HDCN Article Review/Hyperlink

Feldman HI, Kinosian M, Bilker WB, Simmons C, Holmes JH, Pauly MV, Escarce JJ

Effect of dialyzer reuse on survival of patients treated with hemodialysis

J Am Med Assoc (Aug) 276:620-625 1996

The authors report that patients new to dialysis in the years 1986 and 1987, and dialyzed in free standing facilites that reprocessed their dialyzers with Peracetic/Acetic Acid (PA/A) had a higher probability of death than other patients.

The authors used the Medicare PMMIS data base and the Centers for Disease Control annual facility survey for their data. They assumed for their primary analysis that patients once assigned to a facility that reported reuse of dialyzers, that the patient was consistently and continuously "exposed" to reuse ("intention to treat" assumption). They used a Cox model to define relative risks of death. Other covariables were age, sex, primary diagnosis, and race. They looked at two models. The first had reuse (yes or no) and all three types of sterilants (peracetic acid-hydrogen peroxide-acetic acid (PA/A), formaldehyde, and glutaraldehyde) in the equation. The second had only reuse (yes) and the three sterilants.

For both models, the risk of PA/A reuse was only significant in the free standing facilities. In the first model, there was a 10% increase in the relative risk of death (CI 1.01 - 1.18, p = .02) for patients dialyzing in free standing facilities using PA/A vs. no resuse. In the second there was an 8% increase in the relative risk of death (CI 1.01 - 1.14, p = .02) for patients dialyzing in free standing facilities reprocessing with PA/A vs. formaldehyde or glutaraldehyde.

In their sensitivity analysis, the authors "undid" the intention to treat assumptions and censored patients when they moved to a different dialysis unit. The relative risk persisted for PA/A reuse in free standind dialysis units.

The expected impact of race, age, diagnosis, and sex was seen in both their models. There was a slight improvement in survival in free standing facilities where there was "more frequent" water system disinfection and where reuse was an automated procedure.

Comment: The most significant aspect of this report is that it validates the findings of Held and colleagues (AJKD,23:692-708, 1994). It uses a larger data set and comes to essentially the same conclusions. Held's data referred to prevalent patients in 1989 and 90 followed for one year. This data referred to incident patients followed from 1986-7 to death, transplant, change to PD, or June 30, 1991. The time frames overlap, so assumptions about dialysis technology, and dialysis adequacy would be similar.

Since 1992, when the preliminary report of Held's study was published, there has been a significant improvement in the design, surveillance, and practice of reuse. It is not possible to extrapolate the relative risk from the Feldman study to current practice since much has changed in the industry both in the delivery of dialysis and the practice of reuse. Since the delivery of dialysis has an important and measureable impact on survival, it would be necessary to recast the Cox model to include that covariable to see if the reuse effect is still significant.

Feldman's data is reported at the facility level. The authors do not know with certainty that a given patient was actually "exposed" to reuse, but only that the patient was dialyzed in a facility reporting the practice of reuse. They do not know what average reuse numbers were, what the dialyzer was, what the causes of death were, or what the dialysis prescription or delivery were. Finally, since they do not report the R squared values, it is not possible to know how much of the variation in the survival remains "unexplained" by their models.

Feldman does explain the cause of their observation, but argues for the statistical validity of the association. He does not explain why the effect is not observed in hospitals. He suggests that the co morbidity is higher in hospitals so the reuse effect is not discoverable. He does not however provide standard mortality ratios for comparison. Feldman's study has a smaller number of patients on reuse in hospitals than in the free standing units so it may lack sufficient power to find a difference if it exists, though he does not provide a power analysis.

That the authors are not able to explain the effect does not vitiate their study. It is as important to understand the "fact" of this effect as it is to understand the "why" of this effect. The "fact" is that practices like reuse that are widely applied in an inconsistent manner can have important consequences on patient survival. More importantly, it takes large numbers of patients studied over years to detect this impact. In the average sized dialysis facility with about 70 patients and a crude mortality of 20% (14 deaths) an 8% increase in mortality would result in only 1 additional death.

This study reinforces the importance of the recommendations made by the manufacturer of PA/A and of AAMI to validate and document each step in the reuse process. It also reinforces the observation that there are facility based "risk" factors for mortality.

The application of reuse may only be an "indicator" of facility or corporate policies and procedures that have the unintended effect of increasing patient mortality. This study is a strong argument for facilities and dialysis corporations to cooperate in large data bases that allow the review of industry wide practices in a statisically powerful and timely manner. It took 6 years from the end of the Feldman study for the results to be published on a cohort of patients who started dialysis 10 years ago. What unexpected effects are we going to be reviewing in 2006 about our patients starting dialysis in 1996? (Peter B. DeOreo MD, Case Western Reserve University, Cleveland)

The full abstract is available at the JAMA site. You will need to register with the JAMA before being allowed to access this material. Once you have registered, click on the links below:
Abstract
Press Release