Influence of Pretransplant Panel-Reactive Antibody on Outcomes in 8,160 Heart Transplant Recipients in Recent Era

CARDIOVASCULAR Influence of Pretransplant Panel-Reactive Antibody on Outcomes in 8,160 Heart Transplant Recipients in Recent Era Lois U. Nwakanma, MD, Jason A. Williams, MD, Eric S. Weiss, MD, Stuart D. Russell, MD, William A. Baumgartner, MD, and John V. Conte, MD Divisions of Cardiac Surgery and Cardiology, The Johns Hopkins Medical Institutions, Baltimore, Maryland Background. Panel-reactive antibody (PRA) screening to detect HLA antibodies is an important part of evaluation for potential heart transplant recipients. We sought to determine how different levels of PRA affect outcomes in heart transplantation. Methods. A retrospective cohort study of using data reported to the United Network for Organ Sharing /Organ Procurement and Transplantation Network (UNOS/ OPTN) registry from January 1, 2000, to December 31, 2004, was performed. The association between PRA at transplant and primary end points, allograft and patient survival, as well as a secondary end point, rejection within 1 year, was analyzed. Results. Pretransplant PRA was reported for 8,160 (79.4%) of the 10,279 first heart transplant recipients during the study period. Panel-reactive antibody was 0% in 6,481 (79.4%) patients (group 1), 1% to 10% in 930 (11.4%) patients (group 2), 11% to 25% in 309 (3.8%) patients (group 3), and greater than 25% in 440 (5.4%) patients (group 4). Actuarial survival was significantly different among the four groups by Kaplan Meier method (p < 0.001). Furthermore, using PRA cutoffs of 0%, 10%, or 25%, the group with lower PRA had significantly better patient and allograft survival. Cox proportional hazard modeling revealed increasing PRA as a significant predictor of mortality (p < 0.001). However, when each group (2, 3, and 4) was compared with group 1 (PRA 0%), only group 4 (PRA > 25%) had worse survival on multivariate analysis. Patients with PRA greater than 25% confirmed by the flow cytometric technique had the worst overall survival. Rejection rate within 1 year after transplantation also significantly increased with increasing PRA. Propensity-matched patients demonstrated similar results. Conclusions. This large series of patients from the United Network for Organ Sharing database has demonstrated that elevated PRA remains a significant risk factor in a recent cohort of heart transplant recipients. Patients with PRA greater than 25% are at a particularly high risk. (Ann Thorac Surg 2007;84:1556 63) 2007 by The Society of Thoracic Surgeons Panel-reactive antibody (PRA) screening among patients awaiting transplantation is used to determine the presence of circulating antibodies to a random panel of donor lymphocytes. These are antibodies to HLA A, B, and DR in the serum. The recognition and measurement of the degree of sensitization is currently an important part of the evaluation of potential heart transplant recipients. This condition is termed humoral sensitization, and has been associated with both increased frequency of acute rejection and decreased survival [1 3]. The PRA result is usually presented as the percent of panel reactivity (ie, the number of wells with positive reactivity divided by the total number of wells tested 100). What value constitutes a significantly elevated PRA is variably defined, and may prompt a prospective crossmatch, that is, direct testing of the recipient s sera against Accepted for publication May 29, 2007. Presented at the Fifty-third Annual Meeting of the Southern Thoracic Surgical Association, Tucson, AZ, Nov 8 11, 2006. Address correspondence to Dr Conte, Division of Cardiac Surgery, The Johns Hopkins Medical Institutions, 600 N Wolfe St, Blalock 618, Baltimore, MD 21287; e-mail: jconte@csurg.jhmi.jhu.edu. the donor s lymphocytes. Generally, a positive donorspecific cross-match is considered a contraindication to solid organ transplantation. Although current standards in kidney transplantation mandate a routine cross-match between donor recipient pairs before transplantation, constraints of prolonged ischemic time do not allow for routine prospective cross-matches in heart transplantation (HTx). Cross-matching in HTx is generally performed for recipients with an elevated pretransplant PRA. A recent survey of adult heart transplant programs showed that there is considerable variation among centers about the timing of PRA determination, importance of PRA values, threshold for treatment or prospective cross-match, and the clinical implications of sensitization [4]. According to the survey, the most common criterion used for elevated PRA to be considered positive was a value of 10% or more, although there is considerable variation. Furthermore, different methods are used at different centers for PRA determination. The complement-dependent cytotoxicity assay is the older and most commonly used method. During recent years, many laboratories have applied more-sensitive serum screen- 2007 by The Society of Thoracic Surgeons 0003-4975/07/$32.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2007.05.095

