Municipal water used to prepare dialysate must be purified to ensure that no contaminants remain that would be toxic to patients receiving dialysis. Reverse osmosis is the most common procedure to remove ions, but deionization systems, alone or after reverse osmosis, were used by nearly one half of the hemodialysis centers in the United States in 1991 [1]. Paradoxically, these deionization systems pose an additional hazard. When their ion exchange resin becomes exhausted, toxic ions previously removed from municipal water and bound to the resin may be displaced into the effluent stream of purified water [5, 6]. Johnson and Taves [8] found that high concentrations of fluoride could be released in this manner into effluent, but no cases of acute fluoride intoxication attributable to this mechanism have been reported.

We recently investigated in a hemodialysis unit an outbreak of acute illness and death caused by fluoride intoxication. We describe the epidemiologic characteristics of the outbreak and the pattern of deionization system use that resulted in massive fluoride efflux from an exhausted resin.

The Outbreak

On 16 July 1993, several patients treated at a long-term hemodialysis unit became ill during or soon after hemodialysis. The predominant symptoms were severe pruritus, headache, nausea, and chest or back pain. One patient had cardiac arrest immediately after completion of dialysis. Patients continued to receive dialysis treatment until that afternoon, when the unit was notified that two other patients had had cardiac arrests after leaving the unit. Dialysis treatments were stopped during the third shift of patients, and patients who had received hemodialysis on 16 July were brought to the University of Chicago Hospital for evaluation. Those who had any symptoms were admitted to the hospital, and patients whose scheduled treatment had been interrupted resumed dialysis at the hospital's acute care dialysis unit. An investigation was then initiated.

Methods

Top Methods Results Discussion Author & Article Info References

Clinical and Epidemiologic Characteristics of the Patients

We interviewed patients who received long-term hemodialysis treatment in the unit and reviewed their records to identify symptoms during the period 26 April to 16 July 1993, when a temporary water purification system was used. We considered that patients had severe hemodialysis-associated illness if within 5 hours after the start of dialysis either severe pruritus developed or cardiac arrest ensued, or if at least three of the following symptoms developed: burning or feverish feeling; headache; nausea or vomiting; syncope or near syncope; pain in the chest, back, or abdomen; or diarrhea. We reviewed medical and dialysis unit records to determine patient age, sex, primary underlying illness, results of physical examination and clinical laboratory tests, number of years of long-term hemodialysis, duration of the hemodialysis session on 15 or 16 July, hemodialysis blood flow rate, type of dialyzer, and dialyzer reuse. We considered that patients had cardiac disease if they had coronary artery disease, cardiomyopathy, valvular heart disease, or clinically significant electrocardiographic abnormalities such as bundle-branch block. For statistical analysis, we compared each continuous variable using the t-test and the Wilcoxon test, and proportions were compared using the Fisher exact test [9].

Dialysis Unit Design and Water Supply

We surveyed the municipal water supply, dialysis water purification systems, types of dialyzers and dialysis machines, dialysate preparation, and dialyzer reprocessing procedures. We also reviewed records of daily municipal water treatment and periodic chemical analyses by the City of Chicago Department of Water during 7 June to 16 July 1993.

Sample Collection and Analysis

Blood specimens were collected on 16 July from all surviving patients who had had hemodialysis treatment that day. Complete blood count and routine blood chemistry testing were done. A portion of each sample was stored at –70°C and later assayed for fluoride. Samples of purified water used for dialysate preparation were collected on 16 July 1993 for analysis of 18 chemicals for which maximum allowable levels were established by the Association for the Advancement of Medical Instrumentation [10]. Samples of municipal water, purified water used to prepare dialysate, acid concentrate, bicarbonate concentrate, and bicarbonate dialysate also were collected on 16 July and assayed for fluoride concentration using an ion-selective electrode system (Orion Research, Cambridge, Massachusetts) [11]. Additional samples of purified water from each of the two sets of deionization tanks in the room (room 1) where sick patients had received hemodialysis treatment were collected on 17 July after approximately 1000 L of water had flowed through the systems when the dialysis machines were tested. These samples and dialysate collected from room 1 on 16 July were assayed by the Illinois Department of Public Health for fluoride and for toxic organic compounds by gas chromatography-mass spectrometry and by ultraviolet spectrometry. Postmortem serum specimens were obtained and assayed for fluoride by the Cook County Medical Examiner's Office.

