Introduction

Sepsis is characterized by the dysregulated host response to infection, resulting in life-threatening acute organ dysfunction1. Nowadays, due to the high morbidity and high mortality, sepsis has become a major global health challenge and economic burden2. The liver plays a vital role in regulating immune defense during the development of sepsis, while it is also a susceptible target of sepsis-related injury3. In cirrhotic patients, characterized by systemic hemodynamic alterations, gut dysbiosis, bacterial translocation, and immune dysfunction, sepsis is a common complication and a primary reason for ICU admission, contributing to a higher mortality rate and organ dysfunction compared to those without cirrhosis4,5. In recent decades, cirrhotic patients with infections have faced significant epidemiological shifts. The widespread use of quinolones for infection prophylaxis has led to a rise in gram-positive bacterial infections, while the extensive use of third-generation cephalosporins has increased infections caused by enterococci6,7,8. What is more, there has been a notable increase in the prevalence of multidrug-resistant and extensively drug-resistant bacteria, which are more challenging to treat and pose higher risks of unsuccessful antibiotic treatment, sepsis and mortality9,10. Currently, the prevalence of bacterial infections among cirrhotic patients hospitalized for acute decompensation ranges from 25–46%10. A large retrospective study found that infections in cirrhotic patients were associated with a fourfold increase in mortality11. Additionally, the mortality rate from sepsis in cirrhotic patients is at least twice as high as in those without cirrhosis, and the ICU mortality rate for cirrhotic patients with severe septic shock can reach an alarming 65%5. Despite the well-established management protocols for sepsis in the general population, there is a significant knowledge gap in the population of cirrhotic patients12,13. Therefore, exploring potential predictive factors correlated with adverse clinical outcomes is crucial for this special high-risk population.

The anion gap (AG) serves as a well-established indicator for evaluating both the clinical acid-base balance and the severity of the condition. Elevated AG is commonly observed in critically ill patients and is often associated with an adverse clinical prognosis14. However, as the anion gap (AG) is significantly influenced by albumin levels, hypoalbuminemia can lead to a false-negative AG result, impacting the accurate judgment of results15. Therefore, albumin-corrected anion gap (ACAG), calculated by adjusting AG based on albumin levels, was utilized to more accurately reflect the presence of unmeasured anions. The liver plays a pivotal role in maintaining acid-base balance, and liver diseases could contribute to acid-base disorders through a variety of mechanisms16. Metabolic acidosis is common in septic patients with cirrhosis due to significant hemodynamic instability, insufficient tissue perfusion, metabolic abnormalities, and hepatic dysfunction17. Previous studies have found a correlation between elevated ACAG levels and increased mortality among patients with various illnesses, including sepsis. However, there is no research exploring the connection between ACAG levels and clinical progress among septic patients with cirrhosis. Therefore, our study aimed to investigate whether ACAG could help predict the clinical outcomes in this special sepsis patient population.

Methods

Database

The MIMIC-IV database is a publicly available database for medical research, and this database provides de-identified clinical data from critical care units at Beth Israel Deaconess Medical Center. After successfully completing the online course and examination, one of the authors gained authorized access to the database and was responsible for data extraction (certification number: 48410426). Since the database has already undergone anonymization, there is no need for additional ethical approval procedures18.

Study population

We extracted data on patients diagnosed with liver cirrhosis, utilizing diagnosis codes from the International Classification of Diseases, versions 9 and 10. Inclusion criteria included (1) cirrhotic patients diagnosed with sepsis according to Sepsis 3.0 criteria (Supplemental Method)19, (2) hospitalization for 24 h or more, (3) diagnosis of sepsis between 12 h before and 24 h following admission to the ICU, (4) serum anion gap and albumin were recorded during the first day of ICU admission, (5) only the data from the first ICU admission were included.

Data extraction

In this study, demographic data, etiology of cirrhosis, vital signs, laboratory parameters, illness severity scores, comorbidities, and life-sustaining interventions were obtained from the MIMIC-IV database using Structured Query Language. Vital signs, laboratory measurements, and life-sustaining treatments extracted were limited to the initial 24 h following admission to the intensive care unit. Demographic characteristics included age and sex. Vital signals included temperature, mean arterial blood pressure, heart rate and respiratory rate. Comorbidities included ascites, hepatorenal syndrome and malignant tumor. Laboratory parameters included white blood cell, hemoglobin, platelets, aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin, serum creatinine (SCr), anion gap, international normalized ratio (INR), albumin, blood urea nitrogen(BUN). Illness severity scores included the Model for End-Stage Liver Disease (MELD) and Sequential Organ Failure Score (SOFA). life-sustaining interventions included vasopressors and invasive mechanical ventilation. Albumin-corrected anion gap (ACAG) was calculated as 4.4 - albumin (g/dl) * 2.5 + AG20,21. The data for vital signs and laboratory parameters were obtained using standardized and calibrated medical equipment. In addition, diagnoses of diseases and complications were obtained using ICD-9 and ICD-10 codes. The code for data extraction is available from GitHub (https://github.com/MIT-LCP/mimic-iv).

