Conversion of Vascular Access Type Among Incident Hemodialysis Patients: Description and Association With Mortality

Dialysis Conversion of Vascular Access Type Among Incident Hemodialysis Patients: Description and Association With Mortality Brian D. Bradbury, DSc, MA,1 Fangfei Chen, MPH,1 Anna Furniss, MS,2 Ronald L. Pisoni, PhD, MS,3 Marcia Keen, PhD,4 Donna Mapes, PhD,3 and Mahesh Krishnan, MD, MPH4 Background: Limited data exist describing vascular access conversions during the first year on dialysis therapy or the effect of converting to and from a catheter on subsequent mortality risk. Study Design: Retrospective cohort study. Setting & Participants: We studied a random sample of incident US hemodialysis patients (initiated long-term dialysis ⬍ 30 days before study entry) in the Dialysis Outcomes and Practice Patterns Study (DOPPS; 1996-2004). Predictors: At dialysis therapy initiation, we assessed vascular access type in use (arteriovenous fistula [AVF], arteriovenous graft [AVG], or catheter) and other patient characteristics. We characterized changes in vascular access type (conversions) by using regularly collected functional status information. Outcome & Measurements: We assessed time to all-cause mortality. We first described conversions, then used time-dependent Cox regression to estimate mortality hazard ratios (HRs) for conversions from a catheter to a permanent vascular access (versus no conversion) and conversions from a permanent vascular access to a catheter (versus no conversion). Results: The study included 4,532 patients; 69.2% were dialyzing with a catheter; 17.6%, with an AVG; and 13.1%, with an AVF. In patients initiating therapy with an AVF or AVG, 22% experienced a conversion (failure), and median times to first failure were 62 and 84 days, respectively. In catheter patients, 59% converted to an AVF/AVG (predominantly AVG [57%]); median times to first conversion were 92 and 66 days, respectively. Conversion to a permanent access was associated with an adjusted mortality HR of 0.69 (95% confidence interval, 0.55 to 0.85). The effect was similar for conversion to an AVF or AVG, and these persisted across demographic groups and facilities with different conversion practices. Conversion from a permanent vascular access to a catheter was associated with an adjusted mortality HR of 1.81 (95% confidence interval, 1.22 to 2.68). Limitations: Potential for residual confounding because of unmeasured factors influencing decision to convert. Conclusion: Vascular access conversions are common in incident patients. Continued efforts to increase early nephrologist referral and permanent vascular access placement may help decrease mortality risk in incident dialysis patients. Am J Kidney Dis 53:804-814. © 2009 by the National Kidney Foundation, Inc. INDEX WORDS: Vascular access; hemodialysis; mortality.

M

ajor initiatives are underway in the United States to increase the proportion of hemodialysis (HD) patients with a permanent arteriovenous fistula (AVF) access and decrease the number of catheter placements.1 Mounting evidence describes the increased risk of infectious and cardiovascular complications2-6 and mortality associated with catheter access.7-14 The most 1

From the Department of Biostatistics and Epidemiology, Amgen Inc, Thousand Oaks, CA; 2Colorado Health Institute, Denver, CO; 3Arbor Research Collaborative for Health, Ann Arbor, MI; and 4Medical Research Management, Amgen Inc, Thousand Oaks, CA. Received May 13, 2008. Accepted in revised form November 20, 2008. Originally published online as doi: 10.1053/j.ajkd.2008.11.031 on March 6, 2009. 804

recent US Renal Data System (USRDS) Annual Data Report15 estimated the prevalence of catheter use in incident dialysis patients at approximately 65%, an estimate that has remained relatively constant during the past decade. Studies restricted to incident dialysis patients have shown that catheters are associated with increased mortality risk compared with AVF or arteriovenous Address correspondence to Brian D. Bradbury, DSc, MA, 1 Amgen Center Dr, MS: 24-2-A, Thousand Oaks, CA 91320. E-mail: [email protected] © 2009 by the National Kidney Foundation, Inc. 0272-6386/09/5305-0013$36.00/0 doi:10.1053/j.ajkd.2008.11.031

American Journal of Kidney Diseases, Vol 53, No 5 (May), 2009: pp 804-814

Vascular Access Conversions and Mortality

graft (AVG) accesses.9,12-14 However, in prevalent patients, approximately 27% are dialyzing with a catheter at any given time,16 suggesting that incident patients are either converting to other vascular access types or are not surviving. However, there are few data sources that adequately capture detailed longitudinal vascular access data and clinical outcomes to enable characterization of conversions between vascular access types and evaluation of the potential effect of failing to convert patients from catheters to more sustainable accesses (AVFs or AVGs) in the incident HD population. One data source that does is the Dialysis Outcomes and Practice Patterns Study (DOPPS). We used DOPPS data to describe and quantify the frequency of vascular access conversions in US incident HD patients during their first year on dialysis therapy and evaluate the mortality risks associated with conversion to and from a catheter.

METHODS Data Source Data from US facilities participating in DOPPS phase I (1996-2001) and phase II (2002-2004) were used for this analysis. Detailed descriptions of data collection and elements are available.17,18 Briefly, DOPPS is an international cohort study of practice patterns and outcomes in HD patients. Facilities were eligible for inclusion in DOPPS if they were treating more than 25 patients. A stratified randomselection process was used to identify facilities for participation. Within each of these facilities, an initial cohort of 20 to 40 HD patients older than 18 years was randomly selected for participation. In DOPPS I, the initial study cohort size was maintained by replenishment of departing patients with randomly selected patients new to each facility. In DOPPS II, replenishment was not performed; rather, the cohort was supplemented with less than 15 consecutive patients per facility who started end-stage renal disease (ESRD) within 30 days of study entry. At enrollment, site investigators collected patient information, including demographic characteristics, predialysis care, and laboratory and medical history information from medical records. Throughout follow-up, site investigators collected treatment, laboratory, hospitalization, and mortality information from patient records at 4-month intervals. Primary and secondary causes of hospitalization and mortality and reasons for loss to follow-up were collected. Vascular access information, including vascular access type (AVF, AVG, temporary catheter, or cuffed percutaneous catheter), surgical procedures (placement and removal), functional status (first use, functioning, and in place but not used), and dates were collected for vascular access– related events during the study. The DOPPS received institutional review board approval in all participating countries, and patient consent was obtained.

