Comparative efficacy of Chinese herbal injections for treating chronic heart failure: a network meta-analysis

A study was considered eligible if it suited for these criteria: 1) randomized controlled trial (RCT); 2) patients enrolled were diagnosed as CHF according to “Guidelines on the Diagnosis and Treatment of Heart Failure” conducted by The Chinese medical association cardiovascular epidemiology branch in 2014 [9] or “Clinical Guideline of New Drugs for Traditional Chinese Medicine” released by CFDA in 2002 [10]. Both of them contained both western diagnostics, the latter included TCM diagnostics as well; 3) patients receive WM treatment (e.g. cardiotonic, diuretic, ACEIs, β-blocker and so forth), meanwhile patients needed relevant therapy if they had complications during therapeutic process. On the basis of it, the treatment group received one of the included CHIs, the control group received another or just adopted WM. Besides, the dosages of CHIs were reported; 4) RCTs tested the clinical effective rate. The clinical effective rate calculated by this formula: (number of remarkable recovery patients + number of basic recovery patients) / total number of patients * 100%. Cardiac function classification was conformed to the standard issued by New York Heart Association (NYHA) in the United States. Clinical symptoms disappeared and cardiac function improved 2 levels at least was deemed as the class of remarkable recovery, clinical symptoms relieved and cardiac function increase 1 level was classified into the part of basic recovery, clinical symptoms and cardiac function was unaltered or worse belonged to deterioration. Besides, the incidence of left ventricular ejection fraction (LVEF), cardiac output (CO), stroke volume (SV), 6-min walk test (6MWT), brain natriuretic peptide (BNP), left ventricular end-diastolic dimension (LVEDD), left ventricular end- systolic dimension (LVESD), adverse drug reactions/adverse drug events (ADRs/ADEs) were also evaluated. The clinical effective rate and LVEF were regarded as dominating outcomes of the study, because the clinical effective rate can inflect the efficacy directly and LVEF was a main indicator for CHF. And others were counted as secondary outcomes. A study was excluded when it met these following criteria: 1) the study without full text; 2) duplicated reports; 3) RCTs with incomplete or inaccurate data; 4) RCTs with wrong sequence generation method. For example, sequence generated by odd or even date of birth, some rules based on date (or day) of admission and so forth; 5) patients received physiotherapy, acupuncture and moxibustion therapy, and Chinese materia medica preparation.

A comprehensive literatures searching was carried out in seven database including Embase, the Cochrane Library, Pubmed, Chinese Biological Medicine Database (CBM), China National Knowledge Infrastructure (CNKI), Wanfang Database, Chinese Scientific Journal Database (VIP) from their inception up to June 12nd 2017. In addition, there was no restriction on language. The method that incorporated the medical subject headings (MeSH) term and the free text was applied in searching process, and it would vary from different databases. Each searching item included three parts of terms that chronic heart failure, CHIs, and randomization. Detailed searching strategies were illustrated in Additional file 2.

After literatures duplicate checking, the rest literatures were firstly screened by titles and abstracts, reviews, irrelevant literatures and animals’ experiments reports were excluded. Literatures passed the initial filtration were read full text in order to sort out the eligible RCTs. Two reviewers undertook literature selection respectively, any divergences resolved by discussion or the third reviewer.

Information from the eligible RCTs was extracted based on a custom-made form. The data consisted of the following items: 1) basic information of the eligibility: the first author, nationality, publication year, study desgin; 2) basic characteristics of patients: sample size, gender composition, average age, course of disease, primary diseases, cardiac function classification; 3) detail of RCTs’ intervention; 4) outcomes results and RCTs’ quality assessment.

The quality analysis was assessed with the Cochrane Collaboration’s tools (version 5.1.0 the Nordic Chchrane Center, the Cochrane Collaboration, 2012 Copenhagen, Denmark) by two reviewers independently. The tool comprised following these 7 items: 1) the method of randomization; 2) the concealment of random allocation; 3) the blinding method for patients and clinicians; 4) the blinding method for assessor; 5) the integrality of outcomes data; 6) the condition of selective reporting; 7) others bias. Each item was rated as “high risk”, “low risk” and “unclear”. And any difference between two reviewers settled by discussion or the third reviewer.

It is not necessary for this meta-analysis to obtain an ethical approval, because this study was the procedure that just gathered the clinical data in each RCT without any leak of patients’ information.

