Background

Since Niederhuber et al. first introduced the totally implantable venous access device (TIVAD) at the MD Anderson Cancer Center in 1982 [1], TIVAD systems have gained worldwide popularity in oncology patients [2]. The number of implanted TIVADs is increasing, with more than 400,000 sold each year in the USA [3]. The use of a TIVAD allows for the long-term administration of venotoxic compounds, reduces the risk of infection, markedly alleviates the burden of intravenous therapy and thereby improves these patients’ quality of life, as this device does not require any external dressing. [3–5] Nevertheless, approximately 15 % of patients experience catheter-related complications [6]. The implantation of a TIVAD can be performed by different methods, such as percutaneous insertion and surgical venous cut-down [5, 7]. Even through, percutaneous TIVAD insertion has become the preferred method of implantation worldwide [5].

Several meta-analyses [8, 9] and the latest review [10] have recommended the routine utilization of ultrasound guidance in practice. With the help of ultrasound guidance, the percutaneous approach has the lowest rate of early complications [11]. Oncologists are most concerned with long-term complications occurring during the use of TIVADs [12]. Because the internal jugular vein (IJV) and subclavian vein (SCV) are the most common access sites to implant catheters in the superior vena cava (SVC) for long-term use [13, 14], it would be helpful to know which site is associated with fewer complications in the long-term follow-up.

Although several studies comparing the IJV and the SCV have been reported, most are small series of patients with conflicting results [15–18]. To date, neither valid recommendations nor guidelines concerning the choice of access site and long-term complications of TIVADs have been elaborated. In this meta-analysis, we sought to assemble the most robust dataset currently available to address a single focused clinical question: which access site, the IJV or the SCV, has fewer late complications for the percutaneous insertion of TIVADs?

Methods

Search strategy

We performed the meta-analysis in accordance with the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [19, 20]. Eligible studies were searched in online databases including PubMed, Embase, Web of Science, and OVID EMB Reviews, from inception to December 2015. A variety of synonyms for “totally implantable venous device”, “internal jugular vein”, and “subclavian vein” were combined. The complete search process is presented in Table 1. A manual search of the citations and references in the articles retrieved for full review was conducted to identify the potentially eligible studies. No limitations were placed on the time period of the trial or the reporting language. Authors were contacted for additional information if necessary.

Table 1 Search process
Full size table

Inclusion and exclusion criteria

All available randomized controlled trials (RCTs), non-randomized cohort studies that compared the IJV with the SCV as the puncture site for a TIVAD in all age groups, were included. Letters, editorials, case reports, review articles, and animal experimental studies were excluded. In order to make the clinical heterogeneity between studies smaller, the studies with follow-up less than 180 days were excluded. If a study investigated multiple access sites (IJV, SCV, and cephalic vein) [16, 18, 21], only the data from the IJV and the SCV were included.

Data collection

Data extraction was performed by two independent authors (SYW and JXH). Agreement between the two reviewers was measured using the k statistic. Any discrepancies were resolved by discussion with the remaining authors. Demographics, clinical characteristics (age, brand of TIVAD used) and technique used (IJV percutaneous insertion, SCV percutaneous insertion, with or without ultrasound guidance or fluoroscopy) were collected. The complications of TIVAD were categorized into infectious complications, thrombotic complications, and mechanical complications [22, 23].

The primary outcomes were the incidence of TIVAD-related infections and thrombotic complications from the time of TIVAD insertion to TIVAD removal or the end of study; TIVAD-related infections were defined according to updated guidelines by the Infectious Diseases Society of America [24] and included pocket infection, local infection, and catheter-related bloodstream infection [25]. Catheter-related thrombotic complications were defined as a mural thrombus extending from the catheter into the lumen of a vessel and leading to partial or total catheter occlusion, with or without clinical symptoms (including fibrin sheath, deep vein thrombosis, major and complete thrombosis), [26, 27] which would be diagnosed using Doppler ultrasound, [15] follow-up chest radiography or chest computed tomography [17]. The secondary outcome was the rate of major mechanical complications after insertion of the TIVAD and follow-up. Major mechanical complications were defined in accordance with the Clavien-Dindo classification of surgical complications (grade III /IV/V) [28], including catheter malfunction (including infusion malfunction, aspiration malfunction, a combination of both, namely catheter occlusion [29]), catheter dislocation (also called malposition/migration; namely, the tip of catheter lying in a different vein from the intended superior vena cava [30]), catheter fracture (breakage or fracture of the catheter, including the breakage or disconnection of junction between the catheter and the reservoir, with or without embolism by catheter fragments), pinch-off syndrome, port rotation, port extrusion, hemorrhage, and extravasation. In addition, if there were more than three included studies and the complication was common, the data for a single major mechanical complication was pooled for meta-analysis. Late complications were unlikely to be due to the port implantation procedure itself [4], so immediate mechanical complications, such as pneumothorax, arterial puncture, and hematoma, which were procedure-related, were excluded in this meta-analysis. Other immediate mechanical complications such as primary malposition, which could be solved immediately with or without fluoroscopic control [15], were not included in this study. Above all, immediate mechanical complications were not included in this study. In addition, minor mechanical complications (Clavien-Dindo grade I/II), such as catheter looping, [31] were also excluded.

