Researchers propose algorithm to assess kidney transplant injury phenotypes, severity
Click Here to Manage Email Alerts
A data-driven algorithm, RejectClass, may provide a “holistic and quantitative” measurement of kidney transplant injury phenotypes and severity, according to investigators.
“Chronic lesions constitute an independent risk factor for late kidney graft failure. However, there is currently no validated data-driven system to realistically describe the chronic pathology of kidney transplants,” Thibaut Vaulet, PhD, from ESAT Stadius Center for Dynamical Systems in Belgium, and colleagues wrote.
To fill this knowledge gap, researchers trained and validated a semi-supervised consensus clustering algorithm, RejectClass, to cluster post-transplant kidney biopsies based on their chronic Banff lesions score.
The training cohort consisted of 3,549 kidney transplant biopsies from an observational cohort of 937 recipients at University Hospitals Leuven between March 2004 and February 2013.
Researchers reweighted the chronic histologic lesions based on their time-dependent relationship with graft failure and conducted consensus k-means clustering analysis. Additionally, researchers measured total chronicity as the sum of the weighted chronic lesions scores, scaled to the unit interval.
Researchers also validated their findings with an external cohort of 4,031 biopsies from 2,054 kidney transplant recipients.
Analyses revealed a total of four chronic clusters correlated with graft outcome, and the two clusters with the worst outcome were driven by interstitial fibrosis/tubular atrophy and by transplant glomerulopathy. Researchers noticed the chronic clusters overlapped with the existing Banff IFTA classification and were allocated independently of the acute lesions.
Overall, total chronicity significantly correlated with graft failure, regardless of the total activity scores.
“We demonstrated the complementarity of assessing total chronicity on top of the estimation of disease activity from acute lesion scores. This approach delivers a holistic and quantitative assessment of kidney transplant injury phenotypes and severity, as supplement to the current Banff classification. Separating the degree of (often nonspecific) chronicity from rejection activity can significantly reduce the complexity of the description of kidney transplant histology,” Vaulet and colleagues wrote. “Further validation of the system, eg, via the web application, is necessary to gain more widespread adoption by the transplant community, on the way to the eventual integration of these algorithms into accessible clinical decision support systems.”
Collapse
Read more about
Click Here to Manage Email Alerts