Cognitive effects of body temperature during hypothermic circulatory arrest (GOT ICE trial): A randomized clinical trial comparing outcomes following aortic arch surgery

Abstract

Background:

Deep hypothermia has been the standard for hypothermic circulatory arrest (HCA) during aortic arch surgery. However, centers worldwide have shifted towards lesser hypothermia with antegrade cerebral perfusion (ACP). This has been supported by retrospective data, but there has yet to be a multicenter, prospective randomized study comparing deep versus moderate hypothermia during HCA.

Methods:

Randomized single-blind trial (GOT ICE trial) of patients undergoing arch surgery with HCA+ACP at 4 United States referral aortic centers (8/2016–12/2021). Randomization to one of three hypothermia groups: DP: Deep (≤20.0°C); LM: Low-moderate (20.1–24.0°C); HM: High-moderate (24.1–28.0°C). Primary outcome: composite global cognitive change score (GCCS) between baseline and 4-weeks postoperatively. Intention-to-treat analysis to evaluate if: (1)LM non-inferior to DP on GCCS; (2)DP superior to HM. Secondary outcomes: domain-specific cognitive change scores, neuroimaging findings, quality of life (QOL), adverse events.

Results:

308 patients consented; 282 met inclusion and randomized. 273 completed surgery and 251 4-week follow-up (DP: 85 (34%); LM: 80 (34%); HM: 86 (34%)). Mean GCCS from baseline to 4-weeks in the LM group was non-inferior to DP group; likewise, no significant difference was observed between DP and HM. Non-inferiority of LM versus DP, and lack of difference between DP and HM, remained for domain-specific cognitive change scores, except structured verbal memory, with noninferiority of LM versus DP not established and structured verbal memory better preserved in DP versus HM (P=.036). There were no significant differences in structural or functional MRI brain imaging between groups postoperatively. Regardless of temperature, patients who underwent HCA demonstrated significant reductions in cerebral grey matter volume, cortical thickness and regional brain functional connectivity. Thirty-day in-hospital mortality, major morbidity, and QOL were not different between groups.

Conclusions:

This randomized multi-center study evaluating arch surgery HCA temperature strategies found low-moderate hypothermia non-inferior to traditional deep hypothermia on global cognitive change 4-weeks post-surgery, although in secondary analysis structured verbal memory was better preserved in the deep group. The verbal memory differences in low- and high-moderate groups and structural and functional connectivity reductions from baseline merit further investigation and suggest opportunities to further optimize brain perfusion during HCA.

Clinical Trial Registration:

ClinicalTrials.gov Identifier: NCT02834065.

Introduction

Hypothermic circulatory arrest (HCA) protects the brain and visceral organs from ischemic injury and allows aortic arch reconstruction in a bloodless field. As the brain is the organ least tolerant of ischemia, cooling strategies have historically utilized profound (≤14°C) or deep (14.1–20.0°C) hypothermia¹ so as to maximize cerebral metabolic suppression and ischemic tolerance.^2,3 However, the limits of hypothermia alone for cerebral protection have become apparent with accumulating clinical experience and laboratory research.⁴ Because of these concerns, adjunctive cerebral perfusion techniques, including retrograde⁵ and antegrade⁶ (ACP) cerebral perfusion, gained popularity, and ACP has now become the predominant neuroprotective adjunct worldwide during HCA.^7–9 Despite this, recent work^10,11 has found evidence of neurologic injury with even short durations (≤20 minutes) of HCA plus ACP.

Additionally, concerns regarding the consequences of marked temperature reductions, including coagulopathy, systemic inflammatory response, and direct neuronal injury, also led some to advocate for lesser degrees of hypothermia during HCA.^12,13 A large comparative effectiveness analysis¹⁴ suggested similar outcomes with deep or moderate (20.1–28.0°C) HCA, as well as significant variability in current practice between centers and surgeons. However, multicenter randomized controlled trials (RCT) are lacking,¹² and the need for higher quality evidence that includes brain imaging and neurocognitive assessments at longer-term endpoints has been widely recognized.^12,14–16 As such, the purpose of this multicenter prospective RCT was to comprehensively evaluate the neurologic effects of deep versus moderate hypothermia strategies during aortic arch surgery. We hypothesized that: 1) low-moderate (LM; 20.1–24.0°C) hypothermia is non-inferior to deep (DP; ≤20.0°C) hypothermia during HCA with ACP with regards to impact on postoperative neurocognitive function; 2) DP is superior for global cognitive preservation compared to high-moderate (HM; 24.1–28.0°C) hypothermia; and 3) structural and functional brain network connectivity are preserved to a greater degree after DP and LM hypothermia when compared with HM hypothermia.

Methods

The authors declare that all supporting data are available within the article and its online supplementary files. Data collected for the study will not be made available to others.

