Warm versus cold cardioplegia in cardiac surgery: A meta-analysis with trial sequential analysis

Objective This meta-analysis aimed to compare clinical outcomes of warm and cold cardioplegia in cardiac surgeries in adult patients, with trial sequential analysis (TSA) used to determine the conclusiveness of the results. Methods Electronic searches were performed on PubMed, Medline, Scopus, EMBASE, and Cochrane library to identify all studies that compared warm and cold cardioplegia in cardiac surgeries. Primary end points were in-hospital or 30-day mortality, myocardial infarction, low cardiac output syndrome, intra-aortic balloon pump use, stroke, and new atrial fibrillation. Secondary end points were acute kidney injury, hospital length of stay, and intensive care unit length of stay. Prespecified subgroup analyses were performed for (1) studies published since publication of Fan and colleagues in 2010, (2) randomized controlled studies, (3) studies with low risk of bias, (4) coronary artery bypass graft surgeries, and (5) studies with cold blood versus those with cold crystalloid cardioplegia. TSA was performed to determine conclusiveness of the results, using on all outcomes without significant heterogeneity from studies of low risk of bias. Results No significant differences were found between post-operative rates of mortality, myocardial infarction, low cardiac output syndrome, intra-aortic balloon pump use, stroke, new atrial fibrillation, and acute kidney injury between warm and cold cardioplegia. TSA concluded that current evidence was sufficient to rule out a 20% relative risk reduction in these outcomes. Conclusions Concerning safety outcomes, current evidence suggests that the choice between warm and cold cardioplegia remains in the surgeon's preference.


Favours Cold
Favours Warm risk of membrane destabilization, intracellular edema, calcium sequestration, and time for heart rewarming; and decrease the risk of reperfusion injury. Besides, blood was considered to be better than crystalloid cardioplegia due to its greater oxygen-carrying and buffering capacity, better microvascular flow secondary to rheologic effects, and less associated intracellular edema. 1 The Warm Heart Investigators 2 conducted a randomized controlled trial (RCT) of 1732 patients in 1994. They demonstrated a significant reduction in postoperative low cardiac output syndrome (LCOS) in the warm cardioplegia group, without significant differences in 30-day all-cause mortality, postoperative myocardial infarction (MI), and stroke. A meta-analysis on RCTs by Fan and colleagues 3 showed no significant difference in the clinical outcomes investigated. However, it was unclear whether the results were conclusive. This systemic review and meta-analysis aimed to compare clinical outcomes of warm versus cold cardioplegia in adult cardiac surgeries, updating the metaanalysis by Fan and colleagues 3 with more recent evidence, further analyzing the conclusiveness of the results.

METHODS
This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and methods stipulated in the Cochrane Handbook for Systematic Review of Interventions. 4,5 It has been submitted to PROSPERO with a registration number of CRD42020171613 but had not been approved as of the time of submission.

Search Strategy and Selection Criteria
Electronic searches were performed on PubMed, Medline, Scopus, EM-BASE, and Cochrane library to identify all studies comparing warm and cold cardioplegia in cardiac surgeries regardless of publication type or language. All databases were searched since the search of previous metaanalysis (Fan and colleagues 3 ) on the topic, up until June 27, 2020. A search was also conducted on ClinicalTrials.gov to identify ongoing or unpublished clinical trials. The search string used was ([warm OR normothermia OR normothermic OR cold OR hypothermia OR hypothermic] AND [cardioplegia OR "myocardial protection"] AND [valve OR valvular OR AVR OR MVR OR DVR OR TVR OR PVR OR "coronary artery bypass graft" OR "coronary artery bypass grafting" OR CABG OR "vein graft" OR "bypass graft" OR "surgical revascularization"]). All search terms searched as both key words and Medical Subject Headings terms to maximize sensitivity. Reference lists of papers found in the literature search were manually searched to assess suitability for inclusion in this review.
Three reviewers performed literature screening (T.K.M.K., J.S.K.C., Shaik Ashraf Bin Shaik Ismail). Articles were first screened based on their titles and abstracts. Full texts of all identified articles were then retrieved and systemically assessed using the inclusion and exclusion criteria for further study. Conflicts over inclusion were resolved by consensus. Articles were deemed eligible for inclusion if warm cardioplegia was compared against cold cardioplegia in cardiac surgeries. Noncomparative studies, conference abstracts or papers, articles involving fewer than 5 patients, and studies including patients younger than 18 years of age were excluded. Studies not reporting any of the end points specified herein were also excluded. Warm cardioplegia was defined as 28 C to 37 C, whereas cold cardioplegia was defined as 4 C to 15 C.
Primary end points were in-hospital or 30-day mortality, MI, LCOS, intra-aortic balloon pump (IABP) use, stroke, and new atrial fibrillation (AF). Secondary end points were acute kidney injury (AKI), hospital length of stay (LOS), and intensive care unit (ICU) LOS. Summary estimates were extracted manually from included studies. Only the most updated data were included wherever duplicate data existed. Study authors were contacted where necessary. Data reported by previous meta-analysis by Fan and colleagues in 2010 3 were also extracted from published Forest plots. Conflicts over data extraction were resolved by consensus.

