Outcomes of Initial Hemodialysis Vascular Access in Patients Initiating Dialysis with a Tunneled Catheter

Abstract

Objective:

To determine factors that influence time to removal of tunneled hemodialysis catheter (THC), probability of repeat vascular access creation and time to repeat vascular access.

Methods:

The Optum Clinformatics database was queried from 2011–2017 for patients who initiated hemodialysis with a THC. Time from initial arteriovenous fistula (AVF)/graft (AVG) to THC removal and time to repeat AVF/AVG were analyzed using Cox proportional hazards. Likelihood of repeat AVF/AVG was analyzed using logistic regression.

Results:

8941 vascular access met the inclusion criteria: 6913 (77%) AVF and 2028 (23%) AVG. Median follow-up was 595 days among AVF patients (range 1–2543) and 579 days among AVG patients (range 1–2529). Patients undergoing AVF were younger, more likely to be male, White race and obese. Patients undergoing AVF were also slightly less likely to have diabetes, cardiac arrhythmia, congestive heart failure and peripheral vascular disease than patients undergoing AVG. At 90 days and at 180 days after index access creation, significantly more patients who underwent index AVG had their THC removed compared to patients who underwent index AVF. By day 365, 78% of patients in both AVF and AVG had their THC removed. 2550 (28.5%) patients underwent a repeat vascular access creation during the follow up period: 30% of index AVF and 24% of index AVG. At 90 days, 180 days and 365 days, significantly more patients in the index AVF group underwent a repeat vascular access creation than those in the index AVG group. Multivariate analysis demonstrated a significant interaction between vascular access type and age≥70 (P<0.001) for time to THC removal, likelihood of repeat vascular access and time to repeat vascular access. In the age<70 group, patients who underwent AVG were 60% more likely to have a shorter time to THC, had a 50.4% lower odds of repeat vascular access and were 47% more likely to have a longer time to repeat vascular access compared to patients who underwent index AVF. In the age≥70 group, patients who underwent AVG were 98% more likely to have a shorter time to THC removal, had 69.7% lower odds of repeat vascular access and were 66% more likely to have a longer time to repeat vascular access.

Conclusion:

Creation of AVG vs AVF significantly decreases the time to THC removal in dialysis-dependent patients, with a larger difference in patients aged ≥70 vs <70. Initial AVG was associated with lower odds of repeat vascular access and longer time to repeat vascular access. These results suggest that the dictum of “Fistula First” is not appropriate for all patient populations and supports judicious use of AVG in achieving the more recent shift towards “Catheter Last”.

Table of Contents Summary

This study used claims data from a national insurance carrier for 8,941 vascular access creations and found that compared to creation of arteriovenous fistula, arteriovenous graft significantly decreased time to tunneled hemodialysis catheter removal, with patients aged ≥70 years experiencing greater benefit from selecting graft versus fistula.

Introduction

In 2015, there were over 700,000 prevalent cases of end-stage renal disease in the United States with over 124,000 of them being new cases. Of the incident cases, 87.3% were treated with hemodialysis. Out of those patients initiating hemodialysis in 2015, 80% used a catheter for vascular access.¹ Ninety days after the initiation of hemodialysis, 68.5% of those patients were still using a catheter.¹

Tunneled hemodialysis catheters (THC) for hemodialysis vascular access are associated with increased risk of bacteremia and mortality, compared to the alternatives of arteriovenous fistula (AVF) and arteriovenous graft (AVG).^{2, 3} Subsequently, both the Kidney Disease Outcomes Quality Initiative (KDOQI) and the Fistula First Breakthrough Initiative (FFBI) strongly discourage the long-term use of THC for vascular access.^{4, 5} At the same time, both encourage the use of arteriovenous fistula (AVF) over arteriovenous graft (AVG). While the prevalent use of AVF has successfully increased to 63% as of 2016,¹ there is increasing evidence that the emphasis on AVF has resulted in increasing THC dependence due to high rates of AVF failure.^{6, 7}

AVG has been demonstrated to be associated with earlier catheter removal compared to AVF.⁸ However, other factors associated with THC removal have not been investigated. In addition, other authors have suggested the benefit of earlier catheter removal with AVG is balanced by increased need for repeat access procedures after AVG due to reduced patency.⁸ The objective of this study is to determine factors that influence time to removal of tunneled hemodialysis catheter (THC), probability of repeat vascular access creation and time to repeat vascular access after index AVF compared to AVG in patients initiating hemodialysis with a THC.

