Angioplasty for dysfunctional arteriovenous stulas: a meta-analysis of recent randomized controlled trials compared paclitaxel-coated balloon versus conventional balloon angioplasty

Background Stenosis in arteriovenous stulas (AVF) due to neointimal hyperplasia is one of the most common causes of hemodialysis vascular access dysfunction. Treating patients with dysfunctional AVF with drug-coated balloon (DCB) angioplasty may potentially improve outcomes. This systematic review aimed to compare the effectiveness and safety of DCB angioplasty versus conventional balloon angioplasty by pooling evidence from the most recent randomized controlled trials. Methods We conducted a comprehensive literature search in the Medline, Embase, and Cochrane central databases. Two independent researchers screened the article, extracted interest and evaluated included studies for risk of bias. Pooled estimation was conducted in terms of 6-month target-lesion primary patency (TLPP) and target-lesion reintervention (TLR), as well as other outcomes. Results were expressed with odds ratio (OR) and 95% condence interval (CI). pooled estimation of an effect size of 6-month TLPP and other outcomes, was expressed with odds ratio (OR) and 95% condence interval (CI). Mantel-Haenszel methods with a random-effects model were used for pooling analyses accounted for the dichotomous outcomes and inter-study heterogeneity. Heterogeneity across included studies was assessed using Cochran's Q test and I 2 statistics. A p-value lower than 0.1 or I 2 statistic above 50% indicates heterogeneity among included studies. A visual judgment of funnel plots assessed publication bias (or small-study effect) for symmetry. We also performed a sensitivity analysis using the leave-one-out method. All statistical analyses were performed using R (version 4.0) with the 'meta' and 'dmetar' package (17, 18).


Introduction
Hemodialysis is the most typical renal replacement modality for end-stage renal disease (ESRD) patients, and arteriovenous stulas (AVF) have been recommended as the optimal modality of vascular access (1). Nonetheless, its vascular access dysfunction contributed most to patients' mortality and hospitalization (1,2). Stenosis in AVF due to neointimal hyperplasia is one of the most common causes of hemodialysis vascular access dysfunction (3). Percutaneous transluminal angioplasty with a conventional plain balloon is a routine therapy for stenosis in AVF; however, its durability is a signi cant concern and repeated intervention is constantly needed. It has been reported in some small studies that the use of drug-coated balloons (DCB) with a layer of antiproliferative drug paclitaxel may have the potential to improve outcomes. In contrast, other studies reported a contradictory result (4)(5)(6)(7)(8)(9)(10).
Several randomized controlled trials (RCTs) that compared DCB angioplasty with conventional balloon angioplasty for the treatment of dysfunctional hemodialysis vascular access with a larger sample size have been published (11)(12)(13)(14). Results of these trials were less consistent and led to an ongoing debate on whether DCB angioplasty could indeed improve outcomes. Unfortunately, there is no updated meta-analysis regarding this topic. Therefore, we designed this comprehensive systematic review and meta-analysis of most recent RCTs to evaluate whether DCB angioplasty for the treatment of dysfunctional hemodialysis vascular access is potentially bene cial compared to conventional balloon angioplasty. In addition, the effectiveness and safety endpoints of the DCB angioplasty against traditional angioplasty of the balloon were assessed.
According to Cochrane's Collaboration recommendations and Preferred Reporting Items for Systematic Reviews and Metaanalysis (PRISMA) statement (15), we followed the PRISMA checklist, which can be found in the supplements.

