Complications of the Arteriovenous Fistula: A Systematic Review

Abstract

The implementation of patient-centered care requires an individualized approach to hemodialysis vascular access, on the basis of each patient’s unique balance of risks and benefits. This systematic review aimed to summarize current literature on fistula risks, including rates of complications, to assist with patient-centered decision making. We searched Medline from 2000 to 2014 for English-language studies with prospectively captured data on ≥100 fistulas. We assessed study quality and extracted data on study design, patient characteristics, and outcomes. After screening 2292 citations, 43 articles met our inclusion criteria (61 unique cohorts; n>11,374 fistulas). Median complication rates per 1000 patient days were as follows: 0.04 aneurysms (14 unique cohorts; n=1827 fistulas), 0.11 infections (16 cohorts; n>6439 fistulas), 0.05 steal events (15 cohorts; n>2543 fistulas), 0.24 thrombotic events (26 cohorts; n=4232 fistulas), and 0.03 venous hypertensive events (1 cohort; n=350 fistulas). Risk of bias was high in many studies and event rates were variable, thus we could not present pooled results. Studies generally did not report variables associated with fistula complications, patient comorbidities, vessel characteristics, surgeon experience, or nursing cannulation skill. Overall, we found marked variability in complication rates, partly due to poor quality studies, significant heterogeneity of study populations, and inconsistent definitions. There is an urgent need to standardize reporting of methods and definitions of vascular access complications in future clinical studies to better inform patient and provider decision making.

Current literature suggests the arteriovenous fistula (fistula) to be the preferred type of vascular access for hemodialysis.^1–3 Once established, fistulas have longer patency and lower rates of complications compared with arteriovenous grafts and catheters.^4,5 However, with the increasing proportion of elderly and frail patients on hemodialysis, the rate of failure to mature has increased with a decrease in patency rates.⁶ The current focus on patient-centered care requires individualized approaches to therapy, including the choice and use of vascular access, on the basis of each patient’s unique balance of risks and benefits. As such, complications related to vascular access also deserve careful consideration when individualizing the choice of vascular access type, facilitating informed consent, and providing an appropriate implementation plan (creation, monitoring, and use).

Fistula complications are associated with morbidity, mortality, and a high economic burden.^7–10 Although there have been systematic reviews and meta-analyses on fistula patency and failure to mature, to our knowledge, no published systematic review has evaluated fistula complication rates related to aneurysm, infection, ischemic steal syndrome, thrombosis, and venous hypertension. We conducted this systematic review to concisely summarize contemporary published information between 2000 and 2014 on the rates for the above-mentioned complications, as well as to identify knowledge gaps in the existing literature.

Results

The literature search yielded 2292 citations. All citations were screened by title and abstract, resulting in 172 articles reviewed in full-text. Thirty-five articles met eligibility criteria. Details of the study selection are shown in Supplemental Figure 1. Twelve additional studies were identified through manual search of bibliographies of selected articles and eight were included in our analyses. Thus, 43 studies (61 unique cohorts; n>11,374 fistulas—not all studies reported number of fistulas included) were included from 19 countries. Most published studies were from the United States (13 studies), Canada (five studies), and Italy (four studies). One study reported outcomes of interest across European countries and another reported on infections internationally.^11,12 All studies were published between 2001 and 2014 with patient recruitment beginning between 1986 and 2012. The characteristics of each study are described in Table 1.

Table 1.

