The creatinine-to-body weight ratio predicts mortality in critically ill patients with heart failure: a retrospective cohort study of the MIMIC-IV database

Abstract

Background

Heart failure (HF) is a clinical syndrome with high global incidence and mortality, imposing a substantial economic burden. While creatinine (Cr) and body weight (BW) individually influence HF progression, the prognostic role of the creatinine-to-body weight ratio (CWR) remains unclear. This study investigates the association between CWR and mortality in HF patients, aiming to identify high-risk individuals and inform prognosis.

Methods

Patient data were extracted from the MIMIC-IV database. Participants were first stratified into CWR index quartiles to categorize the cohort. The primary endpoints were 30- and 365-day all-cause mortality, while secondary endpoints included 90- and 180-day mortality. Kaplan-Meier curves with log-rank tests were then used to compare survival across quartiles. Next, Cox proportional hazards regression (with sequential model adjustment) and restricted cubic spline (RCS) analysis assessed the association between CWR and prognosis. Sensitivity analyses were subsequently conducted to examine the robustness of the non-linear relationship, and subgroup analyses explored potential effect modifications. Finally, time-dependent receiver operating characteristic (ROC) curve analyses compared the predictive performance of CWR against traditional markers.

Results

Among 4,371 participants (median age: 75 years; 54.8% male), higher CWR index values were significantly associated with increased all-cause mortality risks at 30, 90, 180, and 365 days, as demonstrated by Kaplan-Meier survival curves (log-rank P < 0.01). Building on these results, Cox regression analysis further revealed that individuals in the highest CWR index quartile had an elevated risk of death compared to those in lower quartiles. Additionally, restricted cubic spline (RCS) analysis showed a robust biphasic nonlinear association between the CWR index and mortality, identifying 0.05 as a key threshold where the risk relationship changes. Specifically, for most patients (CWR ≤ 0.05), mortality risk increases with rising CWR values until it plateaus, whereas for those with CWR > 0.05 (n = 320), a distinct pathophysiological state may emerge, marked by a sharp increase in risk. The underlying mechanisms require further investigation.

Conclusion

CWR demonstrates a robust biphasic association with mortality, identifying 0.05 as a critical threshold separating a risk-plateau phase from extreme high-risk. Its stable prognostic value suggests CWR may integrate acute hemodynamic and chronic metabolic stress, though prospective validation is warranted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05509-1.

Introduction

Heart failure (HF) is a complex clinical syndrome [1]. It is initiated by cardiac dysfunction, typically arising from structural or functional cardiac abnormalities—such as myocardial infarction. These changes compromise ventricular systolic or diastolic performance. As the terminal stage of various cardiovascular diseases (CVD), the onset and progression of HF are intimately connected to global epidemics. These include obesity, diabetes mellitus (DM), metabolic syndrome, ischemic heart disease (IHD), and infectious etiologies such as rheumatic heart disease (RHD) [2]. These widespread conditions have contributed to a significant rise in CVD prevalence. Consequently, the incidence of HF has increased on a global scale. HF is a progressive disorder. It constitutes a major public health concern worldwide, with reported prevalence rates ranging from 0.1% to 6.7%. The annual mortality rate is approximately 10% [3]. Prognosis is especially poor for patients in critical care settings. Chronic HF results in hypoperfusion of vital organs—including the liver, kidneys, and lungs. This leads to multi-organ dysfunction, metabolic derangements, and disruption of homeostasis. These complications markedly increase the risk of life-threatening arrhythmias and cardiac arrest [4]. HF is one of the leading causes of hospitalization and incurs substantial healthcare costs. In the United States alone, more than one million patients are hospitalized due to HF each year. The 30-day readmission rate exceeds 25% [5]. The total medical cost attributed to HF was estimated at $30.7 billion in 2012. Projections indicate a dramatic increase to $69.8 billion by 2030—an escalation of 127% [6]. Given the high prevalence, considerable mortality, significant economic impact, and increased clinical risks among critically ill patients, HF represents a significant and growing public health challenge. There is an urgent need to develop accurate prognostic tools. These tools will help enable effective risk stratification and management of severe HF cases.

The creatinine-to-weight ratio (CWR) is a readily available and stable clinical parameter. Recent research has demonstrated a significant correlation between CWR and mortality across various diseases [7]. These include cardiovascular disorders. An elevated CWR is thought to reflect low body fat and diminished energy reserves. This condition may increase mortality risk in both acute and chronic illnesses [8–10].

Creatinine (Cr) is an anhydride form of creatine derived primarily from muscle metabolism [11]. It is widely used as an indicator of renal function. As a biomarker that reflects both cardiac and renal status, serum creatinine levels are influenced by impaired organ function. They are also affected by the accumulation of toxic metabolites and neuroendocrine regulation. Several studies, including one by Linda et al., indicate that elevated creatinine is associated with a higher risk of mortality in cardiovascular diseases (CVD) and chronic HF [12, 13].

Body weight (BW) is a composite measure of muscle mass and fat content. It forms the basis for calculating the body mass index (BMI), a standard tool for assessing obesity. Substantial evidence supports a clear correlation between obesity and the risk of developing CVD [14]. A high BMI is an independent risk factor for HF onset. Paradoxically, however, recent studies have identified an “obesity paradox.” Overweight and obese patients with established HF demonstrate a survival advantage. They have significantly lower all-cause and cardiovascular mortality compared to normal or underweight counterparts [15, 16].

Despite the significant association between CWR and cardiovascular diseases, particularly in HF patients, its prognostic value in critically ill HF patients remains unclear. Compared to their stable counterparts, this high-risk subgroup experiences markedly higher mortality [11]. This underscores the urgent need to establish a reliable risk assessment tool.

Methods

This retrospective cohort study utilized data from the Medical Information Mart for Intensive Care (MIMIC-IV) database. The methodological framework was adapted from the analysis protocol established by Huang et al. (2025) in BMC Nephrology [17], with modifications specific to a heart failure (HF) patient population. The following sections summarize the key methodological components.

Data source and ethical approval

The analysis was conducted using the MIMIC-IV database, a collaborative initiative between the Beth Israel Deaconess Medical Center (BIDMC) and the Massachusetts Institute of Technology (MIT). This publicly available database contains de-identified clinical data from patients admitted to intensive care units (ICUs) [18]. Researchers gain access by completing the Collaborative Institutional Training Initiative exam (certification number 61042076 for author Huang) and signing a data use agreement for human subjects research. Ethical approval for the original data collection was granted by the institutional review boards (IRBs) of both MIT and BIDMC, waiving the requirement for individual patient consent and further ethical approval for this secondary analysis.

