Central Venous Catheterization (CVC) Site Choice Based on Anatomical Landmark Technique: A Systematic Review and Meta-analysis Short title:CVC Site Choice Strategy

BACKGROUND: Internal jugular vein (IJV) and subclavian vein (SCV) have been the most preferred central venous catheterization (CVC) sites clinically. Based on individual preference and institutional routine, however, the choice of IJV catheterization (IJVC) or SCV catheterization (SCVC) is lack of high-level evidence. We sought to provide useful clinical strategy by comparing the catheterization time, success rate and the mechanical complications involved of the two CVCs. METHODS: We systematically reviewed eligible studies from PubMed, OVID, Cochrane and ClinicalTrials.gov till February 2020. The primary outcomes were catheterization time and overall success rate, and the secondary outcomes were the rst-attempt success rate and the instant mechanical complications. RESULTS: A total of 3378 patients from 7 studies were included in the analysis. Neither difference was found on the catheterization time (SMD 95% CI: -0.095-0.124, p=0.792), nor any difference on the overall success rate (RR=1.017, 95% CI: 0.927-1.117, p=0.721, I 2 =89.6%) between the 2 procedures. However, subgroup analysis showed overall success rate of IJVC was signicantly lower than that of SCVC (RR=0.906, 95%CI:0.850-0.965, p=0.002) in adult patients. The rst-attempt success rate of IJVC group was higher than SCVC group in the adult patients (RR=1.472, 95%CI:1.004-2.156, p=0.047). No signicance was detected in arterial injury (RR=1.137, 95% CI: 0.541-2.387, p=0.735) and pneumothorax (RR=0.600, 95% CI: 0.32-1.126, p=0.112) between the two procedures. Hematoma was signicantly more in IJVC group than that in SCVC group (RR=2.824, 95% CI: 1.181-6.751, p=0.02). CONCLUSIONS: Compared with IJVC, SCVC has a higher overall success rate while a lower rst-attempt success rate in adults, and has involved with less hematoma. However, more high-quality studies are essential to provide better evidence especially in pediatric patients. TRIAL REGISTRATION: This meta-analysis was pre-registered in PROSPERO (CRD42020165444).


Introduction
Subclavian vein (SCV) and internal jugular vein (IJV) catheterizations were rst described in the 1950s and 1960s 1 . Since then, various studies have shown other routes to the central venous system. Still, the internal jugular vein catheterization (IJVC) and subclavian vein catheterization (SCVC) are the most preferable ways for uid and medication therapy, transfusion, blood volume monitoring, and parenteral nutrition 2 .
However useful it is, the central venous catheterization (CVC) itself and associated adverse events shouldn't be neglected. The longer time took in CVC, the more discomforts the patients will experience 3 , and a higher insertion success rate correlates with lower mechanical complications 4 , which are expensive to treat and hazardous to patients 2 . Most of the CVC practice choice is currently based on the experience and institutional routine that are lack of high-level evidence. Besides, since the ultrasound assisted catheterization was introduced into clinics, doctors tend to resort to ultrasound assistance when confronted by di cult catheterizations 3 , which seemingly dwarfs the traditional landmark method. Yet, no de nitive evidence has shown its superiority over traditional anatomical ways in decreasing catheterization time, complications and cannulation failure, let alone the essential training, personnel and budget [5][6][7][8][9] . In fact, the landmark-based anatomical CVCs are overwhelmingly used in the less developed countries compared to ultrasound assistance 10 .
Hence, whether ultrasound technique should be widely used among all kinds of patients and all CVCs is still uncertain 2,7,11,12 . Plus, perioperative management of CVC counts not only on the most frequent instant complications that include arterial injury, pneumothorax and hematoma 2, 13 , also on the success rate and time when e ciency matters a lot.
So far, no guidelines or high levels of evidence concerning the choice of IJVC or SCVC have been published. Therefore, we conducted this systematic review and meta-analysis, comparing the landmark-based catheterization time and the success rate, as well as the instant mechanical complications between IJVC and SCVC, hoping to provide useful strategy for CVC management.

