Optimal temperature management in aortic arch surgery: A systematic review and network meta‐analysis

Abstract Objectives New temperature management concepts of moderate and mild hypothermic circulatory arrest during aortic arch surgery have gained weight over profound cooling. Comparisons of all temperature levels have rarely been performed. We performed direct and indirect comparisons of deep hypothermic circulatory arrest (DHCA) (≤20°C), moderate hypothermic circulatory arrest (MHCA) (20.1–25°C), and mild hypothermic circulatory arrest (mild HCA) (≥25.1°C) in a network meta‐analysis. Methods The literature was systematically searched for all papers published through February 2022 reporting on clinical outcomes after aortic arch surgery utilizing DHCA, MHCA and mild HCA. The primary outcome was operative mortality. The secondary outcomes were postoperative stroke and acute kidney failure (AKI). Results A total of 34 studies were included, with a total of 12,370 patients. DHCA was associated with significantly higher postoperative incidence of stroke when compared with MHCA (odds ratio [OR], 1.46, 95% confidence interval [CI], 1.19–1.78) and mild HCA: (OR, 1.50, 95% CI, 1.14–1.98). Furthermore, DHCA and MHCA were associated with higher operative mortality when compared with mild HCA (OR 1.71, 95% CI, 1.23–2.39 and OR 1.50, 95% CI, 1.12–2.00, respectively). Separate analysis of randomized and propensity score matched studies showed sustained increased risk of stroke with DHCA in contrast to MHCA and mild HCA (OR, 1.61, 95% CI, 1.18–2.20, p value = .0029 and OR, 1.74, 95% CI, 1.09–2.77, p value = .019). Conclusions In the included studies, the moderate to mild hypothermia strategies were associated with decreased operative mortality and the risk of postoperative stroke. Large‐scale prospective studies are warranted to further explore appropriate temperature management for the treatment of aortic arch pathologies.


| INTRODUCTION
Different hypothermia regimens alone or combined with selective cerebral perfusion can be used for cerebral protection during surgical interventions involving the aortic arch. Historically, cooling of the brain to profoundly hypothermic levels seemed safe. However, transitions to warmer hypothermic temperatures in conjunction with antegrade cerebral perfusion has become the trend in the last decades. [1][2][3] Despite favorable clinical results reported with newer techniques, a variation in hypothermia regimes persists to exist between centers worldwide.
Evidence on outcome differences between the proposed techniques and different levels of hypothermia during aortic arch surgery is scattered across many publications. Over the last decades, several meta-analyses have compared different levels of hypothermia in different combinations in a pairwise fashion. [4][5][6][7] However, no metaanalysis has compared all three levels of hypothermia, deep, moderate and mild.
Network meta-analysis (NMA) allows comparisons between more than two treatment arms. The advantage of network meta-analysis is that it facilitates indirect comparisons of multiple interventions that have not been studied in a head-to-head fashion. 8

| Assessment of the quality of individual studies and overall quality of evidence
In the studies with a within-study comparison of the hypothermia regimens, the Newcastle-Ottawa quality assessment scale was used to assess bias in observational studies. 11 In case of a randomized controlled trial, the Cochrane risk of bias tool was used. 12

| Outcome measures
The primary outcome was operative mortality. Secondary outcomes were postoperative incidence of stroke and AKI.

| Statistical analysis
Sample-sized weighted pooled baseline patient and procedural characteristics were calculated for each hypothermia regimen group. Early event risks were pooled using inverse variance weighting and pooled in a random-effects model using the Der Simonian and Laird method to estimate the between-study variance. 13 Random-effect meta-analysis was performed using 'metafor' and 'meta' packages in R (version 4.0.5., R Project for Statistical Computing).

| Network meta-analysis
Odds ratios (ORs) were used for the early outcome and were calculated by extracting the raw data from the studies. NMA was performed using the frequentist method (generic inverse variance method) with the 'netmeta' statistical package in R (version 4.0.5., R Project for Statistical Computing), as described by Rücker et al. 14 Random effect network meta-analyses were performed to make direct and indirect comparisons of two-and three-arm studies comparing different levels of hypothermia in aortic arch surgery. 15 Inconsistency in NMA was evaluated by conducting conventional pairwise meta-analyses and comparing direct and indirect OR, also called node-splitting. Additionally, quadratic net heat plots were computed to investigate inconsistency. Heterogeneity was reported as low (I 2 = 0-25%), moderate (I 2 = 26-50%), or high (I 2 > 50%).

