Impact of Mild Hypothermia As Adjunctive Therapy in Patients With ST‐Elevation Myocardial Infarction: A Meta‐Analysis and Trial Sequential Analysis of Randomized Controlled Trials

ABSTRACT

Background

The prevention of reperfusion injury remains an unmet need in ST‐elevation myocardial infarction (STEMI) patients. Several randomized controlled trials (RCTs) evaluated mild hypothermia as adjunctive therapy during STEMI, with conflicting results.

Aims

To summarize the evidence about the efficacy and safety of mild hypothermia in patients with STEMI, as well as its conclusiveness through a trial sequential analysis (TSA).

Methods

PubMed and Scopus electronic databases were screened for eligible studies until August 12, 2024. Efficacy endpoints were all‐cause death, infarct size (IS), left ventricular ejection fraction (LVEF), the occurrence of microvascular obstruction (MVO), thrombolysis in myocardial infarction (TIMI) flow grade 3, and the resolution of ST‐segment elevation (i.e., > 50−70% from baseline) after the procedure. Safety endpoints included: the incidence of atrial fibrillation (AF), infections, any bleeding, major bleeding, acute and subacute stent thrombosis (STh), cardiogenic shock/pulmonary oedema, and ventricular fibrillation/tachycardia. “Door‐to‐balloon time” was indicated as the procedural endpoint. Two pre‐specified subgroup analyses were planned according to the mean ischemic time and the site of hypothermia induction (intra‐coronary vs. extra‐coronary). A TSA was run to explore whether the effect estimate of each efficacy outcome could be influenced by further studies.

Results

Ten RCTs were included. Hypothermia did not provide a benefit for any of the specified efficacy endpoints. Furthermore, it enhanced the risk of infection, the risk of STh in patients with a mean ischemic time of less than 4 h, and the risk of AF in patients undergoing extra‐coronary hypothermia. Finally, it was also associated with an increased “door‐to‐balloon time”, and a trend toward an increased risk of any bleeding. No significant difference was found for the other endpoints. TSA showed conclusive evidence of an absence of benefit of hypothermia on IS, MVO, LVEF, and TIMI three flow.

Conclusions

Mild hypothermia is not beneficial and causes relevant delays in clinical management of STEMI patients, raising safety issues mainly related to the occurrence of STh, AF, and infections.

1. Introduction

Despite considerable advancements in revascularization strategies and medical therapies, the prognosis of patients with ST‐Elevation Myocardial Infarction (STEMI) remains poor [1]. The damage occurring during acute myocardial infarction is not only attributable to ischemia but also to the subsequent reperfusion, which further exacerbates tissue injury in the ischemic area [2]. Most treatments incorporated into daily clinical practice have focused on tackling the first of these two mechanisms, through the timely re‐opening of the culprit vessel. Conversely, strategies to reduce reperfusion injury and prevent adverse left ventricular remodeling have shown conflicting results [2, 3, 4, 5]. The latter may be due to the multifactorial and self‐amplifying nature of the phenomenon, which renders difficult to restore tissue homeostasis with a single treatment at a time, once the pathophysiological cascade has been triggered [2, 6]. In the pre‐clinical setting, one of the most successful strategies was therapeutic hypothermia (TH) [7]. In animal models of coronary artery ligation, TH reliably and markedly reduced infarct size (IS), attenuating adverse left ventricular remodeling [8]. Consequently, several randomized controlled trials (RCTs) have been conducted in STEMI patients to evaluate the clinical efficacy and safety of this approach [9, 10, 11, 12, 13, 14, 15, 16, 17, 18]. Due to the small number of patients enrolled, most of the studies resulted largely underpowered to detect any statistically significant differences for clinical endpoints, and inconclusive in terms of efficacy and safety. Several meta‐analyses have been conducted in this research area with conflicting results [19, 20, 21]. In addition, a trial sequential analysis (TSA) has never been conducted with this approach. The purpose of this meta‐analysis and TSA was to comprehensively assess the efficacy and safety of TH in STEMI patients without cardiac arrest, as well as the impact of procedural variables and baseline patient characteristics on pre‐specified endpoints.

2. Methods

This systematic review was conducted according to the current Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidelines [22]. The PRISMA checklist is available in “Supporting Information S1: Materials, Table 1.” The study protocol was registered within the PROSPERO International Prospective Register of Systematic Reviews with the following ID: CRD42024582111.

2.1. Search Strategy and Selection Criteria

We performed a comprehensive search of two electronic databases (i.e., PubMed and Scopus) from inception until August 12, 2024. The systematic search strategy and search terms are reported in “Supporting Information S1: Method 1.” Studies were eligible based on the following criteria: (1) RCT design; (2) comparing mild hypothermia (irrespective of the cooling method) as an adjunct treatment to standard of care versus conventional therapy in patients with STEMI, undergoing primary percutaneous coronary intervention; (3) reporting at least one of the outcomes of interest; (4) articles of each language were allowed. Two different investigators (R.L. and D.A.P.) independently screened titles and abstracts for relevance. The reference lists of reviews and relevant articles was also checked. Discrepancies in study selection were resolved by consensus.

2.2. Data Extraction and Quality Assessment

Two authors (R.L. and D.A.P.) extracted data from eligible studies using a predefined form. Data extracted included study characteristics (first author's name or trial name, year of publication, and design), patient demographics, interventional protocol details, and endpoints of interest. The risk of bias assessment was conducted using the RoB2 (revised tool for assessing the risk of bias in randomized trials 2.0) and is available in “Supporting Information S1: Table 2” [23].

2.3. Study Endpoints

Outcomes of interest were divided into three categories: efficacy endpoints, safety endpoints, and procedural endpoints. Efficacy endpoints were all‐cause death, IS, left ventricular ejection fraction (LVEF) after percutaneous revascularization, microvascular obstruction (MVO), the achievement of a thrombolysis in myocardial infarction (TIMI) flow grade 3 and the resolution of ST‐segment elevation (i.e., > 50/70% from baseline) after percutaneous revascularization. Safety endpoints included the incidence of atrial fibrillation (AF), infections, any bleeding, major bleeding, acute and subacute stent thrombosis (STh), cardiogenic shock/pulmonary oedema, and ventricular fibrillation/tachycardia (VF/VT). “Door‐to‐balloon time” was the procedural endpoint. All these data were collected at the longest follow‐up available. Further details on endpoints are provided in “Supporting Information S1: Method 1, and Supporting Information S1: Tables 3 and 4.”

2.4. Statistical Analysis

Risk ratios (RRs) with 95% confidence intervals (CIs) were calculated for discrete variables, whereas for continuous variables mean difference (MD) together with 95% CIs and p values represented the summary statistics. The extent of heterogeneity among the included studies was assessed using the Cochran's Q test and Higgins' I ² statistics, with I ² < 20%, 20% < I ² < 75%, I ² < 75%, respectively representing mild, moderate, and severe inconsistency. p values less than 0.05 (two‐tailed) were considered statistically significant. The choice of statistical model for the analysis was determined by the degree of heterogeneity: the fixed‐effects model in case of low heterogeneity and the random‐effects model (restricted maximum likelihood) in case of at least moderate heterogeneity. Leave‐one‐out analysis was performed to evaluate the influence of removing individual studies on the pooled effect size. In case of at least moderate heterogeneity, Galbraith plot analysis was performed to assess the source of heterogeneity. Two pre‐specified subgroup analyses were performed according to the site of hypothermia induction (i.e., intra‐coronary vs. extra‐coronary) and the mean ischemic time (less than 4 h vs. more than 4 h). Furthermore, a post‐hoc subgroup analysis comprising only anterior STEMI patients was planned for efficacy endpoints. Subgroup effects were compared using the Borenstein and Higgins test, and the credibility of subgroup differences was assessed using the ICEMAN tool [24]. A p value less than 0.05 (two‐tailed) was considered statistically significant for subgroup analyses. Meta‐regressions were performed to assess the influence of study‐level covariates [i.e., age, diabetes, male sex, time to target temperature (min), and patients at target temperature (%)] on efficacy endpoints. Publication bias was assessed using funnel plot visual examination and Egger's test. Statistical analyses were performed using Stata 18 (64 bit; StataCorp, College Station, TX). The Hozo's method was used to estimate mean and standard deviation when studies did not report them [25].

