The Trials and Tribulations of a) developing and b) commercializing a non-invasive biomarker for subAR
a) Development
The TruGraf® Blood Gene Expression Test (Transplant Genomics, Inc, Mansfield, MA) is a microarray-based assay that analyzes gene expression profiles (GEP) in the peripheral blood. Our initial strategy was to develop a “rule in”’ test, whereby a positive test would be highly predictive of a positive biopsy (subAR). We used a locked support vector machine (SVM) based classifier with a bootstrap to prevent over-fitting of the discovery set for internal validation as the bio-informatics approach (20). We found two interesting observations: first, at different thresholds, we traded PPV for NPV to the point that a “rule in” test was not possible using this approach. We then switched to Random Forest (RF) as the bio-informatics approach (21) and used a different threshold, but again it was evident that the intended use of the biomarker would need to change. Because the performance metrics were better with RF, we proceeded to use RF but picked thresholds more favorable for a ”rule out” test (21). The product was a GEP classifier that associates with either a normal protocol kidney biopsy (Transplant eXcellence– TX) or the absence of a normal biopsy (not-TX) in patients with stable renal function. All aspects of discovery and external validation of the TruGraf test were performed on blood samples paired with biopsies from prevalent cohorts. For the purpose of validation, the model derived from pre-selected bio-informatics and the threshold used to test performance on the discovery cohort were locked. These data led us to use this approach for external validation in an early access program (EAP) for patients (22). The external clinical validation from seven EAP transplant centers defined the key clinical performance parameters for this assay, as summarized in Table 1 and Figure 1. In this study, the high negative predictive value (NPV) of TruGraf was demonstrated in clinical use, making it a strong rule-out test. Over 90% of stable patients who received a TX results were confirmed to have an immune quiescent phenotype, meaning that a physician can have a high degree of confidence that a patient who tests as TX does not harbor silent subclinical rejection. Importantly this study also found that up to 65% of surveillance biopsies could be avoided in the cohort tested. Unpublished data involving analysis of an additional 129 biopsy-confirmed blood samples provided by Northwestern University (originally used for the CTOT-08 study) revealed identical performance metrics for TruGraf (NPV of 90%). A fourth publication described the impact of TruGraf results on physician decision making for clinical decisions (23). This study highlighted the high degree of confidence physicians place in the ability of TruGraf to provide valuable, added information that could lead to avoidance of unnecessary surveillance biopsies as summarized in Table 2.
As a result of these experiences, we changed the proposed COU from replacing surveillance biopsies for detecting subAR, to reducing the number of necessary biopsies in stable patients which should lead to many less invasive procedures (Table 1) as well as significantly less negative or unnecessary biopsies. The COU proposed in the recent approval from CMS states that “The TruGraf test is intended for use in kidney transplant recipients with stable renal function as an alternative to surveillance biopsies in facilities that utilize surveillance biopsies”. While primarily used to rule out subAR, it is expected that both centers that perform or do not perform surveillance biopsies can use the test to assess the need for a surveillance biopsy in stable patients (24).
Figure 2 illustrates a proposed approach for implementation of TruGraf into clinical care for kidney transplant recipients. For patients with stable renal function, a TruGraf result of “TX” identifies those who have a high likelihood of immune quiescence and a low likelihood of histologically defined rejection at the borderline level or higher. A result of “Not-TX” identifies those in whom silent rejection cannot be confidently ruled out, and thus carry a higher risk of immune activation and borderline or higher rejection. Patients with a “Not-TX” result might benefit from further evaluation and possibly a change in therapy. Early identification of these patients potentially allows better allocation of physician resources, and potential reversal of the process before permanent damage to the donated kidney occurs.
b) Pathway to Commercialization of TruGraf
Developed in 2011, the Molecular Diagnostic Services (MolDX) program is run by Palmetto GBA, a Centers for Medicare and Medicaid Services (CMS) Medicare administrative contractor. It performs the following functions:
CMS approved reimbursement for commercial TruGraf testing on November 25, 2019.