Groups and clinical outcome

The X-tile software was utilized to establish the optimal ACAG cutoff value (ACAG = 19mmol/l). Following this threshold value, patients were stratified into two groups: normal ACAG group and high ACAG group. The primary clinical outcome was the 28-day mortality after ICU admission.

Statistical analysis

Continuous variables were reported as median and interquartile range (IQR) and compared using the Mann-Whitney U-test, whereas categorical variables were expressed as counts and percentages and compared using the Chi-square test. Kaplan-Meier survival analysis was used to assess the cumulative survival rates among groups categorized by levels of ACAG, and differences between the curves were compared using the log-rank tests. To assess whether ACAG could be an independent predictor for adverse clinical outcomes in septic patients with cirrhosis, Cox regression modeling was carried out. Model 1 was unadjusted, while Model 2 was adjusted for age, gender, and type of cirrhosis. In Model 3, covariates demonstrating significance at P < 0.05 in the univariate Cox analysis (supplementary Table 1), along with those selected based on clinical experience, were included. Further, to investigated the potential nonlinear association between ACAG and 28-day mortality, a multivariate restricted cubic spline regression model with four knots was performed. To assess and compare the predictive abilities of ACAG and other indicators for adverse outcomes, receiver operating characteristic (ROC) analyses were performed. Finally, stratified analysis was applied to assess the impact of ACAG on distinct subgroups, encompassing age, sex, type of cirrhosis, MELD score, The requirement for vasopressors and invasive mechanical ventilation. All the statistical analyses were performed by SPSS software (v26.0), X-tile (V.3.6.1) and R software (v 4.3.2), P value < 0.05 was considered to be statistically significant22.

Results

Population and baseline characteristics

In total, 1340 eligible patients were enrolled for further analysis (Fig. 1). In the study, 474 patients (35.37%) were female, and the average age was 58.74 ± 7.52 years. The non-survivor group had higher values of WBC, Scr, BUN, INR, ALT, AST, Toal bilirubin, AG, ACAG, SOFA score, MELD score and higher requirement of life-supportive therapy. However, the survivor group had higher albumin levels, Hb levels than the non-survivors. In addition, survivors were generally younger and exhibited a lower incidence of ascites, hepatorenal syndrome, and malignancy, compared to non-survivors (Table 1).

Fig. 1
figure 1

The fowchart of patient selection.

Full size image
Table 1 Baseline characteristics of the study populations.
Full size table

Kaplan-Meier survival curve analysis

During the 28-day follow-up period, 395 patients (29.5%) experienced mortality. K-M curves of ACAG for 28-day mortality (Fig. 2), indicated a notably higher mortality rate in the high ACAG group (log-rank test, χ^2 = 175.638, P < 0.001).

Fig. 2
figure 2

The fowchart of patient selection.

Full size image

Association between ACAG and all-cause mortality

To investigate the relationship between ACAG and adverse outcomes, Cox regression modeling were applied. In the unadjusted model, we observed a significant association between ACAG and 28-day mortality in septic patients with cirrhosis (HR:1.113, 95% CI 1.097–1.129, P < 0.001). Even in the fully adjusted model, ACAG persisted as an independent risk factor. When ACAG was categorized as a dichotomous variable, elevated ACAG also significantly correlated with a higher risk of 28-day mortality (Table 2). Moreover, in the multivariable restricted cubic spline model (Fig. 3), a linear association was observed between ACAG and the risk of 28-day mortality (P-nonlinear = 0.089, P-overall = 0.001).

Fig. 3
figure 3

Restricted cubic spline analysis of the association between ACAG and the risk of 28-day mortality in patients with sepsis and cirrhosis.

