Donor-Derived Cell-Free DNA as a Prognostic Biomarker in Severe Cardiac Allograft Rejection: From Detection to Risk Stratification

Highlights

Percent donor-derived cell-free DNA (%dd-cfDNA) has evolved from a rejection detection tool into a prognostic biomarker that can stratify risk of death and prolonged left ventricular dysfunction after heart transplantation.
The 2026 GRAFT study (Genomic Research Alliance for Transplantation; NCT02423070) shows that AR with %dd-cfDNA ≥0.15% at diagnosis identifies a high-risk subgroup, whereas AR with lower %dd-cfDNA may not confer the same adverse prognosis.
Earlier studies established strong diagnostic performance for dd-cfDNA in acute cellular rejection and antibody-mediated rejection, but later work clarified key confounders, test thresholds, and biologic correlates of graft injury.
The emerging literature supports a transition from surveillance based on whether rejection is present to surveillance that also asks how biologically severe the episode is and whether the graft is likely to recover.

Background

Acute rejection remains one of the most consequential complications after heart transplantation, contributing to graft dysfunction, heart failure, hospitalization, retransplantation, and death. Although endomyocardial biopsy (EMB) remains the reference standard for diagnosing acute cellular rejection (ACR) and antibody-mediated rejection (AMR), biopsy has important limitations: invasiveness, sampling error, interobserver variability, and imperfect alignment between histology and clinical injury. These limitations have driven intensive development of noninvasive surveillance biomarkers, particularly donor-derived cell-free DNA (dd-cfDNA).

dd-cfDNA is released into the recipient circulation when donor cardiomyocytes are injured and undergo cell death. Because the signal is graft-derived, dd-cfDNA is conceptually different from immune cell activation markers or donor-specific antibody assays: it reflects ongoing tissue injury regardless of whether the injury is triggered by T-cell-mediated, antibody-mediated, ischemic, infectious, or multifactorial processes. In heart transplantation, this biological property has made dd-cfDNA attractive not only as a rejection detector, but also as a quantitative marker of graft injury severity.

The new Circulation: Heart Failure report by Safiullah and colleagues extends this field in an important direction. In the prospective multicenter GRAFT study, %dd-cfDNA measured at the time of AR diagnosis stratified long-term risk of prolonged LV dysfunction and death. This finding is clinically significant because existing risk models have been limited, especially for patients with severe rejection phenotypes that do not map neatly onto biopsy categories alone. The study suggests that a positive rejection diagnosis is not homogeneous: the biomarker burden at diagnosis helps identify which episodes are most dangerous.

Key Content

1) Chronological development of evidence: from rejection detection to prognostic enrichment

The evidence base for dd-cfDNA in heart transplantation developed in stages. The foundational multicenter prospective validation study published in Circulation in 2021 (PMID: 33435695; NCT02423070) showed that %dd-cfDNA rose with AR versus controls, with excellent discrimination for rejection (AUC 0.92) and very high negative predictive value at a 0.25% threshold. Importantly, dd-cfDNA rose before histopathologic diagnosis, suggesting that blood-based surveillance may detect injury earlier than scheduled biopsy. AMR showed stronger dd-cfDNA elevations than ACR, indicating that assay magnitude may encode phenotype as well as binary rejection status.

Subsequent studies in 2022–2024 refined interpretation. Investigators reported that elevated dd-cfDNA can occur without biopsy-proven rejection, especially in settings such as infection, de novo donor-specific antibodies, right ventricular dysfunction, and pericardial effusion (PMID: 34863042; 33653997). These observations broadened the biological meaning of dd-cfDNA from “rejection marker” to “graft injury marker” and emphasized the need to interpret results in clinical context.

By 2023–2025, studies compared dd-cfDNA with other surveillance strategies such as gene-expression profiling (GEP) and peripheral gene-expression profiling (pGEP), generally finding that dd-cfDNA adds specificity and may improve biopsy stewardship. The paired dd-cfDNA/GEP strategies reduced biopsies without obvious compromise in short-term outcomes in retrospective cohorts (PMID: 35559582; 37036133; 41085484). In parallel, mechanistic studies showed that dd-cfDNA correlates not only with rejection-associated transcripts but also with injury and macrophage-related programs in EMB tissue (PMID: 38538559), supporting a model in which dd-cfDNA reflects both alloimmune activity and downstream parenchymal damage.