Quality assessment

The quality of RCTs was assessed using the Cochrane Collaboration’s tool for assessing risk of bias guided by the Cochrane Handbook for Systematic Reviews of Interventions (version 5.1.0) [32]. Six domains were evaluated: sequence generation, allocation sequence concealment, blinding, incomplete outcome data, selective outcome reporting, and other sources of bias. The overall risk of bias in each study was assessed using the following judgments: low, moderate, or high, which was specified in the study by Ata-Ali [33].

The methodological quality of each nonrandomized observational study was evaluated by the Newcastle-Ottawa scale, which consists of three domains: patient selection (0–4 points), comparability of the study groups (0–2 points), and assessment of outcome (0–3 points) [34]. A quality score of 0–9 points was allocated to each nonrandomized study. RCTs with low risk of bias and nonrandomized studies achieving ≥ 7 points were considered to be of high quality.

Statistical analysis

All of the available data were binary outcomes; therefore, they were combined as pooled odds ratio (OR) with 95 % confidence intervals (CIs). Heterogeneity of outcomes was diagnosed by Q statistics (with a significance level set at P = 0.10) and I2 statistics (>75 % indicating high heterogeneity) [35, 36]. The random-effects model was used in all analyses to produced more conservative and cautious estimates [9].

Subgroup analyses were conducted for the outcomes of TIVAD-related infections, catheter-related thrombotic complications, and total major mechanical complications. Data stratified according to patient’ age, whether antibiotic prophylaxis was used, whether ultrasound guidance was used, were analyzed to investigate clinical factors affecting our outcomes. Sensitivity analyses were conducted to examine the robustness of the effect by alternatively using a fixed-effects model. We also did sensitivity analyses according to two different study designs (RCT and non-randomized cohort study). Only outcomes with more than one studies were included in the sensitivity analyses. Publication bias was assessed using Egger regression asymmetry test [37]. A two-tailed P value < 0.05 was considered statistically significant, except otherwise specified. Statistical analysis was performed using R software (https://www.r-project.org; last access 29 March 2016) and Stata software version 12 (StataCorp LP, College Station, USA).

Results

A total of 2106 potentially eligible studies were initially identified, and 2078 were excluded after screening the titles and abstracts. The remaining 28 articles were fully reviewed. Of these, 14 were excluded because the data were not extractable, two studies were duplicate reports with different outcomes, and one was a RCT with only 30 days follow-up which did not fulfill the criteria of minimum follow-up of 180 days in this meta-analysis. In addition, one study [23] was identified from the citations of the study by Araujo [15].

Therefore, 12 studies [15–18, 21, 23, 31, 38–42] including 3905 patients (1824 patients in the IJV group and 2081 patients in the SCV group) published from 2008 to 2015 were included (Fig. 1). Agreement on study selection between the two reviewers was high (k = 0.94). Among the included studies, there were three RCTs [16, 39, 41] and two prospective non-randomized controlled trials [15, 31]. The remaining seven studies [17, 18, 21, 23, 38, 40, 42] were retrospective. The characteristics of the included studies are summarized in Table 2.