805

Study Population We limited our study population to incident US HD patients, defined as initiating long-term dialysis therapy less than 30 days before study entry. All patients were required to have more than 1 day of study follow-up.

Access Conversion Patients were classified according to the vascular access in use at enrollment (AVF, AVG, or temporary or cuffed catheter). Temporary and permanent catheters were combined to simplify the conversion assessment. We used separate algorithms for defining conversion from a catheter to an AVF or AVG and from an AVF or AVG to a catheter. We required that patients have only 1 access in use at all times. We defined conversion from a catheter to an AVF or AVG if a “first use” or “functioning” status linked to the AVF/AVG access appeared in the record of a patient previously dialyzing with a catheter. The date corresponding to a newly functioning AVF/AVG was assigned as the conversion date. However, if a patient had “placement” of an AVF/AVG followed by catheter “removal” followed by “first use” or “functioning,” we assigned the date of the catheter removal as the conversion date. We defined conversion from a functioning AVF/AVG to a catheter when a record of catheter “placement,” “first use,” or “functioning” occurred. Conversions from an AVF to an AVG and vice versa were observed; for these situations, we used the same rules used to define conversion from a catheter to an AVF/AVG.

Covariates Patient characteristics were assessed at DOPPS enrollment. Demographic characteristics included age, sex, race (white and other), body mass index (kg/m2), primary cause of ESRD (diabetes, hypertension, glomerulonephritis, other, and unknown), and insurance status (Medicare, Medicaid, other, and no insurance). Pre-ESRD medical care information included receipt of nephrologist care more than 30 days before dialysis therapy initiation (yes, no, and unknown) and permanent vascular access placement (yes, no, and unknown). Other information included urea reduction ratio (%) and predialysis systolic and diastolic blood pressure (mm Hg). Medical history was categorized as yes (which included suspected) or no for the following conditions: coronary artery disease, cancer, other cardiovascular disease, cerebrovascular disease, congestive heart failure, diabetes mellitus, gastrointestinal bleeding in the past 12 months (gastrointestinal bleed), human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS), hypertension, lung disease, neurological disease, psychiatric disorder, peripheral vascular disease, and recurrent cellulitis. The following laboratory parameters were assessed: serum albumin (g/dL), total serum calcium (mg/dL), hemoglobin (g/dL), serum phosphorus (mg/dL), and white blood cell count (⫻103/␮L).

Person-Time Participants were followed from study entry until the first of the following: date of death; loss to follow-up for reasons including transplantation, transfer to another facility, renal recovery, conversion to peritoneal dialysis therapy, or with-

806 drawal; or 365 days. Patients who withdrew from dialysis therapy were followed for an additional 21 days to assess mortality.

Statistical Analysis Descriptive We used basic descriptive statistics for continuous variables (mean ⫾ SD, median, and 25th and 75th percentiles) and categorical variables (frequency [number of patients] and percent) to characterize patients according to baseline vascular access type and according to conversion status (converter versus nonconverter). Comparisons between converters and nonconverters were limited to conversions to and from a catheter. We used Student’s t-test and MantelHaenszel ␹2 analysis to estimate P values and conducted all analyses stratified by the baseline vascular access type because conversion from a catheter to a permanent vascular access is recommended, but conversion from a permanent vascular access to a catheter is not (ie, generally performed when the permanent vascular access is failing). We estimated time to first conversion and to the end of follow-up for all patients according to their conversion pattern and used Kaplan-Meier methods to estimate the cumulative probability of conversion during the first year.

Time Dependent We used time-dependent Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between time to death during the first year and first conversion to and from a catheter. In these analyses, person-time contribution was based on the vascular access type in use; when a patient converted, the remainder of his or her person-time contribution was attributed to the new vascular access type in use. We also examined conversions between specific vascular access types. The reference group for each comparison was patients who did not convert to a different vascular access type. For example, in the analysis examining conversion from an AVF to a catheter, patients who converted to a catheter were compared with patients who continued dialyzing with an AVF. We present unadjusted and multiplevariable adjusted HR estimates and 95% CIs. We adjusted for all available baseline covariates, as well as calendar year, and accounted for the correlation within facilities using the sandwich estimator.19 We examined the interaction between various baseline demographic characteristics and conversion status on the risk of death by using interaction terms; P less than 0.05 was considered evidence of interaction. Finally, we conducted a separate analysis limited to participants who initiated dialysis therapy with a catheter and who had a permanent vascular access placed before dialysis therapy initiation. For all continuous variables with missing data, we imputed the mean value from participants with nonmissing data. In all cases, missing values did not exceed 10%.

Facility Stratified Facility-level vascular access practices also may influence the likelihood and type of conversions in incident patients. To investigate this, we examined the relationship between

Bradbury et al time to death and conversion from a catheter to a permanent vascular access stratified on the facility’s observed percentage of conversions (catheter to permanent vascular access) at 120 days after dialysis therapy initiation. Evaluating patient-level effects within categories of facility practices can provide a useful framework for estimating effects in the presence of unmeasured confounding.20 For this analysis, we identified all patients who initiated dialysis therapy with a catheter and survived to 120 days. We selected 120 days because greater than 50% of patients had converted to a permanent vascular access by this time (Table 2). In each facility, we calculated the percentage of patients dialyzing with a permanent vascular access as of 120 days (facility permanent vascular access percentage) and categorized them into quartiles (0% to 18%, 19% to 32%, 33% to 48%, and ⬎ 48%). Facilities with fewer than 10 patients were excluded from this analysis. We then used Cox regression to estimate the association between the vascular access type in use at 120 days (permanent vascular access versus catheter) and mortality through the end of the first year, adjusting for baseline covariates and stratified according to quartiles of the facility permanent vascular access percentage. We did not conduct an equivalent analysis for conversion from a permanent vascular access to a catheter because of the small number of these conversions observed in the data. All analyses were conducted using SAS, version 9.1 (SAS Institute, Cary, NC).