Trial Design and Oversight

The trial was approved by the Institutional Review Boards of the participating institutions and all patients provided informed written consent before trial inclusion. The GOT ICE trial was a prospective, randomized, multi-center clinical trial evaluating the neurologic effects of deep and moderate hypothermia strategies during aortic arch surgery, utilizing clinical, neurocognitive function, and quality of life data from all participants, and anatomic and functional neuroimaging data on participants from a single site (Duke University). Study participants, neurocognitive and neuroimaging assessors, and statistician were blinded to patient group assignment.

Patients

Patients aged 18 years or older scheduled for elective aortic arch surgery (hemi- or total arch) via median sternotomy with HCA and ACP were approached for enrollment (Supplementary Methods S1). Assessment of trial eligibility, surgical procedures, and postoperative follow-up were performed at the four enrolling sites (Baylor Scott and White, The Heart Hospital, Plano, TX; Duke University, Durham, NC; Emory University, Atlanta, GA; University of Pennsylvania-Penn Presbyterian Medical Center, Philadelphia, PA), all of which are regional aortic surgery referral centers. Duke University was the trial coordinating center.^17–20

Randomization

Patients were randomly assigned to one of three treatment groups: Group DP: Deep (≤20.0°C); Group LM: Low-moderate (20.1–24.0°C); Group HM: High-moderate (24.1–28.0°C), all in conjunction with cold unilateral ACP.^1,14 Randomization was additionally stratified by age (<40, 40–50, 50–60, >60), sex, surgery type (total- vs. hemi-arch), and clinical site, which were key factors intended to create balance across treatment groups. Randomization schedules were generated by applying random permuted blocks with mixed block sizes and 1:1:1 proportion for the three treatment groups and were generated by an independent statistician at Duke using nQuery (v 7.0, Statistical Solution Ltd., Boston, MA). Details of assignment are described in Supplementary Methods S2.

Interventions

Anesthetic and cardiopulmonary bypass (CPB) management are detailed in Supplementary Methods S3. Right axillary (n=209; 77%) or central aortic (n=64; 23%) arterial cannulation was used for CPB inflow at the discretion of the operating surgeon. Proximal aortic repair was initiated during the period of cooling. When target cooling temperature was reached, CPB was stopped, and the open arch reconstruction (hemi- or total) portion of the case carried out. Unilateral ACP was provided via the right axillary or innominate artery at the discretion of the operating surgeon with the innominate and left common carotid arteries clamped; the left subclavian artery was additionally clamped at surgeon discretion. ACP target flow rate was 5–15 mL/kg/min with an inflow temperature of 12°C-14°C to a target right radial arterial line pressure of 50–70 mmHg. Transition of unilateral to bilateral ACP (n=2/273; 0.7%) was allowed at the discretion of the operating surgeon based upon intraoperative electroencephalography or near-infrared spectroscopy findings suggestive of asymmetry between the right and left cerebral hemispheres. After completing the arch reconstruction, CPB was reinstituted, and the patient rewarmed after a 5-minute period of cold reperfusion.^21,22

Primary Outcome

In accordance with the consensus statement on assessment of neurobehavioral outcomes after cardiac surgery,²³ a battery of standardized cognitive tests were administered and analyzed using factor analysis (Supplementary Methods S4). Factor analysis is a standard method used to derive cognitive measures in the literature.^20,24–26 In this study, a five-factor solution was employed and accounted for 80.4% of the variability of the original 14 test scores. These factors represent five cognitive domains: 1) structured verbal memory (i.e., ability to remember from a list), 2) executive function, 3) visual memory, 4) unstructured verbal memory (i.e., ability to remember from a narrative), and 5) attention and concentration. Further, we derived the continuous cognitive index (CCI), the mean of the five domain factor scores; a larger CCI implies higher cognitive function. The primary outcome was the global cognitive change score (ΔCCI), which was the intra-patient difference between baseline and follow-up CCI (i.e., ΔCCI=CCI_4wk-CCI_baseline). A negative ΔCCI indicates cognitive decline, and a positive ΔCCI indicates cognitive improvement. A reduction in CCI (cognitive decline; negative ΔCCI) has been shown in prior studies to be directly correlated with worsened quality of life, depressive symptoms, and self-reported cognitive difficulties after cardiac surgery.^25,26

Secondary Outcomes

Multiple secondary outcomes were assessed including: (1) domain-specific cognitive change scores, calculated as the difference between the baseline score and 4-week score for each factor; (2) change in neurologic function – change from baseline to postoperative day 4 and 4-weeks postoperatively, assessed using the National Institutes of Health Stroke Scale (NIHSS); (3) CAM-ICU²⁷ (intubated patients) or 3D-CAM²⁸ delirium assessment preoperatively and on postoperative days 1, 2, and 3; (4) a packet of self-report measures administered at baseline and 4-weeks postoperatively to assess health-related quality of life (QOL) outcomes (Supplementary Methods S5); (5) neuroimaging procedures (Duke University participants only) performed at baseline and 4-weeks postoperatively consisting of high-resolution anatomic and functional resting-state magnetic resonance imaging (rs-fMRI) sequences acquired on a 3-Tesla General Electric Discovery MR750 magnetic resonance scanner (Supplementary Methods S6). All neurologic outcomes were adjudicated by a board-certified neurologist (MLJ).

