Abstract
Background
To evaluate the association between lactate-to-albumin ratio (LAR) levels and clinical outcomes in patients with protein-energy malnutrition (PEM).
Methods
A retrospective analysis was conducted using the MIMIC-IV database. Patients with PEM were identified and stratified into four groups based on LAR quartiles. Cox regression models and restricted cubic spline (RCS) curves were employed to investigate the association between LAR and both 28-day ICU mortality and ICU mortality. Subgroup analyses were performed to compare the prognostic value of LAR across different populations.
Results
A total of 2,750 patients were included. Patients in the highest LAR quartile exhibited significantly higher 28-day ICU mortality and ICU mortality compared to those in the lowest quartile (both P < 0.001). After comprehensive adjustment for covariates, LAR remained significantly associated with 28-day ICU mortality (HR: 1.11, 95% CI: 1.01–1.22) and ICU mortality (HR: 1.19, 95% CI: 1.04–1.36) (both P < 0.05). Restricted cubic spline (RCS) curves indicated a non-linear relationship between LAR and both mortality outcomes. Sensitivity analyses further supported the robustness of the results.
Conclusion
LAR was independently associated with increased 28-day ICU and ICU mortality in patients with PEM, exhibiting a non-linear relationship. These findings highlight the potential of LAR as an indicator for risk stratification in the PEM population.
Supplementary Information
The online version contains supplementary material available at 10.1186/s40795-026-01399-w.
Keywords: Protein-energy malnutrition (PEM), Lactate-to-Albumin Ratio (LAR), MIMIC-IV, ICU, Mortality
Introduction
Protein-energy malnutrition (PEM) is characterized by the depletion of body protein mass and energy reserves, representing a severe and prevalent complication in numerous chronic and acute disease states [1, 2]. The etiology involves not merely inadequate nutrient intake, but typically encompasses the complex interplay of metabolic dysregulation, chronic inflammation, and hypermetabolism, particularly in the presence of underlying diseases. Among patients with chronic kidney disease (CKD), the prevalence of PEM increases significantly as renal function declines, affecting a substantial proportion of those undergoing maintenance dialysis [2, 3]. In this population, PEM often coexists with a persistent inflammatory state, forming a synergistic deleterious condition known as Malnutrition-Inflammation Complex Syndrome (MICS). This syndrome is a primary driver of accelerated cardiovascular disease progression and poor survival in this group [4–6]. Similarly, PEM is highly prevalent among patients with malignant tumors, where tumor-induced metabolic alterations, anorexia, and the side effects of anticancer therapy collectively contribute to cachexia [7].
The environment of the Intensive Care Unit (ICU) amplifies the threat posed by PEM. Critically ill patients experience a dramatic increase in muscle protein breakdown and energy expenditure, which more than doubles their protein requirements [8, 9]. This hypercatabolic state is driven by a systemic inflammatory response, where cytokines such as interleukin-6 and tumor necrosis factor-α suppress appetite, promote muscle breakdown, and lead to hypoalbuminemia [10]. Therefore, conventional nutritional support often fails to completely prevent the loss of lean body mass. The PEM observed in ICU patients is far from a passive state of nutritional deficiency; rather, it is an active, pathology-driven process that directly impacts clinical recovery. Studies indicate that PEM diagnosed upon ICU admission using tools such as the Subjective Global Assessment is independently associated with higher mortality, increased complication rates (e.g., pressure injuries), prolonged ICU and hospital stays, and higher readmission rates [11].
The Lactate-to-Albumin Ratio (LAR), as a composite index combining lactate and albumin, reflects metabolic stress, systemic inflammation, and nutritional status. In critical illnesses such as sepsis, multiple studies have confirmed that an elevated LAR at admission is a superior predictor of short-term mortality compared to lactate or albumin alone [12–14]. In addition to sepsis, LAR has been shown to be significantly and independently associated with increased mortality in patients with acute myocardial infarction, out-of-hospital cardiac arrest, and acute exacerbation of chronic obstructive pulmonary disease, as well as in critically ill patients with underlying conditions such as liver cirrhosis, coronary heart disease, and atrial fibrillation [15–20]. However, the prognostic value of LAR in the PEM population remains unclear. This study utilized the MIMIC-IV database to investigate whether the LAR measured within the first 24 h of ICU admission could predict clinical outcomes in patients who were diagnosed with PEM upon ICU admission through ICD coding. Our primary objective was to determine if this readily available metabolic-nutritional biomarker could serve as an effective risk stratification tool for this high-risk population with established malnutrition at the time of critical illness onset.
Methods
Data sources and study population
This study is a retrospective analysis based on the MIMIC-IV database (version 3.1). The MIMIC-IV database contains information on all patients treated at the Beth Israel Deaconess Medical Center (Boston, Massachusetts) between 2008 and 2022. The use of this database for research has been approved by the Institutional Review Boards of both the Massachusetts Institute of Technology and the Beth Israel Deaconess Medical Center. The requirement for written informed consent was waived due to the inability to identify patients through the health information in the database. The investigator (Zhang) has completed the training program provided by the collaborating institution (Certificate No. 69760166) and is qualified to use the database and extract data.
This study included adult patients admitted to the ICU for the first time with a length of stay of at least 24 h and a clinical diagnosis of protein-energy malnutrition as identified by ICD diagnostic codes(E43, E46, E440, 2639, 262, E441, 2638). The exclusion criteria were as follows: (1) age < 18 years or > 100 years; (2) ICU length of stay < 24 h; (3) missing lactate or albumin values in the first laboratory tests within 24 h of ICU admission.