Search strategy
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations during the preparation of this review. And the protocol was registered with PROSPERO (registration no. CRD42020165444). Eligible studies were searched from the inception to February 2020 in PubMed, OVID, Cochrane and ClinicalTrials.gov, with key words "internal jugular vein/venous", "subclavian vein/venous", "central venous catheterization", "central venous cannulation", "central venous insertion", "central line catheterization/cannulation/insertion". Potentially eligible studies were also manually searched from the references in the included studies. The limits to these searches were clinical trials, English language and humans.

Inclusion and exclusion criteria
All randomized controlled trials (RCTs) and non-randomized prospective cohort studies (PCs) that compared the two primary outcomes (catheterization time, overall success rate) or the secondary outcomes ( rst-attempt success rate, arterial injury, pneumothorax, hematoma) based on landmark technique between IJVC and SCVC were included. If a study included more than IJVC and SCVC, only the data of IJVC and SCVC were included, and if only one of the two procedure was involved, it was excluded. Case reports, review articles and editorials were excluded. If ultrasound technique was resorted to after a failed landmark-based catheterization, it was considered a failure, and if used under any circumstances, it would be excluded.

Data collection
Quali ed studies were included by the consensus of two authors (WTX and QQZ), if disagreement occurred, the third author (XL) was asked to make the nal decision. Data extraction from the quali ed studies was carried out by two authors (ZGC and DF) in duplicate, a team discussion would be performed and nal decision made in case of discrepancies. Data regarding the rst author, published year, demographics of the patients, technique used (with or without ultrasound assistance), research type, overall sample size, number of patients in each study arm, catheter lumen, the catheterization mean time and standard deviation (SD), the rst-attempt success, the overall success, the failures, and the incidents of arterial injury, pneumothorax and hematoma were all recorded in EXCEL 2010. Original authors were contacted if associated data was ambiguous or missed. The data information about the authors, magazines, publication years was concealed for later evaluations.
The primary outcomes were the catheterization time and overall success rate. The catheterization time was de ned as the time spent from the start of the skin preparation till the catheter was inserted in the right place, recorded as mean time ± SD. The overall success rate was successful catheterizations among all participants in one group. Successful CVC was de ned as the cannulation was achieved within three attempts by landmark technique 14 . The secondary outcomes were the rst-attempt success rate and the incidents of arterial injury, pneumothorax and hematoma. The rst-attempt success rate was de ned as the catheter syringe penetrated into the aimed vein and successfully catheterized during the rst try among all the patients in a group.

Risk of bias assessment
Our primary intention was to extract data from randomized trials, but only 3 studies were quali ed. Hence, we made a group discussion and included non-randomized studies too. For randomized studies, the qualities were independently assessed by JW and XHJ using The Cochrane Collaboration's tool 15 . Six domains were included: random sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting and other bias. A score was assigned to each domain by consensus. The risk of bias in each RCT was ranked as low (over 5 domains with "yes"), moderate (3 or 4 domains with "yes"), and high (less than 2 domains with "yes") 16 (Table 1).
For non-randomized studies, the qualities were evaluated by HPZ and JS independently with the Newcastle-Ottawa scale 17 (Table 2). Three main domains were included: patient selection, comparability and outcome assessment. The quality score of each study was summed, and the study was considered to be of high quality if the score was over 7 points 16 . The corresponding author would be consulted and made the nal decision if disagreement occurred during these evaluations.

