Sequential Biomarker Testing in Kidney Transplant Surveillance

In this issue of CJASN, Singla et al. describes a two-step testing strategy to inform parsimonious use of blood-based biomarkers and potentially reduce unnecessary kidney transplant biopsies among stable, adult kidney transplant recipients.¹ The authors leveraged two cohorts of adult kidney transplant recipients who underwent surveillance biopsies and had stored blood and urine samples. The discovery cohort was the prospective multicenter Clinical Trials in Organ Transplantation 08 with 476 biopsies from 226 recipients. The validation cohort was a northwestern single-center cohort of 144 biopsies from 134 recipients.

The study has important implications for clinical practice in the current era in which there is a growing armamentarium of diagnostic tools to detect subclinical allograft injury but limited data to inform how and when to use varying tests to optimize utility, reduce cost, and ultimately improve outcomes.

The short-term survival of kidney transplantation has improved significantly over the past decades with a dramatic decline in the 1-year rejection rates approaching 8%. By contrast, long-term graft survival remains suboptimal with nearly 50% of deceased donor kidney transplants failing within 10 years, primarily because of alloimmune injury.² Routine strategies for monitoring graft function, including serum creatinine, urine protein, immunosuppressive drug levels, and donor-specific antibodies (DSAs), lack the sensitivity and specificity to diagnose rejection. By the time these markers are abnormal and “for-cause” biopsy is performed, significant and irreversible kidney damage is often present. Thus, preemptive strategies to detect early immune activation and prevent irreversible graft injury are critically needed.

Subclinical rejection (SCR) occurs in 20%–25% of stable kidney transplant recipients, and by definition, this type of rejection is only diagnosed by surveillance or protocol biopsy. SCR is associated with worse long-term graft outcomes. This is particularly true for subclinical antibody-mediated rejection (ABMR), which has the worst outcomes and is associated with a 3.5-fold increase in graft loss.³ Although protocol biopsies are the current gold standard to detect SCR, they are only performed in 20% of transplant centers because of prohibitive cost, patient burden, potential complications, and limitations in sampling. In an ideal clinical setting, biopsies would be performed only in those patients with the greatest likelihood of having clinically actionable findings. To this end, several novel noninvasive biomarkers have been incorporated into clinical practice, and some have received Centers for Medicare & Medicaid Services (CMS) coverage approval.⁴

Singla et al.¹ developed a two-stage diagnostic strategy combining three noninvasive biomarkers obtained from stable, kidney transplant recipients who underwent surveillance biopsies. They used urinary chemokines as a stage 1 rapid low-cost test to rule out rejection. C-X-C motif ligand 9 (CXCL9) and C-X-C motif ligand 10 are urinary chemokines excreted into urine by tubular epithelial cells and infiltrating leukocytes in response to IFN-γ‑mediated intragraft inflammation. CXCL9, the stage 1 biomarker used in the present study, was validated as a rule-out test in the Clinical Trials in Organ Transplantation-01 multicenter cohort with a negative predictive value (NPV) of 0.92.⁵ Currently, the urinary chemokine assay is commercially available as a research platform without an established CMS reimbursement pathway.

If a participant had a stage 1 positive test, then stage 2 blood-based testing was conducted with donor-derived cell-free DNA (dd-cfDNA) and/or gene expression profiling (GEP). dd-cfDNA is released into recipient's plasma from injured endothelium and has been shown to perform best as a rule-out test for ABMR as demonstrated in the Circulating Donor-Derived Cell-Free DNA in Blood for Diagnosing Acute Rejection in Kidney Transplant Recipients study where it had a NPV of 0.96.⁶ Blood-based GEP detects peripheral transcript signatures of rejection and is also used as a rule-out assay. Its accuracy has been discordant across validation cohorts and depending on clinical settings as well as the type of assay used.

The dd-cfDNA and GEP platforms are available as laboratory-developed tests with CMS coverage at approximately $2500–3000 per assay.

The two-stage sequential biomarker strategy was compared with rejection diagnoses on histology from surveillance biopsies. The authors sought to investigate whether the two-step approach could reduce utilization of expensive blood-based testing and unnecessary biopsies. It is a retrospective secondary analysis using stored blood and urine samples paired with surveillance biopsies from two cohorts. The CXCL9-then-dd-cfDNA strategy performed similarly in the discovery and validation cohorts (area under the receiver-operating characteristic curve [AUC] of 0.82 in both). In the validation cohort, it achieved a positive predictive value of 0.39, NPV of 0.94, and reduced stage 2 testing to 79% of patients, a 21% reduction relative to universal blood-based testing. The combined dd-cfDNA+GEP strategy was marginally better (validation AUC 0.85, NPV 0.96), while GEP alone underperformed (validation AUC 0.66 in the two-stage configuration with confidence intervals crossing 0.5). The authors conclude that the two-stage strategy maintains diagnostic accuracy and reduces utilization of expensive blood-based noninvasive biomarkers by 21%. They propose it as a more efficient strategy for monitoring of stable kidney transplant recipients. It is worth noting that the authors directly measured a reduction in blood testing, not in biopsies; because all patients were biopsied in this study, fewer biopsies is an expected benefit of the strategy's high NPV rather than a measured outcome.