The 2024–2025 literature also explored higher-order clinical questions: Can dd-cfDNA identify disease beyond classic biopsy thresholds? Can it detect cardiac allograft vasculopathy with reduced myocardial blood flow (PMID: 39655433)? Can it help redefine AMR by incorporating DSA plus graft dysfunction (PMID: 39584219)? These studies collectively moved the field beyond diagnosis toward phenotyping, prognostication, and longitudinal risk assessment.

Against that backdrop, Safiullah et al. (PMID: 42290367) provide the most direct evidence yet that %dd-cfDNA at AR diagnosis has prognostic value for hard outcomes. This is a crucial step in biomarker maturation: from detecting who has rejection, to identifying who is likely to fail clinically.

2) The GRAFT study: prognostic stratification at the time of rejection

The prospective multicenter GRAFT study enrolled 277 heart transplant recipients and generated 3218 dd-cfDNA measurements over a median follow-up of 4.9 years. AR was broadly defined to include ACR, pathological AMR, and biopsy-negative AMR characterized by donor-specific antibody positivity with LV dysfunction. This definition is clinically relevant because it captures a spectrum of severe alloimmune injury that may be missed by pathology alone.

During follow-up, 75 patients (27%) developed AR, including 43 with ACR, 18 with pathological AMR, and 14 with DSA-positive/LV dysfunction episodes. Fifty-three patients experienced the composite outcome of death or prolonged LV dysfunction (defined as LVEF ≤50% for ≥90 days). In time-dependent Cox models, AR was associated with a markedly increased risk of the composite outcome (HR 4.47, 95% CI 2.42–8.26; P<0.001).

The key advance was biomarker stratification at the time of diagnosis. Using a data-driven threshold of 0.15%, patients with AR and %dd-cfDNA ≥0.15% had substantially higher risks of prolonged LV dysfunction, death, and the composite endpoint compared with patients who had not developed AR at the same follow-up time; the composite hazard ratio was 6.28 (95% CI 3.04–13.0; P<0.001). In contrast, patients with AR but %dd-cfDNA <0.15% did not show statistically significant increases in death or prolonged LV dysfunction. This suggests that the biomarker is not merely a diagnostic adjunct but a severity marker identifying biologically more aggressive rejection.

3) How this study integrates with prior evidence

The GRAFT findings are most meaningful when interpreted alongside the 2021 foundational validation study. Earlier work demonstrated that dd-cfDNA is highly sensitive for rejection and usually rises before biopsy confirmation, with AMR tending to generate higher levels than ACR. However, the early literature did not establish whether a positive dd-cfDNA result predicts future prognosis after a rejection episode. Safiullah et al. close that gap by linking biomarker magnitude at diagnosis to long-term outcomes.

Two related 2024 studies also support this broader concept of severity. In the GRAfT race analysis (PMID: 38375637), higher %dd-cfDNA was independently associated with the composite of AR, graft dysfunction, or death, and Black recipients had higher dd-cfDNA levels and higher adverse event rates than White recipients. Although race is not a biologic mechanism, the study underscored that persistent dd-cfDNA elevation may track with cumulative graft injury and adverse longitudinal outcomes. In the DSA/LV dysfunction study (PMID: 39584219), median %dd-cfDNA was higher during pAMR+ and DSA-positive/LV dysfunction episodes than in unaffected patients, and the DSA+/LV dysfunction phenotype carried a strikingly elevated hazard for death or prolonged LV dysfunction (HR 26.2). Together, these reports support the idea that dd-cfDNA magnitude reflects clinically meaningful injury burden.

The 2025 study of elevated dd-cfDNA despite negative EMB (PMID: 41025234) adds nuance. Patients with dd-cfDNA ≥0.20% and negative biopsy had increased mortality (HR 4.6) and possible graft dysfunction risk, suggesting that pathology-negative biomarker elevation may represent sampling error, early injury, or non-histologic graft pathology. This is especially relevant to the GRAFT report: the prognostic signal of dd-cfDNA likely extends beyond biopsy-confirmed rejection and may capture episodes of injury not fully represented on EMB.

4) Diagnostic performance, thresholds, and why the 0.15% cutoff matters

Thresholds remain assay- and context-dependent. The literature includes commonly used cutoffs around 0.12%, 0.15%, 0.20%, 0.25%, and 0.35%, reflecting differences in platforms, study design, and intended use. In the 2021 validation study, a 0.25% threshold had excellent NPV for AR and could have reduced biopsies substantially. In pediatric practice, thresholds such as >0.2% retained high NPV for AMR and donor-specific antibodies (PMID: 39412485). In the 2025 AMR-focused study, 0.125% and 0.20% were evaluated as early predictors of AMR risk (PMID: 40478999).

The new GRAFT threshold of 0.15% is notable because it was data-driven and used for prognostic stratification rather than diagnosis alone. That means the threshold is not necessarily intended to replace diagnostic cutoffs used in surveillance programs. Instead, it appears to identify a subset of rejection episodes with more intense tissue injury and worse downstream outcomes. Clinically, this supports a two-step interpretation:

Step 1: Is rejection or graft injury present?
Step 2: If present, how severe is the biologic injury signal, and does it predict persistent dysfunction or death?

This distinction is important because a low-positive dd-cfDNA during rejection may represent earlier or more limited injury, whereas a higher value may indicate broader cardiomyocyte damage, greater inflammatory burden, or delayed recognition.

5) Mechanistic and translational insights

Mechanistically, dd-cfDNA is best understood as a damage-associated molecular signal released from donor cardiomyocytes. Because it originates from the injured graft, it integrates multiple upstream processes: alloimmune attack, endothelial injury, microvascular inflammation, myocardial necrosis, and possibly hemodynamic stress or infection-related injury. The Trifecta-Heart translational study (PMID: 38538559) showed correlations between dd-cfDNA and transcripts linked to interferon signaling, rejection, macrophage infiltration, and injury responses such as HMOX1 and SERPINA1. This suggests dd-cfDNA may be a surrogate for an inflamed, injured graft microenvironment rather than a rejection-specific analyte in the narrowest sense.

The elevated dd-cfDNA signal in AMR compared with ACR is biologically plausible because AMR often produces endothelial injury and microvascular damage that can be extensive even when conventional histology is equivocal. Similarly, the DSA+/LV dysfunction phenotype likely represents clinically important antibody-mediated graft injury that biopsy may miss. In this context, dd-cfDNA could serve as a quantitative bridge between immunology and ventricular performance.

Another translational implication is serial measurement. A single dd-cfDNA result is informative, but trajectories may be more informative than thresholds. Rising or persistent elevation may better identify ongoing injury, while falling levels may indicate response to treatment. This aligns with the 2024 paper on absolute cfDNA quantification, which found that absolute donor- and recipient-derived cfDNA trajectories can better reflect treatment response and help disentangle rejection from infection (PMID: 39422356). The field may be moving toward multimodal, kinetic surveillance rather than isolated test interpretation.

6) Adjacent evidence: confounders, multimodality surveillance, and special populations

Several studies define the boundaries of clinical use. The 2022 cohort on confounding factors highlighted pericardial effusion and recent biopsy sampling as potential sources of elevated dd-cfDNA independent of rejection (PMID: 33653997). The 2022 report of elevated dd-cfDNA without clinical rejection showed frequent infection and DSA positivity, reinforcing the need to avoid overcalling rejection on biomarker data alone (PMID: 34863042).

Combined surveillance approaches remain appealing. The paired dd-cfDNA/GEP study (PMID: 35559582) found similar 1-year survival and rejection-free survival versus GEP alone, with fewer biopsies. A larger 2025 analysis of dd-cfDNA plus pGEP found that concordant positivity was associated with the highest rejection rates and a markedly increased mortality risk (PMID: 41085484), suggesting that biomarker concordance may help identify high-risk phenotypes. This supports a layered diagnostic model where dd-cfDNA contributes specificity for injury and GEP contributes systemic immune activation information.

