The JX-ADHF1 study is a retrospective cohort study initiated by medical professionals. Its primary objective is to establish a high-quality cohort of ADHF patients, effectively utilizing clinical records during hospitalization to explore new methods for early risk stratification and improve the prognosis of ADHF patients. Specifically, the JX-ADHF1 study consecutively included 1,790 ADHF patients admitted to Jiangxi Provincial People’s Hospital from January 2019 to December 2022. The diagnosis of ADHF was based on the latest European Society of Cardiology guidelines for acute and chronic heart failure available at the time of admission. The exclusion criteria for the study population were as follows: (i) 23 subjects with liver cirrhosis and 99 subjects with stage 5 chronic kidney disease or a history of hemodialysis were excluded due to the potential adverse impact of non-heart failure related fluid retention on the study factors and prognosis. (ii) 42 subjects who underwent percutaneous coronary intervention within the last three months were excluded due to the significant impact of reperfusion therapy on short-term prognosis. (iii) 73 subjects with malignant tumors were excluded due to their potentially life-limiting prognosis impacting the study results. (iv) 63 patients with pacemakers were excluded due to expected autonomic regulatory deficits. (v) 12 minors and 1 pregnant woman were also excluded. Finally, for the current study, 229 subjects with missing baseline AIP information were excluded. Ultimately, the current study evaluated 1,248 ADHF patients, and Figure 1 depicts the detailed screening process for the entire study population.

The JX-ADHF1 study data, owned by Jiangxi Provincial People’s Hospital, is accessible to researchers who have signed a data use agreement. The JX-ADHF1 study protocol was approved by the Ethics Committee of Jiangxi Provincial People’s Hospital (IRB: 2024–01), and consent for data use was obtained from the subjects and their families. The research dataset anonymized personal identifiers with identifiable information, adhering to the ethical principles of the Declaration of Helsinki and reported results according to the STROBE guidelines Supplementary Text 1.

Demographic (gender and age) and clinical data [comorbidities (hypertension, diabetes, cerebral infarction, coronary heart disease), New York Heart Association (NYHA) classification at admission, systolic and diastolic blood pressure (SBP and DBP), echocardiogram results at admission, and blood sample laboratory parameters] of the subjects at baseline were independently collected and cross-checked by two trained researchers. The blood pressure measurements recorded in the current analysis were the first measurements after admission, taken in a calm environment or bedside using an Omron automatic sphygmomanometer (HBP-1300). Comorbidities were determined based on patient self-report, ongoing medication treatment, or records in the patient’s medical history.

Laboratory parameters were measured within 24 hours of admission at the Jiangxi Provincial People’s Hospital laboratory center by professional medical laboratory personnel using automatic analyzers. The biochemical indicators measured included albumin, alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), creatinine (Cr), uric acid (UA), total cholesterol (TC), TG, low-density lipoprotein cholesterol (LDL-C), and HDL-C, routine blood parameters [white blood cell count (WBC), red blood cell count (RBC), hemoglobin (HGB), platelet count (PLT)], and the cardiac function indicator N-terminal pro B-type natriuretic peptide (NT-proBNP). It is important to note that lipid and liver enzyme-related indicators were measured from venous blood samples taken on an empty stomach at admission or the next morning after admission.

AIP = log10 (TG/HDL-C) (22).

The primary outcome of the study is the all-cause 30-day mortality rate. The start of the follow-up for all ADHF patients is set at the time of admission, and their 30-day survival status is obtained by trained medical workers through text messages, phone calls, and face-to-face follow-ups in outpatient and inpatient settings.

Data analysis in this study is performed using R language version 4.2.1 and Empower(R) version 2.20 statistical software. Baseline characteristics of the study population were described as counts (%), mean (standard deviation), or median (interquartile range) based on the type and distribution of variables. Differences between groups were compared using t-tests, one-way ANOVA, and non-parametric tests, with a two-sided P < 0.05 set as the threshold for statistical significance.