Full size image
Table 2 Cox proportional hazard ratios (HR) for all-cause mortality.
Full size table

ROC curve analysis

ROC curves were performed to evaluate the predictive value of albumin, AG, ACAG, SOFA and SOFA combined with ACAG for 28-day mortality in septic patients with cirrhosis (Fig. 4 ). Our study revealed that ACAG demonstrated a superior predictive ability for 28-day mortality compared to AG (Z = 4.024, p < 0.01). Furthermore, the integration of ACAG into the SOFA score, a widely utilized tool for prognostic assessment in sepsis, resulted to a better predictive capability when compared to the original score, in septic patients with cirrhosis. (Z = 3.558, p < 0.01).

Fig. 4
figure 4

ROC curves for the ACAG, AG, ALB, SOFA score, and ACAG + SOFA for predicting 28-day mortality in patients with cirrhosis and sepsis.

Full size image

Subgroup analysis

Subgroup analysis indicated a significant correlation between ACAG and worse prognosis in the majority of subgroups, including gender, age, MELD score, and the use of vasopressors and invasive mechanical ventilation (Fig. 5). Notably, the predictive significance of ACAG differed significantly based on the etiology of cirrhosis, showing a higher predictive value in patients with non-alcoholic cirrhosis (P for interaction = 0.014) (Fig. 5).

Fig. 5
figure 5

The subgroup analysis between ACAG and 28-day all-cause mortality.

Full size image

Discussion

To our knowledge, this study is the first investigation into the relationship between ACAG and prognostic implications among septic patients with cirrhosis. The results indicated a significant correlation between higher ACAG levels and 28-day mortality within this special high-risk population. Importantly, even with adjustments for potential confounders, ACAG remained strongly linked to all-cause 28-day mortality and demonstrated predictive value for adverse clinical prognosis. Therefore, our study suggested that ACAG could provide clinicians with valuable insights for guiding interventions in this special high-risk population.

 To our knowledge, this study is the first investigation into the relationship between ACAG and prognostic implications among septic patients with cirrhosis. The results indicated a significant correlation between higher ACAG levels and 28-day mortality within this special high-risk population. Importantly, even with adjustments for potential confounders, ACAG remained strongly linked to all-cause 28-day mortality and demonstrated predictive value for adverse clinical prognosis. Therefore, our study suggested that ACAG could provide clinicians with valuable insights for guiding interventions in this special high-risk population.

The acid-base balance is crucial for normal bodily functions, with severe acid-base disorders often indicating a poor prognosis. Alongside the kidneys and lungs, the liver plays a vital role in maintaining acid-base equilibrium through various regulatory mechanisms17. While liver diseases frequently result in various forms of acid-base disorders, research exploring their correlation with critically ill cirrhotic patients is limited23,24,25. Previous studies indicated that metabolic acidosis was common among critically ill cirrhotic patient in the ICU and was associated with increased mortality. However, the reason for ICU admission can influence the clinical progression in critically ill patients with cirrhosis26. In cirrhotic patients, sepsis is a common complication and is one of the primary reasons for ICU admission among these patients. Due to the immunosuppressive state and intricate organ functional changes, this population is more susceptible to developing multi-organ dysfunction and experiencing higher mortality rates27. Furthermore, previous studies have identified sepsis as a significant independent predictor of prognosis in this cohort of patients. Hence, in this study, we focused on cirrhotic patients with sepsis, a special high-risk group and investigated the association between metabolic acid load and the clinical outcomes among these population.

AG is the difference between the concentrations of unmeasured plasma anions and cations and is widely used to assess acid-base balance in critical ill patients28.

High AG is frequently associated with increased acid production and diminished excretion of anions. In septic patients with cirrhosis, hyperlactatemia is a common acid-base disorder. The elevated lactate levels reflect the severity of tissue hypoxia, circulatory dysfunction, metabolic abnormalities, and deterioration of liver function, are linked to unfavorable clinical outcomes16. In addition, acute kidney injury (AKI) is a common complication in septic patients with cirrhosis and is also indicative of poor prognosis29. The impairment of kidney function leads to the accumulation of various acids in the body, consequently contributing to the elevation of serum AG levels. Nevertheless, albuminate constitutes an essential element of the AG, and hypoalbuminemia results in decreased plasma albuminate levels. In such scenarios, a normal AG may indicate that the decrease in albuminate levels has obscured the presence of elevated plasma acids30,31. Given the high prevalence of hypoalbuminemia in cirrhosis patients with sepsis, utilizing the ACAG could offer a more precise evaluation of acid load.

