Dialyzer Repair: A Disastrous Exercise in Cost Effectiveness

The repair of capillary membrane dialyzers that are found to be leaking during production has always been part of standard operating procedure in the manufacturing of these devices. It was introduced more than 30 years ago and involves a technique known as “the reverse air pressure test.” Essentially, after the capillary bundle is housed and potted, the dialyzer can be pressure tested for capillary leaks in one of several ways, including a simple measurement of the rate of pressure drop after inflation of the blood compartment. Whichever method is used, a percentage of dialyzers will fail the pressure test and are then either scrapped or repaired. The repair of a leaking capillary dialyzer requires identification and sealing of the leaking fibers. One way of achieving this is by filling the blood compartment with a liquid and then applying air pressure to the dialysate compartment with one port closed. The leaking fiber is identified at both ends by the emergence of air bubbles from its lumen. This fiber can then be marked at both its ends; hence the descriptive name “the reverse air pressure test.” Once the leaking fiber has been identified, it can be sealed with the potting material used for routine potting of the bundle. This manual technique requires a skilled person who is able to seal an individual fiber at both its ends by introducing the sealing material into the lumen of the fiber with a fine glass needle. However, if several fibers adjacent to the marked one are also accidentally sealed, nothing in the standard operating procedure requires the dialyzer to be scrapped; the resultant loss of surface area in a capillary dialyzer with more than 10,000 fibers would be undetectable (Figure 1). The frequency with which accidental sealing of intact fibers adjacent to the leaking fiber occurs as a result of the potting sealant spilling out of the lumen of the leaking fiber is presumably a function of manual operator skill and experience.Figure 1: Schematic diagram of possible explanation of how dialyzer repair with PF5070 can allow PF5070 to enter the patient’s blood stream during dialysis, despite adequate rinsing with air and saline.As no bubble will emerge from the open end of the lumen of the intact fiber when the dialyzer is inverted and pressure reapplied to the dialysate compartment, the possibility of having intact fibers sealed only at one end is very high. In addition, there is no way that these partially sealed fibers can be identified with certainty. Thus, whereas it is possible to seal the individual leaking fiber and control it with a repeat standard pressure test, as long as no quality assurance tests exist for detecting partially sealed intact fibers, the scene is set for the disaster which befell the Baxter Company in September 2000, when to increase productivity and reduce the 10% to 20% dialyzer scrap rate, the company decided to introduce a dialyzer repair program into its manufacturing process for the Althane cellulose diacetate capillary dialyzer at its recently acquired plant in Ronneby, Sweden. 1 To perform the “reverse air pressure test” with cellulose diacetate, it is necessary to use a hydrophobic fluid, as the pores of the partially hydrophobic cellulose acetate membrane are prevented from collapsing during the sterilization process by glycerol, which would dissolve in a hydrophilic liquid with resultant collapse of the pores. The state of the art hydrophobic liquid was a fluorocarbon manufactured by 3 M and known as PF5070. Previously, Freon had been used, but its use had been discontinued for ecologic reasons as its vapor was contaminating the environment. PF5070 was an apparently nontoxic fluid, which had no adverse effects on cells and was considered to be completely safe from a biocompatible viewpoint, although the 3 M data sheet states that it is “for industrial use only, not intended for use as a medical device.”2 Nevertheless, Gambro AB had been using it for years, presumably to “repair dialyzers”, and no untoward reactions had been reported. 3 The possibility that this claim is overoptimistic will be discussed later. The use of PF5070 to aid in the repair of the Althane Dialyzer has resulted in more than 50 known deaths of patients on dialysis from many countries, including Spain, Croatia, the United States, Taiwan, and Columbia; it has also attracted two invited editorials in leading journals. 4,5 Both editorials are remarkable for what they do not state and also for their attempts to share the responsibility for this epidemic between the manufacturer and the end user. A more informed explanation of the mechanisms involved in this tragedy would demonstrate that the end user, a dialysis nurse, technician, or doctor, could not in any way be held responsible, even if inadequate air rinsing was the cause, as a standard low pressure saline rinse would not remove hydrophobic PF5070 residues. Furthermore, the possibility that there were many patients exposed to the estimated 30,000 “repaired” Althane Dialyzers and 150,000 “repaired” Gambro Dialyzers (based upon the Gambro statement that PF5070 was used in only 0.3% of its dialyzer production over 5 years 2) that did not die directly from exposure to PF5070, but may have sustained gas emboli in the cerebral or coronary arteries with consequent infarction of the affected cerebral or myocardial