This study retrospectively enrolled patients diagnosed with ACLF according to the criteria of the Asian Pacific Association for the Study of the Liver (APASL) and treated at West China Hospital during 2018–2023. All patients undergone nutritional risk assessment by nurses upon admission and received a standardized medical management during hospitalization. Notably, according to international guidelines (12–14), nutritional support is recommended as standard supportive care for liver failure. In this study, patients who did not receive nutritional intervention (n = 2,896) were a highly selected subgroup, with enrollment criteria restricted to three scenarios: (1) Very mild ACLF (Child–Pugh A grade, no overt complications, and autonomous oral intake meeting ≥80% of daily energy requirements); (2) Terminal-stage ACLF (Child–Pugh C grade with multi-organ failure, where nutritional support was deemed clinically futile by the multidisciplinary team); (3) Patient/family refusal of nutritional intervention after full informed consent. To mitigate inherent selection bias, we adjusted for disease severity in multivariate logistic regression model. Additionally, subgroup analyses stratified by ACLF severity were performed to validate the robustness of associations between nutritional support and complications.

The Nutritional Risk Screening-2002 (NRS-2002) was used to screen for nutritional risk in patients with liver failure. The NRS-2002 screening tool, which has been validated in liver failure populations, is designed to identify individuals at nutritional risk (score ≥3) who may potentially benefit from nutrition intervention to optimize clinical outcomes (15, 16). We selected the NRS-2002 for nutritional risk screening based on its proven applicability in ACLF populations and advantages over liver-specific tools. First, the NRS-2002 can be completed within 24 h of admission without complex auxiliary examinations, adapting to the acute clinical scenario of ACLF (16). Second, unlike RFH-NPT (primarily validated in alcoholic cirrhosis) and LDUST (with limited ACLF data), the NRS-2002 has been validated in HBV-related, alcoholic, and cryptogenic ACLF subgroups (12, 17). Third, it effectively predicts nutritional support responsiveness and clinical outcomes in ACLF, as demonstrated by RCT-derived evidence (16) and ACLF-specific studies (11, 17). These data support its suitability for our heterogeneous ACLF cohort. Malnutrition was defined based on the NRS-2002 criteria (score ≥3) combined with objective indicators: serum albumin <35 g/L, unintentional weight loss >5% in 3 months, or reduced food intake <50% of recommended daily energy for >3 days. Nutritional intervention was initiated if patients met either: (1) NRS-2002 score ≥3 (nutritional risk) plus one objective malnutrition indicator; or (2) NRS-2002 score 1–2 (low nutritional risk) but with persistent food intake <60% of target for >5 days. Patients with NRS-2002 score <3 and stable food intake (≥80% of target) were not recommended for intervention, except for those with physician-documented high risk of malnutrition progression. For patients receiving nutritional support, dietitians customize daily personalized nutritional support plans based on their medical conditions to ensure that their daily energy requirements are met. The personalized nutritional support plan was defined as a tailored strategy integrating multi-dimensional assessments (clinical severity, metabolic status, gastrointestinal tolerance, nutrient absorption) to optimize nutrient delivery, with real-time adjustments based on disease progression. Dietitians conducted baseline and weekly assessments of metabolic markers (serum albumin, prealbumin, glucose, triglycerides) and nutrient levels (fat-soluble vitamins, zinc, selenium) to individualize energy and nutrient targets.