Ann Thorac Surg NWAKANMA ET AL 2007;84:1556 63 PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION ing and cross-match techniques. The anti-human globulin augmented lymphocytotoxicity test, the complementindependent enzyme-linked immunosorbent assay, and the flow cytometric (FCM) assay are prime examples of these more sensitive techniques. Some advocate that the use of FCM technique is more predictable of outcomes [5, 6]. Only a few laboratories use more than one technique to assess PRA. The development of sensitization may occur as a result of previous blood transfusions, previous pregnancies, and more recently, the use of a ventricular assist device (VAD) as a bridge to heart transplantation. Some studies have suggested that although VAD support leads to increased sensitization this does not translate into worse outcomes [7, 8]. Thus, the acceptable threshold and the overall significance of elevated pretransplant PRA in HTx has not been clearly defined. To evaluate the effect of pretransplant PRA on outcomes in a large and recent cohort of heart transplant recipients, we analyzed the influence of different PRA cutoff levels, 0%, 10%, and 25%, using data reported to the United Network for Organ Sharing /Organ Procurement and Transplantation Network (UNOS/OPTN) registry, between 2000 and 2004. Patients and Methods Study Design and Patient Population This was a retrospective cohort study in primary heart transplant recipients (all retransplantations were excluded), analyzing the association between PRA at transplant and three outcome measures: patient and allograft survival, and rejection within 1 year after transplantation. All data reported to the UNOS/OPTN registry for HTx performed between January 1, 2000, and December 31, 2004, were analyzed. This period was chosen to represent a cohort of heart transplant recipients in recent era. The follow-up period was up to September 2005, which is when the Standard Transplant Analysis and Research (STAR) files were created. Individual patients are not identified in this multicenter registry report; therefore, the need for patient consent was waived by our institutional review board. Of the 10,279 first heart transplant recipients during the study period, most recent pretransplant PRA data were available on 8,160 patients (79.4%). Patients were then stratified into four groups: PRA level of 0% (group 1), 1% to 10% (group 2), 11% to 25% (group 3), and greater than 25% (group 4). 1557 Student s t test if values were normally distributed; if not, a Wilcoxon rank sum test was used. Categorical variables were tested using Pearson s 2 test. Survival estimates were based on the Kaplan Meier method, and statistical differences between survival curves were assessed by the log-rank (Mantel-Cox) test. Multivariate analyses were performed by multiple logistic and Cox proportional hazards regression. During the analyses for model selection, only variables with more than two thirds of available data were considered. The only exception made was VAD support because only about half of the patients have information available. To more accurately estimate confidence intervals, we made the assumption that the missing data were missing at random and performed imputation of missing data using the single regression method. Our findings proved to be robust across models that included both imputed and nonimputed (ie, casewise deletion of observations with missing data points) data. Our final model was chosen to reflect the effect of clinically relevant potential cofounders, namely recipient s age, sex, body mass index, donor s age, ischemia time, transfusions between listing and transplantation, life support (which includes any of the following: VAD support, ventilator support, intraaortic balloon pump, and extracorporeal membrane oxygenation), HLA mismatch, and transplant year. Another model was also analyzed in which VAD support was considered as a cofactor separately from other forms of life support only after imputation of missing data, as fewer than two thirds of the patients had data available about VAD support. In addition, to reduce bias and increase precision, propensity scores were generated by developing a model that estimates the probability of being in any of the patient groups based on clinically relevant variables listed above. Propensity scores were analyzed using the regression adjustment technique. That is, both the generated propensity scores and comparative PRA groups were included in a Cox regression model. Subsequently, we also obtained an additional dataset from UNOS that contains information about the techniques used for PRA determination and cross-match results and incorporated these data in our analyses. CARDIOVASCULAR Statistical Analyses and Survival The STAR files from the UNOS/OPTN registry were imported into Stata version 9.0 (StataCorp. College Station, TX), which was used for statistical computations. For all statistical tests, probability values less than 0.05 were considered significant. Continuous variables are summarized as median with interquartile range when they are not normally distributed, and categorical variables are expressed in frequencies and percentages. Bivariate comparison of continuous variables was performed using a Fig 1. Distribution of patients (pts) by most recent pretransplant panel-reactive antibody percent.

