Impact of arteriovenous fistulas versus arteriovenous grafts on vascular access performance in haemodialysis patients: A systematic review and meta-analysis

Abstract

Background

Controversy exists regarding the best-performing vascular access type for patients undergoing haemodialysis. We aimed to compare outcomes of starting dialysis on arteriovenous fistulas (AVFs) versus arteriovenous grafts (AVGs) in haemodialysis patients.

Methods

We conducted a systematic search of multiple electronic information sources and bibliographic reference lists. The following outcome parameters were evaluated at 1, 2 and 5 years: primary failure, defined as access never used for dialysis; primary patency, defined as intervention-free access survival; primary-assisted patency, defined as uninterrupted access survival with interventions; and secondary patency, defined as cumulative access survival.

Results

We identified 15 comparative studies reporting a total of 118,434 patients who initiated haemodialysis with AVF (n = 95,143) or AVG (n = 23,291). Our analysis demonstrated that AVF was associated with significantly higher primary failure rate (OR: 2.05, p = .0005) but significantly higher rate of primary patency at 1 year (OR: 1.91, p < .00001), at 2 years (OR: 2.52, p < .00001) and at 5 years (OR: 2.59, p < .00001); and primary-assisted patency at 1 year (OR: 1.71, p < .00001), at 2 years (OR: 2.13, p < .00001) and 5 years (OR: 2.79, p < .00001). There was no significant difference in secondary patency at 1 year (OR: 1.08, p < .00001) but AVF had better secondary patency at 2 years (OR: 1.26, p < .00001) and 5 years (OR: 1.60, p < .00001) than AVG.

Conclusions

The meta-analysis of best available comparative evidence (Level 2) demonstrated that AVFs may be associated with significantly higher primary failure rate but higher primary patency, primary-assisted patency and secondary patency at 1, 2 and 5 years compared to AVGs. However, the available evidence is subject to significant selection bias and confounding by indication.

Keywords

Arteriovenous fistula arteriovenous graft vascular access haemodialysis

Introduction

Controversy persists regarding the best permanent vascular access type in patients undergoing haemodialysis more than 50 years after Brescia and Cimino et al.¹ described haemodialysis on a surgically created arteriovenous fistula (AVF). There is ongoing debate on the merits of arteriovenous grafts (AVGs) versus AVFs and continuing uncertainty over the performance of different types of fistulas in spite of vascular access outcomes designated as a core outcome measure for haemodialysis patients alongside fatigue, cardiovascular disease and mortality in the global Standardized Outcomes in Nephrology–Hemodialysis (SONG-HD) initiative.²

Access function is considered the most important vascular access outcome measure in haemodialysis patients,² but high-level evidence regarding the best functioning access type is lacking.³ Guidelines unanimously recommend autologous fistulas as the best form of access,^4–6 but the debate regarding the best access still continues.^7,8 Proponents of fistula firstly argue that prosthetic grafts should only be used if no autologous options are available,^9,10 whilst advocates of the catheter last, challenge the benefits of AVFs over AVGs.^11,12

Widespread variation therefore exists in vascular access use between countries and within facilities in the same country, with associated differences in patient survival.^13–16 This is similarly true regarding the type of autologous fistula use.¹⁷

Contributing to the weak evidence base is the observation that it is difficult to compare access function between different types of permanent vascular access. Access function, although critically important for patients and clinicians, is reported inconsistently in randomised trials, making it difficult to make informed decisions.¹⁸ A systematic review of vascular access outcomes, including 168 haemodialysis trials, showed that although access function was reported in the vast majority of the trials, it was the most heterogeneous outcome due to the variety of ways it was assessed. Access function was measured in 489 different ways and at 46 different time points.¹⁸

To our knowledge, no meta-analysis has evaluated comparative outcomes of AVFs and AVGs in haemodialysis patients. We aimed to conduct a comprehensive systematic review and meta-analysis of outcomes to compare initiating haemodialysis on AVFs or AVGs in terms of primary failure rate as well as long-term access patency, in order to add to the evidence base and contribute to the debate on reporting outcomes of vascular access in haemodialysis patients.

Methods

Design and study selection

The inclusion and exclusion criteria, methodology, and investigated outcome parameters of this review were highlighted in a review protocol. Our methodology respected the standards of Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.¹⁹

Inclusion criteria

• All comparative studies including:

• Patients aged > 18 years and of any gender

• Patients undergoing placement of any type of AVF for haemodialysis as intervention of interest

• Patients undergoing placement of any type of AVG for haemodialysis as comparison of interest

Exclusion criteria

• Single arm observational studies

• Case series

• Case reports

• Letters to editors

Primary failure, primary patency, primary-assisted patency and secondary patency at 1, 2 and 5 years were the evaluated outcome parameters. Primary failure was defined as an access that has never been used for dialysis. Primary patency was defined as the time between access placement and any intervention designed to maintain or re-establish patency, or to access thrombosis, or the time of measurement of patency and is similar to intervention-free access survival. Primary-assisted patency was defined as the interval from time of access placement to access thrombosis or time of measurement of patency, including intervening manipulations (surgical or endovascular interventions) designed to maintain the functionality of a patent access and is similar to uninterrupted access survival. Secondary patency was defined as the interval from time of access placement to access abandonment due to access failure and is similar to cumulative access survival.