Data collection
Data regarding patient demographics, comorbidities, clinical scores, initial vital signs in the ICU, laboratory examinations, and outcome variables were extracted from the MIMIC-IV database using the pgAdmin PostgreSQL tool. The specific data included: (1) Demographic data, including age, sex, weight, and height; (2) Comorbidities, including hypertension, diabetes mellitus, stroke, malignancy, chronic kidney disease, and sepsis; (3) Clinical scores, including the Sequential Organ Failure Assessment (SOFA) score and the Acute Physiology and Chronic Health Evaluation (APACHE) II score; (4) ICU vital signs, including heart rate, respiratory rate, peripheral oxygen saturation, mean arterial pressure, temperature, and initial 24-hour urine output; (5) Laboratory tests, including: Complete blood count: white blood cell count, lymphocytes, neutrophils, red blood cell count, platelets, and hemoglobin; Liver function tests: Glutamic-pyruvic transaminase, Glutamic-oxaloacetic transaminase, total bilirubin, and albumin; Renal function tests: creatinine and blood urea nitrogen; Electrolytes: calcium, potassium, sodium, chloride, and anion gap; Coagulation profiles: prothrombin time and activated partial thromboplastin time; Blood gas analysis: pH and lactate; Fasting blood glucose.
The primary and secondary outcomes were 28-day mortality and ICU mortality, respectively.
Statistical analysis
Variables with a missing rate > 20% were excluded, and remaining missing data were imputed using the multiple imputation method via the ‘missRanger’ package. Variance Inflation Factor (VIF) was used to check for multicollinearity among variables, and those with VIF values > 5 were excluded. Variables conforming to a normal distribution were described as mean ± standard deviation, and differences between the two groups were analyzed using independent sample t-tests. Continuous variables not conforming to a normal distribution were reported as median with interquartile range and analyzed using the Mann-Whitney U test. Categorical variables were reported as frequencies and percentages, and differences were analyzed using the Chi-square test.
To evaluate the association between LAR and 28-day ICU mortality as well as overall ICU mortality, four Cox proportional hazards models were constructed: Model I (adjusted for no covariates), Model II (adjusted for demographic features), Model III (adjusted for demographic features and comorbidities), and Model IV (adjusted for all covariates). Kaplan-Meier curves were generated to compare survival probabilities across different LAR quartiles.
The prognostic value of LAR and SOFA scores for 28-day ICU mortality was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC), and a combined analysis was performed.
Restricted cubic spline (RCS) curves with 4 knots were employed to explore the potential nonlinear relationship between LAR and both 28-day ICU mortality and ICU mortality. If a nonlinear relationship was identified, the “segmented” package was utilized to determine the optimal inflection point, and a two-piecewise Cox proportional hazards model was subsequently applied to examine the associations between the variable and the outcomes on either side of the threshold.
Subgroup and interaction analyses were performed to determine whether the associations between LAR and 28-day ICU mortality, as well as ICU mortality, differed across subgroups defined by age, sex, chronic kidney disease, malignant tumors, stroke, diabetes, and sepsis.
Sensitivity analyses were performed using complete data (after excluding missing values) to further investigate the association between LAR and both 28-day ICU mortality and overall ICU mortality, as well as to explore potential non-linear relationships. To validate the inflection point of LAR for 28-day ICU mortality identified by the segmented package and to establish a more robust clinical threshold, we performed a sensitivity analysis by maximizing the log-rank statistic.
All statistical analyses were performed using R version 4.5.1, and a P-value < 0.05 was considered statistically significant.
Result
Data processing
Based on the extent of missing data in this study (Table S1), three variables with a missing rate exceeding 20% were excluded: height, lymphocyte count, and neutrophil count. Multicollinearity among covariates in the imputed data was assessed using the Variance Inflation Factor (VIF) (Table S2), leading to the exclusion of two variables with a VIF greater than 5: red blood cell count and lactate.
Baseline and outcomes characteristics
A total of 4,343 confirmed patients were identified, and 2,750 patients were finally included after applying the exclusion criteria (Fig. 1). Table 1 presents the baseline characteristics of the cohort stratified by LAR quartiles. There were no significant differences in age, sex, or weight across the quartiles (all P > 0.05). The incidence of stroke was lower in the fourth quartile (P = 0.005), whereas the incidence of sepsis was higher (P < 0.001). Mean arterial pressure (MAP), body temperature, and 24-hour urine output were lower in the fourth quartile compared with the first quartile (all P < 0.001). Compared with the first quartile, the fourth quartile had higher SOFA and APACHE Ⅱ scores (all P < 0.001). Regarding laboratory results, there were no significant differences in serum chloride or sodium levels across the cohort (all P > 0.05), while the remaining parameters showed significant differences among the four quartiles (all P < 0.05). Regarding outcomes, the 28-day ICU mortality and ICU mortality in the fourth quartile were significantly higher than those in the first quartile (both P < 0.001) (Table S3).
Fig. 1.
The flowchart of study patients
Table 1.
Baseline characteristics of lactate to albumin ratio (LAR) quartiles
| Characteristic | Overall (N = 2750) |
Q1 (N = 689) |
Q2 (N = 690) |
Q4 (N = 685) |
Q4 (N = 686) |
p-value |
|---|---|---|---|---|---|---|
| Demographic features | ||||||
| Gender, male n (%) | 1,532 (55.7) | 362 (52.5) | 408 (59.1) | 390 (56.9) | 372 (54.2) | 0.0692 |
| Age, (years) | 67 (56, 76) | 67 (56, 77) | 68 (57, 77) | 66 (56, 76) | 66 (54, 75) | 0.1031 |
| Weight, (kg) | 72.4 (60.0, 87.9) | 71.0 (58.3, 87.0) | 71.8 (60.8, 87.2) | 71.8 (60.0, 87.1) | 74.9 (62.1, 88.5) | 0.0621 |
| comorbidities | ||||||
| Hypertension, n (%) | 910 (33.1) | 251 (36.4) | 224 (32.5) | 228 (33.3) | 207 (30.2) | 0.1012 |
| Diabetes, n (%) | 683 (24.8) | 150 (21.8) | 167 (24.2) | 187 (27.3) | 179 (26.1) | 0.0932 |
| Stroke, n (%) | 189 (6.9) | 66 (9.6) | 49 (7.1) | 35 (5.1) | 39 (5.7) | 0.0052 |
| Malignant tumor, n (%) | 474 (17.2) | 112 (16.3) | 120 (17.4) | 119 (17.4) | 123 (17.9) | 0.8692 |
| Chronic kidney disease, n (%) | 509 (18.5) | 127 (18.4) | 142 (20.6) | 113 (16.5) | 127 (18.5) | 0.2832 |
| Sepsis, n (%) | 2,211 (80.4) | 505 (73.3) | 547 (79.3) | 561 (81.9) | 598 (87.2) | < 0.0012 |
| Clinical score | ||||||
| SOFA scores | 6 (4, 9) | 5 (3, 7) | 5 (3, 8) | 7 (4, 10) | 8 (5, 11) | < 0.0011 |
| APACHE II scores | 23 (18, 28) | 21 (16, 25) | 22 (18, 27) | 23 (19, 28) | 26 (21, 30) | < 0.0011 |
| Vital signs | ||||||
| Heart rate, (bpm) | 94 (80, 110) | 88 (75, 103) | 93 (78, 108) | 95 (81, 111) | 101 (87, 118) | < 0.0011 |
| Respiratory rate, (bpm) | 20 (16, 25) | 20 (16, 23) | 20 (16, 25) | 20 (16, 25) | 20 (17, 25) | 0.0261 |
| Peripheral oxygen saturation, (%) | 97 (95, 100) | 97 (95, 100) | 97 (94, 100) | 97 (95, 100) | 98 (95, 100) | 0.3251 |
| Mean Arterial Pressure, (mmHg) | 81 (69, 93) | 84 (73, 96) | 81 (70, 93) | 80 (69, 92) | 77 (67, 90) | < 0.0011 |
| Body temperature, (℃) | 36.78 (36.50, 37.11) | 36.83 (36.50, 37.17) | 36.83 (36.56, 37.17) | 36.78 (36.44, 37.11) | 36.72 (36.44, 37.06) | < 0.0011 |
| Initial 24-hour urine output, (mL) | 1,210 (700, 1,905) | 1,375 (846, 2,150) | 1,300 (795, 1,982) | 1,205 (705, 1,800) | 935 (495, 1,600) | < 0.0011 |
| Laboratory tests | ||||||
| White blood cell, (K/uL) | 11.60 (7.80, 16.60) | 10.50 (7.20, 14.20) | 11.80 (8.00, 16.30) | 11.70 (7.80, 16.40) | 13.20 (8.30, 19.60) | < 0.0011 |
| Hemoglobin, (g/dL) | 9.7 (8.2, 11.4) | 10.0 (8.4, 11.7) | 9.7 (8.4, 11.3) | 9.6 (8.1, 11.5) | 9.4 (8.1, 11.3) | 0.0181 |
| Platelet, (K/uL) | 178 (111, 264) | 198 (139, 277) | 190 (119, 277) | 163 (99, 245) | 156 (89, 243) | < 0.0011 |
| Glutamic-pyruvic transaminase, (IU/L) | 29 (16, 63) | 24 (14, 48) | 26 (15, 54) | 31 (18, 68) | 36 (17, 111) | < 0.0011 |
| Glutamic-oxaloacetic transaminase, (IU/L) | 45 (24, 102) | 33 (20, 66) | 40 (23, 82) | 51 (28, 109) | 69 (29, 193) | < 0.0011 |
| Total Bilirubin, (mg/dL) | 0.8 (0.4, 2.2) | 0.6 (0.3, 1.1) | 0.7 (0.4, 1.6) | 1.0 (0.5, 3.1) | 1.3 (0.6, 3.3) | < 0.0011 |
| Albumin, (g/dL) | 2.7 (2.3, 3.2) | 3.1 (2.7, 3.5) | 2.8 (2.4, 3.2) | 2.7 (2.3, 3.1) | 2.4 (2.0, 2.8) | < 0.0011 |
| Calcium, (mg/dL) | 8.2 (7.6, 8.7) | 8.4 (7.9, 8.9) | 8.2 (7.6, 8.8) | 8.1 (7.6, 8.6) | 7.9 (7.3, 8.6) | < 0.0011 |
| Chloride, (mEq/L) | 102 (98, 107) | 102 (98, 106) | 102 (98, 107) | 103 (98, 107) | 102 (98, 107) | 0.6751 |
| Potassium, (mEq/L) | 4.1 (3.7, 4.7) | 4.00 (3.6, 4.5) | 4.10 (3.7, 4.6) | 4.20 (3.7, 4.7) | 4.20 (3.7, 4.8) | < 0.0011 |
| Sodium, (mEq/L) | 138 (134, 141) | 138 (135, 142) | 138 (134, 141) | 138 (134, 141) | 137 (134, 141) | 0.2381 |
| Plasma prothrombin time, (sec) | 15.3 (13.1, 19.7) | 13.8 (12.4, 16.1) | 15.1 (13.0, 19.3) | 15.6 (13.2, 20.2) | 17.8 (14.5, 22.3) | < 0.0011 |
| Activated partial thromboplastin time, (sec) | 32.5 (28.1, 41.0) | 31.0 (27.2, 37.2) | 31.9 (28.1, 40.6) | 32.4 (28.2, 40.6) | 35.6 (29.6, 46.3) | < 0.0011 |
