An Outbreak of Gram-Negative Bacteremia Traced to Contaminated O-Rings in Reprocessed Dialyzers

Methods

Hemodialysis Unit

The outpatient hemodialysis unit at the University of Chicago Hospital was opened in July 1985 in a building separate from the hospital. At the time of the outbreak, approximately 130 patients were receiving long-term hemodialysis treatment there. The unit operated three shifts per day, 6 days per week, and administered approximately 390 hemodialysis treatments per week. In September 1987, the center introduced high-flux dialysis, and in October 1987 the unit began a program of dialyzer reuse. Reprocessing was done by an automated procedure (Renatron; Minntech, Minneapolis, Minnesota) using a hydrogen peroxide-peracetic acid germicide (Renalin, Minntech). The long-term hemodialysis unit has two rooms for patient treatment and a separate room for the water treatment system, supplies, and dialyzer reprocessing. The main treatment room contains 16 stations, and the other treatment room (annex) contains 3 stations. The use of high-flux dialysis increased dramatically during March 1988, so that at the time of the investigation in July 1988, approximately 50% of all patients at the center were receiving high-flux dialysis. Recovery of Pseudomonas cepacia from blood culture specimens taken during dialysis on 30 June 1988 from 2 patients receiving long-term hemodialysis prompted an investigation that continued from July 1988 to April 1989.

Case Definition and Case Detection

A case was defined as an episode of primary gram-negative bacteremia occurring in a patient receiving long-term hemodialysis. To detect cases, results of all blood culture specimens from patients receiving long-term hemodialysis were reviewed retrospectively for the period 1 July 1987 to 30 June 1988 and then prospectively through April 1989.

Blood Cultures

Specimens for blood culture were obtained during hemodialysis from the venous return port after it was wiped with an iodophor antiseptic and then with 70% alcohol. Specimens were inoculated into an enrichment medium (Bactec, NR660; Becton-Dickinson Diagnostic Instrument Systems, Towson, Maryland). All organisms isolated were identified by an automated method (AutoMicrobic Systems; Vitek, Hazelwood, Missouri), confirmed by conventional biochemical methods, and tested for susceptibility to antimicrobial agents using a broth microdilution method.

Matched-Pair Study

We did a casecontrol study to identify risk factors for the development of primary gram-negative bacteremia. For each case, a matched control of the same sex and closest age was selected from among patients having outpatient hemodialysis on the same day who did not have signs or symptoms of gram-negative bacteremia. Medical records were reviewed to identify underlying renal disease; years on long-term dialysis; type of dialyzer (high-flux or conventional); type of vascular access; dialysis shift, station number, and location; dialysis machine; number of times dialyzer had been reused; dialyzer reprocessing technician; and recent antibiotic use. Data were analyzed using the McNemar chi-square test and the Wilcoxon signed-rank test.

Evaluation of the Hemodialysis Unit and Hemodialyzers

Techniques used to disinfect water-distribution lines and hemodialysis machines, to reprocess dialyzers, and to dialyze patients were observed to determine compliance with written procedures. The results of cultures routinely done each month to monitor treated water were reviewed, and, during the first 3 months of the investigation, additional 100-mL samples of treated water were cultured weekly from seven sites throughout the water-distribution system. Total bacterial counts were determined using a standard membrane filter technique [12]. On 7 July 1988 and 18 October 1988, cultures were taken of moist environmental surfaces and of antiseptics and solutions used in the dialysis unit for hemodialysis and dialyzer reprocessing. Serial volumes (1, 0.1, and 0.01 mL) of solution were plated on blood agar and incubated for 48 hours at 35 C, and growth was assessed quantitatively. Swab samples were plated on blood agar and assessed semi-quantitatively. All isolated organisms were identified using a commercially available kit (API test-strips; Analylab Products, Inc., Plainview, New York) or by conventional biochemical methods. When available, hemodialyzers (associated with patients who had bacteremia) were examined and cultured. Dialysis and blood pathway fluids were cultured using standard membrane filter techniques and were assayed for the presence of Renalin using Renalin residual test strips (Renal Systems, Inc.; Minneapolis, Minnesota). Screw-top headers (when present) were removed, and the fiber-bundle ends and header O-rings were cultured by impression onto blood agar. During 7 July to 6 September, 47 dialyzers used in patients without bacteremia were similarly cultured and assayed for Renalin concentration.

