Update On: Leukocyte Reduction of Blood and Blood Components Public Workshop National Institutes of Health Lister Hill Center Auditorium Building 38A 8600 Rockville Pike Rockville, Maryland 20852 July 20, 2005 [PDF Version] CONTENTS Welcome and Introduction Jay S. Epstein, M.D. Workshop Overview U.S. Regulatory Considerations and International Policies Regarding Leukocyte Reduction Alan E. Williams, Ph.D. Recent Studies Addressing the Value of Pre-Storage Leukoreduction for Non-Targeted Recipients Robertson Davenport, M.D. The Yale-New Haven hospital Universal Leukocyte Reduction Program Edward Snyder, M.D. The Impact of Pre-Storage Leukoreduction Among Transfused Trauma Patients Avery Nathens, M.D., Ph.D., MPH Adverse Events and Manufacturing Failures Associated with Leukoreduction David Stroncek, M.D. Questions for Morning Speakers Practical Aspects of Pre-Storage Leukoreduction in a Blood Establishment Including Use of Pooled Samples for Enumeration of Residual Leukocytes Timothy Malone, MT(ASCP)SBB Practical Aspects of Pre-Storage Leukoreduction in Blood Establishments Dan Waxman, M.D. Fred Walker, Ph.D. FDA Current Considerations: Pre-Storage Leukoreduction: Process Validation, Quality Assurance and Monitoring, Processing, Testing and Licensure Alan E. Williams, Ph.D. Sharyn Orton, Ph.D. Questions for Afternoon Speakers Panel Discussion New Technologies in Filtration Prion Reduction from Blood by Filtration Luisa Gregori, Ph.D. Pall Leukotrap Affinity Filtration System Jerry Ortolano, Ph.D. PROCEEDINGS Welcome and Introduction DR. EPSTEIN: Good morning, everyone. I know people are still taking their seats, but we have a very busy program, so it is important for us to try to start on time and then keep on time. If I could ask everyone, please, to take their seats, so that we can get started promptly, I would appreciate it. My name is Jay Epstein and I am the head of the Office of Blood Research and Review at CBER, FDA, and on behalf of the Food and Drug Administration, which is one of several cosponsors, I would like to welcome you all this morning, and note that we are sharing the organizations' workshop with support from the National Heart, Lung and Blood Institute, and also from the Department of Health and Human Services, Office of the Secretary/Office of Public Health Science. It is very gratifying that we have had over 150 registrants, suggesting that this remains an important topic in transfusion medicine, and although Dr. Williams will outline the agenda in more detail, I just want to highlight the three major themes. We will be reviewing studies regarding the clinical benefit of leukocyte reduction particularly in the non-targeted population and information that has emerged since the most recent public discussion about four years ago. We will then talk about the experience with leukocyte reduction, about FDA's current considerations regarding product standards and quality control, and then some updates on the stage of development of removal of prions by leukocyte filtration. Now, the issue of universal leukocyte reduction, that is to say, non-targeted leukocyte reduction for blood recipients, has been contentious for a number of years. The FDA's consistent point of view has been to encourage universal use of leukocyte reduction, and the reason is because of, first of all, the known benefits in targeted groups including preventing febrile non-hemolytic transfusion reactions, reducing risk of cytomegalovirus transmission, and reducing alloimmunization, but there, the concept is, well, you don't always identify the patient who needs leukocyte-reduced product, so is it not a better precautionary measure just to do it for all, and then there are these potential benefits which are not as well established, such as reducing other cell-associated infections, both known and unknown, the issue of potential benefit in clearance of prions, and then some even more controversial issues reducing postoperative infections, tumor recurrence, et cetera. Now, we brought this issue to the Blood Products Advisory Committee meeting in 1998, it seems like a long time ago, where the general concept was endorsed, and we followed that a few years later with a draft guidance which encouraged leukocyte reduction as a general safety measure, but we were not able to establish a requirement. Our legal counsel advised us that that would require rulemaking because of the fact that non-leukocyte reduced products are already identified in the CFR. Of course, the issue of cost versus benefit became very central to many people's thinking and received a lot of attention in the blood industry. For that reason, the issue was also brought to the Department's Advisory Committee on Blood Safety and Availability, which, in April 2001, essentially endorsed an FDA strategy to develop a rulemaking that would establish leukocyte reduction as a requirement. However, in the interim, we have recognized that there has been an evolving debate in the scientific literature, some studies showing general benefit, some not, and consequently, we have been more in an expectant mode seeking to identify the evidence base that might underpin any ultimate policy. At the same time, we have been moving forward with clarifying the expectations for product standard and quality control, since any product labeled as leukocyte reduced ought to meet a well-defined and a meaningful standard. So, this is where we are and we are hoping that this workshop will shed some light on the current scientific state of the art on these issues. Now, I would like to, before we launch, thank the staff who have been most involved in developing this workshop. Alan Williams has been the chief organizer and will be our moderator with strong support from Sharyn Orton who is his deputy at the Division of Blood Applications in my office, and administrative support from Rhonda Dawson, Marty Edwards, and very active participation and support, as I have said, from persons at NHLBI and DHHS. Many of you are aware of the report of the recent demise of Tibor Greenwalt, who passed away at the age of 91 this last Sunday, July 17th. I think as the beneficiaries of the Titanic legacy that he left in transfusion medicine, it would be fitting to say a few words and perhaps observe a moment of silence. Dr. Greenwalt was born in Hungary on January 23rd, 1914. He immigrated to the United States in 1920 at the age of 6. Earned an Undergraduate Degree and a Medical Degree from New York University and studied hematology at New England Medical Center. His interest in blood diseases remained lifelong. He began his research career in the U.S. Army in India, and after that, he became medical director of what became the Blood Center of Wisconsin, which has been a lead research institution in this field. Dr. Greenwalt served as the Vice President of the American Association of Blood Banks, of which he was a founding member, and probably the first person on record to conceive of the idea, and he served as the National Director of the American Red Cross Blood Program where he was credited with establishing the Rare-Donor Registry both for the Red Cross and the AABB. In his laboratories at the Red Cross in Washington, D.C., he directed research in hepatitis and the storage of red blood cells. He developed the first filter for white blood cells. He did landmark research on blood grouping reagents, and his lifelong work on red cells continued at Hoxworth where he continued his research really up until his terminal hospitalization, and where he was active in developing new storage solutions. He was also the founding editor of Transfusion, which is recognized as the foremost publication in the world for new information regarding transfusion medicine. He contributed to over 200 major books and research papers in the scientific literature. He was elected in 1984 to the Institute of Medicine of the National Academy of Sciences in recognition of his role and his contributions, and in early 2005, Dr. Greenwalt was awarded the most prestigious honor in transfusion medicine, the Karl Landsteiner Memorial Award for lifetime achievement in blood and transfusion sciences. We all could aspire to such accomplishments and probably few of us will ever achieve that. Let me just ask everyone to take a moment in silence in memoriam. [A moment of silence observed.] DR. EPSTEIN: Thank you very much. I will now give the podium over to Dr. Williams who will explain what we are doing today. Workshop Overview U.S. Regulatory Considerations and International Policies Regarding Leukocyte Reduction DR. WILLIAMS: Thanks, Jay. What I wanted to do just firstoff is briefly summarize what the goals of the workshop are because there are just a couple themes that are going to permeate the day and are the areas where I think FDA really wants to update itself, as well as provide data to update you as the audience. The first one, as Jay mentioned, is to review any new evidence regarding leukocyte reduction for non-targeted recipient populations. There have been a lot of discussions previously and we are just going to go ahead and assume that there is an accepted medical value of leukocyte reduction for targeted recipient populations as Jay mentioned. We are interested in updating the current data on leukoreduction failures and adverse events related to the leukoreduction procedure. These can include incomplete filtration, incomplete white cell removal, recipient adverse events, and blood establishment experiences overall. There will be several talks targeted specifically to that area. Finally, an area that poses a bit of a dilemma for process control is the fact that patient populations or subpopulations that need a leukoreduced product for safety reasons need to have a product that really is leukoreduced, and the label needs to reflect the content of the component. On the other hand, for a broader scale or universal product that is being produced, you need the capability of high throughput and efficient quality control, so how to balance those two strategies within one production laboratory is a key issue, and we hope to have some sound discussion on that. So, just to walk through the agenda, I am going to, following the agenda itself, just give a few of the prior regulatory considerations just in terms of stage setting for what FDA's considerations have been in the past, and then later today, we will give current considerations, which then, in turn, might be modified by today's discussions. Our first data review talk with be by Rob Davenport from the University of Michigan addressing recent studies on the value of pre-storage leukoreduction for non-targeted recipients. Following that, Dr. Ed Snyder, a long-term colleague, will be updating us on the Yale-New Haven Hospital Program of leukocyte reduction for all its transfusion recipients that they have several years of experience with this now, and he addresses it both from a patient welfare and I think a cost-benefit relationship. We will also have a talk from Dr. Avery Nathens, who is going to present a currently unpublished study on the impact of pre-storage leukoreduction among transfused trauma patients. Finally, to end this morning's session, an overview talk on adverse events associated with leukoreduction and manufacturing failures, and some data developed at the NIH Department of Transfusion Medicine on sickle cell hemoglobin and several other factors leading to filter failures. Then, we will follow with questions for morning speakers and the open public hearing. In the afternoon, right after lunch, Tim Malone from Florida Blood Services is going to follow up a really very excellent talk given by Herman Leparc at one of our Blood Product Advisory Committee meetings just on the practical aspects of leukoreduction in a blood center and quality control within the production and component laboratories, and also go through some of the data that their laboratory has been doing in evaluating potential use of pooled ABO matched component samples for residual white cell counting. We also, in the interest of presenting current experiences, contrived a survey together with America's Blood Centers primarily and also the American Red Cross, regarding what the current leukoreduction proportion of units are that are being distributed today, what the experiences are with quality control, how white cell counting is done, what rate of filtration failures are occurring, with what products, and so forth. These data will be shared with us by Dan Waxman, representing America's Blood Centers, and Dr. Fred Walker, representing the American Red Cross. Dr. Orton and I will then spend a short time going over FDA's current considerations regarding leukoreduction, including process validation, quality assurance and monitoring, processing, testing, and the licensure process. That will be followed by another open public hearing, which will then go into a panel discussion with a total of seven luminaries that we were able to recruit to give their opinions on a couple of the key major issues that are the theme for the day, namely, universal leukoreduction and quality control strategies. Then, the last session, related but a little bit distinct from those two main themes, Dr. Luisa Gregori from the University of Maryland, working with Dr. Bob Rohwer, is going to give both an overview on some of their latest data on prior reduction by filtration. This will be followed by Dr. Jerry Ortolano from Pall Corporation sharing data with respect to the Pall Leukotrap Affinity Filtration System. So, it is a full day and we are going to move it right along, but there should be a lot of information. We will start sessions on time. If you are five minutes late coming back from lunch, you are going to miss five minutes of content. Sorry, but it's a full day and we really have to do that. We have a five-star cafeteria downstairs here, so you ought to be able to get through the lunch process pretty quickly. I am going to just show a few slides on some of the regulatory history with respect to leukoreduction just in terms of stage setting and a little bit about some of the international practices. The current FDA recommendations, although there have been some draft recommendations in the interim, the current in-force recommendations go back to the 1996 FDA Memo. In terms of quality assurance monitoring, the recommendation is for evaluation of 1 percent of representative products or at least N equals 4 per month. For residual white cells, that memo references a standard of less than 5 x 106 residual white cells per collection or component, and 85 percent retention of original product. For single donor platelets, that value is proportional, so that in a pool of 6, one ends up with the same standard for residual white cells. All evaluated products must meet specifications, and if a failure is observed, the label needs to be revised and the process investigated, and details, methods available for counting at that time, which included the Nageotte manual method, flow cytometry, and