Acute Onset of Decreased Vision and Hearing Traced to Hemodialysis Treatment With Aged Dialyzers

Context A recent event in which 7 patients at 1 hospital developed decreased vision and hearing, conjunctivitis, headache, and other severe neurologic symptoms 7 to 24 hours after hemodialysis drew attention to the issue of the long-term integrity of dialysis machines and materials.

Objective To determine the cause of the adverse reactions that occurred during this event.

Design, Patients, and Setting Retrospective cohort study of all 9 patients who received hemodialysis at hospital A on September 18, 1996, the day of the outbreak. A case-patient was defined as any hospital A patient with acute onset of decreased vision and hearing and conjunctivitis after dialysis on that day. Non–case-patients were all others who underwent dialysis at hospital A on that day but did not develop adverse reactions. In an attempt to reproduce the conditions of the event, cellulose acetate dialysis membranes of various ages were retrieved from other sources and tested for physical and chemical degradation, and degradation products were identified, characterized, and injected intravenously into rabbits.

Main Outcome Measures Clinical signs and symptoms, time to resolution of symptoms, mortality, and dialyzer type and age, for case- vs non–case-patients.

Results Seven of the 9 patients met the case definition. In addition to diminished vision and hearing, conjunctivitis, and headache, some case-patients had blood leak alarm activation (n=6), confusion/lethargy (n=5), corneal opacification (n=4), cardiac arrest (n=2), or other neurologic signs and symptoms. One case-patient died during hospitalization after the event; 5 of 7 case-patients died within 13 months. Resolution of signs and symptoms varied but persisted more than 3 years or until death in 3 of the 6 patients who survived hospitalization. All case-patients but no non–case-patients were exposed to 11.5-year-old cellulose acetate dialyzers (all of these dialyzers were discarded by the hospital before our investigation). Laboratory investigation of field-retrieved 0- to 13.6-year-old dialyzers of similar type indicated significant chemical degradation in the older membranes. In vivo injection of extracts of membrane degradation products produced iritis and hemorrhages in rabbits' eyes.

Conclusions Severe patient injury was associated with exposure to aged cellulose acetate membranes of dialyzers, allowing cellulose acetate degradation products to enter the blood. Clinicians should be aware that aged cellulose acetate membranes may cause severe adverse reactions.

Severe reactions associated with hemodialysis are rare but can be life-threatening.¹ The majority of such reactions are caused by bacterial and/or endotoxin contamination of the dialysate or product water and are manifested by fever and hypotension.^2-8 Oba et al⁹ reported sporadic eye and hearing injuries attributed to degradation products of hemicellulose found in new cellulose acetate hemodialyzers. This early event was traced to improper manufacturing of these devices, and the injuries were manifest as soon as this new product line was introduced to the market. Since that time, similar incidents related to manufacturing or packaging have been reported in many parts of the world.¹⁰ Currently, there are approximately 250,000 patients undergoing dialysis treatment in the United States alone. According to the Centers for Disease Control and Prevention's surveillance system, 43% of centers use cellulose acetate and an additional 14% of centers use cellulose triacetate for high-flux dialysis. Due to this widespread use in dialysis and other blood contacting devices (eg, blood filters), the long-term integrity of cellulose acetate and other medical device polymers has become an important issue.

In September 1996, 7 patients at hospital A (inpatient dialysis unit) developed acute onset of diminished vision and hearing less than 24 hours after hemodialysis using cellulose acetate dialyzers with a previous excellent safety record. Because of the severity of the adverse reactions, the Centers for Disease Control and Prevention and the Food and Drug Administration (FDA) initiated an investigation to identify the source of the adverse events.

A case-patient was defined as any patient with onset of decreased vision and hearing within 48 hours of the initiation of hemodialysis at hospital A on September 18, 1996 (ie, the epidemic day). To identify case-patients, we reviewed patient medical records, laboratory results, and interviewed facility staff caring for all patients receiving hemodialysis at hospital A on the epidemic day. Non–case-patients were all those who underwent dialysis on the same day who did not develop adverse reactions.

To identify risk factors, a retrospective cohort study of all patients who underwent hemodialysis at hospital A on the epidemic day was conducted. Variables examined included demographics, clinical signs and symptoms, laboratory results, dialyzer type, time undergoing dialysis, indication as to whether a hemodialysis machine alarm sounded indicating blood leakage in the dialyzer, visible blood in the dialysis fluid, and medications received.

