This meta-analysis was performed in accordance with the PRISMA guidelines[17]. The protocol was registered with PROSPERO in September 2020, number CRD420202 04670. Ethical approval and patient consent were not required because this study was based on previous studies.
2.1. Literature search strategy
We systematically searched five electronic databases, including PubMed, EMBASE, Web of Science, Cochrane Library, and Clinicaltrials.gov, from their inception until March 2021 for the comparative studies of restrictive transfusion and liberal transfusion in patients with AMI and anemia. Restrictive transfusion was defined as hemoglobin threshold ≤ 8 g/dL or hematocrit ≤ 24%, while liberal transfusion was defined as hemoglobin threshold ≤ 10 g/dL or hematocrit ≤ 30%[18]. In order to systematically search these electronic databases, search terms were constructed as follows: (Transfusion OR Blood transfusion OR Red blood cell transfusion) AND (Myocardial infarction OR Acute myocardial infarction OR ST-segment evaluated myocardial infarction OR Non-ST-segment evaluated myocardial infarction OR Acute coronary syndrome OR Percutaneous coronary intervention). The search was not restricted for trials by type, language, or publication status. To screen for additional studies, the reference lists of the included articles and previous relevant meta-analyses were also carefully scanned. Additionally, the major international cardiology meetings (the European Society of Cardiology, the American Heart Association, and the American College of Cardiology) were also searched for relevant conference abstracts with complete results.
2.2. Study selection
The studies had to satisfy the following criteria to be included: (1) patients suffered from AMI and anemia; (2) one group received the liberal and the other group received the restrictive red blood cell transfusion strategy; and (3) data regarding the risk of in-hospital or follow-up mortality and in-hospital or follow-up major adverse cardiovascular events (MACEs) were available. MACEs were defined as reinfarction, stroke, or acute heart failure.
Studies designed to compare the efficacy between blood transfusion and non-transfusion patients with AMI but without separate data on different transfusion strategies were excluded. In addition, other exclusion criteria were as follows: (1) reviews, meta-analyses, letters, and conferences; (2) in vitro or preclinical animal studies; (3) enrolled pediatric patients; and (4) duplicate data. Two authors (YHL and YSZ) independently screened the articles based on titles and abstracts. We solved the disagreements and reached a consensus through discussion or arbitration by the third reviewer (YMH).
2.3. Outcomes
The primary end point was all-cause mortality, including overall mortality, in-hospital mortality, or follow-up mortality (up to 6 months). The secondary outcomes were follow-up MACEs, including reinfarction, stroke, and acute heart failure.
2.4. Data extraction and quality assessment
Data extraction was independently carried out by two reviewers (YHL and YSZ) on the basis of prespecified extraction criteria. The following information was extracted from each included study: author, publication date, country, study design, sample size, patients’ characteristics, transfusion strategy, and outcomes. The third reviewer (YMH) checked for discrepancies and helped to settle the disagreements.
The quality of included studies was also independently evaluated by two reviewers (YHL and YSZ). The Cochrane Collaboration’s risk of bias tool was used to assess the quality of RCTs[19], which comprised assessment of selection bias and other bias. To evaluate the quality of cohort studies, we used the Newcastle-Ottawa Scale[20], which consists of eight items about sampling methods, comparability, and accuracy of results. The disagreements were resolved through discussion among the reviewers and judged by the third reviewer (YMH).
2.5. Statistical analysis
The statistical analyses were carried out using Review Manager (version 5.3) and R software (version 4.0.2). Heterogeneity of studies was assessed using Q statistics and I2. I2 value < 25% indicated low heterogeneity, 25%–50% denoted moderate heterogeneity, while the values over 50% defined severe heterogeneity. In order to minimize bias due to methodological differences between the studies, random-effects models were selected. Differences in results of all of the included studies were represented in the form of forest plots. Overall mortality was analyzed by subgroup analysis based on the study design and the number of cases. The sensitivity analysis of omission was carried out to evaluate the robustness of the results. Additionally, the funnel plot analysis and Egger’s test were used to evaluate the publication bias. Risk ratios (RR) with 95% confidence intervals (CI) were presented and pooled by random-effects models. All the differences in the two-tailed test with P value less than 0.05 were considered to be statistically significant.