RESULTS We identified 4,687 incident US HD patients in DOPPS, and of those, we excluded 121 (2.5%) who had missing information for baseline vascular access type and 34 (0.1%) who had “other” listed as the vascular access in use, which left 4,532 patients for analysis. Mean age in the study population was 62.5 ⫾ 15.0 years, 56.3% were men, 65.3% were white, 27.5% were African American, and 48.4% had diabetes as a cause of ESRD. At dialysis therapy initiation, 595 participants (13.1%) were dialyzing with an AVF; 800 (17.6%) with an AVG; and 3,137 (69.2%) with a catheter (Table 1). The distribution of patient characteristics differed considerably across baseline vascular access type. Patients initiating therapy with an AVF were more likely to be younger, men, have had predialysis care or prior permanent vascular access placement; were more likely to have higher albumin and phosphorus levels, but lower white blood cell counts; and were less likely to have coronary artery or other cardiovascular disease, heart failure, and peripheral vascular disease. Differences between patients initiating therapy with an AVF versus an AVG were

Vascular Access Conversions and Mortality

807

Table 1. Baseline Patient Characteristics According to Vascular Access Type in Use at Dialysis Therapy Initiation Patient Characteristics

Demographics Age (y) Body mass index (kg/m2) Men (%) White (%) Laboratory values Albumin (g/dL) Calcium (mg/dL) Hemoglobin (g/dL) Phosphorus (mg/dL) White blood cell count (103/␮L) Other characteristics Diabetes as cause of ESRD (%) ⱖ1 Predialysis vascular access placement (%) Medicare as primary payer (%) Diastolic blood pressure* (mm Hg) Systolic blood pressure* (mm Hg) Visited a nephrologist† (%) Comorbidities (% yes) Coronary artery disease Cancer Other cardiovascular Cerebrovascular disease Heart failure Diabetes Gastrointestinal bleeding HIV/AIDS Hypertension Lung disease Neurological disease Psychiatric disorder Peripheral vascular disease Recurrent cellulitis

AVF (n ⫽ 595)

AVG (n ⫽ 800)

Catheter (n ⫽ 3,137)

P

60.8 ⫾ 15.8 27.1 ⫾ 7.1 72.8 69.1

64.7 ⫾ 14 28.1 ⫾ 8 45.4 61.8

62.5 ⫾ 16 27.2 ⫾ 7.3 56.0 65.5

⬍0.001 0.03 ⬍0.001 0.5

3.6 ⫾ 0.5 8.4 ⫾ 1 10.2 ⫾ 1.7 5.8 ⫾ 1.8 7.5 ⫾ 2.6

3.5 ⫾ 0.5 8.4 ⫾ 0.9 10.0 ⫾ 1.5 5.4 ⫾ 1.7 8.0 ⫾ 3.0

3.3 ⫾ 0.5 8.5 ⫾ 0.9 10.0 ⫾ 1.6 5.4 ⫾ 1.8 8.8 ⫾ 3.5

⬍0.001 0.4 0.04 ⬍0.001 ⬍0.001

43.4 88.7 58.7 79.9 ⫾ 15.6 154.1 ⫾ 25.6 80.0

54.0 85.3 62.6 75.3 ⫾ 14.7 150.6 ⫾ 26.7 71.4

47.9 68.3 57.3 76.4 ⫾ 16.3 146.8 ⫾ 26.8 50.6

0.5 ⬍0.001 0.2 ⬍0.001 ⬍0.001 ⬍0.001

46.7 11.9 22.2 13.8 33.4 47.7 7.6 1.2 84.9 13.3 8.4 22.7 20.2 6.1

50.0 13.9 25.3 20.0 43.6 59.4 6.8 0.4 84.6 11.4 8.6 22.5 25.4 7.5

53.6 12.6 32 19.3 50.0 53.3 7.7 1.5 83.2 16.8 10.4 28.5 28.9 9.0

0.01 0.9 ⬍0.001 0.01 ⬍0.001 0.3 0.6 0.1 0.2 0.001 0.06 ⬍0.001 ⬍0.001 0.01

Note: Conversion factors for units: albumin in g/dL to g/L, ⫻10; calcium in mg/dL to mmol/L, ⫻0.2495; hemoglobin in g/dL to g/L, ⫻10; phosphorus in mg/dL to mmol/L, ⫻0.3229. Abbreviations: AIDS, acquired immunodeficiency syndrome; AVF, arteriovenous fistula; AVG, arteriovenous graft; ESRD, end-stage renal disease; HIV, human immunodeficiency virus. *Predialysis. †Visited nephrologist more than 1 month before initiating dialysis therapy.

less pronounced than differences between patients with an AVF versus a catheter. Cumulative probabilities of conversion during the first year were 22.4%, 21.6%, and 59.0% for patients initiating therapy with an AVF, AVG, and catheter, respectively (Fig 1). In patients initiating with an AVF, 86% of first conversions were to a catheter, with a median time of 62 days (25th and 75th percentiles, 12 and 134; Table 2). In patients initiating therapy with an AVG, 94% of first conversions were to a catheter, with a median time of 84 days (25th and 75th percentiles, 29 and 173). In patients initiating therapy with a catheter, 57% of first conversions were to

an AVF, with a median time of 92 days (25th and 75th percentiles, 48 and 168), and 53% were to an AVG, with a median time of 66 days (25th and 75th percentiles, 30 and 131). Few differences were observed between converters and nonconverters for patients initiating dialysis therapy with a permanent vascular access (Table 3). However, there were considerable differences between converters and nonconverters for those initiating therapy with a catheter. Converters were more likely to be younger, be nonwhite, have a greater body mass index, have had more than 1 predialysis permanent vascular access placement, have visited a nephrologist