other validated methods. As Jay mentioned, we issued draft guidance on the topic of pre-storage leukoreduction in January of 2001, and there were a few changes in the proposed standards and quality assurance monitoring scheme at that time. In harmony with the current European standard, we recommend that a product specification of 1 x 106 residual white cells, and introduced a statistically-based quality assurance scheme, and this was based on a binomial consideration, and we put into the draft guidance a candidate specification that 95 percent of products should meet the 1 million specification for residual white cells with 95 percent confidence. This could be achieved by running cycling quality assurance every three months, which would involve 5 for the week, 20 per month, or 60 per quarter, and if one had zero failures within those 60 counts, one would achieve that 95-95 criterion. That guidance also recommended considering the testing of all donors for hemoglobin S, because it was known that that was one of the donor factors that contributed to clogged filters, and made a recommendation for 100 percent quality control of components that were destined to be used for CMV-susceptible recipients and particularly in lieu of CMV antibody-negative units. There were a lot of comments received to that draft guidance and I am just going to summarize them here. Although the European standard is 1 x 106 residual white cells, and FDA certainly supports removing as many of the white cells as possible, there is concern that although current leukoreduction filters are easily capable of removing white cells to that level and far below, whether or not counting methods currently in use for residual white cells could be validated to count at the 1 million level. The comments coming in included that consideration, as well as the fact that when a filter failed, it generally failed big time, and there were a lot of white cell contaminations, so perhaps the difference between 1 million and 5 million was not likely to be medically meaningful for patients who needed a leukoreduced product. There was a lot of discussion and a BPAC public discussion on the feasibility of sickle cell hemoglobin screening and a BPAC vote that, in fact, this was probably not feasible on a universal basis for blood establishments, but certainly could be worked into a quality assurance process. If it was feasible, it could help to identify donors whose collection should not be put through the filtration process. The recommendation for 100 percent quality control of units destined for CMV-susceptible patients was soundly criticized as simply not being practical, because a component laboratory at a blood establishment simply wouldn't know what units were going to that patient population, so it would involve maintaining a dual inventory probably generally not a easy or practical thing to do. So, there was considerable comment about that recommendation and whether or not that was practical despite the fact that, as I mentioned, for quality control for targeted recipients who really need a leukoreduced product, there is potential value certainly in looking at all of those units. There was a lot of discussion about the statistically based quality control paradigm including a BPAC discussion. The general sense of the committee and, in fact, a vote was that 100 percent quality control for all leukoreduction procedures was simply infeasible, and although the statistical-based process control appeared to have merit, there was some concern as to whether even with that cycle of 60 counts, that would, in fact, potentially put a blood establishment into an endless loop or they found one had to requalify, had to do the counts again, and would spend a lot of time doing quality control and trying to recover from failures given the fact that the failure rates at that time were really a very broad range. So, the sense of the committee was we are willing to consider this as a recommendation, but we do have some concerns and it needs some additional thought. As reflected in the summary of the BPAC discussion here, the first one shows the sickle cell vote which was unanimous against that recommendation. There was also discussion at that committee meeting about the filters themselves and whether sufficient parameters regarding their use and performance were elucidated in the product labeling and whether or not labeling needed to be revised to more carefully specify units that were eligible for filtration with the conditions of filtration such as time and temperature to the filtration process should be. At that session, as well, there were a lot of abstentions simply because they felt there was not a lot of data to address that issue, and a mixed vote with 2 voting yes and 4 voting no in terms of label revisions. The committee did recommend better investigations of leukoreduction failures and their causes, so that some of the information could be refined in this area, and did mention specifically, based on the Canadian experience reported at this meeting, that mixing, particularly mechanical mixing with a validated process appeared to really reduce the clogging factor that occurred during collection. That session did not address process control strategies, but a supplement one did. Now, a lot of thought has gone into this quality assurance monitoring factor and I think it has been educational for us, as well as the public discussions I think have really helped to sort of flesh out some of the parameters involved, and as we start thinking about statistical quality control. Defining process failure really depends upon appropriate and distinct control points. It is not usually helpful to lump everything together and consider it a failure and base all of your quality monitoring around that, and candidates here would be that incomplete filtration should be considered independently from white cell contamination, which should be considered independently from therapeutic content of the final product. In 2001, when the last public discussion was held, leukoreduction processes had relatively frequent reported failures, and some of these were poorly understood. Probably because they weren't all investigated, not all of them were process failures, but until it is investigated, you often don't know that. At the time of that Blood Products Advisory Committee discussion, a survey was done among users, and we got a range of 0.3 percent to 13 percent total observed failures of the process, a very broad range. This was in contrast to when we polled the filter manufacturers, they reported a failure rate in their laboratories of 3 per million, so really, quite a broad range in terms of failure of the process. It was known at that time that when a filtration was slower than anticipated, that was often correlated with poor white cell removal. It was known that among donors with sickle cell trait, about 50 percent led to clogging the filter before the process was complete, and of the remaining 50 percent, the unit appeared to filter, but left a lot of white cells in the ensuing product, so that an individual with sickle cell trait, you generally had a successful leukoreduction about 25 percent of the time. It was also recognized that there are other poorly understood donor factors. Dr. Stroncek is going to give some really interesting data on what they have found in terms of their overall failure rates. It is known that donor failure rates appear to be inherent to the donor at least for the most part, that if a donor fails on one occasion, that they are likely to fail in subsequent collections, so there is a message there. It was also recognized that some of the failures are lot specific to the filters being used, and some of the high rates that were observed simply were due to the fact that during that time period of the survey, they had a bad lot of filters that added to their failure rates. As mentioned, Canada implemented universal leukoreduction some time ago and presented data publicly at one of our BPAC sessions, and it showed pretty clearly that mixing particularly using validated procedures reduced clogging the filter and loss of product. What are the implications of failure? This is important in defining a quality assurance monitoring plan. In terms of safety, if a product is labeled as leukocytes reduced, and has high levels of contaminating white cells, patients who need the product can be harmed, for instance, those who really are susceptible to CMV infection and its terrible consequences for an immunocompromised individual. On the other side, there is an efficacy consideration, that if there is undue loss of a therapeutic product, it reduces the potency of the product on one side, but if the filtration is actually incomplete, it results in loss of product and wasted blood resources. So, one simply wants to put all the parameters in place to keep this an efficient and safe process. Some considerations about process control. What I am going to do is just give a couple, a little bit of a structure to it, and Dr. Orton in the afternoon is going to give a little more detail about some of the specific mechanisms that can be used for process control. One needs to consider whether or not it is important to do 100 percent product qualification versus a sampling scheme, and whether or not that should be a statistically based stamped sampling scheme. That really depends on how critical is the final product specification, particularly from a safety aspect. One needs to consider the appropriate distributions for the statistics being used, whether or not the outcome can be dichotomous or needs to be a continuous outcome. For instance, in the use of the binomial, one doesn't necessarily need to quantitate the residual white cells that are there. One could basically make use of that statistic by using a pass/failure scheme if that would make it technically more feasible. Whether or not the distributions need to be log normal and converted to fit into a statistical plan is also a possibility. Use of one-tailed versus two-tailed statistics. I forget where it originated, but one of the examples given is an automobile piston versus a white cell count. If you build an automobile piston and it has to go into the cylinder, it is not going to work if it's too large, but it is also not going to work if it's too small, so you need a two-tail assessment of the quality of that piston. Whereas, for a while cell count, you really don't care if it's too low. What you are interested in is the one-tail as far as whether it exceeds a certain medically important value. The frequency of the QC cycle is important - how long can an out-of-control process be tolerated if your process is out of control, you know, how important is it that you stop distributing units that are out of control before your mechanisms for quality assurance pick it up. I just created a couple of points. Contrasting at 100 percent product qualification versus a statistically based plan. Considering 100 percent product qualification for residual white cells, and this is not unheard of. There are some component laboratories in the country that have been doing this. Whether or not it would be feasible on a large scale for all leukoreduced products is certainly something that would get a lot of discussion, but the advantages are it would produce 100 percent label leukoreduced product which would meet the product standard. It would reduce inappropriate white cell exposure to a susceptible patient subpopulation, such as CMV susceptible, and it would potentially stimulate new technologies that would facilitate cost effective white cell counting. What are the disadvantages? A large part of the country still counts white cells manually, and even current processes are, in the blood establishment's view, very labor intensive. Although it is improving a little bit, there still is a limited selection of automated counting devices which can be used, and blood centers, because of the overhead for the manual counting and the expense related to dedicated staff, blood centers may ultimately choose to provide leukoreduced products, which I think is contraindicated in terms of FDA's overall encouragement of the use of leukoreduced products for the reasons Dr. Epstein mentioned. The use of the binomial was detailed in the January 2001 guidance. If one defines a specification that 95 percent of product needs to meet defined specs with 95 percent confidence, this results in a product that has no 95 percent conformance and is, by definition, a safe and pure product, and 95 percent as a confidence interval is a standard accepted scientific norm. This translates to the probability of less than 5 in 100 that chance nonconformance will exceed 5 percent. The way one reaches that, as I mentioned briefly before, is met by zero counts observed, zero observed failures out of a total of 60 counts, or one can predefine that one is going to count a few more. If you count 93, you can have 1 failure and still be within the tolerance range. This is based on an exact binomial distribution, single tail, it doesn't require log normal distribution, and the white cell counts can be pass/fail. The recommendations of the January guidance did give us an example that 60 consecutive white cell counts would be a suitable scheme for quality assurance monitoring, that ongoing QC could entail 1 percent of the total production, but not less than the random 60 counts per quarter, that QC failure would require a consecutive count of the next 60 units, and that if it had zero failures, the establishment could resume normal QC, but if one experienced 1 or more failures in that consecutive, one would need to define the process as being under control and initiate an investigation and corrective action. FDA, I think in any of its recommendations, will encourage alternate equivalent procedures which would meet the goals of its recommendations, and there are other published schemes specific to quality assurance procedures for leukoreduction, and these could be submitted to the agency if the agency has a license supplement for prior approval and would be carefully considered. The binomial approach assures that 95 percent of products labeled as leukocytes reduced will meet the product standard with 95 percent confidence. The quality control workload at the collection centers, while probably higher than it is today with current existing schemes, would still be considerably less than would be needed to count all products, and it would help to ensure that leukocyte reduced products would more readily available. What are the disadvantages? Again looking at the CMV-susceptible patients, leukoreduced products are currently commonly substituted for CMV seronegative products, and because many of the transfusionists supporting these patients, in fact, don't have a complete degree of confidence in a leukoreduced product, they often call for products that are both CMV seronegative and leukoreduced just to be protective. At the 95 percent range, occasional