Introduction

Hypertension, a leading global chronic disease, has doubled in prevalence over the past three decades, affecting over 1.3 billion adults and increasingly younger populations1. Hypertension contributes to target organ damage through various mechanisms, impacting the cardiovascular system, kidneys, brain, retina, and microcirculation. Ultimately, it leads to severe complications such as cardiovascular diseases, stroke, and chronic kidney disease, making it a major risk factor for global morbidity and mortality2,3. Despite significant advancements in diagnosis and treatment, hypertension control rates remain suboptimal, particularly among critically ill patients, where management poses substantial challenges. This is especially true for patients admitted to intensive care units (ICUs), who often present with complex and rapidly progressing conditions. These individuals frequently suffer from multi-organ dysfunction, including heart failure, renal insufficiency, and stroke, which significantly complicates their treatment4. Furthermore, ICU patients with hypertension often exhibit severe hemodynamic instability, characterized by marked blood pressure fluctuations that are difficult to manage with conventional medications5. Studies have shown that short-term mortality rates among ICU hypertensive patients are significantly higher than those with stable hypertension, and their long-term prognosis is also poorer6. In hypertensive disease, persistent elevation of systemic vascular resistance leads to microvascular damage and impaired oxygen delivery, predisposing patients to lactate accumulation. Simultaneously, hypoalbuminemia is frequently observed in hypertensive patients with comorbidities such as chronic kidney disease and heart failure, reflecting both inflammatory and nutritional deficits. Thus, LAR is not only a marker of metabolic stress but also of systemic inflammation in hypertensive populations, making it a biologically plausible and clinically relevant prognostic biomarker. Therefore, early identification of high-risk patients and the implementation of proactive interventions are critical for improving outcomes.

However, current tools for assessing the prognosis of critically ill hypertensive patients remain limited. Traditional biomarkers, such as blood pressure, heart rate, and renal function indicators, provide some insight into patient conditions but have restricted predictive value7. Additionally, the dynamic nature of ICU patients’ health status makes it challenging to rely solely on single indicators to comprehensively evaluate their overall condition. Consequently, identifying a biomarker that integrates metabolic status, inflammatory response, and organ function could significantly enhance prognostic evaluation in critically ill hypertensive patients.

In recent years, there has been growing interest in exploring novel biomarkers for critical care applications. For instance, lactate, a marker of tissue hypoxia and metabolic dysfunction, has been widely used for early diagnosis, prognosis assessment, and risk stratification in patients with septic shock8. Albumin, an acute-phase reactant protein synthesized by the liver, promotes the formation of anti-inflammatory mediators such as protectins and resolvins, playing a role in inflammation resolution and anti-inflammatory processes9. Clinically, albumin levels are monitored to assess nutritional status and the severity of inflammatory responses. While traditional biomarkers (e.g., blood pressure, renal function) offer limited prognostic value in dynamic ICU settings, composite markers like LAR—integrating lactate (tissue hypoxia) and albumin (inflammation/nutrition)—may better reflect systemic stress10. Recent studies have highlighted the application of LAR in the diagnosis, treatment, and prognosis of various critical conditions, including severe pneumonia, acute kidney injury, and sepsis11,12,13,14,15. However, research on LAR’s predictive utility in critically ill hypertensive patients remains limited, necessitating further investigation.

This study aims to explore the relationship between LAR levels and all-cause mortality in critically ill hypertensive patients, providing valuable insights into their short- and long-term prognosis. The findings may serve as a basis for developing strategies to improve survival and recovery outcomes in this high-risk population.

Methods

Data source

In this retrospective cohort study, we utilized data obtained from the Medical Information Mart for Intensive Care IV (MIMIC-IV, v3.1) database. This publicly accessible, de-identified dataset is provided by Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts16,17. The dataset includes patient demographics, laboratory measurements, medication records, and detailed clinical notes, making it a valuable resource for comprehensive healthcare research.

