Hypothermia for Patients Requiring Evacuation of Subdural Hematoma: A Multicenter Randomized Clinical Trial

This was a prospective, pragmatic, randomized, controlled, multicenter trial to assess the safety and efficacy of intravascular cooling to induce hypothermia in patients with TBI prior to and after surgical evacuation of SDH. The trial enrolled adult (22–65 years of age) patients with TBI within 6 h of SDH who were not following commands (Glasgow Coma Scale [GCS] motor score ≤ 5). Patients were excluded if there was no planned evacuation of the SDH, concomitant injury or history contraindicated hypothermia for patient safety, arrival temperature was < 35 °C, total GCS = 3 and the patient had fixed and dilated pupils, or there was an inability to obtain consent or use the exception to informed consent for emergency research. Investigators at tertiary care medical centers in the United States and Japan enrolled patients under institutional review board approved protocols.

Patients were randomly assigned into two groups. The intervention group received rapid induction of hypothermia to 35 °C followed by maintenance at 33 °C for 48 h up to 5 days. Rewarming occurred at a rate of 0.25 °C/h. Cooling and rewarming interventions were based on recommendations from Clifton et al. [15]. If intracranial pressure (ICP) increased during rewarming temperature was held constant, and rewarming resumed after standard ICP control measures were instituted. The control group received standard care, including temperature maintenance at normothermia (37 °C) for 48 h. Temperature variations of ± 0.5 °C were permissible. Warming of control patients prior to surgery is per standard care. Standard care for rewarming is to warm patients slowly, based on prior indications of poor outcome due to rapid rewarming [14, 22]. Intravascular catheters (Thermogard XP System with Quattro catheter; ZOLL Circulation Inc, San Jose, CA) and fever control were used for hypothermia and to maintain normothermia, as it has been established that fever is detrimental in patients with TBI [23‐25]. Intravascular temperature management has previously been shown to reach the target temperature rapidly and safely [26‐28]. Shivering was managed according to an established protocol used by the intensive care units [29]. Other care was at the discretion of the treating physicians. All centers’ practice incorporated the Brain Trauma Foundation Guidelines [30].

Neuropsychological assessment of patients’ level of recovery was performed at 4 weeks and 6 months post-injury by investigators who were unaware of the treatment group assignment. Dichotomized GOSE at 6 months post-injury was the primary outcome [31]. Good recovery and moderate disability were designated as favorable outcomes; and severe disability, vegetative state, and death were designated as poor outcomes. GOSE is a widely used global outcome score with good interater and intrarater reproducibility [31]. GOSE assesses consciousness, independence, work status, and return of lifestyle via a structured interview.

The trial design aimed for N = 120 patients and allowed for an extension up to 350 patients. The design included multiple interim analyses after 60, 120, 180 and 240 patients were randomized. At each interim analysis, Bayesian predictive probabilities were to be used to determine whether enrollment should stop for either success or futility before the maximum enrollment. The operating characteristics of the study design (type I error, power, and expected sample size) were computed by simulation in a variety of possible scenarios. Treatment effect was based on data from Clifton et al. [16]. The trial had power greater than 90% for a scenario where the proportion of 6-month GOSE score is good in the treatment arm was 0.604 compared to 0.347 in the control arm and the trial was quite likely to stop at or before the N = 120 interim analysis for success. Type I error was controlled at 0.025 and the trial was quite likely to stop for futility at or before the N = 120 interim analysis.

To reduce the likelihood of imbalance of important prognostic factors between centers the study used a blocked randomization scheme that randomized equally at a 1:1 ratio to hypothermia versus control normothermia. Randomization was generated using a computer program by the independent statistical team who provided sites sequentially numbered sealed opaque envelopes. Once eligibility was confirmed site study physicians and nurses enrolled participants, opened the randomization envelope and assigned the participants to the designated intervention.

Participants and clinical care providers were not blind to assignment. Investigators were unaware of treatment assignments and outcomes of other sites’ patients. Outcomes assessors were blinded to the patients’ treatment arm.