Literature search strategy

A comprehensive search strategy was formulated based on thesaurus headings, search operators and limits in MEDLINE, EMBASE, Cumulative Index to Nursing and Allied Health Literature and CENTRAL. Two independent reviewers conducted the literature search via the aforementioned databases and searched the World Health Organization International Clinical Trials Registry http://apps.who.int/trialsearch/, ClinicalTrials.gov http://clinicaltrials.gov/ and the International Standard Randomised Controlled Trial Number Register http://www.isrctn.com/ to identify ongoing and unpublished studies. The same authors independently searched the reference lists of relevant articles and reviews to identify relevant trials. The literature search began on 2 April 2020 and lasted for 6 days. Appendix 1 presents the search strategy that was used for literature search.

Selection of studies

Two reviewers carefully assessed the title and abstract of articles found as the result of literature search. When necessary, the full-texts of relevant articles were retrieved and carefully evaluated against the eligibility criteria of this study. Studies that met our eligibility criteria were included. Discrepancies in this process were resolved by discussion between the authors. However, if the disagreement still existed, an independent author was consulted.

Data extraction and management

An electronic data extraction spreadsheet according to Cochrane’s recommendations for intervention reviews was created and was pilot-tested in randomly selected articles and adjusted accordingly. The following information was extracted from each of the included studies by two independent reviewers:

• Study-related data (first author, publication year, country of origin of the corresponding author, journal in which the study was published, study design, procedure performed, surgical approach and sample size in each group).

• Baseline demographic and clinical information of the study populations (age, gender, body mass index, history of diabetes mellitus, hypertension, hyperlipidaemia, cerebrovascular accident, ischaemic heart disease or myocardial infarction, and peripheral vascular disease).

• Primary and secondary outcome data.

Disagreements during this process were resolved following consultation with an additional author.

Assessment of risk of bias

The methodological quality and risk of bias assessment were carried out by two authors using the Newcastle–Ottawa scale (NOS)²⁰ as all of our included studies were observational studies. The NOS is a star-based scoring system (maximum score: 9) which enables review authors to evaluate an observational study in the following aspects: the selection of the study groups, the comparability of the groups and the ascertainment of outcome of interest. Studies with a score of nine stars were deemed to be at low risk of bias, studies with a score of seven or eight stars were deemed to be at medium risk of bias and those that scored six or less were judged to be at high risk of bias. We resolved discrepancies in risk of bias assessment by discussion between the assessing authors. Nevertheless, if no agreement could be reached, a third reviewer was involved as an adjudicator.

Summary measures and synthesis

For dichotomous outcome variables, we calculated the odds ratio (OR) as the summary measure The OR is the odds of an event in the AVF group compared to the AVG group. In the analysis of primary failure, an OR of less than one would favour the AVF group. In the analyses of primary patency, primary-assisted patency and secondary patency at 1, 2 and 5 years, an OR of more than 1 would favour the AVF group.

The unit of analysis regarding all evaluated outcomes was a vascular access site. Where possible, data regarding dropouts, withdrawals and other missing information were recorded.

One independent review author entered the extracted data into Review Manager 5.3 software for data synthesis.²¹ The entered data were subsequently checked by a second independent review author. Random-effects modelling was used for analysis. We reported the results of our analysis for each outcome parameter in a forest plot with 95% confidence intervals (CIs).

Heterogeneity among the studies was assessed using the Cochran Q test (χ²). We quantified inconsistency by calculating I² and interpreted it using the following guide: 0%–50% might not be important, 50%–75% may represent moderate heterogeneity and 75%–100% may represent substantial heterogeneity. Moreover, where more than 10 studies were available in analysis of an outcome parameter, funnel plots were constructed in order to assess their symmetry to visually evaluate publication bias.

We conducted sensitivity analyses to evaluate the effect of each study on the overall effect size and heterogeneity by repeating the analysis following excluding one study at a time (leave-one-out sensitivity analysis).