Sample Size Calculation

We estimated that 77 patients per group would globally achieve 90% power to test: (1) if LM hypothermia is non-inferior to DP hypothermia, and (2) if DP hypothermia is superior to HM hypothermia for the primary outcome of global cognitive change score (i.e. ΔCCI). Given the two hypothesis testing, significance level was adjusted to 0.025. The power calculations incorporated data from our pilot study, where we observed global cognitive change score means (SD) of 0.26 (0.2) in the DP hypothermia group, 0.17 (0.5) in the LM hypothermia group, and −0.04 (0.4) in the HM group. Non-inferiority between LM and DP was assessed with one-sided two-sample t-test for mean difference at alpha level 0.025 based on the observed variability (the SD estimates) from the pilot data. For 77 patients per group, we achieve 90% power to establish non-inferiority with a margin of 0.20. That is, we reject the null hypothesis if μ_LM−μ_DP≤−0.2, where μ corresponds to the group mean of cognitive change scores. As there are no consensus statements that define a clinically meaningful decline, we used mild cognitive decline defined from consensus guidelines²⁹ as a 1 SD drop from baseline. We posited that a clinically meaningful non-inferiority margin should be < 1 SD. A margin of 0.2 is smaller than the estimated standard deviation for the postoperative 6-week global cognitive change score for the placebo group (SD=0.37, n=209) reported in Klinger et al.¹⁸ Klinger et al. performed the same cognitive battery of outcomes and used factor analysis at a similar postoperative timeframe as our study. For the second study hypothesis, a two-sided two-sample t-test with a 0.025 alpha-level test would have 90% power to detect an effect size difference of 0.67 between the DP and HM groups, which is smaller than the estimated effect (0.99) observed in the pilot data. Enrollment was inflated to 91 patients per group to allow for up to 15% loss to follow-up. Finally, we estimated 25 patients per temperature group by two-sample t-test would provide sufficient power to detect moderate differences (0.68 as seen in a recent CABG study³⁰) for the whole brain voxel-wise connectivity analysis, allowing for up to 15% loss to follow-up.

Statistical Analysis

Descriptive statistics are presented as mean (SD) or median (Q1, Q3) for continuous variables and frequency (percentage) for categorical variables. The variable difference between or among treatment groups was assessed using a two-sample t-test or ANOVA for continuous variables and Chi-square or Fisher’s exact test for categorical variables as appropriate. For non-inferiority testing, a one-sided two-sample t-test using an alpha-level of 0.025 was performed with one-sided 97.5% confidence interval (CI) generated. Non-inferiority is established if the lower bound of one-sided 97.5% CI of the mean difference of global cognitive change score between the LM and DP groups is > −0.20 margin. For the 2^nd hypothesis, a two-sided two-sample t-test was performed to compare the difference in global cognitive change scores between the DP and HM groups. We considered a difference statistically significant if the p-value is < 0.025.

Several secondary outcomes, including domain-specific cognitive change scores, QOL, NIHSS, and incidence of postoperative delirium were analyzed. For domain-specific cognitive change scores, we performed similar analyses to the primary outcome by testing non-inferiority for LM versus DP treatment and cognitive change score (ΔCCI) difference between HM and DP treatments. The domain-specific non-inferiority margin was determined by applying the margin/SD ratio (i.e. 54%) of the global cognitive change score on the SD estimate of domain-specific cognitive change scores from Klinger et al.¹⁸ (Supplementary Methods S7). For other secondary outcomes, we focused on testing differences between HM and DP and between LM and DP treatments. For QOL measures and NIHSS, two-sample t-tests were performed. For postoperative delirium, a chi-square test was performed. As multiple QOL variables were tested, we computed false discovery rate (FDR)³¹ to account for multiple testing, and significance was determined if FDR adjusted p-values were < 0.025. For the NIHSS and delirium outcomes significance threshold was set at 0.025.

The analytic methods for global tissue and white matter hyperintensity volumes, regional perioperative cortical thickness change, and resting-state functional connectivity are detailed in Supplementary Methods S6.^32–35 As noted above, the imaging study was conducted only on study participants at Duke University.

Sensitivity analyses (Supplementary Methods S8) were performed for the following datasets for the primary outcome: 1) all patients enrolled (n=273) using multiple imputation to investigate the impact of missing cognitive test scores; 2) per-protocol cases only. Finally, since patients with total arch repair tended to have longer HCA and ACP times, cognitive outcome difference between patients with total and hemi-arch repair was assessed. Significance threshold for sensitivity analyses was set at 0.025. All analyses were conducted by a statistician (YJL) blinded to treatment groups until analyses were complete using SAS^® version 9.4 (SAS Inc., Cary, NC) and R 4.1.1. The trial was conducted in accordance with the original trial protocol as registered on Clinicaltrials.gov (NCT02834065). A Data Safety and Monitoring Board annually reviewed accumulating safety data, evaluated any adverse effects of treatment, and assessed the continuing validity and scientific merit of the trial.