Statistical Analysis
All included studies were critically appraised by the modified Jadad scale for RCTs or the Newcastle-Ottawa Quality Assessment Scale (NOS) for observational studies. The modified Jadad scale is a numeral scale with components addressing randomization, blinding, selection, adverse effects assessment, and statistical methods. It is described in detail in Table E1. The NOS assessed cohort studies according to selection, comparability, and outcome and is detailed in Table E2. The metaanalysis by Fan and colleagues in 2010 3 was critically appraised by the AMSTAR 2, which is a critical appraisal tool for systematic reviews that includes randomized or nonrandomized studies of health care interventions. 6 All statistical analyses were a priori, specified before the start of data extraction. Odds ratios and 95% confidence intervals (CIs) or weighted mean differences (WMDs) and 95% CIs were used as the main summary measures for baseline characteristics, whereas relative risks (RRs) and 95% CIs or WMD and 95% CIs were used as main summary measures for the outcomes studied. Discrete variables were pooled using the Mantel-Haenszel method with RR as the effect measure. Continuous variables were pooled using the inverse variance method with WMD as the effect measure. Sensitivity analysis is performed by the leave-one-out method. Prespecified subgroup analysis was performed on (1) studies published since publication of Fan and colleagues in 2010 3 ; (2) RCTs; (3) studies with low risk of bias, defined by 5 or 7 score or more in modified Jadad scale or NOS, respectively; (4) coronary artery bypass graft (CABG) surgeries; and (5)  Trial sequential analysis (TSA) can be used to assess conclusiveness of meta-analytical findings. As evidence accumulates, random errors also accumulates and they may incidentally lead to "significant" results reported in meta-analysis. Meta-analyses of cardiovascular and anesthesiologic interventions have many false positions and negative results due to the low statistical power of the meta-analysis when the required number of participants or trials has not been reached, which can be addressed by TSA. 7 Trials were included in chronological order and handled as interim analysis relative to the required information size, which is defined as the number of participants and events necessary to detect or reject an a priori assumed intervention effect in meta-analysis. Statistical techniques were used to adjust the CI of point estimate and to increase the threshold for statistical significance based on effect to be observed, incidence of outcome in control arm, information size, and heterogeneity. 8 It was performed on all outcomes without significant heterogeneity, from studies of low risk of bias. Z-score curve was generated by plotting cumulative Z scores with new study data. A Z-score curve crossing either of statistical significance boundaries (ie, the pair of outer oblique lines) implies that the statistically significant data is conclusive, whereas crossing either of the futility boundaries (ie, inner oblique lines) implies that the statistically insignificant data is conclusive. If the curve crosses the required information size boundary (ie, the vertical line), all observations are said to be conclusive. 8 All available statistical information (Fisher information) was used. The Z-score threshold was adjusted using the O'Brien-Fleming alpha-spending function. Studies reporting no events were handled by adding a constant (1) to both arms. Required information sizes were estimated from an RR reduction of 20%, chosen to represent a clinically meaningful effect. Incidences were calculated from all studies reporting the outcome of interest. Heterogeneity and variance adjustments were estimated from all included studies in TSA. A prespecified permissible 2-sided type 1 error (a) of 5% and type 2 error (b) of 20% were used, therefore giving a power of 80%.
All P values are 2-sided. The meta-analytical component was performed using Review Manager (RevMan), version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). The TSA component was performed using the Copenhagen trial unit, TSA software, version 0.9.5.10 Beta.