Study population and selection criteria

The initial patient cohort was identified from the MIMIC-IV database. Data were extracted using Structured Query Language (SQL, version 14.0) through Navicat Premium software (version 15.0.12) to obtain a comprehensive set of variables, including socio-demographic characteristics, vital signs, laboratory parameters, complication profiles, and microbiological data [19]. Patients in the database who met the following criteria were selected: [1] met the diagnostic criteria for HF. ICD-9 (4280, 4281, 42822, 42830, 42832, 42840, 42842, 42843, and 4289) and ICD-10 (I509, I50, I502, I5020, I5022, I503, I5030, I5032, I5033, I504, I5040, I5042, I508, I5081, I50810, I50812, I50813, I50814, I5082, I5083, I5084, and I5089) codes were used to identify patients with HF in the MIMIC-IV database. [2] First ICU admission at first hospitalization. [3] Length of hospital stay > 48 h and length of ICU stay > 48 h. [4] age > 18 years. [5] Elimination of duplicate patient data. Finally, 4,371 patients met the inclusion criteria. These patients were divided into four groups based on quartiles of CWR (Figure.1).

Fig. 1.

Flowchart of the patient selection process

Variable extraction and definitions

Variables with > 30% missing data in MIMIC-IV, such as height, basophils, C-reactive protein, and other parameters were collected, were excluded from the present study.A total of 49 variables were extracted. The predictors included: (a) demographic information: age, gender and body weight; (b) comorbidities: rheumatic disease, renal disease, peripheral vascular disease, cerebrovascular disease, chronic pulmonary disease, hepatic disease, diabetes, and obesity. (c) first day ICU vital signs: mean heart rate, mean blood pressure (MBP), mean respiratory rate and mean temperature; (d) first day ICU laboratory results: mean anion gap, mean bicarbonate, mean serum chloride, mean serum sodium, mean serum potassium, mean serum calcium, mean hematocrit, mean red blood cell (RBC) count, mean hemoglobin, mean white blood cell (WBC) count, mean platelet count, mean glucose, mean international normalized ratio (INR), mean partial thromboplastin time (PTT), mean prothrombin time (PT), mean blood urea nitrogen (BUN) level, mean serum creatinine (Cr), mean alkaline phosphatase (ALP), mean alanine aminotransferase (ALT), mean aspartate aminotransferase (AST), mean serum lactate, mean urine output; (e) organ scores: Simplified Acute Physiology Score Ⅱ (SAPS-Ⅱ), Acute Physiology Score Ⅲ (APS-Ⅲ), Logistic Organ Dysfunction System (LODS), Oxford Acute Severity of Illness Score (OASIS), Glasgow Coma Scale (GCS), and Charlson Comorbidity Index (CCI); (f) treatment: Renal Replacement Therapy (RRT), Diuretic Drug (Representative Drug: Furosemide). The primary endpoints of this study were 30-day and 365-day all-cause mortality, while secondary outcomes included 90-day and 180-day all-cause mortality.

Statistical analysis

All statistical analyses were performed using SPSS software (version 27.0). The study population (n = 4,371) was stratified into four groups based on quartiles of baseline CWR values, with each group comprising participants in the lowest, second, third, or highest quartile, to facilitate comparative analysis. The normality of continuous variable distributions was assessed using the Shapiro-Wilk test. Data are presented accordingly: normally distributed variables as mean ± standard deviation (SD), and non-normally distributed variables as median with interquartile range (IQR). For group comparisons, Student’s t-test (for two groups) or one-way analysis of variance (ANOVA) (for more than two groups) was used for normal data; the Mann-Whitney U test or Kruskal-Wallis test was applied for non-normal data. Categorical variables are expressed as numbers (percentages) and were compared using the Chi-squared test or Fisher’s exact test, as appropriate. Variables with more than 20% missing data were excluded from the analysis. For variables with ≤ 20% missingness, missing values were imputed using the mean for continuous variables and the mode for categorical variables. A two-sided p-value of less than 0.05 was considered statistically significant. The baseline characteristics and comparative results of the groups are summarised in Table 1.

Table 1.