Study characteristics

Author (Yr Published)	Country	Recruitment Start	Type of Study	Cohort Subset	Follow-Up (Mo)	No. of Fistulas	Upper Arm, %	First Fistula, %	Age, yr	Men, %	DM, %	CAD, %	CHD, %
McCarley et al. (2001)58a,b	United States	1996	Retrospective	NM	11.0	39	.	.	55.3	51	36	.	.
McCarley et al. (2001)58a,b	United States	1996	Retrospective	DVPM	12.0	41	.	.	56.6	54	36	.	.
McCarley et al. (2001)58a,b	United States	1997	Retrospective	VABFM	10.0	43	.	.	56.1	59	34	.	.
Tokars et al. (2001)59	United States	1997	Prospective	Standard	5.2	.	.	.	.	.	.	.	.
Dixon et al. (2002)60	United States	1992	Retrospective	Lower arm	.	88	0	.	50.9	81	47	39	.
Dixon et al. (2002)60	United States	1992	Retrospective	Upper arm	.	117	100	.	57.8	56	56	49	.
Ravani et al. (2002)61	Italy	1995	Prospective	Standard	20.4	197	19	100	65.7	59	22	.	14
Saxena et al. (2002)62	Saudi Arabia	1996	Prospective	Diabetic nasal carriers of MRSA/MSSA	36.0	.	.	.	.	.	.	.	.
Saxena et al. (2002)62	Saudi Arabia	1996	Prospective	Nondiabetic nasal carriers of MRSA/MSSA	36.0	.	.	.	.	.	.	.	.
Stevenson et al. (2002)63	United States	1998	Retrospective	Standard	.	.	.	.	.	.	.	.	.
Taylor et al. (2002)64	Canada	1998	Surveillance	Standard	6.0	.	.	.	.	.	.	.	.
Tokars et al. (2002)65	United States	1999	Prospective	Standard	.	.	.	.	.	.	.	.	.
Elseviers et al. (2003)11b	Europe	1998	Prospective	Standard	12.0	1049	.	.	.	56	.	.	.
Lok et al. (2003)66a	Canada	1997	Prospective	Duplex US Monitoring	9.2	189	30	96	57.5	71	.	.	.
Lok et al. (2003)66a	Canada	1999	Prospective	Transonic surveillance	10.4	241	30	96	57.5	71	.	.	.
Saxena et al. (2003)67b	Saudi Arabia	1997	Surveillance	Standard	.	.	.	.	47.5	54	.	.	.
Bonforte et al. (2004)68	Italy	1991	Prospective	Lower arm	27.0	112	0	.	71	50	22	.	.
Astor et al. (2005)69	United States	1995	Prospective	Standard	35.9	185	.	.	56.8	71	47	42	.
Gilad et al. (2005)70b	Israel	2002	Surveillance	Standard	.	.	.	.	63.5	55	48	44	21
Lok et al. (2005)71	Canada	1995	Retrospective	Nonelderly (<65 yr)	65.0	248	44	100	46	65	29	24	15
Lok et al. (2005)71	Canada	1995	Retrospective	Elderly (≥65 yr)	51.7	196	53	100	74	69	30	44	29
Mallamaci et al. (2005)72	Italy	.	Prospective	Standard	32.5	205	5	.	59.4	57	14	.	.
Shahin et al. (2005)73	United States	1992	Retrospective	Standard	21.0	146	51	100	54.9	58	49	53	.
Shahin et al. (2005)73	United States	1999	Retrospective	Qa surveillance	19.0	76	61	100	57.6	59	57	68	.
Jennings et al. (2006)74	United States	2003	Retrospective	Standard	11.0	134	91	74	61	54	52	.	.
Polkinghorne et al. (2006)75	Australia	2001	RCT	Standard	16.2	68	34	.	56.4	71	28	29	.
Polkinghorne et al. (2006)75	Australia	2001	RCT	Qa surveillance	17.3	69	36	.	60	65	35	28	.
Roozbeh et al. (2006)76a	Iran	.	Prospective	Standard	23.0	171	57	.	53	68	27	.	.
Huijbregts et al. (2008)5	The Netherlands	2004	Prospective	Standard	8.6	285	40	.	64.6	62	33	23	.