Although not routine in the unit, but because of the severity of the visual and hearing abnormality, unit personnel arranged for ophthalmologic and auditory testing of all case-patients. Visual and auditory acuity were measured at multiple times 1 to 8 days after dialysis. Visual acuity was assessed with the standard ophthalmologic descriptions of no light perception, light perception only, hand motion detected, ability to count fingers, and vision better than 20/200. Audiometry results were categorized as profound, severe to profound, severe, moderate to severe, moderate, mild to moderate, mild, and no discernable hearing loss. When multiple tests were done, the worst reading was recorded. Data were collected on standardized forms, entered into a computer, and analyzed using EPI Info statistical software (version 6.02).¹¹ The Fisher exact test was used for comparison of categorical variables and the t test or Kruskal-Wallis test was used to compare continuous variables.

Cellulose acetate is one of the most commonly used dialysis membrane materials. Its properties (eg, melt viscosity, solution solubility, phase equilibrium, tensile strength, and crystallinity) are determined from both average molecular weight and acetyl content. Degradation of the material over time will change these properties. To determine the potential mechanism of the patients' reactions, we assessed material degradation by measuring changes in polymer molecular weight, degree of acetylation, and isolated extractable compounds of cellulose acetate dialyzer membranes of various ages. Staff at hospital A discarded all of the 11.5-year-old dialyzers used for the 7 case-patients as well as 5 unused 11.5-year-old dialyzers before our investigation. Dialyzers were not reused at hospital A. To identify the possible degradation products and to reconstruct the event, the FDA retrieved dialyzers of various ages from commercial sources and dialysis centers nationwide. Some variation in the storage conditions of these devices was anticipated, since the dialyzers were retrieved from warehouses and clinical storage areas; however, these ambient storage conditions were typical of the real-world storage conditions for these devices. We characterized dialyzer membranes and their degradation products by chromatographic, chemical, and spectroscopic methods.

First, to determine changes in molecular weight of the polymers of dialyzer membranes of different ages, we performed gel permeation chromatographic analyses. Cellulose acetate fibers were removed from the dialyzers and dissolved in tetrahydrofuran.

Chromatographic analyses were completed using a Waters 150C-gel permeation chromatograph (Waters, Milford, Mass) equipped with Styragel columns (Waters) (molecular weight ranges, 2000-4,000,000; 5000-500,000; and 500-30,000 g/mol) and a refractive index detector. Number-average molecular weight, weight-average molecular weight, peak-average molecular weight, and polydispersity index were calculated using statistical software (Millennium, Waters). The polydispersity index is equal to the weight-average molecular weight divided by the number-average molecular weight. The Mark-Houwink-Sakurada viscosity parameters used for universal calibration were obtained from the literature,¹² and were in good agreement with our own measurements.

Second, we further assessed the age-related degradation of cellulose acetate by measuring the pendant acetyl groups (total acetyl content) on the polymer chain. Cellulose acetate used for these dialyzer membranes has an acetyl content of 39.9% by weight. As membranes degrade with age, the acetyl content decreases. We measured the acetyl content by a standard saponification and titration procedure.¹³

Third, we measured the dialyzers' extractable compounds as an indication of membrane degradation. The fibers from 40 dialyzers of various ages were removed and rinsed with 1 L of cold deionized water to remove the glycerol, and extracted in 600 mL of deionized water for 2 hours at 50°C. Then, the solutions were filtered through a 0.45-mm filter, the filtrate was evaporated at 70°C, mixed with 10 mL of ethanol, dried at 50°C under vacuum pressure, and then weighed.

Fourth, to confirm the chemical structure of the suspected membrane causative agent, we synthesized model degradation compounds. Cellulose acetate resin (Aldrich, Milwaukee, Wis) was degraded in phosphate-buffered saline (PBS; pH of 7.0) under oxidative conditions using sodium hypochlorite and cobalt as an accelerant.¹⁴ The water-soluble fraction was dialyzed and lyophylized to dryness. Next, we used infrared spectroscopy to compare the chemical structure of the dialyzer extracts and the synthesized degradation compounds with the spectrum of high-purity, cellulose-acetate resin. Samples were deposited on sodium chloride windows and scanned using a Fourier-transformed infrared spectrometer (Nicolet Instrument Corp, Madison, Wis).