Statistical analysis
Most of the data were extracted and calculated from the manuscript. If the data required were not in the manuscript, we tried to contact the corresponding authors. Continuous data in this review was the catheterization time that was recorded as mean ± SD. The dichotomous data included the overall success rate, rst-attempt success rate and the mechanical complications (arterial injury, pneumothorax and hematoma). Heterogeneity of the outcomes was reported using I 2 statistics (>75% indicating high heterogeneity, 25%-50% indicating moderate heterogeneity and <25% indicating low heterogeneity) 18 . Since age has been an important variable in catheterizations, we would set age as an indicator for subgroup analysis. The funnel plot was used to examine the risk of publication bias in the outcomes 19 . All of the analyses were performed using STATA 12.0 (Stata Corp LP, College Station, USA). Two-sided of alpha 0.05 was used for all statistics.
In total, 1554 records were found and screened, 43 were extracted after title screening. Among which, 36 articles were excluded because of intervention or article types. And only 7 studies 4, 13, 14, 20-23 were included in the qualitative and quantitative synthesis ( Figure 1). A total of 4520 patients were included in the seven studies. However, 962 patients in IJVC group and 180 patients in SCVC group were dropped out, because ultrasound guidance was taken since the rst attempt. Finally, 1783 and 1595 patients were analyzed in each group. For studies that included more than just IJVC and SCVC 4, 13, 22, 23 , we extracted data only related to IJVC and SCVC. The details of the included studies are summarized in Table 3.

Primary Outcomes
Catheterization time Only two studies 13, 14 recorded the catheterization time that was not their original established outcome, though, the meta-analysis showed no difference in catheterization time between IJVC and SCVC (SMD 95% CI: -0.095-0.124, p=0.792. Figure 2). The operators and the patients in the two studies signi cantly varied ( Table 3) that were complicated doctors (experienced and non-experienced) to adults 13 and experienced doctors to children (<6 y) 14 . Low heterogeneity (I 2 =0.0%) was detected; however, the clinical heterogeneity could've played a key role in this comparison. The inconsistent and limited studies we retrieved precluded us from any meaningful conclusion between the two procedures. More high-quality randomized clinical trials focusing on the catheterization time would be essential to clarify this issue.
Overall success rate Five studies 4,13,14,20,22 compared the overall success rate were included in the quantitative analysis. No signi cance was detected in meta-analysis (RR=1.017, 95% CI: 0.927-1.117, p=0.721, Figure 2.). This primary outcome was characterized by signi cant heterogeneity (I 2 =89.6%). Subgroup analysis based on age (18y) was performed. The subgroup analysis showed that the overall success rate of IJVC was signi cantly lower than that of SCVC (RR=0.906, 95%CI:0.850-0.965, p=0.002, Figure 3) in adult patient group. It's noteworthy that all catheterizations were nally achieved in a small sample size (n=45) study 21 and it was excluded from the quantitative analysis.
However, the pediatric patients group still posed high heterogeneity (I 2 =91.2%), two studies favored the IJVC for higher success rate 20,22 while the other two favored SCVC 4, 14 .

Secondary Outcomes
First-attempt success rate Only two studies 14,21 contributed to this secondary outcome, however, the heterogeneity was signi cant (I 2 =78.4%). Operated both by experienced doctors, the main differences lie in the age of patients and the catheter lumen (Table 3), for patients under 6 years old 14 , SCVC tended to achieve higher rst-attempt success rate ( Figure 3); while for adults 21 , IJVC had a signi cant higher rst-attempt success rate that was contrary to the overall success rate (Figure 3).

Arterial injury
All the included studies recorded arterial injury complication, probably because it's one of the most common instant mechanical complications both in traditional landmark cannulation and the ultrasound assisted cannulation 2,24 . The seven studies together created a moderate heterogeneity (I 2 =70.2%), but no signi cant statistical difference was detected between the two insertion sites in the quantitative analysis (RR=1.137, 95% CI: 0.541-2.387, p=0.735, Figure 2).

