Author + information
- Received July 22, 2015
- Revision received September 11, 2015
- Accepted October 20, 2015
- Published online March 29, 2016.
- Tara I. Chang, MD, MSa,∗ (, )
- Maria E. Montez-Rath, PhDa,
- Thomas T. Tsai, MD, MScb,c,
- Mark A. Hlatky, MDa and
- Wolfgang C. Winkelmayer, MD, MPH, ScDd
- aDepartment of Medicine, Division of Nephrology, Stanford University School of Medicine, Stanford, California
- bInterventional Cardiology & Institute for Health Research, Kaiser Permanente Colorado, Denver, Colorado
- cColorado Cardiovascular Outcomes Research Consortium, University of Colorado, Denver, Colorado
- dDepartment of Medicine, Section of Nephrology, Baylor College of Medicine, Houston, Texas
- ↵∗Reprint requests and correspondence:
Dr. Tara I. Chang, 777 Welch Road, Suite DE, Palo Alto, California 94304.
Background In patients undergoing percutaneous coronary intervention (PCI), drug-eluting stents (DES) reduce repeat revascularizations compared with bare-metal stents (BMS), but their effects on death and myocardial infarction (MI) are mixed. Few studies have focused on patients with end-stage renal disease.
Objectives This study compared mortality and cardiovascular morbidity during percutaneous coronary intervention with DES and with BMS in dialysis patients.
Methods We identified 36,117 dialysis patients from the USRDS (United States Renal Data System) who had coronary stenting in the United States between April 23, 2003, and December 31, 2010, and examined the association of DES versus BMS with 1-year outcomes: death; death or MI; and death, MI, or repeat revascularization. We also conducted a temporal analysis by dividing the study period into 3 DES eras: Transitional (April 23, 2003, to June 30, 2004); Liberal (July 1, 2004, to December 31, 2006); and Selective (January 1, 2007, to December 31, 2010).
Results One-year event rates were high, with 38 deaths; 55 death or MI events; and 71 death, MI, or repeat revascularization events per 100 person-years. DES, compared with BMS, were associated with a significant 18% lower risk of death; 16% lower risk of death or MI; and 13% lower risk of death, MI, or repeat revascularization. DES use varied, from 56% in the Transitional era to 85% in the Liberal era and 62% in the Selective era. DES outcomes in the Liberal era were significantly better than in the Transitional Era, but not significantly better than in the Selective Era.
Conclusions DES for percutaneous coronary intervention appears to be safe for use in U.S. dialysis patients and is associated with lower rates of death, MI, and repeat revascularization.
- cardiovascular disease
- end-stage renal disease
- ischemic heart disease
- renal insufficiency
Coronary heart disease affects 30% to 60% of patients with end-stage renal disease (ESRD) on dialysis (1–3), and the number of such patients undergoing percutaneous coronary intervention (PCI) has increased by nearly 50% over the past decade (4). Numerous clinical trials in patients with preserved kidney function demonstrate that drug-eluting stents (DES) reduce repeat revascularizations compared with bare-metal stents (BMS), but effects on the risk of myocardial infarction (MI) and mortality are less consistent (5,6). A meta-analysis of 76 randomized clinical trials comparing DES and BMS showed that short-term risk of MI was reduced in DES relative to BMS, but with no benefit on short-term mortality (5). Pooled analysis of data from 3 randomized trials showed similar results consistently across different levels of kidney function, although patients with ESRD were excluded (6).