Finally, we attempted to duplicate the human adverse reactions by injecting rabbits with either the water-soluble material recovered from extraction of a 13.6-year-old dialyzer, the synthetic degradation compounds, or a PBS control solution. The 13.6-year-old dialyzer was chosen since it was the oldest dialyzer available with a diminished acetyl content (33.1% by weight) and was the most similar in type to that used at hospital A at the time of the outbreak. The isolated materials were dissolved in a solution consisting of 70% ethanol and 30% PBS, then diluted in sufficient PBS solution to attempt to obtain a final target concentration of approximately 50 mg/mL filtered. To determine the actual dissolved concentration, 1 mL of each filtered test solution was dried and the recovered material weighed so that the dose to each animal could be accurately determined. The 13.6-year-old dialyzer extract achieved a maximum dose due to solubility of 270 mg per rabbit, whereas the synthetic degradation products achieved a maximum dose of 470 mg per rabbit.

Healthy female New Zealand rabbits (2.5-2.7 kg) were used for the in vivo experiments. The eyes of all the animals were examined with slit-lamp biomicroscopy and indirect ophthalmoscopy following 1% of tropicamide mydriasis 1 week before and 2, 4 to 6, and 24 hours after doses were administered. Four animals each were randomly assigned to 1 of 3 groups. Group 1 received 13.6-year-old dialyzer extract; group 2, synthetic model degradation compounds; and group 3, PBS control solution. Each animal received a 10-mL dose through the right auricular vein at the rate of 2 mL/min or less.

Seven of 9 patients who received dialysis on the epidemic day met the case definition. All case-patients underwent dialysis with conventional capillary flow cellulose acetate dialyzers with the same lot number. A serial number trace revealed that the dialyzers used on the case-patients were manufactured in April 1985. The 2 non–case-patients underwent dialysis with a newer (<1 year old) conventional, capillary-flow, cellulose acetate dialyzer from the same manufacturer. Case-patients ranged in age from 25 to 65 years (median, 51 years) and 5 were women. Duration of hemodialysis using the 11.5-year-old implicated dialyzers ranged from 5 to 167 minutes (median, 100 minutes). Six case-patients were removed from the implicated dialyzer after the blood leak alarm activated. The seventh case-patient was removed from the implicated dialyzer after only 5 minutes of treatment to preempt the blood leak alarm. These case-patients continued undergoing dialysis using the newer, but the same model, dialyzer as the non–case-patients without further incidents. No non–case-patients had any visual or hearing complaints.

Signs and symptoms of case-patients are reported in Table 1. When we compared case-patients with non–case-patients, case-patients were significantly more likely to have diminished vision and hearing, headache, and conjunctivitis (Table 2). Corneal opacification was noted in 4 case-patients (57%). Other neurologic and ophthalmologic findings in case-patients included 1 case each of optic neuritis, optic atrophy, uveitis, seventh cranial nerve palsy, and vertical nystagmus.

Initial signs and symptoms in case-patients appeared 7 to 48 hours (median, 9 hours) after hemodialysis on the epidemic day. In all case-patients, headache and conjunctivitis presented within 24 hours, while decreased vision, decreased hearing, tinnitus, and vertigo presented 24 to 48 hours after hemodialysis (Table 3).

There was 1 case-patient for each of the following visual acuity scores: no light perception, light perception only, hand motion detected, and more than 20/200 vision. Three case-patients had visual acuity scores classified as count fingers. Audiometry revealed hearing loss was profound in 2 case-patients, moderate to severe in 4, and mild to moderate in 1.

After onset of symptoms, 2 case-patients suffered cardiopulmonary arrests during hospitalization after the event, 1 was 24 hours after the epidemic day and the other 5 days later. Both case-patients required transfer to the intensive care unit for further treatment; 1 patient died. These case-patients had a history of heart disease, and the cardiac arrest could not be clearly linked to the events on the epidemic day. However, the incidence of cardiac arrest during or after dialysis was significantly higher for case-patients who received dialysis on the epidemic day than for patients who received dialysis during the prior 23 months at hospital A (2/7 vs 13/2099; P=.001). Postarrest mortality during hospitalization was not significantly different for patients who underwent dialysis during these 2 periods (1/2 vs 7/52; P=.28).

Case-patients' median white blood cell count predialysis and postdialysis increased from 9.5×10⁹/L to 28.4×10⁹/L. There were no other significant changes in blood chemistry or hematologic values from hemodialysis on the epidemic day until 6 days after the epidemic day. Five case-patients had blood cultures within 5 days of dialysis on the epidemic day; all were negative. Head magnetic resonance imaging was performed on 1 case-patient and showed no acute changes.

There were 43 different drugs administered to the 7 case-patients, no drug was common to all of the case-patients, and there were no significant differences in medications given to case- and non–case-patients.

