American Journal of Kidney Diseases
Volume 53, Issue 3 , Pages 359-362, March 2009

Vascular Access Practice in Hemodialysis: Instrumental in Determining Patient Mortality

  • Kevan R. Polkinghorne, MBChB, M Clin Epi, FRACP, PhD

      Affiliations

    • Corresponding Author InformationAddress correspondence to Dr Kevan R. Polkinghorne, PhD, Department of Nephrology, Monash Medical Center, 246 Clayton Rd, Clayton, Melbourne, Victoria, Australia 3168

Monash University, Monash Medical Centre, Melbourne, Australia

Article Outline

 

Related Article, p. 475

Important epidemiological studies and findings often arise from very simple clinical observations. In a letter to Nephrology Dialysis and Transplantation in 1997, de Almeida and colleagues1 noted reduced survival in hemodialysis patients who required more than 1 vascular access over a period of 2.5 years of follow up. This short letter was (unusually) accompanied by an editorial comment by Woods and Port broadening the hypothesis and suggesting that commencing hemodialysis without permanent vascular access could be associated with increased patient morbidity and possibly mortality via an increased risk of infection and the delivery of a lower dialysis dose.2 They suggested that the effect of vascular access “on morbidity and mortality among incident haemodialysis patients should be tested in large prospective epidemiological studies. Ideally such studies should be based on a random national sample so that they are nationally representative rather than simply reflecting the practice of single centres.”

Since the publication of the editorial, a number of large observational studies have been published assessing the link between vascular access type and mortality.3, 4, 5, 6, 7, 8, 9, 10 All but 1 study7 found an association between vascular access type and patient mortality. Three early studies examined US-based cohorts demonstrating increased risks of death for both arteriovenous grafts (AVG) and catheters compared with native arteriovenous fistulas (AVF), respectively.3, 4, 6 However, these studies did not adjust for the presence of late referral, which is not only a marker of poor vascular access preparation and whether a patient receives a catheter,11, 12 but also is a predictor of early mortality in incident end-stage renal disease patients.13, 14 Subsequent studies, 1 each from the United States,8 Canada,10 and Australia/New Zealand,5 again suggested an increased risk of death with catheter but not necessarily AVG use, even after accounting for the presence of late referral and patient comorbidity. Finally, an analysis of data from the Hemodialysis (HEMO) Study again suggested a relationship between access type and mortality but also demonstrated a reduction in the mortality risk when there was a change from a catheter to either an AVF or AVG.9

Twelve years after the original hypothesis, Pisoni et al15 present an analysis of the Dialysis Outcomes and Practice Patterns Study (DOPPS) in this issue of the American Journal of Kidney Diseases exploring the relationship between vascular access and mortality in the United States, Europe, and Japan. Why, then, do we need another study given the consistency of the previous data? The major problem in examining any relationship between vascular access type and mortality is the presence of “treatment-by-indication bias” whereby patients with catheters are older and have greater levels of comorbidity, thus making it very difficult to exclude any residual confounding even after standard statistical adjustment. Researchers are able to make adjustments for the data they have, but often in observational cohort studies, other important known confounding factors have not been measured and therefore cannot be accounted for in any analysis. Late referral and other indicators of predialysis care are good examples of possible unmeasured confounding factors. In addition, there are likely other as-yet unknown confounding factors that cannot be accounted for. Randomized trials are able to deal with this issue but, clearly, a randomized trial of vascular access in hemodialysis comparing AVF to catheters, for example, cannot be ethically justified.

In recent times, various statistical techniques have been proposed as a way to deal with the issue of residual confounding in observational studies, including the use of propensity scores,16, 17 marginal structural models,18 and instrumental variables.19 The propensity score describes the probability for any given patient, based on their characteristics, of receiving one treatment versus another. The estimated propensity scores can then be used either as a stratification variable, as a controlling variable in a multivariable model, or as a tool to derive a propensity score–matched cohort in which all the variables used to derive the propensity score are balanced.16 Polkinghorne et al utilized propensity scores to control for additional confounding when assessing the effect of catheter and AVG use on mortality.5 Modest attenuation of the risk of death for both catheter and AVG use compared to AVF was seen. However, the issue of residual confounding remains a concern with this technique especially given recent analyses suggesting that, in the majority of studies, the addition of propensity scores failed to provide additional adjustment or advantage over standard regression models.20, 21

