Introduction

Non-traumatic intracerebral hemorrhage (ICH) is a serious medical condition characterized by the sudden bleeding into the brain tissue without any history of trauma1. It represents a significant cause of morbidity and mortality worldwide. According to the Global Burden of Disease Study 20,192, among all stroke cases globally, intracerebral haemorrhage constituted 27.9% (3.41 million [2.97–3.91]) and subarachnoid haemorrhage constituted 9.7% (1.18 million [1.01–1.39])1,2,3,4. ICH is associated with a high risk of adverse outcomes, including neurological deficits, disability, and death3.In patients with non-traumatic ICH, the occurrence of acute kidney injury (AKI) is a common complication that further exacerbates the clinical course and prognosis. Studies have shown that up to 23% of patients with ICH develop AKI during their hospitalization, and this condition is associated with increased mortality and prolonged hospital stay5,6,7,8. The pathophysiology of AKI in the setting of ICH is multifactorial and may include hypoperfusion, systemic inflammation, and nephrotoxic medication use6.Treatment options for patients with non-traumatic ICH and AKI are limited, and the use of osmotic agents such as mannitol has been a cornerstone of management to reduce intracranial pressure and improve cerebral perfusion. Mannitol is an osmotic diuretic that works by drawing water out of the brain tissue, leading to a decrease in intracranial pressure and improved cerebral blood flow9. However, the use of mannitol is not without potential risks, including electrolyte imbalances, dehydration, and renal dysfunction. Previous studies have shown that mannitol can lead to renal dysfunction through several potential mechanisms. For instance, Lin et al.10 found that mannitol infusion was associated with an increased risk of AKI in patients with acute stroke, potentially due to its osmotic effects on the renal tubules, leading to cellular damage and inflammation. Fang et al.11 reported that mannitol was an independent risk factor for AKI following cerebral trauma. The current literature has established that mannitol can be effective in reducing intracranial pressure and improving cerebral perfusion in patients with non - traumatic ICH. However, the impact of mannitol therapy on the prognosis of patients with non - traumatic ICH who also develop AKI remains unclear. Most existing studies have focused on the efficacy of mannitol in reducing intracranial pressure or its potential nephrotoxic effects in isolation, but there is limited research exploring how mannitol use influences clinical outcomes in patients with the combined conditions of non - traumatic ICH and AKI. This knowledge gap makes it challenging for clinicians to make informed decisions about the use of mannitol in this specific patient population. In this context, the present study aims to investigate the effect of mannitol use on the clinical outcomes of patients with non - traumatic ICH and AKI, using a retrospective cohort design, in order to fill this gap and provide more evidence - based guidance for clinical practice. We hypothesize that the use of mannitol may be associated with worse clinical outcomes in patients with non - traumatic ICH and AKI, potentially due to its nephrotoxic effects. This study aimed to evaluate the association between mannitol use and 28 - day mortality among patients with non - traumatic ICH and AKI. Secondary outcomes measures included ICU length of stay and the need for dialysis.