2.5. TSA

A TSA was planned to assess precision, uncertainty and conclusiveness of the current evidence regarding the pre‐specified efficacy endpoints. Such analysis allows the calculation of the sample size needed to define a result as “conclusive” or not, and to adjust for the risk of type I error due to repeated significance testing, by plotting the relationship between the Z‐curve of cumulative evidence and a set of benefit or futility boundaries, similarly to an interim analysis in clinical trials [26]. The O'Brien‐Fleming α‐spending function was used, setting a two‐sided type 1 error of 5.0% with a power of 80%. It was run with TSA software version 9.0.5.10 beta.

3. Results

A total of 1863 potentially relevant articles were identified through the literature search. After assessing eligibility, 10 studies met the inclusion criteria and were included in the meta‐analysis, with a total of 806 patients involved (i.e., 414 patients in TH group and 392 in the control group) [9, 10, 11, 12, 13, 15, 16, 17, 18, 27]. The complete flow chart of the study analysis is available in “Supporting Information S1: Table 5.” The methods of hypothermia induction varied across included studies, comprising intracoronary cooling infusion (2 RCTs, 260 patients), intravenous catheter cooling (7 RCTs, 492 patients), and peritoneal cooling (1 RCT, 54 patients). The target temperature ranged from 33°C to 35°C, with duration of cooling varying from 1 to 3 h post‐reperfusion. The cooling infusion was started before revascularization in the studies with endovascular and peritoneal approach, whereas the hypothermia induction was provided during the revascularization procedure for intracoronary cooling protocols. Median follow‐up was 1 month (interquartile range 1−6 months). Table 1 shows the main features of the included studies.

Table 1.

Baseline clinical and procedural characteristics of included RCTs.

Trial/First name and year of publication	Type of STEMI	Methods of hypothermia	Procedural description of hypothermia	Mean ischemic time	Imaging method for IS measurement	Imaging method for LVEF measurement	Clopidogrel (%)	Prasugrel/Ticagrelor (%)
RAPID‐MI‐ICE Trial (2010)	Anterior: 13/18 (72%), Non‐anterior: 6/18 (33%)	Endovascular	Hypothermia was induced before PCI by an intravenous infusion of 4°C cold saline using pressure bags, then maintained through an endovascular catheter (14 Fr) placed in IVC (target temperature 33°C). Hypothermia was continued for 3 h after PCI.	174 min	CMR	N/A	100%	0%
COOL‐MI InCor Trial (2021)	Anterior: 16/50 (32%), Inferior: 31/50 (62%)	Endovascular	Hypothermia was induced before PCI by an intravenous infusion of 4°C cold saline using pressure bags, then maintained through an endovascular catheter (12 Fr) placed in IVC (target temperature 32°C). Hypothermia was continued for 1−3 h after PCI.	370 min	CMR	CMR	46%	52%
COOL‐MI Trial (2002)	Anterior: 19/42 (45%), Inferior: 23/42 (55%)	Endovascular	Endovascular cooling before PCI with an endovascular catheter placed in the IVC via the femoral vein (target temperature 33°C). Hypothermia was continued for 3 h after PCI.	204.5 min	99 mTc‐sestamibi SPECT imaging.	N/A	N/A	N/A
CHILL‐MI Trial (2014)	Anterior: 51/120 (42,5%), Inferior: 68/120 (56,6%)	Endovascular	Hypothermia was initially induced by forced infusion of 4°C cold saline using pressure bags. The hypothermia was then maintained by an endovascular catheter (14 Fr) placed into the IVC with the tip of the catheter at the level of the diaphragm (target temperature was set to 33°C). Cooling was maintained for 1 h after reperfusion.	130.5 min	CMR	CMR	14%	86%
VELOCITY Trial (2014)	Anterior: 25/58 (46.2%), Non‐anterior: 28/54 (51,8%)	Peritoneal hypothermia	Hypothermia to was induced before PCI by lavaging the peritoneal cavity with temperature‐controlled lactated Ringer's solution using a of a peritoneal lavage kit. Hypothermia was rapidly induced (target temperature ≤ 34.9°C, controlled by an esophageal temperature probe) before intervention and maintained for 3 h post‐PCI.	169.75 min	CMR	CMR/TTE	37%	39%
EURO‐ICE Trial (2024)	Anterior: 200/200 (100%)	Intracoronary	Hypothermia was started 7−10 min before reperfusion. Hyperthermia was induced by an OTWB, inflated to 4 atm at the location of the occlusion to prevent reperfusion, connected to two parallel contrast infusion pumps (one filled with saline at room temperature and the other filled with saline at 4°C). Distal temperature was monitored by a coronary PW capable of also measuring temperature positioned distal to lesion. Tympanic temperature was measured to monitor systemic temperature.	152.5 min	CMR	CMR	N/A	N/A
COOL‐AMI EU (pilot) trial (2017)	Anterior: 50/50 100%	Endovascular	Hypothermia was induced before PCI by an intravenous infusion of up to 1 L of 4°C cold saline using pressure bags and continued through an endovascular catheter placed in IVC (target temperature 32°C). Hypothermia was continued for 3 h after PCI.	238 min	CMR	CMR	20%	80%
COOL‐AMI EU (pivotal) Trial (2021)	Anterior 100%	Endovascular	Hypothermia was induced before PCI by an intravenous infusion of up to 1 L of 4°C cold saline using pressure bags and continued through an endovascular catheter placed in IVC (target temperature 32°C). Hypothermia was continued for 3 h after PCI.	201 min	CMR	CMR	0%	100%
Testori et al. [16]	Anterior: 52/101 (51,5%), Inferior: 49/101 (48.5%)	Endovascular + external cooling pads	Hypothermia was induced before PCI by cold saline intravenous infusion and external cooling pads application. In the cath‐lab the hypothermia was maintained with an endovascular cooling catheter (14 Fr) placed in IVC (target temperature 34°C). Hypothermia was for 60 min after reperfusion.	186 min	CMR	CMR	0%	100%
Wang et al. [17]	Anterior: 27/60 (45%), Non‐anterior: 33/60 (55%)	Intracoronary	Intracoronary hypothermia was performed before reperfusion through an aspiration catheter connected to an infusion pump (infused cold saline [4°C]). Target distal coronary temperature was 6°C below the body temperature (detected by a sensor‐tipped coronary pressure wire).	274.5 min	CMR	CMR	N/A	N/A

