Impact of 25% albumin administration on intracranial pressure

Intravenous fluid choices may influence intracranial pressure (ICP) in patients with severe brain injury. In post-hoc analyses of patients with traumatic brain injury (TBI) from the large Saline versus Albumin Fluid Evaluation (SAFE) trial, resuscitation using 4% albumin was associated with higher mortality [1] and higher ICP values [2] in severe TBI compared to 0.9% saline. These findings were incorporated into the Brain Trauma Foundation (BTF) Guidelines and ESICM consensus recommendations, which advise against albumin use in severe TBI. However, the albumin solution used in the SAFE trial was a hypotonic (< 270 mOsm/L) 4% formulation, and results from a subsequent experimental investigation suggested that the raised ICP may have been due to the solution’s hypotonicity rather than to the albumin per se: when isotonic 4% albumin was used, there was no increase in ICP [3]. Use of hyperoncotic (20–25%) albumin has thus been proposed within the Lund concept of TBI management, because it may expand plasma volume with smaller fluid volumes, potentially limiting cerebral edema without worsening ICP [4].

We sought to re-evaluate the effects of 25% albumin on ICP using real-world data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database [5]. We selected patients aged ≥ 18 years who had invasive ICP monitoring during an ICU admission, received at least one dose of 25% albumin, and had at least three ICP measurements both within six hours before and within six hours after the first albumin dose. Pre- and post-infusion ICP and mean arterial pressure (MAP) values were summarized as medians [interquartile ranges, IQR]. Paired comparisons were performed using the Wilcoxon signed-rank test, given the small sample size and non-normal distribution of ICP values. Patients were stratified a priori into two subgroups based on baseline ICP: elevated (> 15 mmHg) and normal (≤ 15 mmHg). This threshold was chosen because it reflects the upper limit of normal physiological ICP in adults.

Nineteen patients met the inclusion criteria (Table 1); one was excluded for data artifacts, leaving 18 patients for analysis. The most frequent diagnosis was non-traumatic subarachnoid hemorrhage (SAH; n = 8), followed by thoracic or thoracoabdominal aortic dissection (n = 5); other conditions included cerebral edema, toxic/metabolic encephalopathy, and TBI (Supplemental Material). Five patients (27.8%) died during their hospital stay. A total of 220 reliable ICP measurements (11–16 per patient) were available over the 12-hour window (individual ICP values 1–36 mmHg; overall median 9 mmHg [IQR 7–15]; Fig. 1). At the cohort level, median ICP values before and after albumin administration were 8 [IQR 7–17] mmHg and 10 [IQR 8–14] mmHg, respectively (p = 0.96). Analyzable MAP data were available for 16 patients; median MAP increased from 74.8 [71.8–82.8] mmHg before to 84.5 [74.1–93.2] mmHg after albumin administration.

Table 1.

Demographic and clinical characteristics of the included patients (n = 18)

Variable	Value
Age, years	64 [52–71]
Men, n (%)	13 (72.2%)
GCS on ICU admission	15 [11–15]
Time from ICU admission to first administration of 25% albumin, hours	102 [43–218]
ICP over the 12-hour window, mmHg MAP over the 12-hour window, mmHg SOFA score on ICU admission	9 [7–17] 78 [73–86] 8 [4–10]
ICU length of stay, days	14 [9–25]
In-hospital mortality, n (%)	5 (27.8%)

Data, referring to the entire 12-hour study window, are presented as median [IQR] or n (%), as appropriate. GCS = Glasgow Coma Scale; ICU = intensive care unit; ICP = intracranial Pressure; MAP = mean arterial pressure; SOFA = Sequential Organ Failure Assessment; TBI = traumatic brain injury

Fig. 1.

