Author + information
- Received January 22, 2015
- Revision received April 7, 2015
- Accepted April 18, 2015
- Published online September 1, 2015.
- Ryan J. Butts, MD∗∗ (, )
- Andrew J. Savage, MD, MSCR∗,
- Andrew M. Atz, MD∗,
- Elisabeth M. Heal, DO∗,
- Ali L. Burnette, FNP†,
- Minoo M. Kavarana, MD‡,
- Scott M. Bradley, MD‡ and
- Shahryar M. Chowdhury, MD, MSCR∗
- ∗Division of Cardiology, Department of Pediatrics, Medical University of South Carolina, Charleston, South Carolina
- †Department of Transplant Services, Medical University of South Carolina, Charleston, South Carolina
- ‡Division of Pediatric Cardiothoracic Surgery, Department of Surgery, Medical University of South Carolina, Charleston, South Carolina
- ↵∗Reprint requests and correspondence:
Dr. Ryan Butts, Department of Pediatrics, Medical University of South Carolina, 165 Ashley Avenue, MSC 915, Charleston, South Carolina 29425.
Objectives This study aimed to develop a reliable and feasible score to assess the risk of rejection in pediatric heart transplantation recipients during the first post-transplant year.
Background The first post-transplant year is the most likely time for rejection to occur in pediatric heart transplantation. Rejection during this period is associated with worse outcomes.
Methods The United Network for Organ Sharing database was queried for pediatric patients (age <18 years) who underwent isolated orthotopic heart transplantation from January 1, 2000 to December 31, 2012. Transplantations were divided into a derivation cohort (n = 2,686) and a validation (n = 509) cohort. The validation cohort was randomly selected from 20% of transplantations from 2005 to 2012. Covariates found to be associated with rejection (p < 0.2) were included in the initial multivariable logistic regression model. The final model was derived by including only variables independently associated with rejection. A risk score was then developed using relative magnitudes of the covariates’ odds ratio. The score was then tested in the validation cohort.
Results A 9-point risk score using 3 variables (age, cardiac diagnosis, and panel reactive antibody) was developed. Mean score in the derivation and validation cohorts were 4.5 ± 2.6 and 4.8 ± 2.7, respectively. A higher score was associated with an increased rate of rejection (score = 0, 10.6% in the validation cohort vs. score = 9, 40%; p < 0.01). In weighted regression analysis, the model-predicted risk of rejection correlated closely with the actual rates of rejection in the validation cohort (R2 = 0.86; p < 0.01).
Conclusions The rejection score is accurate in determining the risk of early rejection in pediatric heart transplantation recipients. The score has the potential to be used in clinical practice to aid in determining the immunosuppressant regimen and the frequency of rejection surveillance in the first post-transplant year.
The first post-transplant year is the most likely time for rejection to occur in pediatric heart transplantation (1). Early rejection is a significant cause of morbidity and mortality in pediatric heart transplantation (1,2) as evidenced by its association with increased risk of late rejection, graft loss, and patient death (3–5). Despite a declining rate of early rejection due to the use of induction therapy, tacrolimus, and new immunosuppressive regimens, early rejection continues to occur in approximately 20% to 30% of transplantation recipients (1,6–8).
Multiple risk factors have been identified as increasing the risk of early rejection (2). However, a simple method to risk stratify pediatric heart transplantation recipients has not been developed. The object of this study was to design and validate a clinically derived rejection score that could be used to assess risk of early rejection in pediatric heart transplantation recipients.
A retrospective analysis was performed using data obtained from the United Network for Organ Sharing (UNOS) Standard Transplant Analysis and Research files. Heart transplantations performed in the United States from January 1, 2000, to December 31, 2012 were included for analysis. The study end date of December 31, 2012 was chosen to allow for 1 year of follow-up. The database was queried for pediatric heart transplantations in patients who were age 17 or younger. Transplantations were included if they had valid reporting of the presence of treated rejection within the first post-transplant year; biopsy confirmation of rejection was not required. Transplantations were excluded if they were performed in patients who were age 18 years or older, were not isolated heart transplantation, or had <1 year of follow-up reported. A validation group was randomly selected from 20% of the transplantations done between 2005 and 2012. The validation cohort was removed from the derivation cohort. The primary endpoint was early rejection, which was defined as rejection that required treatment in the first post-transplant year. The Medical University of South Carolina Institutional Review Board approved the study.