Results

Patients

From August 2016 to December 2021, 308 patients were consented; 282 met all inclusion criteria and were randomized. Nine patients withdrew before treatment administration, 273 patients completed surgery. A total of 251 patients (Group DP=85, LM=80, HM=86) returned for the follow-up visit at 4-weeks postoperatively (Figure 1). The sample sizes were well-balanced across the three treatment groups. There was some sparsity among the large number of randomization strata (4 sites, 2 procedures, 2 genders, and 4 age groups). Randomization was reasonably balanced for both the demographic and clinical characteristics for the treatment groups except for higher smoking history in both HM (48.9%) and LM (51.2%) than DP (34.4%), and more patients with previous CABG/valve surgery in LM than DP (23% vs. 8.6%). There was also a higher proportion of diabetes in DP (17.2%) than HM (8.7%) and LM (10.3%) (Table 1). All surgical characteristics were balanced among the three treatment groups except for longer cooling time in the DP group (Table 2).

Figure 1:

CONSORT diagram showing flow of study participants.

Table 1.

Demographic and clinical characteristics of treated patients

Clinical Variables	Group HM (N=92)	Group LM (N=87)	Group DP (N=94)
Age, mean (SD), years	63 (11.4)	62 (10.4)	62 (11.8)
Sex, n (% female)	19 (20.7)	19 (21.8)	21 (22.3)
Race, n (%)
White	85 (93.4)	78 (89.7)	76 (82.6)
Black	5 (5.5)	8 (9.2)	11 (12.0)
Asian	0 (0)	1 (1.1)	5 (5.4)
Other	1 (1.1)	0 (0)	0 (0)
Years of Education, median [Q1, Q3]	16 [13, 17]	16 [13, 17]	16 [13, 18]
Preoperative cognitive index, mean (SD)	−0.01 (0.67)	0.06 (0.64)	−0.05 (0.74)
Body mass index, mean (SD), kg/m2	29.8 (5.2)	29.1 (5.6)	28.6 (4.8)
Hypertension, n (%)	69 (75)	61 (70.1)	63 (67.7)
Hypercholesterolemia, n (%)	47 (51.1)	56 (64.4)	52 (55.9)
Diabetes, n (%)	8 (8.7)	9 (10.3)	16 (17.2)
Smoking history, n (%)	45 (48.9)	44 (51.2)	32 (34.4)
COPD, n (%)	3 (3.3)	7 (8.0)	5 (5.4)
Coronary artery disease, n (%)	41 (44.6)	45 (51.7)	44 (46.8)
Previous neurologic event, n (%)	9 (9.8)	6 (6.9)	6 (6.4)
Prior myocardial infarction, n (%)	7 (7.6)	3 (3.4)	2 (2.1)
Prior CABG/valve surgery, n (%)	12 (13.0)	20 (23.0)	8 (8.6)
LVEF, median [Q1, Q3], (%)	55 [54.5, 55]	55 [55, 55.8]	55 [55, 60]
Aortic stenosis ≥moderate, n (%)	34 (37.4)	35 (40.2)	33 (35.1)
Aortic insufficiency ≥moderate, n (%)	39 (42.4)	31 (35.6)	39 (41.5)
Bicuspid Aortic Valve, n (%)	42 (45.7)	43 (49.4)	50 (53.2)
Heritable Thoracic Aortic Disease, n (%)	6 (6.5)	6 (6.9)	7 (7.4)
Preoperative Creatinine, median [Q1, Q3], mg/dL	1.0 [0.9, 1.1]	1.0 [0.9, 1.13]	1.0 [0.86, 1.10]

COPD = chronic obstructive pulmonary disease, CABG = coronary artery bypass grafting, LVEF = left ventricular ejection fraction

Table 2:

Surgical characteristics of treated patients

Variable	Group HM (N=92)	Group LM (N=87)	Group DP (N=94)
Site, n (%)
Duke	42 (45.7%)	42 (48.3%)	44 (46.8%)
Emory	29 (31.5%)	25 (28.7%)	25 (26.6%)
Penn	12 (13.0%)	13 (14.9%)	13 (13.8%)
Baylor	9 (9.8%)	7 (8.1%)	12 (12.8%)
Surgical Procedure, n (%)
Hemi-Arch Repair	85 (92.4%)	81 (93.1%)	85 (90.4%)
Total Arch Repair	7 (7.6%)	6 (6.9%)	9 (9.6%)
Concomitant CABG	12 (13.0%)	7 (8.0%)	16 (17.0%)
Concomitant Aortic Valve	56 (60.9%)	59 (67.8%)	57 (60.6%)
Concomitant Aortic Root	19 (20.7%)	16 (18.4%)	17 (18.1%)
Cooling time, median [Q1, Q3] * , minutes	31 [23, 44]	35 [26, 43]	42 [34, 58]
Cross clamp time, median [Q1, Q3], minutes	123 [97, 154]	117 [91, 150]	121 [108, 159]
CPB duration, median [Q1, Q3], minutes	158 [133.5, 192]	156 [131, 194]	161 [141, 197]
ACP time, median [Q1, Q3], minutes	17 [13, 23]	15 [12, 24]	15 [12, 22]
Lower body HCA time, median [Q1, Q3], minutes	17 [13, 24]	16 [12, 24]	16 [12, 24]
Rewarming time, median [Q1, Q3] ** , minutes	102 [77, 127]	96 [76, 125]	100 [83, 125]
Surgery duration *** , median [Q1, Q3], minutes	300 [247, 376]	282 [238, 354]	305 [259, 373]

^*Cooling time was significantly different between the high-moderate (HM) and deep (DP) temperature groups and between the low-moderate (LM) and DP temperature groups (both P<.001).

^**Rewarming time defined as time from start of full flow CPB after completion of lower body HCA to separation from CPB. CABG = coronary artery bypass grafting, CPB = cardiopulmonary bypass, ACP = antegrade cerebral perfusion, HCA = hypothermic circulatory arrest;

^***Surgery duration defined as anesthesia start time to anesthesia end time.

Primary Outcome

Of the 273 patients consented and completing surgery, 254 had complete baseline cognitive assessment. Among them, 251 patients returned for 4-week follow-up, and 228 (DP: n=85 (34%), LM: 80 (34%), and HM: 86 (34%)) had complete cognitive testing data. For the primary outcome of global cognitive change score (ΔCCI) from baseline to four weeks postoperatively, the means (SD) were −0.01 (0.36)) for HM, 0.004 (0.34) for LM, and 0.05 (0.35) for DP hypothermia, respectively (Table S1). Non-inferiority was established for LM compared to DP hypothermia, as the lower bound of the one-sided 97.5% confidence interval for their mean difference was −0.15, exceeding the −0.2 margin (Figure 2, Table S2). We likewise did not observe a significant difference on global cognitive change (ΔCCI) between DP hypothermia and HM hypothermia (mean difference (97.5% CI): 0.06 (−0.07, 0.19), P=.32) (Table S2).

Figure 2.

Non-inferiority results for low moderate (LM) hypothermia versus deep hypothermia (DP) on global and domain-specific cognitive change scores. The blue vertical lines indicate the non-inferiority margin. The details of the results and the margin are listed in Supplementary Table S2.

Secondary Outcomes

Among domain-specific cognitive change scores, non-inferiority of LM versus DP treatment was established for each domain except structured verbal memory with a lower one-sided 97.5% confidence limit of −0.51, which was less than the −0.48 margin (Supplementary Methods S7, Figure 2, Table S2). Structured verbal memory was also better preserved in Group DP as compared to HM (mean difference (95% CI) =0.31 (0.02, 0.61); P=.036) (Table S2). There were no significant differences between treatment groups for the 10 QOL and NIHSS change scores (Table 3). Rates of postoperative delirium trended higher in the HM (35.9%) and LM (38.4%) groups, but were not statistically different from the DP (29.7%) group (P=.37 and .22, respectively) (Table 3).

Table 3.

Secondary outcomes comparison between deep (DP) and high-moderate (HM) hypothermia groups and between low-moderate (LM) and DP hypothermia groups.

	DP vs. HM			LM vs. DP
Change Scores	Mean diff**	97.5% CI	P-value*	Mean diff**	97.5% CI	P-value*
NIHSS	0.12	(−0.05, 0.29)	.15	−0.11	(−0.25, 0.03)	.11
OARS-IADL	0.58	(−0.87, 2.03)	.43	−0.77	(−2.18, 0.64)	.28
CES-D	−1.23	(−3.89, 1.42)	.36	−1.00	(−3.49, 1.50)	.43
DASI	0.10	(−5.27, 5.47)	.97	3.47	(−1.84, 8.79)	.20
STAI	−1.71	(−4.98, ,1.56)	.30	−0.07	(−2.88, 2.73)	.96
Health	−0.12	(−0.46, 0.22)	.50	−0.30	(−0.65, 0.05)	.09
MA	3.25	(−1.61, 8.11)	.19	−0.91	(−5.28, 3.45)	.68
SA	0.04	(−1.06, 1.14)	.94	−0.41	(−1.58, 0.76)	.49
SOCSUP	−3.47	(−7.40, 0.45)	.08	1.43	(−2.82, 5.68)	.51
SCL-90	−0.59	(−1.81, 0.62)	.34	−0.49	(−1.69, 0.72)	.42
Work Activity (WA)	0.41	(−0.90, 1.71)	.54	−0.76	(−2.01, 0.48)	.23
	DP (N=91)	HM (N=92)		LM (N=86)	DP (N=91)
Postoperative Delirium	27 (29.7%)	33 (35.9%)	.37	33 (38.4%)	27 (29.7%)	.22

