Association of Initial Illness Severity and Outcomes After Cardiac Arrest With Targeted Temperature Management at 36 °C or 33 °C

Key Points Question What is the optimal target temperature for targeted temperature management (TTM) in comatose patients after cardiac arrest? Findings In a cohort study of 1319 patients, of whom 911 did not have severe cerebral edema or highly malignant electroencephalogram, TTM at 33 °C was associated with better survival than TTM at 36 °C for patients with the most severe post–cardiac arrest illness, but TTM at 36 °C was associated with better survival in patients with mild- to moderate-severity illness. Patients with severe cerebral edema or highly malignant electroencephalogram had poor outcomes regardless of TTM strategy. Meaning The findings of this study suggest that measuring initial illness severity in patients resuscitated from cardiac arrest may guide selection of the optimal TTM strategy.


Introduction
Cardiac arrest and resuscitation often result in brain injury that impairs functional recovery for survivors. 1 Severe brain injury contributes to in-hospital death for many other patients. Failure to awaken leads to withdrawal of life sustaining therapy (WLST) for most patients with out-of-hospital cardiac arrest who have pulses restored. 2 Targeted temperature management (TTM) in a mild hypothermic range (32-36°C) is an intervention that mitigates brain injury in the laboratory. 3 In clinical trials, treating patients with TTM at 32 to 34°C resulted in higher survival and better functional recovery than not regulating temperature after out-of-hospital cardiac arrest. 4,5 However, a trial reported in 2013 did not demonstrate the superiority of TTM at 36°C vs 33°C in patients with mild to moderate illness. 6 Current practice includes TTM for comatose patients after cardiac arrest with a target temperature of 32 to 36°C. 7,8 There are few data to guide selection of the target temperature, which relies largely on clinician preference or institutional protocol. 9,10 Registries reported worse outcomes in cohorts of patients treated after 2013 with TTM at 36°C compared with patients treated with TTM at 33°C 10 or when TTM was not used at all. 11 A 2019 trial 12 found TTM at 33°C to be superior to normothermia in patients with more severe illness. Because the association of TTM at 33°C with outcomes appears to differ between patients with mild to moderate illness 6 and severe illness, 12 we speculated that TTM strategies might have different associations with outcomes among patients with different magnitudes of post-cardiac arrest injury. Therefore, we tested the hypothesis that outcomes differed between patients with TTM at 33°C and those with TTM at 36°C, stratified by illness severity.

Methods
We maintain a quality improvement database of all patients treated for cardiac arrest at our institution. We performed a retrospective cohort study of consecutive patients in this database who were comatose after resuscitation from cardiac arrest and who were treated with TTM. We initiated our study after noting decreased survival in subsets of patients treated with TTM at 36°C after 2013.
The University of Pittsburgh Human Research Protection Office approved retrospective analysis of this database and determined this research to be exempt from the requirement to obtain informed consent. This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Our institution is a regional referral center for post-cardiac arrest care. 13,14 In addition to receiving patients directly from the scene of their cardiac arrest, many patients are initially transported to a local hospital via emergency medical services and are then transferred to our regional center for specialized care. Within our center, a team of physicians provides consultation on most cardiac arrest cases to direct cardiac arrest-specific aspects of care. This group included physicians with prior training in emergency medicine, critical care medicine, or neurology with multiple years of experience as a post-cardiac arrest consultant. Specific domains for which we regiment treatment include TTM, hemodynamic goals, seizure detection and management, ordering and interpretation of testing for neurologic prognostication, referral for rehabilitation, and secondary prevention. 14,15 Most patients receive electroencephalography (EEG) monitoring during TTM given the incidence of malignant EEG patterns in the post-cardiac arrest population and their association with neurologic outcomes. 16 Most patients also have routine computed tomography (CT) of the brain within the first hours of admission to assess for early signs of cerebral edema and for neurologic etiologies of cardiac arrest. 17 Neurologic prognostication uses clinical examination, EEG, evoked potentials, and repeated imaging as ordered and interpreted by our consultation team, according to principles we have previously described. 18 TTM, initiated as quickly as possible in a mild hypothermia range for 24 hours followed by rewarming at 0.25°C/h, is routine for all comatose patients (with comatose defined as not following commands), unless there are family or patient directives to limit aggressive critical care. TTM at 33°C was the routine treatment for all comatose patients after cardiac arrest from 2010 to 2013. From 2014 to 2018, our clinicians selected TTM at 33°C or 36°C based on individual preference, anecdotally reporting that their choices were influenced by illness severity.
We use several different body temperature-control devices based on nursing, clinician, and unit preferences. During this period, we used both surface cooling with gel-adhesive pads or water-filled blankets and endovascular cooling devices. Comparisons of devices revealed little difference in performance. 19 Esophageal temperature was used as the standard measurement site.
Sedation and analgesia with propofol and fentanyl was the most common strategy to suppress shivering, with pharmacologic paralysis as needed. Clinicians used other sedatives, such as midazolam, ketamine, and dexmedetomidine, for individual patients when propofol was not tolerated.