products with levels of residual white cells that exceed the product standard may unknowingly be transfused to susceptible patients. I think as we found in the course of the Blood Products Advisory Committee discussion, the quality control strategy proposed may be too complex for training and ongoing implementation by blood center staff, and if that happens, that would contribute to reduced compliance, which is always a concern. There was a BPAC discussion December 13, 2001, and the Committee voted on the question, does the Committee recommend Option 1, i.e., that FDA should recommend to industry that all products labeled as "leukocytes reduced" meet the defined standard as demonstrated by evaluating all such products for residual white cell content. The Committee voted a unanimous No to that question. Keep in mind that the question as phrased would encompass both manual and automated procedures for producing the product. Question 2. If No to Question 1, does the Committee concur with the modified statistical quality control strategy, the binomial, as outlined? There was general support for this with considerable discussion as I mentioned earlier as far as its practical use within an ongoing component laboratory. I wanted to close by just discussing some of the standards and procedures in place elsewhere in the world. I think some of the best available data is part of the report issued by the Council of Europe with their annual survey. They do a survey for products produced and the characteristics of these products. In general, in Europe, the processes for leukoreduction are a combination of buffy coat removal and subsequent filtration using fully validated procedures, and while not absolute, they do recommend pre-storage filtration within 48 hours. Council of Europe Standards are for a residual white cell count of less than 1 million, minimum hemoglobin content of 40 grams, and a hemolysis of the units less than 0.8 percent of the original red cell mass, and this is monitored by looking at 4 units per month. Process control. Their quality control assessment is currently 1 percent of all units collected with a minimum of 4 units per month, and the standard is met if 90 percent of the units tested fall within indicated values. The Council of Europe, as well, has had discussions and is moving in the general direction of statistical process control. I think you will see some evidence of this in subsequent versions of the Guide to the Preparation, Use, and Quality Assurance of Blood Components. So, I think as well as being considered here in the United States, the same considerations are underway in Europe. In terms of proportion of blood components that are leukodepleted, which is the term used in Europe, there are currently according to the survey 34 percent of countries that are doing 100 percent leukodepletion of red cells. This is based on the survey of which 45 countries are eligible, 29 completed the survey, and 10 reported 100 percent red cell removal. These are Austria, Finland, France, Germany, Ireland, Luxembourg, Netherlands, Norway, Switzerland, and the UK. I also added in here that Canada has also been doing this for some time and we are very fortunate to have Dana Devine here who heads the R&D program for Canadian Blood Services, and she is going to in the course of the panel, and hopefully, also give some of the Canadian perspective on leukoreduction. As far as 100 percent leukoreduction or leukodepletion of platelets, a slightly higher value, 46 percent of European respondents, and the list I mentioned before, Belgium, Latvia, and Iceland also leukoreduced their platelet components. That is the end of my introductory discussion. I want to emphasize again, and I am ending a little earlier to set the stage, we have a lot of material being presented. I know our speakers have some good material and a lot of slides. I encourage healthy discussion, but we do want to stay within the time limits, so we can get everything presented. At this point, I will introduce Dr. Rob Davenport, who is head of the Transfusion Medicine Unit at the University of Michigan. Rob has kindly agreed to present an overview talk on leukocyte reduction in non-targeted populations. Recent Studies Addressing the Value of Pre-Storage Leukoreduction for Non-Targeted Recipients DR. DAVENPORT: Thank you. I appreciate the invitation to speak here today. I think it is important to start out with a couple of minutes of sort of what the background is, how we got to where we are right now in particular with respect to what would be the generally more agreed-upon indications for leukocyte reduction, that being reduction of CMV transmission, reduction of HLA alloimmunization in reduction of febrile transfusion reactions. In terms of CMV transmission, the landmark study is the Bowden study, which was a randomized clinical trial in the setting of stem cell transplantation in Seattle, that randomized subjects to receive either CMV seronegative components or components that were leukocyte reduced at bedside. These are familiar to you, so I am not going to spend a lot of time on it. The main outcome measure was CMV infection within 100 days from transplantation. This study reported equivalent rates of CMV infection between the two arms, however, there was a higher rate of CMV disease in the group that received the bedside leukocyte reduced. This has been a very contentious issue. With respect to alloimmunization, again, there is a landmark study, which is the TRAP study, which you are probably very familiar with. This is a randomized clinical trial in the setting of acute myelogenous leukemia that randomized subjects to receive four different platelet products, all the red cell products were leukocyte reduced, and the platelets were either unmodified random donor platelet concentrates, filtered random donor concentrates, UVB-irradiated random donor concentrates, or filtered single donors. The outcome measures were alloimmunization monitored by lymphocytotoxic antibody testing and platelet refractoriness. There was a significant difference between the control group receiving the non-leukocytoreduced products in terms of both development of HLA antibodies and refractoriness to platelet transfusion, however, the three study arms were really equivalent, and there was no additional benefit seen in the single donor platelets. The greatest benefit was seen in the subjects who had not been previously exposed to HLA antigens either through pregnancy or through transfusion. For those who had been pregnant previously, there was minimal benefit. In terms of febrile non-hemolytic reactions, there are a multiplicity of studies out there, a number of observational studies, and some prospective case-controlled studies that all used a variety of methods of leukocyte reduction, but all pretty much reached a similar conclusion of the potential benefit. That is where we stand in terms of where the basic data come from for what would be considered the targeted populations. While we could probably argue all day exactly what the targeted groups are, I think this pretty well summarizes the individuals who are at significant risk of some adverse outcome from CMV transmission or HLA immunization or are at higher risk of recurrent febrile reactions. So, presumably, non-targeted is everybody else. What I want to try to do is to review clinical studies, I am not going to focus on laboratory studies, particularly in the last four years of the impact of leukocyte reduction in terms of CMV transmission, in terms of alloimmunization, in terms of febrile transfusion reactions, and in terms of clinical outcomes. There is a variety of data sources and I will be the first to admit I did not do an absolutely exhaustive review, but I believe that what I have here to talk about today is truly representative. Some of these are randomized clinical trials, some are cohort "before and after" studies, and a couple of meta-analyses of both randomized trials and "before and after" studies. If you are interested, the references are all at the end of the presentation, and the numbers in square brackets refer to those references. So, to begin with, CMV transmission, since the Bowden study, I am not aware of a randomized clinical trial of sufficient size that would really be equivalent to the Bowden study. There was, however, a follow-up to that study from the same institution, which was a prospective cohort study that looked at two groups of individuals. All of these were CMV seronegative donor recipient pairs who were undergoing stem cell transplantation, and the two periods that they looked at were differentiated by the kinds of compounds that they received. During the first period, all of the red cells were leukocyte reduced pre-storage filtration, and random donor platelet concentrates and single donor concentrates were also leukocyte reduced by filtration by post-storage methods. During the second part, the only difference was that the single-donor concentrates were leukocyte reduced by process, so were not filtered. The main outcome measure was CMV antigenemia by day 100 from transplantation. The patients were monitored weekly for CMV antigen. They gave data on the quality in terms of number of leukocyte reduction failures. Many studies do not provide these data. These were sampling data, but they reflect a very low rate. I think it is, though, important to recognize that some of the single-donor platelet concentrates had really quite high levels, and these would not be considered leukocyte reduced. In addition, I think it is important to recognize that there was a very low rate of CMV seropositive donors in this. About 8 percent of the single-donor concentrates were, but just over 1 percent of donors of either random donor platelet concentrates or red cells were seen of the seropositive. So, there clearly is some pre-selection going on in this population unless Seattle is incredibly clean combined to Ann Arbor. It must be all the coffee. So, the incidence of CMV antigenemia was significantly higher during the second period as compared to the first. It was about 4 percent of individuals in the second period. It was about 1.7 percent cumulative in the first period, and it appeared to be somewhat earlier in that that there was a steeper initial part of the curve here. In univariate analysis of the 24 individuals who developed CMV antigen compared to those who did not, the significantly different factors were the total number of units that the patients received, the total number of those that were from CMV-positive donors, and within that, it was red cells from CMV-positive donors and single-donor platelet concentrates from CMV-positive donors. Random donor concentrates were not significantly different. Out of the individuals who received at least one component from a CMV seropositive donor, 14 out of 235 developed CMV infection compared to 4 out of 194 individuals who received only CMV seronegative components. Now, to my arithmetic, those don't add up to all of the patients, so I had a little bit of trouble interpreting these, but it looks like out of those who only received CMV seropositive components, there was about a 2 percent baseline rate. That pretty well agrees with the Bowden study and with other experiences. That seems to be the basic underlying rate of development of CMV. Out of those who only received CMV seronegative components, there were more during the first period than during the second period. In fact, about one-quarter as many individuals during the second period received only CMV seronegative components as compared to the first period. The outcome of these individuals, overall there was a 3 percent rate of development of CMV antigenemia. Most of those individuals received preemptive antiviral therapy. There were some who were autologous transplants with low levels that elected not to. No individual developed CMV disease. So, on the basis of this, these authors suggested that it may be not prudent to abandon CMV seronegative for certain populations, particularly those who would be at risk of disease and would not be monitored and given the opportunity to have viral therapy. However, this is an association, it is not a causal relationship. It doesn't directly compare CMV seronegative components versus leukocyte reduced, and in multivariate analysis, the one factor which appeared to be important, that fell out statistically between those who got CMV infection and those who didn't was the number of CMV seropositive red cells that the individual received, not platelets. This, I have a little trouble interpreting given that presumably the same process is being used in both of these populations, so I can't quite interpret what that means. Then, I don't know whether these can be generalized populations where the donors have a much higher incidence of CMV seropositivity. This is clearly a pre-selected population, and with most of us dealing with donor populations, the 50 percent or more that are CMV seropositive, presumably, one would expect to see a higher rate of CMV transmission, but we really don't know. To touch briefly on the VAT study of viral activation in HIV, this was a double-blind, randomized study which enrolled individuals who were HIV-positive and had concomitant CMV infection, the issue being would transfusion of a non-leukocyte reduced component predispose to activation of either viral infection. So, the individuals were randomized to receive either unmodified red cells or leukocyte reduced by filtration red cells. The outcome measures reported were levels of HIV RNA, peripheral blood levels of CMV DNA, and overall survival. There was no difference in the baseline characteristics between the two groups or in terms of transfusion treatment. There was also no difference in overall outcome in terms of survival between the two groups. Looking at levels of viral nucleic acid following the first transfusion, there was no difference in terms of peripheral blood HIV RNA, and when broken out to patients who were not receiving any kind of antiretroviral therapy, there similarly was no difference between the two study groups, and there was no difference in the two study groups in terms of the amount of CMV DNA in peripheral blood following the initial transfusion of red blood cells. So, this study pretty well lays to rest the issue of whether transfused leukocytes in the setting of HIV infection might be resulting in either HIV or CMV reactivation. It does not, of course, address individuals who do not have HIV. Turning to alloimmunization, this is a large retrospective study reported out of British Columbia that looked at individuals undergoing stem cell transplantation for acute leukemia. This was a "before and after" study, the first period being prior to the introduction of universal leukocyte reduction, the second being afterwards. Importantly, between these two study periods, there was a significant change in practice in that the platelet transfusion threshold for prophylactic transfusions was