Association Between Vascular Access Creation and Deceleration of Estimated Glomerular Filtration Rate Decline in Late-Stage Chronic Kidney Disease Patients Transitioning to End-Stage Renal Disease

Keiichi Sumida; Miklos Z. Molnar; Praveen K. Potukuchi; Fridtjof Thomas; Jun Ling Lu; Vanessa A. Ravel; Melissa Soohoo; Connie M. Rhee; Elani Streja; Kunihiro Yamagata; Kamyar Kalantar-Zadeh; Csaba P. Kovesdy


Nephrol Dial Transplant. 2017;32(8):1330-1337. 

In This Article

Materials and Methods

Cohort Definition

We analyzed data from the Transition of Care in CKD (TC-CKD) study, a retrospective cohort study examining US veterans transitioning to dialysis from 1 October 2007 through 30 September 2011.[17] A total of 52 172 US veterans were identified from the US Renal Data System (USRDS)[1] as an initial cohort. The algorithm for the cohort definition is shown in Figure 1. At first, patients without any vascular access procedure codes, as identified by the International Classification of Diseases, Ninth Revision Clinical Modification (ICD-9-CM) procedure codes and Current Procedural Terminology (CPT) codes (Supplementary data, Table S1), were excluded (n = 18 711). Of the remaining 33 461 patients with vascular access procedure codes, 11 574 patients with AVF/AVG creation procedure codes prior to dialysis initiation were identified. In order to quantify the trajectory (slope) of eGFR over time, we used outpatient serum creatinine measurements available from Veterans Affairs (VA) medical centers because of the potential fluctuation of serum creatinine levels among hospitalized patients. Therefore, patients without serum creatinine measurement(s) in the VA medical system or those with only inpatient serum creatinine measurement(s) were excluded (n = 3448). Patients were also excluded if they had fewer than three outpatient serum creatinine measurements either before the AVF/AVG creation or during the interval between the AVF/AVG creation and dialysis initiation (n = 4922). The final cohort consisted of 3026 patients with an AVF/AVG (Figure 1). We also identified 3514 patients without a pre-dialysis AVF/AVG creation as comparators, who started dialysis with a tunneled catheter and had at least three outpatient serum creatinine measurements both before and after the 6-month index date prior to dialysis initiation (Figure 1).

Figure 1.

Flow chart of the study population. AVF, arteriovenous fistula; AVG, arteriovenous graft; eGFR, estimated glomerular filtration rate; VA, Veterans Affairs.

Data Collection

Data from the USRDS Patient and Medical Evidence files were used to determine baseline demographic characteristics and type of vascular access at the time of dialysis initiation. Laboratory variables including serum creatinine were collected as previously described.[18,19] Baseline values for laboratory variables (except for serum creatinine) were defined as the last quarterly average of each variable before dialysis initiation or the second quarterly average from the last if the last one was not available. Data regarding medication exposure were obtained from both Centers for Medicare & Medicaid Services (CMS) Data (Medicare Part D) and VA pharmacy dispensation records.[20] Patients who received at least one dispensation of outpatient medications within 6 months prior to dialysis initiation were recorded as having been treated with these medications. Information on angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin receptor blockers (ARBs) use and systolic BP were collected for the entire evaluation period as time-dependent variables. Information about vascular access procedures and comorbidities was extracted from the VA Inpatient and Outpatient Medical SAS Datasets,[21] using the ICD-9-CM diagnostic and procedure codes and CPT codes, as well as from VA/CMS data. Cardiovascular disease was defined as the presence of diagnostic codes for coronary artery disease, angina, myocardial infarction or cerebrovascular disease. We calculated the Charlson comorbidity index score using the Deyo modification for administrative data sets, without including kidney disease.[22] eGFR was calculated based on the Chronic Kidney Disease Epidemiology Collaboration equation,[23] from all available outpatient serum creatinine measurements starting not more than 7 years before dialysis initiation.

Statistical Analysis

Data are presented as number (percent) for categorical variables and mean ± standard deviation or median [interquartile range (IQR)] as appropriate. Continuous variables were compared using t tests or Mann–Whitney U tests as appropriate. Categorical variables were analyzed with χ2 test. Progression of CKD was assessed by estimating slopes of eGFR (annual change in eGFR) from mixed-effects models with random intercepts and slopes using the XTMIXED command in STATA.[24] This model estimates the slopes of eGFR over time, taking into account the varying number and spacing of eGFR measurements, as well as the variable follow-up for each subject.[24] The effect of potential confounders on eGFR slopes was analyzed in an adjusted multilevel mixed-effects model, which included fixed (age, sex, race, diabetes mellitus and Charlson comorbidity index) and time-dependent variables (systolic BP and ACEIs/ARBs use).

In order to assess change in eGFR slopes, we defined a cutoff time point as the date of the pre-dialysis AVF/AVG procedure for patients who underwent such a procedure. Among patients without an AVF/AVG, we used an index date of 6 months prior to dialysis initiation, which corresponded to the median time interval from AVF/AVG creation to dialysis initiation. eGFR slopes were calculated separately for these two time periods in each patient and compared using paired t tests. To account for potential differences in the effect of AVF/AVG on eGFR slopes by AVF/AVG maturation status, we stratified patients with an AVF/AVG into two groups according to whether or not an AVF/AVG was used as the primary vascular access at the time of dialysis initiation (per USRDS records). Analyses were repeated in a propensity score-matched cohort to account for dissimilarities in clinical characteristics between the groups with and without an AVF/AVG, including pre-AVF/AVG or pre-index date eGFR slopes and eGFR levels at the time of AVF/AVG creation or index date.

The changes in eGFR slope were also examined separately for AVF and AVG and in subgroups of patients categorized by age, race, body mass index, and the presence of diabetes mellitus or cardiovascular disease. Of the variables included in the multivariable-adjusted mixed-effects model, none was missing in the final cohort. Analyses were conducted using STATA MP Version 14 (STATA Corporation, College Station, TX, USA). The study was approved by the Institutional Review Boards of the Memphis and Long Beach VA Medical Centers, with exemption from informed consent.