Baseline characteristics of patients grouped according to CWR index quartiles

Variables	Total (n = 4371)	Q1 (<0.011) (n = 1094)	Q2 [0.011,0.015) (n = 1092)	Q3 [0.015,0.024) (n = 1093)	Q4 (≥ 0.024) (n = 1092)	p
Demographics
Age(year)	75.00(65.00,84.00)	69.00(60.00,78.00)	75.00(65.00,83.00)	79.00(70.00,86.00)	77.00(67.00,84.75)	<0.001
Gender						<0.001
Male(%)	2394(54.80)	550(50.30)	593(54.30)	605(55.40)	646(59.20)
Female(%)	1977(45.20)	544(49.70)	499(45.70)	488(44.60)	446(40.80)
Weight(kg)	80.00(66.85,96.00)	93.98(77.10,113.03)	80.65(68.09,95.23)	75.50(63.70,89.25)	73.70(62.30,88.24)	<0.001
Vital signs
Temperature	36.76(36.53,37.00)	36.77(36.64,37.09)	36.76(36.58,37.02)	36.76(36.52,37.02)	36.68(36.40,36.92)	<0.001
Heartrate(bpm)	82.93(73.62,94.00)	84.43(75.80,96.00)	82.59(74.21,93.54)	81.93(73.18,94.24)	82.51(71.62,92.22)	<0.001
Respiratory rate (insp/min)	19.32(17.17,22.08)	19.23(17.22,21.70)	19.08(16.89,21.77)	19.57(17.30,22.66)	19.42(17.28,22.22)	0.013
MBP(mmHg)	74.00(68.38,80.41)	75.83(70.76,82.14)	74.55(68.88,80.82)	73.75(67.97,79.69)	71.84(65.90,78.50)	<0.001
Organ dysfunction scores
APS-Ⅲ	51.00(39.00,70.00)	41.00(33.00,57.00)	45.00(35.00,63.00)	53.00(41.00,69.00)	65.00(52.00,82.00)	<0.001
SAPS-Ⅱ	40.00(32.00,50.00)	35.00(28.00,42.00)	38.00(31.00,46.00)	42.00(35.00,51.00)	48.00(30.00,56.00)	<0.001
LODS	6.00(4.00,8.00)	4.00(2.75,6.00)	5.00(3.00,7.00)	6.00(4.00,8.00)	7.00(6.00,10.00)	<0.001
OASIS	35.00(29.00,42.00)	34.00(28.00,40.00)	34.00(28.00,41.00)	35.00(29.00,41.00)	37.00(30.00,44.00)	<0.001
GCS	14.00(10.00,15.00)	14.00(10.00,15.00)	14.00(10.00,15.00)	14.00(9.00,14.00)	13.00(9.00,14.00)	0.003
CCI	7.00(6.00,9.00)	6.00(5.00,7.00)	7.00(6.00,8.00)	8.00(6.00,9.00)	9.00(7.00,10.00)	<0.001
Comorbidity
Rheumatic disease[n (%)]						0.984
Yes	187(4.30)	45(4.10)	48(4.40)	46(4.20)	48(4.40)
No	4184(95.70)	1049(95.90)	1044(95.60)	1047(95.80)	1044(95.60)
Renal disease[n (%)]						<0.001
Yes	1574(36.00)	63(5.80)	223(20.40)	467(42.70)	821(75.20)
No	2797(64.00)	1031(94.20)	869(79.60)	626(57.30)	271(24.80)
Peripheral vascular disease[n (%)]						<0.001
Yes	753(17.20)	131(12.00)	174(15.90)	213(19.50)	235(21.50)
No	3618(82.80)	963(88.00)	918(84.10)	880(80.50)	857(78.50)
Cerebrovascular disease [n (%)]						0.193
Yes	691(15.80)	162(14.80)	189(17.30)	182(16.70)	158(14.50)
No	3680(84.20)	932(85.20)	903(82.70)	911(83.30)	934(85.50)
Chronic pulmonary disease [n (%)]						0.004
Yes	1746(39.90)	469(42.90)	459(42.00)	423(38.70)	395(36.20)
No	2625(60.10)	625(57.10)	633(58.00)	670(61.30)	697(63.80)
Liver disease[n (%)]						<0.001
Yes	465(10.60)	85(7.80)	99(9.10)	115(10.50)	166(15.20)
No	3906(89.40)	1009(92.20)	993(90.90)	978(89.50)	926(84.800
Diabetes[n (%)]						<0.001
Yes	1730(39.60)	409(37.40)	377(34.50)	416(38.10)	528(48.40)
No	2641(60.40)	685(62.60)	715(65.50)	677(61.90)	564(51.60)
Obesity[n (%)]						<0.001
Yes	664(15.20)	326(29.80)	136(12.50)	109(10.00)	93(8.50)
No	3707(84.80)	768(70.20)	956(87.50)	984(90.00)	999(91.50)
Laboratory parameters
Anion gap_mean(mmol/L)	14.50(12.50,17.00)	13.00(11.19,14.67)	13.67(12.00,15.50)	15.00(13.00,17.00)	17.33(15.00,20.00)	<0.001
Bicarbonate_mean(mmol/L)	25.80(23.76,27.95)	27.00(25.09,29.44)	26.19(24.50,28.07)	25.39(23.51,27.67)	24.50(22.21,26.70)	<0.001
Chloride_mean(mmol/L)	102.10(99.22,104.67)	101.72(99.06,104.05)	102.31(100.05,104.66)	102.60(99.60,105.17)	101.60(97.93,104.95)	<0.001
Sodium_mean(mmol/L)	138.75(136.67,140.80)	138.78(136.93,140.67)	138.88(136.79,140.73)	138.90(136.70,141.16)	138.41(136.22,140.67)	0.033
Potassium_mean(mmol/L)	4.17(3.98,4.39)	4.07(3.91,4.26)	4.14(3.96,4.32)	4.21(4.00,4.43)	4.30(4.08,4.57)	<0.001
Calcium_mean (mmol/L)	8.61(8.28,8.92)	8.60(8.25,8.91)	8.62(8.30,8.90)	8.61(8.30,8.94)	8.60(8.23,8.95)	0.355
Hematocrit_mean(%)	31.70(28.00,36.15)	32.66(28.95,37.31)	31.80(28.48,36.73)	31.80(28.10,36.18)	30.10(26.77,34.20)	<0.001
RBC_mean(109/L)	3.50(3.08,4.00)	3.65(3.21,4.15)	3.55(3.16,4.08)	3.53(3.08,4.04)	3.27(2.88,3.75)	<0.001
Hemoglobin_mean(g/dL)	10.38(9.20,11.86)	10.82(9.57,12.20)	10.60(9.42,12.07)	10.44(9.20,11.90)	9.79(8.61,11.16)	<0.001
Platelets_mean(109/L)	189.20(139.75,250.50)	196.00(146.38,255.13)	186.58(137.67,244.67)	186.00(140.00,249.50)	188.00(135.38,254.25)	0.016
WBC_mean(109/L)	11.40(8.40,15.20)	11.66(8.73,15.08)	11.44(8.55,14.79)	11.20(8.13,15.40)	11.20(8.24,15.59)	0.769
Glucose_mean(mg/dL)	134.00(116.13,164.60)	131.16(116.09,156.20)	132.00(117.42,159.00)	136.20(117.00,170.04)	136.88(112.38,174.98)	0.013
INR_mean	1.30(1.17,1.61)	1.30(1.15,1.46)	1.30(1.17,1.53)	1.33(1.20,1.62)	1.40(1.20,1.85)	<0.001
PTT_mean(s)	34.55(29.10,44.18)	33.44(28.57,41.50)	34.07(29.08,43.05)	34.60(29.00,45.43)	36.52(29.75,46.99)	<0.001
PT_mean(s)	14.70(13.00,17.40)	14.30(12.89,15.92)	14.45(12.90,16.73)	14.80(13.10,17.52)	15.33(13.18,19.68)	<0.001
BUN_mean(mg/dL)	26.33(17.50,42.60)	16.00(12.50,21.00)	21.00(16.50,28.50)	30.75(23.00,41.33)	53.58(37.67,76.46)	<0.001
Creatinine_mean(mg/dL)	1.20(0.87,1.90)	0.75(0.60,0.90)	1.03(0.85,1.20)	1.43(1.78,1.70)	2.85(2.10,4.30)	<0.001
Alp_mean(U/L)	92.20(69.80,119.16)	86.71(66.63,108.27)	87.73(67.74,111.16)	91.40(70.00,117.10)	105.12(76.21,140.82)	<0.001
Alt_mean(U/L)	27.33(17.22,63.08)	27.00(17.37,50.36)	27.69(17.25,55.36)	27.75(18.00,72.90)	28.00(16.50,74.92)	0.101
Ast_mean(U/L)	35.71(24.00,78.00)	32.00(22.75,61.97)	35.50(23.91,65.87)	37.50(25.06,90.23)	38.76(24.81,101.45)	<0.001
Lactate mean(mmol/L)	1.83(1.40,2.29)	1.78(1.38,2.10)	1.86(1.42,2.27)	1.85(1.43,2.32)	1.86(1.38,2.44)	0.003
Urine output_mean(ml)	1469.00(866.00,2205.00)	1710.00(1153.00,2571.25)	1484.00(986.25,2343.75)	1506.00(930.00,2230.00)	989.00(383.50,1688.69)	<0.001
Treatment
RRT [n (%)]						<0.001
Yes	178(4.07)	6(0.55)	22(2.01)	23(2.10)	127(11.63)
No	4193(95.93)	1088(99.45)	1070(97.99)	1070(97.90)	965(88.37)
Furosemide [n (%)]						<0.001
Yes	3287(75.20)	861(78.70)	859(78.66)	859(78.59)	708(64.84)
No	1084(24.80)	233(21.30)	233(21.34)	234(21.41)	384(35.16)
Outcome
30-day mortality [n (%)]						<0.001
Yes	911(20.80)	121(11.10)	187(17.10)	263(24.10)	340(31.10)
No	4360(79.20)	973(88.90)	905(82.90)	830(75.90)	752(68.90)
90-day mortality [n (%)]						<0.001
Yes	1228(28.10)	188(17.20)	247(22.60)	351(32.10)	442(40.50)
No	3143(71.90)	906(82.80)	845(77.40)	742(67.90)	650(59.50)
180-day mortality [n (%)]						<0.001
Yes	1437(32.90)	221(20.20)	297(27.20)	404(37.00)	515(47.20)
No	2934(67.10)	873(79.80)	795(72.80)	689(63.00)	577(52.80)
365-day mortality [n (%)]						<0.001
Yes	1673(38.30)	269(24.60)	340(31.10)	469(42.90)	595(54.50)
No	2698(61.70)	825(75.40)	752(68.90)	624(57.10)	497(45.50)