O-Ring Contamination

To determine whether O-rings were adequately decontaminated during reprocessing, O-rings from Hemoflow F-80 hemodialyzers (Fresenius AG; Bad Homburg, Germany) Figure 1 were deliberately contaminated by being dipped in a bacterial suspension and were then replaced in dialyzers that were reprocessed. Strains of bacteria used in these experiments were the isolates of P. cepacia, Xanthomonas maltophilia, Citrobacter freundii, Enterobacter cloacae, and Acinetobacter calcoaceticus var. anitratus from patients receiving hemodialysis who had gram-negative bacteremia. Bacterial concentrations ranged from 10² to 10⁵ colony-forming units (CFU)/mL. After the dialyzers had been disinfected with Renalin and stored for 48 hours, sterile saline was flushed through the blood and dialysis compartments and was then cultured quantitatively by plating 0.1-, 1.0-, and 10-mL aliquots onto blood agar. Also, O-rings were removed and cultured by impression onto blood agar. Growth was assessed semi-quantitatively at 48 hours as light when a faint or partial outline of a ring was visible, moderate when a complete circle was present with bleeding onto the surrounding agar, and heavy when a lawn of growth was present. Dialyzers showing persistent bacterial contamination after initial disinfection were processed a second time using the standard method. Several dialyzers with positive cultures after a second attempted disinfection were disinfected a third time with the additional step of removing the O-rings, dipping them in Renalin, and replacing them in the header before reuse.

Simulated Dialysis

Simulated hemodialysis was done after F-80 dialyzer O-rings were contaminated by dipping them in a suspension of 10⁴ to 10⁵ CFU/mL of bacteria. Dialyzers were reprocessed and stored at room temperature for 48 hours. To simulate dialysis, each dialyzer was connected to a dialysis machine, and 800 mL of sterile saline was circulated through the blood compartment, whereas 600 mL of bicarbonate solution was circulated through the dialysis compartment. This simulation was continued for approximately 20 minutes for each dialyzer. The saline circulating through the blood pathway was cultured, the dialyzers were then disconnected, and the O-rings were cultured. Next, these dialyzers were reprocessed with the additional step of removing the O-ring and dipping it in Renalin before automated reprocessed. The O-rings were replaced, and the dialyzer was reprocessed by the automated method. After storage at room temperature for 48 hours, the dialyzers again were used for simulated dialysis, and cultures were repeated.

Results

Description of Cases

During January to October 1988, 12 episodes of primary gram-negative bacteremia were identified in 11 patients receiving long-term hemodialysis (Table 1). One episode was caused by two different organisms, and the pathogens in the 12 episodes were P. cepacia (6 episodes), X. maltophilia (4 episodes), C. freundii (1 episode), Acinetobacter calcoaceticus var. anitratus (1 episode), and E. cloacae (1 episode). During the course of the outbreak, gram-negative bacteremia occurred at a rate of about 0.77 episodes per 1000 hemodialysis sessions.

The most common clinical manifestations of bacteremia were chills (11 episodes) and temperature greater than 37.5 C (8 episodes). In 11 episodes, symptoms began about 30 minutes to 3 hours after the start of treatment. In the remaining episode, chills began shortly after the patient had completed treatment and left the dialysis unit. Intravenous antibiotic treatment was administered for 11 episodes, and 3 episodes required that the patient be admitted to the hospital. All of the patients recovered, including one who received no treatment for X. maltophilia bacteremia. One patient had bacteremia first with C. freundii and then 6 weeks later with P. cepacia. Pseudomonas cepacia bacteremia was documented during three consecutive dialyses with the same dialyzer despite administration of intravenous gentamicin after the second dialysis. No further relapses occurred after a new dialyzer was substituted.