reduced. The outcome measures were lymphocytotoxic antibody production and clinical platelet refractoriness. The patients were monitored weekly for HLA antibody and clinical refractoriness was defined as the occurrence of a corrected count index of less than 5 on two subsequent transfusions. There were some significant differences between the two groups, particularly in terms of the number of platelet transfusions that were received in the second group and the number of donor exposures that were in the second group, and this reflects primarily that they had become much more stringent in their threshold for prophylactic transfusions. As a surrogate marker and associated marker of refractoriness, individuals who received any HLA-matched single-donor platelets product were less in the second group, and the total number of such products that were transmitted were less in the second group. The overall rate of alloimmunization was approximately one-half of what was seen in the pre-leukocyte reduction period. The most significant impact was in individuals whose previous exposure had only been through transfusion. We don't know exactly what the type of product that they were transfused with. Presumably, it was non-leukocyte reduced although we really don't know. There was also a reduction in individuals who had been neither pregnant nor transfused, but in individuals who had been previously pregnant, there was not really a significant reduction, and that is similar to what was seen previously with the TRAP study. In terms of platelet refractoriness, there was an overall decrease to about one-half of the number of patients who became clinically refractory. Out of those, a smaller proportion were defined as alloimmune refractory, and that being the finding of both a poor corrected count index and development of positive or presence of positive tests for lymphocytotoxic antibodies within two weeks one way or the other of the transfusion event. Again, this is most significant in those who had had prior exposure only through transfusion, not very significant in terms of those who had been previously pregnant. A randomized study reported out of Europe from the Netherlands of alloimmunization in the setting of cardiac surgery found some slightly different data. These were randomized individuals who were undergoing cardiac surgery, receiving both intra-operative and post-operative transfusions, so they were randomized prior to the first transfusion. They received either buffy-coat depleted red cells, pre-storage filtered red cells, or post-storage filtered red cells. The clinical outcome measures were lymphocytotoxic antibodies developing at two intervals that they looked at 3 to 10 weeks following operation and 20 to 30 weeks. There was little difference between the three groups in terms of their transfusion requirements and prior exposure. About 40 percent of individuals had had prior exposure, and that is not surprising in a cardiac surgery population which tends to be older. All individuals in all arms received at least 2 transfusions. The mean was 4 with a range being 2 to 6, the range here shown in parentheses. Less than 10 percent of individuals received platelet transfusions. They don't specifically state whether or not those platelet transfusions were leukocyte reduced. They divided the outcome in terms of several groups. The first is those who were negative on lymphocytotoxic antibody testing before and then were negative afterwards, and that constituted around 65 to 70 percent of the groups, and there were overall no differences between the control and the two study arms. Those who were negative at the time of randomization and then became positive constituted about 10 percent, again no difference in terms of the groups. Those who were positive initially and then became stronger were about 6 to 7 percent. Those who were positive initially and then became negative were about 8 to 9 percent, and those who were positive initially and remained positive at approximately the same strength were again about 6 to 7 percent. So, they did not observe any statistically significant differences among these groups. So, this suggests that at least in the setting of a relatively small- and short-term exposure in cardiac surgery, that there may not be a very large effect in terms of development of alloimmunization. Turning to the data on febrile reactions, these are somewhat difficult to interpret because of both the way that the data are reported and how do you know that a patient has a febrile reaction. It is a diagnosis of exclusion, and even when very carefully performed, there are certainly more than a few instances where it is difficult to tell. As one who regularly signs out these reactions, I have to admit there are times when I am virtually flipping a coin trying to decide if this individual, who has been previously febrile, but isn't febrile today, has neutropenia, is on antibiotics, and cultures are negative, what kind of reaction he is having. In addition, the way the data are reported is convenient for blood bankers, but not necessarily convenient for really data analysis. They are in terms of numbers of reactions reported per total number of units transfused. Few studies actually report the number of patients transfused. Few studies report the number of patients transfused who didn't have reactions, and all of these rely on some kind of a passive reporting system. For all of us in blood bank, we know that these reactions are under-reported. So, those make it somewhat difficult to interpret. I extracted data from 6 retrospective cohort studies and 1 randomized controlled study where I could get these reactions in terms of total units transfused. I calculated confidence intervals based on those published data, and these are what I found. Out of these studies, all of them had lower rates following the implementation of leukocyte reduction, however, two of them what would here be labeled as 3 and 6 did not find statistically significant differences, the other ones did. So, this is certainly supportive of decreased rate of febrile reactions, but it is a little hard to interpret given the difficulties with making a diagnosis and the reporting data. The TRAP study in the initial publication has one line that says that there was no difference in terms of reaction rates between the groups, however, more recently that has been re-analyzed and from the reported reactions in the more recent analysis, I extracted those that were either febrile or showed rigor reactions, and pooled those together. Just comparing the group that received the pooled random donor unmodified concentrates versus the filtered random donor concentrates as the most comparable groups, the reaction rate was significantly reduced in those who had received the filtered concentrate, about 1.6 percent compared to 2.5 percent in the control group. In parentheses there is given the confidence interval. The reactions were most associated with components that were greater than 5 x 106 total white cells and storage period that was longer than two days. These are not surprising data. So, it does appear that leukocyte reduction decreases the rate of febrile reactions, but it certainly doesn't completely eliminate them. In terms of outcome studies, a randomized, controlled trial in the general hospital population, which is the Dzik trial, has received a great deal of attention. This specifically excluded individuals who had an indication for leukocyte reduction, so it was looking at presumably the non-targeted populations. The patients were randomized to receive unmodified red cells and pooled random donor platelet concentrates versus pre-storage reduced red cells and process-reduced single-donor platelet concentrates. The primary outcome measures were in hospital mortality and post-transfusion length of stay. They also did some cost analysis, but I am not going to be discussing that today. The patient characteristics between the two were relatively well balanced. About 60 percent were surgical and 40 percent non-surgical.Within the surgical group, the largest were cardiothoracic and orthopedic. So, this reflects a large tertiary care population. The primary outcome measures in terms of both mortality and length of stay, they demonstrated no difference. In terms of subgroup analysis, looking specifically at patients who underwent cardiac surgery, patients who underwent colorectal surgery, as two populations where there is suggestion that leukocyte reduction might reduce mortality or length of stay, they found again no statistically significant difference. That also was true for other surgical cases, as well as non-surgical cases. A number of concerns have been raised about this study, one being the age of the red cells that were used in the two arms, as that they were older in the leukocyte reduced group, however, these are still comparatively relatively fresh. Certainly compared to what we commonly have on the shelf at the University of Michigan, these would be considered pretty much equivalent. The source of the platelets in the two groups was different. There are pools of 6 random donor platelet concentrates in the control group, whereas, it was process reduced single donor concentrates in the leukocyte reduced group. Thus, there were more donor exposures in the control group, and potentially, there could have been a higher content of white cells in single donor platelet concentrates that would not have met leukocyte reduction criteria. There was an exclusion of individuals who had leukocyte reduced indications, and there was a fair number of protocol violations occurring in both groups. When the data were analyzed, rather than on the intention-to-treat basis, but in terms of the actual transfusion groups, again, they found no difference between these two populations. Much of the data come out of European studies, there have been relatively few within the United States. This is a randomized, controlled study out of Europe and the Netherlands, which looked at patients undergoing aortic aneurysm surgery, both emergent and elective, and patients undergoing gastrointestinal surgery. They were randomized to receive either buffy coat depleted red blood cells or filtered leukocyte reduced red blood cells. This was a double-blinded study. The principal outcome measures were mortality and ICU stay, and the secondary outcome measures being incidence of multi-organ failure, infection, and length of hospital stay. Whenever evaluating studies that look at infection as an outcome, there is always a problem with how those are defined. These investigators used CDC criteria for post-operative infection, which include nonculture-positive infection, such as pneumonia can be diagnosed on the basis of chest x-ray findings and clinical findings in the absence of a positive culture. So, there always is a question of whether or not a bias could be introduced when you are looking at outcomes that are not strictly based on objective criteria, such as a positive culture. On the other hand, undoubtedly, cultures are not always positive on individuals who are truly infected for a variety of reasons including that they often receive antibiotics prior to the time that those cultures are taken. So, this study randomized individuals at entry, so that about half of the randomized subjects actually got transfused. This means that on the intention-to-treat analysis, you are looking at a population where about half of them do not undergo the intervention, and when you do the analysis according to transfusion, you are eliminating about half of the individuals in both populations who were, in fact, randomized. So, looking at the data in terms of intention to treat, there was a significant difference in terms of length of stay in hospital, which favored leukocyte reduction, mortality within the group of individuals undergoing gastrointestinal surgery also was statistically better, favoring the group receiving leukocyte reduction in the other two groups, and overall, that did not reach statistical significance. In terms of multi-organ failure, while the total population was significantly better, favoring leukocyte reduction, the subpopulations did not achieve statistical significance. Then this was analyzed in terms of the subgroups who actually received transfusion, it was about, as I said, 50 percent of the individuals, and the overall mortality rate was not significantly different, and in particular, that group who had undergone gastrointestinal surgery where, in the intention-to-treat analysis, there was a significant difference, this no longer was, which suggests that individuals who are not transfused did better. There was a difference that remained in terms of lower length of stay. Overall, that favored leukocyte reduction. Another randomized, controlled trial coming out of the UK looked at cardiac surgery patients who underwent coronary artery bypass grafting, aortic valve replacement, mitral valve replacement, or a combination of those. They were randomized into three groups, one that received plasma-depleted red blood cells, one that received buffy coat depleted red blood cells, and one that received filtered leukocyte reduced red blood cells. The main outcome measures were hospital-acquired infections and length of stay and development of fever. This was not a blinded study, but the treating clinicians were unaware, so it's a semi-blinded study, similar to the Dzik trial, where no specific effort was made to either blind or inform the treating physicians, so it was in a sense a semi-blinded study. Again, the criteria for infection were similar in terms of they are about the same as the CDC criteria, again, for individuals in particular with pneumonia, there is a possibility for a clinical diagnosis. The number of percent of patients who actually were transfused within these two groups is presumably 100 percent although it is not specifically reported as the randomization occurred at the time of the order for the first transfusion. In terms of overall infection rates between the filtered group, the buffy coat depleted group, and the plasma reduced group, there was no significant difference, however, when they analyzed in terms of patients who actually were transfused according to protocol, they did find a difference in that the filtered and buffy coat depleted groups appeared to have a lower rate compared to the plasma reduced group although the infection rate in the plasma reduced group is relatively high, being about 20 percent. In terms of length of stay, they found no difference, being a median of 6 to 7 days, the range of 3 to 5 days. In terms of patients who had fever, there was a difference in that there was a lower rate of fever among those who received either the filtered or buffy coat depleted groups compared to the plasma reduced group. The number of patient days with fever compared to the number of potential days at risk were similar in the two groups. Fever is, of course, a surrogate marker of infection, but can mean