery, great anterior segmental medullary artery, artery of the lumbar enlargement, arteria r a d i c u l a r i s a n t e r i o r m a g n a , a n d g r e a t a n t e r i o r radiculomedullary artery. A search through the references of the initially selected articles was conducted to identify any potential studies that were omitted. The authors adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines throughout this metaanalysis (Supplement 1).

Eligibility assessment
An eligibility assessment was conducted by two independent reviewers. Studies were included in this meta-analysis if they (1) provided complete data on the prevalence of the AKA or (2) provided data on the anatomy of AKA. The following exclusion criteria were employed: case, case-series, conference abstracts, letters to editors, and studies not published in peer-reviewed journals. Studies that were originally published in languages other than English were translated by medical professionals who are fluent both in English and the original language of the manuscript. All differences of opinion among the reviewers concerning the eligibility of the studies were resolved by consensus through consultation with the author of the respective study.

Data extraction
Two reviewers carried out data extraction independently. The following data was extracted: publication year, country of origin, study type (cadaveric, CTA, MRA, DSA), prevalence data of AKA, number of AKAs per patient, laterality of the AKA, origin of the AKA based on the vertebral level, side of origin, and morphometric data. In cases of incomplete data,

Quality assessment
The AQUA tool [11] was used by the reviewers to evaluate quality and reliability of the included studies. In brief, the tool was devised to probe for potential risk of bias. Five domains were evaluated in the analysis: (1) objective(s) and subject characteristics, (2) study design, (3) methodology characterization, (4) descriptive anatomy, and (5) reporting of results; and each domain was categorized as either of BLow,^BHigh,ô r BUnclear^risk of bias. Decision was made that a BNoâ nswer in whichever signaling question within each of the categories arbitrated the domain to be of BHigh^risk of bias, whereas all answers BYes^suggested that it presented a BLow^r isk of bias. BUnclear^option was chosen when the study with incoherent data did not permit for a clear scrutiny.

Statistical analysis
The prevalence analysis was conducted using MetaXL version 5.8 by EpiGear Pty Ltd. (Wilston, Queensland, Australia). Morphometric analysis using Comprehensive Meta-Analysis version 3.3 yielded the pooled mean diameter of the AKA. Single and multi-categorical pooled prevalence rates were calculated using a random effects model. Heterogeneity was assessed using a chi-squared test and the I 2 statistic. For the I 2 statistic, the values of 0-40% indicated that heterogeneity might not be important; values of 30-60% could indicate moderate heterogeneity; values of 50-90% could indicate substantial heterogeneity; and values of 75-100% indicated considerable heterogeneity. A p value below 0.10 for Cochran's Q suggested significant heterogeneity [12].
An analysis of the subgroups was conducted to determine the source of heterogeneity. The difference between the groups was considered to be insignificant if the confidence intervals (CIs) of specific rates overlapped [13]. Subgroups according to study type, gender, and geographical location were analyzed.

Study identification and characteristics of included studies
The study identification process is presented in Fig. 4. An initial search yielded 747 entries. After thorough analysis, The characteristics of the included studies are presented in Table 1. A total of 60 studies (4317 subjects with AKA) published between 1989 [14] and 2017 [15] were included [1][2][3][4][5][6][7][8][9][10]. The studies originated from North America, Asia and Europe, and from ten different countries.

Quality assessment
The majority of studies included in this meta-analysis, evaluated by the AQUA tool, revealed domain one (objective(s) and subject characteristics) and domain three (methodology characterization) to be at BHigh^risk of bias, owing to missing demographic data of the research group and no information regarding experience of the researchers. All studies had a BLow^risk of bias found in domain two (study design) and domain five (reporting of results), and almost all studies had a BLow^risk of bias found in domain four (descriptive anatomy). The AQUA tool evaluation can be found in Table 2.

Prevalence of the artery of Adamkiewicz
A total of 60 studies (n = 5437 subjects) reported data on the prevalence of the AKA. The pooled prevalence estimate (PPE) of the AKA was 84.6% (95% CI 79.7-89.0) ( Table 3).

Continuity of the artery of Adamkiewicz
A total of seven studies (n = 375 patients with AKAs) were included in an analysis of the continuity of the AKA from the aorta to the anterior spinal artery. The results showed that AKA continued from the aorta to the anterior spinal artery in 71.3% of patients (95% CI 45.8-91.6; I 2 95.6%, 95% CI 92.8-97.2; p < 0.001).

Morphometric analysis of the artery of Adamkiewicz
Five studies (n = 324 patients with AKA) analyzed the morphometric data of the AKA. The analysis showed a pooled mean diameter of 1.09 mm (95% CI 0.69-1.50; I 2 36.2%; p < 0.001).

