Correlation between donor-derived cell-free DNA and tissue gene expression in heart transplant patients undergoing...

The introduction of donor-derived cell-free DNA (dd-cfDNA) and the Molecular Microscope Diagnostic System (MMDx) is changing how we diagnose rejection following heart transplantation (HT). This study aims to assess the accuracy of dd-cfDNA in detecting rejection as identified by MMDx and histology, with a focus on determining an optimal dd-cfDNA threshold to improve diagnostic performance.

Single-center prospective study of HT recipients undergoing for-cause biopsies with paired MMDx results and dd-cfDNA levels. We employed a receiver operator curve to evaluate the performance of dd-cfDNA levels to detect rejection assessed by both histology and MMDx. We also assessed the correlation between dd-cfDNA levels and MMDx rejection scores. A mixed-effects model was applied to account for repeated samples when appropriate.

247 for-cause biopsies were identified with a median of 21 months from HT and a median of 11 days between dd-cfDNA and biopsy. 56.7% of the samples had dd-cfDNA levels ≥0.20%. MMDx identified rejection in 27.1% of biopsies, compared to 7.7% identified by histology. Elevated dd-cfDNA levels were associated with a 4-fold increase in rejection rates by MMDx, mainly driven by a 5-fold increase in antibody-mediated rejection detection when compared to histology. Dd-cfDNA demonstrated superior performance in predicting rejection by MMDx (area under the curve [AUC] of 0.77; optimal cutoff dd-cfDNA value 0.30%). When incorporating a mixed-effects model, the predictive performance improved further (AUC of 0.89; optimal cutoff value 0.26%). In contrast, prediction based on histology resulted in a lower AUC of 0.64.

In a for-cause biopsy population, elevated dd-cfDNA levels were more predictive of rejection on MMDx than on histology, suggesting that molecular techniques may detect rejection at earlier stages than traditional histological methods. A dd-cfDNA cutoff of 0.26% provided the highest predictive accuracy for rejection by MMDx when applied within a mixed-effects model for repeated measures.