Changes in ICP at the individual patient level before and after administration of 25% albumin. For each patient, ICP measurements collected within six hours before and six hours after the albumin infusion are displayed. The vertical dashed line represents the moment of albumin administration. Patients were numbered sequentially, with the 5 who had a pre-infusion median ICP > 15 mmHg shown first

Baseline ICP was > 15 mmHg in 5 patients and ≤ 15 mmHg in 13. In those with elevated ICP, median ICP decreased from 24 [19.5–24.0] mmHg in the six hours preceding infusion to 17.5 [14.0–20.5] mmHg post-infusion (p = 0.106), accompanied by an increase in MAP (median 76.8 [70.8–80.8] to 86.8 [79.2–88.8] mmHg). In patients with normal baseline ICP, median ICP increased slightly after albumin (from 7.0 [5.5–8.0] to 8.0 [6.5–9.5] mmHg), with a modest MAP increase (74.8 [71.0–92.5] to 83.0 [74.0–98.0] mmHg).

These findings have several implications. First, the trend toward ICP reduction in patients with elevated baseline ICP is consistent with the theorized oncotic effects of hyperoncotic solutions, which may be expected to be more pronounced in the presence of intracranial hypertension, an observation analogous to the pressure-dependent efficacy of osmotic agents such as mannitol. Second, the modest MAP increase in both subgroups suggests that 25% albumin produces the expected volume-expanding effect, improving systemic hemodynamics without exacerbating intracranial hypertension. Third, these observations align with prior experimental data showing ICP elevation only with hypotonic, and not isotonic, albumin formulations [3], and with the ALISAH trial, which demonstrated safety and possible neuroprotective effects of hyperoncotic albumin in SAH [6].

There are, however, important limitations that should be considered when interpreting the data. First, the sample size is small, reflecting general clinical reluctance to administer hyperoncotic albumin in brain-injured patients, consistent with current guidelines, thus limiting the statistical power to detect differences. Second, the indications for albumin administration and use of concurrent ICP-lowering interventions were not retrieved from the MIMIC database; the observed changes in ICP values cannot therefore be causally attributed to albumin administration. Third, the heterogeneity of diagnoses, including aortic pathologies in which ICP measurements may be inaccurate, may impact interpretation. The indications for ICP monitoring and the devices used were also not consistently recorded, introducing additional measurement variability. Finally, our stratification threshold of 15 mmHg is more conservative than the BTF treatment threshold of 22 mmHg: patients with ICP values between 15 and 22 mmHg may not have been receiving active ICP-lowering therapy at the time of albumin administration, which further complicates interpretation. Despite these limitations, this exploratory analysis of real-world ICU data suggests that 25% albumin does not worsen intracranial hypertension and may be associated with improved hemodynamic status. These findings add to the growing body of evidence questioning the extrapolation of SAFE-TBI data, obtained using a hypotonic formulation, to hyperoncotic albumin preparations, and underscore the need for larger prospective evaluation of 25% albumin in more homogeneous neurocritical care populations.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

CJW conceived the study. All the authors contributed to the study design. Data collection and analysis were performed by MS and CJW. CJW, FST and JLV provided critical revisions and contributed to the interpretation of the findings. CW drafted the first version of the manuscript, and all the authors reviewed and commented on the subsequent versions. All the authors read and approved the final manuscript.

Funding

No funding was received for this study.

Data availability

Given that the MIMIC-IV database contains deidentified health data, approval from our institutional review board (IRB) was not needed. Access to the database was granted to the authors who performed data extraction after completing the required training (CITI Program) and who completed the required training for human subject research and were certified through the CITI ‘Data or Specimens Only Research’ module. MIMIC-IV version 3.1 was accessed through BigQuery.

Declarations

Ethics approval and consent to participate

Given that the MIMIC-IV database contains deidentified health data, approval from our institutional review board (IRB) was not needed. Access to the database was granted to the author who performed data extraction after completing the required training (CITI Program) for human subject research and were certified through the CITI ‘Data or Specimens Only Research’ module.

Consent for publication

Not applicable.

Competing interests

CJW has received fees for speaking and consulting from CSL Behring and Grifols. JLV is a journal editor of Critical Care. MS and FST declare that they have no conflicts of interest.