Methods

Subjects

This study utilized data from the MIMIC-IV (Medical Information Mart for Intensive Care IV) version 3.0 database. MIMIC-IV 3.0 contains comprehensive data from multiple ICUs within the Beth Israel Deaconess Medical Center, a large tertiary care hospital in Boston, USA. The hospital’s critical care department includes several specialized ICUs, such as Medical ICU (MICU), Cardiac ICU (CCU), Neurological ICU (NICU), and Surgical ICU (SICU). These ICUs admit a diverse range of patients. The dataset encompasses approximately 220,000 distinct hospital admissions of over 120,000 adult and pediatric patients admitted to the ICU between 2008 and 2022, covering a wide array of data types, such as demographics, vital signs, laboratory results, and medications12.The author(WC) joined and completed online the course on protecting human research participants proposed by the US National Institutes of Health’s, and therefore obtained permission to access the dataset. The certificate number is 10,311,970. The use of the database for research was then authorized by the review committee of MIT and Beth Israel Deaconess Medical Center, and a waiver of informed consent was also granted to us. We followed the STROBE guidelines for observational research13.We enrolled ICH patients on the basis of the International Classification of Diseases (ICD)-9/10 guidelines, including ICD-9 code 431 and ICD-10 codes I610–I619 and I62.9 for ICH. Acute kidney injury is defined according to KDIGO criteria14. Specifically, it is based on an increase in serum creatinine level of at least 0.3 mg/dL within 48 h compared to baseline, or urinary output below 0.5 mL/kg/h for 6 h. Baseline serum creatinine was determined as follows: for patients with prior outpatient creatinine measurements, the most recent value within the past 365 days was used as the baseline. For patients without prior lab results, the lowest creatinine level measured during the first 7 days of hospitalization was considered the baseline creatinine. This approach aligns with KDIGO guidelines and ensures accurate identification of AKI. To ensure the study’s integrity and the robustness of its findings, we adhered to stringent exclusion criteria. (1) Individuals under the age of 18 at the time of admission were excluded. (2) For patients with multiple ICU admissions, only the first admission was considered. (3) Patients with end-stage renal disease were not included. (4) Those with ICU stays of less than 24 h were excluded due to often incomplete records and insufficient time to assess the impact of mannitol therapy. (5) patients missing critical admission-day data, such as demographic information, vital signs, laboratory test results (e.g., serum creatinine levels for AKI diagnosis), and treatment details (e.g., mannitol usage), were not included. These variables are essential for accurately evaluating the study outcomes and the effects of mannitol therapy.

Data extraction

In our study, we utilized Structured Query Language (SQL) scripts obtained from the GitHub repository (https://github.com/MIT-LCP/mimic-iv) to extract data from the MIMIC-IV database. The specific scripts used for data extraction were the standard SQL scripts provided by MIMIC-IV on GitHub. No modifications were made to these scripts.We collected comprehensive patient data, including characteristics like age, sex, and BMI. Our focus included vital signs within the first 24 h of ICU admission, such as heart rate (HR), mean arterial pressure (MAP), SpO2, temperature, and respiration rate (RR), as well as laboratory tests including hemoglobin, platelet count, white blood cell (WBC) count, sodium, potassium, calcium, blood urea nitrogen (BUN), creatinine, glucose, and blood gas analysis. We also assessed neurological status using the Glasgow Coma Scale (GCS), disease severity scores like the Simplified Acute Physiology Score (SAPS) II and the Sequential Organ Failure Assessment (SOFA) score, and whether the patient had sepsis. Additionally, we recorded details about the patient’s ICU stay, specifically noting if they required mechanical ventilation or underwent surgical procedures. Comorbidities such as hypertension, coronary heart disease, congestive heart failure, diabetes mellitus, and liver disease were also extracted to provide a complete picture of the patient’s health status.

Mannitol exposure

Mannitol exposure is defined as all prescriptions containing mannitol after admission to the ICU, including details such as dosage, frequency, duration of treatment, and the time relative to ICU admission.

Main results

The primary endpoint of this study is the 28-day all-cause mortality rate after admission to ICU. Secondary outcomes include in-hospital mortality.

Statistical analysis

This retrospective analysis was conducted without a predefined statistical analysis plan, given the nature of the study design. A power calculation was not deemed necessary due to the observational nature of the research, and the sample size was determined by the extent of data availability extracted from the existing database.

The study cohort was divided into two groups: those who received mannitol therapy (mannitol group) and those who did not (non-mannitol group). To handle missing data while preserving the dataset’s integrity and statistical power, we used a multiple imputation strategy. Variables with a missingness rate exceeding 50% were excluded from analysis (see Table S1 for details). For variables with a missingness rate of less than 50%, we applied the chained equations (MICE) method for multiple imputation. This covered vital signs (HR, MAP, SpO₂, temperature, RR) and lab tests (hemoglobin, platelet count, WBC count, sodium, potassium, calcium, BUN, creatinine, glucose, blood gas analysis). We created five imputed datasets and pooled results using Rubin’s rules to ensure analysis reliability and validity, accounting for the uncertainty from missing data. To safeguard against multicollinearity among the predictive variables, which could compromise the validity of the regression models, the variance inflation factor (VIF) was calculated for each variable included in the analysis. A VIF threshold of less than five for each variable served as evidence that multicollinearity was not a concern in the dataset, ensuring that the relationships between variables were appropriately modeled without distortion caused by redundant predictors.

