The Beginning of Clinical Hemofiltration: A Personal Account

I have been asked by our Editor to set down my recollections of the first clinical hemofiltration. I would first like to note that there are three other publications that address these early days in a more formal way. 1–3 In 1963, I came from my renal fellowship with John P. Merrill at the Peter Bent Brigham Hospital in Boston to join Lewis W. (Bill) Bluemle in the Fluid and Electrolyte Section chaired by J. Russell Elkinton of the Department of Medicine at the Hospital of the University of Pennsylvania in Philadelphia. Bill had stimulated my interest in dialysis when, as a medical student rotating through his third year medicine course, he spoke about his work on the artificial kidney. Bill had oversight responsibility for the Clinical Research Center (CRC) in the Maloney Building and the new two bed chronic dialysis unit that it housed. At the time of my arrival, Bill was investigating the transport properties on a novel sheet plate dialyzer that was disc shaped and featured a radial flow path and cone supports for the membrane. The latter was cast from elastomeric material with cones staggered in orientation to maximize turbulent flow of dialysate. 4 My laboratory space was adjacent to the CRC. Shortly after joining Bill, I had identified convective transport of urea across the peritoneal membrane using the CRC inpatient facilities to study patients on maintenance peritoneal dialysis. 5 This was prior to the availability of Henry Tenkhoff’s fine catheter and when the only hypertonic solution contained 7.5% glucose. As convective transport across peritoneal capillaries appeared to be analogous to glomerular filtration, I wanted to explore this modality for Cuprophane membrane as a more physiologic way to cleanse uremic blood. This was not successful because of the low hydraulic permeability of this membrane and its proclivity to burst when in twin-coil format it was exposed to transmembrane pressures of 1–2 atmospheres. Allen Michaels, Ph.D. had recently left a professorial appointment in the Chemical Engineering Department at MIT to found the Amicon Corporation, bringing with him Assistant Professor Peter Bixler. Allen had created a novel set of polyelectrolyte membranes (polysulphone among them) and was exploring their use in the laboratory for separation of solutes with different molecular sizes using convective transport. He had the ability to adjust the mean pore area and radius of the membrane to achieve a range of water and solute transport properties. Bill was instrumental in establishing our collaboration with Allen and Peter. Allen had submitted an NIH contract to explore the use of this new membrane for the artificial kidney. Bill, as the reviewer of this proposal, shot it down for a variety of technical reasons. Bill’s subsequent discussions with Allen opened the door for our collaboration. Early bench work was carried out both at Amicon in Boston and in the CRC labs in Philadelphia. Clark Colton, a newly minted Chemical Engineer from the Chemical Engineering Department at MIT, spent summers at Amicon conducting a series of bench experiments with this polysulphone membrane. Figure 1 shows Clark, Norman Jacobs, the CEO of Amicon, and myself, at an ASAIO meeting at a time when medical meetings had a light side. At that time Tony Besarab, a medical student with an undergraduate engineering degree, came to work in the laboratory for the summer with Bill and me. The three of us constituted the research team looking at convective solute transport at the CRC. Figure 2 shows Tony and myself at a birthday celebration for Hedi Bowman, a CRC technician. The bench studies and in vivo dog experiments were conducted in the CRC laboratories and the Maloney Building animal facility.Figure 1: Norman Jacobs, CEO of Amicon, myself, and Clark Colton, taken at an ASAIO meeting.Figure 2: Hedi Bowman, Tony Besarab and myself at a CRC birthday celebration for Hedi.I was the lead author on our first publication with Tony, Allen, and Bill. 6 Bill joined with Peter Bixler and Lita Nelson (an MIT engineering student at the time) on the second. 7 Both were laboratory studies on transport kinetics. From these bench studies, it was apparent that a sheet plate format was undesirable for a clinical convective transport device. This was due to the high inherent capacitance of this configuration when scaled up to clinical size and subjected to high transmembrane pressures. Richard Stuart and Ben Lipps of Cordis Dow, under contract with Ben Burton, Ph.D., who ran the Artificial Kidney Chronic Uremia program (AKCUP) for the National Institute of Diabetes, Digestive Diseases, and Kidney (NIDDK), had recently been successful in spinning hollow fibers of cellulosic membrane that were dimensionally stable even at high transmembrane pressures. 8 At Amicon, polysulphone membrane was spun into hollow fiber format under the guidance of Michael Lysaght and Cheryl Ford and provided on a subcontract basis to Bill and myself for animal testing at the University of Pennsylvania. The ease with which this technically challenging task was accomplished at Amicon was due in no small part to the able technical assistance of Dean Paulson, whose father spun hollow fibers of licorice for the candy industry. Bench studies performed with these high flux fibers using outdated human blood to which inulin (5200 daltons) had been added displayed a solute removal spectrum that was far wider than conventional diffusive dialysis across membranes then in common clinical use (Cuprophane). Inulin passed the polysulphone membrane unrestrained in a manner analogous to its passage across the native glomerular basement membrane, i.e., with a sieving coefficient of unity. Bill also guided our application for funding to Benjamin Burton’s AKCUP at the NIDDK. I had initially dubbed this proposed technique “diafiltration” to acknowledge the likely joint roles of convection and diffusion in the overall transport process. 