Abbreviation: NIHSS: NIH Stroke Scale; OARS-IADL: Duke Older American Resources and Services Procedures – Instrumental Activities of Daily Living; CES-D: Center for Epidemiologic Studies Depression Scale; DASI: Duke Activity Status Index; STAI: State Trait Anxiety Inventory; Health: Health score from short form 36; MA: Cognitive difficulty scale; SA: Social activities; SOCSUP: Perceived social support scale; WA: Work activities;

^*These are raw p-values based on two-sided two-sample t-test for QOL and NIHSS outcomes and chi-square test for postoperative delirium. The minimum FDR adjusted p-value for QQL was 0.675 for DP vs. HM and 0.638 for LM vs. DP.

^**Mean diff = mean (DP) – mean (HM) or mean (LM) – mean (DP) for each variable.

Baseline preoperative raw and total intracranial volume (TICV) (adjusted grey or white matter volumes, white matter hyperintensity (WMH) volumes, or WMH lesion burden) and postoperative changes were not statistically different between groups (Table S3). When all groups were combined, there was a significant reduction in grey matter volume (change score mean (SD)=−15.08 (13.22), P<.001). There were no noteworthy differences in other TICV measures (Table S3).

Similarly, no between group difference was observed in presurgical baseline or 4-week postoperative regional cortical thickness, after controlling for HCA time. However, using the combined data, significant regional postoperative cortical thickness changes were observed in 18 regions-of-interest (generally in the 0.05–0.10 mm range; Table S4). Most of these differences were bilaterally represented in inferior frontal and dorsolateral prefrontal cortices (Figure 3A1). Post-hoc analyses using Pearson partial correlation, controlling for HCA time, identified a significant association between right prostriate/posterior cingulate cortex thickness loss and a reduction in the structured verbal memory domain; Bonferroni corrected P=.003 (Figure 3A2).

Figure 3:

A1. Brain regions demonstrating significant postoperative cortical thickness reduction in all HCA groups combined. Surface-based morphometry region-of-interest analysis of mean regional cortical thickness change using a mixed-model ANCOVA with HCA time as a model nuisance variable (P<.001, Holm-Bonferroni corrected; colors representing strength of association). A2. Postoperative regional cortical thickness reduction and cognitive change associations. Pearson partial correlation, controlling for HCA time, revealed significant association of right prostriate/posterior cingulate cortex thickness change and structured verbal memory domain change. B1. Default mode network-associated brain regions demonstrating significant postoperative intrinsic functional connectivity reduction change in all HCA groups combined. Multivariate pattern analysis (MVPA) mixed-model ANCOVA with HCA time as a model nuisance variable. Statistical significance determined using cluster-forming voxel-level P<.001, two-tailed and cluster mass significance p-FWE<0.01. Left/right posterior cingulate (PCC) MVPA cluster (A.3) denoted in yellow with regions accounting for the PCC functional connectivity reduction denoted in green (A.3.1 = right/left middle & superior frontal gyrus locus; A.3.2 = left/right precuneus & posterior cingulate locus; A.3.3 = left/right anterior paracingulate cortex locus; A.3.4 = right angular gyrus locus; A.3.5 = right insula/inferior frontal gyrus locus). B2. Association of postoperative intrinsic functional connectivity reduction (Fisher-Z) between MVPA cluster (yellow region in left/right PCC; A.3) and contributory locus (green region in right/left middle & superior frontal gyrus; A.3.1) and postoperative attention/concentration factor change. Pearson partial correlation, controlling for circulatory arrest time (P<.05). HM = high-moderate hypothermia (24.1–28.0°C), LM = low-moderate hypothermia (20.1–24.0°C), DP = deep hypothermia (<20.0°C).

There were no between group differences at presurgical baseline in resting-state functional connectivity (RSFC), or in postoperative RSFC change after controlling for HCA time. In the combined data, multivariate pattern analysis revealed four large brain regions with a significant reduction in RSFC 4-weeks post-surgery. Three of the regions were located in posterior occipital/parietal regions, while the fourth was centered in the right inferior frontal/insular cortex (Table S5; Figure 3B1). Postoperative RSFC reduction between the posterior cingulate (Cluster 3) and left/right middle and superior frontal gyrus regions (Foci 3–1) was significantly associated with a postoperative reduction in the attention and concentration cognitive domain (P<.001; Figure 3B2).