RESULTS
The literature search is summarized in a PRISMA diagram ( Figure 1       Critical appraisal of the included studies was performed using the modified Jadad scale or NOS, as summarized in Table E1 and Table E2, respectively. Overall, 18 of 35 RCTs scored 5 points or greater in modified Jadad scale, with all observational studies scoring 7 points or greater in NOS, and were classified as low risk of bias. Significant proportion of studies included before 2009 were classified as having high risk of bias, mainly due to inappropriate randomization methods and nonblinded studies.
Critical appraisal of meta-analysis by Fan and colleagues in 2010 3 was performed using the AMSTAR 2 tool, 6 as summarized in Online data supplement. It showed that the systemic review was of moderate quality.
Baseline characteristics of included patients in studies after previous meta-analysis were summarized in Table E3. Other related baseline characteristics (smoker, European System for Cardiac Operative Risk Evaluation, European System for Cardiac Operative Risk Evaluation II, dyslipidemia, peripheral vascular disease, chronic kidney disease, previous AF, chronic obstructive pulmonary disease, previous stroke, previous MI) were not reported, as they were included by fewer than 10 studies.
A pairing table (Table E4) was constructed to indicate outcomes reported by individual studies. All primary outcomes were supported by at least 15 studies (mortality 31, MI 32, LCOS 15, IABP use 20, stroke 17, new AF 17), whereas secondary outcomes were supported at least 7 studies (AKI 7, hospital LOS 9, ICU LOS 10).
There were no statistically significant differences in all outcomes (mortality, MI, LCOS, IABP use, stroke, new AF, AKI, hospital LOS, and ICU LOS) between warm and cold cardioplegia, with results summarized in Table 2. Forest plots of outcomes reported by most studies (ie, mortality and MI) were shown in Figures 2 and 3, respectively. None of the primary outcomes exhibited significant heterogeneity. Only hospital LOS and ICU LOS had significant heterogeneity.
A prespecified subgroup analysis was performed on primary outcomes for studies published since Fan and colleagues in 2010, 3 with results summarized in Table E5 and forest plots included in Figures 2 and 3 and Figures E1-E4. All of the primary outcomes remained statistically insignificant without significant heterogeneity.
Subgroup analysis were also performed on (1) randomized controlled studies, (2) studies of low risk of bias, (3) CABG surgeries, and (4) studies with cold blood versus those with cold crystalloid cardioplegia, with results summarized in Table 3, Table E6, Table E7, and Table E8, respectively. Most outcomes remained statistically insignificant, with heterogeneity qualitatively unchanged.  Sensitivity analysis was performed for all the outcomes using the leave-one-out method. Removal of individual studies from the analysis did not alter the statistical significance, except for the exclusion of Nardi and colleagues 9 in hospital LOS, which would result in statistically significant (P ¼ .04) shorter LOS in warm cardioplegia arm.
TSA was performed for all the outcomes without significant heterogeneity. The Z value is the test statistic and jZj ¼ 1.96 corresponds to a P ¼ .05, with greater Z values corresponding to lower P values. The Z-score curve for mortality (adjusted RR 1.0 [0.77-1.31], P ¼ .98; I 2 ¼ 0%; Figure 4      . Trial sequential analysis of (A) mortality, (B) myocardial infarction, (C) low cardiac output syndrome, and (D) atrial fibrillation. Z value is the test statistic and jZj ¼ 1.96 corresponds to a P ¼ .05. The required information size to detect or reject the 20% relative risk reduction found in random-effects model meta-analysis is calculated using diversity found in meta-analysis, with double-sided a ¼ 0.05 and b ¼ 0.20 (power of 80%).