Subgroup analyses
We conducted subgroup analyses to investigate the in uence of patients' age on all outcomes. For the catheterization time, neither adult patients were detected of any statistical difference between the two procedures (SMD=-0.005, 95%CI: -0.123-0.113, p=0.929), nor were any statistical difference in pediatric patients (SMD=0.141, 95%CI: -0.145-0.437, p=0.349). Since this result was only recorded but not pre-designated as their outcome in the original two studies, as one of our primary outcomes, it seemed embarrassing. For overall success rate, the subgroup analysis by age has shown a signi cant lower overall success rate in IJVC group than that in SCVC group in adult patients (RR=0.906, 95%CI:0.850-0.965, p=0.002, Figure 3), however, high heterogeneity was shown in pediatric patients (I 2 =91.2%) and no signi cant difference was detected (RR=1.056, 95%CI:0.931-1.198, p=0.395, Figure 3). The rst-attempt success rate, when analyzed in subgroups, IJVC group achieved a higher rst-attempt success rate than SCVC group in the adult patients (RR=1.472, 95%CI:1.004-2.156, p=0.047, Figure 3), whereas in the pediatric patients, no signi cant difference was detected (RR=0.931, 95%CI:0.788-1.100, p=0.399, Figure 3). The subgroup analyses of arterial injury didn't detect any statistical differences neither in adult patient group (RR=2.767, 95%CI:0.729-10.511, p=0.135, Figure 3) nor in pediatric patient group (RR=0.940, 95%CI:0.406-2.177, p=0.149, Figure 3) between the two procedures. Likewise, the subgroup analyses of pneumothorax incidents didn't nd any statistical differences in different age groups between the two CVC ways (Figure 3). The hematoma incidence rate, however, when analyzed by age subgroups, IJVC showed signi cant higher rate than SCVC in adult patient group (RR=7.235, 95%CI:1.534-33.919, p=0.012, Figure 3), whereas in the pediatric patients, IJVC posed a higher risk trend but no signi cance was found (RR=1.870, 95%CI:0.689-5.077, p=0.219, Figure 3).

Sensitivity analyses
We conducted sensitivity analyses with a xed effects model alternatively. The outcome of the catheterization time and rst-attempt success rate were also performed regardless of small number of studies. The results of sensitivity analyses were largely consistent with the primary analysis. (supplemental data)

Publication bias
The funnel plots for the outcomes of overall success rate, arterial injury, pneumothorax and hematoma were conducted. Publication bias on the catheterization time and rst-attempt success were not calculated because of limited studies included. Since the number of included studies for any outcome is small, the funnel plot could have limited power to detect the bias. (supplemental data)