In pediatrics, real-world implementation data are reassuring: dd-cfDNA testing showed high NPV for AMR and DSAs, supporting its use in long-term surveillance (PMID: 39412485). However, multi-organ transplant recipients represent a special challenge because dd-cfDNA may be chronically elevated from noncardiac organs. The 2024 MOT study showed substantial baseline elevations and variability, especially in heart-liver and heart-lung recipients, limiting interpretability for cardiac surveillance in this population (PMID: 39422086).

Outside pure rejection surveillance, dd-cfDNA may also reflect broader graft pathology. Elevated levels were associated with reduced myocardial blood flow in recipients with cardiac allograft vasculopathy, but not necessarily with angiographic CAV alone (PMID: 39655433), implying that dd-cfDNA tracks functional ischemic injury more than luminal anatomy. Meanwhile, primary graft dysfunction was linked to later CAV but not to acute rejection or de novo DSA during the first year (PMID: 36801852), reminding clinicians that not all graft injury pathways are immunologically equivalent.

7) Comparison of major studies

Study	Design / population	Key finding	Clinical implication
Fang et al., Circulation 2021 (PMID: 33435695)	Multicenter prospective, 171 heart transplant recipients	%dd-cfDNA detected AR with AUC 0.92; 0.25% threshold NPV 99%	Established dd-cfDNA as a high-performing noninvasive rejection biomarker
Agbor-Enoh et al., Transplantation 2022 (PMID: 33653997)	Prospective cohort, 87 patients	dd-cfDNA associated with rejection; pericardial effusion and post-biopsy sampling were confounders	Highlighted interpretation pitfalls and context dependence
GRAFT study, Circ Heart Fail 2024 (PMID: 38375637)	Prospective multicenter cohort	Higher dd-cfDNA associated with composite AR/graft dysfunction/death; racial disparities observed	Linked biomarker burden to longitudinal clinical outcomes
Safiullah et al., Circ Heart Fail 2026 (PMID: 42290367)	Prospective multicenter GRAFT analysis	AR with %dd-cfDNA ≥0.15% predicted death/prolonged LV dysfunction; low dd-cfDNA AR did not	Introduced prognostic stratification at diagnosis
Fleischer et al., Circ Heart Fail 2024 (PMID: 39584219)	Prospective multicenter	DSA+/LV dysfunction had high %dd-cfDNA and very high risk of death/prolonged dysfunction	Supports redefining clinically important AMR beyond biopsy positivity alone

Expert Commentary

The central clinical message from the new GRAFT report is that dd-cfDNA is maturing from a binary surveillance test into a risk-stratification tool. This is an important conceptual change. In practice, clinicians do not need only to know whether rejection exists; they need to know whether the episode is likely to resolve with standard therapy or to progress to persistent ventricular dysfunction or death. The 0.15% threshold appears to identify a subgroup with more destructive graft injury, likely reflecting a larger biologic burden at the time of diagnosis.

Still, several caveats matter. First, the threshold was data-driven and requires external validation. Its performance may differ by assay platform, timing after transplant, treatment status, and case mix. Second, the broad AR definition used in GRAFT improves clinical relevance but complicates mechanistic specificity. Third, dd-cfDNA is a marker of injury, not a diagnosis of rejection by itself; infection, hemodynamic stress, pericardial effusion, biopsy timing, and multiorgan grafts can all confound interpretation. Fourth, whether biomarker-guided intensification of therapy improves outcomes remains unproven.

Guideline-wise, EMB remains the reference standard in current practice, and dd-cfDNA should be viewed as complementary rather than replacement technology. The strongest near-term use case is likely a multimodal surveillance algorithm in which dd-cfDNA helps reduce unnecessary biopsies, prompts closer evaluation when elevated, and—based on emerging data—helps decide which rejection episodes warrant more aggressive treatment and follow-up imaging.

From a translational perspective, future studies should test whether integrating dd-cfDNA with DSA, echocardiography, hemodynamics, GEP, and perhaps cardiac MRI or PET-derived functional measures can improve prediction of persistent dysfunction. Serial kinetics, absolute quantification, and machine-learning risk models may further refine accuracy. It will also be important to define how quickly dd-cfDNA should fall after treatment and whether failure to normalize predicts relapse or chronic graft injury.