Numbers (percentages) were utilized to present categorized data, while mean ± standard deviation or median (interquartile range) were used for continuous data. The Shapiro-Wilk test was used to assess the normality of continuous variables. Based on the results, parametric tests (analysis of variance test) were used for normally distributed data, whereas non-parametric tests (rank sum test) were applied for non-normally distributed data. The characteristics of the subjects in the outcome group were compared using the Chi-square test or Fisher’s exact test for categorical variables. KM survival curves were plotted for various subgroups to display survival rates. These curves were compared using the log-rank test. Univariate Cox regression analyses were performed to preliminarily explore the associations of mannitol use with 28-day all-cause mortality rate after admission to ICU and in-hospital mortality in ICH patients with AKI. Variables with statistical significance (P < 0.05) in the univariate analyses were subsequently included in the multivariate Cox regression models. The outcomes presented hazard ratios (HRs) along with 95% confidence intervals (CIs). The multivariate Cox regression was adjusted for potential confounders, including age, gender, Heart Rate (HR), Mean Arterial Pressure (MAP), Respiratory Rate (RR), Platelet count, Sodium, Calcium total, Chloride, PH, Lactate, Urea nitrogen, Creatinine, The Sequential Organ Failure Assessment (SOFA), Simplified Acute Physiology Score II (SAPS II), Glasgow Coma Score (GCS), Charlson Index, Ventilation. Additionally, subgroup analyses, stratified by relevant covariates, were conducted to further reduce survival bias and enhance the robustness of the findings. Subgroup analyses were stratified according to relevant covariates to reduce the impact of survival bias. All analyses were performed using Free statistics software version 1.6 and the statistical package R 4.1.1. A two-tailed test indicating P < 0.05 is considered statistically significant15.

Subgroup analyses

Subgroup analyses in the cohort were based on age (< 65 vs. ≥ 65 year), sex (female vs. male), GCS(< 9 vs. ≥ 9)16,AKI stage(1,2,3)14,Charlson Comorbidity Index (< 6 vs. ≥ 6)17, and Simplified Acute Physiology Score II (< 40 vs. ≥ 40)18.

Results

Patient selection

In the MIMIC-IV (version 3.0) database, a total of 1,685 patients were admitted to the ICU due to ICH with AKI. After excluding patients younger than 18 years old, those with non-first ICU admissions, ICU stays shorter than 24 h, and patients with uremia, the final cohort consisted of 1,042 ICH patients with AKI. Among these, 115 patients (11.0%) were exposed to mannitol, while 927 (89.0%) were not (Fig. 1).

Fig. 1
figure 1

Schematic diagram of study sample selection steps. MIMIC, Medical Information Mart for Intensive Care; ICU, intensive care unit; ICH, Intracerebral Hemorrhage; AKI, acute kidney injury.

Full size image

Cohort characteristics

Table 1 presents the patients’ baseline characteristics. In the entire cohort, patients who received mannitol were generally younger and more likely to be female. They also had a higher Charlson comorbidity index, lower WBC counts and platelet counts, and higher BUN levels. Mean arterial pressure was significantly higher in the mannitol group, whereas oxygen saturation was significantly lower. Additionally, patients in the mannitol group had significantly higher serum calcium levels and were more likely to require mechanical ventilation, with longer ventilation durations. In contrast, patients not receiving mannitol had significantly higher Glasgow scores and serum sodium levels and included more patients with hypertension. Regarding mannitol usage in the mannitol group, the median (IQR) time from ICU admission to first mannitol dose was 0.7 (0.3, 1.9) days. The total median (IQR) dose of mannitol administered was 100.0 (75.0, 200.0) grams. The median (IQR) daily medication frequency was 1.0 (1.0, 3.5) doses per day.