Variables are initial values unless otherwise specified. _mean means the average variable during ICU. WBC white blood cells, RBC red blood cells, PTT partial thromboplastin time, PT prothrombin time, INR international normalized ratio, MBP mean blood pressure, BUN blood urea nitrogen,SAPS-Ⅱ simplified acute physiology score Ⅱ,APS-Ⅲ acute physiology score Ⅲ, LODS logistic organ dysfunction system, OASIS oxford acute severity of illness score, GCS Glasgow coma scale, CCI Charlson comorbidity index, RRT Renal Replacemert Therapy

Survival outcomes were evaluated using Kaplan-Meier curves, stratified by CWR index quartiles, to visualize differences in event rates for primary (30-day and 365-day) and secondary (90-day and 180-day) endpoints. Log-rank tests were applied to assess statistical significance across strata. The association between the CWR index and mortality was further quantified through Cox proportional hazards regression. Initially, univariable Cox models were fitted to examine the crude association between CWR (continuous and quartiles) and mortality at each time point. Subsequently, multivariable models were constructed to adjust for potential confounders. Three sequential models were employed:

Model 1: Not adjusted.

Model 2: Adjustments: Aniongap_mean, Bicarbonate_mean, Chloride_mean, Sodium_mean, Creatinine_mean, Age, Gender, Weight, Potassium_mean.

Model 3: Adjustments: Aniongap_mean, Bicarbonate_mean, Chloride_mean, Sodium_mean, Creatinine_mean, Hemoglobin_mean, Hematocrit_mean, RBC_mean, Renal_disease, SAPS-Ⅱ, Urineoutput, RRT, Furosemide.

In all models, the lowest quartile of the CWR index served as the reference group. This approach allowed for a comprehensive assessment of the independent association between CWR and mortality, while sequentially controlling for physiological domains. A distinct, quartile-stratified analysis was conducted to explore potential effect modification by CWR level on the relationship between mean Cr and mortality. Within each CWR quartile, the Cr-mortality association was evaluated using: (1) univariate Cox regression (Model 1) to assess crude associations; (2) multivariable-adjusted models corresponding to Models 2 and 3 described above. This analytical strategy enabled examination of whether the prognostic value of creatinine varied with underlying CWR levels.

To explore the potential non-linear association between CWR and in-hospital mortality, we employed restricted cubic splines (RCS) with CWR as a continuous variable. If a non-linear relationship was confirmed, a recursive algorithm was used to identify its inflection point for mortality outcomes at 30, 90, 180, and 365 days. We subsequently conducted sensitivity analyses to examine the robustness of the observed pattern, which involved sequentially excluding patients with extreme CWR values and varying the number of knots in the spline models (without covariate adjustment). Subgroup analyses were further performed to evaluate potential effect modifications by age, sex, peripheral vascular disease, chronic pulmonary disease, liver disease, diabetes, and obesity. Finally, we compared the predictive performance of CWR with that of glomerular filtration rate (eGFR), serum creatinine (Cr), and body weight (BW) using time-dependent receiver operating characteristic (ROC) curve analyses.

All statistical analyses were performed utilizing SPSS version 27.0 and R software (package version 4.3.2), with the significance level established at P < 0.05.

Results

Baseline demographic and clinical characteristics

A total of 4,371 patients diagnosed with HF were finally included in our study. The mean age of patients enrolled in the study was 75.00(65.00,84.00) years, and the percentage of males was 54.80%. Based on the quartiles of the CWR index at enrolment (Q1: < 0.011, Q2: 0.011–0.015, Q3: 0.015–0.024, Q4: ≥0.024), the baseline characteristics of the study participants were analyzed and are presented in Table 1.

Compared to patients in the lowest CWR group, those in the highest CWR group exhibited significant differences in clinical characteristics, including: (1) the high-CWR group had a higher prevalence of renal disease, chronic pulmonary disease, and diabetes.(2)lower body temperature and body weight.(3)increased mean anion gap levels, mean creatinine, mean serum potassium, mean glucose, mean BUN, mean ALP and AST levels, along with decreased mean bicarbonate levels; (4) prolonged mean INR, PT and PTT; (5) reduced hemoglobin, hematocrit, WBC count and RBC count; (6) higher severity scores (APS-III, SAPS-II, LODS, OASIS and CCI); and (7) a higher proportion of patients received renal replacement therapy, while a lower proportion received furosemide. Meanwhile, in the group Q4, 30-day mortality (11.10%vs. 17.10% vs. 24.10% vs. 31.10%, P<0.001)、90-day mortality(17.20% vs. 22.60% vs. 32.10% vs. 40.50%, P<0.001), 180-day mortality (20.20% vs. 27.20% vs. 37.00% vs.47.20%, P<0.001༉, and 365-day mortality (24.60% vs. 31.10% vs. 42.90% vs. 54.50%, P<0.001) were all higher than in the other three groups.

Study outcomes

The Kaplan-Meier curves (Figure. 2) demonstrated significant disparities in survival rates across the four CWR quartiles at 365 days. Patients in the highest CWR quartile (Q4) had significantly lower survival rates at all these time points compared to those in the lower quartiles (log-rank test, P < 0.001). Furthermore, pairwise comparisons revealed statistically significant differences in survival among all quartiles (Q1, Q2, Q3, and Q4) at each time point (all P < 0.05).

Fig. 2.

Kaplan-Meier curve of 365-day all-cause mortality stratified by CWR index

Relationship between CWR and clinical outcomes in critically ill patients with HF

Four Cox proportional hazards models were employed to examine the independent effects of the CWR index on mortality (Tables 2 and 3; Supplementary Tables 1 and 2). Model 2 adjusted for demographic and clinical covariates, including Aniongap_mean, Bicarbonate_mean, Chloride_mean, Sodium_mean, Creatinine_mean, Age, Gender, Weight and Potassium_mean. The hazard ratios (HRs) and 95% confidence intervals (CIs) for the CWR index categories (Q1: < 0.011, Q2: 0.011–0.015, Q3: 0.015–0.024, Q4: ≥ 0.024) for 30-day all-cause mortality were as follows: 1.000 (reference), 1.253 (0.987–1.587), 1.421 (1.117–1.809) and 1.817 (1.360–2.428). In Model 3, more covariates were adjusted, including Anion gap_mean, Bicarbonate_mean, Chloride_mean, Sodium_mean, Creatinine_mean, Hemoglobin_mean, Hematocrit_mean, RBC_mean, Renal_disease, SAPS-II, Urineoutput, RRT and Furosemide. The HRs and 95% CIs for 30-day all-cause mortality were: 1.000 (reference), 1.352 (1.072–1.705), 1.662 (1.318–2.069) and 2.022 (1.533–2.665). For 365-day all-cause mortality, the HRs in Model 2 were: 1.000 (reference), 1.083 (0.917–1.280), 1.369 (1.156–1.621) and 1.864 (1.517–2.290). In Model 3, the HRs were: 1.000 (reference), 1.192 (1.013–1.402), 1.628 (1.383–1.915) and 2.100 (1.724–2.559). These findings suggest that patients with a CWR index of ≥ 0.024 have a significantly higher risk of both 30-day and 365-day all-cause mortality compared to those with a CWR index of < 0.024. Similar trends were observed for 90-day and 180-day all-cause mortality.