DISCUSSION
In this study, we compared operative and clinical outcomes of warm and cold cardioplegia. No significant differences were found between both arms for all outcomes. TSA showed that current evidence was conclusive to rule out 20% RR reduction in the following outcomes: mortality, MI, LCOS, IABP use, stroke, new AF, and AKI ( Figure 6).
Overall, our results confirmed the findings by Fan and colleagues 3 that warm and cold cardioplegia were not significantly different in efficacy and safety, further providing a broader look at clinical and operative outcomes. Despite not exhibiting statistically significant subgroup differences, diverging trends were found upon subgroup analysis, suggestive of subtle differences between the subgroups. However, when analyzing only studies with low risk of bias, these numerical trends disappeared. This suggests that such trends might have been the result of bias, possibly due to unclear or inappropriate randomization methods and a lack of blinding in some trials. Furthermore, when we compared cold blood with cold crystalloid cardioplegia, the outcomes of mortality, LCOS, IABP use, stroke, and AKI showed trends in opposite directions, with subgroup differences of P ¼ .27, P ¼ .21, P ¼ .009, P ¼ .09, and P ¼ .08, respectively. The 2014 metaanalysis by Zeng and colleagues 10 suggested that subtle subgroup differences can lead to drastically different outcomes. In their study, there were significantly less postoperative MI in cold blood cardioplegia; however, there were no significant differences in mortality, AF, and stroke between cold blood versus crystalloid cardioplegia. This may have been the reason for the statistically insignificant trends in our subgroup analysis, as current studies may not have been designed to specifically compare cold blood versus cold crystalloid cardioplegia. Nonetheless, the results by