| Creatinine, (mg/dL) | 1.1 (0.7, 1.8) | 0.9 (0.6, 1.5) | 1.0 (0.7, 1.7) | 1.1 (0.7, 1.8) | 1.3 (0.8, 2.1) | < 0.0011 |
| Urea nitrogen, (mg/dL) | 23 (14, 42) | 19 (12, 38) | 23 (14, 41) | 25 (15, 43) | 27 (16, 45) | < 0.0011 |
| Fasting plasma glucose, (mg/dL) | 130 (104, 170) | 120 (100, 148) | 127 (103, 159) | 137 (111, 180) | 141 (106, 197) | < 0.0011 |
| pH | 7.37 (7.30, 7.42) | 7.39 (7.33, 7.43) | 7.38 (7.33, 7.43) | 7.37 (7.31, 7.42) | 7.34 (7.26, 7.40) | < 0.0011 |
| Anion gap, (mEq/L) | 14 (12, 18) | 14 (11, 16) | 14 (11, 17) | 14 (12, 17) | 16 (13, 20) | < 0.0011 |
| Lactate to albumin ratio | 0.69 (0.45, 1.11) | 0.35 (0.29, 0.40) | 0.57 (0.50, 0.63) | 0.85 (0.76, 0.96) | 1.64 (1.30, 2.40) | < 0.0011 |
1Kruskal-Wallis rank sum test, 2Pearson’s Chi-squared test
Association between LAR levels and 28-day ICU and ICU mortality
Cox regression analyses were performed to evaluate the association between LAR levels and both 28-day ICU and ICU mortality (Table 2). In the fully adjusted model (Model 4), when analyzed as a continuous variable, the hazard ratio (HR) for LAR was 1.11 (95% CI: 1.01–1.22, P = 0.023) for 28-day ICU mortality and 1.19 (95% CI: 1.04–1.36, P = 0.013) for ICU mortality. When analyzed as a categorical variable, compared with the first quartile, the second, third, and fourth quartiles of LAR showed statistically significant associations with 28-day ICU mortality (Q2 HR: 1.47, 95% CI: 1.13–1.91; Q3 HR: 1.42, 95% CI: 1.08–1.86; Q4 HR: 1.90, 95% CI: 1.43–2.52; all P < 0.05). Similarly, for ICU mortality, the higher quartiles showed significant risks compared to the lowest quartile (Q2 HR: 1.69, 95% CI: 1.11–2.56; Q3 HR: 1.83, 95% CI: 1.21–2.79; Q4 HR: 2.91, 95% CI: 1.87–4.52; all P < 0.05).
Table 2.
Hazard ratios of LAR for the four cox proportional hazard models
| Characteristic | Model 1 | Model 2 | Model 3 | Model 4 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | p-value | HR | 95% CI | p-value | HR | 95% CI | p-value | HR | 95% CI | p-value | |
| 28-day ICU Mortality | ||||||||||||
| LAR as continuous | 1.25 | 1.18, 1.33 | < 0.001 | 1.27 | 1.20, 1.35 | < 0.001 | 1.27 | 1.20, 1.35 | < 0.001 | 1.11 | 1.01, 1.22 | 0.023 |
| LAR_ group as categorical | ||||||||||||
| Q1 | — | — | — | — | — | — | — | — | ||||
| Q2 | 1.70 | 1.32, 2.19 | < 0.001 | 1.67 | 1.29, 2.15 | < 0.001 | 1.66 | 1.28, 2.14 | < 0.001 | 1.47 | 1.13, 1.91 | 0.004 |
| Q3 | 1.71 | 1.33, 2.21 | < 0.001 | 1.74 | 1.34, 2.24 | < 0.001 | 1.71 | 1.32, 2.21 | < 0.001 | 1.42 | 1.08, 1.86 | 0.011 |
| Q4 | 2.60 | 2.04, 3.30 | < 0.001 | 2.66 | 2.09, 3.39 | < 0.001 | 2.58 | 2.03, 3.29 | < 0.001 | 1.90 | 1.43, 2.52 | < 0.001 |
| ICU Mortality | ||||||||||||
| LAR as continuous | 1.27 | 1.17, 1.37 | < 0.001 | 1.29 | 1.19, 1.39 | < 0.001 | 1.30 | 1.20, 1.41 | < 0.001 | 1.19 | 1.04, 1.36 | 0.013 |
| LAR_ group as categorical | ||||||||||||
| Q1 | — | — | — | — | — | — | — | — | ||||
| Q2 | 1.73 | 1.15, 2.59 | 0.008 | 1.67 | 1.12, 2.51 | 0.013 | 1.76 | 1.17, 2.64 | 0.007 | 1.69 | 1.11, 2.56 | 0.014 |
| Q3 | 1.95 | 1.31, 2.89 | < 0.001 | 1.92 | 1.29, 2.85 | 0.001 | 1.98 | 1.33, 2.95 | < 0.001 | 1.83 | 1.21, 2.79 | 0.005 |
| Q4 | 3.15 | 2.17, 4.57 | < 0.001 | 3.15 | 2.17, 4.57 | < 0.001 | 3.21 | 2.21, 4.67 | < 0.001 | 2.91 | 1.87, 4.52 | < 0.001 |
Abbreviations: CI Confidence Interval, HR Hazard Ratio
Kaplan–Meier survival curves (Fig. 2) demonstrated that in terms of 28-day ICU mortality and ICU mortality, survival rates in the second, third, and fourth LAR quartiles were significantly lower than those in the first quartile, with a statistically significant difference (P < 0.0001).
Fig. 2.
Kaplan-Meier survival curves for LAR quartile groups. (A 28-day ICU mortality, B ICU Mortality)
ROC analysis showed that the AUC (95% CI) for LAR, SOFA score, and their combination in predicting 28-day ICU mortality were 0.62 (0.60–0.65), 0.61 (0.59–0.64), and 0.64 (0.61–0.66), respectively (Figure S1).