Table 1 Baseline characteristics between mannitol and no mannitol.
Full size table

Univariate Cox regression analysis

Table 2 illustrates the findings of a univariate Cox regression analysis, which identified several significant variables associated with increased 28-day mortality risk in patients with ICH and AKI post-ICU admission, as compared to the non-mannitol user reference group. These variables include elevated heart rate, mean arterial pressure, respiratory rate, sodium and chloride levels, pH values, lactate levels, urea nitrogen, creatinine levels, and higher SAPS II and SOFA scores, as well as the utilization of ventilation.

Table 2 Univariate Cox regression analysis for predictors of 28-day mortality Post-ICU admission in patients with ICH and AKI.
Full size table

Outcome

The total cohort exhibited a 28-day mortality rate of 25%. Notably, the mannitol group reported a significantly higher mortality rate of 50.4% (58 out of 115), compared to 21.9% (203 out of 927) in the control group. Figure 2 provides a graphical representation of the Kaplan-Meier survival curve for 28-day all-cause mortality, stratified by mannitol use. The association between mannitol use and elevated 28-day all-cause mortality was confirmed through both multivariable analysis (HR 1.94, 95% CI 1.38–2.72, p < 0.001) and univariable analysis (HR 2.42, 95% CI 1.80–3.25, p < 0.001). Furthermore, the proportional hazards assumption was tested using Schoenfeld residuals. The global test was significant (p = 0.004), suggesting a possible violation of the proportional hazards assumption. This finding should be interpreted cautiously and considered in the context of the model’s limitations. Detailed findings are tabulated in Table S2 and Table 3.

Fig. 2
figure 2

Kaplan-Meier survival analysis curves for all-cause mortality within 28-d of hospital admission.

Full size image
Table 3 Multivariate COX analysis of risk factors for death in patients within 28-d and in-hospital by logistic regression analysis.
Full size table

Subgroup analyses

Table 4 displays the findings from analyses examining the 28-day mortality rates across various subgroups within our study cohort. These analyses accounted for a range of factors, such as age, gender, HR, MAP, RR, platelet count, total calcium, chloride, pH, lactate, urea nitrogen, creatinine, SOFA, SAPS II, GCS, Charlson Index, and ventilation, to mitigate potential confounding influences. The outcomes are categorized based on pre-defined subgroups, including age, gender, GCS, AKI stage, SAPSII, SOFA, and Charlson index. The interaction analysis summaries are provided in Table S3. The data indicates that the use of mannitol is associated with increased mortality rates in specific subgroups. Particularly, patients with KDIGO stage 1 AKI who received mannitol had a significantly elevated mortality risk (HR = 2.52; 95% CI: 1.18–5.39; p < 0.05). A similar pattern was observed in patients with KDIGO stage 2 AKI, where mannitol use was associated with higher mortality rates (HR = 1.89; 95% CI: 1.17–3.05; p < 0.05). Conversely, among patients with KDIGO stage 3 AKI, mannitol use tended towards lower mortality rates, albeit not significantly (HR = 0.76; 95% CI: 0.30–1.87; p = 0.74). Notably, the interaction analysis revealed a statistically significant difference in the impact of mannitol across different AKI stages (P for interaction = 0.035), suggesting that the relationship between mannitol use and mortality varies considerably depending on the AKI stage.

Table 4 Subgroup - Specific analysis of mannitol use and mortality.
Full size table

Furthermore, a significant interaction was identified regarding the SOFA score (P for interaction = 0.009). Patients with a SOFA score < 5 experienced a significantly increased mortality risk with mannitol use (HR = 2.96; 95% CI: 1.91–4.58; p < 0.001), whereas those with a SOFA score ≥ 5 exhibited a non-significant increase in mortality risk (HR = 1.32; 95% CI: 0.75–2.32; p = 0.32).

For other subgroups, such as age, gender, GCS, SAPSII, and Charlson index, the P-values for interaction did not reach statistical significance (all P > 0.05), indicating that the effect of mannitol does not significantly vary across these subgroups after accounting for multiple covariates.