Pisoni et al present the use of instrumental variables to reduce the treatment-by-indication bias inherent to observational studies assessing the vascular access mortality relationship.15 The instrumental variable groups patients in a “pseudo random” way into a regional treatment group, for example within a dialysis facility or a given distance from a treating center, and is able to produce 2 matching groups similar to a randomized trial.19 While the use of instrumental variables is not a new statistical technique, its use in the biomedical arena is recent.22, 23, 24 In using this approach, a number of important issues need to be kept in mind. Like all statistical techniques, the application of an instrumental variable approach requires that a number of underlying properties be satisfied for the results to be valid and free of bias. Specifically, the chosen instrumental variable must have 2 properties.19 First, it should affect (cause variation in) the independent variable of interest. In the current study, the “instrument” chosen was the facility-level catheter, AVG, or AVF use and the authors clearly demonstrated that the facility-level vascular access practice accounted for more of the variance in patient vascular access use than did the patients' demographic characteristics and comorbid conditions. Second, the instrumental variable should have no direct effect on the outcome measure, in this case patient mortality and hospitalization. Can we be sure that facility practice itself is unrelated to patient outcome? One could argue that some dialysis centers provide better care and/or practice than others and that this could indeed affect patient outcomes. However, the authors provide evidence that, in this case, the properties have been met, with minimal change in the vascular access/mortality relationship seen when adjusted for numerous facility-based practices. However, some variation was seen when adjustment was made for facility dialysis dose, control of hemoglobin, and serum albumin level. As noted, each of these variables can arguably be a consequence of vascular access practice and therefore could mediate their effects on mortality via the “treatment variable” (vascular access type). If so, the independence of the instrumental variable to the outcome would be maintained. Finally, interpretation of the hazard ratios in the instrumental variable analysis are also different. Unlike patient-based analyses, which define risk for individual patients, the hazard ratio in the instrumental variable model defines the survival of patients within a facility with the specified vascular access practice patterns.

The study by Pisoni et al provides additional, strong evidence for the increase in both mortality and hospitalization for patients dialyzing with catheters or AVGs compared with AVFs.15 For every 20% increase in adjusted facility catheter and AVG use, there was a significant 20% and 9% increase in the risk of death compared with AVF use, respectively. In an alternative but complementary analysis, compared with facilities with an adjusted catheter and AVG use of less than 10%, patients dialyzing in a facility with greater than 20% catheter use and 60% or greater AVG use had a 60% higher mortality risk. Even in facilities with low catheter use (<10%) but high AVG use (≥60%), the risk of death was 42% higher compared with facilities with both low catheter and AVG use. The nature of DOPPS also provides an opportunity to assess mortality differences between countries and continents. It is well known that mortality rates in the US hemodialysis population are higher compared to those in Europe, Japan, and Australia/New Zealand. US-based DOPPS participants had up to a 40% greater mortality compared with the 5 included European countries despite adjusting for differences in patient demographics. This excess in mortality was significantly reduced when differences in vascular access use within facilities were incorporated, suggesting that a large part of the mortality may be mediated by differences in vascular access practice patterns. Similar results were seen for comparison between Japan and the United States.

The study by Pisoni et al provides important insights into the assessment of the vascular access mortality relationship using novel statistical techniques. The under-representation of randomized trials in nephrology25 ensures the reliance on well-designed observational studies to help guide nephrological practice. It is likely that the use of instrumental variables and other statistical techniques such as marginal structural models will be increasingly employed to answer important clinical questions not easily amenable to randomized trials. Comparing mortality of patients on hemodialysis versus peritoneal dialysis is one example.26

Despite the recent “Fistula First” initiative, which has seen an increase in AVF prevalence rates, incident and prevalent catheter rates remain high in the United States27 as well as in Canada10 and countries with higher AVF prevalence rates such as Australia and New Zealand.28 While we know a great deal about the characteristics of the patients dialyzing with a catheter or an AVG,29, 30, 31, 32, 33 it is clear that they alone cannot account for the large variation seen in catheter use across geographic areas.29, 30, 34 More research is needed to understand why similar patients commence dialysis with a catheter in one facility but with an AVF in another.34 Vascular access practice, I believe, is the most important modifiable practice pattern in hemodialysis. A reduction in catheter use could deliver significant benefits not only for patients, but also in reducing costs and strain on an already stretched healthcare system.

Back to Article Outline

Acknowledgements 

Financial Disclosure: None.