Study selection and data extraction
A comprehensive literature search in the Medline, Embase and Cochrane central databases was conducted from inception to 2021 April. The detailed search syntax is presented in table S1. We restricted the studies to human trials. Additional literature were manually searched by screening relevant references. Studies that meet all the following criteria were included: (1) in hemodialysis patients with dysfunctional AVF, (2) with a randomized controlled design, (3) a comparison between DCB angioplasty and conventional balloon angioplasty, (4) a sample size of at least 50 patients per group. A study was excluded if: (1) results were not published in a peer-viewed journal or data of interest was not available, (2) was not published in English or Chinese, (3) enrolled patients was partly with arteriovenous graft, (4) did not report outcomes of interest. The primary clinical outcome of interest was target lesion primary patency (TLPP) six months after the indexed procedure. We also assessed secondary endpoints, including clinically-driven target lesion revascularization (TLR), access circuit primary patency, access circuit reintervention, severe adverse events and all-cause mortality. In the case of the multiple publications of the same trial, we included the one with a longer follow-up.
Two independent researchers (Y.Z and Q.Y) rst screened the title and abstracted it to identify relevant literatures and then viewed the entire manuscript to check if a study met the prede ned selection criteria. Next, the study setting, sample size, clinical characteristics, interventional procedures, follow-up periods and clinical outcomes were extracted from eligible RCTs using predesigned data extraction forms.
The quality and risk of bias of each selected trial were assessed independently by two researchers (Y.Z and Q.Y) using the assessment tool recommended by the Cochrane Collaboration (16). Discrepancies between two researchers were discussed with a third researcher (C.X) and solved by reaching a consensus.

Statistical analyses
The primary outcome, i.e., pooled estimation of an effect size of 6-month TLPP and other outcomes, was expressed with odds ratio (OR) and 95% con dence interval (CI). Mantel-Haenszel methods with a random-effects model were used for pooling analyses accounted for the dichotomous outcomes and inter-study heterogeneity. Heterogeneity across included studies was assessed using Cochran's Q test and I 2 statistics. A p-value lower than 0.1 or I 2 statistic above 50% indicates heterogeneity among included studies. A visual judgment of funnel plots assessed publication bias (or small-study effect) for symmetry. We also performed a sensitivity analysis using the leave-one-out method. All statistical analyses were performed using R (version 4.0) with the 'meta' and 'dmetar' package (17,18).

Included studies
A total of 4 RCTs were identi ed and included in the meta-analyses, with 956 participants (11)(12)(13)(14). An overview of the study ow of the selection process is presented in Fig. 1. The study design of the identi ed trials and clinical characteristics of included patients are shown in Tables 1 and 2. All trials had a multicenter design and a relatively large sample size. Enrolled patients in the trial conducted by Yin et al. were younger than in the other three trials. Moreover, in their study, the study population had fewer complications, including a history of stroke, coronary artery disease and peripheral vascular disease.   Fig. 2. In general, all trials were well performed according to prede ned study protocols to minimize bias.

Primary clinical outcomes
As present in Fig. 3A, half of the included trials provided evidence supporting that DCB angioplasty had higher rates of 6-month TLPP compared to conventional balloon angioplasty. In the pooling analyses, there was no signi cant increase in rates of 6month TLPP in patients who received DCB as compared to those who were allocated to the conventional balloon angioplasty  Fig. 3B).

Other outcomes
We also assessed access circuit primary patency at 6 months, as present in Fig. 4A. Patients who received DCB were not found to have a higher rate of access circuit primary patency at 6 months than those who received conventional balloon angioplasty (OR 1.41, 95%CI 0.52-3.85, p = 0.36). Access circuit reintervention rate within 6 months was reported in three studies. In the pooled analysis, event rates of 6-month access circuit reintervention were not signi cantly lower in the DCB group (Fig. 4B).

Publication bias and Sensitivity analysis
Visual inspection of funnel plots revealed a likely asymmetric distribution among included trials regarding all outcomes of interest, indicating a potential publication bias due to a small study effect ( gure S1A-D). The contribution of each trial to the overall heterogeneity is also presented by the Baujat plot ( gure S2A-D). I 2 statistic revealed considerable heterogeneities for analyses of TLPP, TLR, access circuit primary patency and reintervention at 6 months (Fig. 3,4).
We further conducted a sensitivity analysis to evaluate whether an individual trial strongly drove the results by removing a single trial at one time. We found the trial conducted by Karunanithy et al. contributed mostly to the overall heterogeneity regarding 6month TLPP. After removing this trial, the pooled estimate of the effect size of 6-month TLPP increased to be 2.83 but remained statistically insigni cant ( gure S3A). In the case of 6-month access circuit primary patency and reintervention, the study of Lookstein et al. contributed mostly to the overall heterogeneity. After removing this trial, inter-study heterogeneity reduced substantially to zero.