other things including febrile non-hemolytic reactions, so it is a little bit difficult to interpret this. Another randomized trial coming out of Europe focused on again cardiac surgery in the setting of aortic or mitral valve replacement with or without concomitant coronary artery bypass grafting. This was a double-blinded study. Patients were randomized at the time of the first transfusion. About 90 percent of patients in both arms actually received transfusions, about 10 percent did not. Randomized again to a control arm, which was buffy coat depleted red cells versus those who, in the intervention arm, received filtered leukocyte reduced red blood cells. The main outcome measures were mortality at 90 days. Secondary outcome measures were in-hospital mortality, length of stay, and infections. Again, infections defined basically according to the CDC criteria. So, analyzed on the basis of intention to treat, so this includes individuals who were not transfused, there was no difference in terms of 90-day mortality, however, there was a difference in in-hospital mortality favoring the leukocyte reduced group. There was also a lower rate of documented infections in the leukocyte reduced group. There was no difference in terms of multi-organ dysfunction scores. Analyzing the patients who received transfusions, so this is excluding about 10 percent of the randomized individuals, again, there was no difference in the 90-day mortality, in-hospital mortality remained statistically significant, and a higher difference was seen in those individuals who received more units. So, those who received four or more units had an odds ratio that was favoring leukocyte reduction. There was also a difference seen in terms of total infections between the two groups, favoring leukocyte reduction. In a retrospective cohort study coming out of France that looked at individuals who underwent abdominal aortic surgery in two-year time frames, one prior to the implementation of universal leukocyte reduction and one after universal leukocyte reduction. The prior group received either unmodified or buffy coat depleted red cells, and the latter group received filtered leukocyte reduced red blood cells. Again, randomization occurred at the time of the first transfusion, so the transfusion rate in this study was 100 percent. This study also reported individuals who received autologous transfusions, however, I am not including that in this particular presentation, so I don't believe that it is particularly germane. The main outcome measures were 30-day mortality and again infections, similar criteria for diagnosis of infection as with the other studies, so that there is some clinical criteria. In terms of the two groups, in the leukocyte reduced group, there was a higher incidence of hypertension and the use of diuretics, so these might favor or predispose to worse outcome. On the other hand, there were a good deal less patients who had a diagnosis of coronary artery disease, had prior intervention for coronary artery disease, either angioplasty or bypass grafting, and a lower rate of patients with respiratory insufficiency, so those would favor better outcomes presumably within the second group, the leukocyte reduced group. So, there were some significant differences in baseline characteristics between the two groups. In terms of the major outcomes, there was not a statistically significant difference in short-term mortality between the two groups and in terms of documented infections between the two groups although there did appear to be a trend toward lower rates within the leukocyte reduced group. Given the differences between these two populations, it is a little hard to make a causal relationship there. A large retrospective cohort study of pre-and post-leukocyte reduction was reported out of Canada. This involved individuals who had received red cell transfusions for cardiac surgery, for hip arthroplasty, or for admission to a surgical ICU including patients who had trauma. It evaluated two years, one year that was prior to the implementation of leukocyte reduction, one year post implementation, with a period of time in between allowing for conversion, so the two groups were really quite distinct. They received either unmodified red blood cells during the first period or filtered leukocyte reduced red cells during the second period. The number of patients who received platelet transfusions is not explicitly stated. About 10 percent of individuals in both groups appeared to have had previous transfusions, so it is a little bit difficult to tell from the way that the data are presented. It was analyzed in terms of in-hospital mortality and nosocomial infections, and the criteria for diagnosis of infection were really quite tight and were well described within the paper. There were a few significant differences between the study populations in that there was a lower incidence of severe lung disease although that was a relatively small group. There was higher use of beta blockers, aspirin, and ACE inhibitors in the second group, the post-URL group, which I believe is important because the breakdown of the patients was about 65 percent of them were cardiac surgery, 25 percent were general surgical ICU patients, 10 percent were hip replacement, and also there was a slightly lower rate of transfusion within the post-URL group. Looking in terms of mortality in the unadjusted group, achieved statistically significance. In the adjusted group, as reported here was not different, however, I have a little bit of trouble understanding this given that as stated in the paper, when adjusted for the use of cardiac medications including aspirin, beta blockers, angiotensin- converting enzyme inhibitors, mortality shifted from significant to non-significant association. So, this figure does not quite seem to correspond with the text of the paper. There was not a significant association with infection, but there was a lower rate of fever and antibiotic use both in the adjusted and non-adjusted groups. Since fever is, of course, only an indirect marker of infection and antibiotic use tends to be driven by fever, these are somewhat indirect. A couple of meta-analyses of randomized clinical trials comparing allogeneic leukocyte reduced and allogeneic non-leukocyte reduced transfusions have been reported. They have reported somewhat different results. This was a Canadian group that included in their analysis 10 surgical studies that were divided among cardiac surgery, colorectal surgery, GI surgery, and a couple that included mixed surgical populations. There were large differences between these in terms of the number of patients who were transfused ranging from 2 percent not transfused to 73 percent transfused. Only one of these studies were the physicians blinded to the intervention, and a variety of leukocyte reduction methods were used. The authors of this study did not address the issues of homogeneity of studies. They did not do statistical analysis for that point of view, so it is unknown from this analysis whether or not non-homogeneity issues may have been introduced. They analyzed the outcome both in terms of intention to treat, that is, all patients randomized and a subgroup analysis of those who were only transfused. So, overall, in the intention-to-treat analysis, there was no statistically significant difference in mortality, however, within the group that was cardiac surgery, that did achieve statistical significance favoring leukocyte reduction. Within the subgroup analysis, looking at only patients who received transfusion, there was overall reduction in--I am sorry, I am looking at infection rates, I am getting ahead of myself--these are infection rates, overall, was not different, was lower in the subgroup analysis of individuals who were only transfused. Again, this held true for those who underwent gastrointestinal surgery. In terms of mortality rate, overall, did not achieve statistical significance, but for those who underwent cardiac surgery, it did lower rate favoring leukocyte reduction, and when this was broken down in terms of only patients transfused, while there still was not a statistically significant difference overall, it did appear that individuals who underwent cardiac surgery had a lower rate of mortality. There is some controversy about how one should analyze such groups, whether you should do intention to treat or subgroup analysis, and how to interpret these. When you do intention to treat, you are directly comparing the randomized population, so it is the cleanest analysis, whereas, that is not necessarily true when you are doing the subgroup analysis. When you do see positive effects in the trial and intention to treat, you can be quite confident that that is a real effect and that is why it is, for instance, in drug studies, is the most desirable way to do an analysis, because the confounders, such as individuals who are not treated, not transfused, would tend to dilute the power of the study, but if you see a positive effect, you can still believe in it. On the other hand, if you see a negative effect, the dilutional effect of having untreated individuals in both arms then becomes a problem. Subgroups may or may not be representative of the randomized population. If it is a blinded study, then, you can be relatively confident that the subgroups were not influenced by the treatment choice or treatment allocation. On the other hand, unblinded studies may be particular susceptible to the introduction of such bias. So, there are reasons to stick with intention-to-treat analysis, there are reasons not to. A second meta-analysis of randomized clinical trials that compared, on the intervention arm, allogeneic leukocyte reduced transfusions, but also included autologous red cell transfusions as in the intervention arm, compared to non-leukocyte reduced either whole blood or red cell transfusions in the control arm came up with slightly different types of studies and slightly different conclusions. This also included studies that were addressing cardiac surgery, colorectal surgery, and others including the Dzik study of mixed general hospital populations. Four of these studies were, in fact, studies of autologous blood rather than leukocyte reduced blood, so it would need to be looked at differently. The author of this study only looked at intention-to-treat analysis. Overall, combining these, there was not a statistically significant difference although a trend toward favoring leukocyte reduction. In particular, breaking these out in terms of cardiac surgery, though, there was a difference in short-term mortality that favored leukocyte reduction. Ignoring the studies which looked at autologous transfusion and picking out only those that used pre-storage leukocyte reduction in the intervention arm, the summary odds ratio just barely failed to achieve statistical significance although there was a trend that favored leukocyte reduction. The same author did a subsequent meta-analysis of "before and after" studies of universal leukocyte reduction, so these were cohort studies that reported either post-operative infection rate or mortality prior to or after the implementation of universal leukocyte reduction. A couple of these studies are ones that we have just gone over. These included two that were cardiac surgery, one aortic surgery, one orthopedic and cardiac surgery, which were broken down separately. Also included was a mixed surgical population and one that addressed the neonatal population. Again, in terms of looking at all of these and in terms of summary odds ratio just failed to achieve statistical significance although a trend appeared to favor leukocyte reduction. Of those that reported adjusted post-operative infection rates, presumably adjusting for other variables, again, it did not achieve statistical significance in terms of the overall odds ratio although a trend that appeared to favor leukocyte reduction. In terms of unadjusted mortality, similarly, it is notable that one study, which was the Williamson study out of the UK, which separately reported orthopedic surgery and cardiac surgery, the cardiac group did report a higher rate of unadjusted mortality in those who received leukocyte reduced than those who received the control group, and to my knowledge, this is the only such study which has suggested possibly adverse effect of leukocyte reduction. The authors attributed this to mortality at one particular center, so there may be a center effect. This study has reported meeting proceedings, but to my knowledge, has not been published yet in a peer-reviewed journal. So, those things make it a little bit more difficult to interpret. Of the studies that reported adjusted mortality rates, presumably adjusting for other factors, again, summary odds ratio did not achieve statistical significance, but did appear to favor leukocyte reduction. In terms of the use of leukocyte reduction in the neonatal population, there is very little data. There is one systematic review of clinical trials of leukocyte reduction in the neonatal setting. The authors, however, were unable to do a formal meta-analysis due to basically lack of data. They were able to identify two randomized controlled trials, one "before and after" study, and one non-randomized controlled trial. Two of these studies looked at CMV infection and two of them looked at alloimmunization. When these data were pooled, neither CMV infection nor alloimmunization achieved statistical significance although these was a general trend that appeared to favor leukocyte reduction. These are, as you can see, very small numbers of patients, so it is a little bit hard to interpret these. Basically, I don't think that we have enough data to really make a conclusion. So, based on all this, what do I take away from what I believe is the state of the current data? CMV transmission by leukocyte reduced components is low. Whether or not it is exactly equivalent to seronegative still I believe is not entirely answered. I believe the best data available remain the Bowden study as being the largest study that has directly compared those two populations. I don't believe that since then we have data that significantly contradicts that. I think it is clear that in the setting of HIV infection, leukocyte reduction doesn't alter the course or result in reactivation of viral infection. In terms of HLA alloimmunization and platelet refractoriness, they are both reduced by leukocyte reduction, they are not eliminated, and the power of leukocyte reduction appears to be greatest in the population that is heavily transfused and has not been previously exposed particularly through pregnancy, which is not surprising to us. So, whether the effect occurs or how significant it is if it does occur in the very brief transfusion episodes, such as isolated surgical transfusions remains to be shown. Leukocyte reduction does reduce, but it doesn't eliminate febrile transfusion reactions. I find it rather difficult to interpret the