Methods

Data source and population

A secondary analysis of the Optum Clinformatics Data Mart claims database from 2011 through 2017 was conducted. The database contains de-identified commercial claims from a single national insurance carrier. Over 47 million unique individuals are represented in the database, with 15–18 million unique covered individuals within each year of claims. Medicare Advantage enrollees in the database represent approximately 25% of all Medicare Advantage enrollment. Member information outside of claims included age, ethnicity, and geographic location. The study was approved by the Institutional Review Board of the University of California, Los Angeles, and informed consent was waived.

All patients over the age of 18 with at least one claim for hemodialysis vascular access creation, as identified by Current Procedural Terminology (CPT) codes for AVF (36818, 36819, 36820, 36821) and AVG (36830) were identified. Patients were included only if they had at least one claim for outpatient hemodialysis within 90 days of permanent vascular access creation, if they had at least twelve months of continuous insurance enrollment without any claims for hemodialysis or vascular access prior to the index AVF/AVG, and if they had placement of a THC without removal any time prior to the initial outpatient hemodialysis claim. Removal of the THC and replacement within seven days was treated as continuous THC dependence.⁹ Patients were excluded if they did not have at least one claim with an International Classification of Diseases, 9th edition (ICD-9) or 10th edition (ICD-10) diagnostic code for chronic kidney disease.

Claims were available from 2011 through 2017; however, only index AVF/AVG between January 1, 2012 and December 31, 2016 were included to allow for a minimum 12-month hemodialysis and vascular access-free period of continuous plan enrollment prior to the index AVF/AVG, and a potential of 12-month follow-up after index AVF/AVG. Patients with discontinuous insurance enrollment post-index AVF/AVG were excluded. No minimum amount of insurance enrollment after index AVF/AVG was imposed. This was done in order to avoid biasing the sample towards patients who are more likely to survive for a pre-determined amount of time. Co-morbidities were identified using ICD-9 and 10 codes.

Statistical Analysis

The primary outcome was days until THC removal (CPT 36589) post-index AVF/AVG. Again, THC removal with replacement within seven days was treated as continuous THC dependence. The secondary outcome was repeat permanent vascular access creation after index AVF/AVG. The primary regressors were index permanent vascular access type and age. For ease of interpretation, age was dichotomized into age <70 and age ≥70. The choice of 70 years was based on the median age in the study sample being 70 years. In addition, while there is not an accepted age cutoff for “elderly”, 70 is well above the Medicare qualification age of 65..

In order to assess whether relative outcomes of AVF and AVG differ among younger (<70) and older (≥70) patients an interaction term between access type and age was explored. A significant interaction between access type and age would imply that the difference in THC removal and repeat vascular access between AVF and AVG vary as a function of age. A significant interaction term between access type and age also essentially creates four categories for comparison: 1) AVF in age <70, 2) AVF in age<70, 3) AVG in age ≥70 and 4)AVG in age ≥70. In the results tables, the effect of the interaction term is reported by using various appropriate reference classes. This allows for clear comparisons of between group and within group variation for both age and access type.

Other covariates included in the multivariable models were race (categorized as non-Hispanic White, non-Hispanic Black, non-Hispanic Asian, Hispanic of any race, and unknown), sex, cardiac arrhythmia (CA), congestive heart failure (CHF), peripheral vascular disease (PVD), diabetes, and obesity. Hypertension was not included in the model as nearly all patients had hypertension.

Descriptive statistics were given as frequencies and percentages. Univariate differences between access types were assessed with chi-squared tests, t-tests, and Wilcoxon rank-sum tests as appropriate. Likelihood of repeat vascular access was analyzed using logistic regression. Survival analysis was performed using the life-table method and AVF and AVG were compared using log-rank test. Cox proportional hazards models were used to determine factors associated with time to THC removal and time to repeat vascular access. Patients were censored at the time they underwent kidney transplant or terminated enrollment. Logistic regression allows us to estimate the likelihood of repeat vascular access, however it does not account for varying lengths of follow up. Cox proportional hazards accounts for varying lengths of follow up using censoring, however, it measures the likelihood of an event accounting for time to event. Both approaches were used to evaluate repeat vascular access and the results are considered in conjunction. Sample selection was performed using SAS 9.2 (SAS institute, Cary, NC) and the remainder of statistical analysis was performed using Stata 15 (StataCorp, College Station, TX).

Results

8941 vascular access met the inclusion criteria: 6913 (77%) AVF and 2028 (23%) AVG. Median follow-up was 591 days overall (range 1–2543), 595 days among AVF patients (range 1–2543), and 579 days among AVG patients (range 1–2529). 6,391 patients had Medicare advantage insurance, 69% of AVF patients (n=4,802) and 78% of AVG patients (n=1,589). Patients undergoing AVF were younger, more likely to be male, White race and obese (Table I). Patients undergoing AVF were also slightly less likely to have diabetes, cardiac arrhythmia, congestive heart failure and peripheral vascular disease than patients undergoing AVG (Table I). However, the incidence of these co-morbidities was high in both groups and typical for vascular access patients.

