Arteriovenous Fistulae for Haemodialysis: A Systematic Review and Metaanalysis of Efficacy and Safety Outcomes
Background: Arteriovenous fistulae are the currently recommended gold standard vascular access modality for haemodialysis because of their prolonged patency, improved durability, and low risk of infection for those that mature. However, notable disadvantages are observed in terms of protracted maturation time, associated high rates of catheter use, and substantial abandonment rates. The aim of this study was to quantitatively summarize the outcomes of fistula patency, infection, maturation, and abandonment published in the scientific literature/
Methods: This was a systematic review and meta-analyses of studies evaluating fistula outcomes. Literature searches were conducted in multiple databases to identify observational and interventional studies of mean fistula patency rates at 1 year, infection risk, maturation time, and abandonment. Digitisation software was used to simulate individual patient level data from Kaplan—Meier survival plots.
Results: Over 8000 studies were reviewed, and from these, 318 studies were included comprising 62,712 accesses. For fistulas the primary unassisted, primary assisted, and secondary patency rates at one year were 64%, 73% and 79% respectively, however not all fistulas reported as patent could be confirmed as being clinically useful for dialysis (i.e. functional patency). For fistulas that were reported as mature, mean time to maturation was 3.5 months, however only 26% of created fistulas were reported as mature at 6 months and 21% of fistulas were abandoned without use. Overall risk of infection in fistula patients was 4.1% and the overall rate per 100 access days was 0.018.
Conclusions: Reported fistula patency rates may overstate their potential clinical utility when time to maturation, maturation rate, abandonment and infection are considered. Protracted maturation times, abandonment and infection all have a significant impact on evaluating the clinical utility of fistula creation. A rigorous and consistent set of outcomes definitions for hemodialysis access are necessary to clarify factors contributing to fistula success and the clinical consequence of fistula failure.
Keywords: Arteriovenous fistula, Haemodialysis, Patency, Maturation, Meta-analysis
The use of chronic haemodialysis (HD) as renal replacement therapy (RRT) in patients with end stage renal disease (ESRD) is a prevalent practice worldwide and in the USA, and HD accounts for approximately 63% of RRT.1 The US Renal Data System estimated that there were 120,688 incident and 678,383 prevalent cases of ESRD in the USA in 2014.1 The three modalities of vascular access used for chronic HD are central venous catheters (CVC), arteriovenous grafts (AVG), and autologous arteriovenous fistulae (AVF). Although AVF had been recommended since 1997, in 2003, the National Kidney Foundation (NKF) set forth the Fistula First Breakthrough Initiative (FFBI), recommending fistula rates of 50% for incident (first placed access), and 40% for prevalent (patient had previous surgically created accesses) patients undergoing HD. The current goal for fistula use among prevalent patients is 68%.2 In the USA, 80% of patients initiate HD using a catheter, which has gone unchanged for nearly a decade, and only 16.9% initiate dialysis with an AVF. However, at 1 year post-HD initiation, 65% of patients dialyse exclusively using an AVF.
The NKF recommends that AVF be placed at least 6 months prior to initiation of HD treatment to allow sufficient time for access creation and evaluation, vein maturation, and if necessary, maturation enhancing interventions prior to cannulation. It is therefore recommended that chronic kidney disease (CKD) patients in the fourth or fifth stages be educated on vascular access modalities to allow sufficient time for access placement.3 While this is a noble goal, it has been difficult to implement because of the unpredictability of renal failure progression, patient referral patterns, and financial disincentives for early fistula creation. It is common for patients to progress to ESRD and initiate HD before the fistula has either been created or matured. In such cases a CVC is commonly placed to be used for vascular access, placing patients at high risk of complications and infection, resulting in increased patient costs and burden,4 and resulting in a mentality of both “fistula first” and “catheter last.
The use of AVFs as vascular accesses for haemodialysis has independent risks. Not all patients are candidates for AVF and many of the studies that highlighted the advantages of autologous access discounted and/or did not acknowledge the unacceptable rate of early failure and fistula non-maturation.6 The FFBI inadvertently exposed the disadvantages of autologous access and the related cascade of unintended consequences from early AVF failure or prolonged catheter exposure because of the need for several maturation enhancing procedures. The resulting undesired situation (high AVF placement and low initial AVF HD rates) is the result of high levels of primary fistula failure, either non-maturation caused by early thrombosis or insufficient dilation to support repetitive cannulation. Fistula non-maturation rates ranging from 20% to 60% have been reported.