Long-term follow-up for the case-patients is shown in Table 4. Case-patients 1, 2, and 3 nearly fully recovered within 1 month of the event. For the others, some signs and symptoms lingered and some resolved. Five of the case-patients died of cardiac or renal disease, which could not be fully attributed to the event.

The FDA retrieved dialyzers ranging in age from 0 to 13.6 years (storage conditions varied and were undocumented). We examined dialyzer membranes from all the suppliers of cellulose acetate dialyzers marketed in the United States. Degradation rates were similar between membranes of different suppliers.

Gel-permeation chromatography was used to evaluate changes in molecular weight of the dialyzer membrane polymer. After 13.6 years in storage, the number-average molecular weight of the dialyzer membrane polymer decreased from 40,000 to 30,000 g/mol (25% by weight; Figure 1), and the polydispersity index (weight-average molecular weight divided by number-average molecular weight) increased from about 1.7 to 1.9.

To evaluate dialysis membrane degradation by deacetylation, we measured the total acetyl content. Of the 40 membranes evaluated that were 0 to 13.6 years old, only 2 (aged 5.8 and 13.6 years) were significantly deacetylated (31% and 33% by weight, respectively). All of the other dialyzer membranes showed only a gradual decrease in acetyl content with age in the range of 40% to 37% by weight.

The extractable component of the dialyzers depends on age and acetyl content. For newer dialyzers, typically less than 1 mg of total material was recovered from the fibers. For older dialyzers, the amount and type of material recovered depended on both the age and the acetyl content of the fibers. For example, a 13.3-year-old dialyzer with a 38.0% weight acetyl content yielded 40 mg of extractable material, whereas a 13.6-year-old dialyzer with 33.0% weight acetyl content yielded more than 6000 mg of extractable material.

Infrared spectra of purchased high-purity, cellulose-acetate resin, the extract from a 13.6-year-old dialyzer, and the synthesized, degraded model compounds are shown in Figure 2. Both the old dialyzer extract and the synthesized degraded model compounds spectra were consistent with virgin cellulose acetate, but each spectra had some differences due to different mechanisms of degradation. The old dialyzer extract showed a diminished peak at 1375 cm⁻¹ (acetate stretch), consistent with deacetylation. Since deacetylation events produce more O-H (oxygen-hydrogen) bonds, the broader absorption at 3347 cm⁻¹ is attributed to increased O-H stretching in this material. By contrast, the synthesized degraded model compounds, as well as the degradation products identified by Oba et al⁹ in 1984, did not show nearly as much diminished acetate stretch at 1375 cm⁻¹. Thus, the acetyl content of this material is considerably higher than the old dialyzer extract. This is also evident in the less broad O-H stretching absorption in this spectrum. The model compound and the material identified by Oba et al⁹ also had an additional peak at 1625 cm⁻¹, which is characteristic of the carboxylate ion group absorption.

Since the older deacetylated dialyzers gave the most water-extractable material, we used the extract from a 13.6-year-old dialyzer with 33% weight acetyl content in the group 1 animal tests. We used the synthesized degraded model compounds in the group 2 animal tests.

Both the extract from the 13.6-year-old dialyzer and the model compounds caused eye injuries similar to the case-patients at hospital A in our in vivo experiments (Table 5).

This outbreak was severe and unusual because of the serious neurologic signs and symptoms and the association with old (ie, >10 years) dialyzer membranes. All patients exposed to the old dialyzers developed acute onset of diminished vision and hearing. The adverse exposure was associated with severe morbidity and mortality as 4 case-patients never fully recovered and 5 of 7 case-patients died within 13 months. Once this unusual outbreak was traced epidemiologically to old dialyzers, we sought to identify and isolate the inciting agent and to reproduce the adverse effects in our animal experiments.

When blood comes in contact with dialysate through a hollow-fiber membrane, material-degradation products can be directly transferred to the blood. Since the membranes are thin (30 µm), and porous (30% porosity),¹⁵ there can be a rapid mass transfer. One of our case-patients was connected to the dialyzer for only 5 minutes yet developed severe sequelae. The wide bore of the fiber lumen (200 µm) will allow any potential contaminant from the membrane itself to be rapidly returned directly back to the patient in the venous catheter.