CARDIOVASCULAR 1558 NWAKANMA ET AL Ann Thorac Surg PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION 2007;84:1556 63 Table 1. Clinical Characteristics of First Heart Transplant Recipients 2000 to 2004 From the United Network for Organ Sharing Database Stratified by Most Recent Panel-Reactive Antibody at Transplant Most Recent PRA at Transplant Clinical Characteristics 0% (N 6,481) Group 1 1% 10% (N 930) Group 2 11% 25% (N 309) Group 3 25% (N 440) Group 4 Overall (N 8,160) Combined Age (y) (IQR) e 54 (41 60) 53 a (39 59) 50 b (33 59) 50 b (30 58) 53 (39 59) Sex (% Female) c 24% 27.7% 38.5% 52.5% 26.5% Donor s age (y) (IQR) 27 (19 41) 27 a (19 41) 27 a (18 42) 28 a (18 40) 27 (19 41) Pretransplantation diagnosis c Ischemic cardiomyopathy 38.5% 38.1% 31.4% 30.2% 37.7% Idiopathic dilated cardiomyopathy 33.1% 34.2% 35.3% 30.2% 33.2% Congenital heart disease 6.9% 5.2% 10.4% 16.6% 7.34% Others f 21.5% 22.6% 23% 23% 21.7% BMI (kg/m 2 ) (IQR) 25.5 (22 29.1) 25.4 a (22 29) 25.4 a (21.1 29.4) 24.5 b (20.8 28.3) 25.4 (22 29) PVR (Wood units) a (IQR) 2 (1.3 3) 2.1 (1.4 2.9) 2 (1.4 3.1) 2.1 (1.3 3.2) 2 (1.4 3) HLA mismatch ( 4) d 56.6% 57.7% 60.2% 57.4% 56.9% Pretransplant transfusions c 20.9% 28.7% 41.2% 44.6% 23 Waiting list time (days) 84 (26 252) 85 a (27 266) 87 a (31 221) 117 b (42 326) 86 (27 257) UNOS status 2 at transplant c 26.4% 22.3% 15.9% 16.4% 25% Positive retrospective cross-match c 2.8% 4.8% 9.6% 31.9% 86.1% Preoperative support Hospitalized c 53.8% 58.5% c 62.1% c 56.8% d 54.8% ICU d 33.5% 37.2% 36.9% 34.8% 34.1% Ventilator support c 4.6% 6.9% 7.1% 9.1% 5.2% VAD support g 39.3% 48.9% 60.1% 62.6% 43.1% IABP support c 4.7% 5.8% 9.4% 4.6% 5% Inotropes d 50.4% 46.6% 47.9% 46.4% 49.7% ECMO d 1% 1.4% 0.7% 2.1% 1.1% Ischemia time (h) a (IQR) 3.1 (2.4 3.8) 3.2 (2.5 3.9) 3.2 (2.4 4) 3.3 (2.4 4) 3.1 (2.4 3.8) Follow-up time (mo) (IQR) 26.2 (12.7 46) 25.1 a (12.3 45.7) 24.7 a (12 38.6) 24.4 b (12 38.6) 25.6 (12.6 45.6) a p 0.05 compared with PRA 0% (group 1) by Wilcoxon rank sum test. b p 0.001 compared with PRA 0% (group 1) by Wilcoxon rank sum test. c p 0.001 compared with PRA 0% (group 1) by Pearson 2 test. d p 0.05 by Pearson 2 test. e Data that are not normally distributed are reported as median with IQR ( 2 ) (25% 75%). f Others include many forms of cardiomyopathy, coronary artery disease, and valvular heart disease. g VAD support is the only variable analyzed in which less than two thirds of patients have data available. IQR interquartile range; ICU intensive care unit; VAD ventricular assist device; ECMO extracorporeal membrane oxygenation; IABP intraaortic balloon pump; PRA panel-reactive antibody; BMI body mass index; PVR peripheral vascular resistance; UNOS United Network for Organ Sharing. Rejection Comparisons A subgroup analysis was performed on recipients with at least 365 days of follow-up to evaluate the association between the development of rejection and PRA level. Patients were excluded from this analysis if they had fewer than 365 days of follow-up as determined by subtracting the date of transplant from the date that the data were captured in the dataset. A total of 1,684 patients (20.6%) were excluded with follow-up of fewer than 365 days. The chosen variable from the database was treatment for rejection within 1 year, which was used to define the rejection events. Other variables in the database that define rejection such as acute rejection episodes had too many missing data points (more than one third) and were not uniformly reported; thus, those were not used for analysis. Adjusted probabilities of rejection within 1 year after HTx accounted for the same clinically relevant cofounders used for survival analysis. Results Demographics A total of 8,160 first heart transplant recipients had data on their most recent PRA before HTx. The patients were stratified into four groups: PRA level was 0% in 6,481(79.4%) patients (group 1), 1% to 10% in 930 (11.4%) patients (group 2), 11% to 25% in 309 (3.8%) patients (group 3), and greater than 25% in 440 (5.4%) patients (group 4). Figure 1 depicts the distribution of these groups. The preoperative characteristics of all four groups of patients are shown in Table 1. Some of the notable differences among the groups with higher PRA are statistically younger age group, higher percentage of female patients, lower body mass index, more patients who received blood transfusions sometime between listing and transplantation, more patients with pretransplant diagnosis of congenital heart disease, and more patients who spent a longer time on the waiting list.