Pflederer et al. (2008)77	United States	2004	Retrospective	Standard	8.0	321	37	.	64.5	65	43	.	.
Pflederer et al. (2008)77	United States	2004	Retrospective	AVF-T	8.0	161	100	.	63.3	61	45	.	.
Qasaimeh et al. (2008)78b	Jordan	2004	Retrospective	Standard	12.0	104	.	.	.	52	32	.	.
Tessitore et al. (2008)79c	Italy	2002	Prospective	Standard	31.0	97	18	.	65.1	64	19	62	.
Tessitore et al. (2008)79c	Italy	2002	Prospective	Qa Surveillance	30.0	62	21	.	63.4	52	31	69	.
Koksoy et al. (2009)80	Turkey	2003	RCT	AVF-T	43.6	50	100	0	54.66	52	32	.	.
Koksoy et al. (2009)80	Turkey	2003	RCT	Upper arm	39.5	50	100	0	54.78	60	24	.	.
Papanikolaou et al. (2009)81	Greece	1986	Retrospective	Lower arm	.	.	100	.	.	45	8	.	.
Papanikolaou et al. (2009)81	Greece	1986	Retrospective	Upper arm	.	.	.	.	.	45	8	.	.
Korkut et al. (2010)15	Turkey	2004	Retrospective	AVF-T	48.0	350	100	.	57.8	44	51	15	.
Paul et al. (2010)82	United States	2006	Retrospective	Endo/AVF-T	14.0	98	100	47	60	40	49	.	.
Paul et al. (2010)82	United States	2006	Retrospective	Open/AVF-T	18.0	78	100	41	62	27	56	.	.
van Loon et al. (2010)13	The Netherlands	2007	Prospective	Rope ladder	.	70	39	84	65	67	21	80	.
van Loon et al. (2010)13	The Netherlands	2007	Prospective	Buttonhole	.	75	51	68	67	59	27	61	.
Jennings et al. (2011)83a	United States	2003	Retrospective	Elderly (≥65 yr)	17.0	461	38	77	73	49	60	.	.
Labriola et al. (2011)84	Belgium	2001	Retrospective	Standard	35.1	.	37	.	70.4	66	33	.	.
Ng et al. (2011)85	United States	1996	Prospective	Standard	6.0	476	.	.	61.1	73	49	47	34
Ayez et al. (2012)86	The Netherlands	2000	Retrospective	Upper arm	19.0	173	100	68	62.61	52	46	3	57
Renaud et al. (2012)87	Singapore	2008	Retrospective	Nonelderly (<65 yr)	12.0	191	41	100	52	60	70	41	.
Renaud et al. (2012)87	Singapore	2008	Retrospective	Elderly (≥65 yr)	13.0	89	39	100	72	55	71	49	.
Kandil et al. (2013)88	United Kingdom	2007	Retrospective	Buttonhole	.	227	.	.	63.8	61	46	.	.
Lin et al. (2013)14	Taiwan	2008	RCT	Infrared	12.0	60	15	100	63.2	53	47	17	.
Lin et al. (2013)14	Taiwan	2008	RCT	Standard	12.0	62	18	100	63	57	37	12	.
Macrae et al. (2013)19	Canada	2006	RCT	Buttonhole	.	70	84	83	77.2	73	49	61	.
Macrae et al. (2013)19	Canada	2006	RCT	Standard	.	69	75	86	66.1	67	54	42	.
Ravani et al. (2013)12	International	1996	RCT	Standard	.	3352	.	100	62.2	68	38	36	29
de Albuquerque et al. (2013)89	Brazil	2010	Surveillance	Standard	.	.	.	.	.	.	.	.	.
Agarwal et al. (2014)90	United States	2005	Retrospective	AVF-T (1-stage)	.	61	100	.	59.1	.	.	.	.
Agarwal et al. (2014)90	United States	2005	Retrospective	AVF-T (2-stage)	.	83	100	.	61.5	.	.	.	.
Badawy et al. (2014)91b	Kuwait	2012	Surveillance	Standard	.	.	.	.	58.1	40	.	.	.
Lok et al. (2014)92	Canada	2000	Retrospective	Rope ladder	29.6	457	.	.	58.8	58	36	29	16
Olsha et al. (2014)93	Turkey	2005	Retrospective	Elderly (≥80 yr)	17.6	128	71	.	85	66	40	45	24