Table 2.

Cox proportional hazard models for 30-day all-cause mortality

Variables	Model1		Model2		Model3
	HR(95%CI)	P	HR(95%CI)	P	HR(95%CI)	P
CWR quantile
1	1.000		1.000		1.000 (Reference)
	(Reference)		(Reference)
2	1.593	<0.001	1.253	0.064	1.352	0.011
	(1.267–2.002)		(0.987–1.587)		(1.072–1.705)
3	2.343	<0.001	1.421	0.004	1.662	<0.001
	(1.889–2.906)		(1.117–1.809)		(1.318–2.069)
4	3.180	<0.001	1.817	<0.001	2.022	<0.001
	(2.584–3.914)		(1.360–2.428)		(1.533–2.665)

Table 3.

Cox proportional hazard models for 365-day all-cause mortality

Variables	Model1		Model2		Model3
	HR(95%CI)	P	HR(95%CI)	P	HR(95%CI)	P
CWR quantile
1	1.000		1.000		1.000 (Reference)
	(Reference)		(Reference)
2	1.334	<0.001	1.083	0.347	1.192	0.034
	(1.137–1.566)		(0.917–1.280)		(1.013–1.402)
3	2.005	<0.001	1.369	<0.001	1.628	<0.001
	(1.726–2.329)		(1.156–1.621)		(1.383–1.915)
4	2.795	<0.001	1.864	<0.001	2.100	<0.001
	(2.420–3.229)		(1.517–2.290)		(1.724–2.559)

HR: Hazard Ratio, CI: Confidence Interval

Model 1: Not adjusted

Model2: Adjustments:Aniongap_mean, Bicarbonate_mean, Chloride_mean, sodium_mean, Creatinine_mean, Age, Gender, Weight, Potassiummean

Model3:Adjustments:Aniongap_mean, Bicarbonate_mean, Chloride_mean, Sodium_mean, Creatinine_mean, Hemoglobin_mean, Hematocrit_mean, RBC_mean, Renaldisease, SAPS-Ⅱ, Urineoutput, RRT, Furosemide

Detection of non-linear relationships

Restricted cubic spline (RCS) analyses revealed a nonlinear association (P for nonlinearity < 0.001) between the CWR index and all-cause mortality at 30, 90, 180, and 365 days (Figure. 3 A–D). A threshold effect was identified at a CWR value of 0.05 across all time points. At CWR levels ≤ 0.05, risk rose relatively steeply at lower CWR values and then plateaued as CWR approached 0.05. In contrast, for CWR > 0.05—a subgroup comprising 320 patients (7.3%) in the raw dataset—the risk curve entered an extreme high-risk interval.

Fig. 3.

Association of the CWR Index with All‑Cause Mortality across Different Time Points (A) 30‑day, (B) 90‑day, (C) 180‑day, (D) 365‑day mortality

To examine the robustness and pattern of this nonlinear relationship, a series of sensitivity analyses were performed. First, patients with extreme CWR values—specifically, the highest (CWR > 0.05, n = 320) and lowest (CWR < 0.005, n = 57) values—were sequentially excluded before refitting the RCS models. For illustrative purposes, results based on 365-day mortality are shown (Supplementary Fig. 1A–E; outcomes at 30, 90, and 180 days exhibited similar patterns and are not displayed for brevity). After excluding patients with high CWR values (> 0.05), the risk exhibited a monotonic increase that plateaued near CWR = 0.05 within the ≤ 0.05 range, and the initial descending segment was no longer observed (Supplementary Figure A). Exclusion of patients with the lowest CWR values (< 0.005) did not substantially alter the initial rising trend (Supplementary Figure B). When both extreme high and low subgroups were excluded, the risk continued to display a monotonic increase that plateaued near the 0.05 threshold, again without a descending phase (Supplementary Figure C). Second, varying the number of knots (3 and 5) in the spline function demonstrated that the biphasic risk pattern and the threshold near CWR ≈ 0.05 remained consistent across model specifications (Supplementary Figures D, E). Finally, models with different levels of covariate adjustment were fitted, and the core nonlinear pattern—characterized by an initial rise and plateau at CWR ≤ 0.05 and a distinct extreme-risk phase for CWR > 0.05—was consistently observed.

Stratified analyses

To further assess the consistency of the association between CWR levels and all-cause mortality at 30, 90, 180, and 365 days across different populations, subgroup analyses were conducted based on age, gender, peripheral vascular disease, chronic pulmonary disease, liver disease, diabetes and obesity. The subgroup analyses demonstrated that CWR levels were significantly associated with all-cause mortality in HF patients across all subgroups, with the exception of liver disease, at 30-day, 180-day, and 365-day follow-ups (P < 0.05). For the 90-day all-cause mortality outcome, a significant association was observed in all subgroups except for age ≤ 65 years and liver disease (P < 0.05; Fig. 4A, B and C, and 4D). Meanwhile, the interactive analysis showed that the association between CWR levels and all-cause mortality in heart failure patients remained consistent across most predefined subgroups at 30, 90, 180, and 365 days. The results of the interaction tests indicated that, apart from age and obesity, no significant effect modification was observed for the other covariates examined. Specifically, the association was not significantly modified by gender, peripheral vascular disease, chronic pulmonary disease, liver disease, or diabetes mellitus across all evaluated time points (Fig. 4A, B and C, and 4 D).

Fig. 4.

Forest plots of stratified analyses for the CWR index (per 0.01-unit increase) and its association with 30-day, 90-day, 180-day, and 365-day all-cause mortality

Performance comparison

To evaluate the unique predictive value of CWR in the critically ill heart failure cohort of this study, the predictive performance of CWR, estimated glomerular filtration rate (eGFR), serum creatinine (Cr), and body weight (BW) for mortality was compared. ROC curve analysis indicated that CWR demonstrated the highest AUC values across all observed time points (detailed data can be found in Supplementary Materials Figs. 2–5)

Discussion

This study constitutes the inaugural retrospective analysis examining the prognostic association between CWR levels and all-cause mortality in critically ill patients with HF. Our analysis demonstrated that patients in the highest quartile (Q4) of the CWR index had significantly lower survival rates at 30, 90, 180, and 365 days compared to those in the lower quartiles.