Results

Our literature search yielded 3316 articles. After further evaluation of the identified articles, 23 articles were shortlisted for potential inclusion. A further eight studies were excluded as four were single-arm studies, three were reviews and the remaining one did not provide enough data. Therefore, 15^22–36 comparative observational studies, two prospective and 13 retrospective were included (Figure 1). The included studies reported on a total number of 118,434 patients, of whom 95,143 patients initiated haemodialysis with AVFs and 23,291 initiated haemodialysis with AVGs.

Figure 1.

Study flow diagram.

Table 1 presents the date of publication and country of origin, journal and study design of the included studies. The patients in the AVF and AVG groups were of comparable age. The proportion of male patients was significantly higher in the AVF group when compared with the AVG group (62.1% vs 55.2%, p = .02). Moreover, there was no significant difference between patients in the AVF and AVG groups regarding diabetes mellitus (51.2% vs 51.9%, p = .75), hypertension (90.8% vs 90.1%, p = .57) and ischaemic heart disease (38.7% vs 40.8%, p = .39). The baseline characteristics of the study populations regarding body mass index, hyperlipidaemia, cerebrovascular accident and peripheral vascular disease were poorly reported by the included studies.

Table 1.

Summary characteristics of included studies.

Author	Year	Country	Journal	Study design	AVF	AVG
Kalman²²	1999	Canada	Journal of Vascular Surgery	Prospective observational study	235	231
Ascher²³	2000	USA	Journal of Vascular Surgery	Retrospective observational study	99	122
Oliver²⁴	2001	USA	Kidney International	Retrospective observational study	56	80
Gibson²⁵	2001	USA	Journal of Vascular Surgery	Retrospective observational study	492	574
Pflederer²⁶	2008	USA	Seminars in Dialysis	Retrospective observational study	321	285
Snyder²⁷	2008	USA	The American Journal of Surgery	Retrospective observational study	47	53
Maya²⁸	2009	USA	Clinical Journal of the American Society of Nephrology	Prospective observational study	322	289
Kim²⁹	2011	USA	Korean J Thorac Cardiovasc Surg	Retrospective observational study	92	37
Harms³⁰	2015	USA	Journal of Vascular Surgery	Retrospective observational study	289	310
Marques³¹	2015	France	Journal of Vascular Access	Retrospective observational study	136	102
Cui³²	2016	USA	Journal of Vascular Surgery	Retrospective observational study	138	44
Pham³³	2016	USA	Journal of Vascular Surgery	Retrospective observational study	29	32
Jadlowiec³⁴	2016	USA	Annals of Vascular Surgery	Retrospective observational study	149	37
Arhuidese³⁵	2018	USA	Journal of Vascular Surgery	Retrospective observational study	73,884	16,533
Arhuidese³⁶	2019	USA	Journal of Vascular Surgery	Retrospective observational study	19,173	4,840

Note: AVF: arteriovenous fistulas; AVG: arteriovenous grafts.

Methodological appraisal

The methodological appraisal of all included 15 observational studies is presented in Table 2. The risk of bias was judged as low in two studies and moderate in the remaining 13.

Table 2.

Methodological quality of the observational studies assessed with the NOC.

Author	Representativeness of the exposed cohort	Selection of the non-exposed cohort	Ascertainment of exposure	Demonstration that outcome of interest was not present at start of study	Comparability of cohorts on the basis of the design or analysis	Assessment of outcome	Was follow-up long enough for outcomes to occur?	Adequacy of follow up of cohorts	Total score
Kalman²²	*	\|*	*	*	*	*	*		7
Ascher²³	*	*	*	*	*	*	*		7
Oliver²⁴	*	*	*	*	*	*	*		7
Gibson²⁵	*	*	*	*	*	*	*	*	8
Pflederer²⁶	*	*	*	*	*	*	*	*	8
Snyder²⁷	*	*	*	*	*	*	*	*	8
Maya²⁸	*	*	*	*	*	*	*	*	8
Kim²⁹	*	*	*	*	*	*	*	*	8
Harms³⁰	*	*	*	*	*	*	*		7
Marques³¹	*	*	*	*	*	*	*		7
Cui³²	*	*	*	*	*	*	*		7
Pham³³	*	*	*	*	*	*	*		7
Jadlowiec³⁴	*	*	*	*	*	*	*		7
Arhuidese³⁵	*	*	*	*	**	*	*	*	9
Arhuidese³⁶	*	*	*	*	**	*	*	*	9

Note: NOC: Newcastle–Ottawa Scale.

Outcome synthesis

Outcomes are summarised in Figures 2 and 3.

Figure 2.

Forest plots of comparison of (a) primary failure, (b) primary patency at 1 year, (c) primary patency at 2 years, (d) primary patency at 5 years, (e) primary-assisted patency at 1 year, (f) primary-assisted patency at 2 years, (g) primary-assisted patency at 5 years, (h) secondary patency at 1 year, (i) secondary patency at 2 years and (j) secondary patency at 5 years. The solid squares denote the ORs. The horizontal lines represent the 95% CIs, and the diamond denotes the pooled effect size. Note: M-H: Mantel–Haenszel test; OR: odds ratios; CIs: confidence intervals.