Ann Thorac Surg NWAKANMA ET AL 2007;84:1556 63 PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION 1559 CARDIOVASCULAR Fig 2. Kaplan Meier estimates of mortality stratified by four groups of panel-reactive antibody (PRA) percents. The groups with higher PRA also had fewer patients with UNOS status 2 at transplant, and the additional dataset obtained from UNOS showed that more patients with higher PRA had positive retrospective cross-match. Several variables of pretransplant support were variable among the four groups; however, VAD support was more prevalent among patients with higher PRA. Patient and Allograft Survival As shown in Figure 2, Kaplan Meier estimate of survival was statistically different among the four groups with worst survival among the patients with PRA greater than 25% (p 0.001). The 30 days, 1 year, 3 years and 5 years actuarial survival was 95%, 87%, 80%, and 74%, respectively, for group 1 (PRA 0%); 94%, 84%, 76%, and 71%, respectively, for group 2 (PRA 1% to 10%); 92%, 87%, 80%, and 73%, respectively, for group 3 (PRA 11% to 25%); and 89%, 80%, 72%, and 65%, respectively, for group 4 (PRA 25%). When the cohort of patients are divided into just two groups using PRA cutoffs of 0%, 10% or 25%, the group with the lower PRA had statistically significant better patient survival (p 0.001, p 0.007, and p 0.001, respectively; Fig 3). Fig 4. Kaplan Meier estimates of mortality conditional on survival to 30 days stratified by panel-reactive antibody (PRA) level. After propensity score matching of the comparative groups, results of survival estimates were still better for the patients in the lower PRA groups. Subsequently, when all the above analyses were performed evaluating for allograft survival, the patient groups with higher PRA emerged with worse allograft survival. However, when patients were analyzed conditional on survival to 30 days, the effect of PRA on patient and allograft survival was no longer present (Fig 4). Multivariate Analyses of Patient Survival In a multivariate Cox proportional hazard model that was adjusted for recipient s age, sex, body mass index, donor s age, ischemia time, transfusions between listing and transplantation, life support, HLA mismatch, and transplant year, PRA emerged as a significant predictor of overall mortality when analyzed as a continuous variable. When each of the PRA groups 2, 3, and 4 are individually compared with the reference group 1 (PRA 0%), only group 4 (PRA 25%) was a significant predictor of mortality in the Cox proportional multivariate model (Table 2). Multivariate logistic regression analysis was also performed, and PRA also emerged as a predictor of 30 days mortality (p 0.001). Another model was also analyzed in which VAD support (after imputation of missing data) instead of life support was analyzed separately, and similar results were obtained. Table 2. Multivariate Cox Proportional Hazard Regression Analyses of Patient Survival Risk Factor a Hazard Ratio 95% Confidence Interval p Value PRA (continuous variable) 1.005 1.002 1.009 0.001 PRA 1% 10% b 1.17 0.96 1.42 0.12 PRA 11% 25% b 0.94 0.99 1.00 0.73 PRA 25% b 1.4 1.09 1.77 0.007 Fig 3. Kaplan Meier estimates of mortality stratified by panel-reactive antibody (PRA) 25% versus panel-reactive antibody 25%. a Each category was analyzed separately with potential cofounders: recipient s age, sex, body mass index, donor s age, ischemia time, transfusions between listing and transplantation, life support, HLA mismatch, and transplant year. b PRA 0% is used as the reference group. PRA panel-reactive antibody.