Current guidelines recommend using DES rather than BMS in patients with ESRD (7), based on extrapolation of data from patients with normal or near-normal kidney function. However, uremia, inflammation, or dialysis itself may lead patients with ESRD to have more diffuse coronary disease and more vascular calcification (1), making them more prone to post-procedural complications than are patients without ESRD (8). There has been only 1 randomized clinical trial directly comparing DES with BMS that did not explicitly exclude patients with ESRD on dialysis, and it showed lower rates of target vessel revascularization at 12 months with DES (9). Only 22 participants in that trial had ESRD, and 1-year mortality was much lower than the national annual death rate. Those results may therefore not be generalizable to the overall ESRD population. Observational studies have yielded mixed results; some show a benefit of DES over BMS (8,10), whereas others do not (11–17). Many of the studies involved small cohorts, were in largely Asian populations, or did not adequately adjust for potential confounders—perhaps accounting for some of the heterogeneity in outcomes.
Despite the paucity of evidence, use of DES rose rapidly in ESRD patients after approval of the first DES in the United States in 2003 (10,18). When reports suggested higher rates of MI, cardiovascular mortality, and stent thrombosis with DES versus BMS in 2006 (19–24), use of DES in patients with ESRD declined markedly until 2007, after which the use of DES stabilized following publication of studies showing longer-term benefit and safety of DES (25).
We studied patients with ESRD on maintenance dialysis undergoing PCI with stenting, hypothesizing that DES, compared with BMS, would be associated with lower risks of death, MI, and repeat revascularization. Due to concerns about residual confounding by indication in this study, we leveraged changing patterns of DES use in a temporal analysis that compared outcomes among 3 DES eras (26): Transitional (April 24, 2003, to June 30, 2004); Liberal (July 1, 2004, to December 31, 2006); and Selective (January 1, 2007, to December 31, 2010). We hypothesized that patients undergoing PCI during the Liberal DES era would have better outcomes than patients in the 2 eras with more restrictive DES use.
Data on all fee-for-service claims from Medicare parts A and B were linked to the USRDS (United States Renal Data System), the national registry of patients with ESRD (27). As mandated in the Social Security Amendments of 1972 (28), almost all patients with ESRD qualify for federal health benefits through Medicare, irrespective of age or disability status.
Study patients were ≥18 years of age, had ESRD, and underwent PCI with stenting (ascertained from International Classification of Diseases-Ninth Edition procedure codes 36.00, 36.01, 36.02, 36.05, 36.06, 36.07, 36.09, or 00.66 reported in the institutional detail datasets) after approval of the first DES by the U.S. Food and Drug Administration on April 24, 2003 (Figure 1) and through December 31, 2010.
For patients with multiple PCI, the first recorded PCI was the index PCI. We required that patients have continuous Medicare parts A and B coverage as primary payer for at least 6 months prior to the index hospitalization admission date to allow determination of baseline comorbid conditions. Patients undergoing any type of heart surgery during the index hospitalization were excluded, as were patients who received both BMS and DES or who had no record of stent placement during the index PCI. All patients were required to be on maintenance dialysis during the index PCI, per USRDS-defined modality file (rxhist60).
Follow-up and outcomes
We examined 3 outcomes within 1 year after index PCI: 1) death from any cause; 2) death or MI; and 3) death, MI, or repeat revascularization (Online Table 1). Death was determined from the USRDS. MI occurring during the index hospitalization was not considered an outcome, because it may have occurred prior to the revascularization. Because determination of MI and repeat revascularization required claims information, follow-up for composite outcomes was censored at the time of loss of Medicare parts A and B coverage. We conducted additional sensitivity analysis censoring patients at the time of loss of Medicare parts A and B coverage or kidney transplantation.
Data included age, sex, race, Hispanic ethnicity, dialysis modality (hemodialysis or peritoneal), and presumed cause of ESRD from the USRDS patient and treatment history files at the index date. We included time since first treatment for ESRD (i.e., dialysis “vintage”) as a covariate, given its association with mortality and other outcomes in this patient population (29). We also included kidney transplant waitlist status (obtained from the waitlist file), because patients with an upcoming kidney transplant may be more likely to receive a BMS, which requires shorter duration of dual antiplatelet therapy. Waitlist status is also an aggregate indicator of the overall health of the patient and associated with outcomes (30).