Figure 3.

Funnel plots of comparison of (a) primary patency at 1 year and (b) primary patency at 2 years.

Primary failure

Eight studies (2969 accesses) were included in the analysis of primary failure. The primary failure rate in the AVF group was 32.3% compared to 20.3% in the AVG group. AVFs were associated with significantly higher failure rates than AVG (OR: 2.05, 95% CI: 1.37–3.07, p = .0005). The between-study heterogeneity was moderate (I²: 75%, p = .0005).

Primary patency at 1 year

Thirteen studies (117,968 accesses) were included in the analysis of primary patency at 1 year. The primary patency rate at 1 year in the AVF and AVG groups were 42.5% and 31.9%, respectively. AVFs were associated with significantly higher primary patency at 1 year than AVGs (OR: 1.91, 95% CI: 1.65–2.23, p < .00001). Significant heterogeneity existed among the included studies (I²: 80%, p < .00001).

Primary patency at 2 years

Twelve studies (118,237 accesses) were included in the analysis of primary patency at 2 years. The primary patency rate at 2 years in the AVF and AVG groups were 33.8% and 20.5%, respectively. AVFs were associated with a significantly higher rate of primary patency at 2 years than AVGs (OR: 2.52, 95% CI: 2.08–3.95, p < .00001). However, significant heterogeneity existed among the included studies (I²: 87%, p < .00001).

Primary patency at 5 years

Three studies (114,559 accesses) were included in the analysis of primary patency at 5 years. The primary patency rate at 5 years in the AVF and AVG groups were 19.6% and 8.9%, respectively. AVFs were associated with a significantly higher rate of primary patency at 5 years than AVGs (OR: 2.59, 95% CI: 2.20–3.04, p < .00001). Significant heterogeneity existed among the included studies (I²: 75%, p = .02).

Primary-assisted patency at 1 year

Four studies (114,729 accesses) were included in the analysis of primary-assisted patency at 1 year. The primary-assisted patency rate at 1 year in the AVF and AVG groups were 52.6% and 39.3%, respectively. AVFs were associated with a significantly higher rate of primary-assisted patency at 1 year than AVGs (OR: 1.71, 95% CI: 1.66–1.77, p < .00001). Low heterogeneity existed among the included studies (I²: 15%, p < .00001).

Primary-assisted patency at 2 years

Four studies (115,134 accesses) were included in the analysis of primary-assisted patency at 2 years. The primary-assisted patency rate at 2 years in the AVF and AVG groups were 46.2% and 28.9%, respectively. AVFs were associated with a significantly higher rate of primary-assisted patency at 2 years than AVGs (OR: 2.13, 95% CI: 2.06–2.20, p < .00001). The between-study heterogeneity was low (I²: 0%, p = .86).

Primary-assisted patency at 5 years

Two studies (114,430 accesses) were included in the analysis of primary-assisted patency at 5 years. The primary-assisted patency rate at 5 years in the AVF and AVG groups were 34.4% and 15.8%, respectively. AVFs were associated with a significantly higher rate of primary-assisted patency at 5 years than AVGs (OR: 2.79, 95% CI: 2.68–2.90, p < .00001). The between-study heterogeneity was moderate (I²: 32%, p = .23).

Secondary patency at 1 year

Thirteen studies (116,406 accesses) were included in the analysis of secondary patency at 1 year. The secondary patency rate at 1 year in the AVF and AVG groups were 58.9% and 57.2%, respectively. There was no significant difference in the secondary patency rate at 1 year between two groups (OR: 1.08, 95% CI: .98–1.19, p = .12). Moderate heterogeneity existed among the included studies (I²: 54%, p = .03).

Secondary patency at 2 years

Twelve studies (116,811 accesses) were included in the analysis of secondary patency at 2 years. Secondary patency rate at 2 years in the AVF and AVG groups were 53.7% and 48.3%, respectively. AVF was associated with a significantly higher rate of secondary patency at 2 years than AVG (OR: 1.26, 95% CI: 1.23–1.30, p < .00001). Moderate heterogeneity existed among the included studies (I²: 40%, p = .10).

Secondary patency at 5 years

Three studies (114,559 accesses) were included in the analysis of secondary patency at 5 years. The secondary patency rate at 5 years in the AVF and AVG groups were 43.2% and 32.8%, respectively. AVFs were associated with significantly higher rate of secondary patency at 5 years compared to AVGs (OR: 1.60, 95% CI 1.37–1.87, p < .00001). Significant heterogeneity existed among the included studies (I²: 89%, p = .0001).