This study has several strengths. The derivation and validation cohorts share identical inclusion and exclusion criteria with large sample sizes. The use of a decision sequence is an advance in screening over prior studies that have used biomarkers independently or concurrently. The study also has a well-defined clinical context in which stable kidney transplant recipients were undergoing protocol biopsies to screen for SCR.

There are some limitations to note. First, the GEP showed only modest discrimination in the discovery cohort (AUC, 0.68) and failed to maintain statistical significance in the validation cohort (AUC, 0.60, with a confidence interval crossing 0.5), likely because of the platform change from microarray to quantitative PCR between cohorts. Second and more importantly, the two-step testing strategy detects subclinical ABMR reliably but performs poorly for subclinical cellular rejection (validation AUC 0.90 versus 0.63, with a confidence interval crossing 0.5).

That result is biologically unsurprising. In ABMR, dd-cfDNA is released from injured allograft endothelial cells into the bloodstream. On the other hand, in T-cell‑mediated rejection (TCMR) in which tubular injury is dominant, the inflammatory milieu is likely most evident the urine. Notably, with higher grades of cellular rejection, detection can improve with dd-cfDNA, especially with vascular involvement. This grade-dependent pattern was demonstrated in the multicenter Assessing Donor-derived cell-free DNA Monitoring Insights of kidney Allografts with Longitudinal surveillance study (N=1092), where median dd-cfDNA was 0.20% in borderline TCMR (below the 0.5% positivity threshold), 0.78% in TCMR 1A, 1.3% in TCMR 1B, and 3.68% in TCMR 2A.⁷ Because surveillance biopsies in clinically stable patients will identify predominantly borderline and low-grade TCMR, the limitations of dd-cfDNA in this population are particularly relevant to the proposed two-stage strategy. Adding GEP as a second-stage test improved ABMR detection further (AUC, 0.96) but offered only a marginal gain for cellular rejection (AUC, 0.69) and reduced specificity overall.

Perhaps the most clinically informative observation comes from the authors' sensitivity analysis. When DSA status was added to the stage 1 model, the proportion of patients requiring stage 2 testing decreased from 79% to 31%, with DSA carrying an odds ratio of 7.05 for rejection. The combined model performed better than CXCL9 alone, but the magnitude of DSA's contribution suggests that much of the discriminatory work was being done by DSA. This raises a question the study does not directly address: How much additional value does the urinary chemokine assay provide over a DSA-guided surveillance strategy alone? The randomized trial by Akifova et al. offers a partial answer⁸: selective dd-cfDNA monitoring in recipients with de novo DSA (no urinary chemokine involved) achieved a positive predictive value of 0.77 for ABMR, more than twice that of the staged strategy applied to unselected surveillance patients. The same biomarker applied to a different population yields fundamentally different clinical utility.

Yet, sole reliance on DSA to guide biopsy indication is also likely insufficient. Recent evidence has identified microvascular inflammation as a key driver of poor graft outcomes even in DSA-negative recipients,⁹ highlighting that important rejection-like injury occurs outside the DSA-positive subgroup. These findings suggest that the most useful role for assays like dd-cfDNA or urinary chemokines may ultimately be to identify the DSA-negative patients who merit closer attention.

A further consideration is that more sensitive detection of SCR has not yet translated into better clinical outcomes: The Hirt-Minkowski trial, which was the first randomized trial of urinary C-X-C motif ligand 10‑guided intervention (elevated values triggered biopsy with therapeutic adaptations according to results), did not improve its primary clinical outcome at 1 year.¹⁰

In conclusion, the work by Singla et al. offers a thoughtful foundation for using noninvasive biomarkers more efficiently in kidney transplant surveillance. They showed that a two-stage strategy was practical to reduce reliance on costly blood-based tests while preserving diagnostic accuracy for ABMR. Questions remain about cellular rejection detection, optimal patient selection, and clinical utility. Future work should evaluate whether biomarker-guided monitoring improves long-term graft outcomes, particularly as emerging therapies for ABMR may finally make earlier rejection detection clinically actionable. Until then, the staged workflow offers a reasonable bridge between current surveillance practice and biomarker-driven monitoring of the future.

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C782.

Author Contributions

Conceptualization: Sandra Amaral, Patrick Hallak.

Formal analysis: Sandra Amaral, Patrick Hallak.

Methodology: Sandra Amaral, Patrick Hallak.

Writing – original draft: Patrick Hallak.

Writing – review & editing: Sandra Amaral, Patrick Hallak.

Funding

None.