Daniel Liu *;
University of Wisconsin School of Medicine and Public Health
John Garrett;
University of Wisconsin School of Medicine and Public Health
Ryan Zea;
University of Wisconsin School of Medicine and Public Health
Adam Kuchnia;
University of Wisconsin School of Medicine and Public Health
David Ji;
University of Wisconsin School of Medicine and Public Health
Ronald Summers;
NIH Clinical Center
Perry Pickhardt;
University of Wisconsin School of Medicine and Public Health
Objective:
Body composition measures from preoperative CT can provide prognostic value in kidney transplant surgery. The purpose of this study was to determine the performance of fully automated CT-based body composition metrics to predict postoperative outcomes in kidney transplant recipients.
Materials and Methods:
In this retrospective study, validated and fully automated artificial intelligence-based body composition tools were applied to pretransplant abdominal CT scans of kidney transplant recipients from 2005 - 2021. Algorithms for abdominal aortic calcification, muscle attenuation (at L3), and bone mineral density (at L1) were applied to all scans. Age- and sex-corrected univariate hazard ratios (HRs) comparing the worst quartile to the remaining quartiles and area under the receiver operating characteristic (AUROC) curve analyses were performed.
Results:
There were 596 kidney transplant recipients (mean age 61.2 ± 12.8; 376 men/219 women) were included in this study. Mean clinical follow-up interval after transplant was 7.8 years with 183 deaths (5.7 ± 3.9 years posttransplant) and 77 graft failures (mean interval, 4.7 ± 4.1years posttransplant). HRs (with 95% CI) predicting mortality were 2.53 (1.78 - 3.59) for abdominal aortic calcification, 1.80 (1.27 - 2.53) for muscle attenuation, and 1.62 (1.15 - 2.26) for BMD. Five-year mortality AUROC were 0.725, 0.653, and 0.638, respectively. HRs predicting graft failure were 2.23 (1.61 - 3.08) for abdominal aortic calcification, 1.66 (1.21 - 2.27) for muscle attenuation, and 1.46 (1.07 - 2.00) for BMD. Five-year graft failure AUCs were 0.654, 0.596, and 0.584, respectively.
Conclusion:
Automated preoperative CT measurements of abdominal aortic calcification, muscle attenuation, and BMD were predictive of graft failure and mortality in kidney transplant recipients. CT imaging, whether performed specifically for preoperative transplant evaluation or for other indications, can provide additional risk stratification for transplantation.
