Part 2: Evidence Evaluation and Management of Conflicts of Interest
2015; Lippincott Williams & Wilkins; Volume: 132; Issue: 18_suppl_2 Linguagem: Inglês
10.1161/cir.0000000000000253
ISSN1524-4539
AutoresLaurie J. Morrison, Lana M. Gent, Eddy Lang, Mark Nunnally, Melissa Parker, Clifton W. Callaway, Vinay Nadkarni, Antonio R. Fernandez, John E. Billi, Jonathan R. Egan, Russell E. Griffin, Michael Shüster, Mary Fran Hazinski,
Tópico(s)Cardiac, Anesthesia and Surgical Outcomes
ResumoHomeCirculationVol. 132, No. 18_suppl_2Part 2: Evidence Evaluation and Management of Conflicts of Interest Free AccessResearch ArticlePDF/EPUBAboutView PDFView EPUBSections ToolsAdd to favoritesDownload citationsTrack citationsPermissions ShareShare onFacebookTwitterLinked InMendeleyReddit Jump toFree AccessResearch ArticlePDF/EPUBPart 2: Evidence Evaluation and Management of Conflicts of Interest2015 American Heart Association Guidelines Update for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care Laurie J. Morrison, Lana M. Gent, Eddy Lang, Mark E. Nunnally, Melissa J. Parker, Clifton W. Callaway, Vinay M. Nadkarni, Antonio R. