Validation of Donor Fraction Cell-Free DNA with Biopsy Proven Cardiac Allograft Rejection in Children and Adults

Abstract

Background

Donor specific cell-free DNA (cfDNA) shows promise as a non-invasive marker for allograft rejection, but as yet has not been validated in both adult and pediatric recipients

Objectives

To validate donor fraction (DF) cell-free DNA (cfDNA) as a non-invasive test to assess for risk of acute cellular rejection (ACR) and antibody mediated rejection (AMR) following heart transplantation in pediatric and adult recipients

Methods

Pediatric and adult heart transplant recipients were enrolled from seven participating sites and followed for ≥12 months with plasma samples collected immediately prior to all endomyocardial biopsies. DF cfDNA was extracted and quantitative genotyping performed. Blinded DF cfDNA and clinical data were analyzed and compared to a previously determined threshold of 0.14%. Sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and ROC curves were calculated.

Results

A total 987 samples from 144 subjects were collected. After applying pre-defined clinical and technical exclusions, 745 samples from 130 subjects produced 54 rejection samples associated with the composite outcome of ACR ≥grade 2R and/or pAMR ≥2 and 323 healthy samples. For all participants, DF cfDNA at a threshold of 0.14% had a sensitivity of 67%, a specificity of 79%, a PPV of 34% and a NPV of 94% with an area under the curve (AUC) of 0.78 for detecting rejection. When analyzed independently, these results held true for both pediatric and adult cohorts at the same threshold of 0.14% (NPV 92% and 95% respectively).

Conclusions

DF cfDNA at a threshold of 0.14% can be used to assess for risk of rejection following heart transplantation in both pediatrics and adults with excellent NPV.

Graphical Abstract

Introduction:

Despite historically high survival rates for children and adults undergoing heart transplant, allograft rejection remains a significant source of morbidity and mortality (1,2). While endomyocardial biopsy remains the gold-standard for the diagnosis of cardiac allograft rejection, biopsy is an invasive procedure and not without risk (3). Further, while there are ISHLT pathology grading systems, pathologic evaluation still relies upon subjective components (4) and suffers from sampling bias. The Rejection may heterogeneously affect the myocardium, potentially leading to the pathological diagnosis from biopsy not aligning with overall allograft health. Because of these concerns, a quantitative, systemic, non-invasive diagnostic test continues to be of great interest to the field.

The use of circulating cell-free DNA (cfDNA), and more specifically, donor-derived cfDNA holds promise to detect allograft rejection and is not subject to many of the limitations of biopsy (5–9). However, the use of donor-derived cfDNA has not been widely validated in the cardiac transplant literature. The DTRT-1 study (10) showed that donor fraction (DF) cfDNA, the ratio of donor-derived cfDNA to total cfDNA, was highly associated with acute allograft rejection in pediatric and adult patients and determined optimal cut-off values. However, that study showed limitations associated with the use of whole blood collection and shipping methods and the potential confounder of cell lysis. With the goal of validating a clinically viable assay to assess DF cfDNA, we prospectively enrolled both adult and pediatric patients in a validation cohort (DTRT-2) using the refined methodology of DF cfDNA assessment to further describe the association of DF cfDNA with biopsy proven acute cellular (ACR) and antibody-mediated rejection (AMR).

Methods:

Study Design:

A prospective observational study (DNA-Based Transplant Rejection Test Extension Protocol [DTRT-2]) was conducted among seven heart transplant centers, each obtaining local Institutional Review Board approval (CW DTRT-2 846823; ClinicalTrials.gov ID NCT02109575, initial approval 2/3/2016).