Model Deionization System

To confirm that sustained high levels of fluoride in effluent from the deionization system in the dialysis unit could have been released by operating the system after the resin was exhausted, a laboratory model of the system was constructed. An activated carbon tank and four newly regenerated deionization tanks were supplied by the water treatment contractor who maintained the system in the dialysis unit. The four tanks were arranged in series as follows: first, a cation resin tank containing sulfonated polystyrene (Amberlite IR-120 Plus, Rohm and Haas, Philadelphia, Pennsylvania) in position 1; next, an anion resin tank containing a type II quaternary ammonium resin (Amberlite IRA-410) in position 2; then two mixed-bed tanks containing both cation and anion resins in positions 3 and 4. This system was used to treat 27 500 L of municipal water, a volume estimated to be the same as that which flowed through each of the two sets of deionization tanks in room 1 from 26 April to 7 June 1993. A second deionization system was then constructed with newly regenerated tanks in positions 1, 2, and 4; the mixed-bed tank from position 4 in the first deionization system was moved into position 3, as had been done in the dialysis unit. The second deionization system was operated beyond exhaustion of the resins to simulate use from 7 June to 16 July 1993. During operation of the two model deionization systems, effluent was collected to measure fluoride concentration and resistivity, which decreases as electrolyte concentration increases.

Results

Top Methods Results Discussion Author & Article Info References

Clinical and Epidemiologic Characteristics of the Patients

At the time of the outbreak, 56 patients were receiving long-term, high-flux hemodialysis. One half of the patients were treated on Monday, Wednesday, and Friday, and the other half were treated on Tuesday, Thursday, and Saturday. Twenty-one patients received at least a portion of their scheduled dialysis on 16 July before treatments were stopped. Nine of the 21 patients treated on 16 July and 3 of the 25 patients treated on 15 July were classified as cases. Review of dialysis unit records for the previous 12 weeks identified only 1 additional patient who met the case definition. That patient, who had a cardiac arrest on 9 June, was not considered to be part of this outbreak and thus we did not include him as a case-patient.

Table 1 summarizes the characteristics of both the 12 patients classified as case-patients and the 20 patients not considered cases who were treated during the same shifts. Symptoms other than cardiac arrest began 1 to 3 hours after the start of hemodialysis in all cases and resolved by the next morning. Case-patients who reported pruritus described it as the most severe they had ever experienced. Diphenhydramine hydrochloride was taken by 5 patients, none of whom noted any relief. Pruritus lasted more than 6 hours in the surviving patients who became ill on 16 July and less than 3 hours in the three patients who became ill on 15 July.

All six surviving patients who became ill on 16 July were hospitalized that evening. None had new abnormal findings on physical examination or complete blood count, and none had hyperkalemia. Three patients had low serum calcium concentrations of 7.8 to 8.3 mg/dL (1.95 to 2.54 mmol/L).

Patients in the dialysis unit received high-flux dialysis in two treatment rooms using individual self-proportioning, volumetrically controlled ultrafiltration machines (COBE Centrysystem 3, COBE Laboratories, Lakewood, Colorado). The 12 patients who became ill received dialysis treatment for a mean of 2.8 ±0.7 hours (range, 1.1 to 3.9 hours) using any of three new (four cases) or reprocessed (eight cases) hemodialyzers. The concentration of potassium in the bicarbonate dialysate for each case ranged from 1 to 4 mEq/L (1 to 4 mmol/L), and the concentration of calcium was 3 to 4 mEq/L (1.5 to 2.0 mmol/L).