Investigators used enzyme-linked immunosorbent assays (ELISA) to evaluate the plasma levels of GFAP and UCH-L1 on the 24 patients with blood samples (13 hypothermia, 11 normothermia) that were available at three time points. Time 1 (T1) was collected at less than 6 h post TBI and precooling. T2 was collected 6 to 48 h post TBI (within the time of temperature management); and T3 was collected 5 to 14 days post TBI (post cooling). Control standards and samples from six healthy volunteers were measured. Blood was collected in K2 EDTA vacutainer tubes (BD, Franklin Lakes, NJ), processed within an hour of draw and frozen at − 80 °C. The concentration of GFAP was measured using a sandwich enzyme immunoassay (BioVendor, Ashville, NC) according to manufacturer’s instructions. The lower limit of detection for this assay is.045 ng/mL. UCH-L1 sandwich ELISAs were developed using the UCH-L1 DuoSet ELISA (R&D Systems, Minneapolis, MN). The lower limit of detection for the UCH-L1 ELISA was 19.5 pg/mL. All samples were analyzed in duplicate.

A secondary objective was to evaluate the safety of intravascular cooling in the management of acute traumatic SDH. Adverse events were monitored for the 6-month study period. Serious adverse events were graded according to the United States Department of Health and Human Services (USDHHS) Common Terminology Criteria for Adverse Events V4.0 [32]. Predefined adverse events of interest that were of concern with hypothermia treatment were selected to be monitored and reported regardless of grade. These included cardiac arrhythmias, thromboembolic events, pneumonia, bleeding/hemorrhage, infection (culture positive, e.g., blood stream infection, urinary tract infection, ventriculitis), and death.

An independent DSMB consisting of a neurosurgeon, a critical care physician and a statistician all of whom are experts in the fields of TBI and hypothermia monitored the study. The study was registered on ClinicalTrials.gov as NCT02064959 and UMIN000014863 in Japanese UMIN Clinical Trials Registry. The exception to informed consent for emergency research provision was used if permitted by the local institutional review board and local law. Berry Consultants performed the statistical design and independent statistical analysis.

The analysis was performed using a modified intent-to-treat population of all randomly assigned patients having no exclusion criteria. Demographic and baseline data were summarized using means, standard deviations, or count and percentage. Adverse events are reported by type, temperature group and frequency. Descriptive data comparing groups were tested for normality using the Shapiro–Wilk test and analyzed using Student’s t-test and reported as means and standard deviation, or if not normally distributed, using the Mann–Whitney Rank Sum test and reported as medians and interquartile ranges (IQR). A one-sided Fisher’s exact test with a predefined significance level of p < 0.02 to compare proportions of patients with good outcomes as measured by GOSE at 6 months is reported as the primary outcome. ELISA data were log transformed to normalize values and analyzed using two-way repeated measures analysis of variance (ANOVA). The ELISA analyses were exploratory, and significance was assessed at the.05 level.

There were 14 of 16 patients managed per intent at hypothermia. Figure 3 displays the variance of time from injury to hypothermia induction, maintenance and rewarming compared to the mean daily low temperature of the normothermia group (Fig. 3).

Intensive care and hospital length of stay data were available for all but one patient with normothermia. Intensive care unit length of stay did not differ between groups, 12.95 days (IQR 9.2–17.7) hypothermia versus 11.4 days (IQR 7.95–32.4) normothermia (p = 0.7). Hospital length of stay was not different between hypothermia and normothermia groups 20.7 (IQR 17.8–29.8) versus 18.2 (IQR 8.2–45.8) days, p = 0.9, respectively.

To examine whether differences in clinical variables explain the lack of differences in outcome in our treatment groups, post hoc analysis of blood pressure, ICP, blood gases, and blood glucose levels was performed. Daily high and low mean arterial pressure (MAP), mean partial pressure of carbon dioxide, mean ICP, and mean blood glucose levels were examined by treatment group over 8 days post TBI. We observed no group differences in these clinical variables (Supplementary Fig. 1).