CARDIOVASCULAR 1560 NWAKANMA ET AL Ann Thorac Surg PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION 2007;84:1556 63 Table 3. Multivariate Logistic Regression Analyses of Rejection Within the First Year After Transplantation Risk Factor a Odds Ratio 95% Confidence Interval p Value PRA 0% 0.9 0.76 1.08 0.23 PRA 1% 10% 0.91 0.72 1.14 0.41 PRA 11% 25% 1.15 0.81 1.64 0.43 PRA 25% 1.4 1.07 1.92 0.017 a Each category was analyzed separately with potential cofounders: recipient s age, sex, body mass index, donor s age, ischemia time, transfusions between listing and transplantation, life support, HLA mismatch, and transplant year. PRA panel-reactive antibody. Rejection Within 1 Year After Transplantation Unadjusted rates of rejection were 35.3%, 32.2%, 37.7%, and 45.5% in groups 1, 2, 3, and 4, respectively (p 0.001 for group 4 [PRA 25%] compared with reference group 1 [PRA 0%]). Multivariate adjustment of the probability of rejection within the first year after transplantation was performed to account for potential cofounders using multiple logistic regression model as shown in Table 3. Once again, patients with PRA greater than 25% demonstrated a statistically significant increase in the rate of rejection within 1 year. Peak Panel-Reactive Antibody at Transplant All the above analyses were repeated after replacing the variable most recent PRA at transplant with peak PRA at transplant, which is also well represented in the UNOS database, and comparable results were obtained. That is, among all comparative groups, Kaplan Meier estimate of survival was significantly better for the group with lower peak PRA. Peak PRA was a predictor of 30-day mortality, and the effect of peak PRA on patient and allograft survival was also lost when only patients who survived more than 30 days were analyzed. In the Cox proportional hazard model, increasing peak PRA was a significant predictor of mortality; however, unlike the result described above, PRA greater than 25% compared with PRA 0% did not emerge as a predictor in the multivariate model. Peak PRA greater than 25% remained a significant predictor of rejection within 1 year of transplantation in the unadjusted and adjusted models. Techniques Used for Panel-Reactive Antibody Determination We further requested an additional dataset from UNOS containing information about the technique used for PRA determination and results of retrospective cross-match. These data were merged into the original STAR files. Of the 8,160 patients, 7,003 (86%) patients had information about the technique used for the determination of the most recent pretransplant PRA. The complementdependent cytotoxicity technique was used in 2,576 patients (37%), the anti-human globulin augmented lymphotoxicity test was used in 2,761 patients (39%), and the FCM technique was used in 1,664 patients (24%). The enzyme-linked immunosorbent assay technique was reported in only 2 patients and thus was not included in these analyses. Among the patients with PRA greater than 25%, the anti-human globulin augmented lymphotoxicity test or FCM technique was the method used for PRA determination in 70% of the patients. All the above multivariate analyses were repeated controlling for techniques used for PRA determination, and only PRA greater than 25% still emerged as a significant predictor of worse survival. Additional subanalyses of the techniques revealed that the patients with PRA greater than 25% determined by the FCM technique had the worst overall survival. Furthermore, when all the patients with positive retrospective cross-match were excluded (350 patients), PRA greater than 25% remained a significant predictor of worse survival only among those patients for whom PRA was determined by the FCM technique. Comment Although PRA is an important component of pretransplant screening in HTx, there is still no clear consensus regarding the importance of PRA values, threshold for treatment or prospective cross-match, and the clinical implications of sensitization. This study of a large cohort of first heart transplant recipients from 2000 to 2004 demonstrated significant differences in survival and rejection in four distinct PRA groups. Increasing PRA as a continuous variable continues to be a significant predictor of mortality in univariate and multivariate analyses. The standards of the American Society of Histocompatibility and Immunogenetics and UNOS recommend that pretransplant cross-matches be performed only when the PRA activity of potential cardiac allograft recipients is 10% or greater [9, 10]. Even though PRA activity does not necessarily predict that a donor-specific cross-match will be positive or that the heart will be lost to hyperacute rejection if transplanted without a prospective cross-match, several studies report that a PRA of 10% or greater affects the duration and incidence of rejections after transplant, as well as patient survival [1 3]. Loh and colleagues [1] evaluated 125 heart transplant recipients and reported that an elevated PRA value of 25% or greater at the time of HTx may be a risk factor for decreased long-term survival. A trend toward an increased risk of death caused by rejection was also observed. Lavee and associates [2] demonstrated in a cohort of 463 heart transplant recipients that PRA greater than 10% is a risk factor for rejection-related events, and a negative lymphocytotoxic cross-match in patients with an elevated PRA does not reduce the risk of death resulting from acute or chronic rejection. These findings were supported by the report of 311 patients from Kobashigawa and colleagues [3], who also found that patients with PRA of 11% or greater at the time of transplant appear to have earlier and more severe rejection with significantly lower survival after transplant surgery, despite negative donor-specific cross-match. One of the major issues about determining truly sensitized patients is the fact that different approaches are used in different laboratories for PRA screening [4]. Historically,