Further analysis suggests that the CWR-mortality relationship is not a simple monotonic or uniform nonlinear association, but instead follows a biphasic risk pattern with a distinct threshold at CWR = 0.05. In most patients (CWR ≤ 0.05), mortality risk rises with increasing CWR until plateauing, possibly reflecting a saturable stress response due to imbalance between renal clearance and metabolic demand in acute heart failure. By contrast, CWR > 0.05 may signal a transition to a distinct high‑risk phenotype, potentially marked by severe circulatory compromise, irreversible organ injury, or profound catabolism. Thus, beyond its continuous prognostic value, CWR offers a clinically actionable binary threshold (0.05) that could help identify patients at extreme risk who may benefit from intensified monitoring.

Among 4,371 critically ill heart failure patients, 320 (7.3%) exhibited CWR values > 0.05. This subgroup reflects the most severe mismatch between renal function and body mass, demonstrating the ability of CWR to identify patients at the extreme end of the risk spectrum—a key aim of prognostic tools in critical care. While peak risk was concentrated in this high‑CWR subgroup, a smooth, continuous dose‑response pattern was observed across the full cohort on both sides of 0.05, indicating that the nonlinear association reflects a general trend rather than an artifact of sparse or outlying data.

Furthermore, subgroup analyses were performed across age, sex, peripheral vascular disease, chronic pulmonary disease, liver disease, diabetes, and obesity. The association between CWR and mortality was not statistically significant in patients with liver disease at any time point, nor in those aged ≤ 65 years at 90 days. In all remaining subgroups, CWR showed statistically significant associations with mortality across time points. These results may help inform clinical strategies to reduce mortality in this population.

The relationship between CWR index and the mortality rate of critically ill patients with HF

CWR, an easily accessible clinical parameter that has been established as an independent predictor of adverse outcomes across various diseases [20, 21]and demonstrates documented prognostic value in cardiovascular and endocrine disorders, remains inadequately studied in heart failure populations. Our study first applies this index for prognostic assessment in HF, confirming its significant association with all-cause mortality and offering a novel stratification perspective. Subgroup analyses reveal that CWR maintains prognostic significance across multiple timepoints (30-, 180-, 365-day), except in liver disease patients; at 90 days, significance persists except in those ≤ 65 years or with liver disease. Patients in the highest CWR quartile demonstrate significantly reduced short- and long-term survival (Fig. 2). RCS analyses further identify a robust biphasic pattern: risk increases with CWR until plateauing around 0.05, beyond which a sharp mortality escalation suggests a distinct high-risk phenotype (Figure. 3 A-D).

When CWR is ≤ 0.05, the observed rise in mortality risk to a plateau may reflect the combined contribution of both elevated creatinine and low body weight. Serum creatinine is a well-established biomarker for predicting adverse outcomes in patients with heart failure, as impaired renal function, indicated by elevated creatinine, is consistently associated with increased mortality in congestive heart failure populations [22]. This association was further corroborated by Grace L et al. in an analysis of 80,098 heart failure patients, which demonstrated that each 0.5 mg/dl increase in serum creatinine was linked to a 15% rise in mortality risk (95% CI: 14% to 17%) [23].

Serum creatinine, a core indicator of renal function and a key integrator of hemodynamic and metabolic stress, also serves as a significant marker of myocardial status. Vedat Cicek et al. [24] identified creatinine as a crucial biomarker for predicting myocardial injury after non-cardiac surgery (MINS) in elderly patients; when incorporated into the CLASHED model, it may help identify high-risk individuals and guide preoperative interventions such as renal optimization and hemodynamic management, potentially reducing MINS incidence. Similarly, in the study by Hayıroğlu et al. [25], serum creatinine elevation—a core criterion for defining acute kidney injury (AKI)—was shown to strongly predict long-term mortality in patients with ST-segment elevation myocardial infarction (STEMI) complicated by cardiogenic shock, further supporting the link between CWR and long-term outcomes. Moreover, within the validated ACEF risk score, the highest-risk tertile (T3) exhibited markedly elevated creatinine levels (mean 1.42 mg/dL) and the greatest in-hospital mortality (86.8%), underscoring the established role of creatinine in predicting short-term mortality in cardiogenic shock [26] and reinforcing the relevance of CWR for short-term risk stratification in comparable high-risk settings.

Obesity increases the risk of developing heart failure (HF), yet in established HF it is associated with improved survival - a phenomenon termed the “obesity paradox.” [27] Compared to normal-weight patients (BMI 18.5–25 kg/m²), overweight (BMI 25–30 kg/m²) and class I obese (BMI 30–35 kg/m²) individuals exhibit lower all‑cause and cardiovascular mortality, suggesting that higher body weight may confer a prognostic advantage, while lower weight is linked to greater risk [28]. This evidence indicates that the combination of high Cr and low BW may underlie the positive association between the CWR index and mortality risk in HF patients when CWR drops below a certain threshold.

Comparison of predictive performance: CWR vs. eGFR, serum Creatinine, and body weight

To assess the distinctive predictive value of CWR in this critically ill heart failure cohort, we directly compared it with commonly used clinical indicators of renal function: eGFR (calculated using the MDRD formula), Cr, and BW. ROC curve analysis (Supplementary Figs. 2–5) demonstrated that, for all‑cause mortality at 30, 90, 180, and 365 days, the AUC for CWR consistently exceeded those of the other three measures (Figure. 2). For instance, in predicting 365‑day mortality, the AUC for CWR was 0.637, compared with 0.617 for eGFR, 0.602 for serum Cr, and 0.593 for BW.

These results suggest that, in critically ill heart failure patients, the composite CWR index may offer superior risk discrimination compared with either of its individual components (Cr or BW) or even the eGFR formula - which incorporates age, sex, and race. This advantage may stem from CWR’s integration of creatinine-based renal information and weight-based nutritional-metabolic information, potentially aligning with two central pathophysiological pathways in critically ill heart failure: cardiorenal interplay and energy-reserve depletion. Therefore, CWR could serve as a convenient bedside risk-stratification tool, though its clinical utility warrants validation in prospective studies. While eGFR remains a well-established and superior standard in stable chronic kidney disease populations, CWR may provide a simpler yet robust alternative for risk stratification in the heterogeneous, critically ill heart failure population encountered in the ICU setting.

Differentiated prognostic role of creatinine in short-term vs. long-term mortality

Our findings indicate that the prognostic value of creatinine—and consequently of the CWR—may vary between short-term (30-day) and long-term (365-day) outcomes. In fully adjusted models, the hazard ratios associated with the highest CWR quartile were numerically higher for 30-day mortality (HR 2.022, 95% CI 1.533–2.665) than for 365-day mortality (HR 2.100, 95% CI 1.724–2.559), though both remained statistically significant. This temporal pattern likely reflects different underlying pathophysiological mechanisms:

In the short term—specifically within the initial 30 days—elevated creatinine often signals acute cardiorenal stress, such as hemodynamically mediated renal hypoperfusion, acute kidney injury, or volume overload. These conditions pose an immediate threat in critically ill heart failure patients.