Table I.

Demographics and Co-morbidities

	Fistula (%) n=6486	Graft (%) n=1869	P-value
Mean age (SD)	67.2 (13)	70.6 (12)	<0.001
Age≥70	3412 (49)	1222 (60)	<0.001
Male	4184 (61)	895 (44)	<0.001
Race			<0.001
Asian	226 (3)	59 (3)
Black	1278 (18)	553 (27)
Hispanic	1043 (15)	243 (12)
White	3305 (48)	877 (43)
Unknown	1061(15)	296(14)
Hypertension	6421 (99)	1850 (99)	1.0
Diabetes	5732 (83)	1737 (86)	0.004
Cardiac Arrhythmia	5285 (76)	1604 (79)	0.01
Congestive Heart Failure	5573 (81)	1687 (83)	0.009
Peripheral Vascular Disease	5025 (73)	1551 (76)	0.001
Obesity	2928 (42)	792 (39)	0.008

At 90 days and at 180 days after index access creation, significantly more patients who underwent index AVG had their THC removed compared to patients who underwent index AVF (Figure 1). By day 365, 78% of patients in both AVF and AVG had their THC removed (Figure 1). Among the patients who had their catheter removed within 365 days after access creation (n=6,184), 4085 (91%) of the AVF patients and 1426 (92%) of the AVG patients had their catheter removed as an outpatient.

Figure 1.

Unadjusted time to tunneled catheter removal

Multivariate analysis of time to THC removal demonstrated a significant interaction between vascular access type and age (P<0.001) (Figure 2). In the age<70 group, patients who underwent AVG were 60% more likely to have a shorter time to THC removal than patients who underwent AVF. (Table II) In the age≥70 group, patients who underwent AVG were 98% more likely to have a shorter time to THC removal than patients who underwent AVF. (Table II) Male sex and Hispanic ethnicity were associated with significantly shorter time to THC removal. (Table II) Diabetes and obesity were associated with significantly longer time to THC removal. (Table II)

Figure 2.

Adjusted time to tunneled catheter removal

Table II.

Multivariable Analysis of Time to Tunneled Catheter Removal

Co-variate	Hazard Ratio	95% Confidence Interval
Arteriovenous Fistula, Age ≥ 70 (vs Fistula, Age < 70)	0.94	0.89, 1.00
Arteriovenous Graft, Age <70 (vs Fistula, Age <70)	1.60***	1.46, 1.75
Arteriovenous Graft, Age ≥70 (vs Fistula, Age ≥70)	1.98***	1.84, 2.14
Arteriovenous Graft, Age > 70 (vs Graft, Age < 70)	1.17**	1.06, 1.30
Male sex (vs Female)	1.14***	1.09, 1.2
Race
White (ref)	1
Asian	1.02	0.89, 1.18
Black	0.98	0.92, 1.04
Hispanic	1.1*	1.06, 1.18
Unknown	1	0.98, 1.13
Diabetes	0.93*	0.86, 0.99
Cardiac Arrhythmia	0.95	0.89, 1.0
Congestive Heart Failure	0.98	0.92, 1.05
Peripheral Vascular Disease	0.97	0.91, 1.03
Obesity	0.93**	0.88, 0.98

^*significant at P=0.05

^**significant at P=0.01

^***significant at P=0.001

2550 (28.5%) patients underwent a repeat vascular access creation during the follow-up period: 30% of index AVF and 24% of index AVG (p<0.0001). Multivariate analysis of likelihood of repeat vascular access demonstrated a significant interaction between index vascular access type and age (P=0.001). In the age <70 group, index graft reduced the odds of

repeat vascular access by 50.4% compared to index fistula. (Table III) In the age ≥70 group, index graft reduced the odds of repeat vascular access by 69.7% compared to index fistula. (Table III) Male sex significantly reduced the odds of repeat vascular access. (Table III) Black race, peripheral vascular disease and obesity significantly increased the odds of repeat vascular access. (Table III)

Table III.