reports due to the subjective nature of the diagnosis due to the reliance on passive reporting, so it is hard to say what exactly that rate is. I have noticed, though, that what we are not seeing is the bed-shaking or rigor reactions to platelet transfusions that used to be pretty common in the Hem/Onc population. I can't attribute that directly to leukocyte reduction because I don't have the data, but it certainly appears to be pretty striking to me. Short-term mortality in cardiac surgery does appear to be reduced by leukocyte reduction. The effect is relatively modest, but does appear to be real. Whether or not long-term mortality is reduced is still controversial and remains to be addressed. Obviously, in cardiac surgery, there are a lot of other issues that are affecting long-term mortality. I don't think that it has been shown that leukocyte reduction has a significant effect on post-operative infections, however, interpreting these studies is somewhat difficult due to varying definitions of infection, which is very hard to do within the clinical setting. The beneficial effect of leukocyte reduction in the general hospital population has not yet been demonstrated, however, I think it is notable that no one has, with the possible exception of the Williamson study, shown an adverse effect of leukocyte reduction in the non-targeted populations. So, it does appear to be quite a safe procedure, so whether or not it has additional benefit in non-targeted populations outside of cardiac surgery, I don't believe has yet been demonstrated. The references are there for those who are interested and I think I have stayed on time. Thank you very much for your attention. [Applause.] DR. WILLIAMS: We have a little time, so if anyone has questions on this excellent summary. DR. EPSTEIN: Thank you for this very comprehensive overview. I guess it continues to trouble me that study after study, especially meta-analysis after meta-analysis shows a trend to favor leukocyte reduction for a number of parameters including infection and mortality. It has always raised in my mind the question of whether any single study is large enough to show statistical significance for a modest but real effect. I am just wondering, you know, I think you drew the correct conclusion where there is lack of statistical significance, but does it mean anything to you that you have seen so many different approaches show essentially the same trend with essentially the same magnitude. DR. DAVENPORT: Yes. My personal take on it is I do believe that there is a modest effect, but that it is going to require such an enormous study as to be virtually impossible to perform to absolutely prove it, and so we are left with the practical question of given that it appears to be a pretty safe procedure and the trend certainly looks like it is favorable, should we simply go ahead and implement it, and I guess that is exactly the problem that you are facing, and I defer that to the FDA. AUDIENCE PARTICIPANT: Rob, do we know what the risk is for the acquisition of CMV infection in a hospitalized immunocompromised non-transfused patient? DR. DAVENPORT: No, we don't, and that is an excellent question. From both the Nichols retrospective study and the Bowden prospective study, the rate of CMV infection in those who received CMV seronegative components, were CMV-negative to begin with, received transplants that were CMV-negative, so they are about as clean as we know how to get, is about 2 percent. So, that appears to be what our floor is, about 2 percent of individuals who, as far as we know, are not getting clear exposures through things that we are doing, are developing CMV infection. AUDIENCE PARTICIPANT: On your slides, I think it is the slide that says, "Concerns raised about the study," I don't know whether it is referring to the TRAP study or not. DR. DAVENPORT: I am sorry, I am referring to the Dzik study of the randomized controlled study in the general hospital population. AUDIENCE PARTICIPANT: So, the age of the red cells there, you said that there was concern about the difference between the control and the leukocyte reduced, 11.5 days and 18 days. When were the red cells leukoreduced, were they leukoreduced upfront or were they leukoreduced at day 18? DR. DAVENPORT: They were pre-storage leukocyte reduced, so presumably within 24 hours. So, if there is an effect, then, it is due to the storage of a leukocyte reduced component. DR. KLEIN: Thank you for a very nice overview and your extensive review. Did you find any data on other cell-associated viruses, such as Epstein-Barr virus, HHV-8, HHV-6? Also, while we all know that you don't eliminate graft-versus-host disease, did you see any evidence of any effect on graft-versus-host disease in any of your reading by reducing the number of leukocytes transfused? DR. DAVENPORT: I didn't see any clinical studies addressing this. There is a number of laboratory studies that indicate that other cell-associated viruses, particularly HTLV-I, EBV, are reduced by leukocyte filtration. I didn't see any clinical studies, but I am open to correction if you know of such. In terms of GVHD, we certainly have case reports of GVHD occurring in individuals who have received filtered products. We don't have, to my knowledge, data from either before or after studies or any other randomized controlled studies on the incidence of GVHD or the severity of GVHD, whether or not it is impacted by leukocyte reduction. I think that would be a terribly difficult study to do given how complex GVHD is, although I think it is a very worthwhile one to try to address. If anyone wishes to correct me? Okay. AUDIENCE PARTICIPANT: Rob, the European studies generally use as their control arm, buffy coat depleted cells, which we know have somewhat fewer white cells than what we use as non-leukocyte reduced in this country. Do you think that the degree of leukocyte reduction in buffy coat depleted cells sort of prejudices those studies against a significant outcome? DR. DAVENPORT: As far as any data I can speak to, I can only say that it appears that buffy coat depletion reduces the rate of febrile reactions, but I am not aware of any data that specifically address either outcomes, CMV infection, or alloimmunization. I believe that the number of residual white cells in a buffy coat depleted unit is too high to really make any difference, but yes, it is a potential confounder as we use a different method of routinely preparing our red cells than the Europeans do. DR. BIANCO: Besides the CMV studies, that is, where you were looking at an infection, before any of the other things that we expect from leukoreduction, all the studies that you have shown and that I have read, are population studies, that is, comparing one population to another population. Do you know of any longitudinal study of patients that will show that after a certain period of time of being exposed to a non-leukoreduced product, they will develop more fever, more alloimmunization, but on individual patients as we can do with an infectious disease? DR. DAVENPORT: No, I am not aware of any such study. The CMV studies have looked at out to 100 days, and they have only looked in the specific transplant populations. I am not aware of any study that has any systematically followed out individuals. MR. ENGEL: Rob, by the way, I think this was an outstanding review of a lot of the studies out there. The one study that raised a flag for me was Baron's study where you have the differences in the study groups between the hypertension, the diuretics. It appears that there is a significant difference between the control and the leukoreduction groups there. Is there any explanation for that? DR. DAVENPORT: I agree that that is a major confounder in that particular study. The authors did not address it. In doing any of these "before and after" studies, obviously, medicine marches on, and fortunately, outcomes are getting much better. AUDIENCE PARTICIPANT: Seeing that overall trends it seems favoring leukoreduction, wouldn't you say that the standard of care would favor giving a leukoreduced product versus not? DR. DAVENPORT: That is for the audience in this room to decide. I would suggest that there are some good reasons to move in that direction, and there doesn't appear to be a really compelling clinical argument against it. AUDIENCE PARTICIPANT: I want to re-emphasize something that you alluded to, and that is the studies that use CDC or NNIS definitions of infections invariably come up with infection rates that any hospital epidemiologist would close the unit for. DR. DAVENPORT: Yes, they appear to be alarmingly high. AUDIENCE PARTICIPANT: They are awful and they are not I am pretty sure from a clinical standpoint to be taken at face value. So, I think it is critical as you look at these studies to look at the hard endpoints, length of stay, and those sorts of issues. I would close any unit that had some of those infection rates at the hospital I am responsible for, and I think it is critically important not to rely upon the infection data to make a decision, which is different than saying which side of the fence or which top of the fence I am sitting on at this point. DR. DAVENPORT: Good point. CDR GILSTAD: Colleen Gilstad from the United States Navy. I have a question. If there is a loss of 15 percent of the potency of a red cell unit or a platelet unit via leukocyte reduction, why in the studies did they not show an increased number of unit usage in the two arms? DR. DAVENPORT: Presumably, because the way that we prescribe red cells is by units rather than by the actual content. Now, one of my other soapboxes is that we really ought to be labeling our units for what is in them and prescribing them on the basis of what is in them. I think it is a matter of the way our practice has evolved to simply give two units rather than to actually target it for a therapeutic dose. I think it is a failure on our part to practice transfusion medicine as scientifically as we could. Thank you very much. [Applause.] DR. ORTON: Thank you again, Dr. Davenport. Before I introduce our next speaker, I would ask that everybody please turn off your cell phones for those of you who haven't or put it on vibrate. Thank you. Our next speaker is Dr. Ed Snyder, who is Professor of Laboratory Medicine at Yale University School of Medicine. He is going to describe the Yale-New Haven Hospital Universal Leukocyte Reduction Program for us. The Yale-New Haven Hospital Universal Leukocyte Reduction Program DR. SNYDER: Thank you very much. It is a privilege to be here. I would like to just comment that we all have had our experiences with Tibi, and one of the poignant moments that I had with him that I often remember is when he told me--I asked him how he was doing, and he said fine. He said, "You know, Ed, there is three phases of a man's life. There is childhood, adulthood, and you are looking good." Three people today told me I was looking good, so thank you, Tibi, very much, and I am a firm believer that when you speak of someone and you talk about them, that they are alive at least in your heart. So, I think Tibi will live in our hearts for many, many, many years to come. I am here to talk about the leukoreduction process at Yale, what we have been doing, and I am going to rely heavily on the data that I have generated. First, my conflict of interest statement. I have to say this. I was told not to spend more than 10 minutes on it, so I will try not to, but I am a member of the Pall Board of Directors. I have been that for about four or five years. I do clinical trials for companies that are involved with leukocyte reduction, and those are the ones in white, plus we do studies for everybody because that is what I do for a living. However, I do have no corporate equity, I do not get paid by any of these companies. Any monies that come in for the fee for services go to Yale University, which pays my salary, and I have no--other than the Pall Board, being on that, and again I have no stock in any of the companies as shown in the proxy statement. So, let me just say that upfront. Hopefully, the data that I show you will justify my view of leukoreduction. At Yale-New Haven Hospital, just to give you a brief review, the total patients transfused are about 7,400 patients a year, so we transfuse about 620 or so patients a month, and that has been reasonably stable for a while. The total products transfused have dropped. There was a peak somewhere in the mid-nineties, which was our Liver Transplant Program, which I am pleased to announce has started again as of a week ago. So, we will hopefully see how that goes, but our total product usage has dropped, so we are about 48,000 components a year. It is about 6.7 products per patient transfused over the years. Our red cell usage is about 20,000 units per year. Our sole source supplier is the Connecticut Red Cross. Our random donor platelet usage has dropped dramatically. We use pooled random donor platelets. We had used originally about 12 when I first came to Yale a while ago and then we dropped it for a variety of medical and budgetary reasons, and are now using a 4-unit pool. We are using about 15,000 units a year. Again, they are 4-unit pools. So, we use about 5,000 pools a year roughly. Our single donor product use has spiked in 2004, because we realized that we get less loss of money by using single donor platelets in the outpatient because of the CMS reimbursement, so you lose less, that is income. So, we use about 1,000 units a year in the outpatient area only, and the inpatient, which is still under a DRG, gets the random donors. With that as a little bit of a background, the accepted benefits of leukoreduction, there are only three that I consider. I don't think we are ever going to see decreased length of stay. I don't think we are going to cure cancer, and I don't think we are going to end poverty by using leukoreduction filters. All we need to do from my perspective is decrease febrile reactions. I think that, in and of itself, is lovely. I spend more money giving factor VIIa into the sewer system at Yale-New Haven Hospital for patients who don't need it than I ever will spend on leukoreduced blood filters, and I will have a little bit of data to show you that. I think decreased CMV transmission occurs. We have not done a CMV test at Yale since the year 2000--knock on pressboard, no problems--and HLA alloimmunization, I also agree with the TRAP study, and I think Rob Davenport did an absolutely superb job of going through a huge amount of data in a very coherent way, and I thank you very much for that effort. Decreased alloimmunization is very difficult to test at Yale. We give very few HLA matched platelets anymore. For those patients who need them, we use cross-matched compatible, which is not an HLA typing process. We have published a paper in Transfusion, which is what I am going to talk about, and then I am going to update the paper. This was a paper that was listed as Paglino in Rob's review, his