DM, diabetes mellitus; CAD, coronary artery disease; CHD, Congestive Heart Disease; NM, no monitoring; ., data not reported; DVPM, dynamic venous pressure monitoring; VABFM, vascular access blood flow monitoring; standard, all fistula locations; lower arm, a cohort consisting of patients only fistulas located in the lower arm; upper arm, patients with fistulas located in the upper arm, including transposed brachiobasilic fistulas; MRSA, methicillin‐resistant Staphylococcus aureus; MSSA, methicillin‐sensitive Staphylococcus aureus; duplex US, duplex ultrasound; RCT, randomized controlled trial; AVF-T, transposed arteriovenous fifistula; endo/AVF-T, a transposed arteriovenous fistula created using an endoscopic procedure; open/AVF-T, a transposed arteriovenous fistula created using a long open incision technique; rope ladder, rope ladder cannulation; buttonhole, buttonhole cannulation.

^aPercentage of patients whose diabetes was the cause of kidney failure.

^bBaseline data reported for fistulas and other accesses grouped together.

^cCAD was defined as presence of cardiovascular disease.

Patient Population

Patient demographic data, comorbid conditions, and site of fistula creation were not always reported. Three out of the 39 observational studies were prospective surveillance studies and generally did not report baseline characteristics specific to patients using a fistula. However, when the data were reported, the median age was 61 years (ranging from 46 to 85 years; interquartile range [IQR] =7.6 years). The median proportion of men and patients with diabetes was 59% (range 27%–81%; IQR=13%) and 37% (range 8%–71%; IQR=20%), respectively. The median proportion of patients who had an upper arm fistula was 44% (range 0%–100%; IQR=70%). A summary of characteristics is presented in Table 2.

Table 2.

Summary of study characteristics (61 cohorts in total)

Variable	N (%)	Median	Min	Q1	Q3	Max
Proportion of patients with first fistula, %	23 (38)	96	0	68	100	100
Mean age reported in study, yr	50 (82)	61.05	46	56.8	64.6	85
Time from fistula creation to use, mo	15 (25)	3.49	1.11	1.73	4.87	5.92
Body mass index, kg/m2	15 (25)	25.5	23.39	24.9	26.8	28
Coronary artery disease, %	27 (44)	42	3	28	53	80
Congestive heart disease, %	9 (15)	24	14	16	29	57
Diabetes, %	47 (77)	37	8	29	49	71
Men, %	52 (85)	59	27	53	66	81
Peripheral vascular disease, %	22 (36)	16	0	9	23	53
Patients with a fistula, %	59 (97)	100	38	100	100	100
White, %	23 (38)	74	44	63	90	100
Upper arm, %	41 (67)	44	0	30	100	100

There were 61 cohorts in total. N, total number of cohorts that reported the variable of interest; Min/Max, minimum/maximum reported value; Q1 and Q3, the values which 25% and 75% of the data set lie below, respectively.

Study Quality Assessment

Methods were inadequately reported and definitions were inconsistent across studies. Definitions were not reported for five of nine studies reporting aneurysm, seven of 23 studies reporting infections, nine of 12 studies reporting ischemic steal syndrome, 13 of 18 studies reporting thrombosis, and the one study reporting venous hypertension. Loss to follow-up was not reported in 28 out of 38 observational studies and one of the four randomized controlled trials. However, when reported, the median loss to follow-up was <10% for all studies. For observational studies, 38 of 39 studies were at moderate or high risk of bias in at least one domain. For randomized trials, three of four studies were at moderate or high risk of bias in at least one domain. The distribution of the components that described study quality is summarized in Supplemental Tables 2 and 3.

Incidence of Primary Outcomes

All complication rates reported by study are presented in Supplemental Table 4. Below we report summary statistics. When ≥2 event rates were reported for a specific subgroup, we calculated the median event rate for subgroups of nonelderly and elderly patients, lower and upper arm fistulas, as well as buttonhole and rope ladder cannulation (Supplemental Table 5).

Aneurysm

The median rate of aneurysm was 0.04 events per 1000 patient days (ranging from 0 to 3.01; IQR=0.06) among 14 unique cohorts (n=1827 fistulas; Supplemental Figure 2). Van Loon et al.¹³ reported the highest rate of aneurysm (3.01 per 1000 patient days) among patients using rope ladder cannulation. The next highest rate was reported by the same study for buttonhole cannulation at a rate of 0.13 events per 1000 patient days.¹³

Infection

Twenty-three studies reported the incident rate of infection for 28 unique cohorts. The median rate of total infections was 0.11 infections per 1000 patient days (range 0–0.98; IQR=0.29) among 16 unique cohorts (n>6439 fistulas; Figure 1). However, Ravani et al. reported an event rate of 3, 1.7, and 0.9 infections per 1000 patient days in the first 1–3, 3–6, and 6–12 months of starting dialysis, respectively. Only the latter event rate was entered in the overall median rate.