Multivariable Analysis of Likelihood of Repeat Vascular Access

Co-variate	Odds Ratio	95% Confidence Interval
Arteriovenous Fistula, Age ≥ 70 (vs Fistula, Age < 70)	0.83**	0.74, 0.93
Arteriovenous Graft, Age <70 (vs Fistula, Age <70)	0.50***	0.40, 0.61
Arteriovenous Graft, Age ≥70 (vs Fistula, Age ≥70)	0.30***	0.24, 0.38
Arteriovenous Graft, Age ≥ 70 (vs Graft, Age < 70)	0.50***	0.38, 0.66
Male sex (vs Female)	0.75***	0.98, 0.84
Race
White (ref)	1
Asian	1.02	0.75, 1.40
Black	1.56***	1.37, 1.79
Hispanic	0.99	0.84, 1.16
Unknown	1.05	0.90, 1.22
Diabetes	1.06	0.92, 1.24
Cardiac Arrhythmia	1.09	0.96, 1.24
Congestive Heart Failure	0.97	0.84, 1.11
Peripheral Vascular Disease	1.18**	1.04, 1.34
Obesity	1.18**	1.06, 1.31

^**significant at P=0.01

^***significant at P=0.001

At 90 days, 180 days and 365 days, significantly more patients in the index AVF group underwent a repeat vascular access creation than those in the index AVG group. (Figure 3) Multivariate analysis of time to repeat vascular access demonstrated a significant interaction between index vascular access type and age (P=0.002). (Table IV) In the age <70 group, patients who underwent index AVG were 47% more likely to have a longer time to repeat vascular access compared to patients who underwent index AVF. In the age ≥70 group, patients who underwent index AVG were 66% more likely to have a longer time to repeat vascular access compared to patients who underwent index AVF. Male sex was associated with a significantly longer time to repeat vascular access. Black race and obesity were associated with a significantly shorter time to repeat vascular access.

Figure 3.

Time to repeat vascular access

Table IV.

Multivariable Analysis of Time to Repeat Vascular Access

Co-variate	Hazard Ratio	95% Confidence Interval
Arteriovenous Fistula, Age ≥ 70 (vs Fistula, Age < 70)	0.83***	0.75, 0.91
Arteriovenous Graft, Age <70 (vs Fistula, Age <70)	0.53***	0.44, 0.64
Arteriovenous Graft, Age ≥70 (vs Fistula, Age ≥70)	0.34***	0.28, 0.42
Arteriovenous Graft, Age ≥ 70 (vs Graft, Age < 70)	0.53***	0.41, 0.69
Male sex (vs Female)	0.78***	0.71, 0.86
Race
White (ref)	1
Asian	0.98	0.74, 1.29
Black	1.43***	1.28, 1.61
Hispanic	0.95	0.82, 1.1
Unknown	1	0.87, 1.15
Diabetes	1.04	0.91, 1.19
Cardiac Arrhythmia	1.05	0.93, 1.17
Congestive Heart Failure	0.96	0.85, 1.08
Peripheral Vascular Disease	1.05	0.94, 1.17
Obesity	1.11*	1.01, 1.22

^*significant at P=0.05

^**significant at P=0.01

^***significant at P=0.001

Discussion

The ideal vascular access has been described as one which “delivers a flow rate to the dialyzer adequate for the dialysis prescription, has a long use-life, and has a low rate of complications.”⁵ The KDOQI guidelines refer to the “surgically created fistula” as coming “closest to fulfilling these criteria.”⁵ These conclusions were based on older data that showed AVF have superior patency to AVG and required the fewest interventions.^{10, 11} These older studies did not address all hemodialysis sub-populations which may have contributed to the results demonstrated. Since then, a number of studies have shown that AVF is not necessarily superior to AVG in all groups, including the elderly and non-white race.^{12, 13}

The Fistula First Initiative has been renamed “Fistula First, Catheter Last”, to emphasize the poor outcomes associated with long-term THC use and perhaps, place less emphasis on AVF.¹⁴ However, neither the KDOQI nor the Fistula First guidelines provide guidance regarding factors that should influence the selection of the most appropriate vascular access type for a given patient.

These analyses demonstrated that consistently across the outcomes of time to THC removal, likelihood of repeat vascular access and time to repeat vascular access, AVG results in superior outcomes compared to AVF. These results can all likely be attributed to the high AVF maturation failure rate that has been repeatedly reported in the literature.^15–17 AVG do not require a maturation period and in the case of early cannulation grafts, can be punctured as early as 24–48 hours after creation, allowing for earlier THC removal.^{18, 19} Further, a non-maturing AVF is typically abandoned within three to six months after creation and a second permanent vascular access is attempted. At the same time, the secondary patency of AVG at twelve months is 70–80%.^20–22 Thus, the first repeat vascular access for the majority of AVG will not occur until after twelve months. The combination of these two phenomena is consistent with a shorter time to repeat vascular access after index AVF.