Case-Control Study

Comparison of case-patients with matched controls showed that case-patients were more likely to have received high-flux dialysis with Hemoflow F-80 hemodialyzers (11 of 12 cases compared with 5 of 12 controls; odds ratio 11). No significant differences were noted in underlying disease, years of hemodialysis, type of vascular access, dialysis room, dialysis machine, dialysis shift, reprocessing technician, or recent antibiotic use. The mean reuse number for bacteremic patients was 13.2 (range, 2 to 46) compared with 8.8 (range, 0 to 24) for matched controls (P > 0.05, Wilcoxon signed-rank test).

Hemodialysis Unit Investigation

In the water-treatment system for the unit, chloramine-treated municipal water passes through a carbon filter and then through reverse osmosis and deionization before entering a closed-circuit system composed of a storage tank and a continuously circulating loop supplying the 19 dialysis stations and the reprocessing room. The circuit is opened only to disinfect the system. When cultures of the treated water show a bacterial count exceeding the standard of 200 CFU/mL or less set by the Association for the Advancement of Medical Instrumentation [13], the system is disinfected. The system had not been disinfected in the year before the first case. Records of monthly cultures of treated dialysis water during January to May 1988 showed bacterial concentrations of 100 CFU/mL or less. No cultures were done in June. As part of this investigation, treated water cultures were taken weekly during July to October 1988. Bacterial concentrations exceeded 200 CFU/mL on four occasions, and sustained concentrations of less than 200 CFU/mL were not achieved until late July when larger-diameter pipe was installed to improve flow rates.

In the long-term hemodialysis unit, 15 Cobe Century 3 hemodialysis machines (Cobe Laboratories, Lakewood, Colorado) were available for either high-flux or conventional hemodialysis, whereas 4 Travenol SPS 450 hemodialysis machines (Baxter, Deerfield, Illinois) were used for conventional dialysis only. The external surfaces of dialysis machines were disinfected with sodium hydroxide after each treatment; the internal circuits of dialysis machines were disinfected each night with sodium hydroxide. Every 2 weeks on a Saturday evening, formaldehyde was instilled into the dialysis machines and was allowed to dwell until Monday morning. Patients receiving high-flux dialysis used Hemoflow F-80 polysulfone dialyzers; patients receiving conventional dialysis used CA-90 or CA-110 cellulose-acetate dialyzers (Baxter) or CF 12.11, CF 15.11, and CF 23.08 cellulosic dialyzers (Baxter). Bicarbonate dialysate was used for all hemodialysis. The bicarbonate concentrate was prepared individually for each patient from commercial powder and treated water in a volume sufficient for a single dialysis treatment.

Initial cultures of fluids and surfaces in the hemodialysis unit, obtained on 7 July 1988, grew aerobic gram-negative bacilli from the following specimens: untreated water (10²CFU/mL of P. cepacia), treated water (10³ CFU/mL of Alcaligenes species and Pseudomonas species), hoses used in the reprocessing room to rinse hemodialyzers with deionized water (P. cepacia), and a faucet aerator (Alcaligenes species). Cultures of Renalin, povidone-iodine, sodium hydroxide, and chlorhexidine disinfectant solutions, cleaning solution sprayers, saline rinse buckets, bicarbonate concentrate containers, sink surfaces, dialysis machine surfaces, and Renatron machine hoses and connectors were negative. Repeated cultures taken on October 18 yielded moderate X. maltophilia and P. acidovorans from treated water hoses in the reprocessing room; all other cultures were negative.