Cellulose acetate-hemodialysis fibers degrade over time as indicated by a decrease in fiber integrity; number-average molecular weight decreases, and the polydispersity index increases with dialyzer age. The events at hospital A indicate that fiber integrity was compromised because the blood leak alarms were activated on most case-patients. The rate of blood leakage was much higher than expected in a dialysis unit of similar size (6/7 vs 1/1265; P<.001).¹⁶ Further evidence of membrane degradation included the decrease in number-average molecular weight over time from 40,000 to 30,000 g/mol in the dialyzers tested (Figure 1). This decrease in molecular weight can be attributed to either chain scission or deacetylation. For deacetylation alone, a change of that magnitude would require removal of more than 50% of the acetyl groups. Since even the oldest dialyzers did not have this much deacetylation, most of the molecular weight decrease occurred because of chain scission of the cellulose polymer. Chain scission, which can occur by a number of mechanisms (eg, oxidation, hydrolysis, or exposure to radiation), results in cleavage of the 1,4-β-D glycosidic linkage in the polymer.¹⁷ Thus, when chain scission occurs, the polymer chain is transformed into 2 lower-molecular weight pieces, which reduces the average molecular weight of the entire membrane.

Our results indicate that the combination of chain scission and deacetylation is a relatively rare event (only 2 of 40 dialyzers had acetyl content reduced by >10%). Most of the retrieved dialyzers only had a minor amount of deacetylation over the period studied, which is a result of the manufacturers' practice of conditioning the membranes to about pH 5 to minimize deacetylation.¹⁸ The additional recommendation that dialyzers be stored under cool and dry conditions will also minimize degradation.^19-23 Under such conditions, chain scission is the major degradation route, but deacetylation can occur over time if the dialyzer membrane was inadvertently exposed to a solution of pH less than 4 during manufacturing. At ambient temperatures, these reaction rates are slow, so long periods are required to allow the deacetylated degradation products to accumulate. Our animal studies show that material released by membranes degraded through both chain scission and deacetylation most likely produced some or all of the symptoms observed in patients at hospital A.

There has been a previous report of similar eye injuries associated with use of cellulose acetate dialyzers.⁹ However, the mechanism of the previously reported outbreak is different from the mechanism identified at hospital A. The degradation products identified by Oba et al,⁹ as well as by our own synthetic model compounds, were produced by a combination of chain scission and oxidation, not deacetylation. The oxidative stress could have resulted in either oxidation of a pendant group or a ring-opening reaction as indicated by the appearance of the carboxylate ion group in the degradation products' spectrum. If the rings in some of the monomer units of cellulose acetate open, the chain flexibility of the molecule is considerably increased. Either this change or deacetylation (as was the case at hospital A) can increase the aqueous solubility of the degradation products, making these products more likely to be leached into the blood during dialysis. Oba et al⁹ identified the suspect impurity in the material as hemicellulose. The hemicellulose was more susceptible to chain scission and oxidative stress during membrane fabrication^24,25 and sporadic and transient patient injuries occurred immediately on first use of these products. Since that incident, dialysis manufacturers have taken great pains to avoid contamination of dialysis membranes with hemicellulose. These type of injuries are now a rare occurrence. In contrast, the material that most likely caused the event at hospital A resulted from long-term degradation due to both chain scission and deacetylation, and produced much more severe and persistent injuries in every exposed patient. The events at hospital A were unusual since a combination of deacetylation, chain scission, and long storage time are unlikely combinations. There were no previous incidents like this reported for this particular model of dialyzer or lot number.

Regardless of the factors that caused degradation, our data suggest that cellulose acetate degrades with age, and the combination of chain scission and deacetylation will produce large quantities of leachable degradation products, which can cross into the dialyzer blood compartment and cause severe neurologic symptoms in humans. Before this event, no outdating or expiration dating on dialyzers was required by the FDA. As a result of the event at hospital A, the FDA and Centers for Disease Control and Prevention sent a letter to all US dialysis centers advising that all dialyzer stock be properly rotated (first in, first out) and that any suspected old dialyzers be cleared by the manufacturer before use.²⁶ In addition, the FDA now requests a date of manufacture to be indicated on the labels of all new dialyzers.²⁷ Also, the FDA is in the process of developing shelf-life criteria for these devices in cooperation with industry, standards organizations, and health care providers.

The events at hospital A vividly demonstrate that material degradation is not only a performance issue, but also a safety issue. Clinicians should be aware of the age of the medical devices that they use and should not attempt to use products (especially those containing cellulose acetate) beyond the expiration date. The device manufacturer should be consulted if there is any doubt whether to use an old medical device. Finally, this outbreak illustrates the value of combined epidemiologic and laboratory investigations of unusual adverse events, which can lead to identification of the source and prevention of the events at other facilities.