Ann Thorac Surg NWAKANMA ET AL 2007;84:1556 63 PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION PRA analysis to detect HLA antibodies has been performed using cell-based complement-dependent cytotoxicity techniques and more recently the anti-human globulin augmented lymphocytotoxicity test. Enzyme-linked immunosorbent assay and FCM assay were introduced as an alternative approach to detect HLA antibodies. The latter seem to overcome the drawbacks of complementdependent cytotoxicity as a more sensitive technique and the ability to detect class II antigens. Flow-cytometric technique detectable pretransplant antibodies have been shown to better differentiate sensitized versus nonsensitized patients in terms of outcomes after HTx [5, 6]. With these developments it appears logical to evaluate the influence of PRA in a recent cohort of patients to see whether they are comparable to previous reports when only the complement-dependent cytotoxicity technique was used, although some major centers started using FCM techniques in earlier years. Furthermore, the proportion of these so-called sensitized patients has been rising with the increased use of VAD as a bridge to HTx [7, 11]. A single center report and another report using the International Society for Heart and Lung Transplantation (ISHLT) registry have suggested that although VAD use does increase sensitization, these patients have comparable outcomes to nonbridged patients [7, 8]. This raises the important and persistently controversial question regarding the overall influence of elevated PRA on outcomes in HTx. A potential explanation for the lack of influence on outcome in VAD patients maybe the finding from some studies, which have shown a temporal pattern of HLA sensitization during VAD support [12]. This temporal pattern consisted of a rapid increase followed by a rapid progressive decrease, and several mechanisms have been proposed [12]. One may rationalize from these findings that although VAD support increases PRA, the level may be normalized at the time of transplantation. This is also suggested by our findings that poorer survival among patients with PRA greater than 25% on multivariate analysis is no longer significant when most recent PRA at transplant is replaced with peak PRA at transplant. Thus, the reports that elevated PRA in patients supported with VAD does not influence HTx outcomes may not necessarily translate to other patients with elevated PRA. Modern improvements and better outcomes in HTx have emboldened more programs to offer HTx to patients with elevated PRA. We hypothesized that despite advancements in HTx, humoral sensitization as reflected by PRA is still a significant predictor of mortality and rejection in the recent era. We explored our hypothesis using data available on first HTx recipients reported to UNOS from 2000 to 2004 (all retransplantations were excluded). Unlike several reports in the literature that use a 10% cutoff to determine positive PRA result, we analyzed PRA as a continuous value from 0% to 100%. We used all available information about PRA in the database and selected most recent PRA at transplant as the most complete and reliably reported variable. Furthermore, we also divided the patients in four clinically relevant 1561 groups: group 1 is PRA 0% (which is also used as the reference group in most of the analyses), group 2 is PRA 1% to 10%, group 3 is PRA 11% to 25%, and group 4 is PRA greater than 25%. This allowed for evaluation of outcomes in these separate groups. Additionally the entire patient cohort was divided into two groups using PRA level cutoff points of 0% or 10% or 25%. Pertinent clinical characteristics that were more prevalent in sensitized patients included younger age (may be related to enhanced immune responsiveness), female recipient (likely associated with pregnancies), transfusions between listing and transplantation (presence of HLA antigens on blood cells), more likely to be on VAD support (possibly related to interactions between the ventricular assist device surface and host immune system), and having spent longer time on the waiting list (could be related to duration of immunomodulatory therapy). These findings are in agreement with predictors of sensitization reported in the literature [12 14]. Some other findings are the higher prevalence of congenital heart disease and fewer patients in UNOS class 2 status. Our analyses clearly demonstrated that elevated PRA (as a continuous value greater than 0%) was a predictor of mortality by Kaplan Meier unadjusted estimates as well as in an adjusted multivariate Cox regression model. When the cohort was analyzed at different cutoff levels described above, the group with lower PRA always had better survival by log rank test. Comparison of each of the other three groups (2, 3, and 4) to reference group 1 (PRA 0%) showed that group 4 (PRA 25%) remained a significant predictor of mortality in a multivariate model. Similar results were obtained when the above analyses were also performed for allograft survival. Notably, these effects of PRA on allograft and patient survival were no longer present when only patients who survived more than 30 days were analyzed. Likewise, PRA also emerged as a significant predictor of 30-day mortality. These findings suggest that the influence of pretransplant