Data on the effect of target temperature management at 32–34 °C in cardiac arrest patients considering assessment by regional cerebral oxygen saturation: A multicenter retrospective cohort study

This data article contains raw data and supplementary analyzed data regarding to the article entitled “Effect of target temperature management at 32–34 °C in cardiac arrest patients considering assessment by regional cerebral oxygen saturation: A multicenter retrospective cohort study”. We examined the effectiveness of target temperature management (TTM) at 32–34 °C considering degrees of patients’ cerebral injury and cerebral circulation assessed by regional cerebral oxygen saturation (rSO2). The research is a secondary analysis of prospectively collected registry, in which comatose patients who were transferred to 15 hospitals in Japan after out-of-hospital cardiac arrest (OHCA), and we included 431 study patients. Propensity score analysis revealed that TTM at 32–34 °C decreased all-cause mortality in patients with rSO2 41–60%, and increased favorable neurological outcomes in patients with rSO2 41–60% in the original research article. With regard to the balance of covariates of propensity-score matching (PSM) and inverse-probability weighting (IPW) analyses, some covariates were not well balanced after the analyses between groups. The overlap plots indicate the overlap of densities of the propensity scores are low in group rSO2 41–60% and group rSO2 ≥ 61%. When patients were limited to those who achieved return of spontaneous circulation (ROSC) until/on hospitals arrival, TTM still tended to decrease all-cause mortality and increase favorable outcomes in group rSO2 41–60%.

groups. The overlap plots indicate the overlap of densities of the propensity scores are low in group rSO 2 41-60% and group rSO 2 Z 61%. When patients were limited to those who achieved return of spontaneous circulation (ROSC) until/on hospitals arrival, TTM still tended to decrease all-cause mortality and increase favorable outcomes in group rSO 2  Effect of target temperature management at 32-34°C in cardiac arrest patients considering assessment by regional cerebral oxygen saturation: A multicenter retrospective cohort study (in press)

Value of the data
The data contain raw data and supplementary contents of our original paper, and these are important information for interpretation the results of original research.
TTM at 32-34°C could be still effective when patients with rSO 2 41-60% were limited to who achieved ROSC until/on hospital arrival, excluding patients achieved ROSC after hospital arrival.
The covariates of PSM and IPW analysis were not well balanced, and the overlap plots indicate the overlap of densities of the propensity scores are low in group rSO 2 41-60% and group rSO 2 Z 61%.
The use of TTM at 32-34°C could be effective in patients with specific degrees of cerebral injury, but the result should be interpreted carefully.

Data
We examined the effectiveness of TTM at 32-34°C considering degrees of patients' cerebral injury and cerebral circulation assessed by regional cerebral oxygen saturation (rSO 2 ). This is a secondary analysis of prospectively collected registry [1,2], in which comatose patients who were transferred to 15 hospitals in Japan after out-of-hospital cardiac arrest (OHCA), and we included 431 study patients (Table S1) [3]. In original research article, propensity score analysis revealed that TTM at 32-34°C decreased all-cause mortality in patients with rSO 2 1 and 2) show the overlap of densities of the propensity scores are low in group rSO 2 41-60% and group rSO 2 Z 61%, this indicates the overlap assumption on the treatment effect on the potential-outcome models may be violated. Table 5 shows that TTM could be still effective when patients with rSO 2 41-60% were limited to who achieved ROSC until/on hospital arrival, excluding patients achieved ROSC after hospital arrival.

Study design and data source
The original research article is a secondary analysis of prospectively collected registry, the Japan-Prediction of Neurological Outcomes in Patients Post-cardiac Arrest Registry [UMIN trial ID 000005065] [2,3], in which OHCA patients transported to 15 tertiary emergency hospitals in Japan from May 2011 to August 2013 were consecutively registered. The database consists of pre-hospital and in-hospital data collected from the Japanese emergency medical service (EMS) system and medical charts of each hospital by using the Utstein style [4].

Study population
Comatose patients after OHCA were included in this study if they achieved ROSC. Exclusion criteria were trauma, accidental hypothermia, age o18 years, completion of "Do Not Resuscitate [5]" orders, and a Glasgow coma scale (GCS) score of 48 on arrival at the hospital.
After arriving at hospital, two disposable probes of NIRS (INVOS TM 5100C, Covidien, Boulder, CO, USA) were attached to the patient's forehead. rSO 2 was monitored at least for 1 minute with the probes after several seconds of stable monitoring, and the lowest rSO 2 value was used.

Treatment and outcome measurement
The treatment, TTM with 32 to 34°C (12-24 h) was conducted by the discretion of the attending physician.
We defined the primary outcome as all-cause mortality at 90 days after cardiac arrest, and the secondary outcome as favorable neurological outcome evaluated according to the Cerebral Performance Category (CPC) [12]. The CPC is a 5-point scale ranging from 1 (good cerebral performance) to 5 (dead). We defined favorable neurological outcome as a CPC 1 or 2 by reference to the international guidelines [13,14]. Both all-cause mortality and neurological outcome are core elements in the guidelines. In principle, CPC in individual patients were determined by the physician-in-charge, but in cases of missing data, the main researcher who developed the database determined CPC by contacting patients or family members; both were blinded to rSO 2 readings.

Covariates
We used patient characteristics as covariates, including demographic characteristics (sex, age), pre-hospital status (location of arrest, witnessed arrest, bystander CPR, first monitored rhythm), prehospital resuscitation attempts by EMS (airway management by intubation or laryngeal mask airway device, intravenous injection of adrenaline, usage of Automated External Defibrillator [AED]), patient    status at emergency unit (time from emergency call to hospital arrival, rhythm of electrocardiogram on rSO 2 measurement), cardiac origin or not (presumed by attending physician retrospectively), and procedures after hospitalization (ECPR, coronary angiography, primary percutaneous coronary intervention).

Statistical analyses
In original research article, effectiveness of TTM was evaluated by group according to rSO 2 . Risk ratios and risk differences were obtained by univariate analyses. In multivariate logistic analysis,   Treatment effect estimation was also performed by propensity-score matching (PSM) and inverse-probability weighting (IPW), in order to adjust for differences in baseline characteristics [15][16][17][18]. All analyses were performed with Stata SE, version 14.0 (Stata Corp., College Station, TX, USA). Tests of statistical significance were two-tailed with an alpha of 0.05. Potential-outcome models, also known as Rubin causal models, were used to estimate the distribution of individual-level treatment effects, i.e., changes in outcome caused by receiving one treatment over another [17,18]. We used the average treatment effect (ATE: average effect of the treatment in the population) and average treatment effect on the treated (ATT: average treatment effect among those who received the treatment).
In PSM analysis, we performed nearest neighbor matching within caliper [16]. We basically used age, sex, witnessed arrest, PaO2, PaCO2 and first monitored rhythm (shockable / non-shockable) as covariates for estimating the propensity score (PS), and if possible, more variables relating to patient characteristics observed before TTM were also used to increase the accuracy of the PS model. We used calipers of width 0.2*(SD of log PS) for matching and also included interaction and higher order terms. In IPW analysis, we basically used same covariates as PSM, and if possible, more variables observed before TTM were used, including interaction and higher order terms. We showed balances of covariates (Tables 1-4) and overlap plots (Figs. 1 and 2) of PSM and IPW analysis. Sensitivity analyses were performed by limiting patients to those who achieved ROSC upon hospitals arrival (excluding patients with ROSC after arrival) ( Table 5).