Figure 1.

Studies reporting the incident rate for all types of infections per 1000 patient days. Dashed line refers to the median event rate per 1000 patient days. AVF-T, transposed arteriovenous fistula; buttonhole, all patients cannulated using a buttonhole technique; elderly, ≥65 years; rope ladder, all patients cannulated using a rope ladder technique; standard, no specific cohort followed.

The median rate for bloodstream infection was 0.05 events per 1000 patient days (range 0–0.55; IQR=0.08) among 9 cohorts (n>753 fistulas; Supplemental Figure 3). The median rate for vascular access site infections was 0.03 events per 1000 patient days (range 0–0.26; IQR=0.05) among 14 unique cohorts (n>771 fistulas; Supplemental Figure 4). Using buttonhole cannulation, the median rate of infection was 0.17 (two studies; N fistulas=302) compared with 0.001 (two studies; N fistulas=527) events per 1000 patient days for rope ladder cannulation.

Ischemic Steal Syndrome

Eleven studies reported event rates of ischemic steal syndrome for 15 unique cohorts (n>2543; Figure 2). The median incident rate of ischemic steal syndrome was 0.05 events per 1000 patient days (range 0–0.27; IQR=0.04). The highest rates were 0.18 and 0.27 events per 1000 patient days reported for elderly patients (≥80 years) and those with an upper arm fistula, respectively.

Figure 2.

Studies reporting the incident rate for ischemic steal syndrome per 1000 patient days. Dashed line refers to the median event rate per 1000 patient days. AVF-T, transposed arteriovenous fistula; elderly, ≥65 years unless otherwise specified; endo AVF-T, fistula transposition of a deep vein through endoscopic procedure; lower arm, cohort made up of all lower arm fistulas; open AVF-T, fistula transposition of a deep vein through a long open incision; standard, no specific cohort followed; upper arm, cohort made up of all upper arm fistulas.

Thrombosis

Fifteen studies reported event rates of thrombosis in 26 unique cohorts (n=4232 fistulas; Figure 3). The median thrombosis rate was 0.24 events per 1000 patient days (range 0.07–0.71; IQR=0.25). Patients undergoing vascular access surveillance (e.g., Qa surveillance) had a median thrombosis rate of 0.33 events per 1000 patient days (seven groups; 721 fistulas). Lin et al.¹⁴ reported 1.43 thrombotic events per 1000 patient days rate at 3 months compared with 0.49 events per 1000 patient days at 12 months (only the event rate at 12 months was included in the median calculation above). A number of studies included prevalent patients and thus did not report thrombotic events occurring before or shortly after starting dialysis.

Figure 3.

Studies reporting the incident rate for thrombosis per 1000 patient days. Dashed line refers to the median event rate per 1000 patient days. AVF-T, transposed arteriovenous fistula; buttonhole, all patients cannulated using a buttonhole technique; DUS monitoring, duplex ultrasound monitoring of blood flow; DVPM, dynamic venous pressure monitoring; elderly, ≥65 years unless otherwise specified; infrared, patients received far infrared therapy 3 times weekly for a year; lower arm, cohort made up of all lower arm fistulas; NM, no monitoring (standard care); Qa surveillance, access blood flow monitoring; standard, no specific cohort followed; TUS surveillance, transonic ultrasound monitoring of blood flow; upper arm, cohort made up of all upper arm fistulas; VABFM, vascular access blood flow monitoring.