Fig. 1
figure 1

Flowchart of the literature search and selection process

Full size image
Table 2 Baseline characteristics of studies included in the meta-analysis
Full size table

The risk of bias of the RCTs included in this meta-analysis is summarized in Table 3. Of the three RCTs, two [16, 39] were considered to have low risk of bias, and one [41] had a moderate risk of bias. Among the nine nonrandomized studies, seven [15, 17, 18, 21, 23, 40, 42] were considered to be of high quality, and two [31, 38] were regarded as being of low quality (Table 4).

Table 3 Cochrane summary assessment of risk of bias for included RCTs
Full size table
Table 4 Newcastle-Ottawa Scale for nonrandomized cohort studies
Full size table

Primary outcomes

The pooled data from 11 studies [15–18, 21, 23, 38–42] that assessed TIVAD-related infections (Fig. 2a) in 3767 patients showed no significant differences between the IJV and SCV groups (2.53 % and 3.77 %; OR 0.71, 95 % CI 0.48–1.04; P = 0.081) with no significant between-study heterogeneity (I2 = 0.0 %; P = 0.963). Catheter-related thrombotic complications were reported in 11 studies [15–18, 23, 31, 38–42] that investigated 3802 patients (Fig. 2b). There were no significant differences between the IJV and SCV groups (2.05 % and 2.05 %; OR 0.76, 95 % CI 0.38–1.51; P = 0.433), with no significant between-study heterogeneity (I2 = 30.2 %; P = 0.159).

Fig. 2
figure 2

Forest plot and meta-analysis of TIVAD-related infections (a) and catheter-related thrombotic complications (b)

Full size image

Secondary outcomes

Data on major mechanical complications were available in 11 studies, [15, 17, 18, 21, 23, 31, 38–42] which evaluated 3665 patients (Fig. 3a). The rate of total major mechanical complications was significantly higher in the SCV group than in the IJV group (3.75 % in the IJV group and 9.70 % in the SCV group; OR 0.38, 95 % CI 0.24–0.61; P < 0.001), with low between-study heterogeneity (I2 = 31.6 %; P = 0.147). Additionally, there were three major mechanical complications that more than three studies reported: catheter dislocation, malfunction, and catheter fracture. In other words, these three complications were common. As a result, the data for the three major mechanical complications were pooled for meta-analysis.

Fig. 3
figure 3

Forest plot and meta-analysis of major mechanical complications, including total major mechanical complications (a), catheter dislocation (b), malfunction (c), and catheter fracture (d)

Full size image

Seven studies [15, 17, 18, 23, 31, 38, 41] that reported on catheter dislocation in 2463 patients showed a significant difference favoring the IJV group (1.08 % in the IJV group and 2.54 % in the SCV group; OR 0.43, 95 % CI 0.22–0.84; P = 0.013) (Fig. 3b). Nine studies [15, 18, 21, 23, 31, 39–42] that assessed 3085 patients reported on malfunction, and the difference was statistically significant in favor of the IJV (2.80 % in the IJV group and 5.56 % in the SCV group; OR 0.42, 95 % CI 0.28–0.62; P < 0.001) (Fig. 3c). Pooling the data of seven studies [15, 17, 18, 21, 39, 40, 42] including 2795 patients that reported on catheter fracture showed no significant difference between the two groups (0.82 % in the IJV group and 2.91 % in the SCV group; OR 0.47, 95 % CI 0.21–1.05; P = 0.065) (Fig. 3d). All of the three major mechanical complications showed no significant heterogeneity (Fig. 3b, c, and d).

Subgroup analyses

Subgroup analyses showed that use of antibiotic prophylaxis did not influence the incidence of TIVAD-related infections (Table 5). In the subgroup analyses of ultrasound guidance, only one study [38] used ultrasound to guide the TIVAD insertion for all patients, and six studies [15, 18, 21, 23, 39, 40] used anatomical landmark technique for all patients (Table 1). The results showed that the use of ultrasound guidance did not affect the risks of TIVAD-related infections and catheter-related thrombotic complications; however, it moderated the effect size of total major mechanical complications (Table 5). In addition, subgroup analyses stratified by the patients’ age showed no change in our conclusions for the outcomes of TIVAD-related infections and catheter-related thrombotic complications; however, in the subgroup of adults, the risk of total major mechanical complications was not significantly different between the two groups with higher heterogeneity (I2 = 56.5 %; P = 0.100) (Table 5), indicating that heterogeneity in the total major mechanical complications was due to other factors, rather than patients’ age.