All 12 patients who became ill received hemodialysis in the same room, designated room 1. The rate of illness in patients receiving treatment in that room during the last two shifts on 15 July was 50% (3 of 6), and the rate during the three shifts on 16 July was 100% (9 of 9). Compared with these 15 patients, none of the 17 patients treated in the other room, designated room 2, during the same five shifts became ill with symptoms that fulfilled the case definition (P < 0.0001; Fisher exact test). The cohort of 15 patients treated in room 1 did not differ significantly from the cohort of 17 patients treated in room 2 in age, sex, underlying illness, years of long-term hemodialysis, duration of hemodialysis session, hemodialysis blood flow rate, type of dialyzer, or dialyzer reprocessing (all P values > 0.5).

To assess risk factors for death, we used case–control methods to compare the three patients with fatal ventricular fibrillation on 16 July with the six patients with nonfatal illness that day. Patients who died were substantially older (mean age, 72 ±14 years compared with 51 ±13 years; P = 0.07 by t-test and Wilcoxon test) and had longer hemodialysis treatment (3.1 ±0.5 hours compared with 2.7 ±0.7 hours; 0.1 < P < 0.2 by t-test and Wilcoxon test). We found no clinically or statistically significant differences in the other variables listed previously or in hemodialysis shift or prevalence of cardiac disease.

Dialysis Unit Design and Water Supply

The central treatment area of the dialysis unit had been closed since 26 April 1993 for renovation. During this period, seven temporary dialysis stations were erected in two rooms (room 1 and room 2) that were not being renovated. Room 1 contained three stations and room 2 contained four stations. During renovation, four shifts of patients (seven patients per shift) were scheduled to have dialysis each day, Monday through Saturday.

Municipal water that was to be purified to prepare dialysate was piped directly to rooms 1 and 2. In each room, a separate, identical, temporary water purification system was installed that consisted of two sets of carbon and deionization tanks. Each set contained in sequence a carbon tank to remove chlorine and organic chemicals, then one cation, one anion, and two mixed-bed tanks to remove ions. The effluent from each set was combined and distributed to the dialysis machines in that room. Resistivity monitors, provided by the water treatment contractor, were located on the effluent lines of the two sets of deionization tanks in each room. The monitors were set to change illumination from a green to an amber light when the quality of the treated water decreased to less than the standard of 1 megaohm/cm [10]. In contrast, the monitoring apparatus previously supplied to the dialysis unit had only an amber light that remained lit as long as resistivity remained greater than 1 megaohm/cm. On 3 June 1993, the contractor replaced the cation, anion, and the first mixed-bed tank from each set in room 2 with newly regenerated tanks. The second mixed-bed tank from each set was not replaced but moved up one position, becoming the first mixed-bed tank. The newly regenerated mixed-bed tank then was positioned as the last tank in the set. On 7 June the same changes were made in the two water purification sets in room 1, after approximately 55 000 L of municipal water had been treated.

When the dialysis unit was surveyed on 16 July, the green light was out and the amber light lit on the resistivity monitors in rooms 1 and 2. The date on which the resistivity lights changed from green to amber could not be determined retrospectively. After the deionization tanks were replaced in June, approximately 52 000 L of water in room 1 (1500 L/d) and 74 000 L in room 2 (2000 L/d) had passed through the water purification systems.

Bicarbonate dialysate was prepared at each dialysis machine by mixing in set proportions water from the room's deionization system with commercial acid concentrate and with a solution of commercial bicarbonate powder mixed in purified water from the dialyzer reprocessing area. If supplemental calcium or potassium was ordered, they were added manually to the container of acid concentrate at the dialysis machine.

Department of Water records showed no increase in fluoride or other chemicals added to municipal water during the week before the outbreak. The concentration of fluoride was about 50 µmol/L on 15 and 16 July 1993.