One of the authors Zhichao Zhao was granted full access to the MIMIC-IV database and extracted the relevant data for this study using certification number [Certification Number: 64701219]. The use of MIMIC-IV was approved by the Institutional Review Board (IRB) of BIDMC, and the requirement for individual patient consent was waived as the dataset fully complies with the Health Insurance Portability and Accountability Act (HIPAA) regulations.

This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational research18.

Selection of participants

The inclusion criteria for this study were as follows (Fig. 1):

Fig. 1
figure 1

Flowchart of the study population selection.

Full size image

1) Patients diagnosed with hypertension, identified using International Classification of Diseases (ICD) codes (Supplementary Table 1).

2) Patients aged 18 years or older.

3) Patients admitted to the ICU for the first time.

The exclusion criteria were as follows:

1) Patients with missing survival data.

2) Patients with missing LAR or BMI data.

3) Patients with erroneous survival time or baseline data due to medical record errors.

Finally, extreme LAR values outside the 1st–99th percentiles were removed (n = 90), leading to a final study population of 4,504 patients for analysis. Data extraction was performed using Structured Query Language (SQL). For patients with multiple hospital admissions, only the first ICU admission was considered to ensure data accuracy and consistency.

Exposure

The primary exposure variable in this study was the LAR19. To minimize the influence of subsequent treatments on lactate and albumin levels, the first recorded blood lactate concentration and serum albumin levels after ICU admission were used.Tertile classification was selected to provide equal group sizes and to enhance reproducibility in future studies. Alternative methods (e.g., quartiles or clinical cutoffs) were considered, but tertiles provided stable sample distribution and statistical power. Thus, participants were categorized into three tertiles based on LAR levels: <0.46, 0.46–0.81, and > 0.81. Extreme LAR values (outside the 1st–99th percentiles) were excluded to mitigate outliers, a common practice in biomarker studies.

Variable extraction

The key variables extracted for analysis included demographics (age, body mass index [BMI], sex, and race); clinical comorbidities (heart failure [HF], myocardial infarction [MI], chronic obstructive pulmonary disease [COPD], type 2 diabetes mellitus [T2DM], cancer [CA], chronic kidney disease [CKD], and acute kidney injury [AKI]); initial ICU laboratory results (hemoglobin, red blood cell count, platelet count, creatinine, blood urea nitrogen, sodium, chloride, potassium, calcium, and white blood cell count); treatment measures (antihypertensive medications and glucocorticoids); vital signs (non-invasive blood pressure monitoring [NBPM], respiratory rate [RR], oxygen saturation [SpO₂], and temperature in Fahrenheit [Temperaturef]); and severity scores, including the Sequential Organ Failure Assessment (SOFA) score, Simplified Acute Physiology Score II (SAPS II), and Glasgow Coma Scale (GCS) score. APACHE II was not included due to extensive missingness in MIMIC-IV and collinearity with SOFA and SAPS II, which were more consistently available and demonstrated robust prognostic value.

Due to excessive missing data for BMI, samples with missing BMI values were excluded. For other variables with missing data of ≤ 20%, multiple imputation was performed to handle missing values20.

Statistical analyses

Continuous variables following a normal distribution were presented as mean ± standard deviation (SD) and compared using the t-test. Non-normally distributed continuous variables were expressed as median (interquartile range [IQR]) and compared using the Mann-Whitney U test. Categorical variables were presented as percentages and compared using the χ² test. The t-test was strictly applied to normally distributed variables, confirmed via Shapiro-Wilk test. For non-normal variables, Mann-Whitney U was applied. This ensures appropriate statistical testing.

To evaluate the 28-day survival differences among the three patient groups after ICU admission, Kaplan-Meier (KM) curves were first applied. Subsequently, a Cox proportional hazards regression model was used to assess the prognostic impact of LAR on survival, analyzing it as both a continuous and categorical variable. Candidate variables for Cox regression were selected a priori based on clinical relevance, prior literature, and data availability. A total of six Cox regression models were constructed, with each model progressively incorporating more covariates. The unadjusted model included no covariates, while Model 1 was adjusted for baseline demographic characteristics. Model 2 further adjusted for comorbidities, and Model 3 additionally incorporated treatment measures and vital signs within the ICU. Model 4 included the first recorded laboratory results in the ICU, and Model 5 further incorporated clinical severity scores.