Kaplan-Meier curve was used to depict the 30-day survival rates of ADHF patients. Cox regression models were constructed to test the association between AIP and the 30-day mortality rate in ADHF patients, in which the variance inflation factors of all covariates were considered in the adjustment of variables (28), and evaluated the suitability of the Cox regression model using Schoenfeld residuals to assess the proportional hazards assumption (29). To test the robustness of the Cox regression analysis, the minimum strength of association needed for an unmeasured confounder to explain the 30-day mortality rate in ADHF patients was calculated based on the final adjusted model (E-value) (30).

Nested within the Cox regression model, RCS with four knots was employed to model the dose-response relationship between AIP and the 30-day mortality rate in ADHF patients. Stratified analysis was further used to explore how the association between AIP and the 30-day mortality rate in ADHF patients varies across different subgroups, with likelihood ratio tests assessing differences between strata.

After establishing the association between AIP and the 30-day mortality rate in ADHF patients, mediation analysis (31) was conducted to explore whether oxidative stress (32), inflammation (33), and nutrition (34) pathways mediate the relationship between AIP and the 30-day mortality rate. The size of the mediation effect was quantified by calculating the ratio of the indirect effect to the total effect, and the significance of the mediation effect was tested using the Bootstrap sampling method (35). Based on previous studies, GGT is selected as a marker of oxidative stress (36), WBC as a marker of inflammation (37), and ALB as an indicator of nutritional status (38).

The analysis included 1,248 ADHF patients with an average age of 68 years and a male-to-female ratio of 1.43:1. During the 30-day observation period, 42 patients (3.37%) experienced mortality events. Figure 2 displays the 30-day survival curve of the study population, indicating a gradual increase in mortality events over time within the 30-day period.

In this analysis, AIP in ADHF patients showed a normal distribution (Supplementary Figure 1), with a median of 0.075. Patients were divided into quartiles based on their AIP, summarizing the baseline characteristics of ADHF patients (Table 1). The results indicated that compared to the group with lower AIP values, those with higher AIP were younger, had lower SBP, left ventricular ejection fraction (LVEF), ALB, and HDL-C levels, and higher levels of WBC, RBC, PLT, ALT, Cr, UA, TC, TG, and LDL-C. Additionally, patients with higher AIP were more likely to have comorbid diabetes and coronary heart disease, a higher proportion of NYHA class IV, and a significantly higher probability of mortality within 30 days (Q1: 1.28%, Q2: 2.88%, Q3: 2.88%, Q4: 6.41%).

Table 2 further displays the baseline characteristics of the study population based on whether or not the subject experienced mortality within the 30-day follow-up period. In summary: (i) In demographic characteristics, ADHF patients who died within 30 days were significantly older at admission, had a higher proportion of NYHA class IV, and a higher prevalence of cerebral infarction. (ii) In terms of measured data, deceased subjects typically had lower baseline levels of blood pressure, RBC, HGB, PLT, ALB, TC, HDL-C, and LDL-C, but higher levels of WBC, liver enzymes, Cr, UA, TG, NT-proBNP, and AIP (Figure 3).

Before analyzing the association, we confirmed through Schoenfeld residual plots of AIP over time (Supplementary Figure 2) that the Cox regression model did not violate the proportional hazards assumption. Additionally, the variance inflation factors of each covariate were calculated, identifying collinearity between HGB, ALB, and TC with other covariates, and thus these were not included in the subsequent multivariable models (Supplementary Table 1).

