. IMPACT OF RACE AND ETHNICITY ON PROGNOSTIC MODELS OF OUTCOMES IN PRIMARY SCLEROSING CHOLANGITIS
Bowlus CL, Assis DN, Wu JL, Levy C, Goldberg DS, Forman L, Schlansky B, Lammert C, Prenner S, Reddy KR, Gordon SC, Ahn J, Zepeda JB, Silveira MG, Boyer JL, and Pollock BH. IMPACT OF RACE AND ETHNICITY ON PROGNOSTIC MODELS OF OUTCOMES IN PRIMARY SCLEROSING CHOLANGITIS. Hepatology 2020; 72:748A-749A.
Background: Primary sclerosing cholangitis (PSC) is a chronic cholestatic liver disease with a median time to liver transplant or death (LT/D) of approximately 20 years. The Mayo Risk Score (MRS), Amsterdam-Oxford Model (AOM), and UK-PSC scores predict LT/D in PSC but have not been validated in a racially and ethnically diverse population. We aimed to validate these models in a diverse cohort and determine if race and/or ethnicity affect their performance. Methods: Patients diagnosed with PSC, including small duct PSC and overlap with autoimmune hepatitis (PSC/AIH), alive without liver transplantation after 2008, were enrolled in the Consortium for Autoimmune Liver Diseases (CALiD) registry at 9 U.S. centers. MRS, AOM, UK-PSC (short term), and model for end-stage liver disease (MELD) were calculated from the earliest available data. Time-to-failure analysis was performed from the date of the earliest available score to either LT/D or hepatic decompensation (HD), defined as first variceal bleed, ascites, or encephalopathy. Accuracy of each model was summarized using Harrell’s C-index. The effect of race and ethnicity on each model for LT/D and HD was determined by Cox regression. Results: A total of 335 patients with a median follow up 6.4 y were analyzed including 39 (11.6%) Black(B) and 37 (11.0%) Hispanic(H) patients. Characteristics did not differ between non-Hispanic White (NHW), B, and H patients [age of diagnosis (median 40.4, IQR 24.6; p=0.52), male sex (60.3%; p=0.84), PSC type (89.0% large duct, 3.6% small duct, 7.7% PSC/AIH; p=0.57), and IBD (72.2%; p=0.14)]. All models predicted both LT/D and HD (Table). However, MELD score demonstrated poor discrimination for both LT/D and HD (C-index<0.7) and MELD prediction of LT/D was significantly lower than MRS (p = 0.013). MRS, AOM, and UK-PSC had good discrimination for LT/D (C-index>0.7) and did not significantly differ. C-index for HD was generally lower than for LT/D and did not significantly differ between the scores. Notably, B race compared to NHW was independently associated with an increased risk of HD in three of four models. Conclusion: Current PSC-specific models performed similarly and better than MELD in an ethnically and racially diverse patient population. Neither race nor ethnicity impacted any of the models’ performance for LT/D, but B race was an independent predictor of HD in some models. Race and ethnicity should be incorporated into new models as they are developed for clinical trials as surrogate endpoints.