Restricted cubic spline (RCS) curves were used to further assess the relationship, revealing a significant non-linear association for both outcomes (P for nonlinearity < 0.05) (Fig. 3). For 28-day ICU mortality, the optimal inflection point was identified at an LAR of 0.47 (Figure S2-A). Below this threshold (< 0.47), LAR was positively associated with mortality (HR: 26.72, 95% CI: 4.46–160.21, P < 0.001), with the LAR < 0.47 cohort comprising 746 patients, of whom 113 experienced 28-day ICU mortality (mortality rate: 15.2%); whereas no significant association was observed above the threshold (≥ 0.47) (HR: 1.08, 95% CI: 0.99–1.19, P = 0.096) (Table 3). For ICU mortality, the inflection point was 2.30 (Figure S2-B). LAR was significantly associated with mortality when < 2.30 (HR: 1.84, 95% CI: 1.42–2.39, P < 0.001), but not when ≥ 2.30 (HR: 0.94, 95% CI: 0.77–1.15, P = 0.552) (Table 3). Likelihood ratio tests (LRT) yielded P < 0.001 for both outcomes, indicating a significant difference in models stratified by the optimal inflection points (Table 3).
Fig. 3.
Restricted cubic splines for LAR. (A 28-day ICU mortality, B ICU Mortality)
Table 3.
Piecewise cox regression model
| Characteristic | HR | 95% CI | p-value |
|---|---|---|---|
| 28-day ICU Mortality | |||
| LAR (continuity) | 1.11 | 1.01, 1.22 | 0.023 |
| Inflection points of LAR | |||
| < 0.47 | 26.72 | 4.46, 160.21 | < 0.001 |
| ≥ 0.47 | 1.08 | 0.99, 1.19 | 0.096 |
| Log likelihood ratio | < 0.001 | ||
| ICU Mortality | |||
| LAR (continuity) | 1.19 | 1.04, 1.36 | 0.013 |
| Inflection points of LAR | |||
| < 2.30 | 1.84 | 1.42, 2.39 | < 0.001 |
| ≥ 2.30 | 0.94 | 0.77, 1.15 | 0.552 |
| Log likelihood ratio | < 0.001 |
Adjusted for all covariates
Abbreviations: CI Confidence Interval, HR Hazard Ratio
Subgroup analysis
Subgroup analyses were performed to further investigate the association between LAR levels and both 28-day ICU mortality and ICU mortality. In the analysis of 28-day ICU mortality (Fig. 4-A), no significant interaction effects were observed (all P for interaction > 0.05). In terms of ICU mortality (Fig. 4-B), a significant interaction effect was observed among patients with malignant tumors (P for interaction = 0.028).
Fig. 4.
Forest plot of LAR stratified by subgroup. (A 28-day ICU mortality, B ICU Mortality)
Sensitivity analysis
To assess the robustness of the findings, a sensitivity analysis was performed. After excluding missing data, 2,183 cases with complete data were included for Cox regression analysis (Table S4). After adjusting for all covariates, LAR levels remained statistically significant for 28-day ICU mortality (HR: 1.12, 95% CI: 1.01–1.24) and ICU mortality (HR: 1.21, 95% CI: 1.04–1.41) (both P < 0.05). A significant non-linear relationship was also observed (both P for nonlinearity < 0.05) (Figure S3).
For 28-day ICU mortality, we employed the method of maximizing the log-rank statistic to re-determine the optimal cut-off value for LAR and identified a new LAR inflection point of 1.18 (Figure S4). Below this threshold (< 1.18), LAR was positively associated with mortality (HR: 1.73, 95% CI: 1.28–2.35, P < 0.001), with the LAR < 1.18 cohort comprising 2138 patients, of whom 422 experienced 28-day ICU mortality (mortality rate: 19.7%); whereas no significant association was observed above the threshold (≥ 0.47) (HR: 1.03, 95% CI: 0.92–1.15, P = 0.645) (Table S5). The likelihood ratio test (LRT) yielded a P-value of 0.003, indicating a significant difference between the models stratified by 1.18(Table S5).
Discussion
This retrospective analysis of a large critical care database provides new evidence that elevated LAR is independently associated with increased short-term mortality in patients diagnosed with protein-energy malnutrition. After comprehensive adjustment for demographics, comorbidities, severity of illness scores, and laboratory parameters, each unit increase in LAR was associated with an 11% increased risk of 28-day ICU mortality and a 19% increased risk of overall ICU mortality. Compared to patients in the lowest LAR quartile, those in the highest quartile exhibited a 1.9-fold increase in 28-day ICU mortality and a 2.91-fold increase in overall ICU mortality. The relationship between LAR and mortality was non-linear, characterized by a distinct inflection point, suggesting a threshold effect. These findings indicate that LAR serves as an accessible and efficient prognostic predictor specifically for this high-risk population of critically ill malnourished patients.
Our study design warrants important methodological considerations regarding the identification of PEM and the use of albumin within the LAR metric. First, we identified PEM patients through ICD diagnostic codes rather than relying on biochemical markers alone. This approach reflects clinical practice where PEM diagnosis integrates multiple factors including patient history, physical examination findings, and laboratory parameters. However, we acknowledge that ICD-based identification may still have limitations compared to comprehensive nutritional assessments performed prospectively.
Regarding albumin as a component of LAR, we recognize that hypoalbuminemia in critical illness is multifactorial. This complexity is precisely why LAR's strength lies in its composite nature—by combining lactate (reflecting metabolic stress and tissue hypoperfusion) with albumin (reflecting both nutritional status and systemic inflammation), LAR captures a more comprehensive physiological picture than either marker alone. The extensive multivariable adjustment for comorbidities, organ dysfunction markers, and severity of illness provides robust statistical control for potential confounding factors. This analytical approach suggests that the prognostic association observed with LAR is likely independent of these underlying pathologies, though residual confounding cannot be entirely excluded in observational studies.