Another consistent finding in this study is the interaction between index access type and age category with a larger difference between index AVG and AVF in the age ≥70 group compared to the age <70 group for all outcomes. This can be attributed to the higher incidence of AVF maturation failure and lower AVF patency that has been demonstrated in the elderly compared to the non-elderly.^23–26 Higher rates of maturation failure in the elderly lead to longer times to THC removal, higher rates of repeat vascular access and shorter time to repeat vascular access after index AVF. This interaction between index access type and age category is particularly relevant in the elderly. In the age ≥65 hemodialysis population, all-cause mortality is 31% per year.^{1, 27} This emphasizes the importance of considering short-term vascular access outcomes for the elderly. If the elderly patient does not survive to experience the long-term benefits of AVF over AVG, then the negative short-term consequences of multiple attempts to achieve a functional AVF may not be justified.

Male sex was consistently associated with improved outcomes in this analysis. There are conflicting data in the literature suggesting that females have lower maturation rates and lower patency rates compared to males.^{6, 28–30} This may be due to smaller vein diameters in females compared to males; however this has not be definitively demonstrated. In this claims database, vein diameters were not available and could not be included in the analysis.

Obesity was consistently associated with worse outcomes in this analysis. Obese body habitus is often associated with deeper veins and increased difficulty in cannulating a vascular access. AVG can be constructed such that the graft is more superficial. Nevertheless, obese body habitus can create torque and tension on the graft that can make cannulation difficult and negatively influence hemodynamics and patency. The challenge is compounded in AVF where the anatomic depth of the vein may require surgical elevation, lipectomy and/or liposuction.^31–33 These factors can all increase AVF/AVG failure, resulting in dependence on THC and increased risk of repeat vascular access.

Black race was associated with increased risk of repeat vascular access and shorter time to repeat vascular access but was not associated with time to THC removal. There is some literature to suggest that Black race is associated with worse outcomes after vascular access, regardless of access type.^34–36 The influence of race and ethnicity on vascular access outcomes has not been well studied and should be a focus of future investigation.

The primary limitation of this study is related to the constraints of the data that are available through a claims database. A number of factors associated with vascular access outcomes, including vein diameter and arterial inflow quality, are not available in this dataset. Other important outcomes, namely vascular access revisions and vascular access-related hospitalizations are difficult to measure in a claims database due to the vagaries of billing codes. With the advent of ICD-10, these outcomes may be more clearly identified in the future.

Ultimately, this is an observational study with non-random assignment of patients to index AVF and AVG. Although as many co-variates as possible were controlled for in the analysis, it is possible the outcomes observed are a result of other unmeasured factors. A potential unmeasured confounder was tobacco use. While ICD diagnostic codes for personal history of tobacco use exist, the use of this code in billing data is likely to be highly unreliable, thus it was not included as a covariate. It is also not possible to know what proportion of the entire population initiated hemodialysis with a THC or non-THC access due to the clean period of continuous insurance enrollment that we imposed prior to AVF/AVG creation to maximize the likelihood that the index AVF/AVG was truly the patient’s first AVF/AVG. Without a clean period, it is possible that the patient had a different type of insurance previously and another access was created under that insurance.

The benefit of using this dataset is that it captures Medicare Advantage patient events that are not captured in the Medicare data. The great majority of the patients in this study are patients who had private insurance at the time they became dialysis dependent and elected to choose a Medicare Advantage plan when they became Medicare eligible by virtue of ESRD. ESRD patients do not qualify for Medicare until the first day of their fourth month of dialysis. Thus, vascular access events that occur prior to that are not measured in Medicare claims for patients that qualify for Medicare by virtue of ESRD only, which applies to most ESRD patients younger than age 65. Thus, this dataset allows us to examine patients who become dialysis dependent at age <65 and those who elect for a Medicare Advantage plan.

Conclusion

Creation of AVG vs AVF significantly decreases the time to THC removal in dialysis-dependent patients, with a larger difference in patients aged ≥70 vs <70. Initial AVG was associated with lower odds of repeat vascular access and longer time to repeat vascular access. These results suggest the dictum of “Fistula First” is not appropriate for all populations and support the judicious use of AVG in achieving the more recent shift towards “Catheter Last”.

Article Highlights.

Type of Research:

Retrospective cohort study of national commercial claims

Key Findings:

At age <70 and age ≥70 years, patients with index arteriovenous graft creation were more likely to have a shorter time to tunneled catheter removal (60% and 98%, respectively) and a longer time to repeat vascular access (47% and 66%, respectively) compared to patients with index arteriovenous fistula.

Take Home Message:

“Fistula First” may not produce optimal outcomes for all patient populations, and judicious use of arteriovenous graft should be considered.