Table 5 Subgroup analyses comparing IJV versus SCVa
Full size table

Sensitivity analyses

Sensitivity analysis by alternatively using a fixed-effects model did not show any relevant influence on all of the outcomes except catheter fracture, which showed a reduced risk in the IJV group (OR 0.38, 95 % CI 0.18–0.78; P =0.008) with low heterogeneity (I2 = 0.0 %; P = 0.436) (Table 6). In the sensitivity analyses, RCTs and non-randomized studies showed the same results for the overall OR estimates for TIVAD-related infections, catheter-related thrombotic complications, total major mechanical complications, and malfunction (Table 6).

Table 6 Sensitivity analyses comparing IJV versus SCV
Full size table

Publication bias

Publication bias was assessed by Egger regression asymmetry test, which did not suggest any significant publication bias for TIVAD-related infections (P = 0.343), catheter-related thrombotic complications (P = 0.147), total major mechanical complications (P = 0.502), catheter dislocation (P = 0.959), malfunction (P = 0.265), and catheter fracture (P = 0.730) among the included studies. Egger funnel plots for TIVAD-related infections, catheter-related thrombotic complications, and total major mechanical complications were shown in Fig. 4.

Fig. 4
figure 4

Egger funnel plots for TIVAD-related infections (a), catheter-related thrombotic complications (b), and total major mechanical complications (c)

Full size image

Discussion

This meta-analysis of three RCTs and nine non-randomized cohort studies, all of which included a total of 3905 patients, compared the efficacy of the IJV and the SCV as the percutaneous access site for a TIVAD. The results suggested that compared with the SCV, the IJV seems to be a safer venous access site with significantly reduced major mechanical complications. To be more specific, the IJV is associated with a lower risk of catheter dislocation and malfunction. We found no significant differences in TIVAD-related infections and thrombotic complications. On subgroup analyses, the use of antibiotic prophylaxis did not influence the incidence of infectious complications; the use of ultrasound guidance did not affect the risks of TIVAD-related infections and catheter-related thrombotic complications, but it moderated the effect size of total major mechanical complications. On sensitivity analyses, the overall estimates of all endpoints except catheter fracture remain robust by alternatively using a fixed-effects model; both RCTs and non-randomized cohort studies showed the same results for TIVAD-related infections, catheter-related thrombotic complications, and total major mechanical complications.

Easy and reliable vascular access is particularly important in cancer patients. The introduction of TIVADs has made the treatment of oncology patients more comfortable and convenient, because dressing is not required and daily activities of the arms need not be restricted once the port is implanted [43]. Compared with external indwelling catheters, advantages of the TIVAD include reduced risk of infection, greater patient acceptance and requiring less maintenance [3, 44]. However, like other short-term central venous catheters, TIVAD also presents risks itself after long-term indwelling.

The rate of TIVAD-related infections in our study was 2.53 % in the IJV group and 3.77 % in the SCV group, which was consistent with the reported results (3–10 %) of a recent review [3]. Subgroup analysis showed that the use of antibiotic prophylaxis did not influence the overall estimates for infections. We did not find a significant difference between the two groups in terms of TIVAD-related infections. Because patients with cancer are susceptible to infections due to immune depression and neutropenia [3, 13], we also suggest that measures should be taken to reduce the risk of infections, including sterile precautions during TIVAD insertion, and optimal aseptic catheter maintenance [12, 45, 46].

The incidences of catheter-related thrombosis in this meta-analysis were both 2.05 % in the IJV and in the SCV group, which were consistent with the results (0.3–28.3 %) of a review by Verso [47]. Thrombosis represents a major problem in oncology practice [48]. Cancer patients are usually at increased risk of venous thrombosis [49]. Although anticoagulant prophylaxis is controversial, routine heparin flushing of the port seems to be sufficient to prevent thrombosis formation [12]. In this meta-analysis, the majority of included studies reported on use of heparinized saline flushing regularly for primary prevention of catheter-associated thrombosis, and only two studies [17, 23] did not mention the use of heparin for routine maintenance of the TIVAD, which, however, did not mean heparin was not used. Actually, prophylactic heparin flushing has become the routine of clinical practice [50]. Consequently, subgroup analysis stratified by whether heparin was used was not conducted. Furthermore, placement of the catheter tip low in the SVC or at the atriocaval junction resulted in a lower risk of thrombosis than placement higher in the SVC [51, 52]. As a result, the use of fluoroscopy after implantation was recommended to identify tip position and ensure adequate catheter length (catheter tip below the T3 level) [52]. When thrombosus occurs, we may resort to medical treatment (anticoagulant agents or thrombolytic drugs) or even remove the TIVAD [48].