Sensitivity analysis

Using random-effects or fixed-effect models did not affect the pooled effect size in any of the outcomes. The direction of pooled effect size remained unchanged when the OR, RR or RD was calculated. Moreover, one-leave-out sensitivity analyses did not change the direction of pooled effect size in any of the outcomes.

Discussion

In view of ongoing debate regarding the best choice of vascular access for haemodialysis, we conducted a comprehensive literature search and identified 15 comparative studies^22–36 reporting a total of 118,434 patients. Our meta-analysis of outcomes demonstrated that AVFs had significantly higher primary failure rates than AVGs. However, initiation of haemodialysis with an AVF was associated with better long-term function as indicated in the significantly higher primary patency and primary-assisted patency at 1, 2 and 5 years, and higher rate of secondary patency at 2 and 5 years compared to AVGs. The between-study heterogeneity in the analyses of primary-assisted patency at 1, 2 and 5 years, and secondary patency at 2 years were low, indicating that these findings are robust. Moderate between-study heterogeneity in the analysis of secondary patency at 1 year indicates variation of reporting in the included studies on this outcome. There was high heterogeneity among the included studies in the analyses of primary failure and primary patency at 1, 2 and 5 years, and secondary patency at 5 years, indicating that our findings about these outcomes may be less robust. However, our findings indicate that AVFs that worked probably delivered more dialysis days and more intervention-free dialysis days than AVGs. Since functional dialysis days were not reported, it is difficult to come to a firm conclusion on which strategy delivers most functional dialysis days and intervention-free dialysis days for our haemodialysis patients.

The definition of primary failure was homogenous among all the included studies as an access that was never used for dialysis although in one study²⁴ some accesses that failed within a month were also included. However, it was poorly reported whether the reason for patients who never used their access was because they never progressed to dialysis or not. The included studies homogenously defined primary patency, primary-assisted patency and secondary patency.

We believe this is the first meta-analysis of comparative studies which evaluated the outcomes of AVFs and AVGs in haemodialysis patients. In 2014, Al-Jaishi et al.³⁷ conducted a systematic review and pooled analysis of all studies evaluating the patency of AVFs in haemodialysis patients. They included 46 studies enrolling a total number of 12,383 AVFs and reported that the pooled primary patency rates at 1 year and 2 years were 60% and 51%, respectively, and a pooled secondary patency rate of AVFs at 1 year and 2 years of 71% and 64%, respectively. Our estimated pooled primary patency rates of AVFs at 1 and 2 years of 42.5% and 33.8%, are lower than those reported by Al-Jaishi et al.³⁷ Similarly, our pooled secondary patency rates of AVFs at 1 year (58.9%) and 2 years (53.7%) were lower. The primary failure rate associated with AVFs in this meta-analysis was 32.3%, which is higher than the primary failure reported by Al-Jaishi et al.³⁷ of 23%. It should be considered that our pooled population size was approximately 10 times larger than the Al-Jaishi study and our findings are less subject to type 2 error. Most importantly, Al-Jaishi et al.³⁷ mainly included single-arm studies in their review. Almasri et al.³⁸ also conducted a meta-analysis of non-comparative studies with similar limitations associated with study of Al-Jaishi et al.³⁷ We only included comparative studies which have more robust methodological quality and provide a higher level of evidence. In 2016, Almasri et al.38 reviewed 200 vascular access related studies and conducted a meta-analysis of outcomes of more than 17 access types to determine their pooled primary and secondary patency. However, despite including some comparative studies, no meta-analysis of comparative outcomes were conducted. The authors only determined the pooled patency rates on each access type from mixed single-arm and few comparative studies. A true comparative meta-analysis should provide an estimate of pooled effect size such as OR, RR or hazard ratio (HR) with associated p value. We believe our meta-analysis is the first to provide such comparative measures.

Our findings of a higher primary failure rate of AVFs but higher long-term patency rates, up to 5 years, is well recognised in the vascular access world. However, the lack of hard evidence to come to firm conclusions on which strategy is best supports the SONG-HD conclusion pertaining to the problems with comparisons between access function of different types of permanent vascular access. Our meta-analysis primarily showed that it was not possible to directly compare access function, with one simple criterion that includes both primary failure and long-term access function. Functional dialysis days per access creation and intervention-free functional dialysis days per access creation include both primary failure and long-term function but often cannot be distilled from comparative studies as time to failure is not reported.