We defined comorbid conditions using International Classification of Diseases-Ninth Edition diagnosis and procedure codes requiring at least 1 inpatient or 2 outpatient encounters separated by ≥1 day using all available historical data prior to (but not including) the index date (Online Table 1) (18), an approach that yields less bias than fixed observation windows (31). We used all inpatient and outpatient physician billing claims included in the USRDS institutional claims detail and physician supplier datasets. To adjust for differences in health care utilization (18), we identified the number of non-nephrology outpatient visits, of hospitalized days, and of nursing home stays for 6 months prior to the index date. We also categorized patients into 1 of 9 U.S. census regions based on the zip code in which they first received ESRD treatment.
Baseline characteristics among patients who received BMS versus DES were compared using standardized differences (32), which are not influenced by sample size (32,33). A standardized difference of >10.0 represents meaningful imbalance between treatment groups.
We estimated exposure propensity scores (34) for each patient from a multivariable logistic regression model with receipt of BMS as the dependent variable that included as predictors all baseline variables listed in Table 1, plus census region. We then used the propensity scores in 2 ways. First, we applied a greedy matching algorithm (35) to tightly match 1 patient who received a BMS to 1 patient who received a DES (maximum difference in propensity scores between matched pairs = 0.1). We required that all matched pairs match exactly by index year (and thus did not include index year in the logistic regression model). Second, propensity scores were used to conduct inverse probability of treatment weighting (IPTW) estimation with stabilized weights (36,37). A strength of IPTW is that the entire cohort remains available for analysis and results are generalizable to the entire population from which the observed sample was derived, in contrast to propensity score matching, which necessarily excludes the unmatched portion of the cohort (32).
We estimated hazard ratios (HR) and corresponding 95% confidence intervals (CI) for outcomes of interest using proportional hazards regression models. Because all baseline variables were well balanced (i.e., standardized differences <10.0) in the IPTW and propensity score–matched cohorts (Table 1), we made no further adjustments to the models. In the IPTW analyses, we used robust standard errors. We tested the proportionality assumption using Schoenfeld residual plots.
Given the possibility of residual confounding by indication, we conducted an analysis leveraging the temporal trends in DES use during the study period. Rather than directly compare DES with BMS use, we divided the cohort into 3 distinct DES eras based on previous studies (26) and based empirically on the proportion of DES use observed in our cohort: Transitional (April 24, 2003, to June 30, 2004); Liberal (July 1, 2004, to December 31, 2006); and Selective (January 1, 2007, to December 31, 2010). We compared baseline characteristics in the Transitional and Selective eras with the Liberal era using standardized differences as was just discussed. Event rates among Transitional, Liberal, and Selective DES eras were estimated using the Kaplan-Meier method and compared using the log-rank estimate. We used proportional hazards regression to compare outcomes in the Transitional and Selective eras relative to the Liberal DES era, adjusting for the variables listed in Table 2. Although previous studies used enrollment year or rates of DES use as an instrumental variable (38,39), we felt that DES era did not satisfy 2 key assumptions of a formal instrumental variable analysis: 1) that DES era was associated with outcomes only through its association with receipt of DES or BMS; and 2) that the instrumental variable randomizes patients such that patients are similar with respect to measured and unmeasured variables across levels of the instrument.
The institutional review board of Stanford University approved the study. All analyses were conducted using SAS Enterprise Guide (version 6.1, SAS Institute, Cary, North Carolina).
Overall, 36,117 patients on maintenance dialysis underwent PCI with stenting between 2003 and 2010 and passed inclusion and exclusion criteria (Figure 1). Sixty-nine percent of the final cohort received a DES. Mean age was 64 years, and there was a high prevalence of diabetes, hypertension, and hyperlipidemia. Patients who received DES had higher prevalence of diabetes mellitus, but patients who received BMS more often had a previous history of MI and more often presented with a ST-segment elevation MI (Table 1).
