Design
We are reporting this systematic review protocol in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) guidelines (25, 26). The PRISMA-P checklist is included as an additional file (see Additional File 1). This systematic review has been submitted to the International Prospective Register of Systematic Reviews (PROSPERO) on June 16th, 2021 (27). Any amendments to this protocol will be depicted and dated in the PROSPERO record.
Eligibility Criteria
The following studies will be eligible for inclusion in the systematic review and meta-analysis: all randomized controlled trials (RCTs) that evaluate various amounts of fluid administered as the primary focus of the study (either in the form of restrictive or liberal fluid administration strategies) in a critically ill patient population (i.e., admitted to an intensive care unit). All relevant studies regardless of language will be included in the final selection process. Only studies that include adults (≥ 18 years old) will be used. We will exclude animal and laboratory studies, observational studies, case series, case reports, review articles, systematic reviews and meta-analyses, commentaries, conference abstracts, and studies without a comparison group. Our research question and eligibility criteria are summarized in Tables 1 and 2.
Table 1
Population | Critically ill adult patients |
Intervention | Restrictive fluid management strategy |
Comparator | Liberal fluid management strategy |
Primary outcome | Major adverse kidney events |
Secondary outcomes | All-cause mortality Persistent renal dysfunction New onset renal replacement therapy |
Timepoints | 30 days, 60 days, 90 days |
Table 2
Inclusion and exclusion criteria of the systematic review
Inclusion criteria | Exclusion criteria |
All RCTs | Animal and laboratory studies |
Studies that include adults (≥ 18 years old) | Observational studies |
Studies with fluid administration amounts as the primary focus | Case series and case reports |
Any language | Review articles |
Critical care | Systematic reviews and meta-analyses |
| Commentaries |
| Studies without a comparator |
| Conference abstracts |
Information sources
Databases that will be used for our search include: Ovid MEDLINE, PubMed, EMBASE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Web of Science, The Cochrane Library, and a manual search of bibliographies from included articles. Articles from inception of these databases to present will be eligible for inclusion.
Search strategy
The search strategy was created with the help of experienced medical librarians (MD, LM) at the Saskatchewan Health Authority. The medical subject headings (MeSH) terms to be used in our search include: “diuresis”, “critical illness”, “critical care”, “intensive care units”, “intensive care”, “critically ill patient”. Furthermore, the following keywords were also incorporated; fluid*restrict*, low* fluid*, less fluid*, min* fluid*, fluid* limit*, conserv* fluid*, reduc* fluid*, negative fluid* balance, diuresis, liberal* fluid*, critical* ill*, intensive care, ICU, critical care, CCU. We will utilize Boolean operators, adjacency operators, as well as variations in stemming in our searches. We will re-run the searches before the final analysis. An example of our search strategy for Medline has been included as an additional file (see Additional File 2). This same strategy will be used for the other databases with minor adaptations.
Study Records
Selection Process
Studies from the initial search will be split into two groups (approximately 50% of the studies in each group). Researchers will work in two groups (group 1: AL, ES; group 2: MH, JM) to independently screen titles and abstracts of studies in their respective group in duplicate. Studies that do not fit inclusion criteria will be excluded. Subsequently, the researchers working independently and in duplicate will read the full-text papers to determine inclusion in the final analysis. Any discrepancies between the researchers will be resolved by the opposite group (i.e., a discrepancy in group 1 will be resolved by the consensus decision of group 2). A PRISMA flowchart illustrating studies that were and were not included in the final selection will be completed and included in the final paper.
Data management and extraction
The internet software Rayyan (Rayyan Systems Inc., Cambridge, USA) will be used to aid in the management and screening of studies and data throughout the revision process (28). A data extraction tool will be created on Microsoft Excel version 16.22 (Microsoft, Redmond, USA) to store the extracted data. Three researchers (MH, AL, JM) will independently derive the following data items from the included studies. A fourth researcher will resolve disagreements (ES). Missing data or unreported details will be resolved by contacting the corresponding authors of the studies.