Discussion
Our meta-analysis of seven studies shows that neither catheterization time nor overall success rate was signi cantly different between IJVC and SCVC, but the overall success rate of IJVC was signi cantly lower than that of SCVC in adult patients under age subgroup analysis. To our knowledge, this is the rst meta-analysis so far that raises the attention to catheterization time and overall success rate between IJVC and SCVC based on anatomical landmark. However, we cannot draw any conclusion from the primary outcomes, because the catheterization time was only recorded but not predesignated as outcomes in two RCTs that makes it less convincing. Less convincing as it is, catheterization time matters clinically.
Perioperative critical and emergent situations sometimes demand fast central lines for monitoring and therapy 25,26 . And the choice of traditional landmark CVC shouldn't be focused only on the complications 27 , prolonged catheterization time could result in severe complications like cardiac arrest, sudden apnea 28 . The shorter time spent in catheterization, the more time reserved for other medical care and the less discomforts the patients will experience 3 . The real clinical operational time in catheterization, however, differs greatly, in the include 2 studies, the mean time for IJVC in adults was 765 seconds 13 , while in pediatric patients in another study, it was only 187 seconds 14 . The existence of clinical heterogeneity like different patients, operators or even different procedures for catheterization does not make this outcome less valuable, instead, it calls for better training for improvement, and most of all, more high-quality randomized studies focusing on catheterization time are essential to elaborate this issue. The overall success rate has been mostly adopted as an outcome in the comparison of ultrasound assisted CVC and landmark CVC 7,11,29 , it's valuable in our comparison of different CVC approaching sites as well. The lower overall success rate of IJVC in adults may be explained by its anatomy variations 30,31 , including the bifurcations that occur at different sites of the internal jugular vein, the bicuspid on the right side of the internal and the tricuspid on the left side. Another explanation could be that IJVC assisted by ultrasound is most recommended and easier to grasp 9 , while for SCVC, the failures and the instant complications weren't be reduced by ultrasound technique 12,32 , the operators would be more inclined to resort to ultrasound guidance when confronted by di cult cannulations during IJVC.
Some other important ndings emerged from our meta-analysis. First, as one of the secondary outcomes, IJVC had a signi cant higher rst-attempt success rate than SCVC in adult patients, contrary to the overall success rate, which also implied more attempts could've happened in the SCVC group. The rst-attempt success rate correlates not only with the catheterization time, also with the mechanical complications 14,33 . Each attempt increases the risk of complications after two attempts 4 . CVC may result in many complications that could be instant mechanical complications, infection, thrombosis and even cardiovascular events 2, 3 . Infection and thrombosis are mainly ascribed to CVC maintenance, while the mechanical complications mostly occur during the catheterization 2 . Second, compared to SCVC, the arterial injury and pneumothorax in IJVC were not signi cantly different neither in adult patients nor in pediatric patients. However, SCVC induced hematoma was signi cantly less than IJVC, unfortunately, the subgroup analysis showed that pediatric patients might not bene t from it. Those ndings indicate that age could be a potential effect modi er, however, due to limited studies, the interpretation of these differences could be questionable. More high-quality and well-designed studies based on different ages will be needed to clarify the differences of the two CVCs.
There are some limitations in our meta-analysis. First, we anticipated including only RCTs, but found only three RCTs were quali ed 13,14,20 , what's more, none of them were consisted of all age groups and the study quality differed (Table 1). For the PCs included, earning a total score of at least 6, though, they were subject to selection bias, information bias and measurement bias 34 . As is known that the strength of the results that can be drawn from a meta-analysis is dependent on the quality of the included studies, the outcomes in our meta-analysis could be confounded and should be treated with considerable caution. Second, the de nitions of the outcomes that used in the studies were not uni ed. For example, the catheterization time and failure were not strictly de ned in the largest RCT since they were not the pre-setting outcomes 13 , while in another RCT, the catheterization failure was de ned as three consecutive unsuccessful attempts at one site 14 .
Third, signi cant heterogeneity was detected among the studies. It might be understandable in the ages of the patients, however, other aspects like the different experience of the operators, different lumens that could've created more heterogeneity were not well-de ned either (Table 3), thus, the subgroup analyses strati ed by these two were not conducted.
Aside from these limitations, our meta-analysis brought the importance of cannulation sites based on landmark method to the light in CVCs, whereas overwhelming studies concluded the superiority of ultrasound technology in CVCs. We are not trying to devalue the signi cance of ultrasound in CVCs, especially in some di cult catheterizations. What we are trying to clarify is that anatomy and experience matter in both landmark and ultrasound assisted CVCs. Besides, it could be tricky when encountered with limited medical resources, short of personnel and less training in ultrasound technique. Further, it could be much bene cial when CVCs are done fast and clean during the perioperative period.
In conclusion, compared with SCVC, IJVC shows no difference in catheterization time and a signi cant lower overall success rate in adult patients. SCVC produces a lower rst-attempt success rate but less hematoma in adult patients. The shortage of high-quality studies in our meta-analysis undermines our conclusions. But, whenever it's di cult to identify the landmarks of one CVC, another route should be considered 2 . For better evidence in choosing CVC sites, more well-designed, randomized trials are required. Ethics approval and consent to participate This study was pre-registered with PROSPERO (CRD42020165444). No informed consent would be necessary for this meta-analysis Consent for publication Not applicable

Availability of data and materials
The datasets generated and/ or analyzed are available from the corresponding author upon reasonable request.

Competing interests
None Funding This meta-analysis was supported by the grant from National Natural Science Foundation of China (NO. 81671947)

Authors' contributions
All authors were involved in this meta-analysis and have approved the nal manuscript  RCT stands for randomized controlled study.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.