Adverse Events

Adverse events, defined according to the consensus statement from the International Aortic Arch Surgery Study Group³⁶ (Supplementary Appendix B), are detailed in Table S6. The rate of any in-hospital serious adverse event was 31.9% for the entire cohort and not different among groups (P=.78). Overall, 30-day/in-hospital mortality was low and not different among groups (DP: 1.1%; LM: 3.4%; HM: 0%; P=.15). Likewise, rates of other major adverse events including clinically diagnosed transient ischemic attack (DP: 1.1%; LM: 2.3%; HM: 2.2%; P=.79), stroke (DP: 4.3%; LM: 3.4%; HM: 3.3%; P=.93), paraplegia (DP: 1.1%; LM: 0%; HM: 0%; P=.38), prolonged ventilation >24 hours (DP: 6.4%; LM: 5.7%; HM: 6.5%; P=.99), and renal failure (RIFLE grade III-V) (DP: 3.2%; LM: 2.3%; HM: 1.1%; P=.62) were not different among groups. Total volume (ml) of blood products transfused was greater in Group DP (DP: 2102 [1416, 3520]; LM: 1744 [1247, 2813]; HM: 1442 [1101, 2579]; P=.04). Length of hospital stay was 6 days [IQR: 5, 8] in DP, 6 days [IQR: 5, 7] in LM, and 5 days [IQR: 5, 7] in HM (P=.63). Hospital readmission at 30-days was not different among the treatment groups (DP: 11.7%; LM: 9.2%; HM: 18.5%; P=.16).

Sensitivity Analyses

Sensitivity analyses to address missing cognitive test scores by multiple imputation and analyzed patients without protocol deviation provided similar results as the primary analysis for global cognitive change scores (ΔCCI), supporting non-inferiority for LM versus DP and no difference between DP and HM on global cognitive changes. Lastly, as expected, total arch repair patients (n=22) had longer lower body HCA (65 [35, 82] vs 16 [12, 21] minutes) and ACP times (65 [34, 82] vs 15 [12, 20] minutes) than those undergoing hemi-arch repair (n=251). There was also a trend towards greater cognitive change score (ΔCCI) reduction (−0.16 [−0.38, 0.14]) versus those with hemi-arch repair (0.02 [−0.18, 0.22]) (P=.08).

Discussion

In this multicenter, prospective, randomized controlled trial comprehensively evaluating the effects of deep versus moderate hypothermia strategies utilizing adjunctive ACP during aortic arch surgery, no significant neurocognitive, neuroimaging, or clinical outcome differences were detected among the three treatment groups investigated. Our results indicated that low moderate hypothermia was non-inferior to deep hypothermia on global cognitive change and domain-specific cognitive changes except for structured verbal memory. Structured verbal memory was also better preserved in the deep hypothermia group compared to the high-moderate group. Regardless of temperature range, patients who underwent HCA demonstrated significant postoperative reduction in cerebral grey matter volume, cortical thickness and regional brain functional connectivity.

Prior work has suggested a very high incidence of MRI diffusion-weighted hyperintensities in the first week following arch surgery.¹⁰ The absence of an increase in WMH lesion volume from pre- to 4 weeks post-surgery in any of the treatment groups in the current study may reflect the extended scanning interval in this trial. Most of the small diffusion-weighted particulate or gaseous emboli lesions detected during the acute perioperative period in prior trials are thought to be transitory and likely resolved by four weeks’ time.^37–39 The absence of WMH changes is encouraging, but tempered by the reductions in gray matter volume and regional cortical thickness. Gray matter volumetric and thickness changes following surgery have been observed in prior cohorts^40,41 and do not appear to be directly related to Alzheimer’s disease neuropathology,⁴¹ but their clinical significance remains uncertain. The detected association between reductions in cortical thickness and verbal memory suggests that cortical thinning in regions proximal to or directly involving limbic cortex may account, in part, for cognitive changes often observed in patients with postoperative cognitive dysfunction (POCD).²⁶ Additionally, cortical thinning in the left dorsolateral prefrontal cortex (DLPFC) was found to be associated with verbal memory change. Left DLPFC function has been closely tied to episodic memory encoding^42,43 and may be related to the common memory complaints in postsurgical patients.^44,45 Residual RSFC reductions were also observed in four different brain regions in patients undergoing HCA, regardless of temperature. Two of the regions involved reciprocal connections between the left and right inferior temporal/lingual gyrus regions, while the remaining two regions involved primary default mode network (DMN) and salience network hubs. Decreased connectivity between the left and right posterior cingulate cortex (PCC; DMN hub) and left and right middle/superior frontal gyrus regions was also significantly associated with attention/concentration decline from preoperative baseline. Postoperative alteration of DMN and salience networks has been previously reported following cardiac^30,46,47 and non-cardiac surgeries,^48,49 reinforcing the importance of DMN functional connectivity across a range of cognitive tasks and domains.⁵⁰

There has been a recent worldwide shift in hypothermia strategies for arch surgery because of concerns regarding increased coagulopathy, systemic inflammatory response, and neuronal apoptosis with the use of deep hypothermia.^7,8,12 Further, data from retrospective non-randomized studies have suggested increased rates of surgical mortality, neurologic and renal dysfunction, and decreased patient reported QOL with the use of deep versus moderate HCA.^9,14,51 However, in our study, no differences were observed in any major clinical outcome between patients randomized to either DP, LM, or HM hypothermia. Patients in the DP group did have a numerically higher volume of blood products transfused, although this did not translate into a higher rate of re-exploration for bleeding, prolonged ventilation, or postoperative infections, thereby calling into question the clinical significance of this finding.