808

Bradbury et al

Figure 1. Cumulative probability of vascular access conversion during the first year on dialysis therapy according to baseline vascular access type. Abbreviations: AVF, arteriovenous fistula; AVG, arteriovenous graft.

before HD therapy initiation, have lower hemoglobin levels, and be less likely to have coronary artery, cerebrovascular, neurological, or peripheral vascular disease. Of note, 62% of patients who had a catheter in use at dialysis therapy initiation, but never converted to a permanent vascular access during the first year, had a permanent vascular access in place before initiating dialysis therapy. In time-dependent analyses, conversion from a catheter to a permanent vascular access was associated with a mortality HR of 0.56 (95% CI, 0.46 to 0.69). After adjustment for confounding, the HR was attenuated to 0.69 (95% CI, 0.55 to 0.85; Table 4). We observed similar adjusted HR estimates when we separately evaluated conversion to an AVF (HR, 0.64; 95% CI, 0.47 to 0.87) and conversion to an AVG (HR, 0.71; 95% CI, 0.55 to 0.92). Our results did not change materially when we limited our analysis to patients who initiated dialysis therapy with a permanent vascular access in place, but not in use (HR, 0.74; 95% CI, 0.58 to 0.95; data not shown). Conversion from a permanent vascular access to a catheter was associated with a mortality HR of 1.63 (95% CI, 1.11 to 2.38), and this effect persisted even after adjustment for confounding (HR, 1.81; 95% CI, 1.22 to 2.68). When evaluated separately, the mortality HR for conversion from an AVF to a catheter was 1.55 (95% CI, 0.84 to 2.87), and for conversion from an AVG to a catheter, 2.11 (95% CI, 1.21 to 3.67). For all time-dependent analyses, we did not observe meaningful differences

in HR estimates associated with conversion within strata of baseline patient characteristics (tests for interaction showed no statistical differences). Of 3,137 patients in 194 facilities initiating dialysis therapy with a catheter, we excluded 96 facilities and 1,330 patients because either the patients did not survive to 120 days or the facility was treating fewer than 10 incident patients, leaving 1,807 patients (58%) in 98 facilities for the facility-stratified analysis. Mean number of incident patients in each facility was 18.4 ⫾ 5.9, and mean percentage of patients using a permanent vascular access at 120 days was 35.5% ⫾ 20.5%. We observed similar HR estimates for patients with a permanent vascular access in use (versus a catheter) in facilities in the lowest (HR, 0.74; 95% CI, 0.25 to 2.21) and highest quartiles of facility permanent vascular access conversion (HR, 0.54; 95% CI, 0.30 to 0.99; Fig 2), although the smaller sample sizes in each strata led to wider CIs. After adjustment for facility differences and patient case-mix, patients dialyzing with a permanent vascular access (versus catheter) had a mortality HR of 0.65 (95% CI, 0.47 to 0.91).

DISCUSSION Using 4,532 randomly sampled incident US HD patients, we characterized vascular access conversions during the first year on dialysis therapy. Nearly 70% of patients were dialyzing with a catheter at therapy initiation, and of those, only 59% were converted to a permanent vascu-

Vascular Access Conversions and Mortality

809

Table 2. Distribution of Conversion Patterns, Length of Follow-up, and Mortality Events by Baseline Vascular Access Type

Conversion Patterns

AVF No conversion AVF to catheter AVF to catheter AVF to catheter to AVF/AVG AVF to AVG AVF to AVG AVF to AVG to catheter AVF to AVG to AVF AVG No conversion AVG to catheter AVG to catheter AVG to catheter to AVF/AVG AVG to AVF AVG to AVF AVG to AVF to catheter AVG to AVF to AVG Catheter No conversion Catheter to AVF Catheter to AVF Catheter to AVF to AVG Catheter to AVF to catheter Catheter to AVG Catheter to AVG Catheter to AVG to AVF Catheter to AVG to catheter

Median Time to First Conversion (d)

Median Time From Conversion to End of Follow-up (d)*

No. of Patients (%)

Median Length of Follow-up (d)*

462 (77.6) 114 (19.1) 49 (43.0) 65 (57.0) 19 (3.2) 14 (73.7) 3 (15.8) 2 (10.5)

365 (225, 365) 365 (277, 365) 319 (181, 365) 365 (365, 365) 365 (365, 365) 365 (297, 365) 365 (365, 365) 365 (365, 365)

— 62 (12, 134) 66 (15, 177) 44 (10, 119) 86 (25, 154) 63 (25, 142) 105 (86, 156) 151 (9, 293)

— 223 (122, 322) 151 (81, 244) 259 (183, 346) 260 (150, 338) 248 (150, 338) 260 (209, 279) 214 (72, 356)

46 13 9 4 2 2 0 0

627 (78.4) 164 (20.5) 76 (46.3) 88 (53.7) 9 (0.1) 4 (44.4) 2 (22.2) 3 (11.1)

365 (216, 365) 365 (316, 365) 365 (225, 365) 365 (359, 365) 365 (365, 365) 365 (365, 365) 365 (365, 365) 365 (264, 365)

— 84 (29, 173) 115 (37, 215) 71 (26, 134) 160 (120, 237) 184 (118, 287) 165 (160, 169) 122 (30, 237)

— 210 (117, 304) 151 (51, 229) 263 (152, 319) 196 (128, 245) 181 (79, 247) 201 (196, 205) 142 (128, 335)

77 21 11 10 0 0 0 0

1,712 (54.6) 611 (19.5) 397 (65.0) 11 (1.8) 203 (33.2) 814 (25.9) 579 (71.1) 6 (0.07) 229 (27.38)

178 (70, 365) 365 (308, 365) 365 (293, 365) 365 (365, 365) 365 (320, 365) 365 (314, 365) 365 (298, 365) 365 (365, 365) 365 (339, 365)

— 92 (48, 168) 105 (58, 182) 52 (35, 123) 70 (42, 131) 66 (30, 131) 75 (33, 143) 31 (10, 49) 58 (25, 107)