Venous Hypertension

Only one study reported an event rate of 0.03 venous hypertensive events per 1000 patient days. However, it should be noted that arm edema was included as an event. This cohort consisted of incident patients (n=350 fistulas) with transposed upper arm fistulas with a mean age of 58 years.¹⁵

Incidence of Secondary Outcomes

Endovascular interventions (34 unique cohorts; n=3930) and surgical (27 unique cohorts; n=2886) revisions were the most commonly reported secondary outcomes. Endovascular interventions included angioplasty (with or without thrombectomy), stenting, and mechanical thrombolysis. The median event rates were 0.82 (range 0–7.42; IQR=1.39) and 0.19 (range 0–0.77; IQR=0.15) per 1000 patient days for endovascular interventions and surgical revisions, respectively. All-cause mortality (13 unique cohorts; n=1880 fistulas), bleeding (seven unique cohorts; n>1672 fistulas), and total fistula complications (six unique cohorts; n>1666) were reported at a median rate of 0.33 (range 0.09–0.64; IQR=0.11), 0.06 (range 0.02–0.11; IQR=0.05), and 0.56 (range 0.13–1.56; IQR=0.79) per 1000 patient days, respectively. Other outcomes such as catheter insertions, fistula ligation due to high cardiac output, difficult cannulations, hematomas, and hospitalizations were infrequently reported (Supplemental Table 4). No study reported on pulmonary hypertension or heart failure.

Discussion

Our review identified 61 unique cohorts that reported on the fistula event rate of aneurysms, infections, steal syndrome, thrombosis, venous hypertension, and our secondary outcomes among patients using a fistula. We identified three important findings: (1) the contemporary rates of complications in fistulae; (2) the critical need for standardized definitions to report complication rates in order to allow proper evaluation of these outcomes across studies; and (3) a clear need to improve reporting and data quality of observational data in the area of vascular access.

Despite the burden of vascular access complications on patients and the healthcare system, there remains a poor consensus on the incidence risk and factors associated with increased rates of fistula complications. The wide variation in complication event rates across studies is attributed to variation in definitions, inconsistent reporting, and differences in patient populations.

With regard to fistula aneurysmal complications, repeated punctures in a clustered area can weaken the vascular access wall and cause aneurysm formation. Aneurysmal dilatation can occur naturally over time because of the higher blood flow and the process may be accelerated by raising the pressure within the fistula. A previous review documented an aneurysm rate between 5% and 6% (no units were provided), a much higher rate than our median rate of 1.5% per year (0.04 per 1000 patient days).¹⁶ Aneurysms generally require surgical repair if there is evidence of loss of skin integrity or ulceration. Left untreated, aneurysms are at risk of rupture and serious hemorrhage, as well as limiting the available sites for cannulation. If inadvertently traumatized, the aneurysm may rupture which can be fatal due to exsanguination.^17,18

The clinical practice guidelines for the National Kidney Foundation Kidney Disease Outcomes Quality Initiative (NKF/KDOQI) and the Canadian Society of Nephrology (CSN) recommend that the infection rate should not exceed 0.01 events per patient year (0.027 per 1000 patient days).^1,2 On the basis of the evidence from the current literature, the rate of infection was higher for 23 out of 30 cohorts compared with recommended rates proposed by practice guidelines. We also found that patients cannulated using the buttonhole technique had a much higher rate of infection compared with rope ladder cannulation. Similar results were found in a recent randomized controlled trial that found 17% of patients had an infection when using buttonhole compared with 0% using the rope ladder technique.¹⁹ Compared with catheters and arteriovenous (AV) grafts, fistulas have lower rates of infection. Previous studies have shown that the rate of catheter infection ranges from 5% to 18% per patient year but this depends on the duration of catheter use.^20,21 Similarly, AV-grafts may be at a higher risk for infections.^2,22 The rate of local and bacteremic infections for AV-grafts can range between 11% and 20% per patient year.^23–26

Ischemic steal syndrome is another important complication of fistulas with significant patient implications including pain and loss of access or extremity function. The NKF/KDOQI and CSN have made no recommendations for target rates of ischemic steal syndrome. Previous studies have reported symptomatic ischemic steal occurrence to range between 1% and 2% for lower arm fistulas and between 5% and 10% among patients using an upper arm fistula.^27–32 Indeed, we found that as the proportion of upper arm fistulas increased, the rate of ischemic steal syndrome increased. We found that the rate of ischemic steal syndrome was similar to estimates reported in the NKF/KDOQI guidelines (1%–4% of patients) and other studies.² Compared with AV-grafts, fistulas have a two-fold lower risk of developing ischemic steal syndrome.^2,22,33