Dialyzer Reprocessing Procedure

Three technicians, one from each shift, were responsible for reprocessing dialyzers. First, the inside of each dialyzer was flushed manually with treated water; during this step, removable headers (if present) were unscrewed to rinse away blood and fibrin trapped underneath. Next, each dialyzer was attached to an automated reprocessing machine that tests dialyzer fiber-bundle volume and membrane integrity. Then the reprocessing machine infused 2.5% Renalin into both the blood and dialysate pathways, the dialyzer ports were closed with caps disinfected in Renalin, and the dialyzer was stored for 48 to 72 hours until next use. Immediately before reuse, the disinfectant was drained and the dialyzer was rinsed with sterile saline.

Renalin being used in the dialysis unit was sent to the manufacturer for testing in August 1988. The disinfectant was reported to have full activity.

Hemodialyzers

From 7 July to 21 October, 51 dialyzers in clinical use were obtained from the unit for culture. Dialyzers were retrieved from four bacteremic patients before being reprocessed. Each was an F-80 dialyzer, and each was culture positive for the same organism responsible for the patient's bacteremia. Specimens that were culture positive included those from the O-ring (4 dialyzers), fiber bundle (2 dialyzers), dialysate compartment (1 dialyzer), and blood compartment (1 dialyzer). Three of the culture-positive dialyzers were reprocessed using standard technique and were recultured 72 hours later. In all 3 dialyzers, O-rings remained culture positive, and in 2, the fiber-bundle ends that were in contact with the O-rings remained positive. The other 47 dialyzers were obtained from patients without documented bacteremia. Forty-one of these were from asymptomatic patients, and all were culture negative. The six remaining dialyzers were obtained from patients with fever during dialysis who had negative blood cultures. Four of the 6 dialyzers were culture positive: Two grew Alcaligenes species or Acinetobacter calcoaceticus from O-rings, 1 grew E. cloacae in the dialysate and blood compartment fluids, and 1 had P. aeruginosa in the dialysate compartment and E. cloacae and Acinetobacter calcoaceticus on end fibers.

Bacterial Inoculation of O-Rings and Simulated Hemodialysis

Inoculation of F-80 dialyzer O-rings with 10³ CFU/mL or more of X. maltophilia, P. cepacia, E. cloacae, or C. freundii strains isolated from patients with bacteremia resulted in persistent O-ring contamination despite attempts to disinfect dialyzers with the Renatron automated reprocessing method (Table 2). Acinetobacter calcoaceticus did not cause persistent O-ring contamination.

Simulated dialysis using dialyzers with contaminated O-rings caused blood pathway contamination despite intervening disinfection with Renalin disinfectant (Table 3). When the method of reprocessing the F-80 dialyzers was altered to include the removal of O-rings and submersion in Renalin disinfectant, O-ring and blood compartment fluid cultures were consistently negative.

Control Measures

In late October 1988, the procedure for reprocessing dialyzers was changed to require that O-rings from F-80 dialyzers be removed and immersed in Renalin disinfectant as part of the routine disinfection procedure. During the ensuing 6 months, no further cases of primary gram-negative bacteremia were detected in patients receiving long-term hemodialysis (Figure 2).

Discussion

We hypothesize that the outbreak of gram-negative bacteremia in patients receiving long-term hemodialysis at our institution was caused primarily by intermittent contamination and inadequate disinfection of O-rings in Hemoflow F-80 dialyzers. When dialyzers were disassembled for reprocessing, the O-ring was sometimes removed from the header, and both parts presumably became contaminated with gram-negative bacteria present at low concentrations in the treated water used to rinse the header. When the O-ring was replaced in the header for disinfection, portions of the O-ring that were compressed against the header or fiber bundle apparently remained shielded from the disinfectant. Organisms on shielded portions of the O-ring entered the blood pathway when the dialyzer subsequently was used for hemodialysis.