Dietitians conducted detailed nutritional assessments for ACLF patients, with a focus on disease severity [ACLF grade 1/2/3 per APASL criteria (2)] and organ failure status (hepatic, renal, coagulation, encephalopathy, circulatory). Energy requirements were set at 25–30 kcal/kg/day for ACLF Grade 1/2 and 20–25 kcal/kg/day for Grade 3, with a 10–15% increase for patients with hypermetabolism (12, 18). Macronutrient ratios were customized: protein intake ranged from 1.0–1.5 g/kg/day (35–40% as BCAAs for hepatic encephalopathy), carbohydrates accounted for 50–60% of total energy (slow-release formulations preferred), and fat (25–35% of total energy) included medium-chain triglycerides (MCTs) to enhance absorption (12). Special nutrients (glutamine, vitamin D, zinc, selenium) were supplemented based on deficiency status (18, 19). For ACLF grade 1 patients, oral diet was prioritized, with oral nutritional supplements (ONS) containing branched-chain amino acids (BCAAs, 0.8–1.0 g/kg/day) added if oral intake met <60% of basal metabolic rate (BMR). For ACLF grade 2 patients, enteral nutrition (EN) via nasogastric tube was initiated within 48 h of admission, with a stepwise increase in BCAA dosage (1.0–1.2 g/kg/day) and routine supplementation of fat-soluble vitamins (A/D/E/K) and trace elements (zinc, selenium) to address malabsorption caused by hepatic dysfunction. For ACLF grade 3 patients, parenteral nutrition (PN) was combined with EN (if gastrointestinal tolerance allowed) to achieve a protein intake of 1.2–1.5 g/kg/day; glucose-lipid ratio was adjusted to 60:40 to reduce hepatic glucose overload, and glutamine (0.3 g/kg/day) was added to maintain intestinal barrier integrity and inhibit bacterial translocation. Daily gastrointestinal tolerance monitoring (gastric residual volume, stool frequency, symptoms) guided adjustments to infusion rate or composition. EN infusion was reduced by 20% for gastric residual volume >500 mL/6 h, and carbohydrate proportion was adjusted for hyperglycemia (18).

BMR was measured using the InBody 770 device (Biospace, Seoul, South Korea), which employs bioelectrical impedance analysis (BIA) with the following standard conditions: morning fasting state (≥8 h), no strenuous exercise in the previous 24 h, no edema or ascites (or ascites volume <1,000 mL, confirmed by abdominal ultrasound), and measurement performed in a temperature-controlled room (22–25 °C). BMR was calculated via the device’s built-in formula for Asian adults: BMR (kcal/day) = 88.362 + (13.397 × weight/kg) + (4.799 × height/cm) − (5.677 × age/years) for males; BMR = 447.593 + (9.247 × weight/kg) + (3.098 × height/cm) − (4.330 × age/years) for females. BMR was set at 25–30 kcal/kg/day for ACLF grade 1/2 and 20–25 kcal/kg/day for grade 3 (to avoid excessive metabolic burden) with daily energy requirements. Dietitians performed daily follow-ups to monitor gastrointestinal tolerance and adjust nutrient delivery: for patients with hepatic encephalopathy, BCAA proportions were increased to 35–40% of total amino acids to improve ammonia metabolism (12); for those with renal failure, non-essential amino acids were reduced to prevent uremic toxin accumulation. All nutritional formulations avoided excess sodium (<2 g/day) to mitigate ascites progression, consistent with ACLF-specific management principles (14).

For continuous variables, when the data followed a normal distribution, they were presented as the mean and standard deviation (SD). In the case of non-normally distributed data, the median and interquartile range (IQR) were used for presentation. The Lilliefors test was employed to conduct normality testing. Categorical variables were described by presenting their percentages to show the distribution. To compare differences among groups, for continuous variables, the Kruskal–Wallis test was utilized, and for categorical variables, the chi-square test was applied. The logistic regression model was used to calculate the odds ratios (OR) and 95% confidence intervals (CI) to explore the relationship between nutritional support and the risk of clinical complications. Meanwhile, subgroups analyses were performed according to age, gender, body mass index (BMI), diabetes, hypertension, smoking, drinking, Child–Pugh score and cirrhosis status. To further investigate any possible correlations between these subgroups and complication risk, interaction studies were also performed. Cox proportional hazards models were used to assess the association between nutritional support and all-cause mortality. All statistical analyses were carried out using R software together with Zstats v1.0.¹ A two-sided p-value less than 0.05 was defined as statistically significant.