Discussion
The main ndings of our study include: (1) 6-month TLPP rate in dysfunction hemodialysis AVF patients who received DCB was higher than those who were treated with conventional balloon angioplasty; however, this was not statistically signi cant; (2) there was no evidence supporting that DCB had a signi cantly lower rate of 6-month TLR; (3) there was no signi cant difference in terms of access circuit primary patency, access circuit reintervention, or severe adverse events between DCB and conventional balloon angioplasty; (4) substantial inter-heterogeneity among included trials were observed.
This meta-analysis failed to provide evidence supporting that DCB has a statistically signi cant higher rate of TLPP and lower rates of TLR, as compared to conventional balloon angioplasty within 6 months after the indexed procedure; however, a trend was observed. This meta-analysis is different from prior meta-analyses in the following perspectives (20)(21)(22): (1) we only included RCTs with generally high quality and relatively larger sample size, which have higher evidence level than observational studies; (2) RCTs included in our study were up to date; (3) we excluded patients who had arteriovenous graft and focused only on AVF.
Nevertheless, in the study of Kennedy et al. and Wee et al., retrospective studies and prospective cohorts were included (21,22).
In the study of Liao et al., RCTs with a small sample size and inconclusive results were included, which are likely to be susceptible to publication bias and may, therefore, be less reliable (20).
The limitation of this systematic review and meta-analyses needs to be addressed. First, the overall heterogeneity and risk-ofbias observed in our meta-analysis were attributed to each included original RCT. However, individual-level data are needed to explore the potential cause of heterogeneity fully, however, they were not available for us. Second, publication bias or smallstudy effect was inspected visually using a funnel plot. Due to the limited number of RCTs, a statistical test for the funnel plot asymmetry was not performed. Third, our pooled results in terms of TLPP should be interpreted with an awareness that in the trial conducted by Yin et al., the de nition of TLPP was different from the other three studies, which included a peak systolic velocity ratio lower or equal to 2.0 as determined using doppler ultrasound. Fourth, it is only feasible for us to assess the effectiveness of DCB in terms of the short-term outcomes, i.e. within 6 months after the indexed procedure, long-term outcomes regarding the e cacy and safety wait for longer follow-up results.
In conclusion, we failed to provide evidence supporting that DCB has a statistically signi cant higher rate of TLPP or lower rates of TLR in the treatment of dysfunctional AVF than conventional balloon angioplasty within 6 months after the indexed procedure. However, DCB was non-inferior to conventional balloon angioplasty in terms of safety. Therefore, further study is needed to clarify whether DCB angioplasty can bene t hemodialysis patients with dysfunction AVF.

Declarations
Funding None.
Con ict of interest: YZ, KC, and YC are employed by the Cardionovum Co, Ltd, Wuhan China. Other authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.
Author contributions: Conception and design: Yufang Chen, Qin Yang, Congying Xia; Administrative support: Yufang Chen; Provision of study materials: Yi Zhou; Collection and assembly of data: Congying Xia, Qin Yang and Yi Zhou; Data analysis and interpretation: Congying Xia, Qin Yang and Yi Zhou; Manuscript writing: All authors; Final approval of manuscript: All authors.

Figure 1
The ow of study selection process.

Figure 2
Risk of bias graph (upper) and summary (lower) for included trials.

Figure 3
Forrest plot of 6-month target-lesion primary patency (A) and 6-month target-lesion reintervention (B) in the drug-coated balloon (experimental) and conventional plain balloon (control).

Figure 4
Forrest plot of 6-month access circuit primary patency (A) and 6-month access circuit reintervention (B) in the drug-coated balloon (experimental) and conventional plain balloon (control).