Two Can Be Better Than One

Despite recent advances in transplant immunosuppression and postkidney transplant care, acute rejection episodes continue to contribute to recipient morbidity and can negatively affect long-term allograft survival. In particular, studies performed over the past 2 decades documented up to a 20%–30% incidence of rejection diagnosed on a surveillance biopsy performed in the absence of a clinical indication (1–3). If unrecognized or untreated, these subclinical rejections can contribute to late graft dysfunction and graft loss (2,3). As the biopsy procedure is costly, is inconvenient, and engenders a small but measurable risk of graft injury or loss, the identification of noninvasive biomarkers capable of reliably diagnosing subclinical rejection episodes in lieu of a biopsy remains a high priority for the transplant field.

Considerable effort by multiple research teams has resulted in the discovery of several promising noninvasive assays for this purpose. Prospective, multicenter studies published since 2010 have shown that urinary and PBMC gene expression profiling approaches (4,5), urinary chemokine assessments (6), and the quantification of donor-derived cellfree DNA (dd-cfDNA) (7), among others, are potentially capable of guiding clinical decision making. Suboptimal accuracy, specificity, and sensitivity of each individual assay for diagnosing subclinical rejection, as well as a lack of availability, have thus far prevented their widespread clinical use, and transplant biopsies currently remain the “gold standard.” Whether combining assay approaches improves diagnostic utility has not been carefully examined.

In this issue of CJASN, Park et al. (8) tested whether combining two mechanistically distinct noninvasive biomarkers for diagnosing subclinical rejection—PBMC RNA expression and plasma dd-cfDNA—would improve their diagnostic utility compared with either assay alone. The first assay uses microarrays to quantify RNA transcripts isolated from recipient PBMCs, yielding a specific gene expression pattern that associates with the presence of subclinical rejection. Data published from the prospective, observational Clinical Trials in Organ Transplantation-08 (CTOT-08) cohort of 812 biopsies from 283 patients and two independent validation sets (4) showed that a negative test result strongly correlates with the absence of subclinical biopsy-proven acute rejection (BPAR), with a negative predictive value (NPV) of 78%–88%. A positive assay result detects subclinical BPAR, but the significant false-positive rate leads to a positive predictive value (PPV) of only 47%–61%. Work by others testing plasma dd-cfDNA assays for clinically evident BPAR (subclinical rejection was not studied) showed that this assay also has a high NPV (84%) but a relatively low PPV (61%) for detecting BPAR in kidney transplant recipients (7).

In the CJASN study, Park et al. (8) reanalyzed a subset of their data from the CTOT-08 cohort (423 paired blood and biopsy samples from 208 subjects) and from their internal biorepository cohort (105 paired blood and biopsy samples from 85 subjects) and added new dd-cfDNA assay results generated from stored plasma samples obtained at the same time points. Their analyses of the CTOT-08 subset confirmed the low PPVs for each of the assays alone (58% for PBMC gene expression and 47% for dd-cfDNA assays) but remarkably noted that when both assays were positive at the same time point, the PPV increased to 81%. The combination approach (when both tests are negative) actually yielded an NPV of 88%, higher than either test alone. The authors confirmed an analogous increase in the PPV without a change in the NPV in their independent biorepository validation set. The implications of these findings, assuming that the biopsy diagnoses are correct, are that if the physicians had made treatment decisions on the basis of the results of the two assays without a biopsy, 20% of subclinical BPAR episodes would not have been treated, and approximately 20% of those patients treated for rejection would have been subjected to unnecessary increases in immunosuppression.

It is important to note, however, that transplant biopsy interpretation by expert pathologists can vary by as much as 30% (9), sampling error can result in the underdiagnosis of pathologic infiltrates, and the long-term effects of borderline subclinical rejection on graft function remain incompletely understood (2,3). Together with the emerging evidence that molecular diagnostics may be more accurate than pathology reads (10), it is conceivable that treating patients on the basis of the assay results would lead to better outcomes than treatment decisions on the basis of the “gold standard” biopsy interpretations.

Major strengths of the study by Park et al. (8) include the use of paired blood samples and kidney biopsies, a prevalence of subclinical rejection in their cohort similar to those observed by others, blinded biopsy reads by the same pathologist, and the inclusion of the independent validation cohort. Nonetheless, several caveats of the work need to be noted. In addition to the biases inherent to post hoc analyses of previously reported cohorts, the authors studied a carefully chosen subgroup of transplant biopsies. The selected biopsy samples showed either (1) outstanding transplant histology without evidence of rejection in patients with excellent kidney function or (2) Banff borderline or higher rejection without other abnormalities. The authors specifically excluded biopsies that contained other common pathologies found in kidney allografts—including infection, chronic tubular injury/fibrosis, calcineurin inhibitor toxicity, recurrent primary disease, and others—that could confound the diagnostic utility of the assays. How the two assays will behave in a more heterogeneous and unselected cohort remains unclear.

Overall, the data from Park et al. (8) serve as a proof of principle that integrating these two mechanistically distinct biomarker assays could allow for the better detection of subclinical acute rejection in kidney transplant recipients than either test alone. Determining whether these biomarkers are adequate to replace transplant biopsies for clinical decision making will require prospective, multicenter, biomarker-guided interventional trials using heterogeneous patient populations, in which outcomes can be attributed to assay-driven versus biopsy-driven treatment decisions. Two noninvasive biomarkers likely perform better than one for the diagnosis of subclinical rejection. However, without additional data, transplant physicians are unlikely to make routine treatment decisions solely on the basis of the results of PBMC gene expression profiling plus plasma dd-cfDNA measurements in clinical practice.

Disclosures

P.S. Heeger reports consultancy agreements with, receiving honoraria from, and serving as a scientific advisor or member of Mallinckrodt Pharmaceuticals. The remaining author has nothing to disclose.

Funding

This work was supported by National Institute of Allergy and Infectious Diseases grant U01 AI63594.