All pre-transplantation variables present in the UNOS database were assessed for association with early rejection using univariate statistics (chi-square for categorical variables and Student t test for continuous variables). Continuous variables were converted to categorical variables using previously published cutoffs for the purposes of the multivariable logistic model (1,9,10). For panel reactive antibodies (PRAs), the most recently reported PRA before transplantation was used. The variables associated with rejection in univariate analysis (p < 0.2) were entered into a multivariable logistic regression analysis. Variables with ≥20% of missing data were excluded from the multivariable model. Covariates were eliminated from the model if they were not independently associated with early rejection, which was defined as a p value of <0.05. The model’s goodness of fit was tested with the Hosmer-Lemeshow test and the receiver-operating area under the curve (c-statistic).
Remaining covariates associated with rejection were assigned points based on the relative weight of their odds ratio in the final multivariable logistic regression model to derive a formula for the “rejection score.” A score was then calculated for all transplantations in the derivation and validation groups. The association of the rejection score with early rejection was assessed using weighted regression analysis and logistic regression in both the derivation and validation cohorts. In weighted regression analysis, correlations between model-predicted rejection rates and actual rejection rates within each respective rejection score were assessed, and weights were given based on the number of patients in each respective rejection score. Rejection scores were subsequently classified as low, medium, and high, and the predicted and actual rates of rejection among these 3 groups were compared.
All statistics was performed using SPSS version 21 (IBM, Armonk, New York).
A total of 4,106 isolated heart transplantations were performed between 2000 and 2012 in pediatric patients. Of those, 3,195 (78%) had adequate reporting of early rejection, and therefore, were used for further analysis. Average recipient age was 6.9 ± 6.2 years. The average donor age was 9.5 ± 9.4 years. Recipient race was Caucasian in 1,822 (57%) transplantations, and donor race was Caucasian in 1,854 (58%) transplantations. An underlying cardiac diagnosis of cardiomyopathy was present in 1,652 (51%) transplantations, and congenital heart disease was present in 1,233 (39%) transplantations. Two hundred fourteen (7%) patients had re-transplantations, and 112 (3%) transplantation patients had other conditions.
One hundred twenty-seven patients (4%) were on extracorporeal membrane oxygenation (ECMO) at the time of transplantation; 412 (13%) had a ventricular assist device (VAD), 469 (15%) were on mechanical ventilation, and 1,537 (48%) were receiving inotropes. Donor–recipient human leukocyte antigen (HLA) matching was high (≥3 matches) in 447 (14%) transplantations, low (1 to 2 matches) in 2,031 (64%) transplantations, and occurred in none in 717 (22%) transplantations. PRAs were <10% in 2,322 (73%) transplantations, 336 (11%) transplantations had a PRA between 10% and 50%, and 537 (17%) transplantations had a PRA of >50%. Mean ischemic time was 3.5 ± 1.2 h. Overall, 1,127 (35%) patients were reported to have early rejection, with a mean follow-up of 5.7 ± 3.9 years.
Comparisons between the derivation and validation cohorts are shown in Table 1. The cohorts did not differ according to recipient age, PRA, or donor–recipient HLA matching, cross match, or underlying cardiac diagnosis. Both groups had similar rates of recipients transplanted from ECMO, mechanical ventilation, or VADs; however, the derivation cohort was more likely to be on inotropes at the time of transplantation. The derivation cohort was more likely to be male (55.7% vs. 50.9%; p = 0.04). The validation cohort was less likely to have early rejection (23.4% vs. 37.5%; p < 0.01).
All pre-transplantation variables were assessed for association with early rejection using univariate statistical methods. Of the 111 pre-transplantation variables investigated, 29 variables had a univariate association with early rejection (Table 2). The final multivariable logistic regression model only included covariates that were independently associated with early rejection. The reasons for covariate removal from the multivariable model are listed in Table 2.