Chemical Analysis

Samples of purified water used for dialysis were within the standards set by the Association for the Advancement of Medical Instrumentation for all chemicals tested except the following: The concentrations of fluoride in rooms 1 and 2 were 1182 µmol/L and 51 µmol/L, respectively (standard, 10.5 µmol/L); the concentration of calcium in room 2 was 7.5 mg/L (standard, 2 mg/L); and the concentration of magnesium in room 2 was 30.6 mg/L (standard, 4 mg/L). Fluoride was not detected in other components of dialysate, and dialysate in rooms 1 and 2 had elevated concentrations of fluoride corresponding to the levels in purified water (Table 2). Assay results were confirmed by the Illinois Department of Health and the Centers for Disease Control and Prevention. Purified water collected separately on 17 July from each of the two sets of deionization tanks in room 1 had fluoride concentrations of 598 µmol/L and 1083 µmol/L. No toxic organic compounds were detected at a threshold level of one part per billion in the purified water or dialysate from room 1. Resistivity of purified water collected on 16 July from room 1 and room 2 was 4630 ohm-cm and 3400 ohm-cm, respectively.

Fluoride Levels in Serum

Serum specimens from the nine case-patients who became ill on 16 July had elevated fluoride concentrations that were inversely related to the interval from hemodialysis until death or serum specimen collection (Table 3). Serum specimens obtained 16 July from the 12 patients who had hemodialysis treatment in room 2 all had concentrations of fluoride of less than 1 µmol/L.

Model Deionization System

When a set of four newly regenerated deionization tanks was operated to simulate use in room 1 during the period 26 April to 7 June 1993, fluoride was adsorbed and then displaced first from the anion tank (position 2) and then from the first mixed-bed tank (position 3). Fluoride efflux from the first mixed-bed tank peaked at 1315 µmol/L, when approximately 23 500 L of water had been treated and resistivity of effluent from the tank had decreased to less than 4000 ohm-cm. After treatment of 27 500 L of water, resistivity of effluent from the second mixed-bed tank (position 4) decreased from 16 megaohm-cm to 54 000 ohm-cm. Fluoride concentration in effluent remained 1 µmol/L or less, indicating that the second mixed-bed tank retained fluoride adsorbed by the deionization system during the cycle of use.

A second set of four deionization tanks, including the fluoride-laden mixed-bed tank from the first cycle, was operated to simulate use in room 1 during the period 7 June to 16 July 1993. Fluoride displacement into effluent from the fourth tank began when 24 500 L of municipal water had passed through the system and resistivity was 70 000 ohm-cm. Fluoride concentration peaked at 1683 µmol/L when resistivity was 3000 ohm-cm and 26 000 L of water had been treated. Centered around this peak, fluoride efflux was sustained at a concentration of 1000 µmol/L or more in 600 L of effluent and at a concentration of 500 µmol/L or more in 2400 L of effluent.

Discussion

Top Methods Results Discussion Author & Article Info References

Fluoride is a low-molecular-weight anion that is absorbed in the gastrointestinal tract, diffuses from plasma into interstitial fluid, is rapidly concentrated in bone, and is eliminated primarily by renal excretion [12]. Like other low-molecular-weight solutes, fluoride diffuses readily across dialyzer membranes [13, 14]. Based on evidence that fluoride reduces tooth decay, fluoride routinely is added to municipal drinking water to achieve a concentration of about 50 µmol/L [15, 16]. Persons with normal renal function generally have a serum concentration of fluoride of less than 2 µmol/L; the concentration may be several times greater in persons with chronic renal failure or in those taking supplemental sodium fluoride orally [12, 17-21]. Uptake of large amounts of fluoride into blood, usually from ingestion or from skin burns with hydrofluoric acid, can cause acute fluoride intoxication, a life-threatening illness in which peak fluoride concentrations in serum have reached or exceeded 500 µmol/L [14, 22, 23]. The only outbreak of acute fluoride intoxication previously reported in patients receiving hemodialysis was caused by accidental overfluoridation of municipal water that was not treated by a deionization system or reverse osmosis before use in dialysis [24].