A total of 6,097 ACLF patients were included in this study, with a flow diagram shown in Figure 1. The median age was 53 years [interquartile range (IQR) 43, 64], and 69.2% of them were male. The proportion of patients with cirrhosis was 19.6% (1197). 555 (9.1%) were diagnosed with ascites, 733 (12.0%) with spontaneous bacterial peritonitis (SBP), 792 (13.0%) with infection, 191 (3.1%) with hepatorenal syndrome, 346 (5.7%) with hepatic encephalopathy (HE), 577 (9.46%) with coagulation disorder, 290 (4.8%) with variceal bleeding. Infection was defined as systemic bacterial or viral infections confirmed by positive microbiological cultures or clinical evidence. SBP was classified as a separate, ACLF-specific complication, defined as bacterial infection of ascitic fluid. Among all patients, 171 (2.8%) died and 3,201 (52.5%) patients were at nutritional risk. The results showed that patients with nutritional risk had a higher median age and a lower BMI (p < 0.001). Levels of aspartate transaminase (AST), total bilirubin (TBIL), and international normalized ratio (INR) were significantly higher in the nutritional risk group, while albumin levels were lower (p < 0.001). In terms of renal function, the nutritional risk group had higher creatinine levels and slightly lower estimated glomerular filtration rate (eGFR) (p < 0.001). Moreover, patients with nutritional risk had lower platelet and hemoglobin levels (p < 0.001). Regarding disease severity, patients with nutritional risk had higher Child–Pugh scores and longer hospital stays (p < 0.001), and significantly higher incidences of complications such as hepatorenal syndrome, HE, infection, and coagulation disorders (p < 0.01). Mortality was also higher in the nutritional risk group (p < 0.001) (Table 1).

Next, we analyzed the relationship between nutritional intervention and complications and all-cause mortality in ACLF patients with or nutritional risk using logistic regression.

After adjusting for age, gender, height, weight, smoking, drinking, hypertension, diabetes, and Child–Pugh score, we found that patients with nutritional risk receiving nutritional intervention were associated with lower frequency of ascites, SBP, and infection, with OR and 95% CI of 0.50 (0.31–0.80), 0.34 (0.24–0.49), and 0.70 (0.52–0.94), respectively. However, nutritional intervention was not associated with lower frequency of hepatorenal syndrome, hepatic encephalopathy (HE), coagulation abnormalities, or variceal bleeding risk with OR and 95% CI of 0.66 (0.38–1.13), 0.69 (0.43–1.11), 0.81 (0.57–1.13), and 0.82 (0.47–1.43) respectively (Figure 2). Cox regression was used to analyze the impact of nutritional intervention on all-cause mortality in patients with ACLF. After adjusting for age, gender, height, weight, smoking, drinking, hypertension, diabetes, and Child–Pugh score, we found that nutritional intervention had no association with all-cause mortality, with a hazard ratio (HR) and 95% CI of 1.00 (0.61–1.65) (Figure 2).

We further analyzed whether nutritional intervention was associated with lower frequency of clinical complications and all-cause mortality in patients without nutritional risk. We found that nutritional intervention was associated with lower frequency of ascites, SBP, and coagulation disorders, with OR and 95% CI of 0.26 (0.12–0.55), 0.30 (0.17–0.54), and 0.35 (0.17–0.72), respectively. However, it was not associated with lower frequency of infection, hepatorenal syndrome, HE, or variceal bleeding with OR and 95% CI of 0.63 (0.39–1.04), 0.42 (0.10–1.87), 0.71 (0.28–1.78), and 0.79 (0.39–1.59), respectively (Figure 2). Cox proportional hazards model showed that nutritional intervention had no significant effect on reducing all-cause mortality, with HR and 95% CI of 1.38 (0.51–3.75) (Figure 2).