Adverse events are summarized as serious adverse events (Table 2) and other, not serious adverse events (Table 3). There was no difference in the number of adverse events per participant by treatment group, 4.5 (IQR 25–9) events per normothermia participant versus 4 (IQR 2–7.8) events per hypothermia participant (p = 0.7). In the normothermia group, 5 of 16 (31%) patients experienced new or increased bleeding. These included one with worsening swelling and hemorrhage during surgery and, post-operatively, two with epidural hematomas, one with intracranial hemorrhage, and one with a worsening contusion. There was no new or increased bleeding observed in the hypothermia treatment group.

At T1, median plasma GFAP and UCH-L1 levels of patients with SDH were elevated compared to healthy controls. GFAP levels at T1 were 3.39 (IQR 1.35–8.66) ng/mL compared with 0 (IQR 0–0.12) ng/mL for healthy comparators (p < 0.001). UCH-L1 levels at T1 were 1.07 (IQR 39–1.82) ng/ml compared to 0 ng/mL for healthy comparators (p < 0.001). When separating the patients by outcome group, two-way repeated measures ANOVA indicated that GFAP (p = 0.036) [but not UCH-L1 (p = 0.26)] levels were lower in the patients with favorable outcome compared with those with unfavorable outcome. The T1 samples showed higher levels of these biomarkers as compared to the levels in T2 or T3 samples (Fig. 4a GFAP and b UCH-L1). Separating the biomarker results by temperature group, plasma levels of both GFAP and UCH-L1 were elevated within the first 6 h of injury (T1) and were significantly higher at T1 compared with T2 and T3. However, two-way ANOVA analysis indicated that neither marker differed by temperature group (Fig. 4c and d). Thus, an effect from the hypothermia treatment on GFAP and UCH-L1 levels cannot be identified in these samples.

Our statistical design planned for enrollment of 120 patients. Therefore, conclusions on the 32 patients must be interpreted with the understanding that our sample size is a limitation. A recent meta-analysis of severe TBI hypothermia trials that utilized protocols similar to the one used in the present study indicated a reduced death rate in those treated with hypothermia (33–35 °C) compared with no cooling (OR = 0.627, p = 0.05) [33]. Although we also had fewer deaths in the hypothermia group, the number of those surviving with good recovery did not differ between groups. We could not verify the protective effect of hypothermia observed in post hoc analysis of National Acute Brain Injury Study: Hypothermia I and II data; however, our results align with the recent POLAR-RCT and Eurotherm3235 trials. The international POLAR-RCT study which investigated prophylactic hypothermia in acute severe TBI demonstrated no difference in favorable outcomes between hypothermia (48.8%) and normothermia (49.1%) patients (absolute risk difference, − 0.4, 95% CI, − 9.4 to 8.7; unadjusted relative risk with hypothermia, 0.99, 95% CI, 0.82–1.19, p = 0.94) [34]. The Eurotherm3235 trial evaluated the effect of hypothermia on elevated ICP and 6-month GOSE outcome after TBI. Eurotherm3235 ended early on recommendation of their DSMB. Final analysis of a dichotomized GOSE favored standard care (OR 1.74, 95% CI 1.09–2.77) [13].

GFAP is an intermediate filament cytoskeletal protein found primarily in astrocytes [35]. GFAP is released into the circulation after TBI and early elevated GFAP levels in the plasma are predictors of poor outcome [30]. Consistent with this, we demonstrated higher GFAP levels in patients with poor outcome. UCH-L1, an abundant protein expressed in neurons, is involved in repair of injured axons and neurons [36]. We did not detect an association between UCH-L1 levels and outcome. Other studies have supported GFAP as being superior to UCH-L1 at predicting outcome [37‐39]. Circulating GFAP and UCH-L1 measured together are biomarkers for severe TBI [20, 40]. Consistent with previously published studies, our patients’ plasma levels of GFAP and UCH-L1 were elevated within 6 h after injury and decreased to levels comparable to healthy comparators by T3 [19, 20].