Clinical and laboratory findings confirm that the outbreak of hemodialysis-associated illness described in this report was caused by fluoride. First, serum concentrations of fluoride were markedly elevated in all of the sick patients tested and not in other patients. Second, the clinical features of the outbreak illness closely match those of fluoride intoxication. Pruritus, chest pain, nausea, vomiting, and diarrhea were major symptoms in a previously reported outbreak [24, 25], and delayed cardiac arrest due to ventricular fibrillation was noted in humans and in an experimental animal model of fluoride intoxication [22-29]. Third, our investigation showed a source of fluoride that explained the temporal and spatial clustering of cases in one of two hemodialysis treatment rooms. The water used for dialysis was the only component of hemodialysis that was different between the two rooms, and the deionization system in the room where patients became ill was releasing high concentrations of fluoride.

Fluoride was released from the deionization system in room 1 because of continued use of the system after the ion exchange resin was exhausted. Deionization systems contain synthetic ion exchange resins that release electrostatically bound hydrogen ions for cations and hydroxyl ions for anions. As exchange sites on the resin become depleted, fewer ions are removed, the ion content of the effluent increases, and resistivity decreases [30]. Continued use of the resin to treat municipal water causes low-affinity ions, such as fluoride, already bound to the resin to be displaced into the effluent by ions with higher affinity [30, 31]. In retrospect, displacement of large amounts of fluoride from the deionization tanks in room 1 apparently began during the afternoon of 15 July, when three patients receiving hemodialysis treatment developed severe pruritus. Fluoride displacement probably peaked on 16 July, when symptoms developed in all of the patients receiving treatment in room 1 and three died of fluoride intoxication.

It is surprising that the outbreak of fluoride intoxication in room 1 was not preceded by outbreaks in room 2, where the deionization system twice had been used beyond exhaustion. The most probable explanation is that the amount of fluoride inadvertently accumulated during two consecutive 6-week cycles then displaced in one sustained surge from the deionization system in room 1 was considerably greater than the amount adsorbed and then displaced separately during each 6-week cycle in room 2. Based on the volume of water treated in each room and on the resin capacity, we estimate that 5.2 mol of fluoride had been retained by the deionization system in room 1, compared with 3.1 and 2.3 mol, respectively, during each cycle in room 2. We wondered whether lesser fluoride efflux in room 2 might have caused mild illness, but we found no evidence of this by retrospective interviews or review of dialysis records. If symptoms occurred, they apparently were not distinctive against the background incidence of nausea, headache, mild pruritus, and other complaints during hemodialysis [2].

The outbreak we investigated was caused by errors in maintenance of the deionization system and by an unfortunate coincidence that the schedule of tank replacement at 6-week intervals maximized fluoride efflux from the system in room 1. Several lessons about the use of deionization systems for hemodialysis can be learned from this investigation. First, adherence to existing guidelines for calculating deionization capacity and monitoring deionization systems can prevent inadvertent use of these systems beyond exhaustion [31]. In particular, audible and visual alarms, as recommended by the Association for the Advancement of Medical Instrumentation [10], are necessary to reliably alert dialysis unit personnel that resistivity has decreased to less than the minimum standard. Second, it seems relevant that this problem occurred during use of a temporary system. Because changes in practice can require personnel to handle unfamiliar tasks or equipment, it is especially important at such times to ensure that personnel are appropriately trained and that adequate attention is devoted to monitoring critical equipment and procedures. Third, even in settings such as hemodialysis, where patient complaints are common, an increase in the frequency or severity of common symptoms may indicate a new and potentially serious problem. Finally, awareness of the potential hazards of deionization systems and of appropriate safety practices is important because of the widespread use of these systems for hemodialysis and the lethal consequences when deionization systems fail.