The objectives of this study were: 1) to perform a systematic literature review of AVF among adults in developed nations including the outcomes of patency, maturation, infection, and abandonment; 2) to digitize available KaplaneMeier curves and simulate individual patient level data; 3) to meta-analytically combine estimates for the outcomes of interest; and 4) to evaluate the summary estimates by various potential confounding or modifying factors.
Study design : This systematic review and meta-analysis was conducted in accordance with PRISMA guidelines8 (the protocol was filed with PROSPERO register [registration number CRD42016040010] on 6/7/2016). The study population consisted of patients with CKD or ESRD either preparing for or on chronic HD treatment with an autologous AVF. Included studies were published in the English language with no lower date limit. Studies with outcomes of primary unassisted patency, primary assisted patency, secondary patency, infection rates, fistula maturation, access abandonment, and bridging catheter time were included. Randomised controlled trial and observational study designs were included. To generalise the analyses to outcomes of vascular accesses in developed nations, only studies conducted in countries classified as “developed” by the United Nations were included.9 After discussion with the clinical team (SMG and JHL), South Africa, Brazil, and Argentina were added to this list given that the level of dialysis health care that ESRD patients in those countries are likely to receive is comparable with that of the “developed” nations. Only standard of care methods were included in this review to produce the most generalisable results to other vascular access methods. The clinical team provided input as to the methods and interventions considered to be “standard of care.”
Case series with fewer than 20 patients, opinion pieces and editorials, guidelines and recommendations, articles without original data, and conference abstracts were excluded from the review. AVFs placed anywhere other than the arm were excluded as they are not considered standard of care. Publications from the same cohort that contained unique study populations or analyses were included in the review.
Literature searches were conducted in April 2016 in multiple scientific databases including Medline, Embase, Cochrane Library, and Clinicaltrials.gov. A flow diagram of the study selection and inclusion process is shown in Fig. 1. The included search terms used in each database are found in Table S1 (supplementary information). Studies were screened for eligibility by two reviewers at the levels of abstract and full text.
Meta-analyses were conducted using random effects models in the statistical software packages of either STATA (StataCorp LP, V.14.1) or Comprehensive Meta-Analysis (Biostat, v.3.0). Individual study weights were created using the inverse of the variance using the method proposed by DerSimonian and Laird.10 Stratified analyses were also performed on variables selected a priori to potentially affect the outcomes of interest, such as gender, age, race, diabetes status, obesity, cardiovascular disease status, access location, time on dialysis, access placed before or after dialysis began, incident or prevalent AVF patients, primary cause of renal failure, and whether vein transposition was involved in access construction.11 Evolving clinical practice patterns of enhanced fistula maturation meant that outcomes by studies conducted before and after 2005 were stratified, as surgical interventions used to prolong patency underwent a pivotal shift toward more aggressive endovascular interventions, which may have contributed to improvement of patency during this time period. Subgroup analyses were performed for outcomes reported in three or more studies.
For many outcomes, the included studies used time to event analysis and published survival curves using the KaplaneMeier method.12 As the gold standard for metaanalyzing time to event data is individual patient level data (IPD), a published algorithm13 was used and DigitizeIt software to reproduce IPD. The software identified each line in a KaplaneMeier curve and the resulting output was used, in coordination with the study’s reported life tables, to estimate survival summary statistics for each study including patency at 1 and 2 years, mean and median patency time, and standard errors, which were then entered into the meta-analyses. If outcome estimates were both reported in the text as well as published in survival curves, the originally reported in-text data were used, provided that a measure of variance was included. Comparisons were made between the digitisation output and the original reported outcomes on a sample of papers with survival curves and in-text data as a quality control measure for the digitisation software Some curves could not be digitized because of lack of numbers at risk or poor digital quality of the publication; others were not compatible with the algorithm and could not be included.
Endpoints and definitions
The patency analysis was categorized using the definitions provided by Sidawy.14 “Primary unassisted patency” is the interval from access placement until any intervention designed to maintain or re-establish patency, access thrombosis, or the time of measurement of patency. “Primary assisted patency” is the interval from access placement until access thrombosis or the time of measurement of patency, including intervening manipulations (surgical or endovascular interventions) designed to maintain the functionality of a patent access. “Secondary patency” is the interval from access placement until access abandonment, thrombosis, or the time of patency measurement including intervening manipulations (surgical or endovascular interventions) designed to re-establish functionality in thrombosed access. These definitions, with the inclusion of the “functionality” modifier, intend to take into consideration clinical usefulness of the access for hemodialysis (i.e. functional patency), but it was apparent that approximately half of reports focused simply on the patency aspect of the definition (i.e., any blood flow through the access) in their patency definition, and were therefore ambiguous as to the clinical utility of the accesses analyzed.