Furthermore, our meta-analysis demonstrated that the best available evidence (Level 2) failed to sufficiently report the comparative outcomes of AVFs and AVGs with respect to baseline characteristic of haemodialysis patients. The included studies rarely reported their outcomes with respect to baseline characteristic of the included patients such as age, gender, diabetes mellitus, etc. Although they demonstrated comparability of some characteristics, the heterogeneous reporting of patient’s clinical characteristics did not allow us to report the outcomes with respect to them. Most importantly, the central issue of observational studies comparing different types of vascular access for haemodialysis is that the patients might have been selected to have placement of grafts rather than fistulas because of clinical characteristics, such as poor vessel quality or urgent need to start haemodialysis. Furthermore, when focusing on patients who initiate haemodialysis with fistulas rather than grafts, patients with non-maturing fistulas who start with a catheter or graft are not analysed according to the initial vascular access strategy. This introduces selection bias or ‘confounding by indication’.

Interestingly, the included studies in our meta-analysis reported mainly surgeon-centred outcomes, such as primary patency, primary-assisted patency or secondary patency, rather than patient-centred outcomes such as primary failure, intervention-free functional dialysis days, total functional dialysis days, satisfaction with access, ease of cannulation or quality of life. In fact, of the patient-centred outcomes, only primary failure, which arguably is both patient- and surgeon-centred, was reported by the included studies. Even though the primary failure has been reported by some of the included studies, the pooled population size in the analysis of primary failure (around 3000 vascular accesses) was much lower than the pooled population size of the other outcomes (at least around 114,000 vascular accesses). This is remarkable as vascular access is created to enable dialysis. Vascular access function may be defined differently by surgeons or patients. Patients may define an access function as the ability to receive uninterrupted and adequate dialysis without the need for interventions rather than using parameters that surgeons are interested in such as flow or patency. Moreover, the vascular access outcomes have been heterogeneously reported with over 1400 different outcome measures used to report 23 different vascular access outcomes. The need for successful implementation of a standardised core outcome measure for vascular access in haemodialysis research that is important to patients, caregivers and health professionals is obvious. The SONG-HD initiatives have recommended core outcome measures for vascular access that are meaningful and feasible to be reported consistently across studies.² Compliance to such recommendations can contribute to more meaningful comparison of function of different access types. Considering our findings and the recommendations from SONG-HD, we strongly recommend future high-quality research to consider standardised patient- and surgeon-centred outcomes of access function with respect to baseline characteristics of their study populations to provide stronger evidence in favour of either access type.

The limitations of this meta-analysis should be taken into account when interpreting its findings. None of the included studies were randomised controlled trials. All of the included studies were observational studies, and selection bias was present. Most of the included studies had moderate to high risk of bias. Moreover, the outcomes reported by the included studies and, subsequently, this meta-analysis were not evaluated with respect to baseline demographics and clinical characteristics of the included studies which hampers making solid conclusions on the superiority of each approach. Furthermore, we were not able to analyse the outcomes with respect to different types of AVFs or AVGs due to significant heterogeneity in the type of vascular accesses used. Also, not all the included studies reported all outcome measures, resulting in different number of included studies in analysis of each outcome. It should be highlighted that two of our included studies,^35,36 accounting for the vast majority of included patients, were conducted by the same group. Although it is not clear whether overlapping patients were included in both studies, there is a possibility that some of the patients were included twice in outcome synthesis potentially skewing the results. However, we conducted one-leave-out sensitivity analysis with exclusion of one study at a time, and the direction of the pooled effect size remained unchanged. Importantly, our evaluated outcomes were better to be investigated as time-to-event outcomes. However, the available data in the included studies did not allow conducting HRs meta-analytical models on the natural logarithm scale. Finally, the patient-centred outcomes were poorly reported by the included studies.

Conclusions

The meta-analysis of best available comparative evidence (Level 2) suggests that initiating haemodialysis on an AVF is associated with significantly higher primary failure rate but higher primary patency, primary-assisted patency and secondary patency at 1, 2 and 5 years compared to AVGs. Failure to directly compare access function, with one simple criterion that includes both primary failure and long-term access function and limitations associated with the available evidence with respect to baseline characteristics, reporting of patient-centred outcomes, and variety of vascular access types used, do not allow making solid conclusions to resolve the existing controversies. We encourage future high-quality randomised studies with standardised patient- and surgeon-centred outcomes of access function.

