We performed a post hoc analysis of data collected prospectively in four published RCTs of neuroprotective interventions after cardiac arrest: HYPERION (hypothermia at 33 °C [5]), COMACARE (modified Mean Arterial Pressure, modified arterial CO2 and O2 levels [13]), TTH48 (hypothermia for 24 h vs. 48 h [14]), and Xe-HYPOTHECA (inhaled xenon [15]). Only HYPERION found a significant improvement in neurological outcome with the trial intervention. These four trials were chosen because they enrolled similar populations of patients with sustained ROSC after cardiac arrest and ICU admission with coma. Also, the process of care and the inclusion window were similar in the four trials.

The four trials enrolled 1172 ICU patients in all between August 2009 and January 2018. For our study, we included those patients admitted to the ICU alive and for whom laboratory results were available then evaluated their identified data.

The study was approved by the ethics committee of the French Intensive Care Society (FICS/SRLF 23/036) and was conducted in compliance with the latest version of the Declaration of Helsinki and good clinical practice guidelines. The protocol of each of the four RCTs was approved by the appropriate ethics committees.

All patients admitted to the participating centres during each trial period were screened for eligibility. Inclusion criteria were age 18 years or older (in the TTH48 trial up to 80 years) and ICU admission after out-of-hospital cardiac arrest (OHCA), or in-hospital cardiac arrest in HYPERION, followed by the return of spontaneous circulation (ROSC) with persistent coma (defined as a Glasgow Coma Scale [GCS] score ≤ 8). We did not include data from patients who withdrew consent after initial inclusion, had randomisation errors, or did not receive the allocated intervention.

We defined serum sodium levels of 135–145 mmol/L as normal natremia [16], < 135 mmol/L as hyponatremia, and > 145 mmol/L as hypernatremia. We chose to divide the natremia values into three categories to facilitate interpretation by clinicians and to increase relevance to any possible future trials comparing targeted normonatremia to dysnatremia. We considered the first serum sodium level obtained after ICU admission.

Osmolality was computed as follows: (natremia·2) + glycaemia [17]. Osmolality < 275 mOsm/kg defined hypo-osmolality and ≥ 295 mOsm/kg hyperosmolality.

For each patient in each RCT, a dedicated study nurse or investigator at each participating centre collected the baseline clinical data and comorbidities; characteristics of the cardiac arrest and resuscitation; clinical and laboratory features at ICU admission; treatments delivered in the ICU; ICU length of stay; invasive mechanical ventilation duration; and vital and functional status at ICU discharge and hospital discharge and at long-term follow-up.

For the present study, the primary outcome was the day-180 Cerebral Performance Category (CPC) [18]. The day-180 CPC was the primary outcome in the TTH48 trial and a secondary outcome in the COMACARE and Xe-HYPOTHECA trials; in HYPERION, the primary outcome was the CPC on day 90, and the value for day 180 was impute using the last-observation-carried-forward method. A poor outcome was defined as a CPC of 3 (moderately severe disability), 4 (severe disability), or 5 (dead), as recommended in guidelines [19].

The secondary outcome for the present study was 6-month survival.

Given the exploratory nature of our analysis, we did not perform a formal sample size estimation. Categorical variables were described as proportions and continuous variables as mean ± SD if normally distributed and as median [interquartile range] otherwise.

To look for associations linking dysnatremia to day-180 CPC, we performed binary logistic mixed regression analyses with a random effect on the trial. Multivariable analysis adjusted on the modified Cardiac Arrest Hospital Prognosis (mCAHP) score [20] was pre-planned. The mCAHP was introduced in the model as a continuous variable. This tool is a summary statistic built to predict brain damage early after cardiac arrest with ROSC, thereby enabling comparisons of different populations of patients with cardiac arrest [20,21,22]. The seven variables are age, setting of cardiac arrest (public place/home), initial rhythm (shockable/not shockable), time from collapse to CPR initiation, time from CPR initiation to ROSC, blood pH at ICU admission, and epinephrine dose (0/1–2 mg/ ≥ 3 mg).

A sensitivity analysis was performed by considering three groups, defined by hypo-osmolality, normal osmolality, and hyperosmolality, respectively, then performing univariate analyses to look for associations with day-180 CPC before and after adjustment on the mCAHP. Finally, all analyses were repeated with day-180 survival as the endpoint.

The CPC in the HYPERION trial [5] was available only until day 90, instead of day 180 in the three other trials. For HYPERION, we used the last-observation-carried-forward method to perform imputation for the day-180 values of the CPC and mortality. However, since this method assumes stability of values over time (here, from day 90 to day 180), we also performed a sensitivity analysis using only the data from the three other trials.

All tests were two-sided, and p values < 0.05 were considered significant. The analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and R version 3.3.1.