In a prior prospective observational study, DTRT-1, the test assay was developed and refined (10,11). The current study, DTRT-2, followed as a prospective observational study focusing on validation of the refined assay. DTRT-2 enrollment was conducted between August 2016 and October 2017 for a total of 147 participants at seven centers. All participants signed written informed consent/assent as appropriate for participation in the study and publication of their data. The inclusion and exclusion criteria for the cohort are as described previously (10). Briefly, patients who underwent biopsy >8 days post-transplant and had a test sample collected prior to biopsy were included in this study cohort. Samples were included only if >8 days post-transplant as prior data show a decline to baseline levels of DF cfDNA by that time (12). Clinical conditions that may invalidate the calculation of cfDNA resulted in exclusion of samples and included pregnancy, multi-organ transplant recipient (including bone marrow), history or active post-transplant lymphoproliferative disorder, history of cancer in the prior two years, and mechanical circulatory support at the time of sample collection. Lab based genotype quality control (QC) measures were then applied (11), and samples that did not meet QC thresholds were excluded.

The study was strictly blinded and the blind was maintained by an independent clinical research organization (CRO) using a 21CFR11 compliant clinical database. Structured patient data were collected prospectively and monitored by the CRO including demographic, clinical pre-transplant, transplant and post-transplant information, and clinical and laboratory data such as rejection, infection, hospitalization and death. Structured diagnostic data including catheterization and biopsy were also collected. The local center read and interpretation of the biopsy was used for this analysis. The CRO monitored the data fields required for analysis of the primary end-point of allograft rejection used in this analysis at 100%. All data were deidentified and coded. The study utilized a secure and validated electronic data capture system for management of all clinical data. The system and procedures for electronic database set-up, entry, review, access, security, and monitoring were designed in specific compliance with 21CFR 11.

Blood sample collection:

Blood samples included in this analysis were all drawn within 24 hours prior to cardiac catheterization/endomyocardial biopsy as previously described (10). Samples were collected in Streck tubes (Streck BCT tubes, Streck, Omaha, NE) at a minimum whole blood draw volume of 3 ml in patients weighing less than 20 kg and 5 ml in patients ≥20 kg. At each site, samples were de-identified except for a study assigned sample code. The collecting site isolated plasma locally using a validated clinical protocol with two slow speed centrifugation steps, freezing and storing the sample until it was shipped overnight air delivery on dry ice (11).

Extraction:

Prior to cfDNA extraction, all plasma samples were thawed and spun with a high-speed centrifuge at 15,000 x g spin for 10 minutes. All plasma was extracted on a Tecan Evo150 liquid handling platform (Tecan, Switzerland) with ReliaPrep HT Circulating Nucleic Acid Kit, Custom (Promega, Madison, WI) and resuspended in 0.1X TE.

cfDNA analysis:

cfDNA concentration of plasma was measured by quantitative real-time polymerase chain reaction (PCR) targeting the nuclear gene RNase P (Applied Biosystems, Foster City, CA) as previously described (10,11,13). PCR analysis was carried out on a Lightcycler 480 (Roche Applied Science, Penzberg, Germany). A dilution series of commercially-sourced human genomic DNA was used to create a standard curve for quantification (Promega, Madison, WI). Mean cfDNA concentration in ng/ml plasma for replicates of each sample were determined. To assure pre-analytical plasma quality, a cfDNA fragmentation test to detect leukocyte lysis by comparing short (115 bp) and long (247 bp) multicopy Alu sequences was performed (TAI Diagnostics, Wauwatosa, WI) (11).

DF analysis:

Donor-specific cfDNA was calculated as a fraction (DF) of the total cfDNA and performed without the requirement of a donor sample using the myTAI-HEARTv3 assay (TAI Diagnostics, Wauwatosa, WI) (11). The myTAI-HEARTv3 assay incorporates reagent quality improvements and other standardizations that produce a lower level of quantification (LOQ) compared to the previous version as described in North et al 2020 (and unpublished). Based on the results of the DTRT-1 study (10), and further work from our group, a predefined cut-off value for DF cfDNA of 0.14% was established for the purpose of assay validation as a potential clinically relevant threshold value for the detection of allograft rejection.

