#NephJC Chat
Tuesday June, 21st 2022 at 9 pm Eastern
Wednesday June 22nd, 2022, at 9 pm IST and 4:30 pm BST
Kidney Int 2022 Apr; 101(4): 793-803.
doi: 10.1016/j.kint.2021.11.034. Epub 2021 Dec 22.
Clinical outcomes from the Assessing Donor-derived cell-free DNA Monitoring Insights of kidney Allografts with Longitudinal surveillance (ADMIRAL) study
Lihong Bu, Gaurav Gupta, Akshta Pai, Sanjiv Anand, Erik Stites, Irfan Moinuddin, Victor Bowers , Pranjal Jain, David A Axelrod, Matthew R Weir, Theresa K Wolf-Doty , Jijiao Zeng, Wenlan Tian, Kunbin Qu, Robert Woodward, Sham Dholakia, Aleskandra De Golovine, Jonathan S Bromberg, Haris Murad, Tarek Alhamad
PMID: 34953773
Introduction
Kidney transplantation is associated with an improved quality of life and survival compared to dialysis and is currently the treatment of choice for kidney failure. Allograft rejection occurs when the recipient’s immune system recognizes the donor kidney tissue as “foreign” leading to inflammation, injury and loss of allograft failure.
Current standard of care for surveillance of kidney allograft function includes monitoring of serum creatinine, immunosuppression medication drug levels, and (in many countries) de novo donor specific antibodies (dnDSAs) titers. Elevated creatinine can be due to many causes in this setting: acute allograft rejection, calcineurin inhibitor (CNI) toxicity, BK virus nephropathy, infections, thrombotic microangiopathy (TMA), or acute tubular necrosis (ATN), to name a few. When allograft rejection is suspected, physicians rely on invasive kidney biopsies which are subjected to sampling errors and variable interpretation among renal pathologists (Furness and Taub, KI 2001).
Donor-derived cell-free DNA (dd-cfDNA) is a non-invasive nucleic acid marker that has been suggested for early detection of allograft dysfunction and response to treatment (Oellerich et al, Nat Rev Nephrol 2021). Dd-cfDNA are cell-free fragments that are released from the injured donor cells into the circulation following cellular inflammation and apoptosis. Measuring dd-cfDNA in the bloodstream can hence indicate allograft injury, which could be due to acute rejection, BK nephropathy, ATN, CNI toxicity, or any cause of allograft injury. Paresh Jadav wrote an excellent dd-cfDNA overview on Renal Fellow Network.
The Assessing Donor-derived cell-free DNA Monitoring Insights of kidney Allografts with Longitudinal surveillance (ADMIRAL) study is a large multicenter study that evaluated the utility of dd-cfDNA in diagnosing subclinical rejection, the relationship between dd-cfDNA and non-immune allograft injury, and predictors of long term allograft survival in kidney transplant recipients.
The Study
Design and Aim:
An observational multicenter study aiming to evaluate the utility of dd-cfDNA in detecting early allograft dysfunction and assessing response to treatment over a period of 3 years.
Study Population:
1092 patients across 7 transplant centers were monitored with dd-cfDNA in addition to routine standard of care. Data from June 1, 2016 to Jan 31,2020 was collected.
Exclusion Criteria included:
Pregnancy
Multiple Organ Recipients
Monozygous twin-twin transplants
Prior bone marrow transplant.
Donor-Derived Cell-Free DNA Methodology:
Dd-cfDNA was measured at regular intervals using a targeted next-generation sequencing assay according to each center’s standard of practice (supplementary table S2)
Biopsy-defined rejection and graft dysfunction:
Results of protocol and for-cause kidney biopsies were collected. Biopsies were interpreted according to the Banff 2019 classification by a local pathologist followed by a central pathologist. Biopsies were paired with dd-cfDNA measured within 30 days preceding biopsy.Indications for for-cause biopsies included one or more of the following:
Worsening in creatinine
Worsening in proteinuria
Positive dnDSA
Treatment of allograft rejection was based on individual center protocols.
Please refer to supplementary table S1 for a full list of data collected.
Statistical Analysis:
Descriptive analyses were used for demographics and dd-cfDNA measurements. Performance characteristics (sensitivity, specificity and predictive values) of dd-cfDNA was calculated based on discriminatory power thresholds of 0.5% and 1%. Patients were further categorized into high (≥ 0.5%) or low (<0.5%) dd-cfDNA and were compared using Fisher’s exact test and Student’s t-test for categorical and continuous variables, respectively. The discriminatory power of dd-cfDNA was assessed using the area under receiver-operating characteristic curve (AUROC). Predictors for high dd-cfDNA were determined using multivariate logistic regression analysis.
Estimated glomerular filtration rate (eGFR) was assessed using the MDRD formula. Machine-learning based Spearman rank correlation was used to assess the relationship between eGFR and dd-cfDNA. Generated clusters included 3 different time intervals for assessment: (1) 0-4 months, (2) 4-12 months, (3) 12-36 months.
The relationship between dnDSAs and dd-cfDNA was evaluated using a multivariate cox proportional hazard model. A positive dnDSA was defined as a mean fluorescence intensity of >500.
The utility of dd-cfDNA as a marker of allograft quiescence (defined as absence of injury) was evaluated retrospectively. Allograft injury included one or more of the following:
Out-of range tacrolimus levels (<4 or >12 ng/ml)
BK viremia
DnDSA positivity
Urinary tract infection
Proteinuria
Allograft rejection
Recurrent focal segmental glomerulosclerosis
Funding:
This study seems to have been sponsored by CareDx, Inc., producers of dd-cfDNA surveillance products under the brand name AlloSure®, though this is not clearly stated anywhere in the manuscript. Clinicaltrials.gov identifier NCT04566055. Six of the authors were CareDx employees, five other authors served on the Advisory Board for the company (and have done consulting work) and two received funding for research studies. Another person (not an author) seems to have performed the statistical analysis, and some authors ‘participated’ in the analysis.
Results
The flow diagram with subsets of the ADMIRAL cohort and patients included in the study are shown in Figure 1 . Baseline characteristics of the ADMIRAL cohort compared to the United Network of Organ Sharing (UNOS) 2020-2021 registry are shown in Table 1.
Figure 1: Flow diagram of the ADMIRAL study cohort
Table 1 for the ADMIRAL study
The ADMIRAL cohort was composed of many more deceased donor kidney allografts (94% vs 68%) and fewer re-transplants (8% vs 13%) compared to the full UNOS registry dataset. The demographic comparisons for those with rejection vs no rejection are shown in Table 2.
Table 2 shows the characteristics of the ADMIRAL cohort comparing those with rejection vs no rejection. There was no statistical significance between both cohorts in terms of median serum creatinine and eGFR.
Note that sex/race percentages for patient are by rows; but sex/race percentages by donor seem to be by the column - with no explanation. The 828% for female donors with no rejection is a mathematical impossibility so likely a typo for 28%. There might be more errors in the table, so interpret the data carefully.
However, the median dd-cfDNA in patients without rejection was significantly lower compared to those with rejection (0.23% vs 1.6%, p<0.0001, Figure 2a). The AUROC for all rejection dd-cfDNA was significantly higher than the AUROC of creatinine (0.798 vs 0.492, p < 0.001, Figure 2b).
Figure 2 Relationship between dd-cfDNA and creatinine levels in patients with and without rejection.
Table 3 shows the performance characteristics of dd-cfDNA in discriminating the presence or absence of rejection after excluding other causes of allograft injury.
Table 3 showing performance characteristics of dd-cfDNA in discriminating allograft rejection. ABMR: Antibody mediated rejection. TCMR: T-Cell mediated rejection. PPV: Positive predictive value. NPV: Negative predictive value
The relationship between eGFR and dd-cfDNA was assessed using Spearman rank correlation. Results, post transplant are shown in Figure 3:
0-4 months: Trends were ‘erratic’ (we presume scattered with no clear trend) with no correlation
12-36 months:Positive correlation between elevation of dd-cfDNA and eGFR decline (R: -0.84, p=0.01)
The relationship between dd-cfDNA and dnDSA was assessed. Dd-cfDNA >0.5% was associated with ~ 3 fold elevation in dnDSA formation (Figure 4).
Figure 4: Free from dnDSA model in relation to dd-cfDNA levels.
The association between dd-cfDNA and composite graft injury was assessed. Table 5 shows the performance characteristics of dd-cfDNA as a molecular marker for graft injury. The median dd-cfDNA in quiescent (non-injury) kidney allografts was 0.21% compared to 0.51% in those with graft injury (p<0.0001, Figure 5). The AUROC for dd-cfDNA was 0.727 (95% CI 0.71-0.88).
Table 5 showing the performance characteristics of dd-cfDNA as a molecular marker for graft injury. PPV: Positive predictive value. NPV: Negative predictive value
Figure 5 showing the relationship between dd-cfDNA and graft injury.
Discussion
The quest for renal troponin continues, and dd-cfDNA is a new armamentarium in the arsenal. What makes it unique is its ability to predict subclinical injury in the kidney allograft which has been successfully demonstrated in clinical trials. The sentinel study (DART study by Brennan et al, in JASN 2017), retrospectively evaluated dd-cfDNA levels, found that a cut off level of > 1% for dd-cfDNA was able to distinguish between active rejection vs no rejection with an AUROC of 0.74. It was even better at distinguishing antibody mediated rejection (ABMR) vs no-ABMR with an AUROC of 0.87. Quite impressive - right? That led to its use by > 150 transplant centers in the United States (Jordan et al, KI 2022), and based on a limited number of studies, also driving its use as a marker for non-specific allograft injury rather than limiting it to rejection (Gupta et al, Am J Transplantation 2020, Sawinski et al, Clin Transplant 2021, Gousosus et al, Transplantation Direct 2020 & Huang et al, Transplantation Direct 2020). The ADMIRAL study aimed to assess the utility of routine use of monitoring of dd-cfDNA in the real world. Let’s discuss if they found it to be useful in routine clinical practice!
The study confirmed the utility of dd-cfDNA to distinguish rejection (both clinically evident and subclinical) from no rejection at a cut-off of both 0.5% and 1.0 % with an AUROC of 0.8. When evaluating dd-cfDNA as a marker of kidney dysfunction, it didn’t really come as a shock when dd-cfDNA outperformed creatinine (don’t we all know the limitations of creatinine as a marker of kidney allograft dysfunction!). This study showed no statistically significant difference in creatinine between patients with and without rejection, while dd-cfDNA was significantly lower in patients without rejection compared to those with rejection.
It also showed that elevated dd-cfDNA level preceded future eGFR decline and formation of dnDSAs. The association of dd-cfDNA with decline in eGFR was especially notable during month 12 and 36 of the post-transplant period. However, only 4.6% patients developed de novo DSA, and therefore the absolute risk appears to be very small.
The study also shows that elevated dd-cfDNA levels correlated with a composite of ‘allograft injury’. Theoretically, this should not be surprising either as dd-cfDNA is released as a result of cell breakdown, which would ultimately translate to decline in allograft function. It must be noted however that ‘Allograft injury’, as defined in the study, encompasses out-of-range tacrolimus levels (<4 or >12 ng/ml), BK viremia, positive serum DSA, urinary tract infection, proteinuria, allograft rejection, or recurrent focal segmental glomerulosclerosis. It should be noted though, and is rightly pointed out by Jordan et al. in the commentary on the article, that 4 of these 7 diagnoses are not universally associated with allograft injury (i.e., out-of-range tacrolimus levels, BK viremia in the absence of nephropathy, dnDSA in the absence of histologic ABMR, and acute cystitis without pyelonephritis may not be associated with allograft injury). In addition, a retrospective study (Bromberg et al, Transplantation Direct 2021) has found that even in patients with normal dd-cfDNA level, non-immune allograft injury was present. A very small number of patients in this study also had rejection (3 of 41) with a normal dd-cfDNA level. Keeping this in mind, the clinical implication of elevated dd-cfDNA in this setting remains uncertain, despite the results of ADMIRAL study.
Lastly, question arises regarding the follow up of ‘elevated dd-cfDNA result’ - surely, it does not provide any guidance regarding the type of rejection, or the precise nature of allograft injury for that matter, and therefore currently cannot replace a kidney biopsy. Additionally, with a positive predictive value of less than 60%, a case cannot be made for putting everyone through a kidney biopsy who turns up a positive test!! On the other hand, a negative result has a negative predictive value of approximately 90% for rejection.
The paper shows a lot of positives for the dd-cfDNA: it is a better marker for allograft injury, it is better able to predict future decline in eGFR, & it is even able to foresee development of dnDNA. Yes, it was able to detect early injury however it was unable to differentiate the type of injury. Almost all centers, barring 1, set the protocol for testing patients 15 times in the first 36 months post-transplant. However, what clinicians and patients are looking for after spending thousands of additional dollars on testing is whether it improves allograft survival and patient outcomes. So far, we do not have anything supporting that. It does not reflect if any earlier interventions, if any made, were helpful in improving outcomes.
The authors mention the complications and inadequacies of a kidney biopsy, and use this argument as one of the reasons for the development and use of the dd-cfDNA test. In spite of that, the study fails to show if dd-cfDNA could replace the ‘need for tissue diagnosis’. With a positive test, you still need ‘tissue’ to obtain the definitive diagnosis to determine the appropriate course of action. Rather, it raises the concern whether we might be subjecting more patients to an unnecessary procedure after getting a positive test result. It needs to be studied in further detail how many additional biopsies will be needed to prevent, or delay, graft failure in one patient.
Some reassurance is provided however when the test is negative, and in the right clinical scenario this may be an indication to decrease long-term immunosuppression, and hence prevent further side-effects from these medications.
It must be noted that almost all authors, even the ones who did the commentary, have various conflicts of interests with CareDx. While no one here is questioning their integrity, we think it needs to be mentioned - and it would be good to replicate these findings by other groups before being widely used.
Conclusion
Like many tests in medicine, the utility of dd-cfDNA is probably in a negative result; a negative test rules out allograft injury with reasonable confidence. This study adds to the growing literature supporting the theoretical benefits of dd-cfDNA. However we believe that the current state of evidence does not support routine use of dd-cfDNA for screening in every patient, especially given high costs (~ $2800 per test). Further studies are needed to categorically show improved graft survival with its use and also to characterize the specific patient population which may benefit from early intervention based on a positive result.
Summary prepared by
Mohamed Hassanein
Assistant Professor of Nephrology
University of Mississippi Medical Center &
Sheikh Bilal Khalid
Attending Nephrologist
Shaukat Khanum Memorial Cancer Hospital
NSMC Interns, Class of 2022
Reviewed by Swapnil Hiremath, Jamie Willows, Jade Teakell, and Anju Yadav