DES versus BMS
In the logistic regression model on stent type (c = 0.6; Hosmer and Lemeshow goodness-of-fit test, p = 0.86) (Online Table 2), female sex, multivessel intervention, and presence of heart failure and diabetes mellitus were associated with higher odds of receiving a BMS versus DES. We matched 96% of BMS patients to a corresponding DES patient. All baseline variables were well balanced among patients receiving BMS and DES after propensity score matching and after applying the IPTW (Table 1).
One-year event rates were high, with 38 deaths; 55 death or MI events; and 71 death, MI, or repeat revascularization events per 100 person-years. Unadjusted rates of death; death or MI; and death, MI, or repeat revascularization were lower in DES patients than in BMS patients (Table 3). In the propensity score–matched cohort, DES, compared with BMS, was associated with an 18% HR (95% CI: 14% to 22%) lower risk of death; 16% HR (95% CI: 13% to 19%) lower risk of death or MI; and 13% HR (95% CI: 9% to 16%) lower risk of death, MI, or repeat revascularization (Central Illustration). These results were similar using the IPTW approach (Central Illustration) and were not materially changed in sensitivity analyses that censored follow-up at the time of kidney transplantation (data not shown).
Temporal analysis: transitional, liberal, and selective DES eras
Average use of DES during PCI changed significantly over the study period, from 56% in the Transitional era, to 85% in the Liberal era, and 62% in the Selective era (Figure 2). Patients who underwent PCI in the Transitional era were less likely to have a non–ST-segment elevation MI on index presentation or require multivessel coronary intervention and had fewer nursing home stays and lower prevalence of most comorbid conditions compared with patients undergoing PCI in the other 2 eras (Table 2).
In unadjusted analyses, patients who underwent PCI in the Liberal era tended to have lower rates of death; death or MI; and death, MI, or repeat revascularization than patients who underwent PCI in the Transitional or Selective eras (Table 3, Figure 3). After adjusting for baseline characteristics, outcomes for patients undergoing PCI in the Transitional era were consistently worse than for patients in the Liberal era, whereas the outcomes of PCI during the Liberal versus Selective eras were similar, with only the composite of death and MI being significantly worse in the Selective era (HR: 1.05; 95% CI: 1.01 to 1.09) (Figure 3). Results were not changed in sensitivity analyses that censored follow-up at the time of kidney transplantation (data not shown).
In this large, representative cohort of patients with ESRD on dialysis, use of DES rather than BMS during PCI was associated with 18% lower risk of death; 6% lower risk of death or MI; and 13% lower risk of death, MI, or repeat revascularization (Central Illustration). DES-associated reduction in need for repeat revascularization among patients with reduced kidney function was recently demonstrated in the RENAL-DES (Randomized Comparison of Xience V and Multi-Link Vision Coronary Stents in the Same Multivessel Patient with Chronic Kidney Disease) trial, which enrolled 215 patients with estimated creatinine clearance <60 ml/min and multivessel coronary disease to receive DES or BMS (9). Mean creatinine clearance was 47 ml/min, with 10% of the cohort (n = 22) on dialysis. Results from RENAL-DES showed that the incidence of ischemia-driven target vessel revascularization at 12 months was 8.7% lower in the DES group (p < 0.001). Differences were even larger for patients with a creatinine clearance <30 ml/min or who were on dialysis (BMS = 24.2% vs. DES = 3.1%; absolute risk reduction = 21.1%; p = 0.005). However, the patient cohort in this trial was highly selected, as reflected in its 1-year death rate of only 3.7%, which is much lower than the annual death rates for patients with chronic kidney disease and coronary heart disease overall (40). Our study extends the results of RENAL-DES to a population better reflecting patients treated in actual clinical practice.
