Restrictive fluid vs. liberal fluid resuscitation among patients with sepsis, which is beneficial? Systematic review and meta-analysis

Data exploration and strategy

We performed a meta-analysis and systematic review of the interventions. The PRISMA Guideline 2020 served as the foundation for this systematic review and meta-analysis. [CRD420251081057] is the registration number for this study using PROSPERO. The PICOS framework, which includes the following: (P) atient, Adult (18–65 years old), Septic OR Severe Septic; (I) intervention, Restrictive Fluid Resuscitation; (C) omparator, Placebo/Control; (O) utcome, Mortality, Days alive without Vasopressor, ICU length of stay, and hospital length of stay, served as a guide for our literature search, and (S)tudy: RCT/CT; we only used articles written in English or Indonesian, and we restricted publication to ten years after publication.

Data screening

A variety of databases, including PubMed, ScienceDirect, Epistemonikos, and Google Scholar were assigned to four authors. Each author allocated to the database used keywords like (((sepsis) OR (septic shock)) OR (severe sepsis)) AND (fluid restriction)) AND (mortality)) to separately search for publications. If the search data could be obtained automatically, the program was used to perform duplication selection on the found searches. Otherwise, manual duplication selection was performed for databases that could not be fully collected. Subsequently, the articles were chosen on their own using the PICOS framework. Full article availability was checked for a few selected articles. If an entire item could not be located, it was excluded.

Data extraction

Appropriate articles were then extracted independently using Google Sheet, followed by a joint evaluation. Disagreements were resolved through mutual consensus. Data that were confirmed for inclusion were extracted for the following data: (1) Author, Year; (2) country; (3) age; (4) sex; (5) description of intervention; (6) description of control; (7) number of samples; (8) mortality number; (9) days alive without vasopressor, (10) days alive without ventilator, (11) days alive without renal replacement therapy (RRT), (12) ICU length of stay, and (13) Hospital Length of Stay. A copy of data extraction form is available as extended data.⁹

Statistical analysis

Meta-analysis of interventions was performed using RevMan 5.4. The random effects model (REM) was selected for population-wide interpretation (not limited to the study alone). The intervention analysis used dichotomous data (n, population) for mortality parameters and numerical data (mean, SD) for numerical data, such as the duration of hospitalization. Subgroup analyses were planned when data could be categorized separately if allowed. The analysis results were displayed using forest plots.

Risk of bias and publication bias

Risk of bias analysis using RoB2 as a guide to analyze manuscripts with an RCT design. The results were displayed using a traffic plot and a summary plot. Publication bias analysis was performed by showing a funnel plot and assessing the symmetry of the study results.

Screening result & characteristics of included studies

A total of 3500 articles were found in various databases, and 21 were excluded as duplicates. After undergoing various processes, one article was excluded after it was found to be a protocol only. At the end of the process, 14 articles were found to be suitable and were included in this study, with a total of 4097 patients included (Figure 1). The complete characteristics of the included studies that evaluated the impact of restrictive fluid resuscitation on mortality in patients with sepsis are presented in Table 1.

Figure 1. PRISMA flowchart.

This systematic review included 14 randomized controlled trials published between 2015 and 2024, primarily enrolling older adults with septic shock across diverse international settings including Europe, the USA, and Zambia. A significant finding from the included studies was the marked heterogeneity in both patient severity and study protocols. Baseline illness severity varied substantially, as indicated by a wide range of Sequential Organ Failure Assessment (SOFA) scores (2.67 to 11.00). Furthermore, the definitions of fluid strategies were inconsistent; “restrictive” protocols ranged from being guided by dynamic parameters to having fixed volume caps, while “standard care” was often non-specific and left to clinician discretion. This considerable variability in patient severity and the operationalization of the intervention and control arms are key factors to consider when interpreting the overall findings of this review.

Mortality among participants

Once extracted, the mortality data had parameters that varied according to the duration of the observation (Figure 2). Aspects of mortality up to day 7 were explored by Boutlet, 2024 who found a 20% lower risk of death in the restrictive regimen than in the control, but this was not statistically significant. Some studies examined up to day 28 mortality and found that restrictive regimens increased day 28 mortality by 1.39 times compared to the control group. These results were statistically significant with very low heterogeneity.^10–15

Figure 2. Forest plot of meta-analysis towards mortality based on duration.