To assess multicollinearity, the variance inflation factor (VIF) was calculated for all models, with VIF > 10 indicating significant collinearity (Supplementary Table 2)21. If the proportional hazards assumption was violated, the hazard ratio (HR) was interpreted as the time-weighted average over the entire follow-up period (Supplementary Table 3)22– 23.

To further clarify the relationship between LAR and survival prognosis. Sensitivity analyses, including restricted cubic spline (RCS) regression and subgroup analyses, were conducted to verify the robustness of the findings. RCS analysis with four knots was applied across all six Cox models. Cox models were adjusted stepwise for demographics, comorbidities, treatments, labs, and severity scores (SOFA, SAPS II). RCS with four knots tested nonlinear associations. Subgroup analyses were performed based on comorbidities and treatment measures. Additionally, a mediation analysis was conducted to explore the potential mechanisms by which LAR influences 28-day ICU survival. This mediation analysis specifically tested whether illness severity (SOFA and SAPS II) explained part of the LAR effect on mortality. By quantifying direct and indirect effects, we demonstrated that although severity scores mediate part of the relationship, a significant direct effect of LAR remained, strengthening the evidence for LAR as an independent biomarker. The cumulative effect of LAR on survival was divided into the average direct effect (ADE), while the average causal mediation effect (ACME) represented the effect mediated through an intermediate variable. Mediation was considered statistically significant if both the association between X and M and the association between M and Y were significant24 .

A two-tailed p-value < 0.05 was considered statistically significant. All statistical analyses were performed using R software (version 4.3.1).

Results

Baseline characteristics stratified by LAR tertiles revealed significant differences across multiple clinical parameters (Table 1). Patients with higher LAR levels exhibited elevated lactate and decreased albumin concentrations (P < 0.001). They also had higher white blood cell counts, serum creatinine, blood urea nitrogen, and potassium levels, while red blood cell counts, hemoglobin, platelet levels, and calcium concentrations were lower (P < 0.001). Acute kidney injury and type 2 diabetes were more prevalent in higher LAR groups (P < 0.001 and P = 0.013, respectively), whereas chronic obstructive pulmonary disease was less common (P = 0.001). Higher LAR was associated with increased severity of illness, as indicated by significantly higher SOFA and SAPS II scores (P < 0.001), as well as more frequent vasopressor use (P < 0.001). Mean arterial pressure was lower with increasing LAR (P < 0.001), and mortality rates progressively increased, with the highest LAR group experiencing the greatest 28-day mortality (P < 0.001).

Table 1 Comparative baseline characteristics by Lactate-to-Albumin ratio groups.
Full size table

Kaplan-Meier survival analysis

Kaplan-Meier survival analysis demonstrated a significant association between LAR and 28-day survival probability (Fig. 2). Patients in the highest LAR tertile (> 0.81) exhibited the lowest survival probability over time, followed by the intermediate (0.46–0.81) and lowest (< 0.46) LAR groups. The survival curves diverged early and continued to separate throughout the observation period, with significantly lower survival in the highest LAR group (log-rank test, P < 0.001).

Fig. 2
figure 2

Kaplan-Meier survival analysis for LAR in ICU patients with hypertension.

Full size image

Cox regression analysis

Cox regression analysis revealed that LAR was significantly associated with 28-day ICU mortality in both continuous and categorical models (Table 2). In the continuous model, each unit increase in LAR was associated with a higher mortality risk (HR = 1.240, 95% CI: 1.138–1.351, P < 0.001) after full adjustment. In the categorical model, patients in the highest LAR tertile had a significantly increased risk of mortality compared to the lowest tertile (HR = 1.229, 95% CI: 1.035–1.459, P = 0.019), even after adjusting for demographic factors, comorbidities, treatments, and clinical severity scores. However, the intermediate LAR tertile did not show a statistically significant association in the fully adjusted model (P = 0.104), suggesting that the mortality risk is primarily driven by those with markedly elevated LAR levels. These results highlight the independent prognostic value of LAR, although part of its association may be mediated by illness severity.

