Effects of Preoperative HbA1c Levels on the Postoperative Outcomes of Coronary Artery Disease Surgical Treatment in Patients with Diabetes Mellitus and Nondiabetic Patients: A Systematic Review and Meta-Analysis

Aims To investigate the effect of preoperative HbA1c levels on the postoperative outcomes of coronary artery disease surgery in diabetic and nondiabetic patients. Methods and Results The MEDLINE (via PubMed), Cochrane Library, Web of Science, Embase, Wanfang Data, China National Knowledge Infrastructure (CNKI), and Chinese Biology Medicine (CBM) databases were used to search the effects of different preoperative HbA1c levels on the postoperative outcomes of coronary artery disease surgical treatment in diabetic and nondiabetic patients from inception to December 2018. Two review authors worked in an independent and duplicate manner to select eligible studies, extract data, and assess the risk of bias of the included studies. We used a meta-analysis to synthesize data and analyze subgroups, sensitivity, and publication bias as well as the GRADE methodology if appropriate. The literature search retrieved 886 records initially, and 23 cohort studies were included in the meta-analysis. In this meta-analysis, we found that there was a reduced incidence of surgical site infections (OR = 2.94, 95% CI 2.18-3.98), renal failure events (OR = 1.63, 95% CI 1.13-2.33), and myocardial infarction events (OR = 1.69, 95% CI 1.16-2.47), as well as a shortened hospital stay (MD = 1.08, 95% CI 0.46-1.71), in diabetic patients after coronary artery disease surgical treatment with lower preoperative HbA1c levels. For nondiabetic patients, a higher preoperative HbA1c level resulted in an increase in the incidence of mortality (OR = 2.23, 95% CI 1.01-4.90) and renal failure (OR = 2.33, 95% CI 1.32-4.12). No significant difference was found between higher and lower preoperative HbA1c levels in the incidence of mortality (OR = 1.06, 95% CI 0.88-1.26), stroke (OR = 1.49, 95% CI 0.94-2.37), or atrial fibrillation (OR = 0.94, 95% CI 0.67-1.33); the length of ICU stay (MD = 0.20, 95% CI -0.14-0.55); or sepsis incidence (OR = 2.49, 95% CI 0.99-6.25) for diabetic patients or for myocardial infarction events (OR = 1.32, 95% CI 0.27-6.31) or atrial fibrillation events (OR = 0.99, 95% CI 0.74-1.33) for nondiabetic patients. The certainty of evidence was judged to be moderate or low. Conclusion This meta-analysis showed that higher preoperative HbA1c levels may potentially increase the risk of surgical site infections, renal failure, and myocardial infarction and reduce the length of hospital stay in diabetic subjects after coronary artery disease surgical treatment and increase the risk of mortality and renal failure in nondiabetic patients. However, there was great inconsistency in defining higher preoperative HbA1c levels in the studies included; we still need high-quality RCTs with a sufficiently large sample size to further investigate this issue in the future. This trial is registered with CRD42019121531.


