Add-On Effect of Chinese Herbal Medicine on Mortality in Myocardial Infarction: Systematic Review and Meta-Analysis of Randomized Controlled Trials

In China, Chinese herbal medicine (CHM) is widely used as an adjunct to biomedicine (BM) in treating myocardial infarction (MI). This meta-analysis of RCTs evaluated the efficacy of combined CHM-BM in the treatment of MI, compared to BM alone. Sixty-five RCTs (12,022 patients) of moderate quality were identified. 6,036 patients were given CHM plus BM, and 5,986 patients used BM only. Combined results showed clear additional effect of CHM-BM treatment in reducing all-cause mortality (relative risk reduction (RRR) = 37%, 95% CI = 28%–45%, I 2 = 0.0%) and mortality of cardiac origin (RRR = 39%, 95% CI = 22%–52%, I 2 = 22.8). Benefits remained after random-effect trim and fill adjustment for publication bias (adjusted RRR for all-cause mortality = 29%, 95% CI = 16%–40%; adjusted RRR for cardiac death = 32%, 95% CI = 15%–46%). CHM is also found to be efficacious in lowering the risk of fatal and nonfatal cardiogenic shock, cardiac arrhythmia, myocardial reinfarction, heart failure, angina, and occurrence of total heart events. In conclusion, addition of CHM is very likely to be able to improve survival of MI patients who are already receiving BM. Further confirmatory evaluation via large blinded randomized trials is warranted.


Myocardial Infarction: Disease Burden and erapeutic
Options. Incoronary artery disease, a critical reduction of the blood supply to the heart may result in myocardial infarction (MI), a phenomenon owing to the formation of an area of necrosis in heart muscles caused by inadequate supply of blood to the muscles, usually as a result of occlusion of a coronary artery. About a quarter of MI patients will die from it due to complications including cardiogenic shock, cardiac perforation, embolism, heart failure, papillary muscle rapture, rhythm disturbances, or autoimmune pericarditis. Current evidence on biomedicine (BM) treatment suggests that aspirin, thrombolytics with or without adding lowmolecular-weight heparin, beta-blockers, ACE inhibitors, and nitrates are bene�cial for improving outcomes in people with MI. Invasive procedures including coronary artery bypass graing (CABG) and percutaneous transluminal coronary angioplasty (PTCA, balloon angioplasty) were also found to be useful. However, their efficacy in preventing death is not without limitations. For instance, beta-blockers have no short-term effect on mortality, and they may increase the risk of cardiogenic shock. rombolytics may cause stroke and major bleeding while reducing mortality, and those who are treated will receive no additional bene�ts from nitrates [1].
Despite these therapeutic advances, coronary artery disease remained to be the foremost leading cause of death in both low-and middle income countries as well as high-income countries, contributed 11.8% and 17.3% of total deaths, respectively [2]. Researchers are evaluating the potential bene�ts and harms of add-on treatments like 2 Evidence-Based Complementary and Alternative Medicine vasodilators and positive inotropes on mortality [3]. Chinese herbal medicine (CHM) is another novel candidate as an add-on treatment. Infarction. In China, CHM is widely prescribed in both outpatient and inpatient settings [4]. Amongst community health clinics, 75% provide both BM and traditional Chinese medicine (TCM) treatments. TCM hospitals comprised 13.8% of all hospitals, and 90% of the BM hospitals are annexed with TCM departments [5]. Given the omnipresence of TCM services within the Chinese healthcare system, it is not uncommon for clinicians to prescribe CHM as an adjunct to BM treatment in the management of potentially lifethreatening conditions including MI [6]. One of the most researched single herbs is Radix Astragali, which exerts its therapeutic effectiveness by inhibiting cardiac �brosis, reducing infarct size, and increasing capillary and arteriole densities [7]. Commonly used Chinese proprietary medicines include Shexiangbaoxin tablets and Tongxinluo capsules. Shexiangbaoxin tablets are found to slow MI pathogenesis by inhibiting hypertrophy related metabolites [8]. On the other hand, Tongxinluo capsules act by promoting local blood supply and thus limit infarct size [9]. CHM injections based on sheng mai san are also widely prescribed. It reduces infarct size via the activation of protein kinase C, opening of the mitochondrial KATP channels, and lowering the concentration of 5-hydroxytryptamine, norepinephrine, methionine-enkephalin, and leucine-enkephalin [10,11].

Synthesizing Chinese Herbal Medicine Trials:
Focusing on Objective Outcomes. e average effect of these CHM formulae as an adjunct to BM could be estimated using random effect meta-analyses of randomized controlled trials (RCTs) [12]. One of the major caveats in conducting systematic reviews on CHM is that existing RCTs are oen prone to high risks of bias, thus limiting their usefulness in elucidating treatment effectiveness [13]. However, results from a recent metaepidemiological study have provided an alternative perspective on this issue. It is suggested that objective outcomes are less susceptible to bias associated with inadequate allocation concealment and blinding [14,15]. Accordingly, by focusing on objective outcomes like mortality, we may partially overcome limitations imposed by the relatively high risk of bias amongst CHM trials.