references, Reduction of Febrile but not Allergic Reactions. I will point out that 0.3 of an FTE Yale research nurse was funded by the Pall Corporation. We needed some support for a research nurse as part of the study. Pall provided the funds to the university that went to supplement this nurse's salary. So, again, there was no money other than just for that, but I do mention that in terms of full disclosure, and Pall had nothing to do with the writing of the paper, the data, or any of that. So, what the study consisted of was we had used selective leukoreduction for a whole period of time since I have been at Yale, and we had about a 30 percent level of leukoreduction. We looked at starting in January of '05 and then we considered that through October 1998 to be the period of about 30 percent, 25 to 30 percent. That was decided by the physician as to whether to use that for selective leukoreduction. Then, we went through a transition phase when we converted to 100 percent leukoreduction, when I decided that was what I wanted to do for the university and the medical center, and then we went for a universal leukoreduction phase, which started January 2000 through December 2002, at the time the paper was written. So, we had a pre-universal leukocyte reduction phase, which was about 30 percent, we had a transition period, and then we had a period where we were 100 percent or 99.9 percent leukoreduced starting about January 2000. We did that so I didn't have to divide the transition phase into, you know half went this way, half went that way, just make it its own unit or just ignore it for most of the calculations, so I could compare pre- versus post. So, this is not zero leukoreduction, it actually starts at 30 percent, so based on that, any improvement or loss of improvement could be for that 70 percent of non-targeted people. Now, the non-targeted people, you could say, I mean that is a euphemism. They are all targeted, it's a question of when. They are either targeted when you are responsible for the hospital budget or when your successor will be responsible for the hospital budget, but it is like we only put seat belts in cars where kids are bad drivers, because they are the ones that are likely to have the accident, you put it in so everybody can benefit from it. That's my philosophy. That's why I give you my conflict of interest way upfront, so you will know I am not a talking head for the corporation. I really believe that leukoreduction is helpful for patients, and I don't have to cure their cancer, but I may be able to make them feel better and prevent some CMV transmission along the way. So, these were the number of units that were evaluated, 85,000 units of red cells in the selective phase, 60,000 units in the universal pre-storage leukoreduction phase. For platelets, there were 100,000 units that we evaluated, and in the post-, there were 37,000 units, and I listed them as pools also for purposes of the evaluation. What we were looking for was febrile reactions which were really defined as febrile reactions that were reported to the blood bank. Again, it is passive, as Rob pointed out. So, evaluations of the reactions per month for the pre-storage period was about 6.3, and then dropped to 2.94 in the universal phase. This had a p-value of 0.001, 53 percent reduction as reactions per month. These are for red cells alone for febrile reactions. When done as reactions over the total, the rate dropped from 0.34 to 0.18, and as you can see, during the transition, there was a drop in the middle, 6.3 reactions went to 5, then, to 2.9, and the rate was 0.34, 0.29, 0.18, was about 47.1, which you would expect there to be comparability. For allergic reactions, there was really no change at all, as expressed the p-value, and you wouldn't expect there to be a change, but the purpose of looking at allergic reactions was to see if there was a problem with reporting of the reactions from the floor. So, that was for red cells. For platelets, we went from 7.6 reactions per month down to 0.43, 94 percent drop, similar with the pooled. That was with the pooled, and this was the reactions, the rate was 0.44 down to 0.04, and it was a 90.9 percent drop in the rate. Allergic reactions, 37 percent, 28 percent decrease. Some of that was due to the fact we had decreased the pool size during this time period from about 6 units to 4, so my thought was there were fewer reactions because there was less platelets being given, so there was less allergens being given. The rate was 0.1 to 0.09, so we didn't see a change really in the rate, there was just a drop that was not considered statistically significant in allergic reactions. Looking at it as a bar graph, this is the selective period of time. This is the transition. This is universal on the upper lefthand corner here with this slowly dying green thing, and as you can see the rate dropping. For the analysis, we did not look at the transition period when we did the analysis. For red cell, for allergic, there was no difference in the upper righthand corner. For platelets, there was a very marked drop. For febrile reactions, for allergic reactions, there really wasn't much change at all. So, then, the odds ratios bore this out, I am going to skip this. So, our conclusions were that--I am sorry, not conclusions--what we decided to do was there was always the concern that maybe our nurses were just doing a bad job, they were off doing their nails or he was looking out the window when the patient was rigoring their merry way, off the bed, onto the floor, and rolled out the door, so nobody caught it. So, we evaluated 500 patient charts, undertaken to see if the decline was due to under-reporting by the nursing staff. This is sort of like, you know, a container comes in, in San Diego, and you pick one bag a week to seek if there is cocaine in there, and you randomize that to is there cocaine anywhere in the shipment. We looked at 500 even though there were thousands of people being transfused because I couldn't devote an entire nurse to this process. So, we randomly selected 300 patients who received red cells and 200 who received platelets, all without a report of a reaction over the period of time. We reviewed the charts, and it says 100 patients were interviewed, but that should be 300, 300 patients were interviewed, 200 red cell out of the 300, and 100 platelet recipients out of the 200, to see if they had symptoms. The bottom line, as we detail in the paper, is that 1 patient had a fever of 100.2 that was not reported, and there was no change in nursing protocols, there was no change in our report of a transfusion reaction. We didn't start putting Tylenol or acetaminophen in the water supply. Everything had been the same, and I had been at Yale that entire period of time, so that nothing really had changed much in the way that we approach febrile reactions. So, my thought was this was not due to under-reporting. The conclusions were that there was a significant decrease in the rate of febrile reactions, the pre-storage universal leukoreduction decreased to platelets and to red cells. There was no change in the allergic reaction post-universal leukoreduction. The decrease was not attributable to nursing inattention, and the universal leukoreduction decreased febrile reactions and provided patient care. Our administration agreed that the improvement in patient care was acceptable as a reason to spend some dollars, so now let's update for you, and let me go through this because all the ensuing 8 graphs will express it similarly. The ordinate on the left is allergic reactions or febrile reactions. This is listed as the number of reactions, so that is the individual number per month, or it will be the rate. On the right ordinate is percent of leukoreduction. So, the ordinate on the right, the leukoreduction is seen in blue. It starts off about 20, 25 to 30 percent starting with 1995, and then moving to about the year 2000 when we slowly converted to 100 percent over several months, and then this period is the period of universal leukoreduction. Each of these black markers on the bottom abscissa is a year, and the red dots are reactions per month, so there are actually 12 red dots for each year period. There is a best fit curve that is put in by the program. So, what you see is for allergic reactions, we have some very banner months with very high allergic reactions up to about 7 or 8 reactions, and then others where we had relatively little, and as we went to full leukoreduction, I think you can see there was really not much change in the number of reactions, for allergic reactions to red cells. Then, the rate of reactions to red cells, the same data only plotted, the left ordinate is rate, and not number, and it should be pretty much identical, and as you can see, the curve is actually going up a little. For some reason, we have had an increase in allergic reactions. We haven't increased the number of red cells. It is that people are getting itchy, I guess, for whatever reason. For febrile reactions to red cells, we did see a change, and as you can see here, there is a drop. This drop was the drop of about 30 percent, that was statistically significant, and, yes, we still see febrile reactions to red cells, and I don't really understand why we see these many. There is something in red cells that is a cytokine that is in an inflammatory cytokine. I am no sure what it is. There is not much else in red cells besides hemoglobin and some RNA and some other things, but it is there, whatever it is, I believe. Actually, if you look, it is sort of a bimodal type of a--you could draw a line through here. This is the pre-storage. Then, you could draw a line through here. So, there is a drop of 30 percent. Have I cured cancer, have I saved some child who can now walk again? No, but we have decreased, I think, this is five years of data, so this is not like yesterday afternoon, and the total number of patients is 7,200 x 5, is about 35,000 patients. I think the data does speak for itself, that there has been a decrease in incidence of febrile reaction by about 30 percent. Again, we started off here with about 25 to 30 percent leukoreduction, so these are the non-targeted people supposedly. Then, the same thing for the febrile reaction, the rate is shown over here, as well. So, allergic reactions to platelets, now, again, we see a slight increase in the allergic reactions to platelets. The number, it is hard to explain exactly what is happening. DR. DUMONT: had a paper talking about sap running in the areas and its relation to CMV and other things. There are lots of allergens that occur during times of the year, both in Colorado, and I think he also did this with I think Dr. Elfath and looking at things in Virginia, and there are lots of things that go on in the allergen world that we don't understand very well, and it may be reflected in the patient population. Then, the allergic reactions for the rate again has been pretty minimal, and you could draw a line across it here, and the rate has been the same. For platelets, this is our data, and it is hard to believe that it had that much of a drop, 94 percent, but it has been sustained. During 2003, we had 1 or 2 reports for the entire year. Now, you could say, well, the nurses are missing something, but they are not missing 500 reactions. I mean no one is that nasty that they just ignore that many number of patients. The increase over here was the onset of the bone marrow transplant program and the transplant program in general, and this was the transition period, so I got on the blue line, the blue line to 100 just in time from my perspective, and most of these bumps are actually the same person who has had more than one reaction, because this is actual numbers. So, this is 1 reaction, this may be 2 over here. This is probably 3 or 4. That is basically one patient that was repeatedly transfused, who responded to a very low level of white cells. Then, if you look at the rate, it is the same. So, I am convinced. When I said there is something in red cells, why are we seeing such a good reduction in platelets, but not in red cells, I don't know, and that has basically been reported by others. So, the cost of leukoreduction, when we started, it was $300,000. It didn't cost us anything to leukoreduce, and these are pooled leukoreduced units prepared by the Red Cross. The cost was zero because of cost shifting, and the cost shifting and we also used reallocation and decreased outdating, which is well known to be voodoo economics, and that is perfectly lovely. Cost shifting and internal reallocation are the same thing, I just used two different words, so it would sound like they were two different things, but what we did was we stopped doing CMV testing, and if we CMV test--she is standing up, that's not a good sign. He had an hour and 15 minutes, I am going just two more minutes, I am sorry. It's 20 to $25 for a CMV unit, and if we did only a third of the units that we transfuse, which is 16,000 units for CMV, that would be $400,000. So, whether you pay me now or pay me later, you are spending the money. It is just a question of what you are spending it for. In 2004 dollars, it went up to about $420,000, again, platelets would be zero because we have shifted it all onto the red cells. So, yes, it does cost money. We have an $8 million budget. I do spend 0.5 percent for leukoreduction, and what am I getting? What is your saving in febrile reaction workups? Now, here is an area we could talk about for weeks, but we are not, just for mere seconds. I put it in, in terms of direct costs and incremental. So, you can look at all of this. We have blood bank, you know, the serology and the labor time and the microbiology, yada yada. The direct of $109 was for those of you who think, yes, we spend a lot of time working it up. For those of you who say give me a break, the tech is there, she is doing the same thing, he has got the same stuff, so I will give you 23 bucks, and if you want to throw in a replaced unit, because they had to send the unit back because of the fever, you can add a couple hundred dollars for the red cells or the platelets. So, I gave it to you both ways. You can have transfusion workup light, or you can have transfusion workup full boat. So, I view, over five years, I prevented 132 reactions a year. I prevented four reactions a month for red cells and seven reactions a month for platelets. I honestly think that we did that. That is 660 patients over five years, 132 a year. Even if it's 100 bucks, it's 13,000 and there is 66,000 I saved over five years, not to mention the number of patients that I think I helped a little bit. I can't think of a way that I could spend a few hundred thousand dollars that would benefit patients as well, when there are people who are spending that much in factor VIIa for people who don't need it, et cetera, et cetera. So, I think this is justification enough plus the fact that we have cost shifted the CMV, and it is essentially cost neutral because