Association between longer duration of citrate accumulation and 90-day mortality of acute-on-chronic liver failure

Regional citrate anticoagulation (RCA) is an optional anticoagulant for plasma adsorption (PA) plus plasma exchange (PE) therapy in patients with acute-on-chronic liver failure (ACLF), but with risk of transient citrate accumulation due to plasma and citrate [1]. Regardless of the anticoagulants: heparin or citrate, some patients would suffer from longer duration of citrate accumulation (LDCA), defined as the presence of citrate accumulation 2 h after PA plus PE therapy with RCA [1, 2]. However, whether citrate accumulation itself would lead to poor prognosis remains uncertain.

We conducted a retrospective study based on medical records to assess the association between LDCA and prognosis of hepatitis B virus (HBV)-related ACLF. Methods and some data from this cohort have been published already [2]. We kept to follow-up these patients for another 90 days after acquiring further ethical approval and registered this study with ChiCTR-OON-17013631. HBV-ACLF was diagnosed according to COSSH ACLF criteria [3]. Citrate accumulation was defined as the ratio of total calcium (Ca_tot) to ionized calcium (Ca_ion), (Ca_tot/Ca_ion), over or equal to 2.5 (Ca_tot/Ca_ion ≥ 2.5) [1, 2]. Cox proportional hazards models were applied to evaluate the association of LDCA with outcome.

From January 2018 to December 2019, we reviewed the data of 258 patients who fulfilled the HBV-ACLF criteria and received PA plus PE therapy with RCA. LDCA patients (N = 76) were more often female and older and had worse severity of disease condition than non-LDCA patients (N = 182) (Table 1). There was no significant difference in indicators, such as intracorporeal and extracorporeal Ca_tot and Ca_ion, representing patients receiving similar RCA during and after the first session of PA plus PE therapy with RCA.

Table 1 Characteristics of ACLF patients with or without LDCA

The 90-day mortality of LDCA patients was much higher than that of non-LDCA patients (63.2% vs. 32.4%, log-rank p < 0.001). Compared with non-LDCA patients, LDCA patients had much higher 90-day mortality risk (crude hazard ratio (HR) (95% confidence interval (CI)), 2.62 (1.79–3.84)) (Table 2). However, no significant differences in 90-day mortality risk were observed with the Cox proportional hazards models established with LDCA, age, gender, liver cirrhosis, HBV DNA, other co-existing liver diseases, comorbidities, and disease severity (Model 1, COSSH ACLF score; Model 2, CLIF-C ACLF score; Model 3, AARC ACLF score; Model 4, MELD score): Model 1 adjusted HR (95% CI), 1.07 (0.66–1.73); Model 2, 1.49 (0.95–2.36); Model 3, 1.41 (0.90–2.22); Model 4, 1.05 (0.65–1.72) (Table 2). Similarly, no significant differences in 90-day mortality risk were observed with similar Cox models established with citrate level indicators (Model 5, Ca_tot/Ca_ion ≥ 2.25; Model 6, Ca_tot/Ca_ion; Model 7, anion gap), disease severity (COSSH ACLF score), and the others mentioned above: Model 5, 1.28 (0.78–2.08); Model 6, 1.56 (0.74–3.27); Model 7, 1.06 (0.97–1.16). The disease severity was the independent risk factor of 90-day mortality (Model 1–7, all adjusted HR > 1, all p < 0.001).

Table 2 LDCA and other factors associated with risk of 90-day mortality in ACLF patients

Our study proved that ACLF patients with LDCA would suffer higher 90-day mortality. This finding was in accordance with the results in critically ill patients undergoing continuous renal replacement therapy with RCA [4]. However, no significant differences in 90-day mortality risk were found in ACLF patients with or without LDCA. As RCA brings no alteration of pro- and anti-coagulation function and ACLF patients have re-balanced but fragile coagulation function [1, 5], our new results would support the use of RCA with caution in ACLF patients. Adequate training, experienced operation, and well-developed safety protocols would further expand indications of RCA [6].

Our study for the first time assessed the association between LDCA and prognosis in ACLF patients treated with PA plus PE therapy with RCA. There were limitations: monocentric retrospective design, only HBV-ALCF cases, and applying Ca_tot/Ca_ion instead of directly measuring plasma citrate concentration to reflect citrate accumulation.