Aortic arch surgery volumes continue to increase worldwide,⁵² at a rate faster than cardiac surgery as a whole.¹⁴ Despite this case growth, the optimal temperature for HCA has remained unclear.^12,53,54 The current study provides much-needed randomized trial data comparing outcomes for the most commonly utilized HCA strategies world-wide, and importantly, the findings raise some concerns regarding the use of HCA in general. The observed postoperative reduction in grey matter volume, cortical thickness, and functional connectivity suggests that regional post-acute changes in brain structure and function following HCA are associated with POCD and may underlie subsequent susceptibility to later cognitive decline.^24,55 As such, less-invasive endovascular methods of aortic arch repair⁵⁶ may gain further traction in the future should they prove less injurious to the brain in similar rigorous analysis.

Loss to follow-up is a limitation of the current study. The rate (8%) was, however, superior to most perioperative studies and non-differential among the study groups. As with most prior studies, patients who did not complete the 4-week follow-up testing were older, had greater co-morbidities, lower baseline cognitive score, and experienced a greater rate of adverse events. Using multiple imputation to account for missing data, we still detected non-inferiority between LM and DP and no difference between DP and HM on global cognitive change scores. Unwillingness of study participants to return during the COVID pandemic created some delays in follow-up testing and a lower number who participated in the imaging component of the study. Sensitivity analyses that adjusted for the interval between cognitive testing did not alter the results. Further, the study enrolled primarily patients undergoing hemi-arch replacement, although the percentage of patients undergoing total arch replacement (10% of the overall study cohort) is higher than national-level data where only 7.5% of all arch replacements reported to the STS-ACSD between 2011–2014 were total arch replacement.¹⁴ Regardless, future studies focused on total arch replacement may be warranted. Finally, although the study protocol mandated cold unilateral ACP, the site of administration (axillary vs. innominate artery) was left to the discretion of the operating surgeon, and believed to be more representative of actual clinical practice¹⁴, as well as to increase generalizability of the findings. Furthermore, recent RCT data from the ACE CardioLink-3 Trial¹¹ has demonstrated no differences in outcomes for patients undergoing arch surgery using either innominate or right axillary cannulation.

The results of this randomized study evaluating temperature strategies during aortic arch surgery found a low-moderate hypothermia strategy utilizing ACP to be non-inferior to traditional deep hypothermia for HCA. However, the verbal memory deficit seen in secondary analyses merits additional investigation. Similarly, the significance of gray matter volume and functional connectivity reductions following HCA need additional evaluation but suggest opportunities to further optimize brain perfusion during HCA.

Clinical Perspective.

What is new?

• Multicenter, prospective randomized trial comprehensively evaluating the effects of deep versus moderate hypothermia strategies during aortic arch surgery.

• HCA strategies utilizing low-moderate (20.1–24.0°C) hypothermia plus ACP appear non-inferior to traditional deep (≤20.0°C) hypothermia on global cognitive change 4-weeks post-surgery, although structured verbal memory was better preserved in the deep group.

• Regardless of temperature, patients who underwent HCA demonstrated a significant reduction in cerebral grey matter volume, cortical thickness and regional brain functional connectivity at 4 weeks postoperatively.

What are the clinical implications?

• For patients undergoing elective aortic arch surgery with HCA plus ACP, the use of low-moderate systemic hypothermia (20.1–24.0°C) may be preferred over deep (≤20.0°C) hypothermia given similar global cognitive outcomes.

• The need for HCA in any particular case should be carefully considered given the observed detrimental effects of HCA, regardless of temperature, on brain volumes and function.

• The gray matter volume and functional connectivity reductions following HCA suggest opportunities to further optimize brain perfusion during HCA.

Non-standard Abbreviations and Acronyms:

HCA

hypothermic circulatory arrest

ACP

antegrade cerebral perfusion

RCT

randomized controlled trial

DP

deep (≤20.0°C) hypothermia

LM

low-moderate (20.1–24.0°C) hypothermia

HM

high-moderate (24.1–28.0°C) hypothermia

CPB

cardiopulmonary bypass

NIHSS

National Institutes of Health Stroke Scale

QOL

quality of life

rs-fMRI

functional resting-state magnetic resonance imaging

LMM

linear mixed modeling

TICV

total intracranial volume

WMH

white matter hyperintensity

RSFC

resting-state functional connectivity

POCD

postoperative cognitive dysfunction

DLPFC

dorsolateral prefrontal cortex

DMN

default mode network