Three progressively adjusted multivariable Cox regression models were constructed to analyze the association between AIP and the 30-day mortality rate in ADHF patients (Table 3). In the first model (Model 1), adjustments were made for age, gender, and comorbidities (hypertension, diabetes, cerebral infarction, and coronary heart disease), showing a positive correlation between AIP and 30-day mortality rate in ADHF patients with an HR of 7.38 (95%CI: 2.55–21.36). Additionally, the HR values corresponding to the AIP quartiles showed a positive trend (P-trend=0.0002). In the second model (Model 2), adjustments were further made for NYHA classification, LVEF, DBP, and NT-proBNP, slightly decreasing the HR to 5.87 (95%CI: 1.77–19.44). In Model 2, despite a continuing positive trend in AIP quartiles, the HR value for the second quartile was slightly higher than for the third (HR: Q2 2.38, Q3 2.15). The third model (Model 3), further considered the impact of WBC, RBC, PLT, ALB, AST, GGT, Cr, UA, and LDL-C, with findings similar to Model 2, showing a slight decrease in HR (HR 3.94, 95%CI: 1.08–14.28). Notably, in Model 3, the HR value for the second quartile was higher than the third but lower than the fourth (HR: Q2: 2.82, Q3: 2.01, Q4: 4.09); suggesting a potential non-linear association before the fourth quartile of AIP. Based on Model 3, the minimum E-value associated with the 30-day mortality rate in ADHF patients was calculated to be 7.34.

Using a RCS with four knots, we further constructed a dose-response relationship curve between AIP in ADHF patients and their 30-day mortality rate (Figure 4). After adequately adjusting for confounding factors, we found that the dose-response curve correlated with the association analysis results in Table 3. It can be observed that the association between the 30-day mortality rate and AIP demonstrated a U-shaped curve before the fourth quartile of AIP (AIP<0.24), with the lowest 30-day mortality risk in ADHF patients occurring at an AIP of approximately -0.1.

We conducted stratified analyses based on age (median), gender, LVEF value (50%), NYHA classification, and comorbidities, assessing the presence of interactions between AIP and stratification factors through likelihood ratio tests. The results (Table 4) indicated no significant interactions in all the subgroups (All P-interaction>0.05), suggesting that the association between AIP and the 30-day mortality rate in ADHF patients was relatively stable and unlikely to be influenced by these external factors.

Mediated analysis was performed to explore the roles of inflammation, oxidative stress and nutritional pathways in the association between AIP and the 30-day mortality rate in ADHF patients. Table 5 presents the detailed results of the mediation analysis, and Figure 5 illustrates the mediation diagram for inflammation (WBC), oxidative stress (GGT), and nutrition (ALB). The exploratory analysis revealed significant mediating effects of inflammation and nutrition in the association between AIP and the 30-day mortality rate in ADHF patients (P-value of proportion mediate < 0.05), while the mediating effect of oxidative stress appeared non-significant (P-value of proportion mediate > 0.05). Specifically, inflammation accounted for approximately 24.29% of the mediation effect, and nutrition for about 8.16% in the association between AIP and the 30-day mortality rate in ADHF patients.

Our study has several notable strengths: (i) This is the first report of the relationship between AIP and the prognosis of ADHF patients. (ii) The discovery of a U-shaped curve association is of significant clinical importance, as the AIP threshold indicating the lowest death risk can provide crucial assistance in risk assessment and treatment for ADHF patients. (iii) The findings from the mediation analysis offer a mechanistic explanation for the association between AIP and 30-day mortality risk in ADHF patients and also provide insights for future treatment directions.

However, our study also has certain limitations: (i) Being observational, it inevitably includes some unmeasured factors leading to residual confounding. Nevertheless, the calculated E-value (7.34) suggests that it is unlikely that any confounding factors could significantly alter our findings. (ii) We lack repeated measurements of AIP, which might be more beneficial for early risk stratification in ADHF patients. (iii) The study evidence is primarily applicable to the population in Jiangxi, and its relevance to other regions and ethnicities should be interpreted with caution. (iv) Due to the limited sample size, we did not observe significant associations in subgroups after further stratification. (v) The observational nature of the study limits our ability to further assess the impact of enhanced nutritional support and anti-inflammatory treatment on adverse outcomes in patients with high AIP. (vi) The causes of ADHF were not distinguished in the current study; considering the significant adverse cardiovascular effects of pre-existent cardiomyopathy, infections, ischemic heart disease, heavy alcohol use or illegal drug use and some chemotherapy medicines in patients with ADHF that were already present prior to the onset of the disease (61–65), this may result in some special populations being unobserved, and further studies are needed.