摘要
Finding the right balance between underimmunosuppression and overimmunosuppression remains one of the holy grails of kidney transplants. In the absence of a reliable laboratory measure of immunosuppression burden (an “immunostat”), clinicians depend upon noninvasive measurements of traditional laboratory biomarkers and allograft biopsies to guide immunosuppression dosage and regimen. An important premise for seeking new non invasive monitoring tools is that long-term graft injury is driven by subclinical events undetected by traditional methods, such as a trend in serum creatinine levels. Studies have reported an incidence of up to 20% to 30% subclinical rejection in seemingly clinically stable patients.1 The last few years have seen an explosion in research focusing on the use of donor-derived cell-free DNA (dd-cfDNA) as the preeminent biomarker for the real-time detection of posttransplant rejection.2 A meta-analyses of all published studies showed that, at a fractional threshold of ≥1%, sensitivity for the diagnosis of antibody-mediated rejection (AMR) was excellent, although it was less so for T-cell mediated rejection (mostly elevated at ≥1% threshold in Banff 1B or higher rejections).3 The major studies utilizing commercially available dd-cfDNA assays reported the following test performance characteristics for all rejections (using the 0.69%–1% cutoff to define test positivity): an area under the curve of 0.71 to 0.85, the sensitivity of 45% to 89%, specificity of 69% to 85%, positive predictive value of 52% to 77%, and a negative predictive value of 66% to 95% depending on the pretest probability of rejection.4 Most of these previous studies were performed in the context of ad hoc testing in patients with a presumed high pretest probability for rejection. The study by Huang et al5 assesses the predictive value of an initial single dd-cfDNA measurement in patients (between 1-m and 4-y posttransplant) with the absence of current or previous renal dysfunction, rejection episodes, and donor-specific antibody (DSA). The authors divided patients into 3 groups based upon dd-cfDNA: (a) high (N = 35/317; >1%); (b) moderate (N = 43/317; 0.5%–1%); and (c) low (239/317; <0.5%) dd-cfDNA. Allograft biopsies were performed on an average of 2 mo after dd-cfDNA measurement and were based on clinician preference rather than any set criteria (76% “for-cause” and 24% based on a moderate/high dd-cfDNA). Among the 62 (20%) patients biopsied, 24 were diagnosed with rejection. Those with high (6/25; 17%) or moderate (5/43; 12%) dd-cfDNA were more likely to have rejection than those with low (13/239; 5%) dd-cfDNA. Despite the lack of biopsies in the remainder of the patients, the authors report that there was no difference in short-term graft outcomes (median follow-up: 1.6 y) as measured via trends in kidney function (estimated glomerular filtration rate [eGFR]) and de novo DSA formation. They conclude that most patients with elevated dd-cfDNA in conjunction with preserved allograft function remained stable over follow-up without deterioration in function (as measured by eGFR) or graft loss. Strengths of this study included large sample size, a well-known research group, and a “real-world scenario” with inherent complexities that clinicians need to navigate to provide care to kidney transplant patients. There are several weaknesses that need to be highlighted, though. The study population is too heterogeneous to derive any definitive conclusions from. There are several other instances of selection bias. This is suggested by the inclusion of patients all the way from 1 mo to 4 y from transplant. The authors suggest that tests were sent for routine surveillance, but the scenarios where a clinician taking care of a patient more than a year posttransplant with preserved graft function and no DSA would send dd-cfDNA are not clear. Allograft biopsies were done for varied reasons in only a small minority of patients and on an average 2-mo postreporting of dd-cfDNA result. What immunosuppressive adjustments were performed in the interim? In the absence of these data, the interpretation that there was no impact of dd-cfDNA measurements on 1-y graft outcomes is overly simplistic. The possibility of a type II error (false-negative outcome) due to inadequate follow-up also needs to be considered. A recently published large multicenter study that included nearly 1100 kidney transplant patients demonstrated a higher risk of loss of eGFR over 3 y and de novo DSA formation over time among kidney transplant patients with dd-cfDNA >0.5%.6 In our single-center study on the calibration of dd-cfDNA with traditional histopathology and molecular tissue gene expression (MMDx, Alberta, CA), sensitized patients at high risk for rejection (preformed or de novo DSA, and/or positive flow crossmatch at transplant, and/or documented medication nonadherence) were found to have a strong correlation with dd-cfDNA >1% and subclinical AMR diagnosed by MMDx but not by histopathology.7 Similar findings were confirmed by the multicenter Trifecta study.8 This study also reported that nearly half of the AMR cases were DSA negative. Other data using tissue gene expression have shown that many DSA positive but histologic and MMDx negative AMR cases do have upregulation of rejection transcripts.9 In addition, the Banff criteria for the diagnosis of T-cell mediated rejection and AMR continue to evolve.10 Finally, it is important to realize the possibility of significant interobserver variation in transplant biopsy interpretation by pathologists. These data indicate the complexity of calibrating noninvasive biomarkers to a flawed gold standard that is a moving target. Notwithstanding the above criticisms, the study by Huang et al is important in that it provides initial data on the uncertain value of routine cross-sectional dd-cfDNA testing for patients with a low pretest probability of rejection. Two ongoing prospective studies (The Ongoing Kidney Allograft Outcomes Registry [KOAR], NCT03984747, and The Prospera Kidney Transplant ACTIVE Rejection Assessment Registry [PROACTIVE], NCT03984747) aim to assess the impact of serial dd-cfDNA surveillance in kidney transplant recipients. Until these data become available, it seems reasonable to use clinical context and personalization of care when considering the utilization of dd-cfDNA surveillance in patients with preserved allograft function at a presumed minimal risk of rejection.