Sensitivity analyses

To account for survival bias, a sensitivity analysis excluding patients who died within 48 h of ICU admission was conducted. Mannitol use remained independently associated with increased 28-day mortality (adjusted HR 1.89, 95% CI 1.35–2.65; p < 0.001), after adjusting for age, gender, HR, MAP, RR, platelet count, calcium, chloride, pH, lactate, urea, creatinine, SOFA, SAPS II, GCS, Charlson Index, and ventilation.(Table 5).

Table 5 Sensitivity analysis after excluding patients who died within 48 h of ICU admission.
Full size table

Secondary outcomes

The in-hospital mortality rate was 47.8% (55 out of 115) in the mannitol treatment group and 16.2% (150 out of 927) in the non-treatment group. Univariate analysis revealed that mannitol use was associated with a higher in-hospital mortality rate (HR = 2.97; 95% CI: 2.17–4.07; P < 0.001). Multivariate analysis further confirmed this association, with an adjusted hazard ratio of 2.69 (95% CI: 1.90–3.80; P < 0.001) (Table 2).

Discussion

The current study findings indicate that the use of mannitol is associated with an increased 28-day all-cause mortality in patients with non-traumatic intracerebral hemorrhage (ICH) and acute kidney injury (AKI). Specifically, the 28-day mortality rate in the mannitol treatment group was 50.4% (58/115), whereas it was 21.9% (203/927) in the group without mannitol. Univariate analysis showed that mannitol use was associated with higher mortality (HR = 2.42; 95% CI: 1.80–3.25; P < 0.001), and multivariate analysis further confirmed this result (HR = 2.31; 95% CI: 1.67–3.19; P < 0.001). In subgroup analyses, mannitol was associated with increased mortality in specific subgroups, particularly in patients with KDIGO stage 1 and 2 AKI. However, in KDIGO stage 3 AKI patients, mannitol use showed a non-significant trend towards reduced mortality, although this result was not statistically significant (HR = 0.76; 95% CI: 0.30–1.87; P = 0.74). Additionally, in patients with a SOFA score < 5, mannitol use was associated with a significantly increased mortality risk (HR = 2.96; 95% CI: 1.91–4.58; P < 0.001), whereas in patients with a SOFA score ≥ 5, the trend of increased mortality with mannitol was not significant (HR = 1.32; 95% CI: 0.75–2.32; P = 0.32). For other subgroups (including age, gender, GCS, SAPSII, and Charlson index), the P-values for interaction were not statistically significant (all P > 0.05), indicating no significant difference in the effect of mannitol across these subgroups after adjusting for multiple covariates.

In the multivariate analysis, we adjusted for a series of potential confounding factors, including age, gender, heart rate (HR), mean arterial pressure (MAP), respiratory rate (RR), platelet count, total calcium, chloride, pH, lactate, urea nitrogen, creatinine, SOFA score, SAPS II score, GCS score, Charlson index, and ventilation status. The selection of these covariates was based on factors previously identified in the literature as being related to the prognosis of ICH and AKI, as well as variables that showed association with mortality in univariate analysis (P < 0.05). Additionally, we considered key confounding factors such as GCS score, ventilation status, pH, and lactate levels, which have significant impacts on prognosis in ICH and AKI patients. Through stepwise selection and multicollinearity testing (variance inflation factor VIF < 5), we ensured the rationality and independence of covariates in the final model. We also conducted a proportional hazards assumption test using Schoenfeld residuals. The global test was significant (p = 0.004), suggesting a potential violation of the proportional hazards assumption. This finding should be considered when interpreting the results of the Cox model.

For the inconsistent results in subgroup analyses, such as the non-significant trend towards reduced mortality with mannitol in KDIGO stage 3 AKI patients, we speculate that this may be due to: (1) insufficient statistical power due to the small sample size; (2) KDIGO stage 3 AKI patients might have received more aggressive renal replacement therapy (RRT), which could have obscured the potential adverse effects of mannitol; (3) in patients with severe renal insufficiency, the metabolism and excretion of mannitol may be significantly affected, altering its pharmacodynamic properties. Additionally, patients with a SOFA score < 5 generally represent those with milder organ dysfunction, and mannitol use may have a more pronounced negative impact on these patients, whereas in patients with a SOFA score ≥ 5, the effect of mannitol may be obscured by other factors due to the severity of underlying diseases.