Footnotes

Author contributions

Conception and design: TW and Shahin H

Literature search and study selection: HB and PG

Data collection: Shahin H, Shahab H, HB and PG

Analysis and interpretation: Shahin H and Shahab H

Writing the article: Shahin H, Shahab H and TW

Critical revision of the article: All authors

Final approval of the article: All authors

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Shahin Hajibandeh

Appendix 1

Search No	Search strategy ^a*
#1	MeSH descriptor: [arteriovenous graft] explode all trees
#2	arteriovenous graft: TI, AB, KW
#3	MeSH descriptor: [arteriovenous fistula] explode all trees
#4	arteriovenous fistula: TI, AB, KW
#5	MeSH descriptor: [radio-cephalic fistula] explode all trees
#6	radio-cephalic fistula: TI, AB, KW
#7	MeSH descriptor: [brachiocephalic fistula] explode all trees
#8	brachiocephalic fistula: TI, AB, KW
#9	MeSH descriptor: [vascular access] explode all trees
#10	Vascular access: TI,AB,KW
#11	#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10
#12	MeSH descriptor: [haemodialysis] explode all trees
#13	haemodialysis: TI, AB, KW
#14	hemodialysis: TI, AB, KW
#15	#11 OR #14
#16	#11 AND #15

^aThis search strategy was adopted for following databases: MEDLINE, EMBASE, CINAHL and the Cochrane Central Register of Controlled Trials (CENTRAL).

References

Brescia

Cimino

Appel

, et al. Chronic hemodialysis using venipuncture and a surgically created arteriovenous fistula. N Engl J Med 1966; 275(20): 1089–1092.

Viecelli

Tong

O'Lone

, et al. Report of the standardized outcomes in nephrology-hemodialysis (SONG-HD) consensus workshop on establishing a core outcome measure for hemodialysis vascular access. Am J kidney Dis 2018; 71(5): 690–700.

Wilmink

. Vascular access: clinical practice guidelines of the european society for vascular surgery. Eur J Vasc Endovasc Surg 2018; 55(6): 753–754.

Schmidli

Widmer

Basile

, et al. Editor’s choice - vascular access: 2018 clinical practice guidelines of the european society for vascular surgery (ESVS). Eur J Vasc Endovasc Surg 2018; 55(6): 757–818.

Tordoir

Canaud

Haage

, et al. EBPG on vascular access. Nephrol Dial Transplant 2007; 22(Suppl 2): ii88–117.

NKF/KDOQI . 2006 Update vascular access. Guideline 2: selection and placement of hemodialysis access. Am J Kidney Dis 2006; 48(Suppl 1): S183–S247.

Drew

Lok

Cohen

, et al. Vascular access choice in incident hemodialysis patients: a decision analysis. J Am Soc Nephrol 2015; 26(1): 183–191.

Wish

. Catheter last, fistula not-so-first. J Am Soc Nephrol 2015; 26(1): 5–7.

Besarab

. Resolved: fistulas are preferred to grafts as initial vascular access for dialysis. J Am Soc Nephrol 2008; 19(9): 1629–1633.

10.

Santoro

Benedetto

Mondello

, et al. Vascular access for hemodialysis: current perspectives. Int J Nephrol Renovascular Dis 2014; 7: 281–294.

11.

Moist

Lok

Vachharajani

, et al. Optimal hemodialysis vascular access in the elderly patient. Semin Dial 2012; 25(6): 640–648.

12.

Gomes

Schmidt

Wish

. Re-envisioning fistula first in a patient-centered culture. Clin J Am Soc Nephrol 2013; 8(10): 1791–1797.

13.

Pisoni

Arrington

Albert

, et al. Facility hemodialysis vascular access use and mortality in countries participating in DOPPS: an instrumental variable analysis. Am J Kidney Dis 2009; 53(3): 475–491.

14.

Polkinghorne

McDonald

Atkins

, et al. Vascular access and all-cause mortality: a propensity score analysis. J Am Soc Nephrol 2004; 15(2): 477–486.

15.

Pisoni

Zepel

Port

, et al. Trends in US vascular access use, patient preferences, and related practices: an update from the US DOPPS practice monitor with international comparisons. Am J Kidney Dis 2015; 65(6): 905–915.

16.

Fluck

Pitcher

Steenkamp

. Vascular access audit report 2012. Care UrraNK. NHS Information Centre, 2012.

17.

Pisoni

Zepel

Fluck

, et al. International differences in the location and use of arteriovenous accesses created for hemodialysis: results from the dialysis outcomes and practice patterns study (DOPPS). Am J Kidney Dis 2018; 71(4): 469–478.

18.

Viecelli

O’Lone

Sautenet

, et al. Vascular access outcomes reported in maintenance hemodialysis trials: a systematic review. Am J Kidney Dis 2018; 71(3): 382–391.

19.

Liberati

Altman

Tetzlaff

, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ 2009; 339: b2700.

20.

Wells

Shea

O’Connell

, et al. The newcastle-ottawa scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp (accessed 2 April 2020).

21.

Higgins

Altman

. Assessing risk of bias in included studies. In: JP

Higgins

Green

(eds) Cochrane handbook for systematic reviews of interventions: John Wiley & Sons, http://hiv.cochrane.org/sites/hiv.cochrane.org/files/uploads/Ch08_Bias.pdf (accessed 21 April 2020).