The independent prognostic value of the Lactate-to-Albumin Ratio (LAR) observed in malnourished patients within our study cohort aligns with and further extends the growing body of literature across various critical care settings. In general ICU populations, LAR has demonstrated superior or comparable performance in predicting mortality compared to lactate or albumin alone [21, 22]. Its utility is particularly pronounced in sepsis, where systematic reviews have confirmed LAR as a predictor of moderate strength for mortality [23, 24]. For instance, one meta-analysis reported a pooled diagnostic odds ratio of 5.23 for LAR in predicting sepsis mortality [3], while another study found an area under the curve (AUC) as high as 0.976 for predicting death versus discharge outcomes [23]. The prognostic significance of LAR also extends to specific organ failures common in critical illness, including acute kidney injury, acute respiratory distress syndrome, and sepsis-associated acute kidney injury [25–28]. Our study reinforces this association by confirming its validity even in a baseline population characterized by the significant comorbidity of protein-energy malnutrition.
Our analysis revealed a nonlinear, threshold-dependent relationship between LAR and mortality. This suggests that the risk attributable to LAR is not constant but increases sharply within specific ranges. Similar nonlinear associations have been reported in other patient populations, such as those with liver cirrhosis, where inflection points were identified for both 28-day and 90-day mortality [29]. In our PEM cohort, the inflection point for 28-day mortality was lower, which may reflect an exquisite sensitivity to subtle perturbations in the balance between metabolic stress and nutritional reserves. For individuals who are already malnourished, a slight increase in lactate relative to albumin may signal early, decompensated metabolic stress, accelerating the transition to poorer short-term survival. Conversely, the higher threshold for overall ICU mortality may indicate that patients who survive the initial 28 days possess distinct physiological reserves, or that other long-term risk factors modulate the LAR relationship.
Furthermore, our initial piecewise Cox regression analysis, which utilized the segmented package to identify the inflection point of LAR for 28-day ICU mortality, yielded a highly unstable hazard ratio (HR) with an exceptionally wide confidence interval in the LAR < 0.47 group. This instability is likely attributable to the limited number of events within the very low LAR range, which challenges the robustness of parametric modeling in that region. Recognizing this limitation, we conducted a sensitivity analysis using the maximally selected rank statistics method, which provided a more robust and clinically plausible threshold of 1.18. We therefore emphasize the latter finding for clinical interpretation, while acknowledging the former as an exploratory result that highlights the statistical challenges inherent in defining thresholds within regions of sparse data.
The pathophysiological basis of LAR's prognostic capability lies in its integration of two key pathways: tissue hypoperfusion/metabolic stress and systemic inflammation/nutritional depletion. Elevated lactate levels serve as a marker of anaerobic metabolism, typically resulting from tissue hypoxia, increased glycolysis, or impaired hepatic clearance [30]. Hypoalbuminemia in critical illness is multifactorial, primarily driven by increased capillary permeability, systemic inflammation leading to albumin redistribution into the extravascular space, and impaired hepatic synthesis [31, 32]. In the context of pre-existing PEM, low albumin levels are further exacerbated by chronic nutritional and protein deficits [33]. Consequently, a high LAR value may reflect a synergistic injury: acute metabolic derangement superimposed on a chronic basis, alongside severely compromised systemic inflammation and nutritional status. This composite metric may reflect overall physiological reserve and injury severity more accurately than its individual components, which explains its consistently superior performance observed in multiple studies [34, 35].
From a clinical perspective, the simplicity and rapid availability of LAR offer practical advantages for risk stratification in the complex ICU setting. For clinicians managing critically ill patients with known malnutrition, our findings suggest that calculating LAR upon ICU admission helps identify those at the highest risk of early mortality. This may prompt more intensive monitoring or earlier consideration of advanced supportive therapies. While established scoring systems such as SOFA or APACHE II provide comprehensive assessments, LAR can serve as a rapid, bedside adjunct, particularly in resource-limited settings or for quick initial triage.
Our study design has several limitations. First, as a retrospective analysis of a single-center database, the results are susceptible to unmeasured confounders, thereby limiting the generalizability of the conclusions. Second, despite our efforts to address missing data through multiple imputation and sensitivity analysis using complete cases, a significant number of patients were excluded due to missing lactate or albumin values. While we employed robust statistical methods to handle missing data under the assumption of missing at random (MAR), we cannot definitively rule out the possibility that missing data patterns were related to underlying patient characteristics or disease severity (i.e., missing not at random). This potential selection bias may limit the external validity of our findings, particularly if patients with missing data differed systematically from those included in the analysis. Third, we utilized the first recorded lactate and albumin values within 24 hours of ICU admission and did not assess the dynamic changes in LAR levels over time. Future studies should prioritize prospective validation in multicenter cohorts of critically ill patients. Investigating the dynamic changes in LAR over time, rather than a single static measurement, could provide insights into therapeutic response. Further evaluating whether LAR can guide specific interventions, and exploring the combination of LAR with other biomarkers or clinical scores, may help develop more refined predictive models for this vulnerable patient population.
Conclusion
The lactate-to-albumin ratio (LAR) was identified as an independent risk factor for 28-day and ICU mortality in patients with protein-energy malnutrition, exhibiting a non-linear relationship. These findings highlight the potential of LAR as a clinically relevant biomarker for risk stratification in this patient population.
Supplementary Information
Acknowledgements
None.