Clinical Endpoint:

The endpoint of allograft rejection was defined as the presence of cellular or antibody mediated rejection on endomyocardial biopsy as reported using the ISHLT cellular and antibody-mediated rejection grading system by the site clinical pathologist (14,15). For the purpose of the study, biopsies with ISHLT cellular rejection (ACR) Grade 2R or higher and/or a pathologic antibody-mediated rejection (pAMR) grade 2 or higher were considered as having the endpoint of acute allograft rejection. Biopsy results of ISHLT Grade 0 with pAMR grade 0, or with no reported pAMR grading due to no clinical suspicion of AMR, were considered as healthy allograft controls. Further clinical exclusions of evidence of potential poor graft health were also applied to samples to ensure that control samples were associated with truly healthy allografts. Samples within 30 days of cardiac arrest or death, with known coronary allograft vasculopathy, within 14 days prior or 21 days after treatment of infection, within one week of mechanical circulatory support or within 30 days of biopsy-proven or clinically treated rejection were all excluded from being considered as healthy allograft samples.

Statistical Analysis:

Continuous variables are summarized as median and interquartile range (IQR) unless otherwise specified, categorical variables as number (%). Generalized linear models for repeated measures were used with maximum likelihood estimation. Subjects were viewed as clusters. We examined which variance covariance matrix provided best fit and the results for compound symmetry are reported. A logit link function was used with log2 donor fraction as the outcome. Receiver Operating Curves (ROC) were used to elect a threshold for the predictive values. All data analyses were performed using SAS9.4, R×64 3.6.0 and SPSS26.0.

Results:

Study Cohort:

Across seven collaborating centers, a total of 987 biopsy samples with available total cfDNA from 144 participants (69 pediatric, 75 adult) were analyzed. Of these, 71 samples were excluded due to clinical exclusions, 5 samples were excluded for not processing to plasma within 5 days, and 144 samples could not be analyzed due to insufficient total cfDNA present in the sample. An additional 8 samples were removed from analysis for failing to meet the pre-analytical plasma quality criteria of an ALU ratio <0.5. Finally, 14 samples were excluded due to inadequate tissue for diagnostic purposes per the pathology report. After all clinical and sample associated exclusions were applied, the final cohort for analysis consisted of 745 samples from 130 patients, with 54 samples associated with ACR grade 2 and/or AMR grade 2 or higher rejection on biopsy (Figure 1).

Figure 1.

CONSORT Flow Diagram illustrating exclusion of samples according to predefined clinical and samples-based criteria, resulting in final cohort of 745 biopsy associated DF cfDNA samples from 130 participants. DF cfDNA=Donor fraction cell-free DNA

Median age at transplant was 24.3 years (0.1 to 68.9) and 46.4% of the cohort were pediatric (<18 years of age at time of transplant). Patients enrolled at the time of transplant comprised 60% of the cohort with an additional 30% enrolled within 30 days of transplant and 10% enrolled post-transplant. 68.8% were male, 20% were Black race/African American descent, and 8% identified as Hispanic/Latin ethnicity (Table 1).

Table 1.

Demographics of the study population

n=144
Age at transplant in years, Median (Range)	19.1 (0.1, 68.9)
Adult ≥18 years, n (%)	75 (52.1)
Male, n (%)	96 (66.7)
Ethnic, n (%)
Hispanic/Latino	11 (7.6)
Non-Hispanic/Latino	119 (82.6)
Unknown	14 (9.7)
Race, n (%)
Black/AA	34 (23.6)
White	95 (66.0)
Asian	1 (0.7)
More than one race	2 (1.4)
Other	2 (1.4)
Unknown	10 (6.9)
Time from transplant at enrollment, n (%)
Prior to transplant	85 (59.0)
Within 30 days post-transplant	45 (31.3)
>30 days post-transplant	14 (9.7)

Association of Donor Fraction cfDNA and Allograft Rejection:

As shown in Figure 2a, median DF cfDNA was significantly higher in samples associated with allograft rejection (ACR and/or AMR) as compared to healthy allograft controls (0.21% vs 0.09%, p<0.0001). ROC analysis (Figure 3a) yielded an AUC of 0.78 and at the predefined threshold of 0.14%, the test sensitivity was 0.67 with a specificity of 0.79, resulting in a Negative Predictive Value (NPV) of 0.94 and Positive Predictive Value (PPV) of 0.34 in this validation cohort. This relationship held when looking at the adult and pediatric cohorts separately with adult participant samples associated with allograft rejection having a median DF cfDNA of 0.17% vs. 0.07% in healthy controls (p<0.001, Figure 2b) and pediatric participant samples associated with allograft rejection having a median DF cfDNA of 0.26% vs. 0.11% in healthy controls, (p=0.02, Figure 2c). ROC analysis of these age-divided cohorts resulted in an AUC for the adult cohort of 0.81 (Figure 3b) and an AUC of 0.79 for the pediatric cohort (Figure 3c). Optimal threshold differed slightly between the two cohorts with an optimal threshold identified at 0.10% for adults and 0.13% for children. However, using the predefined threshold of 0.14% did not drastically alter test performance characteristics and yielded a sensitivity of 0.61 with specificity of 0.88 (NPV=0.92, PPV=0.49) in adults and a sensitivity of 0.76 with specificity of 0.65 (NPV=0.95, PPV=0.25) in the pediatric cohort.

Figure 2.

Box-plot showing the relationship between log DF cfDNA and acute allograft rejection (as defined by ACR ≥2R and/or pAMR ≥2) in all participants (a), adults (b) aad children (c). Median DF cfDNA was significantly higher in samples associated with acute allograft rejection as compared to healthy allografts. The upper and lower borders of the box represent the upper and lower quartiles, respectively. The horizontal line represents the median value. The upper and lower whiskers represent the maximum and minimum values of the nonoutliers, respectively. DF cfDNA=Donor Fraction Cell-Free DNA, ACR=Acute Cellular Rejection, pAMR=Antibody Mediated Rejection

Figure 3.

Receiver operating characteristic curves for detection of acute allograft rejection (as defined by ACR ≥2R and/or pAMR ≥2) using a DF cfDNA threshold of 0.14% in all participants (a), adults (b), and children (c). DF cfDNA=Donor Fraction Cell-Free DNA, ACR=Acute Cellular Rejection, pAMR=Antibody Mediated Rejection, AUC=area under the curve, PPV=positive predictive value, NPV=negative predictive value

Association of Donor Fraction cfDNA and Cellular Rejection

Samples positive for AMR were excluded to reveal the relationship between DF cfDNA and cellular rejection only. Overall test characteristics are seen in Figure 4 and supplemental Figure 1. Median DF cfDNA was higher in samples associated with ACR as compared to those of healthy allografts (0.16% vs. 0.09%, p<0.0001). ROC analysis with an AUC of 0.71 showed the optimal threshold in this validation cohort was identical to the predefined threshold of 0.14% with a sensitivity of 0.57, specificity of 0.79 and NPV of 0.96 and PPV of 0.19 for detection of ACR. The entire validation cohort was then divided into adult and pediatric age groups. ACR resulted in higher median DF cfDNA as compared to healthy controls in both adults (0.15% vs 0.07%, p=0.009, Figure 4b) and children (1.94% vs. 0.11%, p=0.002, Figure 4c). ROC analysis yielded differing optimal thresholds in adults and children of 0.098% and 0.86% respectively (Supplemental Figure 1b and 1c). However, at the predefined threshold of 0.14%, test characteristics were similarly good with sensitivity of 0.54, specificity of 0.88, NPV of 0.94 and PPV of 0.38 in the adult cohort and sensitivity of 0.75, specificity of 0.65, NPV of 0.99 and PPV of 0.06 in the pediatric cohort.

Figure 4.