Exposure
For each patient, we determined whether TTM was targeted at 33°C or 36°C. Body temperature goals were explicitly written in the medical record as physician orders. For this analysis, we grouped patients according to the intended regimen. We considered any target of 35°C or higher part of the TTM at 36°C group. We considered any target of 34°C or less part of the TTM at 33°C group. No included patients had TTM targeted between 34 and 35°C. We excluded patients for whom TTM strategy could not be determined. We determined actual body temperatures for each patient from the electronic health record. To understand how clinicians selected TTM at 33°C vs TTM at 36°C after 2014, 1 of us (C.W.C.) queried each clinician before revealing the scope or findings of this study (eAppendix in the Supplement).

Primary Outcome
Our primary patient outcome was survival to hospital discharge. Secondary outcomes were survival to hospital discharge without severe functional impairment (modified Rankin Scale [mRS], 0-3) or without neurological devastation (cerebral performance category [CPC], 1-3). An abstractor determines mRS and CPC for all patients with cardiac arrest as part of routine quality assurance from review of the medical record using instruments designed for this purpose. 20

Subgroups
We first examined 2 subgroups of patients who rarely survive hospitalization with current medical therapy. First, we examined patients with early, severe cerebral edema on CT of the brain. 16 We defined severe cerebral edema as a gray-white ratio of radiograph attenuation in Hounsfeld units of less than 1.20 at the level of the basal ganglia. Interrater correlation (>0.64) and test-retest correlation (>0.93) of GWR measurement is high. 17,21 Second, we examined patients with EEG findings suggestive of irrecoverable primary brain injury. Based on prior literature, 22,23 we defined highly malignant EEG as the absence of any cortical background activity (<2 μV) with intermittent bursts of epileptiform activity, including burst suppression with identical bursts, with or without associated myoclonus. Readers have nearly perfect agreement recognizing this pattern.
Finally, we analyzed patients with neither cerebral edema nor highly malignant EEG, stratified by initial illness severity. We prospectively quantified initial illness severity using the Pittsburgh Cardiac Arrest Category (PCAC) measured during the initial patient evaluation. We derived this 4-level score in cohorts based on the best neurologic examination (based on the motor and brainstem scores of the Full Outline of Unresponsiveness [FOUR] score) and predominant cardiopulmonary failure (hypotension and hypoxemia, based on the cardiovascular and respiratory subscales of the Sequential Organ Failure Assessment [SOFA] score) within 6 hours of restoration of pulses. 24,25 PCAC is strongly predictive of the probability of survival, multiple organ failure, and awakening, even when calculated by investigators not involved with the patient's clinical care. 25 We defined PCAC 1 as awakening and purposeful (FOUR motor, 4); PCAC 2, comatose with preserved brainstem reflexes (FOUR motor and brainstem, 4-7) and without severe cardiopulmonary failure (SOFA cardiovascular and pulmonary, <4); PCAC 3, comatose with preserved brainstem reflexes (FOUR motor and brainstem, 4-7) and severe cardiopulmonary failure (SOFA cardiovascular and pulmonary, Ն4); and PCAC 4, deeply comatose with no movement and missing some brainstem reflexes (FOUR motor and brainstem, <4). When a patient was comatose but the initial neurologic examination was confounded by medications, intoxicants, or chemical paralysis, we defined PCAC as unknown.