Discussion
Because the AKA originates from the lumbar arteries, it may be prudent to preserve the blood flow from the lumbar arteries when a thoracoabdominal aortic repair is planned [5,63]. Concomitant or previous abdominal aortic repair and extensive thoracic aorta exclusion by means of multiple stent grafts are associated with a significantly higher risk of paraplegia [64]. After the interruption of most of the  intercostal and lumbar arteries, the residual collateral blood supply is marginal, and in some cases, the spinal cord may become extremely prone to injury due to arterial hypotension or low cardiac output from any cause [65]. During aortic repair, preservation, reattachment, or reconstruction of the intercostal or lumbar arteries can maintain the blood supply to the spinal cord [66,67]. Depending on the number of intercostal or lumbar arteries that require reconstruction, the ischemic duration may be prolonged during reconstruction. In our study, in patients with an AKA present, 11.3% had two AKAs, with bilateral AKAs present in 26.7% of these patients. The preoperative identification of the AKA and its anatomical characteristics allows for superior surgical planning, such that the surgical time and postoperative spinal complication risk are decreased [31]. Therefore, AKA identification is of interest for surgeons aiming to reconstruct intercostal or lumbar arteries in order to prevent postoperative spinal ischemic complications [3]. With respect to the continuity between the radicular arteries (including the AKA) and the anterior spinal arteries, the AKA continued from the aorta to the anterior spinal artery in 71.3% of the patients in our study. When this continuity is present, blood may drain away from the spinal cord through the anterior spinal arteries and the radicular arteries, acting as stealing channels by rerouting the blood to be distal to an aortic obstruction [5]. During aortic crossclamping, back-bleeding from the ostia of the posterior intercostal and lumbar arteries may be a clinical   [68]. The detection of the AKA can be difficult because of the various possible levels of origins of the artery, its small size, the amount of time needed to obtain the angiogram, and complications that can occur during surgical procedures [14,26]. In our study, the pooled mean diameter of the AKA was 1.09 mm. Various techniques have been devised to preoperatively identify the location and anatomy of the AKA, such as CTA [55], MRA [7], and DSA [9]. These techniques can be used to identify both the level and the laterality of the artery, which can affect a surgeon's approach to an aneurysm or spinal lesion. We have included three DSA images with injected contrast into left radicular artery at the level of T4 (Fig. 5), T8 (Fig. 6), and T11 (Fig. 7). In our meta-analysis, cadaveric studies had the highest prevalence of an AKA (97.5%), and among the different imaging modalities, MRA and CTA had the highest prevalence rates (88.3% and 88.1%, respectively), while DSA had the lowest prevalence rate (75.4%). In spite of its apparent success in detecting an AKA, MRA has been shown to be inferior to DSA in terms of evaluating vessel continuity, sharpness, and background homogeneity [7]. Furthermore, compared with CTA, a more limited field of view is a major disadvantage of MRA [61]. As a result, MRA may fail to depict the clinically important collateral vessels to the AKA in some patients, when a collateral source  is the internal thoracic artery or the thoracodorsal artery [69]. Despite DSA studies reporting a lower prevalence rate of the AKA than MRA and CTA in our meta-analysis, DSA remains the Bgold standard^for identifying spinal cord vasculature as it is both safe and efficient [9]. A possible reason for this discrepancy could be the small number of patients included in our DSA analysis as compared to the number of patients included in our MRA and CTA analyses.
Future studies should examine the blood supply and the collateral circulation of the spinal cord in the presence of degenerative atherosclerotic or dissecting aneurysm, or after a surgical or endovascular aortic procedure. In these patients, the disease and the surgical procedure may occlude several segmental arteries and promote collateral vessels enlargement, significantly altering the normal patterns of blood supply to the spinal cord [5].
Our meta-analysis was limited by the high amount of heterogeneity between the studies. However, the number of included studies and their large sample sizes mitigate this limitation. As cadaveric dissection is the gold standard for anatomical considerations, more cadaveric studies should assess prevalence of AKA, especially performed on subjects poorly represented in our meta-analysis, such as Africa, South America, and Oceania.
Because of the lower prevalence of AKA in radiological studies, surgeons should keep in mind that these results might be false negative. In this case, the risk of iatrogenic injury to the AKA during thoracolumbar surgical procedures is increased. More accurate imaging methods should be developed to assess the true prevalence of AKA.
To ensure spinal cord safety, preoperative AKA identification and its subsequent reconstruction or preservation are effective adjuncts for more secure protection of the spinal cord, along with other adequate management strategies.

Conclusions
Our main findings revealed that the AKA was found to be present in the vast majority of the general population (84.6%), most often as a single vessel (87.4%) originating between T8 and L1 (89%) on the left side (76.6%). Based on our anatomical findings, we recommend that efforts should be made to identify and subsequently reconstruct or preserve the AKA to prevent postoperative neurological deficit due to spinal cord ischemia in vascular and endovascular surgical procedures in the thoracolumbar spinal cord.