Catheter dislocation (also defined as a secondary malposition) can occur months after implantation of the TIVAD if the catheter tip is dislocated from its original position [2, 13]. Radiological control of the catheter tip using chest fluoroscopy after implantation is mandatory [12, 53]. In fact, all the included studies in this meta-analysis used fluoroscopy to confirm the catheter tip in the right place. The reason why catheter dislocation is more common in SCV group is still unclear. However, according to a retrospective study by Paleczny [14], spontaneous dislocation of the vascular port and catheter tip associated with changes in body position was found by chest radiograph in two patients with the catheters placed only in the SCV rather than in the IJV group. This phenomenon indicates that TIVAD insertion via the SCV route may be more subject to spontaneous dislocation when changing body position in daily life.

The pinch-off syndrome is specifically associated with the SCV approach [54]. Due to the compression of an implantable port between the clavicle and the first rib, the pinch-off syndrome can result in mechanical compression and shearing forces on the catheter lines [55], which may lead to malfunction, damage, and even fracture of the catheter after material fatigue [56], with embolization in the lung vascular bed. Pinch-off syndrome serves as a warning prior to catheter fracture, a rare but serious complication [57]. We confirmed that compared with IJV, SCV was associated with more incidences of major mechanical complications and many (malfunction, damage and catheter fracture) may be due to pinch-off syndrome.

Our meta-analysis is unique and presents important implications for clinicians in that, to our knowledge, it is the first study to systematically summarize the association of venous access sites for percutaneous implantation of a TIVAD and long-term morbidity. We used a comprehensive search strategy and systematic review method, following the MOOSE guidelines and the PRISMA statement. We limited heterogeneity by including only studies with more than 180 days follow-up. Furthermore, we redefined the outcome of malfunction to cover all aspects of catheter malfunctioning, namely infusion and aspiration malfunction as well as a combination of both [29], thereby avoiding potential heterogeneity in the endpoint of malfunction. Moreover, heterogeneity was low to moderate in the analyses of all outcomes, suggesting that variations in findings are compatible with chance alone and not likely to be caused by genuine differences between studies [58].

Our study has the following limitations. First, the majority of included studies were not RCTs and often presented a small sample size. They were carried out in hospitals with different protocols and likely different levels of physician expertise. Second, the definitions of endpoints such as TIVAD-related infections, catheter-related thrombotic complications, were not clearly described in some studies; however, studies were pooled irrespective of their definitions of these endpoints. The heterogeneity in endpoint reporting of the primary studies should be considered as a limitation. Third, in the subgroup analysis, ultrasound guidance diminished the advantage of IJV for the outcome of total major mechanical complications. However, this result should be interpreted with caution, because only one study was included in the subgroup of ultrasound guidance. Fourth, the implantation of a TIVAD can be performed by surgical venous cut-down technique and percutaneous approaches [43], and the results of our meta-analysis only apply to percutaneous approaches. Fifth, some definitions of mechanical complications (port rotation, port extrusion, hemorrhage, and extravasation) were not sufficiently described; these outcomes were included in the outcome of total major mechanical complications and were not individually pooled for meta-analysis. Sixth, because the raw data of the included studies were not available and both arms were comparable in terms of the follow-up period in each study, the results of the analysis did not take into account of the number of catheter days. However, as the cumulative risk of infectious, thrombotic, and mechanical complications increased with increasing catheter exposure, the complications might have been underestimated due to the relatively short follow-up period in some studies [31, 41].

Conclusions

In conclusion, in the present meta-analysis comparing the IJV and the SCV as a venous access site for percutaneous insertion of a TIVAD, we identified a better choice of the IJV in terms of the incidence of major mechanical complications (catheter dislocation and malfunction), but we did not find any statistically significant differences in TIVAD-related infections and thrombotic complications. Given the inherent limitations of the included studies, the findings from our study must be confirmed and updated in a large-scale and well-designed RCT with long-term follow-up.