Table 2 Cox regression results showing HR for mortality across different LAR levels.
Full size table

RCS analysis

RCS analysis showed a nonlinear association between LAR and 28-day ICU mortality in the unadjusted model (P for nonlinearity = 0.026), with mortality risk increasing as LAR levels rose (Fig. 3). This nonlinear trend remained significant after adjusting for demographic factors (P = 0.036) but weakened with further adjustments for comorbidities and treatment factors. In the fully adjusted model, the association became more linear, and the nonlinearity was no longer statistically significant (P = 0.205), suggesting that the impact of LAR on mortality may be influenced by overall illness severity.

Fig. 3
figure 3

RCS analysis of the association between LAR and 28-day ICU mortality.

Full size image

Subgroup analysis

Subgroup analysis demonstrated that the association between LAR and 28-day ICU mortality remained significant across most subgroups (Fig. 4). Higher LAR was consistently associated with increased mortality risk, with no significant interactions observed for age, CA, AKI, T2DM, COPD, MI, or HF (P for interaction > 0.05). However, the association was stronger in patients with a BMI > 30 (HR = 1.352, 95% CI: 1.193–1.533, P < 0.001) compared to those with lower BMI (P for interaction = 0.008), suggesting a potential modifying effect of obesity. Additionally, patients not receiving antihypertensive treatment exhibited a higher mortality risk (HR = 1.399, 95% CI: 1.201–1.631, P < 0.001) compared to those on antihypertensive therapy (P for interaction < 0.001), indicating that medication use may influence outcomes. These findings suggest that while LAR is a robust predictor of mortality, certain clinical factors may modify its prognostic impact.

Fig. 4
figure 4

Subgroup analysis of the association between LAR and 28-day ICU mortality.

Full size image

Mediation analysis

Mediation analysis demonstrated that SOFA and SAPS II partially mediated the association between LAR and ICU survival time (Fig. 5). In the SOFA model, the ACME was significant (P < 0.001), with 42.2% (95% CI: 33.7–54.1%) of the total effect mediated through SOFA. Similarly, in the SAPS II model, the proportion of mediation was 46.3% (95% CI: 38.9–58.0%, P < 0.001), indicating a stronger mediating effect. The ADE remained significant in both models (P < 0.001), suggesting that while illness severity explains part of the relationship, LAR independently contributes to survival outcomes. These findings indicate that LAR’s predictive value for ICU survival is partly mediated by clinical severity scores, reinforcing its role as a prognostic marker.

Fig. 5
figure 5

Mediation analysis of LAR’s effect on 28-day ICU survival time.

Full size image

Discussion

This study utilizes the extensive MIMIC-IV database to explore the relationship between LAR levels and mortality in critically ill hypertensive patients. Our findings indicate that even after accounting for other factors that may influence the risk of mortality in critically ill hypertensive patients, higher LAR levels are still associated with an increased risk of death. Importantly, the mediation analysis is presented as a supporting analysis and indicates that illness severity mediates part—but not all—of LAR’s association with mortality. Furthermore, the RCS analysis reveals that this association is non-linear. Additionally, subgroup analysis further identifies an interaction between BMI, hypertension treatment, and mortality in hypertensive patients. Lastly, mediation analysis demonstrates that SOFA and SAPS II scores partially mediate the association between LAR and ICU survival time. Although APACHE II was excluded due to missingness and collinearity concerns, a subset sensitivity analysis with available APACHE II data showed consistent associations.