5 Subsequently, Ben, on recognizing how little diffusion played a part, suggested that we call it “hemofiltration,” the name by which it is known today. 9 Gil Kliger and Eddy Horn, who were machinists at the University of Pennsylvania’s Physiology Machine Shop, were skilled in translating ideas into medical hardware. They constructed the clinical fluid cycling equipment from basic materials and, where possible, from “off the shelf” components such as motors and electrical switches. The fluid cycler utilized the principle of “paired pumping.” Two pistons arranged in a horizontally opposed configuration with a common drive shaft propelled by a single cam permitted precise matching of the two flows when non-protein containing solutions were used for the bench tests (Figure 3). However, solutions containing protein, such as whole blood, reduced the hydraulic permeability of the polysulphone membrane rather dramatically, causing the ultrafiltrate pump to operate at negative pressure (300–400 mm Hg) while the infusion pump operated at a small positive pressure. This caused degassing of the ultrafiltrate, creating a substantial error in the fluid balance due to gas bubbles displacing some of the ultrafiltered plasma water. A volumetric cylinder was added to the circuit to permit visual quantification of this error and allow manual compensation for this imbalance, as well as to set the net fluid removal for the treatment session (see Figure 3). Figure 3: Schematic diagram of the paired pumping diafiltration fluid cycler. The metering chamber was added to permit measurement of net fluid gain or loss and to permit compensation for the degassing of ultrafiltrate under negative pressure.Clinical nephrologists at the time were weighing a speculation made by Belding (Scrib) Scribner about the importance of middle molecular weight solutes circulating in uremia that, because of their slow diffusivity, were poorly removed by conventional dialysis. Scrib’s comments had been provoked by his observation that patients maintained on intermittent peritoneal dialysis did remarkably well clinically, even though peritoneal dialysis was much less efficient than hemodialysis at removing the commonly considered uremic toxins of small molecular weight as epitomized by urea. The longer treatment time (24 hours or longer) presumably permitted the slowly diffusing middle molecular weight solutes to be adequately removed. Scrib’s comments were duly recorded by George Schreiner, MD, Editor of the Transactions of the ASAIO. 10 It was apparent to Bill and me that hemofiltration would provide a superb tool to test the importance of this “middle molecule” hypothesis, and possibly might offer a new treatment modality for uremia. Bill and his technician Dorothy Senesky had been conducting experiments in the CRC lab to isolate compounds from spent dialysate that were other than the conventional toxins of urea, creatinine, etc. using counter-current fractionation methods. Testing for toxicity of these compounds was done in collaboration with Paul Teschan, MD, at Walter Reed Army Hospital. Paul had developed a baboon model to examine their toxicity. He would introduce these compounds by intravenous infusion with the ureters transiently closed by cinching snares through the flanks of the animal. Behavioral impairment with and without the infusion of these compounds was the “read out.” Parallel investigations at the time were being carried out by Jonas Bergstrom and Peter Furst (again under contract to the AKCUP) applying sophisticated chemical and physical separation methods to uremic plasma, posttransplantation urine, and ultrafiltered plasma water in order to understand the toxin composition of these solutions and how they differed. 11 The generous presence of these solutes (identified only by size) was intimidating because of their large number, the lack of chemical characterization, and the lack of clinical correlates to qualify them as toxins. Clinical hemofiltration appeared to be a way around this dilemma. Our short-term studies in normal mongrel dogs served primarily to qualify the precision and repeatability of the fluid balance equipment, but offered no evidence that the clinical outcome of serially applying this wider solute clearance profile would offer clinical benefit to the patient with end-stage kidney disease. In fact it was duly noted that urea and, to a lesser degree, creatinine were more swiftly removed by dialysis than with hemofiltration. Clinical application of hemofiltration, considered the next logical experiment, was attended by two concerns: 1. Would substitution of hemofiltration for hemodialysis result in inadequate removal of the small solutes? 2. Were there as yet unidentified larger molecular weight “vital” solutes that would be lost and result in morbid and/or mortal events? At that time there were no human experimentation committees (IRBs) to assuage investigator angst about these concerns. After considerable discussion it was decided to employ a strategy of “creeping substitution,” where a small hemofilter would be run in parallel with the Travenol 100 liter tank twin coil dialyzer for maintenance dialysis. We would gradually scale up the area of the hemofiltration membrane, the blood flow rate, and the amount of ultrafiltrate exchanged. Both concerns would thus be addressed. Initial conditions called for a blood flow of 100 ml/min, a membrane size of 0.2 m2, and 500 cc of fluid exchanged in predilution mode (Figure 4A). This contrasted with final values of 300–400 ml/min of blood flow, 1.6 m2 of membrane, and 100 liters of fluid exchanged (Figure 4B). The latter estimates were guided by rough kinetic modeling of hemofilter performance. We were confident that this incremental approach would permit the identification of any deterioration in either the clinical or the laboratory status of the study subject. Today this combined use of convection and diffusion is known as hemodiafiltration. 