that would have--I can't worry about the people who are off 8 West or 8 Onc, who have oncology concerns, who were supposed to get CMV-negative leukoreduced, but I don't know about it, because the house staff is still looking for colonic polyps, their own colonic polyps for the first two months, and don't know what they are doing. We don't have to worry about that. That, to me, is an important issue--I am out of control here, Alan. [Laughter.] Two more slides. CMV testing. We have done 200 allotransplants at Yale over the past five years--I am not talking about autologous where we do about 100 or 200 a year, 150 a year--200 allos. Ten percent of them are CMV seronegative for both donor and recipient. None of them have seroconverted, and these people are monitored with a PP65 weekly for seroconversion. Twenty people is a very small number. Charlie Shiffer would think it is an insignificant number. He has taken blood bankers to task for having small n's in some of their studies. But the point is that they do weekly surveillance and we have had no seroconversion. We do not do CMV testing, and there has been no reported transfusion-transmitted CMV seroconversion of any patient. There was one seroconversion that occurred. It was a breastfeeding mother, and she was CMV-positive. So, I don't believe that we are doing our patients a disservice, even the allotransplantations, and this is in full conjunction with the oncology and the neonatologists. So, the conclusion is at Yale, decisions, as everywhere, as all hospitals, decisions are only made for medical reasons, to improve patient care, and at Yale, I believe universal leukoreduction is an improvement in patient care. I think we have the data. This is ongoing, I may come back every four or five years and update you, to your great dismay. Whether universal leukoreduction is the best thing since sliced bread, I don't know. We think it works at Yale, and I am willing to justify it financially to patients and to Ted Koppel on Nightline if I need to. Thank you very much. [Applause.] DR. ORTON: Thanks a lot, Ed. That's why we invited you. Our next speaker is Dr. Avery Nathens from the Department of Surgery at the University of Washington. He is going to talk about the Impact of Pre-storage Leukoreduction Among Transfused Trauma Patients. The Impact of Pre-Storage Leukoreduction Among Transfused Trauma Patients DR. NATHENS: I am Avery Nathens. I am a trauma surgeon at Harborview Medical Center at Seattle. I am a little out of my league in terms of background here, but it is a privilege to be here and a privilege to present these data. This is a randomized controlled trial that we just completed at the end of 2004. We have just finished data cleaning and data analysis, so these results are preliminary and we haven't quite submitted this for publication yet. In the trauma population and in critically ill surgical patients, there is a very clear association between the transfusion of allogeneic red blood cells and a high likelihood of infection and multiple organ failure, and this holds true whether it's trauma, elective major surgery, cardiac surgery, et cetera. The mechanism is thought to be related to transfusion-related immunomodulation. I won't go through the background here, but the sense is that the passenger leukocytes that contaminate every unit of red cells play a role in the induction of energy, as well, these cells play an important pro-inflammatory role in a number of clinical and experimental studies suggesting that there is this relationship. It is pretty clear and we have heard a lot of evidence supporting this that the transfusion of leukoreduced blood is associated with fewer febrile transfusion reactions, lower rates of platelet alloimmunization, and lower rates of CMV infection. The effects on infection risk and multiple organ failure is really inconsistent across studies depending on whether it is a North American study, a European study, the types of analysis, the risk of transfusion, as well as the baseline rates of infection. These all seem to impact on the likelihood of there being a positive effect. Our goal in this study was to evaluate the effect of pre-storage leukoreduction compared to standard allogeneic transfusion in patients with hemorrhage due to trauma. The reason we chose this population is (a) I am a trauma surgeon, and (b) these patients really are at very high risk for infection and multiple organ failure. The typical baseline rate of infection in a transfused trauma patient is around 30 percent. These patients also have large transfusion requirements. So, this was a double-blind, randomized controlled trial. We managed to get emergency waiver of consent, so patients would come in, they would be immediately cross-matched, and at the time of cross-match, undergo randomization. So, again, this was an emergency waiver of consent. These patients were randomized at the time of type and cross-matched to either standard packed cells or leukoreduced transfusions. These are the inclusion and exclusion criteria. We limited it to enter patients who were over the age of 17 as per IRB. We limited it to those who were transfused within 24 hours of injury. The baseline risk of infection drops dramatically for people who are later transfused, which is why we focused on this to get a very high baseline rate of infection to minimize our sample size. We excluded patients who had active infections, prisoners, an IRB requirement, those expected to die very quickly, those who received blood products at a transferring institution. Also, because of inventory concerns, we were required to exclude patients who were proven to be AB-negative or B negative, those with a positive antibody screen, and those who our regional blood center knew required irradiated, leukoreduced, or CMV seronegative products. So, those patients were unfortunately excluded post hoc. We randomized patients using a stratified scheme based on age and mechanism of injury. This is solely to ensure that there is equal numbers of patients with these risk factors in each group. The randomization was performed by a hospital-based transfusion support service once they received the request for a type and cross. Also, in patients who required uncrossed-matched blood, the transfusion support service would automatically randomize the patient as they brought study blood refrigerators to the emergency room with uncrossed-matched blood, either leukoreduced or regular. We would never know what the next randomization was going to be. That is, this was concealed allocation. So, again, there was no way to guess what the next patient was going to be, and the units were blinded and labeled for research purposes only. So, it was completely double-blinded. The subjects received study blood products for the shorter of the duration of this hospital admission of 28 days. The study products were again pre-storage leukoreduced within 24 hours of collection by a Pall filter, and quality control suggested, but 90 percent of the units were actually less than 1 x 106 with the remainder between a and 5 x 106. I was concerned that the leukoreduced blood, because Puget Sound Blood Center doesn't routinely administer leukoreduced blood, that there would be a differential in the age of the transfused blood, so we actually controlled the inventory, so that all patients received blood that was younger than 25 days. Anybody who required platelets received pre-storage leukoreduced apheresis platelets. The study endpoint is as follows. It is any infectious complication within 28 days of randomization. We actually followed these patients up for 28 days even if they were discharged. Based on some preliminary data, we have performed a sample size estimate with a baseline risk of infection of 30 percent, which is pretty typical for these patients. We expected a relative risk of infection in the leukoreduced group of 0.4. This is based on a study of colorectal surgery patients. It resulted in an absolute risk of 12 percent in the leukoreduced group with a power of 90 percent, an alpha of 0.5. We assessed we need about 117 patients per arm. Because this is a waiver of consent study, we anticipated that some patients would be randomized and would deny us consent, so we inflated the sample size calculation appropriately. We had a variety of secondary endpoints that we were interested in. One was the rates of multiple organ failure as assessed by the Marshall score, which is really a composite score of failure in 6 different organ systems, mortality, ventilator days, ICU and hospital length of stay. So, we randomized a lot of patients, and this addresses the issue of what do you do with patients who are never transfused. A lot of patients were randomized, a lot of patients were excluded post-randomization, and I have the breakdown on the next slide. 324 patients were actually randomized who met the inclusion criteria. These are patients who had no exclusion criteria and received blood within 24 hours; 56 refused consent. We have mortality data only on these patients. The IRB did not allow us to actually pursue obtaining primary endpoint information on these patients. So, using a modified intent-to-treat analysis, we have 268 subjects that form the meat of the rest of the presentation. Why were patients excluded? The vast majority of patients were excluded because they were not transfused within 24 hours, and there was a variety of other exclusions, as well, some because they were too young, a few patients were randomized that were not trauma patients. Prisoners were randomized because we did not know they were prisoners at the time. Some patients had prior transfusions for this injury event, and so on. The majority of patients were excluded because they were not transfused within the first 24 hours, but you will see proportions across both groups are the same. Again, this is blinded, so biases probably are not introduced here. These are the baseline characteristics in the modified intent-to-treat population. This is the average trauma population, typically male, typically in the high 30s, low 40s, equal racial distribution across both groups. The majority of patients were young with limited comorbidities, and almost 20 percent of patients had penetrating injuries. These patients are thought to be at higher risk because of perforated viscus and intra-abdominal problems. So, again, very similar baseline characteristics across the two groups. Injury severity was also similar across groups. Injury severity score is a simple anatomic way of describing how severe the injuries are. Anybody with an ISS of more than 15 is supposed to be severely injured, so severely injured population. Maximum IS score looks at the worst injury in all body regions, 6 being fatal, 5 being fairly severe. Again, equal distributions of severe injuries across both groups. Just over 50 percent of patients were in shock. Looking at some measure of shock, the Emergency Department lactate level was high, so these were sick patients, and the lowest ED hematocrit is demonstrated here, again equal between both groups, so a very nice randomization. We looked at their blood requirements. Their transfusion requirements across the entire hospitalization were similar with almost 6 units of blood. The transfusion requirements in the first 48 hours was also similar across both groups, and no differences in rates of transfusion of FFP, platelets or cryoprecipitate. We did a fairly good job of inventory control. The age of transfused blood across both groups was 16.8 days and 16.9 days, so no difference here. Now, this is the important slide. This looks at infectious complications which, as you recall, was our primary endpoint. If you look at pneumonia, rates were no different. We limited it to patients with an invasive diagnosis of pneumonia only. These were diagnosed as per either bronchoalveolar lavage or protected specimen brushings because of the difficulty in identifying patients with pneumonia. No difference. The confidence intervals here all include 1. Blood human infections no difference. Surgical site infections no difference. UTI is no difference, and the rate of any infection was about 12 percent lower in leukoreduced patients versus other patients, but this was not significant. Now, importantly, we used CDC and NNIS definitions here, but the only infection that did not require culture positivity were superficial surgical site infections. This is as per the CDC. Every other infection here required a positive culture. We then looked at whether there was interaction between study arm and severity of illness. We looked at patients with shock or without shock. Again, no difference in outcome. There is a trend here with patients in shock doing worse, but not significant. We looked at whether there was a dose response, people who are massively transfused versus those who weren't. Again, not much of a difference. We looked at patients with evidence of physiologic shock, looking at ED lactates. Again, no difference. Injury severity score no difference across high ISS versus low ISS. So, no matter how you look at these results, there appears to be no difference in rates of infectious complications in this patient population. What about second endpoints? We understand that infectious complications are difficult readout. Mortality was no different, whether it is hospital mortality or 28-day mortality. The median length of stay was similar across groups, 13 days versus 12 1/2 days. An ICU length of stay was similar. Ventilator days was also similar. We looked at Marshall Multiple Organ Failure Scores, and both groups received a score of approximately 5, so difference in the degree of organ dysfunction. There were some patients who either died early, that is, died before they can get an infection, and about 16 patients out of 268 were lost to post-discharge follow-up. So, we did a time-to-event analysis with the event being infection, and this is Kaplan-Meier analysis here, and you will see that the leukoreduced group, in red, has an identical curve to patients receiving standard blood here, in white. So, again, no difference in the time-to-event analysis either. This deals with the censoring that is inherent and people who die early are lost to follow-up. So, just to summarize, there is really no measurable effect in this high-risk population requiring transfusion whether you look at infectious complications, multiple organ failure, or resource utilization. There is lots of differences across studies, and Dr. Davenport has focused on this particular aspect. It is possible there is differences in the baseline risk of infections in the patient population with regard to their baseline immuno-inflammatory state. Study design and sample size also plays a role here. Given our baseline rates of infection, to achieve significance here, we would need about 996 patients per group. If we truly performed an intent-to-treat analysis where the rates of infection would be lower, because we would actually include patients who w