Abbreviations
- MIMIC - IV
Medical Information Mart for Intensive Care
- PEM
Protein-energy malnutrition
- LAR
Lactate-to-albumin ratio
- RCS
Restricted Cubic Spline
- SMD
Standardized Mean Differences
- ICU
Intensive Care Unit
- HR
Hazard Ratio
Authors’ contributions
Conceptualization, Investigation, Data Curation: Zhenpeng Xiao and Liangliang Zhang; Writing - Original Draft, Writing - Review & Editing: Wei Qiu and Yang Jiang; Data Curation, Methodology, Validation: Xinger Geng and Qian Yang; Review, Supervision: Anhua Tang, Youwei Lu and Huixia Wu; Review, Supervision, Funding Acquisition: Lin Zhang. All authors reviewed and approved the final manuscript.
Funding
None.
Data availability
All the data in this study were derived from the publicly available MIMIC-IV database. According to the database’s licensing agreement, the data involved in this study can be provided under the terms of the relevant agreement. Reasonable requests for data access can be made to the corresponding author.
Declarations
Ethics approval and consent to participate
This study is a retrospective analysis of the publicly available, anonymized MIMIC-IV database. The Institutional Review Boards (IRB) of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center provided ethical approval for the use of these databases in research. The IRBs of the Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center granted a waiver of informed consent due to the retrospective nature of the study and the use of de-identified data. All research operations comply with the principles of the Declaration of Helsinki and the requirements of the US federal regulations.
Core Link: MIMIC-IV Official Documentation (including IRB Approval Statement): https://mimic.mit.edu/docs/iv/.
The Declaration of Helsinki: https://www.wma.net/policies-post/wma-declaration-of-helsinki-ethical-principles-for-medical-research-involving-human-subjects/.
The Common Rule of the United States (45 CFR Part 46): https://www.hhs.gov/ohrp/regulations-and-policy/regulations/45-cfr-46/index.html.
Consent for publication
Not applicable. This study is a retrospective analysis of anonymized data extracted from the Medical Information Mart for Intensive Care (MIMIC‑IV) database. The dataset contains de‑identified patient information with all personal identifiers removed. No individual patient data, images, or other personal or clinical details that could compromise anonymity are presented in this manuscript. Therefore, consent for publication is not required.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Agarwal E, Miller M, Yaxley A, et al. Malnutrition in the elderly: a narrative review. Maturitas. 2013;76(4):296–302. 10.1016/j.maturitas.2013.07.013. [DOI] [PubMed] [Google Scholar]
- 2.Lodebo BT, Shah A, Kopple JD. Is it important to prevent and treat protein-energy wasting in chronic kidney disease and chronic dialysis patients? J Ren Nutr. 2018;28(6):369–79. 10.1053/j.jrn.2018.04.002. [DOI] [PubMed] [Google Scholar]
- 3.Piratelli CM, Telarolli Junior R. Nutritional evaluation of stage 5 chronic kidney disease patients on dialysis. Sao Paulo Med J. 2012;130(6):392–7. 10.1590/s1516-31802012000600006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Merchant RA, Vathsala A. Healthy aging and chronic kidney disease. Kidney Res Clin Pract. 2022;41(6):644–56. 10.23876/j.krcp.22.112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Kalantar-Zadeh K. Recent advances in understanding the malnutrition-inflammation-cachexia syndrome in chronic kidney disease patients: What is next? Semin Dial. 2005;18(5):365–9. 10.1111/j.1525-139X.2005.00074.x. [DOI] [PubMed] [Google Scholar]
- 6.Kalantar-Zadeh K, Mehrotra R, Fouque D, et al. Metabolic acidosis and malnutrition-inflammation complex syndrome in chronic renal failure. Semin Dial. 2004;17(6):455–65. 10.1111/j.0894-0959.2004.17606.x. [DOI] [PubMed] [Google Scholar]
- 7.Sanz Ortiz J, Moreno Nogueira JA. García de Lorenzo y Mateos A. Protein energy malnutrition (PEM) in cancer patients. Clin Transl Oncol. 2008;10(9):579–82. 10.1007/s12094-008-0253-8. [DOI] [PubMed] [Google Scholar]
- 8.Genton L, Pichard C. Protein catabolism and requirements in severe illness. Int J Vitam Nutr Res. 2011;81(2–3):143–52. 10.1024/0300-9831/a000058. [DOI] [PubMed] [Google Scholar]
- 9.Hoffer LJ, Bistrian BR. Why critically ill patients are protein deprived. JPEN J Parenter Enter Nutr. 2013;37(3):300–9. 10.1177/0148607113478192. [DOI] [PubMed] [Google Scholar]
- 10.Kalantar-Zadeh K, Stenvinkel P, Pillon L, et al. Inflammation and nutrition in renal insufficiency. Adv Ren Replace Ther. 2003;10(3):155–69. 10.1053/j.arrt.2003.08.008. [DOI] [PubMed] [Google Scholar]
- 11.Ferrie S, Weiss NB, Chau HY, et al. Association of subjective global assessment with outcomes in the intensive care unit: a retrospective cohort study. Nutr Diet. 2022;79(5):572–81. 10.1111/1747-0080.12767. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Xiao X, Wu JJ, Liu Y, et al. Increased lactate dehydrogenase to albumin ratio is associated with short-term mortality in septic ICU patients: a retrospective cohort study. Med (Baltim). 2024;103(52):e40854. 10.1097/MD.0000000000040854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cakir E, Turan IO. Lactate/albumin ratio is more effective than lactate or albumin alone in predicting clinical outcomes in intensive care patients with sepsis. Scand J Clin Lab Invest. 2021;81(3):225–9. 10.1080/00365513.2021.1901306. [DOI] [PubMed] [Google Scholar]