Box-plot showing relationship between log DF cfDNA and acute cellular rejection (ACR ≥2R) in all participants (a), adults (b), and children (c). Median DF cfDNA was significantly higher in samples associated with acute cellular rejection as compared to healthy allografts. The upper and lower borders of the box represent the upper and lower quartiles, respectively. The horizontal line represents the median value. The upper and lower whiskers represent the maximum and minimum values of the nonoutliers, respectively DF cfDNA=Donor Fraction Cell-Free DNA, ACR=Acute Cellular Rejection

Association of Donor Fraction cfDNA and Antibody Mediated Rejection

To assess performance in detecting antibody mediated rejection (AMR), samples with evidence of ACR on biopsy were excluded and analysis was performed on the remaining samples. Similar to ACR, samples associated with AMR had higher median DF cfDNA as seen in Figure 5a (0.26% vs 0.09%, p<0.0001). ROC analysis (supplemental Figure 2a) with an AUC of 0.84 showed the optimal threshold was 0.17%, however at the predefined threshold of 0.14%, there was sensitivity of 0.77, specificity of 0.78 and NPV of 0.97 and PPV of 0.25 for detection of AMR. The entire validation cohort was again divided into adult and pediatric age groups. AMR resulted in higher median DF cfDNA as compared to healthy controls in both adults (0.15% vs 0.06%, p<0.001, Figure 5b) and children (0.24% vs. 0.11%, p=0.023, Figure 5c). ROC analysis again yielded differing optimal thresholds for the detection of AMR in adults and children of 0.17% and 0.197% respectively (Supplemental Figure 2b and 2c). However, at the predefined threshold of 0.14%, test characteristics continued to be excellent with sensitivity of 0.77, specificity of 0.87, NPV of 0.98 and PPV of 0.29 in the adult cohort and sensitivity of 0.78, specificity of 0.65, NPV of 0.96 and PPV of 0.22 in the pediatric cohort.

Figure 5.

Box-plot showing relationship between log DF cfDNA and Antibody Mediated Rejection (pAMR ≥2) in all participants (a), adults (b), and children (c). Median DF cfDNA was significantly higher in samples associated with biopsy proven antibody mediated rejection as compared to healthy allografts. The upper and lower borders of the box represent the upper and lower quartiles, respectively. The horizontal line represents the median value. The upper and lower whiskers represent the maximum and minimum values of the nonoutliers, respectively DF cfDNA=Donor Fraction Cell-Free DNA, pAMR=Antibody Mediated Rejection

Review of Incongruent Results

When assessing agreement between DF cfDNA and biopsy results, there were 18 samples in which DF cf-DNA was below the set threshold of 0.14%, yet the results from biopsy were positive for rejection, and are detailed in Supplemental Table 1. These samples were from 15 patients, 5 pediatric and 10 adults. Thirteen of 18 were within 6 months of transplant. Cellular rejection at Grade 2R was present in 11 samples with pAMR grade 2 being present in 7. Only one sample had higher grade rejection of 3R. Interestingly only one sample was associated with a biopsy that correlated with treatment of rejection by the clinical team.

Discussion:

In this prospective study of a clinically viable assay using cfDNA in heart transplant recipients, a higher donor fraction of cfDNA was significantly associated with acute allograft rejection. Specifically, this assay performed admirably at a DF-cfDNA threshold level of 0.14%. In this study cohort, below 0.14%, the assay had a negative predictive value of 94%. Furthermore, at this same threshold, the assay had a positive predictive value of 34% whereas a full third of patients with DF cf-DNA >0.14% had evidence of Grade 2 or higher allograft rejection. In this first study to validate the use of DF cfDNA in pediatric heart recipients, the same threshold could be used with nearly identical performance characteristics in both children and adults. Additionally, this assay was able to show association with both cellular and antibody mediated rejection. Overall, the assay was able to predict lack of rejection, or a “healthy allograft” with a single DF cf-DNA threshold regardless of age.