Conclusion

The dd-cfDNA literature in heart transplantation now supports three sequential clinical roles: early rejection detection, biologic injury characterization, and prognostic risk stratification. The 2026 GRAFT study is a milestone because it shows that AR is not prognostically uniform; elevated %dd-cfDNA at diagnosis identifies patients at substantially higher risk of prolonged LV dysfunction and death. This complements earlier diagnostic studies and newer work on biopsy-negative injury, AMR redefinition, and multimodal surveillance.

For clinicians, the immediate implication is not to abandon EMB, but to interpret dd-cfDNA as a quantitative measure of graft danger that can inform intensity of surveillance and follow-up. For researchers, the challenge is to validate thresholds, standardize assay interpretation, and test whether dd-cfDNA-guided management can improve hard outcomes. The field has clearly moved beyond “Does rejection exist?” to a more actionable question: “How sick is the graft, and what is the risk of not recovering?”

References

Safiullah ZN, Su H, Kong H, Jang MK, Shah K, Berry GJ, Shah P, Valantine HA, Tian X, Agbor-Enoh S. Donor-Derived Cell-Free DNA Stratifies Risk of Mortality and Graft Dysfunction in Severe Acute Cardiac Allograft Rejection. Circulation. Heart failure. 2026-06-15:e013718. PMID: 42290367.
Agbor-Enoh S, et al. Cell-Free DNA to Detect Heart Allograft Acute Rejection. Circulation. 2021;143(12):1184-1197. PMID: 33435695.
Graft-derived Cell-free DNA as a Noninvasive Biomarker of Cardiac Allograft Rejection: A Cohort Study on Clinical Validity and Confounding Factors. Transplantation. 2022;106(3):615-622. PMID: 33653997.
Observed elevated donor-derived cell free DNA in orthotopic heart transplant recipients without clinical evidence of rejection. Clin Transplant. 2022;36(3):e14549. PMID: 34863042.
Combining donor derived cell free DNA and gene expression profiling for non-invasive surveillance after heart transplantation. Clin Transplant. 2023;37(3):e14699. PMID: 35559582.
Comparison of two donor-derived cell-free DNA tests and a blood gene-expression profile test in heart transplantation. Clin Transplant. 2023;37(6):e14984. PMID: 37036133.
Racial Differences in Donor-Derived Cell-Free DNA and Mitochondrial DNA After Heart Transplantation, on Behalf of the GRAfT Investigators. Circ Heart Fail. 2024;17(4):e011160. PMID: 38375637.
Comparing Plasma Donor-derived Cell-free DNA to Gene Expression in Endomyocardial Biopsies in the Trifecta-Heart Study. Transplantation. 2024;108(9):1931-1942. PMID: 38538559.
Redefining Cardiac Antibody-Mediated Rejection With Donor-Specific Antibodies and Graft Dysfunction. Circ Heart Fail. 2024;17(12):e011592. PMID: 39584219.
Elevated Donor-Derived Cell-Free DNA Levels Are Associated With Reduced Myocardial Blood Flow but Not Angiographic Cardiac Allograft Vasculopathy: The EVIDENT Study. Circ Heart Fail. 2025;18(1):e011756. PMID: 39655433.
Significance of Elevated Donor-Derived Cell-Free DNA in Heart Transplant Recipients With Negative Endomyocardial Biopsies: A Dawn of a New Era. Circ Heart Fail. 2025;18(11):e012787. PMID: 41025234.
Clinical Utility of Combined Donor-Derived Cell-Free DNA and Peripheral Gene-Expression-Profiling in Heart Transplant Recipients. Clin Transplant. 2025;39(10):e70340. PMID: 41085484.
Increased Donor-Derived Cell-Free DNA as a Predictor for the Early Detection of Antibody-Mediated Rejection Following Heart Transplantation. Clin Transplant. 2025;39(6):e70209. PMID: 40478999.
Progress in Noninvasive Surveillance for Acute Rejection in Pediatric Heart Transplant Recipients: A Real-World Analysis of Donor-Derived Cell-Free DNA-Based Surveillance Protocol. Clin Transplant. 2024;38(10):e15481. PMID: 39412485.