Thrombosis is a common early and late fistula complication that can lead to fistula loss. The current NKF/KDOQI and CSN guidelines recommend that center-specific thrombosis rate for the fistula should not exceed 0.25 events per patient year (0.69 per 1000 patient days).³⁴ Among studies included in our review, the thrombosis rates observed were generally lower (with the exception of two studies) than the target rates recommended by practice guidelines. Furthermore, the rates of thrombosis observed in studies before 2000 were also comparable to rates reported in our study.^35,36 Compared with AV-grafts, fistulas have lower rates of thrombotic events. Previous reports have reported graft thrombosis to exceeded 0.8 events per patient year (2.2 events per 1000 patient days).^37–39

The most common cause of venous hypertension is central stenosis secondary to placement of central venous catheters and devices.^40,41 Symptoms of venous hypertension include finger and hand edema that may progress to limit upper extremity mobility. Beyond swelling, extremely advanced stages of this complication can lead to hand and extremity discoloration and even venous gangrene.⁴² This complication may be under-reported and requires more attention in future studies where they explicitly report its definition, diagnosis, and occurrence.

Studies included in our review varied substantially in quality, outcome definitions, and patient population. Accordingly, the rate of fistula complications varied and may reflect selection bias of study participants, differences in clinical practice, inconsistent levels of vascular access care and monitoring across facilities, and variable case definitions. Despite published recommendations for standardized vascular access reporting,^43–45 the majority of studies failed to report definition of outcomes and only a smaller number of studies used published standardized definitions.

In this review, we do not report on fistula nonmaturation or primary failure. We recently published a separate systematic review and meta-analysis, indicating primary failure occurs in approximately 23% (95% confidence interval [95% CI], 18% to 28%) of created fistulas.⁶ We also found the primary patency rate was 60% (95% CI, 56% to 64%) at 1 year and 51% (95% CI, 44% to 58%) at 2 years. The secondary patency rate was 71% (95% CI, 64% to 78%) at 1 year and 64% (95% CI, 56% to 73%) at 2 years.

Our review has a number of strengths compared with prior narrative reviews on this topic. We conducted a comprehensive search of the literature and systematically identified relevant studies in accordance with published guidelines and a published prespecified protocol. We prespecified a wide range of primary and secondary outcomes that are relevant to patients, physicians, and administrators. To our knowledge, this is the first review that has examined fistula complication in the contemporary dialysis population. Previous narrative reviews have included studies from the 1970s when patient characteristics, clinical focus, and policies differed from the current dialysis environment. Furthermore, previous reviews reported the proportion of patients with an infection, ischemic steal syndrome, and thrombosis events as opposed to the event rate.^46–49

Our study does have limitations. Heterogeneity between studies precluded precise estimates of complication incidence rates and precluded the pooling of risk factors; it has been suggested that pooling results when not warranted may lead to misleading conclusions.⁵⁰ We restricted this review to articles published in English; whether this introduced some bias is controversial.⁵¹ We also did not query prospective databases (e.g., Dialysis Outcomes and Practice Patterns Study, Centers for Medicare & Medicaid Services, etc.) without peer-reviewed published data. Finally, when calculating event rates, we assumed a constant hazard ratio; however, it has been shown that hazard ratios of fistula outcomes can vary over time, with higher hazard rates being observed within the first 6 months of dialysis.⁵²

The quality of primary studies inherently limits the conclusions that can be drawn from this review. We were also unable to differentiate complications that occurred in fistulas that were created in the predialysis period and a number of studies reported outcomes for patients already on dialysis. Risk factors for fistula complications which include patient comorbidities, vessel characteristics, surgeons’ experience, and nursing experience with cannulation were generally not reported. Additionally, factors on timing between fistula creation and fistula use, timing of vascular access interventions, and clinical monitoring or surveillance practices were also not reported. These variables may have explained some of the heterogeneity in the literature. The paucity in reporting of risk factors impairs our ability to identify patients at highest risk for complications.