Introduction
Diabetes mellitus (DM) is a chronic disease with dysfunction in insulin secretion that gradually affects critical organs and tissues, including the heart, nerves, blood vessels, and eyes, and increases the risk of their infection [1]. The International Diabetes Federation reported that 415 million people, i.e., 8.8% of the world population, suffered from DM in 2015 [2]. Among them, 75% were from lowand middle-income countries. China, with 109.6 million people suffering from diabetes, ranked number one in the world.
With a sharply rising trend, the number of global diabetic patients and the global prevalence rate are estimated to reach 642 million and 10.4%, respectively, in 2040. At that time, the number of diabetic patients in China will reach 150.7 million. In 2015, DM resulted in approximately five million deaths in the world. Spending on the treatment of DM was estimated to be between 673 and 1197 billon US dollars, which was approximately 11.6% of the total world spending on health. As one of the most important noninfectious diseases, diabetes brings about a heavy burden on patients and their families, societies, and countries in terms of health and economic outcomes.
With the increase in diabetes prevalence, an increasing number of diabetic patients will have surgical operations. Statistics show that approximately half of DM patients have the chance to receive at least one surgery in life [3]. Current evidence indicates that perioperative blood glucose abnormalities (including hyperglycemia, hypoglycemia, and blood glucose fluctuation) could increase the rate of death and the incidence of complications such as infections, poor wound healing, and cardio/cerebrovascular events; prolong the duration of hospitalization; and affect long-term prognosis [4,5]. For diabetic patients undergoing cardiac surgery, hyperglycemia could increase mortality, cause adverse events in the kidney and lung, increase the incidence of atrial fibrillation, and even lead to a serious threat to life [6][7][8]. Therefore, diabetic patients who undergo surgical treatments, especially those undergoing heart surgery, need strict management of blood glucose.
Glycosylated hemoglobin A1c (HbA1c) is usually regarded as a marker for the average blood glucose level over the three months before the measurement and is used for assessing glycemic control. Currently, many guidelines have provided recommendations on the preoperative HbA1c level for achieving better surgical outcomes, but the levels vary significantly. Some guidelines recommended that the HbA1c level should be below 7% or 6.5% [9,10], but others recommended a level at 8.5% or 8%-9% [11,12]. A systematic review [13] indicated that elevated preoperative HbA1c levels did not increase the morbidity and mortality of operations (including cardiac surgery) in diabetic patients; however, it did not mention the high threshold of preoperative HbA1c explicitly, which is not suitable to support the formulation of a recommendation. For nondiabetic patients, the relationship between outcomes after cardiac surgery and preoperative HbA1c level is still unclear.
Our study focused on the impact of preoperative HbA1c levels on the postoperative outcomes of coronary artery disease surgical treatment in diabetic and nondiabetic patients, and we conducted a meta-analysis by retrieving and reviewing systematically relevant research evidence. In addition, this study will provide guidance for clinical practice by objectively presenting the results with the GRADE system.

Search Methods.
A number of database resources, including MEDLINE (via PubMed), Embase, the Cochrane Library, Web of Science, Wanfang, CNKI, and CBM, were searched from the inception of each resource to December 12, 2018. The WHO International Clinical Trials Registry Platform (ICPTR) and ClinicalTrials.gov were searched as supplements without limits to the publication types. The following keywords were used: diabetes, cardiac surgery, and preoperative. We also searched references from original articles, clinical guidelines, narrative reviews, and previous systematic reviews/meta-analyses to identify additional studies. We followed the PRISMA guidelines for conducting and reporting meta-analyses of RCTs (Supplementary Annex 1) [14]. Annex 2 detailed the search strategy.

Eligibility Criteria
2.2.1. Population. Diabetic patients who met the diagnostic criteria for DM of the American Diabetes Association [15] and who underwent elective cardiac surgery were eligible for inclusion in the study. We mainly focused on coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI) and ignored the other risk factors that most likely affect the outcomes of elective cardiac surgery, such as cardiovascular risk factors and medical history. We also referred to the indications for adult cardiac surgery created by Young [16]. In addition, we included nondiabetic patients who also underwent an elective cardiac surgery to compare with the diabetic patients.

Outcomes.
We analyzed perioperative health-related outcomes. The primary outcome was mortality. The secondary outcomes were hospital stay, the length of ICU (intensive care unit) stay, cardiovascular events (such as atrial fibrillation and stroke), and other adverse events (such as renal failure and sepsis).

Types of Studies.
Prospective randomized controlled trials (RCTs), cohort studies, and retrospective studies were included in this meta-analysis.
2.2.5. Selection of Studies. Two reviewers independently screened titles, abstracts, and the full texts of the identified 2 Journal of Diabetes Research studies. There was a pretest and discussion before the formal screening of the literature to ensure the consistency between reviewers. Two reviewers solved any disagreement by discussion or consultation with a third researcher.
2.2.6. Data Extraction. Data extraction was undertaken independently by two reviewers using standard data extraction templates with the following information: basic information (publication year, first author, institution, and journal) and contents of studies (study design, sample size, research objective, the characteristics of population, preoperative HbA1c levels, and health-related clinical outcomes). A pretest was performed before the formal extraction to ensure that each reviewer was consistent in their understanding of the criteria and process of extraction. Any disagreement was solved by discussion or consultation with a third reviewer. If any important data in the original study provided inadequate information, attempts would be made to acquire the necessary information by contacting the authors, and if we could not obtain a reply within one week, we excluded the study.