Over the past 50 years, mannitol has been widely recognized as an effective treatment for brain edema, capable of reducing intracranial pressure (ICP) in various conditions19,20. It reduces the total water content in brain tissue and cerebrospinal fluid, decreases blood volume by inducing vasoconstriction21,22and improves cerebral perfusion by reducing viscosity or altering red blood cell rheology23. Furthermore, mannitol may offer protective effects against biochemical losses in stroke patients as a free radical scavenger24.While mannitol is effective in treating brain edema caused by various conditions, it is associated with several common side effects, including impaired kidney function25,26alterations in serum electrolytes and osmolality27blood volume overload, and rebound edema28. Most existing research has focused on the risk of AKI associated with mannitol use in patients with stroke or traumatic brain injury, along with the potential mechanisms underlying these effects10,11. However, the prognosis of patients with non-traumatic intracerebral hemorrhage (ICH) and AKI who receive mannitol has received less attention.

Based on previous research and conclusions, we infer that the poor prognosis of patients with ICH and AKI who receive mannitol may be due to the following reasons: (1) Cardiovascular Risk: In patients with pre-existing cardiac dysfunction or increased cardiac load28mannitol’s effect on improving cerebral perfusion can increase the risk of cardiovascular events. This risk is further heightened in patients with impaired renal function, who are more likely to experience heart failure or exacerbated cardiac dysfunction29. (2) Renal Tubular Cell Injury: The hyperosmotic properties of mannitol can cause injury to renal tubular cells, especially in patients with pre-existing renal impairment. This damage can further deteriorate renal function10,11,29,30. (3) Electrolyte Disturbances: Mannitol can lead to electrolyte imbalances, such as hyponatremia and hypokalemia, which can adversely affect cardiac and neurological functions27,30,31. (4) Reduced Efficacy in AKI Patients: In patients with AKI, the effectiveness of mannitol may be limited due to impaired renal function affecting its metabolism and excretion21,22,32.

The current study aimed to evaluate the association between mannitol use and 28-day mortality among patients with non-traumatic ICH and AKI. Secondary outcomes measures included ICU length of stay and the need for dialysis. Our findings suggest that mannitol use may be associated with worse clinical outcomes in patients with non-traumatic ICH and AKI, potentially due to its nephrotoxic effects.

Study limitations

Our study has several limitations that should be acknowledged. First, the data is derived from a single center (MIMIC-IV), which may limit the generalizability of our findings to broader populations. Second, despite our efforts to adjust for potential confounders, residual confounding may still be present, potentially influencing the results. Third, we lack longitudinal data on creatinine changes and mannitol accumulation, which could have provided valuable insights into renal function dynamics over time. Furthermore, our study did not include detailed information on key clinical confounders such as electrolyte trends and volume status, which are important factors in patient management and outcome assessment. Additionally, the absence of data on nephrotoxic medications and contrast examinations is a significant limitation, as their use could have a substantial impact on renal function and subsequent patient outcomes. Particularly, regarding intracerebral hemorrhage (ICH), we were unable to provide specific details on its type, localization, and size. This limitation arises from the complexity of extracting such detailed information and the constraints inherent in the MIMIC-IV database. The lack of these ICH-related details prevents a more nuanced analysis of how these factors might influence patient outcomes, especially in the context of renal function and the use of therapies like mannitol. Moreover, we also acknowledge that the high missingness rate for some variables is another limitation, which may affect the robustness of our findings. We sincerely regret these shortcomings and recognize the importance of these factors in a comprehensive assessment of patient care and prognosis.

Conclusion

Given the methodological limitations and incomplete confounder adjustment, we’ve revised our conclusion to avoid misleading causal interpretations. Our findings suggest an association between mannitol use and increased mortality in patients with non-traumatic ICH and AKI, but further research is needed to confirm this relationship and explore the underlying mechanisms.