Using computer-assisted morphometrics of 5-year biopsies to identify biomarkers of late renal allograft loss

Aleksandar Denic, Martha C. Morales, Walter D. Park, Byron H. Smith, Walter K Kremers, Mariam P Alexander, Fernando G Cosio, Andrew D Rule, Mark D Stegall

Research output: Contribution to journalArticle

Abstract

The current Banff scoring system was not developed to predict graft loss and may not be ideal for use in clinical trials aimed at improving allograft survival. We hypothesized that scoring histologic features of digitized renal allograft biopsies using a continuous, more objective, computer-assisted morphometric (CAM) system might be more predictive of graft loss. We performed a nested case-control study in kidney transplant recipients with a surveillance biopsy obtained 5 years after transplantation. Patients that developed death-censored graft loss (n = 67) were 2:1 matched on age, gender, and follow-up time to controls with surviving grafts (n = 134). The risk of graft loss was compared between CAM-based models vs a model based on Banff scores. Both Banff and CAM identified chronic lesions associated with graft loss (chronic glomerulopathy, arteriolar hyalinosis, and mesangial expansion). However, the CAM-based models predicted graft loss better than the Banff-based model, both overall (c-statistic 0.754 vs 0.705, P <.001), and in biopsies without chronic glomerulopathy (c-statistic 0.738 vs 0.661, P <.001) where it identified more features predictive of graft loss (% luminal stenosis and % mesangial expansion). Using 5-year renal allograft surveillance biopsies, CAM-based models predict graft loss better than Banff models and might be developed into biomarkers for future clinical trials.

Original languageEnglish (US)
JournalAmerican Journal of Transplantation
DOIs
StatePublished - Jan 1 2019

Fingerprint

Allografts
Biomarkers
Transplants
Kidney
Biopsy
Clinical Trials
Case-Control Studies
Pathologic Constriction
Transplantation

Keywords

  • biomarker
  • biopsy
  • clinical research/practice
  • kidney failure/injury
  • kidney transplantation/nephrology

ASJC Scopus subject areas

  • Immunology and Allergy
  • Transplantation
  • Pharmacology (medical)

Cite this

@article{d9024beb551048679f23758bde3a8bed,
title = "Using computer-assisted morphometrics of 5-year biopsies to identify biomarkers of late renal allograft loss",
abstract = "The current Banff scoring system was not developed to predict graft loss and may not be ideal for use in clinical trials aimed at improving allograft survival. We hypothesized that scoring histologic features of digitized renal allograft biopsies using a continuous, more objective, computer-assisted morphometric (CAM) system might be more predictive of graft loss. We performed a nested case-control study in kidney transplant recipients with a surveillance biopsy obtained 5 years after transplantation. Patients that developed death-censored graft loss (n = 67) were 2:1 matched on age, gender, and follow-up time to controls with surviving grafts (n = 134). The risk of graft loss was compared between CAM-based models vs a model based on Banff scores. Both Banff and CAM identified chronic lesions associated with graft loss (chronic glomerulopathy, arteriolar hyalinosis, and mesangial expansion). However, the CAM-based models predicted graft loss better than the Banff-based model, both overall (c-statistic 0.754 vs 0.705, P <.001), and in biopsies without chronic glomerulopathy (c-statistic 0.738 vs 0.661, P <.001) where it identified more features predictive of graft loss ({\%} luminal stenosis and {\%} mesangial expansion). Using 5-year renal allograft surveillance biopsies, CAM-based models predict graft loss better than Banff models and might be developed into biomarkers for future clinical trials.",
keywords = "biomarker, biopsy, clinical research/practice, kidney failure/injury, kidney transplantation/nephrology",
author = "Aleksandar Denic and Morales, {Martha C.} and Park, {Walter D.} and Smith, {Byron H.} and Kremers, {Walter K} and Alexander, {Mariam P} and Cosio, {Fernando G} and Rule, {Andrew D} and Stegall, {Mark D}",
year = "2019",
month = "1",
day = "1",
doi = "10.1111/ajt.15380",
language = "English (US)",
journal = "American Journal of Transplantation",
issn = "1600-6135",
publisher = "Wiley-Blackwell",

}

TY - JOUR

T1 - Using computer-assisted morphometrics of 5-year biopsies to identify biomarkers of late renal allograft loss

AU - Denic, Aleksandar

AU - Morales, Martha C.

AU - Park, Walter D.

AU - Smith, Byron H.

AU - Kremers, Walter K

AU - Alexander, Mariam P

AU - Cosio, Fernando G

AU - Rule, Andrew D

AU - Stegall, Mark D

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The current Banff scoring system was not developed to predict graft loss and may not be ideal for use in clinical trials aimed at improving allograft survival. We hypothesized that scoring histologic features of digitized renal allograft biopsies using a continuous, more objective, computer-assisted morphometric (CAM) system might be more predictive of graft loss. We performed a nested case-control study in kidney transplant recipients with a surveillance biopsy obtained 5 years after transplantation. Patients that developed death-censored graft loss (n = 67) were 2:1 matched on age, gender, and follow-up time to controls with surviving grafts (n = 134). The risk of graft loss was compared between CAM-based models vs a model based on Banff scores. Both Banff and CAM identified chronic lesions associated with graft loss (chronic glomerulopathy, arteriolar hyalinosis, and mesangial expansion). However, the CAM-based models predicted graft loss better than the Banff-based model, both overall (c-statistic 0.754 vs 0.705, P <.001), and in biopsies without chronic glomerulopathy (c-statistic 0.738 vs 0.661, P <.001) where it identified more features predictive of graft loss (% luminal stenosis and % mesangial expansion). Using 5-year renal allograft surveillance biopsies, CAM-based models predict graft loss better than Banff models and might be developed into biomarkers for future clinical trials.

AB - The current Banff scoring system was not developed to predict graft loss and may not be ideal for use in clinical trials aimed at improving allograft survival. We hypothesized that scoring histologic features of digitized renal allograft biopsies using a continuous, more objective, computer-assisted morphometric (CAM) system might be more predictive of graft loss. We performed a nested case-control study in kidney transplant recipients with a surveillance biopsy obtained 5 years after transplantation. Patients that developed death-censored graft loss (n = 67) were 2:1 matched on age, gender, and follow-up time to controls with surviving grafts (n = 134). The risk of graft loss was compared between CAM-based models vs a model based on Banff scores. Both Banff and CAM identified chronic lesions associated with graft loss (chronic glomerulopathy, arteriolar hyalinosis, and mesangial expansion). However, the CAM-based models predicted graft loss better than the Banff-based model, both overall (c-statistic 0.754 vs 0.705, P <.001), and in biopsies without chronic glomerulopathy (c-statistic 0.738 vs 0.661, P <.001) where it identified more features predictive of graft loss (% luminal stenosis and % mesangial expansion). Using 5-year renal allograft surveillance biopsies, CAM-based models predict graft loss better than Banff models and might be developed into biomarkers for future clinical trials.

KW - biomarker

KW - biopsy

KW - clinical research/practice

KW - kidney failure/injury

KW - kidney transplantation/nephrology

UR - http://www.scopus.com/inward/record.url?scp=85065336975&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85065336975&partnerID=8YFLogxK

U2 - 10.1111/ajt.15380

DO - 10.1111/ajt.15380

M3 - Article

JO - American Journal of Transplantation

JF - American Journal of Transplantation

SN - 1600-6135

ER -