Assessment of the Risk of Bias in the Included Studies.
Two reviewers used the Cochrane bias risk assessment tools [17] to assess the included RCTs and used the Newcastle Ottawa Scale (NOS) [18] for cohort studies and retrospective studies. For the Cochrane bias risk assessment tools, the criteria we used included random sequence generation, allocation concealment, the blinding of participants and personnel, the blinding of outcome assessors, incomplete outcome data, selective reporting, and other biases. A judgment of "unclear" indicated an unclear or unknown risk of bias. We judged individual bias items individually for each study as described in the Cochrane Handbook for Systematic Reviews of Interventions. If a cohort study or retrospective study had an NOS score lower than seven, it indicated that there was a serious bias in this study. A pretest was performed before the formal assessment to ensure that each reviewer was consistent in their understanding of the criteria and process of evaluation. Any disagreement was solved by discussion or consultation with a third researcher.

Statistical Analysis.
The random-effect model was used for all analyses. Mean differences and 95% confidence intervals (CIs) were calculated for continuous outcomes. For dichotomous data, odds ratios and 95% CIs were calculated instead. Heterogeneity was identified by using the Q test (P < 0:05, suggesting the existence of heterogeneity). Heterogeneity was also specifically examined by employing the I 2 statistic. Inconsistency across studies was determined to assess the impact of heterogeneity on the meta-analysis. An I 2 statistic of 50% or more indicated a considerable level of inconsistency. The data were summarized statistically provided they were available, sufficiently similar, and of sufficient quality [19]. Statistical analyses were performed according to the guidelines referenced in the 2011 version of the Cochrane Handbook using RevMan 5.3 software. Sensitivity analysis was used to investigate the source of heterogeneity using Stata 14.0 software, provided that obvious or significant heterogeneity was observed. In addition, publication bias was performed if the included studies were more than nine.  [20][21][22] was used to assess the quality of evidence for each outcome. The criteria mainly considered included the risk of bias, indirectness, inconsistency, imprecision, and publication bias. The quality of evidence for each outcome was graded as high, moderate, low, or very low. Finally, we presented the results of the quality of evidence for each outcome in a table that summarizes the findings.

3.1.
Results of the Search. The literature search retrieved 886 records preliminarily, and 701 studies were excluded by screening the title and abstract. After the removal of 46 studies for not meeting the inclusion criteria, 23 studies were eligible for inclusion . Figure 1 shows the flowchart of study selection.

Baseline Characteristics of Included Studies.
Twentythree studies were included. Table 1 details the baseline characteristics of the included studies.

Risk of Bias.
Most of the included studies did not report on the independent blind assessment of outcomes. In terms of follow-up, 9 studies [23,24,26,28,31,36,37,42,43] did not report follow-up time and one study [41] had a 30-day follow-up, which was relatively short for outcomes and had high bias. Other studies were at low risk of bias in other items (see Table 2).

Myocardial Infarction.
Nine studies reported on the incidence of myocardial infarction [24, 28-31, 34, 39, 40, 42]. The meta-analysis results showed that the incidence of myocardial infarction in diabetic patients after cardiac surgery was lower in the group with a lower preoperative HbA1c level than in the group with a higher preoperative HbA1c level (OR = 1:69, 95% CI 1.16-2.47, P = 0:47, and I 2 = 0%), as shown in Annex 3-2.

Atrial Fibrillation.
Five studies reported on the incidence of atrial fibrillation [23,28,35,39,43]. The results of the meta-analysis showed that there was no significant Studies included in quantitative synthesis (meta-analysis) (n = 23)  5 Journal of Diabetes Research difference in the incidence of atrial fibrillation between diabetic patients with lower preoperative HbA1c levels and those with higher preoperative HbA1c levels after cardiac surgery (OR = 0:94, 95% CI 0.67-1.33, P = 0:06, and I 2 = 56%), as shown in Annex 3-4.
3.11. Length of ICU Stay. Four studies reported on the length of time in the intensive care unit (ICU days) [23,26,31,41]. The results of the meta-analysis showed that there was no significant difference in ICU days between diabetic patients with lower preoperative HbA1c levels and those with higher HbA1c levels after cardiac surgery (MD = 0:20, 95% CI -0.14-0.55, P = 0:40, and I 2 = 0%), as shown in Annex 3-5.
3.12. Sepsis. Four studies reported on the incidence of sepsis [26,31,40,41]. The results of the meta-analysis showed that there was no significant difference in the incidence of sepsis between diabetic patients with lower preoperative HbA1c levels and those with higher HbA1c levels after cardiac surgery (OR = 2:49, 95% CI 0.99-6.25, P = 0:77, and I 2 = 0%), as shown in Annex 3-6.
3.14. Subgroup and Sensitivity Analysis. Based on different types of surgery (CABG and PCI), we performed subgroup analyses showing that mortality among those with different preoperative HbA1c levels did not reduce or increase in diabetic patients undergoing CABG (OR = 1:09, 95% CI 0.84-1.43, P = 0:07, and I 2 = 40%) or PCI (OR = 1:21, 95% CI 0.65-2.24, P = 0:55, and I 2 = 75%), as shown in Figure 5. Furthermore, we performed sensitivity analyses with the use of a metaninf command from Stata software for mortality in diabetic patients, and the results were not significantly different from those of the primary analysis, as shown in Figure 6.

