This double-blind, randomized, placebo-controlled trial was conducted at Fundación Valle del Lili (FVL) University hospital between November 2015 and August 2018. The FVL is a fourth-level university hospital equivalent to a US-level I trauma center, and it has 523 beds. Of these, 205 are ICU beds and 10 are reserved for trauma patients. The FVL trauma center admits approximately 700 moderate-to-severe trauma patients per year, serving as one of the largest trauma referral centers in the southwest region of Colombia.

The trial protocol was registered in clinicaltrials.gov (identifier: NCT02542241). The protocol was designed by the Trial Steering Committee and was reviewed and approved by the FVL ethical and biomedical research committee. Written informed consent was obtained from all patients or another surrogate decision-maker. Consent by an independent physician and deferred consent was used as a surrogate to subject consent when appropriate. In these cases, the trial team performed the randomization following pre-specified trial protocol criteria and then requested for the patient’s or representative’s (proxy) informed consent in a later phase.

An independent committee and the staff from the FVL clinical research center regularly monitored the trial to check for protocol compliance and data transparency.

Adult patients with traumatic injuries were eligible for the study if they were to undergo damage control abdominal surgery in the index laparotomy. Specific inclusion criteria were: 1. informed consent obtained before any trial-related activities, 2. age above or equal to 18 years at the time of inclusion, and 3. patients with blunt or penetrating trauma and requiring DCL for abdominal injuries. The decision to perform an emergent laparotomy was taken by the attending trauma surgeon at the emergency department in patients with signs of exsanguination/hemodynamic instability, acute abdomen or evisceration. When in the operating room, patients underwent a DCL (involving surgical control of bleeding and contamination plus packing with laparotomy pads plus leaving the abdominal wall open) if there was evidence of hypothermia, metabolic acidosis or hyperlactatemia (reflecting physiological exhaustion), hemoperitoneum and/or destructive intra-abdominal organ injuries. A detailed description on how we approach and provide surgical treatment to patients with abdominal injuries requiring damage-control surgery at our institution is described elsewhere [9, 10].

Patients were excluded if the time between the occurrence of his/her traumatic injuries and the randomization was longer than four hours, had a severe traumatic brain injury, and had a high probability of death in the first 48 h. Although the treating surgeon subjectively determined this, the decision was based on what is known about the epidemiology of deaths after trauma [11,12,13]. Therefore, patients with a high probability of death during the first 48 h were those with destructive lesions associated with major exanguination leading to refractory shock. For example, patients with abdominal injuries and concomitant/coexistent high grade ≥ 4 solid organ injuries, severe pulmonary injuries causing massive hemothorax, heart injuries, or single/multiple injuries to major named axial torso vessels.

Patients were also excluded if the indications for performing a DCL were different from the initial trauma (i.e., DCL in the context of post-traumatic intrabdominal infections, patients in whom the abdomen was closed at the index surgery but required an emergent relaparotomy for postoperative bleeding or control of contamination). Pregnant women were also excluded.

Patients in both study groups received trauma care in the emergency department and the operating group following current clinical practice guidelines and institutional protocols. After the decision to perform a DCL, patients underwent randomization in a double-blind manner. Randomization was performed in permuted blocks of four and six by an independent statistician using Random Allocation Software® [14]. The assignment was concealed until consent was given. Then, the investigator confirmed the inclusion in an internet-based platform, which revealed the allocation after confirming the consent and the inclusion criteria. Regularly, randomization happened immediately after the index surgery when preparing for the transfer to the ICU or in the first minutes in the ICU.

According to their allocation, the patients received a 72-h infusion (rate: 50 mL/h) of 3 N hypertonic saline or 0.9 N isotonic saline after the index damage control laparotomy. The infusion was administered in the intensive care unit. The presentation of the study solutions was similar. Neither the treating group nor the patients nor the investigators could discern their composition. The treating team made decisions regarding other fluids, blood components, vasopressors, inotropic support, or monitoring according to the usual practice. The study infusion was considered at all times when planning and executing fluid resuscitation. Double-blind was achieved by the use of similar appearing saline bags in the two groups.

Decisions regarding the timing of the reoperations and the abdominal closure were made by the surgical team according to the need for reconstruction of specific lesions, the occurrence of complications, and the fluid balance. In each patient, the surgeons worked to reconstruct the lesions and close the abdomen as early as possible.

The protocol for organ injury management and abdominal wall closure has been previously published [9, 10, 15]. In brief, delayed abdominal closure was considered a bridge to achieve three main goals: 1. compartment syndrome prevention, 2. definitive control of bleeding and contamination, and 3. definitive wound healing. Therefore, the first reoperation was performed as early as possible after correcting the patient’s physiological derangement (control of acidosis, hypothermia, and coagulopathy). It is our surgical practice to perform definitive abdominal wall closure when: all the injuries are correctly managed, the patient physiology is stabilized, and there is no evidence of abdominal contamination focuses/absence of intra-abdominal infections. Also, definitive closure is performed when the abdominal cavity can be closed without tension. Other factors that the surgeon considers to perform definitive closure are the fluid balance, the intestinal edema found at surgery, and the change in the airway pressure or the tidal volume during the procedure.

The primary endpoint of the trial was the proportion of patients with abdominal wall closure in the first seven days after the index surgery.

Secondary endpoints were: 28-day mortality, the fluid balance during the first 72 h, the proportion of patients requiring reoperation for intra-abdominal hypertension, the proportion of patients with abdominal compartment syndrome (defined as sustained intra-abdominal pressure > 20 mm Hg and a new organic dysfunction), the proportion of patients with organic dysfunction, defined as a Sequential Organ Failure Assessment (SOFA score) > 2 [16], the proportion of patients with ARDS [17], and the proportion of patients with hypernatremia.

The trial was designed on the assumption of the superiority principle to have 80% power with a p < 0.05 to detect a 60-to-75% difference in the probability of closing the abdomen on the seventh day after the index laparotomy with the therapy with hypertonic saline. With the addition of 20% for losses, we calculated 200 patients needed in each study group.

The trial team had full access to the data and was responsible for the analysis. All analyses were performed in a blinded manner so that investigators could not know if subjects were assigned to the control or treatment group.

Statistical analyses were performed on Stata 15.1® (College Station, Tx). Categorical variables are presented as frequencies and percentages. The normality of continuous variables was examined by the Shapiro–Wilk test. Afterward, they are presented as mean and standard deviation or median and interquartile rank.

Categorical variables were compared with Chi2 or Fisher’s Exact Test, as indicated. Continuous with Student’s T test or Wilcoxon–Mann–Whitney according to normality. Time-to-event variables were presented as Kaplan–Meier curves and compared with the Logrank test.

Interim analyses were planned at enrollment of 50, 100, 200, and 300 patients, with predefined termination criteria for futility or effectivity [18]. We chose the Haybittle-Peto criteria [19, 20] for stopping for effectivity and the analysis of conditional power for stopping for futility [21, 22].