Aim of is Paper.
Taking into account the methodological considerations above, we performed a systematic review and meta-analysis of RCTs on the efficacy and safety of CHM for MI as an add-on to BM treatment, with a focus on objective critical outcomes including death, recurrent myocardial infarction, and other post-MI cardiac consequences.

Criteria for Considering Studies for is Paper.
We included RCTs comparing the efficacy and safety of CHM plus BM versus BM alone. CHM is de�ned as any preparation containing at least one herb or its extraction referenced in the 2010 Chinese Pharmacopeia [16]. We included RCTs which enrolled adult MI patients regardless of gender, age, ethnicity, or comorbidities. We focused on the primary outcomes of (i) mortality of cardiac origin and (ii) all-cause morality. We also consider the following as secondary outcomes: (i) recurrence of MI and (ii) other nonfatal, post-MI cardiac outcomes including cardiac arrhythmia, heart failure, cardiac rupture, cardiogenic shock, and angina. Adverse events reported by authors were also summarized. We imposed no restrictions on language and publication status.

Selection of Studies, Data Extraction, and Risk of
Bias Assessment. Two reviewers (Y. Qin and C. Mao) independently screened the titles and abstracts to assess their eligibility. Full texts of potentially eligible citations were retrieved for detailed examination. Selection discrepancies were settled through discussions between these two authors. e remaining disagreements were resolved by consulting another author (J. L. Tang). For included RCTs, comprehensive information on patients, CHM interventions, and baseline and control treatments, as well as outcomes, was extracted. Risks of bias amongst included RCTs were evaluated by the Cochrane collaboration's risk of bias assessment tool [17]. e assessment composed of a description and a judgement for each entry in a risk of bias table, including (i) sequence generation, (ii) allocation sequence concealment, (iii) incomplete outcome data, (iv) selective outcome reporting, and (v) other potential sources of bias. Blinding was assessed for the primary outcome of all-cause morality.

Data Analysis.
Analyses were conducted using Stata 11 and R soware. Dichotomous efficacy outcomes were expressed as relative risk reduction (RRR) and relative risk (RR), while RR was used for adverse events. 95% con�dence intervals (CIs) were calculated for all estimates. We performed random-effect meta-analysis separately for each outcome. For primary outcomes of all-cause mortality and cardiac death, funnel plots were drawn for assessing publication bias. In case of asymmetry, random trim and �ll analysis were performed as a sensitivity analysis [18]. Tests for heterogeneity were performed with chi-squared testes, at a signi�cance level of . 2 statistic was calculated to estimate variation across studies. We regarded 2 < 25% as an indicator of low heterogeneity level, 25-50% as moderate level, and higher than 50% as high level [19]. Heterogeneity was explored with random-effect metaregression using baseline risk, mean age, route of drug administration (oral versus intravenous), and treatment duration as covariates, taking into account the sample size requirement of including not more than 1 covariate for every 10 studies [20]. We expected that higher baseline risk and mean age could be associated with a smaller effect [1], while intravenous administration and longer treatment duration could be associated with a larger effect. Figure 1, our search in electronic bibliographical databases yielded 12,666 citations aer removal of duplications, of which 2,660 were classi�ed as potentially relevant and were subjected to a full-text assessment. A total of 65 RCTs published in 63 articles met the inclusion criteria. Details of these studies are presented in Table 1.

Study Characteristics.
A total of 6,036 patients were enrolled in the CHM plus BM group, and 5,986 patients were allocated to the BM only group. e average size of the trials was 185 participants (ranging from 28 to 2735 participants per trial). Fiy trials reported treatment duration and the average duration was 68.9 days, ranging from 3 to 1440 days. Forty-nine trials reported the length of followup. e average follow-up length was 7.1 months, ranging from 0.1 to 84 months.
For diagnostic criteria, 36 (55.4%) studies applied the 1979 World Health Organization criteria, which enrolled patients with at least two of the following three presentations: chest pain or discomfort, an elevation in CK-MB levels, or an ECG with signi�cant ST-segment elevations [84]. Four adopted criteria from the Chinese Society of Cardiology [85] and one used criteria from the European Society of Cardiology [86]. Twelve applied author-de�ned diagnostic criteria, and the remaining 12 did not report criteria used. irty-one standardized Chinese herbal formulae were examined in 63 (96.9%) of the 65 included studies, while the other two studies used an individualized approach. 32 (50.0%) preparations were administrated orally, 30 (46.9%) were prescribed as herbal injections, and 2 (3.12%) trials used both intravenous and oral treatments. Eight formulae were evaluated by three or more trials. In total, these formulae were assessed in 38 studies, constituting 58.5% of all included trials.