The final multivariable logistic regression model consisted of 4 variables: recipient age, underlying cardiac diagnosis, PRA, and transplant year. The c-index of the model was 0.73 (Figure 1), and the Hosmer-Lemeshow test had a p value of 0.16, which indicated that the final model fit the data appropriately. To further test the model, each of the 4 variables used in the final model were used individually to create rejection scores and in combinations (i.e., age, then age + PRA). The performance of scores made from individual variables or combinations of variables was then assessed by c-statistic. No individual variable or combination had a c-statistic >0.62, which indicated that the final model with all 4 variables possessed the greatest association with rejection. The odds ratio of the 4 covariates is shown in Table 3. Whole number scores were assigned to each category based on the relative magnitude of the odds ratio for each individual category. The year of transplantation was independently associated with early rejection, which was consistent with previous reports (Figure 2) (2). However, due to the impracticality of using the transplantation year in a clinical score, it was not included in the final model.
The rejection score (range 0 to 9) was then calculated for all transplantations in the derivation cohort. The score for each variable was calculated by subtracting 1 from the odds ratio, then dividing the difference by 0.2, and then finally rounding to the nearest integer (e.g., PRA >50%, odds ratio: 1.41 − 1 = 0.41 ÷ 0.2 = 2.05, rounded to 2). The mean score was 4.5 ± 2.6. Using regression analysis, which was weighted for frequency of transplantations within an individual score, there was a close association between model-predicted rates of rejection and actual rates of rejection (R2 = 0.92; p < 0.01) (Figure 3). In logistic regression analysis, an increase in rejection score was associated with higher rates of rejection (odds ratio: 1.12; 95% confidence interval: 1.08 to 1.15; p < 0.01).
The mean rejection score for the validation cohort was 4.8 ± 2.7. An increasing rejection score was associated with increasing rates of rejection (score = 0, 10.6% early rejection vs. score = 9, 40%; p < 0.01). Predicted rates of rejection were calculated using the model developed in the derivation group. The predicted rates of rejection and the actual rates of rejection for each rejection score showed close correlation in the weighted regression analysis (R2 = 0.86; p < 0.01) (Figure 4). In logistic regression analysis of the validation cohort, an increasing score was associated with an increased chance of early rejection (odd ratio: 1.21; 95% confidence interval: 1.12 to 1.32; p < 0.01).
To further increase the clinical utility of the rejection score, disjoint categories of low, medium, and high risk were determined. A low-risk score was set as a predicted chance of rejection of ≤20% in the validation cohort; medium risk was 20% to 30%, and high risk was set as a predicted risk of >30%. Low rejection risk scores were 0 to 3, medium risk scores were 4 to 7, and high risk scores were 8 and 9. In the derivation cohort, 941 transplantations (35%) had a rejection score of 0 to 3; 1,445 (54%) had a risk score of 4 to 7, and 291 (11%) had a score of 8 or 9. In the validation cohort, 177 (35%) had a low-risk score, 269 (53%) had a medium-risk score, and 63 (12%) had a high-risk score. Actual rates of rejection and model-predicted rates of rejection were similar for each rejection category in both the derivation (R2 = 0.95; p < 0.01) and validation cohorts (R2 = 0.99; p < 0.01) (Figure 5).
Early allograft rejection in pediatric heart transplantation recipients is associated with significant mortality and morbidity (3,4). Although risk factors for early rejection have been investigated (2), before this study, there was no clinically useful risk stratification tool in pediatric heart transplantation. Using >3,000 pediatric heart transplantation in the modern era, a 9-point rejection score was derived and validated. Table 4 demonstrates how the rejection score can be calculated in 4 different clinical scenarios.
The components of the clinical score are consistent with previous studies that investigated risk factors for early rejection in children. Elevated PRA has been associated with increased mortality in the first post-transplant year, as well as decreased long-term graft survival in pediatric heart transplantation (9,11). The presence of congenital heart disease or re-transplantation is also associated with increased 1- and 5-year mortality (1,12). In a previous multi-institutional database review, increasing age was independently associated with early rejection (2). The risk score developed in this study provides the opportunity for clinicians to place appropriate weight to each individual risk factor that is routinely obtained in the clinical setting.
A previous study of adult heart transplantation recipients found that age, recipient race, sex, and degree of HLA matching were predictive of early rejection (13). Although age was associated with early rejection in our study, unlike the adult population, younger age was associated with less rejection in the pediatric population. Also, race and sex were not associated with rejection in the pediatric population. Although African-American race was associated with worse graft survival in the pediatric population, the 1-year survival for African Americans was similar to all other races (14). Although HLA matching was associated with improved long-term graft survival in pediatric heart transplantation recipients, it was not associated with early rejection as reported in adults (10,15). Therefore, the risk factors for early rejection in the adult heart transplantation population do not apply to the pediatric population.
