ABC | Volume 112, Nº3, March 2019

Original Article Kiyose et al A systematic review for valvular prostheses Arq Bras Cardiol. 2019; 112(3):292-301 that of mechanical prosthesis use in terms of mortality, reoperations, the incidence of thromboembolic events, bleeding, and endocarditis. Methods Search strategy and sources The literature search included the following electronic databases: MEDLINE/PubMed, (from 1950 to 04 November 2014), CENTRAL/Cochrane Library, EMBASE/Elsevier (from 1966 to November 4, 2014), SCOPUS/Elsevier (from 1960 to November 4 2014), Web of Science/Thomson Reuters (from 1898 to November 4, 2014), and LILACS/BVS (from 1980 to November 4, 2014), without language and publication date restrictions. Previous systematic reviews and guidelines were consulted to identify and include relevant studies. Other sources were also consulted to identify relevant studies including Clinicaltrials.gov, conference abstracts; lists of text references related to the topic; review articles; and information letters concerning unpublished or incomplete studies. The search strategies were developed by defining descriptors, synonyms and the use of Boolean logical operators (AND, OR, and ANDNOT) for each database (MeSH/Medline, Emtree/Embase, and DeCs/BVS). 9 The MeSH/Medline subject descriptors were sensitised by the strategy of adding "entry terms" (synonyms). In Medline, the Cochrane Handbook Filter 10 was used, which has high sensitivity for recovery of indexed randomized controlled trials (RCTs). Study Selection We included randomized trials in any language that compared native valve replacement with the biological and mechanical prosthesis, regardless of the follow-up period. Observational studies, studies with children or patients under 18 years of age, and studies with patients who required tricuspid valve replacement were excluded. The study eligibility evaluation process consisted of two steps, both performed independently by pairs of reviewers. The first author (ATK) participated in all pairs. The first step consisted of screening articles by reading the title and abstract. In this step, the article was selected for the next step if at least one of the reviewers deemed the article eligible. In the second step, the full article texts were evaluated and selected based on an eligibility form. The final eligibility of the article was decided by agreement between the reviewers or by the judgment of a third reviewer in the event of a disagreement. In the case of multiple publications of the same study, we considered the manuscript reporting the longest follow-up. Data extraction and risk of bias For the data extraction process, we developed a standard form with the clinical information of each patient, including gender, age, functional class, affected valve, type of implanted prosthesis, follow-up period, and methodological characteristics, for further evaluation of evidence quality. An assessment of the risk of bias of the included studies was based on an evaluation of the following domains: random sequence generation, allocation concealment, blinding of outcome assessors, and incomplete outcome data. Blinding of patients and the healthcare team regarding the prosthesis type was not feasible, and these items were therefore not evaluated. We generated a descriptive table to compare the selected studies by classifying the risk of bias as low, moderate, high, or unclear for each risk of bias domain. Outcomes The outcomes measured included total mortality, defined as death from any cause; embolic events, defined as a systemic embolism; bleeding events (of any magnitude); new surgery, defined as the need to replace the prosthesis implanted in the initial procedure; and episodes of infectious endocarditis. Data synthesis and analysis We determined the risk ratios (RRs) and their respective 95% confidence intervals (CIs) for binary outcomes of each trial. Meta-analyses were performed with random effects models using inverse variance. Subgroup analyses were conducted based on the position of valve replacement (aortic, mitral or combined aortic-mitral). Most trials did not report the number of events, only probabilities of events and their standard errors. Thus we calculated the variance of the logarithm of the RR with the formula used by Kassai et al. 8 SE 2 1 p 2 1 SE 2 2 p 2 2 + Where: p 1 = the probability of an event for a mechanical heart valve p 2 = the probability of an event for a bioprosthesis SE 1 = standard error of p 1 SE 2 = standard error of p 2 We assessed the statistical heterogeneity across trials or subgroups using Cochrane’s chi-squared test. The Higgins inconsistency test (I 2 ) was used to quantify the percentage of the variability in the effect estimates that was due to heterogeneity rather than by chance; 11 we considered values of I 2  ≤ 25% as low heterogeneity and values ≥ 50% as high heterogeneity. We conducted these analyses using Review Manager Version 5.2 software (Cochrane IMS, Oxford, UK ). A p-value lower than 0.05 was considered significant. Quality of evidence assessment We assessed the confidence in the estimates of effect (quality of evidence ) using the GRADE (Grades of Recommendation, Assessment, Development, and Evaluation) system. 12 Results Characteristics of included studies The electronic database search resulted in 7,725 citations (Figure 1). After evaluationof the articles, we identified four original studies including 1,528 patients in total. The clinical characteristics of the four included studies are presented in Table 1. 293