Several observations linked the problem of gram-negative bacteremia to O-rings in F-80 dialyzers. This was the only model that had a removable header and O-ring, cases of bacteremia were significantly associated with this model, bacteremia began soon after the dialysis unit began reprocessing F-80 dialyzers, and the outbreak ended when the disinfection procedure was modified specifically to disinfect the O-rings. Culture studies in the dialysis unit further incriminated O-rings and the reprocessing procedure. The two most common bloodstream isolates, P. cepacia and X. maltophilia, were detected in treated water or hoses used to rinse O-rings; positive cultures could almost always be obtained from O-rings of dialyzers associated with cases of bacteremia or fever; and contaminated O-rings remained culture-positive despite repeated reprocessing. Finally, mock trials with F-80 dialyzers showed that organisms from the O-ring could enter the blood pathway during hemodialysis.

The hemodialysis unit continued to use the Hemoflow F-80 for high-flux dialysis until October 1992. Throughout the period from October 1988 to October 1992, hemodialysis technicians were instructed to use the modified reprocessing procedure, that is, to remove the O-rings from F-80 headers and to disinfect them in Renalin. During that 4-year period, occasional episodes of primary gram-negative bacteremia (approximately 3/y) occurred in patients receiving dialysis. These may have been related to imperfect compliance with the modified reprocessing procedure or to other events. In October 1992, the dialysis unit switched from Hemoflow F-80 to Baxter CT-190G high-flux dialyzers. The CT-190G dialyzers were lower in cost and appeared to be more durable, allowing a greater number of reuses. Like the F-80 dialyzers, CT-190G dialyzers have removable headers and O-rings, but the O-rings are tight fitting and are difficult to remove. Attempts to remove the O-rings from the headers sometimes resulted in damage to the O-rings. As a result, reprocessing technicians independently abandoned the modified reprocessing procedure until a cluster of three patients with primary gram-negative bacteremia in February 1993 prompted another review of reprocessing practices. Given the difficulty in removing the O-rings from the CT-190G dialyzer header, our current approach is to rinse residual blood and fibrin from the header with treated water and then immerse the entire header (with the O-ring in place) in Renalin disinfectant. Whether this will prevent further primary gram-negative bacteremias related to O-ring contamination remains to be determined.

Hemoflow F-80 dialyzers continue to be used in many dialysis centers. The dialyzer reprocessing methods used at the hemodialysis center we investigated are practiced elsewhere and presumably have caused other cases of bacteremia. Such cases may not be recognized to represent an outbreak problem, because they occur at low frequency and involve different pathogens. Also, as shown by Wagnild and colleagues [5], only about 10% of cases of bacteremia cause systemic symptoms that might prompt hemodialysis personnel to obtain blood cultures.

A unique aspect of this investigation was our success in identifying the route by which bacteria gained access to the bloodstream. Investigations of previous outbreaks have frequently identified the causative organism in treated dialysis water, but the mechanism of bloodstream contamination has been largely a matter of conjecture [5, 6, 9]. Several clusters of patients with gram-negative bacteremia in centers practicing dialyzer reuse have been attributed to inadequate disinfectant activity, either due to intrinsic resistance [5], improper preparation [8], or improper delivery [6]. A cluster of Mycobacterium chelonae infections in a hemodialysis center in California was associated with reuse of Hemoflow F-80 dialyzers, as in the present investigation [10]. Laboratory studies documented the failure to eradicate M. chelonae from two contaminated dialyzers using routine reprocessing techniques. The outbreak was attributed to inadequate potency of the Renalin disinfectant. No attempt to culture dialyzer O-rings was described.

Based on reports of bacteremias in patients receiving high-flux dialysis attributed to contamination of dialyzer O-rings, Bland and coworkers [14] at the Centers for Disease Control studied reprocessed F-80 dialyzers after contamination of the header and O-rings with suspensions of X. maltophilia and M. chelonae [14]. They found that low-level O-ring contamination was associated with infrequent blood compartment contamination during simulated dialysis. However, when high-level O-ring contamination was established by reprocessing with saline in place of Renalin, blood compartment contamination occurred consistently.