Nowadays, research has found that high ACAG was associated with adverse prognosis in various disease such as acute kidney injury, cardiovascular disease, acute pancreatitis, and so on32,33,34,35. However, there is limited research exploring the correlation between ACAG and the clinical outcomes of septic patients. Hu et al. observed a significant association between ACAG and in-hospital mortality in septic patients, with ACAG demonstrating superior predictive accuracy compared to AG and albumin36. On the contrary, Zhu et al. reported a higher AUC for serum AG compared to ACAG, indicating a slightly better predictive advantage for AG37. However, in our study which focused on patients with cirrhosis, increased ACAG was related to the increased risk of 28-day mortality. Furthermore, compared to AG, ACAG demonstrated superior predictive value for adverse clinical outcome within our special patient population. Nevertheless, given the limitations in both sensitivity and specificity of a single predictor for sepsis prognosis, scholars recommend utilizing a blend of clinical predictors to effectively detect potential adverse outcomes. In our study, we found ACAG combination with SOFA score improved the predictive value of SOFA. Hence, we proposed that the ACAG could serve as a supplementary measure in conjunction with other clinical data to enhance the precision of predicting adverse outcomes and to effectively stratify their risk in septic patients with cirrhosis.

In addition, we observed a significant interaction between the etiology of cirrhosis and ACAG, with a higher predictive value of ACAG in patients with non-alcohol-associated cirrhosis compared to those with alcohol-associated cirrhosis. This suggests that the predictive performance of ACAG varies notably depending on the underlying cause of cirrhosis, with alcohol abuse possibly playing a significant role in this variation. Chronic alcohol consumption induces significant metabolic changes, notably the increased production of ketone bodies, which are often an overlooked byproduct of alcohol metabolism but can have serious consequences38. A recent study demonstrated that ethanol treatment could upregulate proteins associated with ketogenesis, leading to a significant increase in β-hydroxybutyric acid, the primary circulating ketone body, following chronic ethanol exposure39. Additionally, alcoholic cirrhosis is frequently associated with nutritional deficiencies, particularly thiamine deficiency, which can contribute to lactic acidosis and metabolic imbalances40. These factors lead to an elevation in the ACAG, further complicating its interpretation in septic patients with alcoholic cirrhosis and potentially reducing its predictive accuracy. Previous study indicated that a higher infection risk significantly contributed to increased mortality in alcoholic cirrhosis, compared to non-alcoholic cirrhosis41. Recently, Choi et al. found that patients with alcoholic cirrhosis admitted to the ICU had worse clinical outcomes compared to those with non-alcohol-associated cirrhosis42. Nevertheless, research on the association between cirrhotic etiology and prognosis in septic patients is currently lacking. In this study, we observed higher mortality among septic patients with alcohol-associated cirrhosis, despite their younger age and lower baseline MELD scores (Supplementary Table 2). Alcohol has detrimental effects on the immune system by suppressing the functionality of immune cells and disrupting the balance of immune cell cytokines, resulting in a state of immunosuppression43. Therefore, chronic alcohol consumption may contribute to the increased severity of sepsis in patients with alcohol-associated cirrhosis. Moreover, alcohol use disorder is a common complication among patients with alcohol-associated cirrhosis. This disorder has the potential to cause harm to various organs, including the liver, cardiovascular system, and brain, consequently increasing the mortality risk in septic patients with alcohol-associated cirrhosis44. Therefore, our study emphasizes the need for heightened attention to patients with alcohol-associated cirrhosis. Furthermore, our study revealed a linear correlation between ACAG levels and 28-day mortality, suggesting that ACAG could be a reliable indicator for detecting high-risk patients with cirrhosis and sepsis. In summary, our study suggested that ACAG, an easily accessible and cost-effective indicator, holds significant predictive value for adverse outcomes in septic patients with cirrhosis.

Of course, the present study also had several limitations. Firstly, given the inherent limitations of retrospective research, it is unable to establish a definitively causal relationship. Additionally, despite the use of comprehensive statistical analyses being a strength that enhanced the credibility of our study’s findings, certain variables not controlled for in our study could still influence clinical outcomes. Additionally, the lack of external validation could impact the reliability of our results, and future studies using independent datasets are necessary to confirm these findings. Furthermore, ACAG was not dynamically monitored: only the first-day ACAG in the ICU was analyzed. Therefore, evaluating the predictive power of ACAG changes is crucial for future research.

Conclusion

High ACAG levels emerge as an independent risk factor for 28-day mortality among septic patients with cirrhosis. Integrating ACAG assessment into routine clinical practice could enhance risk stratification and guide interventions for better outcomes in this high-risk patient cohort.