NMA was performed with Stata software (version 12, Stata Corporation, College Station, Texas, U.S.) and Winbugs (version 1.4, MRC Biostatistics Unit, Cambridge, UK) software by using Mantel-Haenszel random-effects model. In Winbugs software, the number of iteration was set as 50,000, the first 20,000 was used for annealing algorithm in order to eliminate the impact of initial value. For binary outcomes, the pooled results were calculated as odds ratios (ORs). For continuous outcomes, mean differences (MD) were used. Both types of outcomes were presented with their 95% credible intervals (95% CIs) as well. Besides, the network graph showed indirect comparative relationship between different interventions was described. The node area of each intervention on behalf of its number of patients, and the thickness between different interventions represented the number of relative RCTs [11]. To rank various CHIs in treatment in CHF, the surface under the cumulative ranking curve (SUCRA) was utilized, which expressed each intervention’s efficacy with percentages. A larger area of SCUAR indicated that corresponding intervention was more preferable in certain outcomes [12]. After that, the funnel plots were depicted to reflect publication bias. Due to non-close loops in this NMA, the assumption of consistency between direct and indirect evidence was not utilized.

A total of 9968 literatures were identified in initial search (Fig. 1). After removing duplicates, there were 4852 remained. By screened titles and abstracts, 1491 literatures were excluded because they were irrelevant literatures, reviews and animals’ experiments reports. 3361 literatures were eligible and then examined respectively, among which 3248 were further excluded, for the following reasons: 1) the RCT’s intervention or diseases missed eligibility criteria (n = 2694); 2) the therapeutic effect standard missed eligibility criteria (n = 256);3) the RCT with wrong randomization (n = 68); 4) the RCT did not divide patients in two groups (n = 68); 5) case reports (n = 40); 6) the RCT without full-text (n = 104); 7) the RCT with duplicated data (n = 18). As results, 113 RCTs that evaluated CHIs combined with WM for CHF were eligible in the NMA, and all of then carried out in China between 2001 and 2017. Meanwhile, 6 types of CHIs were identified, including HQI (12 RCTs), SFI (39 RCTs), SI (31 RCTs), SMI (13 RCTs), SQFZI (12 RCTs) and YQFMI (6 RCTs).

One hundered thirteen [3, 13‐124] RCTs with 9525 patients were accorded with the eligible criteria, among which 4852 patients in the treatment groups and 4673 patients in the control groups. Among patients, the male patients were about 55% of total, and majority of patients were middle aged and elderly people. The intervention of the control groups were WM treatment, for instance, ACEIs, β-blocker, cardiotonic, diuretic. In the meantime, the treatment groups received one of the identified CHIs on the basis of the control groups. HQI, SFI, SI, SMI was a kind of injection that clinicians injected them with 5%–10% dextrose solution or 0.9% normal saline, the specific dosage were determined by clinicians. SQFZI was a kind of already made injection with menstruum. And YQFMI was a powder-injection, clinicians injected them with 5%–10% dextrose solution or 0.9% normal saline as well. All of identified CHIs were injected once a day via mainline. Characteristics of included RCTs can be found in Additional file 3. And the compared connections among each intervention for each outcome were displayed in Fig. 2.

For the eligible RCTs, 19 RCTs [15, 26, 32, 34, 37, 56, 57, 61, 64, 75, 81, 88, 96, 101, 103, 104, 107, 110, 118] used the random number table method or sortation randomization method to generate groups and 1 RCTs [88] utilized double blind method. Thus all of them were assessed as low risk. The rest RCTs were evaluated as high risk due to insufficient information. Besides, none of the included RCTs assessed had incomplete data, so the attrition bias was appraised as low risk. As for the part of reporting bias and others bias, the included RCTs did not provide relevant contents about selective report and mention any factors leading to high risk. Therefore these two items were evaluated as unclear risk. The graphical summary was depicted in Fig. 3.

Nevertheless, there was still insufficient in this study. Frist, the enrolled patients in RCTs were merely Chinese, which may lead to a bias on whether non-Chinese use eligible CHIs effectively or not. Although CHIs was mostly adopted in China, clinicians also can not only recruit Chinese. Next, just ten of included RCTs reported 6MWT in this study. While it is 6MWT, and readmission rate that associate with CHF patients closely and influence patients’ survival quality. Thus, these aspects should be paid more emphasis when RCTs are designed. In addition, the methodological quality was general, and most included RCTs did not mention the details of randomization and allocation concealment, which may generate an overestimate for eligible CHIs. It should be note that clinicians utilize low risk randomization and concealment method as possible. Based on the limitations, the RCTs conducted in the future should be perfected in relevant areas.