Two studies examined mortality up to day 30, concluding that restrictive fluid can reduce mortality by 5% compared with the control group, but the difference was not statistically significant. Finally, mortality was observed up to day 90 in several studies, concluding that restrictive fluid reduced mortality by 1% compared to the control group, but the difference was not statistically significant with low heterogeneity. Overall, there was no statistically significant difference between restrictive fluid and control in septic patients, and heterogeneity in the study was found to be low (I² = 1%).^16,17

Days alive without vasopressor

Compared to a control fluid strategy, a restrictive fluid strategy has no discernible effect on the number of days that patients with sepsis spend without vasopressors (Figure 3). Given the large heterogeneity (I2 = 62%) among trials, the conclusion should be regarded cautiously, even though there was a minor tendency in favor of the restrictive regimen (MD 0.53 days). This variation could be due to variations in each study’s standardized procedures, patient group, or definition of fluid restriction.^{6,10,12,13,15,17}

Figure 3. Forest plot of number of days without vasopressor.

Risk of bias

The results of RoB2 analysis are shown in Figure 4. The results varied across studies, with five articles of low risk, seven articles of medium risk, and three articles of high risk of bias. Three articles exhibited a high risk in Domain 2 due to the unblinded design, reliance on recommendations rather than direct control over the intervention, and poor adherence to key protocol elements in the intervention group. Furthermore, studies exhibited a medium risk, poorly described randomization problems, unclear blinding or open label processes, and several other obstacles that have not been well explained by various studies. However, all studies demonstrated a low risk of bias in terms of selection of the reported results.

Figure 4. Risk of bias result using RoB2 tools for RCT studies.

Publication bias

Publication bias analysis is shown in the funnel plot in Figure 5. Assessing the symmetry of the plot, it appeared symmetrical between the right and left sides but tended to be denser on the left side. There appears to be a potential publication bias at the bottom, indicating the possibility of unpublished studies with a large Standard Error (SE) indicating small unpublished studies.

Figure 5. Funnel plot of publication bias.

Clinical implication

Sepsis is defined as organ dysfunction resulting from a dysregulated body response to infection. This serious clinical syndrome occurs when the body’s immune system responds to an infectious agent in an uncontrolled manner, leading to widespread inflammation, tissue injury, and impairment of vital organ function.¹⁸ Early and adequate fluid resuscitation aims to restore circulating blood volume, correct tissue hypoperfusion, and support organ function, all of which are often compromised by the maladaptive immune response that characterizes sepsis.^19,20 This meta-analysis aimed to synthesize the latest evidence comparing the efficacy and safety of restrictive versus liberal fluid resuscitation strategies in adult septic patients. The key outcomes were mortality and days alive without vasopressors. The findings of this meta-analysis shed light on the ongoing debate regarding restrictive versus liberal fluid resuscitation strategies in patients with sepsis.

Restrictive fluid resuscitation refers to a strategy in which intravenous fluids are administered cautiously, typically using smaller total volumes (often ≤30 mL/kg in the initial phase) with a focus on preventing fluid overload. This approach emphasizes careful, individualized monitoring of the patient’s hemodynamic status and earlier use of vasopressors if needed, rather than automatically administering large fluid boluses. The aim is to maintain adequate organ perfusion while minimizing the risk of complications characteristic of excess fluid, such as pulmonary edema, tissue swelling, and organ dysfunction.^21–23 On the other hand, liberal fluid resuscitation is characterized by the administration of larger volumes of intravenous fluids, typically 50−75 mL/kg (e.g., 4−6 L in an 80 kg adult) within the first several hours of treatment. This approach is designed to rapidly restore the circulating volume, improve cardiac output, and reverse hypotension, with the aim of optimizing tissue perfusion. Liberal fluid strategies are supported by traditional resuscitation guidelines, including the early Surviving Sepsis Campaign recommendations, which advocate for at least 30 mL/kg in the first hours after sepsis recognition.^24,25 Both strategies are widely studied, with restrictive approaches gaining favor due to increasing awareness of the harmful effects of fluid overload. However, the definitive superiority of one approach over the other remains under investigation, and the best practice is likely patient-specific and tailored with ongoing monitoring and dynamic assessment tools.^25,26