Risk of Bias.
Among these 65 RCTs, only 7 were at low risk for bias for allocation sequence generation. Twelve were at high risk and the remaining RCTs did not report their sequence generation procedure clearly. All but one had high risk of bias in terms of allocation concealment and none of the included studies report the use of blinding. However, we regarded the risks of bias associated with lack of blinding and allocation concealment to be minimal, as the primary outcomes were of objective nature. Two of the included studies had high risk of bias for incomplete data and one for selective outcome reporting. Six are at high risk of bias due to other reasons. In summary, we consider the overall risk of bias amongst our included studies to be moderate ( Figure  2). e detailed risk of bias assessment results is presented in Appendix 2 in the supplementary materials. Pooled results demonstrated superiority of combined treatment in preventing all-cause mortality (RRR = 37%, 95% CI = 28%-45%). Funnel plot indicates the presence of publication bias. A�er applying trim and �ll procedure (Figure 3), the RRR remained to be signi�cant (RRR = 29%, 95% CI = 16%-40%, Table 2). Ten RCTs reported death of cardiac origin, and pooled �ndings also favor combined treatment (RRR = 39%, 95% CI = 22%-52%). Funnel plot indicates the presence of publication bias. A�er applying trim and �ll procedure, the RRR remained to be signi�cant (RRR = 32%, 95% CI = 15%-46%).

Metaregression.
We explored these heterogeneities by performing multivariate metaregression analyses using mean age, treatment duration, route of administration (oral versus intravenous), and baseline risk as covariates. None of the four covariates is signi�cantly associated with cardiac arrhythmia (for baseline risk regression coefficient ( ) = 0.46, = .4 ; for mean age = . , = .96; for duration of treatment = . , = .9 ; for route of administration = .2 , = .63), or heart failure (for baseline risk = . 67

Discussion
is systematic review on the add-on effect of CHM on BM in the treatment of MI summarized �ndings from 12,022 patients reported in 65 RCTs. e overall risk of bias amongst included studies was moderate. Despite the lack of allocation concealment and blinding in the majority of included trials, its impact on risk of bias was less critical as we focused on objective outcomes. Random-effect meta-analyses demonstrated that combined treatment is superior to BM alone in reducing the risk of all-cause mortality and death of cardiac origin. Funnel plots indicated the presence of publication bias for both outcomes, and trim and �ll procedures were conducted as sensitivity analyses. e directions of effect did not change aer the adjustment, and the 95% CI of the estimates overlapped with the unadjusted values. e lower 95% CI boundary of the trim-and �ll-adjusted RRR for allcause and cardiac mortality was 16% and 15%, respectively. Conservatively speaking, CHM appeared to offer a protective add-on effect against mortality aer adjusting for the publication bias, a common problem amongst the clinical research literature on CHM [87]. Combined treatment is also found to be more effective than BM alone in lowering the risk of fatal cardiogenic shock. Our analyses did not demonstrate therapeutic bene�ts of combined treatment on other reviewed fatal outcomes including myocardial reinfarction, cardiac arrhythmia, heart failure, and sudden cardiac death. For nonfatal outcomes, our analyses demonstrated that CHM is an effective add-on for lowering the risk of cardiogenic shock, cardiac arrhythmia, myocardial reinfarction, and the occurrence of total heart events. Bene�ts in preventing heart failure and angina were also observed but these �ndings are less robust given the subjective nature of the outcome, and metaregression did not shed light on potential sources of heterogeneity. We have considered including allocation concealment and blinding as covariates in our metaregressions but numbers of trials with low risk in these domains are too small for conducting such analysis. e effect of combined treatment on these two outcomes would need to be further evaluated with methodologically stronger trials. In addition, more comprehensive reporting on BM treatment details and adverse events is expected in future studies, preferably with reference to the CONSORT statement.
Comprehensiveness of search is the major strength of this systematic review. e use of both international and Chinese databases allowed us to locate a much higher number studies compared to seven existing reviews on the topic [88]. We also attempted to synthesize results from trials evaluating heterogeneous CHM using random-effect model. is allowed us to estimate the average effect of adding CHM on top of conventional therapies [12]. e use of the trim and �ll method has also partly circumvented the problem of publication bias. Nevertheless, the robustness of our conclusion depends on the assumption that the objective nature of outcomes was less affected by two major sources of bias: allocation concealment and blinding. While this assumption is tested in metaepidemiological studies [89,90], the generalizability of these �ndings warrants further investigations.

Conclusion
Based on RCTs of moderate quality, this systematic review demonstrated consistent, add-on bene�ts of using CHM on top in BM treatment for preventing all-cause and cardiac mortality amongst MI patients. ese �ndings are in line with the results from seven existing systematic reviews of smaller scope and lower methodological quality. is tentative conclusion warrants further scrutiny using rigorously designed RCT, and a more comprehensive approach in reporting BM treatment details and adverse events is warranted.