— 221 (117, 295) 197 (95, 281) 279 (236, 317) 254 (165, 306) 246 (137, 317) 231 (122, 315) 334 (316, 355) 277 (166, 324)

459 56 41 0 15 103 77 0 26

No. of Deaths

Note: Values shown as number (25th and 75th percentiles). Abbreviations: AVF, arteriovenous fistula; AVG, arteriovenous graft. *End of follow up was defined as the first of the following: date of death; loss to follow-up for reasons including transplantation, switch to peritoneal dialysis, or withdrawal; or 365 days.

lar access in the first year. Catheter patients who converted to a permanent vascular access were considerably different compared with those who did not, particularly with respect to predialysis care, age, race, and degree of comorbidity. After accounting for these differences, we observed a nearly 30% lower risk of death for patients who converted to a permanent vascular access versus those who did not. Of the 30% of patients initiating dialysis therapy with a permanent vascular access, 20% experienced a conversion in the first year. In this group, few differences were noted between converters and nonconverters, and conversion to a catheter was associated with an 80% increased risk of death, although the mortality risk was greater in patients who converted from an AVG to a catheter. We then analyzed conver-

sions from a catheter to a permanent vascular access according to the conversion practices within each dialysis facility and observed similar mortality HR estimates, suggesting that the potential benefits of conversion at the individual patient level exist irrespective of whether the facility where they are receiving dialysis is more or less aggressive in converting catheters to permanent accesses. Accumulating evidence suggests that the type of vascular access in use at HD initiation is strongly related to future infectious complications (eg, bacteremia),15 central vein stenosis,3-7 and mortality risk.8-14 Studies have shown that patients initiating dialysis therapy with a catheter have a markedly different demographic and comorbidity profile compared with those initiating therapy with a perma-

810

Bradbury et al

Table 3. Distribution of Baseline Patient Characteristics by Conversion Status Stratified by Baseline Vascular Access Type AVF (n ⫽ 576)* Nonconverter

Converter

Demographics Age (y) 60.1 ⫾ 16.2 62.6 ⫾ 15.0 27.1 ⫾ 7.2 27.3 ⫾ 7.4 Body mass index (kg/m2) Sex (% men) 73.8 70.2 Race (% white) 68.0 73.7 Laboratory Albumin (g/dL) 3.5 ⫾ 0.5 3.6 ⫾ 0.5 Hemoglobin (g/dL) 10.2 ⫾ 1.7 10.0 ⫾ 1.5 Total calcium (mg/dL) 8.4 ⫾ 1.0 8.4 ⫾ 0.9 Phosphorus (mg/dL) 5.8 ⫾ 1.7 6.0 ⫾ 1.9 White blood cell count 7.5 ⫾ 2.6 7.7 ⫾ 2.6 (⫻103/␮L) Other Diabetes as cause of ESRD 42.0 46.5 ⱖ1 Predialysis vascular access placements 88.1 90.4 Medicare as primary payer 56.9 63.2 Diastolic blood pressure† (mm Hg) 80.5 ⫾ 15.6 76.2 ⫾ 15.6 Systolic blood pressure† (mm Hg) 154.3 ⫾ 25.7 152.1 ⫾ 24.8 Visited a nephrologist‡ 77.7 86.8 Comorbidity (% yes) Coronary artery disease 44.6 50.9 Cancer 12.1 9.6 Other cardiovascular 22.9 15.8 Cerebrovascular disease 13.9 10.5 Heart failure 32.3 37.7 Diabetes 46.1 50 Gastrointestinal bleeding 7.1 10.5 HIV/AIDS 1.3 0.9 Hypertension 83.5 90.4 Lung disease 12.8 14.9 Neurological disease 8.2 8.8 Psychiatric disorder 21.6 25.4 Peripheral vascular disease 19.7 21.9 Recurrent cellulitis 6.1 7.0

AVG (n ⫽ 791)*

Catheter (n ⫽ 3,137)

P

Nonconverter

Converter

P

Nonconverter

0.1 0.9 0.4 0.2

65.2 ⫾ 14 28 ⫾ 7.9 44.8 61.4

63.5 ⫾ 14 28.1 ⫾ 8.4 46.3 62.8

0.2 0.9 0.7 0.7

63.3 ⫾ 16.4 26.7 ⫾ 7.2 55.1 70.3

0.3 0.5 0.5 0.4

3.5 ⫾ 0.5 10.0 ⫾ 1.5 8.5 ⫾ 0.9 5.4 ⫾ 1.7

3.4 ⫾ 0.5 9.8 ⫾ 1.6 8.3 ⫾ 0.9 5.4 ⫾ 1.7

0.03 0.08 0.06 0.8

3.3 ⫾ 0.6 10.1 ⫾ 1.6 8.5 ⫾ 0.9 5.4 ⫾ 1.8

3.3 ⫾ 0.5 9.8 ⫾ 1.6 8.4 ⫾ 0.9 5.5 ⫾ 1.8

0.01 ⬍0.001 0.02 0.07

0.4

8.0 ⫾ 3.1

8.1 ⫾ 2.9

0.7

8.8 ⫾ 3.6

8.7 ⫾ 3.3

0.3

0.3

53.4

57.3

0.4

46.0

50.1

0.02

0.8 0.2

85.8 63.6

83.5 59.1

0.6 0.5

61.6 57.0

76.5 57.6

⬍0.001 0.6

0.01

74.6 ⫾ 14.4

75.7 ⫾ 16.5

77.2 ⫾ 16.0

0.02

0.5 0.4 0.3 0.4 0.09 0.3 0.3 0.5 0.2 0.7 0.09 0.6 0.9 0.4 0.6 0.7

77.2 ⫾ 15.2 0.06

149.9 ⫾ 26.9 152.4 ⫾ 25.7 0.3 69.9 76.8 0.8 52.0 13.2 24.6 20.3 43.2 58.5 6.4 0.5 84.4 11.0 8.5 22.8 25.0 7.3