Abbreviations: CMR, cardiovascular magnetic resonance; IS, infarct size; IVC, inferior vena cava; LVEF, left ventricular ejection fraction; MI, myocardial infarction; N/A, not available; OTWB, over‐the‐wire balloon; PCI, percutaneous coronary intervention; PW, pressure wire; RCT, Randomized controlled trial; SPECT, Single‐photon emission computed tomography; STEMI, acute ST‐elevation myocardial infarction; TTE, transthoracic echocardiogram.

3.1. Risk of Bias

Supporting Information S1: Table 2 showed the risk of bias of RCTs included in the meta‐analysis, with only one study presenting a high risk of bias (i.e., COOL‐AMI EU [pivotal] trial) [15].

3.2. Efficacy Endpoints

TH as an adjunctive therapy did not reduce all‐cause death compared to control group (RR 1.09, 95% CI 0.52−2.28, I ² = 0%, p = 0.82, 10 RCTs, 806 patients) in STEMI patients (Figure 1). This result was consistent in the leave‐one‐out analysis (Supporting Information S1: Figure 1). Neither IS, defined as a percentage of left ventricular mass, nor LVEF after revascularization significantly differed between the two groups (MD −0.52, 95% CI −2.40 to 1.36, I ² = 33.04%, p = 0.59, 9 RCTs, 686 patients) (MD −0.11, 95% CI −1.72 to 1.50, I ² = 24.31%, p = 0.89, 8 RCTs, 746 patients), respectively (Figure 2). Similarly, MVO, defined as a percentage of left ventricular mass at cardiac magnetic resonance (MD 0.11, 95% CI −0.04 to 0.26, I ² = 0%, p = 0.16, 7 RCTs, 654 patients), TIMI flow grade 3 post‐revascularization (RR 1, 95% CI 0.87−1.14, I ² = 0%, p = 0.97, 8 RCTs, 505 patients) and ST‐segment resolution (RR 1.07, 95% CI 0.84−1.36, I ² = 0%, p = 0.60, 3 RCTs, 234 patients) were not affected by the addition of TH to standard treatment (Supporting Information S1: Figures 2–4). These results were consistent at the leave‐one‐out sensitivity analysis (Supporting Information S1: Figure 1). At the Galbraith plot visual examination, the study by Wang et al. resulted the main outsider for IS and LVEF endpoints (Supporting Information S1: Figure 5) [17]. In the TSA for IS, MVO, TIMI 3 flow, and LVEF after revascularization, the Z‐curve crossed the required information size, therefore suggesting that the current evidence was conclusive for the lack of benefit for the above endpoints (Figure 3) (Supporting Information S1: Figure 21). Conversely, for all‐cause death, the required information size was not achieved, and the evidence should be considered inconclusive (Supporting Information S1: Figure 21). Since only three trials reported ST‐segment resolution as an endpoint, TSA was not performed.