Endomyocardial biopsy remains the gold-standard for the diagnosis of allograft rejection. While biopsy has its limitations, pathological examination allows for grading of the degree of rejection and the type: cellular or humoral (3,14,15). As yet, there are no noninvasive tests that allow for this level of granularity in the diagnosis of allograft rejection. However, endomyocardial biopsy is an imperfect standard, and is very much vulnerable to variation in reporting; resulting in questionable reproducibility (4), and is also limited by sampling bias. Any individual biopsy is only representative of that specific area of myocardium, which may or may not be representative of the overall health of the entire allograft. Any given biopsy may under- or overestimate the degree of rejection found throughout the heart. The assessment of donor-derived cfDNA, however, should be more representative of the entire transplanted organ. A high level of donor-derived cfDNA is associated with damage to the allograft and should not be subject to the same sampling bias of biopsy. By indexing the amount of donor-derived cfDNA to the total amount of circulating cfDNA in recipient plasma (i.e. DF cfDNA), the assay becomes more specific for allograft injury, as systemic disease will cause both recipient and donor cfDNA levels to rise in conjunction (16). Despite these benefits, the likely utility of DF cfDNA in heart transplantation is not to replace biopsy completely. Instead, given the excellent negative predictive value of the assay, DF cfDNA is most promising as a screening tool, perhaps entirely replacing the practice of routine surveillance biopsies. Healthy patients with values below the threshold are very unlikely to have significant allograft rejection and would not need to undergo invasive testing. In a superficial review of those samples with value below the threshold but with biopsy results suggestive of rejection, only one sample was associated with an episode of rejection requiring treatment as decided by the clinical team. These findings are interesting as the clinical teams did not have knowledge of the cfDNA results and still declined to treat rejection as graded on biopsy. Additional study of these incongruent results is certainly warranted, as there can be many factors at play responsible for these discrepancies.

Those with elevated levels, would still require biopsy to both classify the type of rejection as cell-mediated or humoral, and to grade the rejection according to current ISHLT standards. To be used in this way, any assay needs to not only be very reliable, but also be able to be collected and run in a standard fashion consistent with other clinical tests. The assay described in this study meets these criteria. Collection and shipping as spun plasma along with stringent QC methodology (10,11) has resulted in excellent test characteristics and an ability to provide a result in 48 hours, similar to the turnaround time for endomyocardial biopsy.

Although the number of pediatric heart transplants each year is overshadowed by adult numbers, these children typically have many more post-transplant years and overall better survival outcomes (2). This results in the potential need for >20 years of endomyocardial biopsy surveillance. Although routine biopsy is considered safe in children, complications do arise (3), and the utility of non-invasive surveillance is arguably more important in children whom often require general anesthesia for the procedure. While our group has previously shown the association between cfDNA and rejection in children (10,17), this study is the first to validate a cfDNA assay in pediatric patients alongside adults, and was able to show excellent performance characteristics in both age groups. The use of cfDNA as a screening tool, and thereby the avoidance of catheterization, in children cannot be understated. Additionally, showing that a single DF cfDNA threshold value can be valid in children and adults, has the additional benefit of one single assay that can be used across the lifespan. Transition of pediatric patients to adult programs can be fraught with problems, and has been associated with increased risk of adverse events (18–20). A single noninvasive assay used by both pediatric and adult programs alike, can eliminate one of the many variations in care that may occur at the time of transition.

When discussing a cfDNA assay as a screening test for acute allograft rejection, in addition to the test performance characteristics, there are certain real-world limitations that must be addressed. For example, this assay was able to be successfully performed with only 3ml of blood collection in patients smaller than 20kg, as volume of blood draw is often a limiting factor in pediatric patients. Despite this refinement in required sample volume, the study still needed to exclude approximately 15% of samples for inadequate volume. As the study blind remained throughout the entire study duration, sample collection practices were not further adjusted during the study period. In post-hoc review however, we determined that 106 of 144 excluded samples were drawn from pediatric patients, with most having less than the required 3ml minimum volume. Certainly, in future studies and clinical use, sample collection volumes will need to be scrutinized, an easily surmountable limitation. More reassuring, is that after excluding samples that had inadequate volume of total cfDNA, only 8 of 767, or about 1% of samples failed the rigorous QC process involved in cfDNA quantification. In comparison, in our cohort, biopsy had a 1.8% fail rate with a read of ‘tissue inadequate to grade.’ With similar fail rates, cfDNA as a replacement for routine biopsy is promising. However, in this study, cfDNA did not have perfect correlation with the local biopsy read. Given the subjectivity of pathologic results, a more comprehensive assessment of the biopsy pathology is desirable, with biopsy results read and scored by a core lab. While such reads would be more standardized, the primary aim of the current analysis was to correlate cfDNA results to the local read in a closer approximation of clinical use.