The following variables will be extracted: study year, date of study publication, ICU type (i.e., surgical, medical, mixed), patient type and diagnosis at admission (i.e., sepsis, AKI, ARDS, etc.), study country, study inclusion criteria, study exclusion criteria, number of patients (total in each arm), definition of restrictive versus liberal fluid exposure, mean age of patients (standard deviation [SD]), mean (and SD) APACHE (Acute Physiology and Chronic Health Evaluation) II score, mean (and SD) SOFA (Sequential Organ Failure Assessment) score, mean (and SD) SAPS (Simplified Acute Physiology Score) II score, proportion of female sex, type of fluids received (i.e., crystalloids only, crystalloids/colloids, colloids only), mean (and SD) fluid balance at 24 hrs, 48 hrs, 72 hrs, 96 hrs and 168 hrs, and diuretic usage (type, dose). Fluid balance is defined as the difference between the fluid intake and fluid output at a given point in time (8). We will collect information, where available, on in-hospital mortality, ICU mortality, hospital length of stay, ICU length of stay, proportion of new onset AKI, proportion of new onset ARDS, and duration of mechanical ventilation. ARDS will be defined using the Berlin definition (29, 30).
Outcomes
The primary outcome for this systematic review is MAKE by 30 days, which is a composite of death, new onset RRT, or persistent renal dysfunction (31). New onset RRT will be defined by the initiation of any type of RRT either by hospital discharge or 30 days (31). Persistent renal dysfunction is defined as serum creatinine greater than or equal to 200% of the baseline creatinine level (31). For the primary outcome, we will extract count data and measures of association (i.e., risk ratios [RR] and their corresponding 95% confidence intervals [CIs]) at 30 days, when and where appropriate. If only count data is presented, we will estimate the measure of association.
Secondary outcomes will include MAKE at 60 and 90 days; mortality at 30, 60, and 90 days; new RRT at 30, 60, and 90 days; and persistent renal dysfunction at 30, 60, and 90 days. Tertiary outcomes include duration of mechanical ventilation; length of hospital stay; length of ICU stay; proportion of new onset AKI; proportion of new onset ARDS; and fluid balance at 24, 48, 72, 96, and 168 hours. The definition of AKI is an increase of serum creatinine by at least 26.5 µmol/L within 48 hours, or an increase in serum creatinine at least 1.5 times baseline (from the prior 7 days); or urine output < 0.5mL/kg/h for 6 hours (32). For continuous outcomes, we will extract the mean and standard deviation (SD). When only median and interquartile range information is available, we will estimate the mean and SD using the technique described by Luo and colleagues (33).
Risk of bias
The quality of all included studies will be assessed with the Cochrane collaborative Risk of Bias (ROB) assessment tool for RCTs by at least two independent researchers (MH, AL, JM) (34). Discrepancies will be resolved by a fourth researcher (ES). The following characteristics will be assessed: random sequence generation; allocation concealment; blinding of participants, personnel, and outcome assessments; incomplete outcome data; selective reporting; other biases (34). The results of this assessment will be used to perform subgroup analysis on the primary outcome based on low versus high risk of bias categories. Furthermore, we will grade the certainty of evidence using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach, which classifies the quality of evidence on a spectrum from very low to high (35).
Data synthesis
We will perform a meta-analysis of the included articles if three or more studies are eligible for inclusion in this study. Statistical analysis will be performed using Review Manager 5.4.1 (The Nordic Cochrane Centre, The Cochrane Collaboration, Odense, Denmark) and Stata 15.1/MP (StataCorp, College Station, United States). For dichotomous outcomes, a DerSimonian and Laird random effects meta-analysis model will be performed using an inverse-variance method to estimate the pooled risk ratio (RR) and 95% CI (36) and visually presenting these as forest plots. For continuous outcomes, a random effects meta-analysis model will be performed using an inverse-variance method to estimate the standardized mean difference (SMD) and corresponding 95% CI. We will estimate statistical heterogeneity using the I2 statistic. An I2 less than 25% will be considered no heterogeneity, 25–50% as low heterogeneity, 51–75% as moderate heterogeneity, and greater than 75% as high heterogeneity (34). Publication bias will also be assessed visually using funnel plots and quantitatively using the Egger test (37).
Subgroup analyses will be performed on the primary outcome for the following categories: patient type (i.e., sepsis/septic shock, AKI, ARDS, heart failure), type of ICU (medical, surgical, cardiac, etc.), type of fluids received (i.e., crystalloids only vs. crystalloid/colloids vs. colloids only), amount of fluid received (i.e., large cumulative fluid balance versus low) and studies of low risk of bias versus high risk of bias.
We plan to perform meta-regression by patient age, study patient type, and study country. Additional sensitivity analyses will include performance of cumulative meta-analyses and influence meta-analyses by omitting one study at a time.