Figure 1.

Forest plot for all‐cause death. The boxes are proportional to the weight of each study in the analysis, and the lines represent their 95% confidence intervals (CIs). The open diamond represents the pooled relative risk, and its width represents its 95% CI. SD, standard deviation. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 2.

Forest plot for infarct size and LVEF. The boxes are proportional to the weight of each study in the analysis, and the lines represent their 95% confidence intervals (CIs). The open diamond represents the pooled median difference, and its width represents its 95% CI. LVEF, left ventricular ejection fraction; SD, standard deviation; REML, Restricted Maximum Likelihood. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 3.

Trial sequential analysis for infarct size. [Color figure can be viewed at wileyonlinelibrary.com]

3.3. Safety Endpoints

On pooled analysis, TH group was associated with a higher risk of infections compared with control group (RR 2.96, 95% CI 1.34−6.51, I ² = 0%, p = 0.01, 7 RCTs, 504 patients) (Figure 4). Furthermore, in the TH group, an increase at the edge of significance was found in the risk of AF (RR 2.25, 95% CI 0.85−5.95, I ² = 48.2%, p = 0.08, 5 RCTs, 531 patients), STh (RR 2.46, 95% CI 0.92‐6.57, I ² = 0%, p = 0.08, 6 RCTs, 525 patients), any bleeding (RR 2.05, 95% CI 0.97‐4.32, I ² = 0%, p = 0.06, 8 RCTs, 645 patients) (Supporting Information S1: Figures 6–8). Conversely, no significant difference was found for VT/VF (RR 1.30, 95% CI 0.85−1.98, I ² = 0%, p = .23, 10 RCTs, 806 patients), major bleeding (RR 1.85, 95% CI 0.60‐5.74, I ² = 0%, p = 0.23, 7 RCTs, 595 patients), and pulmonary edema/cardiogenic shock (RR 1.97, 95% CI 0.89–4.37, I ² = 0%, p = 0.1, 7 RCTs, 633 patients) (Supporting Information S1: Figures 8–10). At the leave‐one‐out sensitivity analysis, the removal of CHILL‐MI and EURO‐ICE trials rendered the result for AF statistically significant, whereas the removal of CHILL‐MI rendered the result for any bleeding statistically significant (Supporting Information S1: Figures 11, 12) [11, 18]. For all the other endpoints, the results were consistent at the leave‐one‐out analysis (Supporting Information S1: Figure 11). At Galbraith plot analysis, no single study was found to be contributor to the heterogeneity for AF (Supporting Information S1: Figure 12).

Figure 4.

Forest plot for infection. The boxes are proportional to the weight of each study in the analysis, and the lines represent their 95% confidence intervals (CIs). The open diamond represents the pooled relative risk, and its width represents its 95% CI. [Color figure can be viewed at wileyonlinelibrary.com]

3.4. Procedural Endpoint

On pooled analysis, TH resulted in a significant prolonged “door‐to‐balloon time” compared to control group (MD 12.12 min, 95% CI 6.35−17.88, I ² = 82.9%, p < 0.01, 10 RCTs, 806 patients) (Figure 5). These results were consistent at the leave‐one‐out sensitivity analysis (Supporting Information S1: Figure 1). According to Galbraith plot analysis, three studies mainly contributed to heterogeneity: COOL‐MI trial, COOL‐AMI EU (Pivotal) trial, and RAPID‐MI ICE trial [10, 12, 15].

