Abstract

Background:

Accurate estimation of glomerular filtration rate (GFR) is important for decisions regarding dual organ transplantation and patient management on the waitlist. Currently, a novel race free equation (CKD-EPI AS, Inker et al. NEJM 2021) is the reference standard for estimating GFR across the US. However, its performance in patients with cirrhosis was never validated. Further, performance of CKD EPI AS 2021 may be suboptimal in cirrhosis patients with low GFR, ascites and frail patients (AASLD 2022). We have shown that liver specific equations (GFR assessment in liver disease, GRAIL, Asrani et al. Hepatology 2019) have better performance as compared to other GFR equations. We sought to develop and validate an updated GRAIL without race (GRAIL_NR).

Methods:

We examined all cirrhosis patients with protocol measured GFR between 1985-2015 using iothalamate clearance. We estimated GFR using novel non-race equations: CKD-EPI 2021, CKD-EPI refit to liver population, as well as GRAIL_NR. Model was developed using cumulative probability models, associations between variables and mGFR were examined via linear association and restricted cubic spline and validated using split sample. The final components were age, sex, albumin, creatinine, and BUN. Addition of interactions or other variables (BSA, serum sodium) did not appreciably improve performance. We compared Concordance Correlation Coefficient, bias, precision, % agreement CKD stages and measurements within 30% of mGFR (p30). We further examined performance in relevant subsets: ascites, sex and low GFR (GFR<40 vs >40 ml/min/1.73m²).

Results:

Updated GRAIL_NR was more precise and had higher concordance (0.82, 0.79-0.85) as compared to CKD-EPI 2021 (0.78, 0.74-0.81) as well as refit CKD-EPI AS_Liver (0.77, 0.73-0.81). As compared to CKD-EPI 2021, updated GRAIL_NR had lower bias (-0.8 vs. +1.56, mGFR-eGFR), higher percent agreement for CKD staging (95.4% vs. 93.3%), higher agreement for low GFR vs high GFR (cutoff 40, 95.4% vs. 93.7%) and higher measurements within 30% of mGFR (78.7% vs. 74.3%). We examined differences in relevant subsets. Percent discordance between estimated CKD stage and actual CKD stage was lower with GRAIL_NR in patients with ascites and females. (Figure) This was especially pronounced when limited to patients with GFR<40ml/min/1.73m²

Conclusion:

A race-free eGFR equation developed and validated in patients with may help guide decision making in patients with decompensated cirrhosis listed for transplantation, especially in subsets where prediction of renal function is suboptimal.