PRA on outcomes in HTx is an early phenomenon. The most reliably reported variable about rejection in the database is treatment for rejection within 1 year after transplantation. Sensitized patients with PRA greater than 25% demonstrated statistically significant higher rejection rates compared with PRA 0%, in an unadjusted and adjusted model. Additionally, patients were propensitymatched according to baseline characteristics, and the same findings were noted for all the outcome variables, namely, allograft and patient survival as well as rejection. Comparable results were also obtained when peak PRA at transplant was substituted for most recent PRA at transplant ; however, none of the groups remained significant on multivariate analyses when compared with PRA 0%. This would suggest that most recent PRA at transplant is a more important predictor of outcome, because some patients with high peak PRA at transplant may have their PRA level normalized or reduced at the time of HTx. The current study has the inherent limitations of the review of a multiinstitutional voluntary registry. Analysis of data may be compromised by lack of uniformity of reporting, nonstandardized method of PRA screening, and other pertinent variations such as organ preservation, surgical CARDIOVASCULAR

CARDIOVASCULAR 1562 NWAKANMA ET AL Ann Thorac Surg PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION 2007;84:1556 63 techniques, and postoperative and immunosuppressive management. Only first heart transplant recipients were included to help achieve a more uniform cohort of patients; moreover, retransplantation may significantly increase sensitization. The 2000 to 2004 era was also selected to reflect the advancements that have recently evolved in HTx. We also do not have enough information to evaluate why these patients died; it could be that they were overimmunosuppressed because of high PRA, which could lead to higher incidence of infection. One of the major disadvantages of database analyses is the large amount of missing data in the reporting of some important variables, such as acute rejection episodes, previous pregnancies, and primary cause of graft failure or death. We compensated for these issues by analyzing only variables with more than two thirds available data except for VAD support as described above; however, statistical imputation of missing data showed that the results of our analyses were robust. Poorer outcome in the sensitized group may also be a reflection of other variables that were not accounted for. Our analysis of the UNOS database from a recent era showed that PRA is still a significant risk factor in the outcomes of HTx, and its effect appears to be an early phenomenon. These data suggest that patients with PRA greater than 0% have worse posttransplant outcomes than those with PRA of 0% and should still receive careful evaluation before undergoing HTx. Patients with PRA greater than 25% are at a particularly high risk, and pretransplant lymphocyte cross-match should be considered, especially if confirmed by the FCM technique. Further advancements in HTx will better define the future role of pretransplant PRA. This work was supported in part by Health Resources and Services Administration contract 231-00-0115. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. Doctor Nwakanma is a Hugh R. Sharp Cardiac Surgery Research Fellow. Doctors Williams and Weiss are Irene Piccinini Cardiac Surgery Research Fellows. References 1. Loh E, Bergin JD, Couper GS, Mudge GH Jr. Role of panel-reactive antibody cross-reactivity in predicting survival after orthotopic heart transplantation. J Heart Lung Transplant 1994;13:194 201. 2. Lavee J, Kormos RL, Duquesnoy RJ, et al. Influence of panel-reactive antibody and lymphocytotoxic crossmatch on survival after heart transplantation. J Heart Lung Transplant 1991;10:921 9. 3. Kobashigawa JA, Sabad A, Drinkwater D, et al. Pretransplant panel reactive-antibody screens: are they truly a marker for poor outcome after cardiac transplantation? Circulation 1996;94(Suppl):II-2947. 4. Betkowski AS, Graff R, Chen JJ, Hauptman PJ. Panel-reactive antibody screening practices prior to heart transplantation. J Heart Lung Transplant 2002;21:644 50. 5. Tambur AR, Bray RA, Takemoto SK, et al. Flow cytometric detection of HLA-specific antibodies as a predictor of heart allograft rejection. Transplantation 2000;70:1055 9. 6. Gebel HM, Bray RA. Sensitization and sensitivity: defining the unsensitized patient. Transplantation 2000;69: 1370 4. 7. John R, Lietz K, Schuster M, et al. Immunologic sensitization in recipients of left ventricular assist devices. J Thorac Cardiovasc Surg 2003;125:578 91. 8. Joyce DL, Southard RE, Torre-Amione G, et al. Impact of left ventricular assist device (LVAD)-mediated humoral sensitization on post-transplant outcomes. J Heart Lung Transplant 2005;24:2054 9. 9. American Society for Histocompatibility and Immunogenetics. Standards for histocompatibility testing. Lenexa, KS: American Society for Histocompatibility and Immunogenetics, 1998. 10. The United Network for Organ Sharing. Standards for histocompatibility testing. Richmond, VA: United Network for Organ Sharing, 1998. 11. Gonzalez-Stawinski GV, Atik FA, McCarthy PM, et al. Early and late rejection and HLA sensitization at the time of heart transplantation in patients bridged with left ventricular assist devices. Transplant Proc 2005;37:1349 51. 12. Kumpati GS, Cook DJ, Blackstone EH, et al. HLA sensitization in ventricular assist device recipients: does type of device make a difference? J Thorac Cardiovasc Surg 2004; 127:1800 7. 