12Figure 4: A, 0.2 m2 fiber bundle. B, 1.6 m2 bundle.Benjamin Forer, a long-term patient and friend on maintenance dialysis at the CRC dialysis unit, was exceptionally well versed in the dialysis technology that had sustained his life for more than 2 years. He was also a practicing lawyer. With Ben, good quality informed consent was a reality in a circumstance where this is not always possible given the barriers of technical complexity, and the hierarchical dependency that attends the nephrologist/end-stage renal disease (ESRD) patient relationship. The incremental protocol was initiated. With time (weeks) as blood flow to the hemofiltration membrane increased and that to the twin coil reduced, we reached a point where the slow flow of blood through both circuits resulted in clotting problems. The ease with which excess body water could be removed was remarkable to us and to Ben, who indicated few to no cramps and no “crashing.” However, 10 to 15 L of fluid exchange over the 5-hour treatment schedule produced a low-grade fever about half the time that resolved an hour or so following treatment. A second patient initiated the protocol and also displayed frequent low-grade fevers during treatment that resolved promptly on their discontinuance. 13 The cause for this fever subsequently turned out to be the use of pyrogen containing water in the spinning process that adsorbed to the luminal surface of the fiber and was eluted by plasma but not by saline. Our plan with Ben was to alternate two purely twin coil treatments with one predilution 15 L hemofiltration treatment per week. The latter did not occur, as clinical hemofiltration was stopped, and we returned to animal work for the remainder of the year to track the cause of this fever. Sadie Williams, a 45-year-old inner city woman with kidney failure due to severe hypertensive cardiovascular disease, was the first to be fully maintained with 100 liter predilution hemofiltration as the sole therapy. By this time the dialysis unit had moved to more luxurious quarters and the number of chairs expanded to 12. Beth Cameron, the Head Nurse, and her second in command Marilyn Neff R.N., in conjunction with Inell Lawson, a nurse’s aide, were the primary caregivers for Sadie. Clinical points of note after 6 months of maintenance hemofiltration were the lack of hypertension in the absence of all antihypertensive agents, the resolution of her cardiomegaly on chest x-ray, and the return to normal of her EKG, which had shown significant voltage criteria for left ventricular hypertrophy. Her biochemical profile as contrasted with that on simple dialysis showed no significant differences. 14 A major ergonomic objection to the technique was voiced by her caregivers as the diluting fluid used was supplied by Abbott Pharmaceuticals in the form of 1-L bottles of Normosol-R. Access to the fluid required piercing a diaphragm with the pointed end of the fluid administration set. For 100 L to be administered in 5 hours required the “spiking” of twenty 1 L bottles per hour, or 1 liter every 3 minutes. This proved taxing to the sturdiest of wrists (Figure 5). Figure 5: Sadie Williams undergoing predilution hemofiltration. Note the impressive rack of liter bottles used for replacement solution.It should be noted that Edward Quellhorst, M.D. from Hanover’s Munden FRG visited with me in 1968 in conjunction with the ASAIO meeting that was held in Philadelphia. Bill Bluemle was President of the Society that year. Edward and I discussed the filtration technology in depth, and it was clear that he was pursuing a similar path in his unit in Germany. 15 While my paper in 1967 3 antedates his in 1971 16 and I am credited with the first clinical application of hemofiltration, it should be noted that Edward was responsible for the first in depth clinical application of postdilution hemofiltration using the commercially available AN-69 membrane from Hospal and fluid cycling equipment of his own design (Figure 6). His series of 100 patients followed for 10 years is a landmark work 17 pointing out, among other things, that survival was comparable to, if not better than, that of his hemodialysis patients who were treated three times per week for 5 hours, even though his hemofiltration patients were receiving a KT/V for urea of only 1.65 per week. In light of Scrib’s speculation, it is interesting to note that this KT/V urea value is quite comparable to that resulting from continuous ambulatory dialysis, i.e., 1.7 per week. 18Figure 6: Edward Quellhorst, MD, showing me his fluid cycler used with the AN 69 membrane for postdilution hemofiltration in his unit in Hanovers Munden, FRG.New fluid cycling technology and the acceptance of online, membrane produced, intravenous quality solution 19 by the regulatory agencies means that hemofiltration can now be conducted for the same cost and with comparable safety to hemodialysis. This has brought renewed interest in this modality to both Europe and the United States. Today, we see a sharp upswing in the prescription of hemodiafiltration with online preparation of diluting fluid, especially in Europe, where it is now the most rapidly growing modality of treatment for ESRD (23% increase in 2001). 20 Given the equivocal outcome of the recent NIH sponsored hemodialysis trial, 21 I await with high interest a rigorous prospective clinical demonstration of the possible role played by middle molecules in the morbid and mortal events that plague ESRD patients. In conclusion, it should be evident that the first clinical hemofiltration treatments were made possible by a remarkable team of talented biomedical engineers, machinists, and research oriented medical professionals. I am proud to have been a member of that team.