Figure 5.

Forest plot for door to balloon time. The boxes are proportional to the weight of each study in the analysis, and the lines represent their 95% confidence intervals (CIs). The open diamond represents the pooled median difference, and its width represents its 95% CI. REML, Restricted Maximum Likelihood; SD, standard deviation. [Color figure can be viewed at wileyonlinelibrary.com]

3.5. Publication Bias

According to Egger's test, publication bias was detected only for IS (Supporting Information S1: Figure 13, Supporting Information S1: Figure 14).

3.6. Subgroup Analyses and Meta‐Regressions

The two pre‐specified subgroup analyses (i.e., site of hypothermia induction and mean ischemic time) showed no significant differences for both efficacy and procedural endpoints (Supporting Information S1: Figure 15 and Supporting Information S1: Figure 16). Furthermore, the post‐hoc subgroup analysis revealed the absence of benefit of hypothermia even in the subset of patients with anterior STEMI (Supporting Information S1: Figure 17).

Conversely, TH was associated with a greater risk of STh in the subgroup of patients with a mean ischemic time less than 4 h (RR 3.64, 95% CI 1.04−12.74, I ² = 0, p = 0.04, p for interaction = 0.2) (Figure 6). Furthermore, TH conferred an increased risk of AF in the subset of patients undergoing extra‐coronary hypothermia (RR 2.90, 95% CI 1.21−6.98; I ² = 37.32%; p = 0.02; p for interaction = 0.06). According to the ICEMAN tool, both subgroups should be considered of low credibility because effect modification was based on between‐trial rather than within‐trial comparisons, and the number of RCTs included was small (Figure 6). No differences were found in the subgroup analyses for the other safety endpoints (Supporting Information S1: Figure 18 and Supporting Information S1: Figure 19). Meta‐regression analyses showed no significant relationship between the covariates and the analyzed outcomes (Supporting Information S1: Figure 20). Due to the low numbers of trials reporting data regarding ST‐segment resolution (i.e., 3 RCTs), subgroup analyses and meta‐regressions were not performed for this endpoint.

Figure 6.

Forest plot subgroup analyses for stent thrombosis and atrial fibrillation. The boxes are proportional to the weight of each study in the analysis, and the lines represent their 95% confidence intervals (CIs). The open diamond represents the pooled median difference, and its width represents its 95% CI. REML, Restricted Maximum Likelihood; SD, standard deviation. [Color figure can be viewed at wileyonlinelibrary.com]

4. Discussion

The main findings of the current meta‐analysis, encompassing 806 STEMI patients from 10 RCTs, were the following: (1) the use of TH did not provide a benefit in terms of all‐cause death, IS size, TIMI flow, MVO according to cardiac magnetic resonance, ST‐resolution, and LVEF, regardless the site of hypothermia induction (intra‐coronary vs. extra‐coronary), mean ischemic time, procedural characteristics (i.e., time until target cooling was achieved and percentage of patients reaching target cooling), and baseline patients characteristics (i.e., age, sex, diabetes, and tobacco use); (3) TH increased the risk of infection, the risk of STh in patients with a mean ischemic time of less than 4 h, and the risk of AF in patients undergoing extra‐coronary hypothermia; (5) a trend toward an increased incidence of any bleeding was also found in the TH group; (6) importantly TH prolonged the “door‐to‐balloon” time.

To our knowledge, the current meta‐analysis was the first to report, through a TSA, solid evidence that TH, if performed with existing protocols, provides no benefit in terms of IS, MVO, LVEF, and TIMI 3 flow after revascularization. Moreover, it raised further safety concerns in some patient subgroups (i.e., STh in those with a mean ischemic time < 4 h, and AF in the subgroup undergoing extra‐coronary hypothermia).