Beyond predicting the presence or absence of any allograft rejection, this study was able to show independent associations with both ACR and AMR at pathologic grades of 2 or higher. Additional analyses showed a positive association with higher DF cfDNA values with higher pathologic grading, even with lower levels of rejection (1R/pAMR 1). While this suggests that a continuum of DF cfDNA values could be used to infer the severity of allograft rejection, further analyses are needed to determine if different thresholds could be applied to predict different types or severity of rejection. In the current study, there were not adequate numbers of each sub-type of rejection to do such analyses, but exploration of this utility of cfDNA is certainly a future goal.

In this validation study of a clinically viable cfDNA assay in pediatric and adult heart transplant recipients, a DF cfDNA threshold of 0.14% resulted in excellent performance characteristics, whereby below this threshold subjects were highly unlikely to have allograft rejection (Figure 6, Video). This study showed the potential for “real-world” applicability by utilizing standard clinically available plasma collection and shipping techniques, and comparing assay results with the patient’s own local biopsy read. While this study shows the promise of cfDNA in the clinical care of heart transplant recipients, there is still much to learn about how other factors besides acute rejection can influence these results. Additional analyses of these data are needed to elucidate how conditions like graft vasculopathy and other causes of graft dysfunction may or may not affect these results. Further clinical trials comparing patient outcomes when using a DF cfDNA surveillance strategy as compared to standard of care will help define some of the benefits and limitations of this approach.

Figure 6.

Summary of the validation of donor fraction cell-free DNA to detect acute allograft rejection. The upper and lower borders of the box represent the upper and lower quartiles, respectively. The horizontal line represents the median value. The upper and lower whiskers represent the maximum and minimum values of the nonoutliers.

Supplementary Material

MMC1

Video. Recording of the presentation of these data at the 101^st Annual Meeting of the AATS Late Breaking Abstract Session: Bench to Bedside: Translational Discoveries on April 30, 2021

Fig1S

Supplemental Figure 1. Receiver operating characteristic curve analysis for detection of acute cellular rejection (ACR ≥2R) using a DF cfDNA threshold of 0.14% in all participants (a), adults (b), and children (c). DF cfDNA=Donor Fraction Cell-Free DNA, ACR=Acute Cellular Rejection, AUC=area under the curve, PPV=positive predictive value, NPV=negative predictive value

Fig2S

Supplemental Figure2. Receiver operating characteristic curve analysis for detection of antibody mediated rejection (pAMR ≥2) using a DF cfDNA threshold of 0.14% in all participants (a), adults (b), and children (c). DF cfDNA=Donor Fraction Cell-Free DNA, pAMR=Antibody Mediated Rejection, AUC=area under the curve, PPV=positive predictive value, NPV=negative predictive value

Central Message:

In both adult and pediatric heart transplant recipients, donor fraction cell-free DNA had a negative-predictive value of 94% for the presence of biopsy-associated acute allograft rejection.

Perspective Statement:

In this validation study of a cfDNA assay in pediatric and adult heart transplant recipients, a donor fraction cfDNA threshold of 0.14% resulted in excellent performance characteristics, whereby below this threshold subjects were highly unlikely to have rejection. This shows the potential for “real-world” applicability by utilizing standard clinically available plasma collection and shipping techniques

Central Figure:

Median DF cfDNA was significantly higher in allograft rejection as compared to controls.

Abbreviations:

cfDNA

cell-free DNA

DF

donor fraction

ACR

Acute Cellular Rejection

AMR

Antibody Mediated Rejection

DTRT

DNA-based Transplant Rejection Test

CRO

Clinical Research Organization

ROC

Receiver Operator Characteristic

PPV

Positive Predictive Value

NPV

Negative Predictive Value