One of the primary concerns in sepsis management is reduction in mortality. In this study, analyses on days 7, 28, 30, and 90 showed that restrictive fluid strategies could reduce mortality; however, in general, there was no statistically significant difference between the restrictive fluid and control groups in septic patients. Restrictive fluid resuscitation strategies are increasingly favored over liberal or controlled approaches. The primary reason for a lower mortality rate with restrictive fluid strategies is the prevention of fluid overload, which can exacerbate tissue edema, impair oxygen delivery, and worsen organ dysfunction.²⁷ For instance, observational and comparative studies have demonstrated that patients in the restrictive fluid group exhibit significantly better outcomes, such as lower Acute Physiology and Chronic Health Evaluation (APACHE II) and Sequential Organ Failure Assessment (SOFA) scores, indicating less severe organ dysfunction after treatment. These patients also tend to have a lower incidence of complications, including myocardial injury and a reduction in the rate of acute respiratory distress syndrome (ARDS), all of which contribute to improved survival rates.^28–30 While several trials, such as the restrictive intravenous fluid trial in severe sepsis and septic shock (RIFTS) Study by Corl et al. and the protocolized reduction of non-resuscitation fluids versus usual care in septic shock patients (REDUSE) study by Linden et al., did not find a statistically significant mortality difference between restrictive and liberal strategies, they did confirm that a restrictive approach does not increase harm and frequently reduces the need for mechanical ventilation and vasopressor support, both of which correlate with better clinical outcomes.^17,31

In terms of days alive without vasopressors, a comparison between fluid restriction strategies showed no significant effect on the number of days that patients with sepsis spent without vasopressors. This is also in line with the CLOVERS trial in 2023, which found that 59% of vasopressor use in the restrictive group and 37.2% in the liberal group, but days alive without vasopressors were similar.⁶ While restrictive fluids may modestly decrease overall fluid balance, and secondary outcomes such as mortality, organ dysfunction, or days free from vasopressor therapy are largely similar. Restrictive fluid resuscitation strategies are characterized by the limitation of intravenous fluid administration, which prompts clinicians to initiate vasopressor therapy earlier and with greater intensity to achieve and maintain adequate blood pressure and tissue perfusion. Although this approach is effective in reducing the risk of fluid overload and its associated complications, it may also be associated with a slower resolution of shock states, as the underlying hypovolemia is not as aggressively corrected. Consequently, patients managed with restrictive fluid strategies may experience a reduced number of vasopressor-free days, reflecting a prolonged dependency on vasopressor support to sustain hemodynamic stability.^5,6,32

Inconsistency of the result, what’s wrong?

The inconsistencies in mortality outcomes between restrictive and liberal fluid strategies for sepsis are attributable to several critical factors. First, the definitions of “restrictive” and “standard” fluid management differ significantly between studies, encompassing different volume caps, dynamic physiological targets, and early vasopressor use, making direct comparisons problematic. Second, the heterogeneity of patient populations, including wide ranges in baseline illness severity (e.g., SOFA scores from 2.67 to 11.0), age, and comorbidities, profoundly influence fluid responsiveness and susceptibility to fluid overload. Third, the timing and duration of the interventions varied significantly across studies, with some focusing on early resuscitation and others tailored to the patient’s condition. This further complicates the analysis of the phase-specific effects.

Furthermore, variations in healthcare resource settings have been shown to contribute to disparate results. Finally, the presence of inconsistencies in primary outcome measures, in conjunction with a general absence of dynamic fluid responsiveness assessments in numerous trials, diminishes the capacity to discern genuine treatment effects. The combination of methodological and clinical heterogeneities collectively obscures a clear and consistent mortality benefit for either strategy, emphasizing the necessity for more standardized and personalized approaches to fluid management in sepsis.

Strength and limitation

The meta-analysis demonstrated a significant strength in its low overall heterogeneity for the total mortality data, with an I2 value of 1%. This indicated a high degree of consistency across the included studies. Furthermore, this study offers a comprehensive view of mortality by analyzing day-specific rates at 7, 28, 30, and 90 days, providing a broad spectrum of time-dependent outcomes. The broad scope of the study is another strength, as it included 14 randomized controlled trials published between 2015 and 2024, involving 4097 patients from diverse international settings.

Despite its low heterogeneity, this study has several key limitations. A major drawback is the use of the mortality rate as a single predictor, which may overlook other important clinical outcomes. The operationalization of “restrictive” fluid resuscitation varied significantly among the included studies, with protocols ranging from dynamic parameter-guided strategies to fixed-volume caps. This inconsistency in defining the intervention makes it challenging to identify the most effective protocol.

Future direction

Based on these findings, we suggest several directions for future research. To address the variability in the study protocols, future studies should aim to specify more consistent fluid strategy protocols. The significant heterogeneity in baseline illness severity, as indicated by a wide range of SOFA scores (2.67 to 11.00), also points to the need for a stratification strategy based on SOFA scores in future research.