We found that the rates of aneurysm, ischemic steal syndrome, and thrombosis rates were similar or less than previously published reports and those reported in NKF/KDOQI guidelines (update 2006). However, the infection rate was much higher than the range recommended by NKF/KDOQI guidelines, although this varied by cannulation technique. We found marked variability in complications rates in part due to the poor quality of studies, significant heterogeneity of study populations, and inconsistent definitions. This information on complication rates is critical to informing patient-physician decision making and patient consent, and guiding resource allocation for vascular access care. Having accurate information on fistula complications and patients at highest risk for complications is important in making informed decisions and choosing the appropriate vascular access on the basis of the risks.

There is an urgent need to standardize reporting of methods, baseline patient and access characteristics, and complications of vascular access in future clinical studies.

Concise Methods

We conducted and reported this systematic review according to published guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] checklist: Supplemental Table 1) using a prespecified protocol (CRD42014010444 and Supplemental Item 1).^53,55 Description of studies eligible for review, data sources, study selection, data abstraction, and quality assessment used for this review have been reported elsewhere.^6,55

Briefly, we included English-language studies indexed in Medline that followed patients prospectively (observational cohort, randomized control trials, or surveillance programs) for a period of at least 3 months. Studies included were published between January 1, 2000 and December 31, 2014 and reported on ≥100 fistulas. Two blinded independent reviewers (A.A.A. and one of A.R.L. or J.C.Z.) selected the studies using a standardized form with a third reviewer adjudicating discrepancies when necessary. We assessed risk of bias using previously reported methods and modified these tools as appropriate,^56,57 and abstracted data on study methodology and cohort characteristics. All studies must have reported on at least one of the primary outcomes: (1) aneurysm, (2) infections, (3) ischemic steal syndrome, (4) thrombosis, or (5) venous hypertension among hemodialysis patients using a fistula. When reported, we also captured the following secondary outcomes: (1) difficult cannulation, (2) bleeding, (3) hematoma, (4) catheter insertion, (5) endovascular intervention, (6) surgical revisions, (7) hospitalization (infection-related), (8) hospitalization (VA-related), (9) hospitalizations (all-cause), (10) pulmonary hypertension, (11) high cardiac output, (12) heart failure, (13) all-cause mortality, and (14) total complications.

Study Definitions

We used outcome definitions in accordance with the Society of Vascular Surgery (SVS) and the American Association of Vascular Surgery as well as the North American Vascular Access Consortium (NAVAC) (below).^43,44 The SVS and NAVAC did not report a definition for aneurysm. When the study definition was not in accordance with previously published definitions, we noted the differences in our tables. Definitions for infections and venous hypertension were intentionally kept broad to include all variations in the published literature.

Primary Outcome Definitions

(1) Aneurysm: Diffuse and progressive degeneration of the vascular access site. Patient has signs of bleeding, infection, or ulceration.^16,45 (2) Infections: Definite or probable local vascular access infections, vascular access–related sepsis, bacteremia, or a composite of these infections. (3) Ischemic steal syndrome (SVS): One or more clinical manifestations of: pain, ischemic neuropathy, ulceration, and gangrene felt to be related to a fistula diverting blood from the distal circulation resulting in a zone of arterial insufficiency in the tissues distal to the fistula.⁴⁴ (4) Thrombosis (SVS and NAVAC): Absence of bruit or thrill, using auscultation and palpation, throughout systole and diastole at least 8 cm proximal to the arteriovenous anastomosis.^43,44 (5) Venous hypertension (SVS): High pressure in veins due to damage to venous system often presenting with arm and hand swelling; dusky, rubber color of the hand; and dilated veins on the arm and/or chest wall.

Summary Statistics

We report the median and range for the event rate (per 1000 patient days) of an outcome. Because of differences in sampled populations, outcome definitions, prevalence of comorbid conditions, and variable sample selection criteria, it was not appropriate to calculate a summary statistic on the basis of the weighted average. When incidence rates were not reported, we calculated the overall follow-up time (denominator) by multiplying the mean follow-up time by the number of patients. We used the overall follow-up time to calculate the event rate per 1000 patient days. Using this method, we assumed that the hazard rate of developing a particular outcome was constant across individuals and over time.

Disclosures

None.