This study is an exploration of the results of the GAS-ART trial recently published (Voldby et al. 2022). The GAS-ART trial was approved by the Ethical committee in Region Zealand (SJ-436), and all enrolled patients provided informed consent. The study was categorized as a drug study and registered at EudraCT (no. 2015–000563-14). The rationale and design of the GAS-ART trial were published before study completion (Voldby et al. 2018).

The reporting of these results adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

Patients

In brief, we included patients between August 2015 and December 2018 undergoing emergency surgery within hours after admittance, for radiologically verified gastrointestinal obstruction or perforation.

The presence of an anaesthesiologist trained in the protocol was mandatory for inclusion. We excluded patients pregnant or younger than 18 years, having a terminal illness (ASA classes 5–6), receiving regular dialysis, with iatrogenic gastrointestinal perforation, unable to give informed consent, or having had intraabdominal surgery within 30 days. Patients were included from five Hospitals located in three regions of Denmark: Odense and Svendborg University Hospitals from Region Southern Denmark, Herlev University Hospital from the Capital Region, and Holbæk and Slagelse Hospitals, both part of Copenhagen University Hospitals from Region Zealand. The randomization was stratified by hospital and by gastrointestinal obstruction or perforation. The number of included patients was based on the power calculation of the original trial. The patients underwent surgery within 6 h of admission; however, in case of ileus, the timing of operation depended on the degree of illness and risk of gangrene and could be postponed until the following day.

Treatment

A GDT-optimized perioperative fluid regimen was compared with a standard regimen. Preoperative fluid volume and administration was identical between the groups aiming at a heart rate below 100 min−1, a systolic blood pressure above 100 mmHg, and venous oxygen saturation above 95%. Intraoperatively, the patients in the GDT-group were given boluses of human albumin 5% in saline based on a stroke-volume algorithm and a maintenance fluid administration ≤ 2 mL kg−1 h−1; after surgery, the fluid administration aimed at a fluid-balance less than 2 L positive or body weight increase below 2 kg. Patients in the standard-group were intra-operatively given crystalloids to ensure a mean arterial pressure > 65 mmHg and diuresis > 0.5 mL kg−1 h−1. In both groups, vasopressors were administered to ensure that a mean arterial pressure > 65 mmHg in case the fluid regimen did not achieve that goal. Haemoglobin was kept above 70 g L−1, in patients with chronic ischemic heart disease above 80 g L−1, or in case of acute ischemic heart disease above 90 g L−1.

On the wards, fluid loss was replaced with a fluid that in volume and electrolyte content resembled the fluid lost, aiming at a zero fluid balance in both groups. This was continued until free oral intake, discharge, or the seventh postoperative day in both groups. The GAS-ART trial found a lower intra-operative fluid administration in the GDT-group (1.5L vs. 2.0L) as expected per protocol. The post-operative fluid administration was comparable between the GDT- and the STD-group.

Explorative analysis

We collected pre-, intra-, and postoperative data prospectively in case report forms.

The fluid input included crystalloids, glucose containing fluids, albumin in saline, packed red blood cells, platelets, fresh frozen plasma, intravenous medicine, and oral fluids. The fluid loss included diuresis, aspirate, vomit, ascites, drain fluid, stoma fluid, perspiration, and blood loss.

In this re-assessment, the exposure variable was the perioperative fluid balance calculated as the difference between the fluid input and loss from induction of anaesthesia and until the end of post-operative day one. We divided the cohort in three groups at a perioperative balance of low < 0L, moderate 0–2L, and high > 2L, because we in a previous study found the 0–2-L interval to be the fluid balance with the lowest complication rate (Voldby et al. 2022).

The primary outcome was cardiopulmonary complications. The secondary outcomes were renal, wound-related, or infectious complications. These outcomes represent our hypotheses for the original trial that the fluid balance influenced cardiopulmonary, wound healing, renal, and infectious complications (Voldby et al. 2018).

The following complications were registered:

Cardiopulmonary complications: pleural effusion, pulmonary congestion or oedema, respiratory failure requiring mechanical ventilation, arrhythmia, acute myocardial infarction, or cardiac arrest.

Renal complications: need for renal replacement therapy or other complications needing intervention. In contrast to the original trial, acute kidney injury (AKI) defined according to KDIGO guidelines (increase in plasma creatinine of more than 27 μmol L−1 or a 50% increase between a pre-operative creatinine value 30 day prior to surgery and a post-operative value within 48 h) was included in this assessment as a renal complication.

Wound-related complications: superficial wound rupture or infection, deep wound infection, and fascia defect or dehiscence.

Infectious complications: superficial or deep wound infection, urinary tract infection, pneumonia, or intraabdominal abscess formation.

The diagnostic criteria of the individual complications are given in the original trial (Voldby et al. 2018; Aaen et al. 2021).

The follow-up for the data used for this assessment was 30 days postoperatively. After discharge, the local investigator contacted the patients by phone if the patient was alive. In addition, the outcomes were registered by blinded assessment of the medical files censored for identity, and all information on fluid therapy, fluid balance, and body weight. Only the assessor blinded outcome data are used in this analysis.

Statistics

Parametric statistics was used for data following a Gaussian distribution; otherwise, non-parametric statistics was used. Numbers and percentages present categorical variables. The outcome was analysed by qhi (Myles et al. 2018) and logistic regression with the Moderate-FB group as reference.

We used a weighted propensity score for each strata of the comparator. The variables included were chosen by the authors based on a priori knowledge of potential confounders. Continuous variables were age and body-mass-index. Categorical variables were: sex, ASA class (grouped in classes 1–3 or 4–5), excess alcohol intake (> 7 units/week for women and > 14 units/week for men), pre-operative sepsis-2-score (classes 0–2 or 3–4), the surgical method (laparotomy or laparoscopy), the type of surgery (resection of intestine with anastomosis or stoma formation, or no resection of intestine), and the diagnosis (gastrointestinal obstruction, upper perforation (gastric, jejunal or ileac), or lower perforation (colonic or rectal)) and “yes or no” for the following: tobacco use, active cancer, cardiac co-morbidity, pulmonary co-morbidity, other co-morbidity including renal disease, liver disease or diabetes, use of vasopressors, and limited postoperative treatment. A planned adjustment for hospital was abandoned due to low numbers at the three hospitals. Multiple imputation was planned to replace missing values > 5%. We present the crude and adjusted results by odds ratio (OR) with 95% confidence interval (95% CI). Additionally, we present the predicted risk of complications related to the fluid balance on a continuous scale. A generalized additive model with smoothing splines and four degrees of freedom was used. The statistical plan was approved by the authors before commencing the analysing of data. A two-tailed p value of less than 0.05 was considered significant. The statistical software was R version 3.5.0 GUI 1.70 El Capitan ©R, 2016 and RStudio version 1.1.453. The codes used can be given on reasonable request.