Statistical Analysis
All patients with data on intended TTM therapy were included. We describe continuous and ordinal variables using median and interquartile range (IQR). We describe categorical data using percentages and 95% CIs.
We present the association between illness severity and choice of TTM strategy using odds ratios (ORs) and binary logistic regression. We tested whether clinical variables were associated with the choice of TTM strategy using ORs and binary logistic regression.
We report the relative risk (RR) for survival, awakening, and functional recovery when TTM at 33°C is selected relative to TTM at 36°C calculated directly or using log binomial regression. 26 We used durations and age whenever possible. However, for some models to converge, age was coded as decades or as equal to or older than 70 years, and duration of cardiopulmonary resuscitation (CPR) was coded in 10-minute bins. We believe that 10-minute bins reflect the precision of reported CPR duration when it is not confirmed by automated monitoring.
In the subgroup of patients with neither severe cerebral edema nor highly malignant EEG, we tested for interactions of PCAC with TTM strategy with outcomes. Initially, we included a binary indicator of time period (ie, 2014-2018 vs 2010-2013). These analyses did not reveal any secular trends or interactions with PCAC. Because this indicator was nearly perfectly associated with the use of the TTM at 36°C strategy, we did not include it in final models. We chose to keep data from 2010 to 2013 because we were certain that comparable patients in that period would have been treated with TTM at 33°C even if they would have been more likely to receive TTM at 36°C in 2014 to 2018.
We conducted 2 sensitivity analyses. First, we calculated RRs adjusted for patient and clinical characteristics measured before the TTM exposure that were plausibly associated with outcomes or the choice of TTM, ie, age, sex, in-hospital vs out-of-hospital cardiac arrest location, presence of corneal reflex, presence of pupil reflex, duration of CPR, epinephrine administration and dose, shockable electrocardiogram rhythm, and number of shocks. In final models, we excluded variables with no independent association with outcomes (ie, with P > .05). Because epinephrine dose was collinear with duration of CPR and absence of pupil reflex was collinear with PCAC 4, we also dropped these in the final adjusted model. Second, we created a propensity score for the likelihood to choose TTM at 33°C vs TTM at 36°C using the same clinical variables measured before TTM exposure. We then reported the RRs of survival, awakening, and functional recovery with TTM at 33°C vs TTM at 36°C in 1:1 propensity matched groups, using calipers of 0.005.
We conducted analyses with Stata version 15.0 (StataCorp). Statistical significance was assessed using 2-sided 95% CIs. PCAC strata differed in association of TTM choice with outcomes (

Discussion
Choice of TTM at 36°C vs TTM at 33°C from 2014 to 2018 was associated with lower survival and functional recovery among patients with the most severe post-cardiac arrest illness, after exclusion of patients with severe cerebral edema and highly malignant EEG patterns. It is perilous to infer a causal connection based on these observational cohort data. However, our observations are consistent with results of a recent clinical trial that randomly assigned patients with nonshockable rhythms to TTM at 33°C or TTM at 37°C 12 and with observational studies noting decreased survival among patients after adoption of a TTM at 36°C strategy. 10 Taken together, these data are consistent with a differential effect of TTM strategy based on illness severity.
The beneficial effect of lower body temperatures for more severe illness severity has biological plausibility. Reducing brain temperature can reduce seizure incidence, 27 cerebral edema, 28 intracranial pressure, 29 and metabolic demand during marginal perfusion. 30   clinicians, although we note that this score correlates with patient outcomes even when calculated asynchronously for patients treated by a separate clinical group in another hospital. 25 The excellent outcome for patients in PCAC 2, in whom survival consistently exceeded 60%, is reassuring. Patients presenting with favorable clinical signs immediately after restoration of pulse may not require or benefit from a TTM at less than 36°C, and a less intensive TTM strategy might reduce complications. This fact may explain some of the neutral findings in randomized clinical trials, which excluded patients with the most severe injuries. For example, the TTM trial excluded patients resuscitated from asystole. 6 The final cohort in that trial had short no-flow times, usually preserved brainstem reflexes, and most closely resembled the patients in PCAC 2 encountered at our center.
Survival for the patients in PCAC 2 in this study is comparable with survival for the TTM trial cohort.
An ongoing clinical trial will better assess whether rigorous fever control is superior to TTM at 33°C in this patient phenotype. 32 Few studies or trials choose therapy based on initial illness severity or even report initial illness severity after cardiac arrest. The present analysis confirms how strongly illness severity is associated with expected survival and outcomes 24,25 and how illness severity can interact with response to TTM. 34 A previous study from our center 36 illustrated how illness severity interacted with outcomes after coronary angiography. In the future, clinicians may select specific therapies based on illness severity. Several measures of illness severity after cardiac arrest are available. 24,25,37 To advance our understanding of pathophysiology and to find optimal therapies, future studies and randomized clinical trials in patients with cardiac arrest must stratify patients according to severity.

Limitations
This study has limitations. We emphasize caution using these observational data to guide clinical practice. Observational data are prone to unmeasured biases, and we documented that physicians are consciously using clinical gestalt to select TTM strategies. Adjusted analyses using measured