Beyond the initial 30 days, during the long-term (up to 365 days), creatinine may instead represent chronic renal impairment and sustained uremic burden. This contributes to progressive myocardial fibrosis, arrhythmia susceptibility, and increased vulnerability to recurrent decompensation.

Therefore, while CWR consistently predicts mortality across time horizons, its clinical interpretation may evolve from reflecting acute hemodynamic compromise in the early phase to indicating chronic organ dysfunction and metabolic dysregulation over extended follow-up.

Potential mechanisms of high levels of cr and poor prognosis in HF

Elevated Cr levels are associated with adverse prognosis in HF, which may be explained in part by the pathophysiology of cardiorenal syndrome. This involves two main processes: (1) hemodynamic abnormalities with neuroendocrine overactivation; (2) accumulation of uremic toxins from renal impairment, which directly inhibits myocardial contractile function.

Hemodynamic changes combine with neuroendocrine activation to form a vicious cycle

On one hand, reduced cardiac output and systemic hypotension can lead to renal hypoperfusion. On the other hand, elevated central venous pressure and increased intra-abdominal pressure contribute to renal venous outflow obstruction and renal interstitial congestion. These hemodynamic abnormalities collectively impair the structure and function of nephrons, resulting in a decreased glomerular filtration rate and consequent elevation in Cr levels [29–31]. Meanwhile, the renal impairment indicated by elevated Cr levels can further activate the renin-angiotensin-aldosterone system (RAAS) through mechanosensitive and chemosensitive signaling at the juxtaglomerular apparatus, specifically via baroreceptor-mediated sensing of reduced renal perfusion pressure and chemoreceptor-mediated sensing of low sodium chloride delivery at the macula densa. Angiotensin II induces widespread arteriolar vasoconstriction, leading to a marked increase in peripheral vascular resistance, elevated cardiac afterload, and heightened myocardial oxygen consumption, thereby accelerating the progression of heart failure. Concurrently, aldosterone promotes sodium and water reabsorption in the distal tubules and collecting ducts, resulting in volume overload, increased venous return to the heart, and elevated ventricular wall stress. These effects collectively exacerbate myocardial injury and contribute to maladaptive cardiac remodeling and deterioration of cardiac function [32, 33].

Therefore, Cr serves not only as a reflective indicator of cardiac and renal function but also a key biomarker linking HF and renal impairment, playing a significant mediating role in their bidirectional regulatory mechanisms.

The accumulation of uremic toxins produces direct toxic effects on the myocardium

Elevated Cr levels have relatively low direct toxicity. However, they often occur alongside the buildup of other uremic toxins like urea, guanidine compounds, and phenols. These substances promote systemic inflammation, speed up atherosclerosis, and directly inhibit myocardial contractility. This can lead to cardiomyocyte death and interstitial fibrosis [34, 35]. Chronic uremia puts a persistent burden on cardiac structure and function. It also triggers pathological changes that promote cell death. For example, secondary hyperparathyroidism can cause left ventricular hypertrophy. Over time, this may progress to poor myocardial adaptation, apoptosis, and even myocardial infarction, rapidly worsening HF [36].

Elevated Cr levels reflect a convergence of the pathophysiological mechanisms linking renal impairment to adverse outcomes in HF, establishing Cr not merely as a sensitive indicator of disease severity but also as a significant biomarker for predicting disease progression.

Potential mechanisms of low levels of BW associated with poor outcomes in HF

Based on the widely documented “obesity paradox,” body mass index (BMI)—a measure of weight adjusted for height—is conventionally used as a body composition indicator. In adults with relatively stable height, body weight (BW) correlates positively with BMI. Consequently, the obesity paradox may also be applicable to BW, as low BW has been associated with poor prognosis in heart failure (HF). This association may be explained by two potential aspects: (1) energy and nutritional imbalances during decompensated HF, a phase characterized by symptom worsening, which is often linked with low BW; and (2) a diminished buffering capacity against harmful metabolic substances due to low BW, reflecting impaired physiological resilience.

Firstly, HF is recognized as a catabolic state. Patients with higher BW, such as those who are overweight or obese, may possess greater metabolic reserves, which could enhance their ability to withstand the effects of this catabolic condition. In contrast, low BW may indicate more severe chronic HF and could reflect underlying cardiac wasting or cachexia, which is itself associated with unfavorable outcomes in HF [37]. Secondly, evidence suggests that higher BW might serve as a “metabolic buffer,” potentially mitigating the severity of catabolic complications and cachexia in advanced HF [38, 39]. In the setting of HF, harmful substances such as inflammatory cytokines can accumulate in the body. Adipose tissue, owing to its lipophilic properties, may temporarily sequester these compounds, thereby reducing their systemic impact and potentially protecting organs, including the heart, from injury. Conversely, lower BW may weaken this buffering capacity.

Nevertheless, the precise mechanisms underlying the obesity paradox in HF remain incompletely understood, and direct clinical evidence supporting this hypothesis remains limited. Despite this, multiple clinical studies consistently report a significant association between severe underweight status and elevated all-cause mortality risk in HF patients compared to those with normal or overweight BMI [40]. Notably, this protective effect is not observed in patients with extreme obesity.

The obesity paradox is a well-recognized phenomenon in HF and critical care medicine. In this study, the proposed CWR integrates renal function (assessed by serum creatinine) with body weight. Its predictive performance for mortality may be influenced by the complex phenotypes associated with the obesity paradox, meaning that the risk signal captured by CWR likely reflects a composite of renal function, nutritional status, and body composition. Clinically, the implications of CWR may vary across different body composition profiles. In non-obese or underweight HF patients, lower BW is often related to sarcopenia or malnutrition. In this context, a higher CWR may more strongly indicate an intensified catabolic state and compromised physiological reserve, serving as a sensitive marker for poor prognosis. In contrast, in obese patients, higher BW may lead to a relatively lower CWR due to a “dilution” effect on creatinine. If obesity itself confers a survival advantage in this HF population, the ability of CWR to accurately reflect renal impairment-related risk may be attenuated.

To evaluate the independent prognostic value of CWR, this study adjusted for BW and the SAPS-II score (an indicator of acute illness severity) in multivariable analyses. This approach aims to control for confounding related to body size and disease acuity, thereby isolating the independent contribution of CWR to mortality risk. It should be noted, however, that because BW is already incorporated in the CWR calculation, including it as a covariate may help partially address confounding from physique. Nonetheless, the physiological basis of the obesity paradox is multifactorial and likely involves aspects beyond BW alone, such as body composition and metabolic adaptations. Therefore, when interpreting the prognostic significance of CWR, careful consideration should be given to its nature as a composite index that integrates information on renal function, nutritional status, and body composition interplay.