tissue, must be considered. The causes of death were finally established after autopsies of Croatian patients showed massive gas formations in the right side of their hearts and pulmonary-alveolar capillary beds. 6 The discovery of residual perfluorocarbon (PF5070-perfluoroheptane C7F16) in certain lots of the Althane dialyzer indicated that this volatile hydrophobic fluid was responsible for the gas formation. The underlying hypothesis explaining these findings requires first that PF5070 is insoluble in plasma water and, hence, fails to traverse the pulmonary alveolar capillary membrane (according to the data sheet of the manufacturer, PF5070 is completely insoluble in water 2). Thus PF5070, once it has entered the circulation, cannot equilibrate across the pulmonary alveolar capillary membrane, and, in consequence, the vapor pressure of PF5070 contributes to the total vapor pressure in the pulmonary capillary blood, but does not make a similar contribution to the vapor pressure in the alveolar air. Second, the vapor pressure of the PF5070 must exceed a limit of approximately 55 mm Hg at body temperature to produce fatal gas embolization. 7 The vapor pressure of PF5070 at 20°C is ca. 79 mm Hg;2 at body temperature, the vapor pressure of PF5070 is even higher. Consequently, the total gas tension in pulmonary capillary blood (containing PF5070) exceeds the total tension of alveolar gases (atmospheric pressure). Bubbles of oxygen, carbon dioxide, nitrogen, PF5070, and water vapor progressively form in the pulmonary capillary bed. Eventually, after some hours, the total volume of gas in the pulmonary capillary bed exceeds the bed’s capacity to expand, and then retro filling of the pulmonary artery and the right heart with this gas occurs. A simplified description of the “reverse air pressure test” was published in El Pais on November 6, 2001, with a clearly understandable diagram to aid the lay reader in understanding a reported interview with Dr. Jose Divino, Medical Director of Baxter Europe. 8 The diagram indicates that if more than five fibers (presumably judged by the rate of loss of pressure) were found to be leaking in the routine air pressure test, the device was discarded. However, it was estimated in the El Pais report that as many as 10% of the dialyzers had leaks of less than five fibers. 8 The leaking fibers in these dialyzers were repaired, and the dialyzer was then reintroduced into the production lot (presumably with the same lot number). “Dialyzer repair” is less common today. Baxter used it only in its Ronneby plant, which represented less than 30% of its global production. 1 As previously stated, the “dialyzer repair” process requires that the leaking fiber or fibers are identified and then sealed at both ends, and there is a probability that both ends of one or more intact fibers adjacent to the leaking fiber were not sealed, i.e., one end of the fiber was sealed, but the other was not (Figure 1). Under such conditions, there is poor accessibility to the PF5070 that is trapped within the fiber. That trapped fluid will only come out very slowly with any rinsing method, whether it be air, solvent, or saline, because the interface for transport comprises only of the cross sectional area of the fiber and that of the break, if any, within the fiber. Thus, it is likely that any residual PF5070 trapped in a fiber sealed at one end would be incompletely removed by an air rinse unless it was continued for a sufficiently long time. A saline rinse would be ineffective because it would have to rely on dissolution of the trapped PF5070 into the aqueous stream, which cannot occur as the hydrophobic PF5070 is insoluble in water. 2 (There is a theoretic possibility of ultrafiltration from dialysate side to lumen, thereby displacing PF5070 out of the lumen, but that too is likely to be very slow). Thus a standard rinse/priming procedure, which is all the end user could be expected to do without explicit warning by the manufacturer to rinse more extensively because of possible contamination, would be highly unlikely to remove trapped PF5070 that remained after an air rinse. It would appear, therefore, to be incorrect to involve the end user in sharing the responsibility for these deaths as was implied by both editorials. 4,5 However, the prolonged contact of blood, and especially albumin, with the open ended fibers would result in the slow passage of PF5070 into the lungs, with a consequent delay of several hours from the end of dialysis until death, as was reported in the Croatian epidemic. 6 The possibility that there are survivors of this global epidemic seems to have been discounted but cannot be excluded. The postmortem evidence from the Spanish outbreak showed multiple organ damage 9 that one could conclude resulted from small gas emboli reaching the microcapillary circulation of the brain and other vital organs. Histologic examination of the organs of rats after IV injections of 100 μL of fluorocarbon revealed extensive brain infarction, with gas emboli in the cerebral arteries. 10 Thus it is reasonable to postulate that nonfatal systemic emboli may also have occurred in some of the population at risk. This possibility can only be evaluated by a retrospective analysis of the incidence of pulmonary and systemic embolization in patients exposed to dialyzers repaired with PF5070 compared with a matched control population.