Publication
Bias. Publication bias was tested by a visual examination of the funnel plots, which were symmetrical and showed no evidence of publication bias for mortality outcome, as shown in Figure 7. 3.16. Grading of Quality of Evidence. Table 4 details the quality of evidence for nine outcomes in the systematic review. ① Representativeness of the exposed cohort. ② Selection of the nonexposed cohort. ③ Ascertainment of exposure. ④ Demonstration that the outcome of interest was not present at the start of the study. ⑤ Comparability of cohorts on the basis of the design or analysis. ⑥ Assessment of outcome. ⑦ Follow-up was long enough for outcomes to occur. ⑧ Adequacy of follow-up of cohorts. 6 Journal of Diabetes Research

Discussion
This systematic review compared the association between the preoperative HbA1c levels and health outcomes of diabetic and nondiabetic patients who underwent cardiac surgery. The results showed that there were fewer surgical site infection events, fewer renal failure events, fewer myocardial infarction events, and shorter hospital stay times after cardiac surgery in diabetic patients with lower preoperative HbA1c levels (≤6.5% or 7%), and the quality of evidence was low to moderate. However, no significant difference was found between higher and lower preoperative HbA1c levels in the incidence of mortality (the quality of evidence was moderate) or stroke (the quality of evidence was moderate), the length of ICU stay (the quality of evidence was moderate), or the incidence of sepsis (the quality of evidence was low) or atrial fibrillation (the quality of the evidence was low) for diabetic patients. For nondiabetic patients who underwent cardiac surgery, we found a high risk of mortality and renal failure for patients with a higher preoperative HbA1c level, but in terms of myocardial infarction and atrial fibrillation, there was no significant difference between the patients with higher and lower preoperative HbA1c levels. Limited by the number of included studies, we did not perform a meta-analysis on other outcomes. A current systematic review showed that perioperative hyperglycemia was not beneficial for patients undergoing cardiac surgery [7]. However, practitioners mainly focused on the relationship between preoperative blood glucose levels and the outcomes of cardiac surgery in patients with diabetes. A systematic review published in 2015 indicated that it was not necessary to measure preoperative blood glucose and HbA1c if the patients who underwent noncardiac surgery were without obvious clinical symptoms or signs of high risk [46]. Another systematic review [13] in the same year also obtained similar results. In addition, many systematic reviews that addressed the association between the preoperative HbA1c level and the outcomes of cardiac surgery in patients with diabetes have been published in the last [13,[47][48][49] years that showed that an elevated preoperative HbA1c level did not increase the postoperative morbidity and mortality in diabetic patients who underwent cardiac surgery. The results of our review were consistent with those of current systematic reviews, namely, the preoperative HbA1c level (we regarded 6.5% and 7% as the lower and higher boundaries, respectively) had no statistically significant effect on the incidence of mortality in diabetic patients     Journal of Diabetes Research after cardiac surgery. However, a systematic review published in 2017 [50] showed contrasting results, mainly because the authors did not include the PCI studies and nondiabetic patients. Our results might be influenced by the different definitions of higher and lower HbA1c levels in the included original studies. For nondiabetic patients, the results of this systematic review were consistent with those of a previous systematic review [51].
For the incidence rates of other complications, our systematic review found that a lower preoperative HbA1c level could reduce the incidence of surgical site infection, renal failure, and myocardial infarction, reducing the hospital stay in patients with diabetes mellitus after cardiac surgery. Several possible mechanisms may explain the results. HbA1c is an indicator of long-term (3-to 4-month) glycemic control and is formed when glucose in the blood binds irreversibly to hemoglobin to form a stable glycated hemoglobin complex. HbA1c is not affected by acute changes in blood glucose levels [52]. In patients with type 2 diabetes, HbA1c > 7% can induce adverse metabolic memory. The higher the HbA1c, the higher the risk of macrovascular complications, microvascular complications, and death in diabetic patients [53]. HbA1c induces dyslipidemia, hyperhomocysteinemia, and hypertension and increases C-reactive protein, oxidative stress, and blood viscosity, which would contribute to the development of cardiovascular diseases [54]. Surgery, stress, and anesthesia can exacerbate oxidative stress and increase blood viscosity. For diabetic patients, the situation worsens [55]. In the case of stress and blood loss, to protect the blood supply of important organs, visceral large blood vessels contract. The kidney is one of the most sensitive organs for blood volume. When the blood volume is insufficient, kidney  9 Journal of Diabetes Research failure can occur at the earliest. Patients with long-term hyperglycemia due to osmotic diuresis may experience a relative lack of blood volume. Insufficient blood volume during the perioperative period is more obvious and may be one of the important causes of renal failure [56]. Increased blood viscosity is one of the important factors leading to the occurrence of large blood clots. The high blood viscosity of diabetic patients can induce myocardial infarction [57]. Decreased wound healing ability, weakened inflammatory cell chemotaxis, and decreased immune function in diabetic patients are the main causes of postoperative infection. High HbA1c can lead to lower wound healing, weakened chemokine chemotaxis, and decreased immune function in diabetic patients, which can increase infection [58]. For these reasons, high HbA1c levels will prolong the hospital stay in patients with diabetes.
This systematic review had the following strengths: (1) To our knowledge, this was the first designed systematic review to investigate the effect of the preoperative HbA1c level on the postoperative outcomes in diabetic and nondiabetic patients who underwent cardiac surgery; the current systematic review mainly focused on coronary artery disease or percutaneous coronary interventions. (2) Our study is registered in PROSPERO, which could improve the overall reporting and methodological quality [59]. (3) We conducted a quality assessment using the GRADE system for each outcome compared with a previous systematic review. The following limitations should be taken into consideration: (1) We included cohort studies that had a potential high risk of bias. (2) The definitions of higher preoperative HbA1c levels were different in the original studies and included multiple values, such as 6.5%, 7%, 7.5%, and 8%. Although we grouped them for analysis, the grouping might not be very accurate and may have caused heterogeneity. (3) The follow-up time for postoperative mortality varied greatly (ranging from before discharge to 7 years after surgery). (4) The postoperative glycemic index is probably a confounding factor that affects the correlation of events after cardiac surgery with HbA1c levels.