Subgroup analysis

To assess the consistency of the association between the CWR index and all-cause mortality across different time points in HF patients, we performed subgroup analyses based on age, gender, peripheral vascular disease, chronic pulmonary disease, liver disease, diabetes, and obesity. The results showed that while the association did not reach statistical significance in patients with liver disease at any time point (P ≥ 0.05), interaction tests revealed no significant effect modification by liver disease status. This suggests that the CWR–mortality relationship remains consistent irrespective of liver disease, with the nonsignificance in this subgroup likely attributable to limited statistical power due to its smaller sample size (n = 465, 10.6% of the cohort). Similarly, in patients aged ≤ 65 years, the association was not significant at 90 days. In all other subgroups, CWR showed statistically significant associations with mortality across time points (P < 0.05). Interaction analyses further indicated that, apart from age and obesity, no other predefined subgroups significantly modified the association between CWR and mortality. The observed effect modification by age may be explained by differing risk profiles: in younger HF patients, CWR may more strongly predict mortality, potentially because younger individuals face fewer competing risks, and nutritional‑muscular status exerts a clearer influence on prognosis—where a low CWR could signal sarcopenia, itself linked to poor outcomes. In older patients, however, the presence of multiple comorbidities (e.g., dementia, cancer, multi‑organ dysfunction) complicates the mortality trajectory, potentially attenuating the standalone explanatory power of CWR. Additionally, the higher baseline prevalence of sarcopenia in the elderly may reduce the discriminative capacity of CWR within this subgroup.

The observed significant interaction effects for age and obesity suggest that the clinical interpretation of CWR may differ across these subgroups. In non-obese patients, a higher CWR may predominantly reflect reduced muscle mass and cachexia, serving as a surrogate marker for sarcopenia and nutritional depletion. In contrast, among obese patients, where body weight is elevated, CWR may be more indicative of renal dysfunction, as the denominator (body weight) is larger, potentially “diluting” the creatinine component. This dual interpretation underscores that CWR is not merely a unidimensional biomarker but a composite measure whose association with mortality varies across subgroups defined by body composition and age.

Limitations

The present study demonstrates an association between the creatinine-to-weight ratio (CWR) and all-cause mortality in heart failure patients based on a comprehensive real-world dataset; however, several limitations should be considered. First, the single-center, retrospective design using the MIMIC-IV database, which primarily includes critically ill ICU patients, may limit the generalizability of our findings. Consequently, the results may not be fully applicable to broader heart failure populations in general wards, outpatient settings, or other healthcare systems. Second, due to the structure of the database, precise causes of death were unavailable, preventing analysis of heart failure-specific mortality and limiting pathophysiological interpretation. Third, missing data led to the exclusion of approximately 20% of variables. Although statistical imputation was applied to reduce information loss, this method cannot completely substitute for actual data and may introduce uncertainty. Fourth, the CWR metric itself has inherent limitations. Similar to body mass index (BMI), CWR is a composite measure influenced by body composition but does not distinguish between muscle and adipose tissue. In individuals with obesity, a high fat mass may artifactually lower CWR values due to a “dilution” effect, potentially misrepresenting true muscle status and reducing the metric’s predictive validity in this subgroup. Fifth, the inclusion criterion requiring ICU hospitalization for more than 48 h, while ensuring data stability for longitudinal analysis, may exclude the most critically ill patients who died within the first two days. Thus, our findings regarding the prognostic value of CWR are most relevant to patients who survive the initial high-risk period and may not generalize to those with immediately fatal presentations. Sixth, although sensitivity analyses—including exclusion of extreme values and variation of model parameters—supported the robustness of the biphasic pattern with a CWR threshold of 0.05, the generalizability of this threshold requires validation in prospective cohorts. Seventh, while CWR showed better predictive performance for mortality than eGFR, serum creatinine, and body weight in this cohort, this comparison was based on internal validation within a single observational dataset. Whether CWR outperforms established multivariable risk scores (e.g., APACHE or dedicated heart failure scores) remains to be examined in external cohorts and prospective studies. Additionally, only the MDRD formula was used to estimate eGFR in this analysis; future studies could compare different eGFR equations with CWR. Finally, although the observed nonlinear relationship is consistent with the “muscle protection hypothesis,” it remains an association derived from observational data. The findings may be influenced by unmeasured confounding, and causal inference requires further validation through prospective studies incorporating direct measures of muscle mass and mechanistic investigations.

Conclusion

Our retrospective cohort study revealed CWR demonstrates a robust biphasic nonlinear association with mortality. A CWR of 0.05 emerged as a key threshold distinguishing a phase of rising risk plateau from an extreme high-risk state, suggesting its potential clinical relevance for risk stratification. Furthermore, the consistent association of CWR with both short- and long-term mortality underscores its utility as a stable prognostic marker across different clinical phases of HF. The differential mechanisms implicated in early vs. late mortality suggest that CWR may capture both acute hemodynamic stress and chronic metabolic burden, reinforcing its potential as a composite indicator in risk stratification. However, it is crucial to emphasize that our findings, derived from an observational study, demonstrate an association​ rather than a causal relationship. Therefore, the CWR index should not be used in isolation to guide treatment strategies at the present time. Future large-scale, multicenter, prospective studies are essential to validate these findings, elucidate the underlying mechanisms, and determine whether incorporating the CWR index into clinical algorithms can genuinely improve patient management and outcomes.

Authors’ contributions

All authors contributed to the study’s conception and design. DF, YH, JH, WH, CZ and XL prepared the material, collected data, analyzed it, and wrote the first draft. ZS critically revised the manuscript. All authors commented on previous versions. All authors read and approved the final manuscript. DF and YH contributed equally to this work and shared the first authorship.

Funding

This work was supported by the Mechanism of Pingtiao Particles in Ameliorating Hypertensive Myocardial Remodeling via Regulating the Keap1/Trx1/GPX4 Pathway to Inhibit Ferroptosis (20251029).

Data availability

MIMIC-IV data were obtained from https://mimic.mit.edu. Observed data of patients with HF are available from the corresponding authors upon reasonable request.

Declarations

Ethics approval and consent to participate

The establishment and use of the MIMIC database adhere to the ethical guidelines stipulated in the U.S. Code of Federal Regulations (45 CFR 46), commonly referred to as the Common Rule, as well as the provisions of the Health Insurance Portability and Accountability Act (HIPAA). The data were de-identified in accordance with the HIPAA “Safe Harbor” method, which involves the removal of 18 categories of personally identifiable information, including names and dates, thereby exempting the requirement for individual patient consent. The MIMIC database meets HIPAA standards for de-identification and aligns with the ethical principles set forth in the Declaration of Helsinki. The collection and dissemination of the data were granted an ethical exemption by the Institutional Review Boards (IRBs) of the Massachusetts Institute of Technology (MIT) (Protocol No. 0403000206) and Beth Israel Deaconess Medical Center (BIDMC) (Protocol No. 2001-P-001699/14). Furthermore, the author Huang of this study has successfully completed the CITI ethics training program (Certification No. 61042076), fulfilling the necessary data access requirements, and thus this study is exempt from further patient consent or ethical review.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.