| Sensitivity analyses
To minimize possible confounding that can be encountered in observational research, data from randomized controlled trials (RCTs) and propensity score matched (PSM) studies were analyzed separately. Tables S1 and S3.

| Study and patient characteristics and clinical outcome
Individual study characteristics are presented in Supporting Information:

| Network meta-analysis
The network graph illustrates that the most common comparison performed in our network analysis was between DHCA and MHCA ( Figure 5).
Differences found in the pairwise comparisons sustained their significance in the network meta-analysis. The use of DHCA and MHCA were associated with significantly higher operative mortality compared with the use of mild HCA (OR, 1.71, 95% CI 1.23-2.39, p value = .0014 and OR, 1.50, 95% CI 1.12-2.00, p value = .007, respectively). Utilization of DHCA was associated with a higher incidence of postoperative stroke compared with the use of MHCA (OR, 1.46, 95% CI, 1.19-1.78, p < .001) and mild HCA (OR, 1.50, 95% CI, 1.14-1.98, p value = .004).
No difference in the postoperative incidence of AKI was found between the different hypothermia levels ( Figure 5). The net heat plot for outcome AKI showed relatively higher inconsistency between the different comparisons compared with other outcome measures (Supporting Information: Figure S3).

| RCTs and PSM studies sensitivity analysis
Two RCTs and seven PSM studies included 5425 patients. Baseline characteristics of the included studies are detailed in Supporting Information: Table 5.  Etiology  The network graph is displayed in Supporting Information:  Figure S6).
Indirect and direct estimates did not differ significantly, with low inconsistency (Figure 6, Supporting Information: Figure S5 and S6).
Differences between effect estimates based on direct evidence were not significant. Heterogeneity/inconsistency were not significant for operative mortality(p = .72), stroke (p = .47) and AKI (p = .45). Quadratic net heat plots for outcomes postoperative stroke and mortality are presented in Supporting Information: Figures S7 and S8. Heterogeneity was low for all outcome measures (I 2 = 0%).

| DISCUSSION
The optimal cerebral protection strategy during aortic arch surgery remains controversial. The main finding of this network meta-analysis is that the application of moderate and mild hypothermia in combination with selective cerebral perfusion is associated with lower incidence of postoperative stroke, when compared with DHCA alone or in combination with any selective cerebral perfusion strategy. To the best of our knowledge, this is the first network meta-analysis on the three most commonly used hypothermia levels.
Historically, the implementation of DHCA seemed to be safe under 40 min. 16 The general impression exists among the surgeons that lower temperatures are safer. Despite well-recognized complications of deep and profound hypothermia, wide-ranged F I G U R E 4 Forest plot of net split results of direct, indirect evidence and network estimates for the postoperative incidence of acute kidney insufficiency.
F I G U R E 5 Network graphs for all comparisons of the reported outcomes. The thickness of the beams indicates how commonly a comparison was found in the network analysis, in relation to the other comparisons.
F I G U R E 6 Forest plot of net split results of direct, indirect evidence and network estimates for the postoperative incidence of stroke (randomized controlled trials and propensity score matched studies).
implementation of moderate temperatures is not fully adopted. This emerges from concerns that still exist due to suboptimal organ protection of non-DHCA approaches, even with the addition of selective cerebral perfusion. 17 Recently the collaborative efforts of the ARCH registry have resulted in more published global data on this topic. Keeling et al. 18 reported studies. 6 They concluded that MHCA significantly reduced the postoperative incidence of renal failure and the need for renal replacement therapy. Unfortunately, we were unable to find any differences in the postoperative occurrence of AKI.
The shift towards warmer temperatures persists but is not uniform. 17 The same restraint exists in the adaptation of more evolved selective cerebral perfusion techniques, such as unilateral cerebral perfusion. 19 Skeptics advocate that it might not sufficiently supply the contralateral hemisphere and cause undetectable transient neurological injuries by conventional imaging methods. 20 As the debate about optimal cerebral protection continues, this comprehensive network-meta analysis incorporating the existing data illustrates that despite the current concerns about warm cooling temperatures, moderate and mild hypothermic circulatory arrests in conjunction with selective cerebral perfusion provide improved data are that may support daily practice as well as direct designs of future multicentric randomized controlled trials in this regard.

| Strengths and limitations
The results of this network meta-analysis must be interpreted in light