13. Massad MG, Cook DJ, Schmitt SK, et al. Factors influencing HLA sensitization in implantable LVAD recipients. Ann Thorac Surg 1997;64:1120 5. 14. Johnson MR, Naftel DC, Hobbs RE, et al. The incremental risk of female sex in heart transplantation: a multiinstitutional study of peripartum cardiomyopathy and pregnancy. Cardiac Transplant Research Database Group. J Heart Lung Transplant 1997;16:801 12. DISCUSSION DR W. STEVES RING (Dallas, TX): It is an excellent study and I would like to congratulate the authors on a very nice presentation and appreciate the opportunity to review the manuscript in advance. Use of these national databases opens an opportunity for quite a bit of clinical research in the field of transplantation, similar to what the STS database does, and a lot to be gleaned from these. Unfortunately, as with the STS database, one of the major problems with the UNOS (United Network for Organ Sharing) database is the lack of complete data in many areas, and this may have hampered some of the analyses for you. I have got several questions for you. First of all, you have shown that PRA (panel-reactive antibody) does have an impact, largely for high PRA, within the first 30 days. Where you did utilize life support as one of your variables, were you able to tease out ventricular assist devices (VADs) separate from other forms of life support such as balloon pump and ventilator status at the time of transplant and able to distinguish that as an independent variable in these patients, because I suspect it will affect the outcomes? In other words, how much of the high PRA is related to the use of VADs, which actually have been shown to have a higher early mortality rate than non-vad patients in the UNOS registry? The second question I have is, PRA is an evolving concept. Historically it has been the cytotoxic antibodies using the classic AHG (anti-human globulin augmented lymphotoxicity test) wash or NIH (National Institutes of Health) protocol well that

Ann Thorac Surg NWAKANMA ET AL 2007;84:1556 63 PANEL-REACTIVE ANTIBODY IN HEART TRANSPLANTATION you describe where it is a percentage of reaction and cell lysis to a panel of something less than 100 random donors that are supposed to represent the population. As we have moved through the ELISA (enzyme-linked immunosorbent assay) and the flow cytometry and more recently to the use of the bead movement bead technology we have been able to distinguish class I from class II antibodies a little bit better, and more specifically, to separate out donor-specific antibodies. Now, I know the UNOS registry doesn t collect a lot of this data currently, but the question I have is related to a study that we recently presented at the American Transplant Congress whereby we were able to show that PRA was, or the outcomes were, related to not the nondonor-specific antibodies but purely to the presence of donor-specific antibodies in the population. So that PRA, per se, may not be the critical point here. It may be the presence of donor-specific antibodies in a population where you are not doing prospective cross-matches. And the follow-up of this question is, I was a little surprised that you did not use either prospective or retrospective cross-match as one of the variables in your analysis, because it may be that it is not the PRA per se that is the problem but it is the presence of a positive donor-specific cross-match that is contributing to the early poor outcomes. Thank you. DR NWAKANMA: Thank you, Dr Ring, for your comments. Your first question was to separate VAD support from life support. We did think about that. As you acknowledged, one of the disadvantages of using the UNOS database is incomplete data. One of the ways we have been able to make reliable conclusions using the database is to use only variables where at least two thirds of the patients have data available, because 1563 whenever we do imputation of missing data, if two thirds of the data are available, then the results are usually similar. Unfortunately, VAD support happens to be one of those variables where less than two thirds of the data were available. However, we did still analyze it and did find, like you would expect, that patients with higher PRA or patients that were on VAD support had higher PRA, and it did not affect the survival. But one needs to be careful to interpret these data because, again, the data were incomplete for that variable. Your next question was about the technique used. We do appreciate the fact that with the new flow cytometric method you are more able to detect class II antigens from class I. Unfortunately again, the UNOS database does not specify the type of technique that was used. We just expect by using a recent cohort of patients we may be able to capture more patients where the new technique is used as opposed to the CDC (complement-dependent cytotoxicity) technique. The next question is about the presence of donor-specific antibodies. That was a very, very important question that we went back to address with UNOS, because the standard transplant research files which they provide whenever one asks for data from the UNOS provides a variable called cross-match done. It does not tell you whether the cross-match was done prospectively or retrospectively or when it was done. It just tells you yes or no. So we are going to go back to have them reload the information and give us exactly when it was done, prospectively or retrospectively; then we will be able to analyze whether it was the donor-specific antibodies that affected survival. So thank you for those comments. We do acknowledge those limitations and we will try our best to include more information in our future analyses as we have them available. CARDIOVASCULAR