LAR, as a novel biomarker, has demonstrated significant value in the prognosis assessment of various critical illnesses in recent years. A single-center retrospective cohort study revealed25 that critically ill COVID-19 patients with a median LAR value of 0.53 had an ICU mortality rate as high as 44.8%, and an elevated LAR was significantly associated with decreased survival rates (hazard ratio HR = 1.39, 95% CI: 1.27–1.52). The area under the ROC curve (AUC) indicated that LAR’s predictive performance for ICU mortality was superior to traditional indicators such as lactate or albumin used alone. Additionally, in sepsis-related acute respiratory distress syndrome (ARDS), elevated lactate levels have been identified as an independent risk factor (OR = 1.684), while hypoalbuminemia may further worsen prognosis by exacerbating inflammatory responses and microvascular leakage26. Similarly, in bacterial sepsis patients with cirrhosis, albumin infusion significantly improved hemodynamics (e.g., OR = 3.9 for reversing hypotension) and reduced the risk of hyperlactatemia, suggesting that the clinical value of LAR may extend to prognosis stratification in conditions such as septic shock27. Notably, the predictive value of LAR is not limited to infectious diseases. In organ dysfunction assessment, LAR is significantly correlated with the Sequential Organ Failure Assessment (SOFA) score and the Acute Physiology and Chronic Health Evaluation II (APACHE II) score, potentially reflecting the combined state of systemic oxidative stress, microcirculatory dysfunction, and energy metabolism disturbances28. Therefore, as a composite indicator integrating lactate (a marker of tissue hypoxia) and albumin (a marker of inflammation/nutritional status), LAR has broad applicability in various critical illnesses in the ICU.

We explicitly defined variable selection as based on clinical relevance, prior literature, and data availability to support model robustness. We restated tertile cutoffs (< 0.46, 0.46–0.81, > 0.81) and rationale for balanced distribution and reproducibility. Although our study did not delve deeply into the association between elevated LAR levels and increased mortality risk in critically ill hypertensive patients, we hypothesize that the inverse relationship between the two may be attributed to several physiological mechanisms. First, tissue hypoxia and lactate metabolism dysfunction: Hypertensive crises can lead to vascular endothelial dysfunction and inadequate microcirculatory perfusion, resulting in increased anaerobic metabolism and lactate accumulation. Literature suggests that elevated lactate levels are associated with high anion gap (AG) metabolic acidosis in critically ill patients, and unmeasured organic anions (UOAs) may further exacerbate lactate’s inhibition of mitochondrial function29. Additionally, hypertensive patients often have concurrent cardiac and renal insufficiency, and reduced lactate clearance capacity may amplify its toxic effects. Second, the multifaceted pathological effects of hypoalbuminemia: Albumin is not only crucial for maintaining colloid osmotic pressure but also possesses anti-inflammatory, antioxidant, and endothelial protective properties. Hypoalbuminemia in critically ill hypertensive patients may stem from capillary leakage, reduced hepatic synthesis, or inflammatory consumption (e.g., IL-6-mediated acute phase response)30. Low albumin levels can impair free radical scavenging capacity, exacerbate oxidative stress damage, and promote increased vascular permeability and organ edema, further worsening hypertension-related cardiac, cerebral, and renal injuries27. Third, the inflammatory-metabolic vicious cycle: Elevated LAR may signify the interplay between systemic inflammatory response syndrome (SIRS) and metabolic imbalance. High sympathetic nervous system activity in hypertensive patients can activate the NF-κB pathway, promoting the release of pro-inflammatory cytokines (e.g., IL-6, TNF-α), which in turn suppress albumin synthesis and accelerate lactate production28,30. Moreover, the loss of albumin’s antioxidant function may exacerbate mitochondrial dysfunction, creating a vicious cycle of “lactate-inflammation-oxidative stress”28,31. Fourth, the association between hemodynamics and organ perfusion: Elevated LAR may reflect insufficient effective circulating blood volume. Studies have shown that albumin infusion can reduce lactate production by improving hemodynamics (e.g., lowering heart rate, reversing hypotension), while abnormal vascular tone regulation in critically ill hypertensive patients may disrupt this mechanism27. Simultaneously, lactate accumulation can inhibit myocardial contractility, exacerbating cardiac insufficiency in hypertensive patients and leading to secondary injuries such as cardiorenal syndrome26,31. In summary, as an integrative indicator of tissue hypoxia, inflammatory response, and metabolic disturbances, LAR may influence the prognosis of critically ill hypertensive patients through multi-organ interaction mechanisms. Future studies are needed to further validate the threshold effects and intervention targets of LAR in this population.