4.1. Pathophysiologic Rationale for TH in Reducing IS

The use of TH stem from preclinical studies showing its protective effect in animal models of coronary occlusion and brain ischemia, with a direct proportionality between temperature decrease and cardio‐protection (i.e., 10% reduction in IS for every 1°C decrease in body temperature) [8, 28, 29, 30]. Nevertheless, its protective role in STEMI patients with and without cardiac arrest is still debated [31, 32]. A critical point is the rapidity of induction of TH, as its cardio‐protective effect in animal models has been shown to increase progressively as normothermic ischaemic time diminishes [33]. Furthermore, animal models have shown that the beneficial effect of TH is nullified if it is induced at or after vessel reopening when reperfusion injury has already started [33]. These considerations may hamper the establishment of a potentially effective TH induction protocol in daily clinical practice.

4.2. IS, MVO, and All‐Cause Death

In the current meta‐analysis, TH provided no benefit in any of the efficacy endpoints. These findings were in contrast to previously published meta‐analyses showing a benefit of TH in reducing IS, especially in patients with anterior STEMI and when a temperature below 35° was achieved [20, 21, 34]. The reason for this difference may lie in the included studies. The previous work only included a subset of studies, some of which were never published, but only presented at conferences [35], and the benefit of TH in reducing IS was nullified by the removal of that study at leave‐one‐out sensitivity analysis [20]. Conversely, in our meta‐analysis, we included all available studies for which we disposed of a full paper, increasing statistical power and accuracy, and reducing reporting bias. Also, it is possible that the benefits of TH on IS may be overestimated. Visual inspection of the funnel plot and Egger's test revealed a publication bias. In the Galbraith plot examination, the principal outliner was the study from Wang et al. [17]. Compared to the other studies, Wang et al. was the only effective in reducing IS, reporting the shortest hypothermia induction time (i.e., 31 s) and the largest temperature drop (i.e., 5.8°) [17]. Due to the relevance of the time to reach target temperature and the intensity of hypothermia achieved, we performed a meta‐regression using as moderators the mean time to reach hypothermia and the percentage of patients who reached target temperature. These variables did not to affect the effect size of the pre‐specified efficacy endpoints. Therefore, TH does not appear to provide any benefit in humans, regardless of the magnitude of hypothermia achieved and the rate at which it is achieved.

4.3. AF

To the best of our knowledge, our meta‐analysis was the first to find an increased risk of AF in patients undergoing extra‐coronary TH. Apart from the immediate reduction in cardiac output caused by loss of atrial contraction and increased irregular heart rate, the occurrence of AF during acute myocardial infarction has been associated with a worse long‐term prognosis and increased risk of heart failure [36]. There may be several reasons why extra‐coronary hypothermia increases the risk of AF. At the level of cardiac cells, mild hypothermia induces conduction delay in atrial and atrioventricular node cells, which could promote the formation of re‐entry microcircuits [37, 38]. Furthermore, extra‐coronary hypothermia through infusion of cold saline solutions could lead to volume overload, ultimately increasing left ventricular filling pressures and promoting atrial myocyte stretching with augmented arrhythmogenicity [39]. Conversely, the volume of saline solution injected in the intracoronary protocol is lower and it is restricted to the obstructed vessel site, which may minimize the systemic atrial pro‐arrhythmogenic effect [17, 18].

4.4. STh

To the best of our knowledge, our meta‐analysis was the first to show an increased risk of STh. This phenomenon reached statistical significance in patients with a mean ischemic time less than 4 h. There could be several reasons for this finding. First, mild hypothermia is known to induce a pro‐thrombotic state via increased platelet aggregation from both preclinical and clinical studies [40, 41]. Second, absorption and initiation of P2Y12 inhibition can be delayed during mild hypothermia, as demonstrated in comatose cardiac arrest survivors subjected to hypothermia with a “P2Y12 inhibition gap” of almost 3 h both with clopidogrel and newer oral agents (i.e., ticagrelor and prasugrel) [42]. Moreover, co‐administration of opioids as part of the anti‐shivering protocol further impairs gastrointestinal mobility and P2Y12 absorption. Therefore, the delayed onset of the antiplatelet effect of P2Y12 inhibitors could explain why the risk of STh was higher in the subgroup with an ischemic time of less than 4 h. Conversely, in the other subgroup (i.e. mean ischemic time > 4 h), the antiplatelet agents would have more time to achieve their inhibitory effect. Of note, cangrelor, an intravenous P2Y12 inhibitor with immediate onset of action, may have represented an alternative strategy to bridge the non‐inhibition gap and reduce the risk of STh, especially in the subset of patients with a short mean ischemic time [43].