Conclusions
The results of this systematic review showed that patients with diabetes mellitus with lower preoperative HbA1c levels showed a reduced incidence of surgical site infection, renal failure, and myocardial infarction, as well as a reduced length of hospital stay, after cardiac surgery compared with those with higher HbA1c levels, and the quality of evidence was

10
Journal of Diabetes Research low to moderate. However, the higher level of preoperative HbA1c had no effect on the incidence of mortality or other adverse events. Our results were based on current available evidence, and the conclusions can provide guidance for clinical practice to some extent. However, there was great inconsistency in the definition of higher preoperative HbA1c levels in the included studies, and we still need high-quality RCTs with large sample sizes to further investigate this issue in the future.

Conflicts of Interest
The authors declare that they have no competing interests.

Authors' Contributions
Jinjing Wang and Xufei Luo contributed to the study concept and design, search of the literatures, data extraction, data analyses, and drafting and review of the final manuscript. Kang Chen, Meng Lv, and Xueqiong Li contributed to the conception and design of the analysis and interpreted the analyzed data. Yiming Mu, Kang Chen, and Yaolong

Low
High: we are very confident that the true effect lies close to that of the estimate of the effect; moderate: we are moderately confident in the effect estimate-the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different; low: our confidence in the effect estimate is limited-the true effect may be substantially different from the estimate of the effect; very low: we have very little confidence in the effect estimate-the true effect is likely to be substantially different from the estimate of effect. a The poor quality of included studies in the independent blind assessment of outcomes and inadequate follow-up time. b Serious inconsistency for the scattered 95% CI. c Wide confidence intervals, serious imprecision.