- 14.Cheng J, Hou J, Wang Y. Prognostic utility of the lactate-to-albumin ratio for predicting 28-day all-cause mortality in critically ill cases with acute sepsis: a retrospective study on the basis of MIMIC-IV critical care database. J Med Biochem. 2026;44(9):1943–50. 10.5937/jomb0-59662. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Chen Y, Lai W, Yang K, et al. Association between lactate/albumin ratio and prognosis in patients with acute myocardial infarction. Eur J Clin Invest. 2024;54(1):e14094. 10.1111/eci.14094. [DOI] [PubMed] [Google Scholar]
- 16.Wang J, Chen X, Qin C, et al. Lactate-to-albumin ratio as a potential prognostic predictor in patients with cirrhosis and sepsis: a retrospective cohort study. BMC Infect Dis. 2025;25(1):223. 10.1186/s12879-025-10601-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Liu Y. Association between lactate/albumin ratio and 28-day mortality in ICU critical patients with coronary heart disease: a retrospective analysis of the MIMIC-IV database. Front Cardiovasc Med. 2024;11:1486697. 10.3389/fcvm.2024.1486697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wu R, Xing B, Zhou Z, et al. Lactate-to-albumin ratio and 28 day mortality in hypertensive patients with atrial fibrillation: a retrospective cohort study. Eur J Med Res. 2025;30(1):845. 10.1186/s40001-025-03170-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ding CW, Huang SS, Xu YH, et al. Lactate dehydrogenase to albumin ratio and prognosis in patients with acute exacerbation of chronic obstructive pulmonary disease: a retrospective cohort study. BMC Pulm Med. 2025;25(1):154. 10.1186/s12890-025-03622-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Liu L, Ma Q, Yu G, et al. The association between the lactate to albumin ratio and all-cause mortality in cardiac arrest patients: an analysis of the MIMIC-IV database. Eur J Med Res. 2025;30(1):712. 10.1186/s40001-025-02986-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gharipour A, Razavi R, Gharipour M, et al. Lactate/albumin ratio: an early prognostic marker in critically ill patients. Am J Emerg Med. 2020;38(10):2088–95. 10.1016/j.ajem.2020.06.067. [DOI] [PubMed] [Google Scholar]
- 22.Suzuki Y, Aoki Y, Shimizu M, et al. Predictive accuracy of lactate albumin ratio for mortality in intensive care units: a nationwide cohort study. BMJ Open. 2024;14(12):e088926. 10.1136/bmjopen-2024-088926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Yoon SH, Choi B, Eun S, et al. Using the lactate-to-albumin ratio to predict mortality in patients with sepsis or septic shock: a systematic review and meta-analysis. Eur Rev Med Pharmacol Sci. 2022;26(5):1743–52. 10.26355/eurrev_202203_28244. [DOI] [PubMed] [Google Scholar]
- 24.Zhao X, Peng Q, Li W, et al. Elevated lactate/albumin ratio is associated with poor prognosis in sepsis patients: a systematic review and meta-analysis. J Med Biochem. 2024;43(4):334–49. 10.5937/jomb0-42284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Shi X, Zhong L, Lu J, et al. Clinical significance of the lactate-to-albumin ratio on prognosis in critically ill patients with acute kidney injury. Ren Fail. 2024;46(1):2350238. 10.1080/0886022X.2024.2350238. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Zhu X, Xue J, Liu Z, et al. The lactate/albumin ratio predicts mortality in critically ill patients with acute kidney injury: an observational multicenter study on the eICU database. Int J Gen Med. 2021;14:10511–25. 10.2147/IJGM.S339767. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Wang HX, Huang XH, Ma LQ, et al. Association between lactate-to-albumin ratio and short-time mortality in patients with acute respiratory distress syndrome. J Clin Anesth. 2024;99:111632. 10.1016/j.jclinane.2024.111632. [DOI] [PubMed] [Google Scholar]
- 28.Wang Y, Yu H. Association between lactate to albumin ratio and mortality among sepsis associated acute kidney injury patients. BMC Infect Dis. 2025;25(1):414. 10.1186/s12879-025-10838-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Ye Y, Huang S, Wang X, et al. Association between lactate-to-albumin ratio and all-cause mortality in cirrhosis patients: analysis of the MIMIC-IV database. Med Intensiva (Engl Ed). 2025;49(10):502145. 10.1016/j.medine.2025.502145. [DOI] [PubMed] [Google Scholar]
- 30.Deulkar P, Singam A, Mudiganti VNKS, et al. Lactate monitoring in intensive care: a comprehensive review of its utility and interpretation. Cureus. 2024;16(8):e66356. 10.7759/cureus.66356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Erstad BL. Serum albumin levels: who needs them? Ann Pharmacother. 2021;55(6):798–804. 10.1177/1060028020959348. [DOI] [PubMed] [Google Scholar]
- 32.Allison SP, Lobo DN. The clinical significance of hypoalbuminaemia. Clin Nutr. 2024;43(4):909–14. 10.1016/j.clnu.2024.02.018. [DOI] [PubMed] [Google Scholar]
- 33.Fukushima R, Nakamura K. Preexisting malnutrition in the ICU. Curr Opin Crit Care. 2026;32(2):200–7. 10.1097/MCC.0000000000001371. [DOI] [PubMed] [Google Scholar]
- 34.Ray CC, Pollack MM, Gai J, et al. The association of the lactate-albumin ratio with mortality and multiple organ dysfunction in PICU patients. Pediatr Crit Care Med. 2023;24(9):760–6. 10.1097/PCC.0000000000003272. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bou Chebl R, Jamali S, Sabra M, et al. Lactate/albumin ratio as a predictor of in-hospital mortality in septic patients presenting to the emergency department. Front Med (Lausanne). 2020;7:550182. 10.3389/fmed.2020.550182. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All the data in this study were derived from the publicly available MIMIC-IV database. According to the database’s licensing agreement, the data involved in this study can be provided under the terms of the relevant agreement. Reasonable requests for data access can be made to the corresponding author.