Required information size = 34
Cumulative Z-Score

Required information size = 34
Cumulative Z-Score

Required information size = 34
Cumulative Z-Score FIGURE 5. Trial sequential analysis of (A) intra-aortic balloon pump use, (B) stroke, and (C) acute kidney injury. Z value is the test statistic and jZj ¼ 1.96 corresponds to a P ¼ .05. The required information size to detect or reject the 20% relative risk reduction found in random-effects model meta-analysis is calculated using diversity found in meta-analysis, with double-sided a ¼ 0.05 and b ¼ 0.20 (power of 80%).
Zeng and colleagues 10 were limited by high risks of bias and other possible confounders; therefore, more studies are needed to evaluate the effects of blood versus crystalloid cardioplegia. All primary outcomes and AKI exhibited insignificant heterogeneity, whereas TSA showed conclusive results. While all primary outcomes were supported by at least 15 studies, suggesting uniformity of the included studies regarding the outcome of interest. In contrast, hospital LOS and ICU LOS displayed significant heterogeneity. This could be contributed by several factors, including differences in local practices, the admission and discharge criteria of ICU, experience of surgeons' etcetera. Heterogeneity remained high despite stratification by subgroups, suggesting that variability was less likely to be caused by differences in publication year, biased studies, type of surgery, or composition of cardioplegia solution. Although such significant heterogeneity limited the strength of our findings, our analysis represented the most up-to-date evidence. However, there is a need for further studies delineating factors affecting the aforementioned outcomes, along with trials controlling for the aforementioned factors.
A survey performed by Ali and colleagues 11  . Warm versus cold cardioplegia in cardiac surgery: a meta-analysis with trial sequential analysis. Forty-six studies, with 15,428 patients were included in analysis (35 randomized controlled trials þ 11 observational studies). No significant differences were found between two arms in post-operative mortality, myocardial infarction, low cardiac output syndrome, intra-aortic balloon pump use, stroke, and new atrial fibrillation as shown in the Forest plots. Trial sequential analysis of mortality was shown signifying current evidences were conclusive. In conclusion, choice of warm versus cold cardioplegia remains surgeon's preference.
The safety and efficacy of different types of crystalloid solutions (eg, histidine-tryptophan-ketoglutarate solution, St Thomas solution) warrants further investigations, as exemplified in the study by Pizano and colleagues. 12 Histidine-tryptophan-ketoglutarate solution is a widely used cardioplegic and organ-preserving solution; however, despite its widespread use, it is seldom studied in comparison with blood cardioplegia. Del Nido solution was initially intended for pediatric surgeries and is now extended to adult cardiac surgeries. Ler and colleagues, 13 in a meta-analysis performed in 2020, compared Del Nido versus St Thomas cardioplegic solution, showing similar postoperative outcomes.
Besides composition of cardioplegic solution, mode of administration should be further explored. Gambardella and colleagues 14 performed a meta-analysis in 2019 comparing single versus multidose cardioplegia, suggesting that more studies were needed to compare effects of different solution types, as current evidence were not yet conclusive. In addition, terminal hot-shot cardioplegia was proposed as a potential way to improve clinical outcomes, yet a systematic review performed by Volpi and colleagues 15 in 2019 concluded that there was insufficient evidence to evaluate its clinical merits. Mallidi and colleagues 16 conducted an observational study, suggesting that warm blood cardioplegia was associated with better long-term survival and less late MI than cold blood cardioplegia. However, there are few studies in the current literature that investigate the long-term effects of warm versus cold cardioplegia, which in fact are important concerns that should be further explored.
In fact, the choice of cardioplegic solutions might be based on other perceived benefits that were not investigated in this meta-analysis. At our unit, which consisted of 22 consultant cardiac and aortic surgeons, choice of cardioplegic temperature remained surgeons' preferences. The majority of our team members preferred cold cardioplegia, aiming for metabolic inhibition and thus less energy consumption. In the meta-analysis by Ler and colleagues, 13 the rate of defibrillation, aortic crossclamp time, and cardiopulmonary bypass time were listed as primary outcomes. This might suggest future directions of investigating for other outcomes.
This meta-analysis has a few limitations. First, one major confounder was the variation in means of administrating the cardioplegia solutions: intermittent versus continuous and antegrade versus retrograde. This meta-analysis was not designed to investigate these factors; thus, it was unclear the impact of these factors on our findings. Second, a significant proportion of the studies included that were published before 2009 had a high risk of bias; however, we attempted to mitigate this issue by performing a subgroup analysis of studies with low risk of bias. Third, the results of observational studies published after 2009 were pooled with RCTs, which may add on to confounders and selection bias. To address this issue, we performed a subgroup analysis for RCTs, which showed no changes of statistical significances of all outcomes. In addition to that, the observational studies included in this meta-analysis were of low risk of bias, and their value lies in their representation of real-world experience. Fourth, this study based on and updated the findings of previously meta-analysis by Fan and colleagues in 2010. 3 The results might therefore be inaccurate if it was of poor quality. Thus, a quality assessment was performed, showing that it was of moderate quality without major methodologic flaws. Fifth, the studies included in this meta-analysis spanned over 27 years, during which time surgical techniques and clinical practices have advanced significantly. To mitigate this issue, we performed a subgroup analysis of studies published after 2009. Nonetheless, this cutoff of publication time was merely based on the publication of the last meta-analysis on this topic, and the time of publication remains a potential confounder. Finally, only studies in English language were included; therefore, it is possible that relevant non-English studies were omitted.
In conclusion, this systematic review and meta-analysis concluded that there were no significant differences in postoperative rates of mortality, MI, LCOS, IABP use, stroke, new AF, and AKI, between the use of warm and cold cardioplegia. The choice of warm versus cold cardioplegia solution remains the surgeon's preference. Nonetheless, further studies should evaluate any differences between various compositions and modes of administrating cardioplegic solutions, with greater exploration on the long-term effects of warm versus cold cardioplegia.

Conflict of Interest Statement
The authors reported no conflicts of interest. The funding was provided by University College London Open Access Team, Main Library; University College London, UK.
The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.
We hereby thank the following individuals for their assistance and effort in this project: Shaik Ashraf Bin Shaik Ismail (University of Liverpool Medical School) for title and abstract screening, full-text screening, data extraction; Wing Yan Elizabeth Wong (Brighton and Sussex Medical School, University of Sussex) for data extraction; and Chun Ming Chiu (Brighton and Sussex Medical School, University of Sussex) for data extraction.            CB, Cold blood; CC, cold crystalloid; RR, relative risk; WMD, weighted mean difference; CI, confidence interval; MI, myocardial infarction; LCOS, low cardiac output syndrome; IABP, intra-aortic balloon pump; AF, atrial fibrillation; AKI, acute kidney injury; LOS, length of stay; ICU, intensive care unit.