Results from observational studies in ESRD examining outcomes of death and MI have been less consistent. Some smaller studies (sample size ranging from 74 to 505) in dialysis patients, conducted mostly in Asian cohorts, showed no differences in death with DES versus BMS (11–17), whereas 2 larger, U.S.-based observational studies in patients with ESRD (8,10) showed benefit associated with DES, consistent with our results. One was a descriptive analysis using USRDS data that showed lower crude 1-year survival rates for patients on dialysis receiving a BMS (63%) than for patients receiving DES (71%) (10). The other included only older patients undergoing PCI and showed a 15% (95% CI: 6% to 14%) lower risk of death with DES versus BMS in the subgroup of dialysis patients (8). Our results expand upon those findings by adjusting for potential confounders using propensity score matching and IPTW analyses and by studying a more diverse patient cohort.
Patients with ESRD on dialysis are at higher risk of serious bleeding events due to chronic heparin exposure, uremia-induced platelet dysfunction, and concomitant use of anticoagulants (41–43). Such patients are also more likely to discontinue clopidogrel or other antiplatelet agents prematurely (44), which can lead to in-stent thrombosis and subsequent MI (45). Concerns about adverse bleeding effects or medication nonadherence may lead physicians to choose BMS for sicker, less adherent patients because DES requires a longer duration of dual antiplatelet therapy. We used different statistical approaches to account for these factors (i.e., propensity score matching and IPTW), but as with any observational study comparing 2 treatment options, residual confounding remained a concern. Rather than directly compare outcomes associated with DES versus BMS, we leveraged changing provider preferences for DES versus BMS during the study period and conducted a temporal analysis by DES era. We found that patients undergoing PCI during the Transitional era, when DES use was at its lowest (56% overall), had a 9% to 12% higher adjusted risk of death and cardiovascular events than did PCI with stenting during the Liberal era, when DES use was at its highest (85% overall) (Central Illustration). These results persisted despite the fact that patients in the Transitional era had a lower prevalence of most measured comorbid conditions and may have been expected to have better outcomes. In contrast, undergoing PCI in the Selective era (DES use 62% overall) compared with the Liberal era was not consistently associated with higher risks of death and cardiovascular outcomes, although patients in the Selective era had a higher prevalence of many comorbid conditions traditionally associated with poorer outcomes (e.g., diabetes mellitus, heart failure, peripheral arterial disease). In 2007 (the start of our Selective era), the recommended duration of dual antiplatelet therapy was extended for DES to at least 1 year and for BMS for up to 1 year (46), which may partly account for the lack of differences we observed in comparisons of the Selective and Liberal eras. Although not definitive, our results suggest no increased risk of death or MI with DES, an important finding, given the potential safety concerns in this population.
Despite our use of complementary analytical strategies to test the robustness of our results, in the absence of randomization to DES or BMS, the potential for residual confounding remains. In addition, our analysis relied on administrative claims data, and we did not have access to laboratory information (e.g., cardiac troponins), clinical variables (e.g., left ventricular ejection fraction), or coronary angiography data (e.g., lesion length and complexity). We cannot exclude the possibility that there were differences in the indication for any PCI (using BMS or DES) across the 3 eras and how such marginal patients may have influenced our results. Although we were able to examine any repeat revascularization procedure, we were unable to distinguish among target vessel revascularization, target lesion revascularization, or a de novo revascularization. Furthermore, we did not have information on the specific type of DES used (e.g., sirolimus-eluting, paclitaxel-eluting, or a newer generation agent), which can differ in efficacy (5). Finally, we did not have information on the concomitant use of medical therapies, including antihypertensive agents and antiplatelet agents.
The use of DES in patients with ESRD on dialysis mirrored trends in the general population: rapidly increasing after its introduction to the U.S. market in mid-2003, decreasing in 2006 over concerns regarding higher risks of stent thrombosis, and stabilizing from 2007 onward. Several analytical techniques suggest that DES were associated with better outcomes in terms of death and cardiovascular events than were BMS in patients with ESRD on dialysis. Until a randomized clinical trial is conducted, our study provides additional evidence to support current guidelines recommending preferential use of DES over BMS in this high-risk patient population.