Moreover, we divided ACLF patients into four groups based on nutritional risk and nutritional intervention: nutritional risk without intervention, no nutritional risk without intervention, no nutritional risk with intervention, and nutritional risk with intervention. After multivariate regression analysis, patients with no nutritional risk with intervention was associated with lower frequency of ascites, SBP, and infection compared to those with nutritional risk without intervention, with OR and 95% CI being 0.34 (0.18–0.67), 0.35 (0.19–0.63), and 0.44 (0.27–0.71), while patients with nutritional risk with intervention also was associated with lower frequency of these outcomes compared to those with nutritional risk without intervention, with corresponding OR and 95% CI of 0.59 (0.41–0.84), 0.35 (0.25–0.50), and 0.71 (0.54–0.94). Additionally, patients with no nutritional risk without intervention was associated with lower frequency of infection than those with nutritional risk without intervention [0.65 (0.50–0.84)]. Moreover, after multivariate regression analysis, the group with no nutritional risk without intervention was associated with lower frequency of hepatorenal syndrome, HE, and coagulation disorders compared to those with nutritional risk without intervention, with OR and 95% CI of 0.44 (0.25–0.76), 0.51 (0.34–0.76), and 0.67 (0.49–0.90), respectively; meanwhile, the group with no nutritional risk with intervention also was associated with lower frequency of these complications compared to those with nutritional risk without intervention, with corresponding OR and 95% CI of 0.22 (0.05–0.93), 0.33 (0.14–0.78), and 0.27 (0.13–0.54) (Figure 3).

Among these intervened patients with nutritional risk, subgroup analysis by administration route confirmed consistent efficacy across all three modalities. EN, PN, and EN + PN all reduced the risks of ascites (EN: OR = 0.52, 95% CI: 0.33–0.82; PN: OR = 0.47, 95% CI: 0.28–0.79; EN + PN: OR = 0.49, 95% CI: 0.30–0.81), SBP (EN: OR = 0.36, 95% CI: 0.25–0.52; PN: OR = 0.31, 95% CI: 0.20–0.48; EN + PN: OR = 0.33, 95% CI: 0.19–0.58), and infection (EN: OR = 0.73, 95% CI: 0.54–0.99; PN: OR = 0.68, 95% CI: 0.47–0.98; EN + PN: OR = 0.70, 95% CI: 0.48–1.02), with consistent but non-significant effect sizes for hepatorenal syndrome, HE, and gastrointestinal bleeding.

Among the patients without nutritional risk who received intervention, subgroup analysis by administration route demonstrated consistent efficacy across modalities. EN, PN, and EN + PN all reduced ascites (EN: OR = 0.27, 95% CI: 0.12–0.59; PN: OR = 0.24, 95% CI: 0.09–0.63; EN + PN: OR = 0.25, 95% CI: 0.08–0.79), SBP (EN: OR = 0.31, 95% CI: 0.16–0.59; PN: OR = 0.28, 95% CI: 0.12–0.65; EN + PN: OR = 0.30, 95% CI: 0.11–0.82), and coagulation disorder (EN: OR = 0.37, 95% CI: 0.17–0.80; PN: OR = 0.32, 95% CI: 0.14–0.73; EN + PN: OR = 0.34, 95% CI: 0.12–0.96), with consistent direction of effect (though non-significant) for infection, hepatorenal syndrome, HE, and gastrointestinal bleeding across all three administration routes.

Moreover, we conducted subgroup analyses in patients with nutritional risk and found that the association between nutritional intervention and reduced risk of complications in patients with ACLF was not consistent in some subgroups. But the association of nutritional intervention and reduced risk of ascites and spontaneous bacterial peritonitis were robust in most subgroups (Table 2).

In ACLF patients without nutritional risk, we conducted subgroup analyses and found that nutritional intervention was associated with lower frequency of ascites, SBP, and disorders in most subgroup populations (Table 3).