22.

Kalman

Pope

Bhola

, et al. A practical approach to vascular access for hemodialysis and predictors of success. J Vasc Surg 1999; 30(4): 727–733. DOI: 10.1016/s0741-5214(99)70112-6.

23.

Ascher

Gade

Hingorani

, et al. Changes in the practice of angioaccess surgery: impact of dialysis outcome and quality initiative recommendations. J Vasc Surg 2000; 31(1 Pt 1): 84–92. DOI: 10.1016/s0741-5214(00)70070-x.

24.

Oliver

McCann

Indridason

, et al. Comparison of transposed brachiobasilic fistulas to upper arm grafts and brachiocephalic fistulas. Kidney Int 2001; 60(4): 1532–1539. DOI: 10.1046/j.1523-1755.2001.00956.x.

25.

Gibson

Gillen

Caps

, et al. Vascular access survival and incidence of revisions: a comparison of prosthetic grafts, simple autogenous fistulas, and venous transposition fistulas from the United States resnal data system dialysis morbidity and mortality study. J Vasc Surg 2001; 34(4): 694–700. DOI: 10.1067/mva.2001.117890.

26.

Pflederer

Kwok

Ketel

, et al. A comparison of transposed brachiobasilic fistulae with nontransposed fistulae and grafts in the Fistula First era. Semin Dial 2008; 21(4): 357–363. DOI: 10.1111/j.1525-139X.2008.00451.x.

27.

Snyder

Clericuzio

Stringer

, et al. Comparison of outcomes of arteriovenous grafts and fistulas at a single Veterans’ Affairs medical center. Am J Surg 2008; 196(5): 641–646. DOI: 10.1016/j.amjsurg.2008.07.013.

28.

Maya

O’Neal

Young

, et al. Outcomes of brachiocephalic fistulas, transposed brachiobasilic fistulas, and upper arm grafts. Clin J Am Soc Nephrol 2009; 4(1): 86–92. DOI: 10.2215/CJN.02910608.

29.

Kim

, et al. Clinical analysis of hemodialysis vascular access: comparision of autogenous arterioveonus fistula & arteriovenous prosthetic graft. Korean J Thorac Cardiovasc Surg 2011; 44(1): 25–31. DOI: 10.5090/kjtcs.2011.44.1.25.

30.

Harms

Rangarajan

Young

, et al. Outcomes of arteriovenous fistulas and grafts with or without intervention before successful use. J Vasc Surg 2016; 64(1): 155–162. DOI: 10.1016/j.jvs.2016.02.033.

31.

Marques

Sadaghianloo

Fouilhé

, et al. Higher patency of transposed brachio-basilic arteriovenous fistulas compared to brachio-axillary grafts for hemodialysis patients. J Vasc Access 2015; 16(6): 486–492. DOI: 10.5301/jva.5000433.

32.

Cui

Steele

Wenger

, et al. Hemodialysis arteriovenous fistula as first option not necessary in elderly patients. J Vasc Surg 2016; 63(5): 1326–1332. DOI: 10.1016/j.jvs.2015.11.036.

33.

Pham

XBD

Kim

Ihenachor

, et al. A comparison of brachial artery-brachial vein arteriovenous fistulas with arteriovenous grafts in patients with poor superficial venous anatomy. J Vasc Surg 2017; 65(2): 444–451. DOI: 10.1016/j.jvs.2016.09.037.

34.

Jadlowiec

Mannion

Lavallee

, et al. Hemodialysis access in the elderly: outcomes among patients older than seventy. Ann Vasc Surg 2016; 31: 77–84. DOI: 10.1016/j.avsg.2015.08.013.

35.

Arhuidese

Orandi

Nejim

, et al. Utilization, patency, and complications associated with vascular access for hemodialysis in the United States. J Vasc Surg 2018; 68(4): 1166–1174. DOI: 10.1016/j.jvs.2018.01.049.

36.

Arhuidese

Cooper

Rizwan

, et al. Vascular access for hemodialysis in the elderly. J Vasc Surg 2019; 69(2): 517–525el. DOI: 10.1016/j.jvs.2018.05.219.

37.

Al-Jaishi

Oliver

Thomas

, et al. Patency rates of the arteriovenous fistula for hemodialysis: a systematic review and meta-analysis. Am J Kidney Dis 2014; 63(3): 464–478. DOI: 10.1053/j.ajkd.2013.08.023.

38.

Almasri

Alsawas

Mainou

, et al. Outcomes of vascular access for hemodialysis: a systematic review and meta-analysis. J Vasc Surg 2016; 64(1): 236–243. DOI: 10.1016/j.jvs.2016.01.053.