COMPETENCY IN MEDICAL KNOWLEDGE: Coronary artery disease is common in patients with ESRD on dialysis and rates of PCI with stenting have risen over the past decade.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Current guidelines recommend use of DES over BMS in patients with ESRD, based largely on extrapolation of evidence from trials in patients without ESRD.
TRANSLATIONAL OUTLOOK: Future interventional studies should consider inclusion of patients across a broad range of kidney function, including patients with ESRD on dialysis.
This work was conducted under a data use agreement between Dr. Winkelmayer and the National Institutes of Diabetes and Digestive and Kidney Diseases. A National Institutes of Diabetes and Digestive and Kidney Diseases officer reviewed this manuscript for research compliance and approved of its submission for publication. Data reported herein were supplied by the USRDS (United States Renal Data System). Interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or interpretation of the U.S. government. Dr. Chang is supported by a grant from the National Institutes of Diabetes and Digestive and Kidney Diseases (K23DK095914). Dr. Winkelmayer is a scientific advisor to Amgen, Relypsa, Bayer, and Vifor Fresenius Medical Care Renal Pharma; and serves on trial committees for Bristol-Myers Squibb, Medtronic, and Zoll. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Anthony Bavry, MD, MPH, served as Guest Editor for this paper.
- Abbreviations and Acronyms
- bare-metal stent(s)
- confidence interval
- drug-eluting stent(s)
- end-stage renal disease
- hazard ratio
- inverse probability of treatment weighting
- myocardial infarction
- percutaneous coronary intervention
- Received July 22, 2015.
- Revision received September 11, 2015.
- Accepted October 20, 2015.
- American College of Cardiology Foundation
- Kumar N.,
- Baker C.S.,
- Chan K.,
- et al.
- U.S. Renal Data System
- Bangalore S.,
- Kumar S.,
- Fusaro M.,
- et al.
- Tsai T.T.,
- Messenger J.C.,
- Brennan J.M.,
- et al.
- Tomai F.,
- Ribichini F.,
- De Luca L.,
- et al.
- Shroff G.R.,
- Solid C.A.,
- Herzog C.A.
- Kim B.K.,
- Oh S.,
- Jeon D.W.,
- et al.,
- for the KOMATE Investigators
- Chang T.I.,
- Shilane D.,
- Kazi D.S.,
- Montez-Rath M.E.,
- Hlatky M.A.,
- Winkelmayer W.C.
- Pfisterer M.,
- Brunner-La Rocca H.P.,
- Buser P.T.,
- et al.,
- for the BASKET-LATE Investigators
- Krone R.J.,
- Rao S.V.,
- Dai D.,
- et al.,
- for the ACC/NCDR Investigators
- Venkitachalam L.,
- Lei Y.,
- Stolker J.M.,
- et al.,
- for the EVENT Registry Investigators
- Saran R.,
- Li Y.,
- Robinson B.,
- et al.
- Morrison L.
- Rosenbaum P.R.,
- Rubin D.B.
- Kosanke J.,
- Bergstralh E.
- Brookhart M.A.,
- Wyss R.,
- Layton J.B.,
- Sturmer T.
- Hernan M.A.,
- Robins J.M.
- Venkitachalam L.,
- Lei Y.,
- Magnuson E.A.,
- et al.,
- for the EVENT Registry Investigators
- Yeh R.W.,
- Vasaiwala S.,
- Forman D.E.,
- et al.
- U.S. Renal Data System
- Holden R.M.,
- Harman G.J.,
- Wang M.,
- Holland D.,
- Day A.G.
- Chang T.I.,
- Montez-Rath M.E.,
- Shen J.I.,
- Solomon M.D.,
- Chertow G.M.,
- Winkelmayer W.C.
- Anderson J.L.,
- Adams C.D.,
- Antman E.M.,
- et al.