Due to improvements in management of pediatric heart transplantation recipients, the incidence of early rejection in pediatric heart transplantation is decreasing in the modern era (Figure 4). The validation cohort consisted of transplantations from 2005 to 2012, although the derivation cohort consisted of transplantations from 2000 to 2012. This led to reduced early rejection in the validation cohort. Despite the disparate early rejection rates between the groups, the rejection score was able to successfully risk stratify transplantation recipients in the more contemporary validation cohort. Therefore, the rejection score will likely continue to be clinically useful as medical therapies improve and rejection rates decrease.
The developed risk score has potential uses in the clinical arena. For example, screening for rejection has important morbidity and financial costs (16). A recent study showed that pediatric heart transplantation recipients are exposed to a significant amount of radiation in the first post-transplant year, and the majority of this exposure occurs during scheduled catheterizations (17). The ability to estimate risk of early rejection using this evidence-based rejection score may allow clinicians providing care to pediatric heart transplantation recipients to tailor rejection surveillance based on their risk score. Further studies will elucidate the role of this rejection score in decreasing the need to perform invasive screening procedures in low-risk populations.
A second possible use of the score is to identify patients at high risk for early rejection and adjust immunotherapy for these patients. Induction therapy and maintenance immunotherapy with tacrolimus and mycophenolate mofetil has been associated with decreased early rejection (6,18,19). However, these therapies are associated with an increased risk of developing anemia and neutropenia, which requires treatment (18). Use of the rejection score has the potential to assist pediatric heart transplantation physicians and care providers in the assessment of the risk/benefit ratio of each medication regimen to individualize immunosuppressive therapy.
There were many variables that were associated with early rejection in the univariate analysis that had too much missing data to be included in the multivariable analysis. It is possible that these variables might be independently associated with rejection. The type of rejection (cellular, antibody-mediated, mixed) could not be determined from the database. It is possible that individual covariates were strongly associated with 1 type of rejection versus another type (i.e., high PRA and antibody-mediated rejection). However, this could not be determined. Continuous variables (age, PRA) were changed to categorical variables to improve clinical feasibility of the rejection score. However, it was possible that when changing continuous variables to categorical variables, the statistical power of the logistic regression model was decreased. The definition of rejection for the study was treated rejection; biopsy confirmation was not always performed. Therefore, some patients who were treated might not have had biopsy-proven rejection.
A 9-point early rejection risk score in pediatric heart transplantation recipients was successfully derived and validated. The risk score accurately distinguished patients at high, medium, and low risk for early rejection using easily obtainable clinical factors. The rejection score could be used to help tailor rejection surveillance in the first year after pediatric heart transplantation.
COMPETENCY IN MEDICAL KNOWLEDGE: The risk for rejection in the first transplant year can be estimated using recipient age, PRA, and recipient cardiac diagnosis in pediatric heart transplantation recipients.
TRANSLATIONAL OUTLOOK: Further studies of early rejection in pediatric heart transplantation recipients should be performed to see if medical interventions could reduce rejection in the high-risk group.
This work was supported in part by Health Resources and Services Administration contract 234-2005-370011C. 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. Data represented is based on OPTN data as of December 31, 2013. Dr. Heal was supported by National Institutes of Health/National Heart, Lung, and Blood Institute Grant 5T32-HL07710-19.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- extra-corporeal membrane oxygenation
- human leukocyte antigen
- panel reactive antibody
- United Network for Organ Sharing
- ventricular assist device
- Received January 22, 2015.
- Revision received April 7, 2015.
- Accepted April 18, 2015.
- American College of Cardiology Foundation
- Swenson J.M.,
- Fricker F.J.,
- Armitage J.M.
- Butts R.J.,
- Scheurer M.A.,
- Atz A.M.,
- et al.
- Kilic A.,
- Weiss E.S.,
- Allen J.G.,
- et al.
- Pophal S.G.,
- Sigfusson G.,
- Booth K.L.,
- et al.