43.3 16.5 28.0 20.1 45.1 64.6 8.5 0.0 86.0 12.8 9.8 22.6 26.8 8.5

0.05 0.3 0.4 0.9 0.7 0.2 0.3 0.4 0.6 0.5 0.6 0.9 0.6 0.6

Converter

P

61.5 ⫾ 15.5 0.003 27.6 ⫾ 7.3 0.002 57.1 0.3 59.6 ⬍0.001

144.6 ⫾ 27.2 149.4 ⫾ 26.0 ⬍0.001 45.6 56.6 ⬍0.001 55.8 13.6 33.9 20.9 50.8 51.6 7.1 1.9 80.4 18.0 11.5 29.7 31.0 9.5

50.8 11.5 29.7 17.3 49.1 55.2 8.6 1.0 86.5 15.2 9.0 27.2 26.4 8.4

0.003 0.06 0.007 0.007 0.3 0.1 0.1 0.05 ⬍0.001 0.02 0.02 0.09 0.003 0.2

Note: Values expressed as mean ⫾ SD or percent. Conversion factors for units: albumin in g/dL to g/L, ⫻10; calcium in mg/dL to mmol/L, ⫻0.2495; hemoglobin in g/dL to g/L, ⫻10; phosphorus in mg/dL to mmol/L, ⫻0.3229. Abbreviations: AIDS, acquired immunodeficiency syndrome; AVF, arteriovenous fistula; AVG, arteriovenous graft; ESRD, end-stage renal disease; HIV, human immunodeficiency virus. *Includes only participants who converted to a catheter. †Predialysis. ‡Visited nephrologist more than 1 month before initiating dialysis therapy.

nent vascular access9,12,13,24 and are less likely to have received care from a nephrologist before dialysis therapy initiation.13,21-23 These differences were apparent in our population. The comorbidity burden (heart failure, cerebrovascular disease, lung disease, and so on) in patients initiating dialysis therapy with a catheter was considerably greater than for patients with a permanent vascular access. Moreover, patients using a permanent vascular access were more likely to have had predialysis

nephrologist care and previous permanent vascular access placement.All these factors have been shown to strongly predict mortality in the first year on dialysis therapy.13,14 Although 70% of patients initiate dialysis therapy with a catheter, only 27% of prevalent patients are dialyzing with a catheter,16 which highlights the ongoing efforts to place and convert patients to permanent vascular accesses. It also may suggest that patients who do not have a

Vascular Access Conversions and Mortality

811

Table 4. HR Estimates for Associations Between Vascular Access Conversion Patterns and Subsequent Mortality During the First Year on Dialysis Therapy Conversion Pattern by Baseline Vascular Access Type

Catheter Catheter to permanent vascular access† Catheter to AVF† Catheter to AVG† Permanent vascular access Permanent to catheter‡ AVF to catheter§ AVG to catheter储

Crude HR (95% CI)

Adjusted HR (95% CI)*

0.56 (0.46-0.69) 0.49 (0.37-0.65) 0.61 (0.48-0.77)

0.69 (0.55-0.85) 0.64 (0.47-0.87) 0.71 (0.55-0.92)

1.63 (1.11-2.38) 1.64 (0.95-2.83) 1.63 (0.99-2.68)

1.81 (1.22-2.68) 1.55 (0.84-2.87) 2.11 (1.21-3.67)

Abbreviations: AVF, arteriovenous fistula; AVG, arteriovenous graft; CI, confidence interval; HR, hazard ratio. *Models were adjusted for age, sex, race, body mass index, primary cause of end-stage renal disease, Medicare as primary payer, systolic and diastolic blood pressure, previous vascular access placement, predialysis nephrologist care; albumin, hemoglobin, calcium, phosphorus, and white blood cell count levels; 13 comorbid conditions measured at dialysis therapy initiation (HIV/AIDS was excluded because of the large number of missing values); and calendar year. †Reference group is patients who do not convert from a catheter. ‡Reference group is patients who do not convert from a permanent access. §Reference group is patients who do not convert from an AVF. 储Reference group is patients who do not convert from an AVG.

permanent vascular access placed or who are not converted represent a subpopulation with greater disease burden and worse prognosis. It therefore is critical to account for these patient differences when estimating the effects of conversion. Fortunately, DOPPS is unique in its extensive collection of patient information, enabling improved

Figure 2. Crude and adjusted mortality hazard ratio (HR) and 95% confidence interval estimates for conversion to a permanent vascular access by 120 days, overall and stratified by quartiles of the percentage of patients converted at the facility level. *Models were adjusted for age, sex, race, body mass index, primary cause of endstage renal disease, Medicare as primary payer, systolic and diastolic blood pressure, previous vascular access placement, predialysis nephrologist care; albumin, hemoglobin, calcium, phosphorus, and white blood cell count levels; 13 comorbid conditions measured at dialysis therapy initiation (human immunodeficiency virus [HIV]/acquired immunodeficiency syndrome [AIDS] was excluded because of the large number of missing values); and calendar year.

control for confounding. However, even after considerable statistical adjustment, the observed protective effect of conversion persisted. Importantly, this adjustment was limited to baseline patient characteristics and did not account for differences in facility practices that might influence the likelihood that a patient would be converted to a permanent vascular access (eg, access to better trained vascular access surgeons25). We attempted to address this issue by examining the effect of conversion for different groups of facilities based on their observed ability to convert patients. The idea was that facilities with many converted patients have targeted practices designed to enable patients to begin dialyzing with a permanent vascular access. In those facilities compared with facilities with fewer converted patients, we would expect to see less of a benefit of conversion if conversion was simply a proxy for prognosis. That is, patients with a poorer prognosis would be converted in facilities that target greater permanent vascular access use compared with selective patient conversions dictated by better prognosis in facilities with lower conversion rates. However, we observed similar HR estimates across facility groupings, suggesting that the effect persists regardless of facility conversion practices. In patients who initiated dialysis therapy with a permanent vascular access in use, we observed increased mortality risk for those who subse-