To examine the effect of induced hypothermia on biomarker levels, we analyzed the plasma concentration over time and did not see a group difference in GFAP or UCH-L1 levels. Contrary to our findings, rodent TBI models have demonstrated a reduction in UCH-L1 levels with hypothermia treatment [41, 42]. However, after cardiac arrest UCH-L1 levels did not differ between comatose patients with cardiac arrest maintained at 36 °C compared with those maintained at 33 °C [43]. Mondello et al. [44] reported higher serum UCH-L1 levels in diffuse TBI compared with patients with mass lesions (p = 0.01) and higher GFAP levels in patients with mass lesions than those with diffuse injury (p = 0.006). We do not have evidence for differing degrees of neuronal injury between our temperature groups. The hematoma size and amount of shift were similar and diffuse axonal injury was absent in comparable proportions of patients. However, it is plausible that there were differing degrees of diffuse injury between groups. It is also plausible that both the hypothermia and protecting from fever with controlled normothermia may have deterred pathological processes associated with diffuse injury.

Our enrollment was slower than expected and resulted in an insufficient sample size to meet our study objectives. Review of our enrollment criteria indicate the key explanations for problems enrolling. Age limits (37%) and lack of SDH or planned surgical evacuation (29%) were the leading exclusion factors. Older age was associated with poorer outcome and more complications in patients with TBI hypothermia [14] and after acute SDH [45]. Younger ages (16–21 years) were excluded from our study because they are classified as pediatric patients by the United States Food and Drug Administration. The Brain Trauma Foundation Guidelines on surgical management of acute SDH recommend surgical evacuation for (1) SDH with thickness > 10 mm or midline shift > 5 mm and (2) on patients with SDH with GCS < 9 or neurodeterioration of 2 or more points on the GCS and/or asymmetric or fixed and dilated pupils and/or ICP > 20 mmHg [46]. Patients who do not meet these criteria may be observed closely and managed non-operatively. If the patient neurologically deteriorates and/or a repeat head computed tomography indicates that the brain injury has worsened, the patient will receive delayed surgery within 2–4 h of clinical deterioration. Guidelines indicate that surgery performed 2–4 h after clinical deterioration result in superior outcome compared to delayed surgery.

We have learned that a large proportion of patients with acute SDH do not meet Brain Trauma Guidelines criteria for surgical intervention. A retrospective chart review revealed that 646 of 869 (74.3%) of patients with acute traumatic SDH at a major level I trauma center were managed without surgical intervention. Only 6.5% of these patients required a delayed surgical evacuation at a median of 9.5 days after injury. GOS at discharge was good in 77% of the non-operatively managed patients [47]. Our criteria required that patients receive surgery within 6 h of injury, and those patients requiring surgery outside of this window would have been excluded.

More than 15% of patients with SDH were excluded because they had non-survivable injuries, conditions contraindicating hypothermia or consent could not be obtained. Our enrollment criteria, based on previous therapeutic hypothermia studies, while restrictive were necessary for patient safety.

The study outcome was a general functional outcome, GOSE at 6 months after injury. It is possible that a more specific measure of cognitive function may have identified a treatment effect. GOSE is the current functional outcome standard in TBI studies. Our findings may not be generalizable to other centers with different management protocols. However, this was a multicenter, pragmatic trial with protocols based on published recommendations for methods of therapeutic hypothermia. We compared therapeutic hypothermia with controlled normothermia. In previous studies [14‐16, 22], standard care normothermia was not controlled with a device. Controlled normothermia may have resulted in a smaller variation in the first 48 h of the normothermia temperature range in our study and thus a potentially protective effect from early fever. A meta-analysis of 39 studies including 14,431 patients indicated fever after neurological injury (traumatic, hemorrhagic, or ischaemic) is associated with worse outcome [48]. The future of targeted temperature management in TBI may focus on tightly controlled normothermia.