4.5. Infection

TH increased the risk of infection. This finding was mainly driven from RCTs including patients with a mean ischemic time of less than 4 h and with induction of extra‐coronary hypothermia. Conversely, the incidence of infection was not assessed in any of the studies using an intracoronary induction protocol. An increased risk of infection has also been reported in previous meta‐analyses and in RCTs conducted in other clinical settings (i.e., traumatic brain injury, stroke, cardiopulmonary bypass) [19, 44]. Hypothermia is known to reduce the secretion of pro‐inflammatory cytokines and to inhibit leukocyte migration and phagocytosis [45]. In this regard, unintended hypothermia during surgery has been strongly associated with an increased risk of wound infection [46]. Furthermore, the access site for the administration of the cold solution and the solution itself could be potential carriers of pathogens, if not properly treated. Infection prevention strategies through antibiotics could have been contemplated.

4.6. Bleeding

In the current meta‐analysis, TH group had a numerically higher rate of any bleeding at the edge of statistical significance (p = 0.06), with no difference in major bleeding. Hypothermia has also been associated with a coagulation dysfunction, probably due to decreased fibrinogen levels and thrombin generation from impaired secondary haemostasis [47, 48]. In addition, the extra‐coronary hypothermia induction protocol requires venous central vascular access to inject cold saline solutions, which could be one of the sites of bleeding. Furthermore, in the early phase after STEMI, the risk of bleeding is inherently increased, as patients require vascular access and are exposed to adjuvant intra‐procedural antithrombotic therapies [49]. The occurrence of bleeding, even when minor, worsens the prognosis of patients undergoing percutaneous myocardial revascularization, thus blunting the benefit of any additional therapy [50, 51].

4.7. Door‐To‐Balloon Time

TH increased the door‐to‐balloon time by an average of 12 min, regardless of hypothermia site of induction. This finding was not unexpected considering the time required to induce hypothermia before flow is restored. Of note, high heterogeneity was observed for this endpoint. Based on visual analysis of the funnel and Galbraith plots, the main contributors were COOL‐MI and COOL‐AMI (pivotal and pilot) trials [9, 10, 15]. COOL‐MI was the study with the shortest door‐to‐balloon time. This may be explained by the fact that cooling was started in the emergency room before arrival in the catheterization room and it only involved the endovascular cooling catheter without cold saline injection [10]. Conversely, COOL‐AMI trials, combining both saline and intravascular catheter cooling and initiated only in the catheterization room, had the longest “door‐to‐balloon” time [9, 15]. Minimizing the time to induction of hypothermia should constitute an area of development for future studies to maximize the potential effectiveness of any strategy, as prolonging the time to ischemia worsens STEMI patient prognosis [52].

4.8. Limitations

Our meta‐analysis has certain limitations. First, the absence of patient‐level data hinders the assessment of baseline clinical and procedural characteristics that may potentially impact safety and efficacy outcomes. Second, the result of subgroup analysis is per se hypothesis‐generating [53]. The different risk of STh according to the mean ischemic time may be due to the smaller number of patients in the subgroup including RCTs with mean ischemic time greater than 4 h, which may render the analysis underpowered, as indicated by a non‐significant p for interaction. Conversely, the differential effect of hypothermia on the risk of AF according to the site of induction (i.e., intra‐coronary vs. extra‐coronary) appears to be more robust, as indicated by a p for interaction at the limits of significance (i.e., p = 0.06). Nevertheless, the influence on outcomes of some covariates with unequal distribution in the two subgroups cannot be excluded. Third, the hypothermia protocol was not standardized among the included studies. However, heterogeneity can be an asset to improve the identification of specific biological effects of TH, while accounting for differences in protocols. Fourth, the number of enrolled patients is small. This may lead to underpowered analyses, especially for hard endpoints. However, the lack of benefit also applies to soft endpoints, for which fewer patients would be needed.

5. Conclusion

The routine use of TH is not supported by RCTs conducted to date, as it leads to delayed reperfusion and increases the risk AF, infection, and STh, without providing any beneficial effect.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting information.

Data Availability Statement

The data underlying this article will be shared on reasonable request to the corresponding author.