Passage of bacteria through high-flux dialyzer membranes is a potential mechanism for bacteremia but is not a recognized problem. Further, pressure testing was routinely done during reprocessing, and dialyzers with evidence of membrane leak were discarded. An investigation of four clusters of patients with gram-negative bacteremia associated with reprocessing cellulosic dialyzers with a chlorine dioxide disinfectant concluded that the disinfectant may have degraded the integrity of dialyzer membranes to the extent that leaks developed [7, 15]. High-flux dialysis uses a synthetic open membrane with a high diffusive and hydraulic permeability allowing for shorter, more efficient dialysis [16]. Patients have a universal preference for shorter therapy sessions, and there are obvious economic incentives to shorten treatment time. There is concern, however, that the more open membranes and the higher flow rates used in high-flux dialysis may allow increased back-filtration of bacteria and endotoxin from dialysate to blood. A recent Centers for Disease Control survey [17] of long-term hemodialysis centers in the United States found that use of high-flux dialysis was associated with an increased risk for pyrogenic reactions, particularly in centers that reused dialyzers. However, no association was noted between high-flux dialysis and septicemia. In addition, a prospective comparison of patients treated with either high-flux or conventional dialysis found no difference in the incidence of pyrogenic reactions [18]. In vitro studies have shown that high-flux membranes are effective barriers to bacteria and endotoxin when compared with conventional membranes; F-80 dialyzer polysulfone membranes, in particular, have shown extremely low permeability to endotoxin [19].

Renalin, the hydrogen peroxide-peracetic acid disinfectant in use during this outbreak, has gained substantial acceptance among dialysis centers. Its nonirritant qualities and reputed germicidal effectiveness have made it increasingly popular among centers reusing dialyzers. Of U.S. hemodialysis centers reusing dialyzers in 1989, 46% used Renalin for reprocessing [17]. Although the effectiveness of Renalin has been questioned in the context of particular clusters of cases of bacteremia [8, 10], extensive use of Renalin has not been associated with increased risks for infection [3, 17]. In the circumstances described in this outbreak, it does not seem likely that an alternative disinfectant would have eliminated persistent O-ring contamination.

In the past, rubber gaskets in primitive hemodialyzers, analogous to F-80 dialyzer O-rings, have been associated with bacterial contamination. Jones and coworkers [20, 21] noted that cemented or tight-fitting gaskets in modified Kiil dialyzers were associated with heavy bacterial contamination of the gaskets and increased bacterial counts in the dialyzer effluent. They postulated that bacteria sequestered beneath the gasket escaped disinfection. The use of loose-fitting gaskets that were removed during the disinfection process appeared to eliminate the problem. Similarly when our reprocessing procedure was altered to include removal and disinfection of the O-ring gasket, our infection problem was almost eliminated. Cases occurring during the subsequent 3 years may have represented contamination from other sources or failure to do disinfection procedures correctly.

In 1989, 68% of long-term hemodialysis centers reused disposable dialyzers; these centers treated 73% of the dialysis patients [17]. Despite this widespread practice, the benefits and risks of this approach have not been well defined. When practiced carefully, dialyzer reuse has not been associated with increased mortality or morbidity rates [2, 5, 22]. In fact, reprocessing appears to confer greater biocompatibility on dialyzer membranes, with a decrease in certain adverse reactions [23]. Moreover, the potential cost savings are substantial: Savings are estimated at $70 to $120 million annually for U.S. dialysis providers [2]. Thus, because reuse of dialyzers will likely continue, strict disinfection techniques must be applied, and devices as well as disinfectants must be carefully evaluated and re-evaluated for the presence of design flaws that permit infection.

Preliminary results of this study were presented in part at the 31st Interscience Conference on Antimicrobial Agents and Chemotherapy (Chicago, 1991).