812

quently converted to a catheter. The finding that patients switching from an AVG to a catheter have greater mortality compared with those switching from an AVF may have its basis in the selection bias associated with the initial vascular access type. That is, patients who were originally selected to receive an AVG rather than an AVF may have had worse vascular status and/or underlying comorbidity (eg, longer exposure to diabetes) that precluded placement of an AVF initially. These same factors that may be influencing the initial permanent access selection may also help explain the greater subsequent mortality risk associated with conversion from an AVG to a catheter. Recently, Allon et al26 evaluated the effect of vascular access conversions on subsequent mortality burden by using randomized clinical trial data from the Hemodialysis (HEMO) Study.27 They related mortality to the vascular access type in use during 2 consecutive 6-month periods and found that persistent catheter use was associated with a markedly greater mortality risk (HR, 3.43) compared with persistent AVF use. However, this study included a selected population of prevalent dialysis patients in which only 5% were dialyzing with a catheter. More recently, Wasse et al28 used Clinical Performance Measures and USRDS data to examine conversions from a catheter to a permanent vascular access in incident HD patients and found similar differences between converters and nonconverters, particularly with respect to race and sex. Findings from our analysis are consistent with those from a recent study by Wang et al29 that sought, but was unable to identify, predictors of AVF failure in a population of incident Canadian dialysis patients. Neither study evaluated other factors that might influence permanent vascular access survival, such as anticoagulant use, which was recently shown to significantly reduce early AVF thrombosis.30 Future studies of this topic should attempt to collect and examine data for concomitant medication use, as well as surgical and radiological interventions,31 factors that may influence permanent vascular access patency and, consequently, conversions. However, to our knowledge, ours is the first study to comprehensively describe the myriad of patterns of vascular access conversions in the HD population, specifi-

Bradbury et al

cally in incident patients, and relate those conversions to mortality. This study has the following limitations. First, we limited our analysis to incident US HD patients because of the high prevalence of catheter use in this subpopulation. Patterns and time to conversion for the various vascular access types therefore should not be extrapolated to the prevalent population, a potentially healthier subpopulation. Second, data for permanent vascular access placements after dialysis therapy initiation were not reliably recorded by all facilities. Thus, we were unable to determine definitively when patients had a permanent vascular access placed after being on dialysis therapy and therefore would be eligible for future conversion. This could have introduced a selection bias that would make conversion appear beneficial, even if it was not. To evaluate whether this potential for misclassification was biasing our estimates, we conducted an additional analysis limited to only patients who had a predialysis permanent vascular access placement and observed similar estimates, providing some reassurance that this did not explain our findings. Third, previous studies suggest that AVF and AVG placements may be particularly difficult in specific patient subgroups, such as patients with diabetes and women, because of inherent differences in vasculature.28,29,32-35 Because DOPPS did not collect data for vein mapping or accessibility, it was not possible to determine whether patients were simply not eligible for permanent vascular access placement as a result of their vascular anatomy. However, our stratified analyses did not suggest a difference in the effect of conversion for different patient subgroups, including those historically identified as having anatomic barriers to vascular access placement. Fourth, we did not evaluate the impact of events occurring subsequent to dialysis therapy initiation that might influence the likelihood of conversion and also predict future mortality risk, including hospitalization for sepsis or infection, worsening underlying cardiovascular disease, or malnutrition. Because we allowed conversions to occur at any point in time during the first year and did not group them into regular intervals (eg, assessing conversions monthly), evaluating the influence of intercurrent hospitalization events was not feasible. Thus, the observed associations

Vascular Access Conversions and Mortality

could be caused by unmeasured confounding. Future analyses should investigate the influence of these time-dependent events on the relationship between vascular access conversion and mortality risk. Finally, one could postulate that such subjective factors as decisions regarding the maturity of the permanent access, appropriate time to cannulate, and expertise of the individual performing the cannulation might all influence conversion decisions and may differ from one facility to another. We attempted to control for these unmeasured factors by stratifying on facility conversion practices and observed similar effect estimates across facilities. In recent years, despite the steadily decreasing mortality rates in prevalent dialysis patients, the mortality rate in the incident patient population has not decreased. Placement of permanent vascular accesses before dialysis therapy initiation and greater attention to converting patients initiating dialysis therapy with catheters to more sustainable permanent accesses may help improve the survival probability of incident dialysis patients.

ACKNOWLEDGEMENTS Support: The DOPPS has been supported by research grants from Amgen and Kirin Pharma with no restrictions on publication. Financial Disclosure: Dr Bradbury, Mrs Chen, Dr Keen, and Dr Krishnan are employees at Amgen Inc.

REFERENCES 1. No author listed: http://www.fistulafirst.org/“Change Concepts”. Accessed September 29, 2008 2. Feldman HI, Held PJ, Hutchinson JT, Stoiber E, Hartigan MF, Berlin JA: Hemodialysis vascular access morbidity in the United States. Kidney Int 43:1091-1096, 1993 3. Nassar GM, Ayus JC: Infectious complications of the hemodialysis access. Kidney Int 60:1-13, 2001 4. Taylor GD, McKenzie M, Buchanan-Chell M, Caballo L, Chui L, Kowalewska-Grochowska K: Central venous catheters as a source of hemodialysis-related bacteremia. Infect Control Hosp Epidemiol 19:643-646, 1998 5. Allon M: Dialysis catheter-related bacteremia: Treatment and prophylaxis. Am J Kidney Dis 44:779-791, 2004 6. Ribot S, Siddiqi SW, Chen C: Right heart complications of dual lumen tunneled venous catheters in hemodialysis patients. Am J Med Sci 330:204-208, 2005 7. Dhingra RK, Young EW, Hulbert-Shearon TE, Leavey SF, Port FK: Type of vascular access and mortality in US hemodialysis patients. Kidney Int 60:1443-1451, 2001 8. Pastan S, Soucie JM, McClellan WM: Vascular access and increased risk of death among hemodialysis patients. Kidney Int 62:620-626, 2002