Back to Article Outline

References 

  1. de Almeida E, Dias L, de Sousa FT, Mil-Homens MC, Pataca I, Prata MM. Survival in haemodialysis: is there a role for vascular access?. Nephrol Dial Transplant. 1997;12:852
  2. Woods JD, Port FK. The impact of vascular access for haemodialysis on patient morbidity and mortality. Nephrol Dial Transplant. 1997;12:657–659
  3. Dhingra RK, Young EW, Hulbert-Shearon TE, Leavey SF, Port FK. Type of vascular access and mortality in U.S. hemodialysis patients. Kidney Int. 2001;60:1443–1451
  4. Pastan S, Soucie JM, McClellan WM. Vascular access and increased risk of death among hemodialysis patients. Kidney Int. 2002;62:620–626
  5. Polkinghorne KR, McDonald SP, Atkins RC, Kerr PG. Vascular access and all-cause mortality: a propensity score analysis. J Am Soc Nephrol. 2004;15:477–486
  6. 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. 2003;42:1013–1019
  7. Di Iorio BR, Bellizzi V, Cillo N, et al. Vascular access for hemodialysis: the impact on morbidity and mortality. J Nephrol. 2004;17:19–25
  8. 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. 2005;16:1449–1455
  9. 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. 2006;47:469–477
  10. Moist LM, Trpeski L, Na Y, Lok CE. Increased hemodialysis catheter use in Canada and associated mortality risk: data from the Canadian Organ Replacement Registry 2001-2004. Clin J Am Soc Nephrol. 2008;3:1726–1732
  11. 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. 2002;55:711–716
  12. Astor BC, Eustace JA, Powe NR, et al. Timing of nephrologist referral and arteriovenous access use: the CHOICE Study. Am J Kidney Dis. 2001;38:494–501
  13. Stack AG. Impact of timing of nephrology referral and pre-ESRD care on mortality risk among new ESRD patients in the United States. Am J Kidney Dis. 2003;41:310–318
  14. Winkelmayer WC, Owen WF, Levin R, Avorn J. A propensity analysis of late versus early nephrologist referral and mortality on dialysis. J Am Soc Nephrol. 2003;14:486–492
  15. Pisoni RL, Arrington CJ, Albert JM, et al. Facility hemodialysis vascular access use and mortality in countries participating in DOPPS: an instrumental variable analysis. Am J Kidney Dis. 2009;53:475–491
  16. D'Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17:2265–2281
  17. Joffe MM, Rosenbaum PR. Invited commentary: propensity scores. Am J Epidemiol. 1999;150:327–333
  18. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology. 2000;11:550–560
  19. Newhouse JP, McClellan M. Econometrics in outcomes research: the use of instrumental variables. Annu Rev Public Health. 1998;19:17–34
  20. Shah BR, Laupacis A, Hux JE, Austin PC. Propensity score methods gave similar results to traditional regression modeling in observational studies: a systematic review. J Clin Epidemiol. 2005;58:550–559
  21. Sturmer T, Joshi M, Glynn RJ, Avorn J, Rothman KJ, Schneeweiss S. A review of the application of propensity score methods yielded increasing use, advantages in specific settings, but not substantially different estimates compared with conventional multivariable methods. J Clin Epidemiol. 2006;59:437–447
  22. McClellan M, McNeil BJ, Newhouse JP. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? (Analysis using instrumental variables). JAMA. 1994;272:859–866
  23. Brooks JM, Irwin CP, Hunsicker LG, Flanigan MJ, Chrischilles EA, Pendergast JF. Effect of dialysis center profit-status on patient survival: a comparison of risk-adjustment and instrumental variable approaches. Health Serv Res. 2006;41:2267–2289
  24. Stukel TA, Fisher ES, Wennberg DE, Alter DA, Gottlieb DJ, Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA. 2007;297:278–285
  25. Strippoli GF, Craig JC, Schena FP. The number, quality, and coverage of randomized controlled trials in nephrology. J Am Soc Nephrol. 2004;15:411–419
  26. McDonald SP, Marshall MR, Johnson DW, Polkinghorne KR. Relationship between dialysis modality with mortality. J Am Soc Nephrol. 2009;20:155–163
  27. US Renal Data System: USRDS. 2007 Annual Data Report: Atlas of chronic kidney disease & end-stage renal disease in the United States. Am J Kidney Dis. 2008;52(suppl 1):S123–S136
  28. Moist LM, Chang SH, Polkinghorne KR, McDonald SP. Trends in hemodialysis vascular access from the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA) 2000 to 2005. Am J Kidney Dis. 2007;50:612–621
  29. Hirth RA, Turenne MN, Woods JD, et al. Predictors of type of vascular access in hemodialysis patients. JAMA. 1996;276:1303–1308
  30. Polkinghorne KR, McDonald SP, Atkins RC, Kerr PG. Epidemiology of vascular access in the Australian hemodialysis population. Kidney Int. 2003;64:1893–1902
  31. Reddan D, Klassen P, Frankenfield DL, et al. National profile of practice patterns for hemodialysis vascular access in the United States. J Am Soc Nephrol. 2002;13:2117–2124
  32. Sehgal AR, Silver MR, Covinsky KE, Coffin R, Cain JA. Use of standardized ratios to examine variability in hemodialysis vascular access across facilities. Am J Kidney Dis. 2000;35:275–281
  33. Stehman-Breen CO, Sherrard DJ, Gillen D, Caps M. Determinants of type and timing of initial permanent hemodialysis vascular access. Kidney Int. 2000;57:639–645
  34. Hopson S, Frankenfield D, Rocco M, McClellan W. Variability in reasons for hemodialysis catheter use by race, sex, and geography: findings from the ESRD Clinical Performance Measures Project. Am J Kidney Dis. 2008;52:753–760

PII: S0272-6386(09)00041-9

doi:10.1053/j.ajkd.2009.01.010

Refers to article:

  • Facility Hemodialysis Vascular Access Use and Mortality in Countries Participating in DOPPS: An Instrumental Variable Analysis , 16 January 2009

    Ronald L. Pisoni, Charlotte J. Arrington, Justin M. Albert, Jean Ethier, Naoki Kimata, Mahesh Krishnan, Hugh C. Rayner, Akira Saito, Jeffrey J. Sands, Rajiv Saran, Brenda Gillespie, Robert A. Wolfe, Friedrich K. Port
    American Journal of Kidney Diseases March 2009 (Vol. 53, Issue 3, Pages 475-491)

American Journal of Kidney Diseases
Volume 53, Issue 3 , Pages 359-362, March 2009

Article Tools