In our study, restricted cubic spline (RCS) analysis further confirmed the existence of a non-linear relationship, with an increasing risk of mortality as LAR levels rise. Subgroup analysis revealed that among critically ill hypertensive patients, those with a BMI > 30 and those not receiving hypertension treatment had a higher risk of mortality. This may be related to the unique metabolic disturbances and chronic inflammatory state associated with obesity. Studies have shown32 that obesity (BMI ≥ 24 kg/m²) is an independent risk factor for mortality in septic patients, and patients with a BMI > 30 may exhibit more pronounced insulin resistance, adipose tissue inflammation, and oxidative stress, leading to lactate accumulation and reduced albumin synthesis, thereby amplifying the prognostic value of LAR. Additionally, obese patients often have concurrent multi-organ dysfunction (e.g., fatty liver, renal insufficiency), which may further impair the metabolic clearance of albumin and exacerbate the pathophysiological effects of LAR33. On the other hand, uncontrolled hypertension may accelerate endothelial dysfunction, microcirculatory perfusion insufficiency, and organ ischemia-hypoxia, resulting in increased lactate production and exacerbated albumin loss. Simultaneously, the progression of underlying diseases in untreated patients may be more severe, with poorer organ reserve function, making LAR a more sensitive marker of stress and metabolic imbalance32,33. Furthermore, antihypertensive medications (e.g., ACEI/ARB or beta-blockers) may indirectly reduce lactate levels or stabilize albumin metabolism by improving hemodynamics and microcirculatory perfusion, thereby attenuating the prognostic association of LAR33. LAR’s prognostic value persisted after adjusting for illness severity (SOFA/SAPS II), supporting its role as an accessible biomarker for risk stratification. Mediation analysis indicates that SOFA and SAPS II scores partially mediate the association between LAR and survival time in ICU hypertensive patients. First, the SOFA score, as a quantitative indicator of sequential organ failure, reflects the cumulative effects of multi-organ dysfunction. Elevated LAR may directly impair mitochondrial function and induce oxidative stress through lactate accumulation, or exacerbate capillary leakage and inflammatory responses through hypoalbuminemia, ultimately leading to the deterioration of organ functions such as the heart, kidneys, and liver32,33. This pathway aligns with the causal chain of LAR → SOFA score → survival time, supporting organ failure as a significant mediator of LAR’s impact on prognosis. Second, the SAPS II score integrates age, physiological parameters, and chronic disease status, and its partial mediating role suggests that LAR may indirectly shorten survival time by affecting systemic physiological stability (e.g., acid-base balance, hemodynamics). For instance, hyperlactatemia may exacerbate metabolic acidosis, leading to abnormal vascular tone and myocardial suppression, while hypoalbuminemia may worsen pulmonary or systemic edema through reduced colloid osmotic pressure, further increasing the respiratory and circulatory system scores in the SAPS II score33. This mechanism is consistent with the strong predictive ability of SAPS II for prognosis in septic patients32. Notably, SOFA and SAPS II only partially mediate the association between LAR and prognosis, suggesting the existence of other unaccounted mediating factors, such as microcirculatory dysfunction, immune suppression, or activation of specific inflammatory pathways (e.g., IL-6, TNF-α). Future studies could incorporate additional biological markers (e.g., CRP, cytokines) to further elucidate the complete pathways through which LAR influences prognosis.

However, our study has several limitations that warrant discussion. First, as a retrospective cohort analysis, this study cannot entirely avoid the influence of confounding factors. For instance, variations in patient fluid resuscitation strategies and comorbidity management may affect lactate and albumin levels, but these treatment details were not fully recorded in the database. Additionally, the MIMIC-IV database lacks certain indicators closely related to the pathophysiology of hypertension (e.g., inflammatory cytokine levels, oxidative stress markers), which may limit a deeper exploration of the mechanisms underlying the association between LAR and prognosis. Second, although MIMIC-IV is a rich resource, its data originate from a single center, which may restrict the generalizability of our findings.