813 9. Xue JL, Dahl D, Ebben JP, Collins AJ: The association of initial hemodialysis access type with mortality outcomes in elderly Medicare ESRD patients. Am J Kidney Dis 42:1013-1019, 2003 10. Rayner HC, Besarab A, Brown WW, Disney A, Saito A, Pisoni RL: Vascular access results from the Dialysis Outcomes and Practice Patterns Study (DOPPS): Performance Against Kidney Disease Outcomes Quality Initiative (K/DOQI) Clinical Practice Guidelines. Am J Kidney Dis 44:S22-S26, 2004 (suppl 2) 11. Polkinghorne KR, McDonald SP, Atkins RC, Kerr PG: Vascular access and all-cause mortality: A propensity score analysis. J Am Soc Nephrol 15:477-486, 2004 12. Astor BC, Eustace JA, Powe NR, Klag MJ, Fink NE, Coresh J: Type of vascular access and survival among incident hemodialysis patients: The Choices for Healthy Outcomes in Caring for ESRD (CHOICE) Study. J Am Soc Nephrol 16:1449-1455, 2005 13. Lorenzo V, Martin M, Rufino M, Hernandez D, Torres A, Ayus JC: Predialysis nephrologic care and a functioning arteriovenous fistula at entry are associated with better survival in incident hemodialysis patients: An observational cohort study. Am J Kidney Dis 43:999-1007, 2004 14. Bradbury BD, Fissell RB, Albert JM, et al: Predictors of early mortality among incident United States hemodialysis patients in the Dialysis Outcomes and Practice Patterns Study (DOPPS). Clin J Am Soc Nephrol 2:89-99, 2007 15. US Renal Data System: USRDS 2006 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 2006 16. Centers for Medicare & Medicaid Services: 2006 Annual Report, End Stage Renal Disease Clinical Performance Measures Project. Baltimore, MD, Department of Health and Human Services, Centers for Medicare & Medicaid Services, Office of Clinical Standards & Quality, 2007 17. Pisoni RL, Gillespie BW, Dickinson DM, Chen K, Kutner MH, Wolfe RA: The Dialysis Outcomes and Practice Patterns Study (DOPPS): Design, data elements, and methodology. Am J Kidney Dis 44:S7-S15, 2004 (suppl 3) 18. Young EW, Goodkin DA, Mapes DL, et al: The Dialysis Outcomes and Practice Patterns Study (DOPPS): An international HD study. Kidney Int Suppl 74:S74-S81, 2000 19. Klein JP, Moeschberger ML: Survival Analysis Techniques for Censored and Truncated Data. New York, NY, Springer, 1997, p 417 20. Johnston SC, Henneman T, McCulloch CE, van der Laan M: Modeling treatment effects on binary outcomes with grouped-treatment variables and individual covariates. Am J Epidemiol 156:753-760, 2002 21. Astor BC, Eustace JA, Powe NR, et al: Timing of nephrologist referral and arteriovenous access use: The CHOICE Study. Am J Kidney Dis 38:494-501, 2001 22. Stehman-Breen CO, Sherrard DJ, Gillen D, Caps M: Determinants of type and timing of initial permanent hemodialysis vascular access. Kidney Int 57:639-645, 2000 23. Avorn J, Winkelmayer WC, Bohn RL, et al: Delayed nephrologist referral and inadequate vascular access in patients with advanced chronic kidney failure. J Clin Epidemiol 55:711-716, 2002

814 24. Saran R, Elder SJ, Goodkin DA, et al: Enhancing training in creation of vascular accesses predicts arteriovenous fistula placement and patency in hemodialysis patients. Ann Surg 247:885-891, 2008 25. Chand DH, Teo BW, Fatic RA, Brier M; Medical Review Board of The Renal Network Inc: Influence of vascular access type on outcome measures in patients on maintenance hemodialysis. Nephron Clin Practice 108:c91c98, 2008 26. Allon M, Daugirdas J, Depner TA, Greene T, Ornt D, Schwab SJ: Effect of change in vascular access on patient mortality in hemodialysis patients. Am J Kidney Dis 47:469477, 2006 27. Eknoyan G, Beck GJ, Cheung AK, et al: Effect of dialysis dose and membrane flux on mortality and morbidity in maintenance hemodialysis patients: Primary results of the HEMO study. N Engl J Med 347;2010-2019, 2002 28. Wasse H, Speckman RA, Frankenfield DL, Rocco MV, McClellan WM: Predictors of delayed transition from central venous catheter use to permanent vascular access among ESRD patients. Am J Kidney Dis 49:276-283, 2007 29. Wang W, Murphy B, Yilmaz S, Tonelli M, MacRae J, Manns BJ: Comorbidities do not influence primary fistula success in incident hemodialysis patients: A prospective study. Clin J Am Soc Nephrol 3:78-84, 2008

Bradbury et al 30. Dember LM, Beck GJ, Allon M, et al, for the Dialysis Access Consortium Study Group: Effect of clopidogrel on early failure of arteriovenous fistulas for hemodialysis: A randomized controlled trial. JAMA 299:21642171, 2008 31. Karakayali F, Basaran O, Ekici Y, et al: Effect of secondary intervention on patency of vascular access sites for hemodialysis. Eur J Vasc Endovasc Surg 32:701-709, 2006 32. Woods JD, Turenne MN, Strawderman RL, et al: Vascular access survival among incident hemodialysis patients in the United States. Am J Kidney Dis 30:50-57, 1997 33. Windus DW: The effect of comorbid conditions on hemodialysis access patency. Adv Ren Replace Ther 1:148154, 1994 34. Mendes RR, Farber MA, Marston WA, Dinwiddie LC, Keagy BA, Burnham SJ: Prediction of wrist arteriovenous fistula maturation with preoperative vein mapping with ultrasonography. J Vasc Surg 36:460-463, 2002 35. Astor BC, Coresh J, Powe NR, Eustace JA, Klag MJ: Relation between gender and vascular access complications in hemodialysis patients. Am J Kidney Dis 36:1126-1134, 2000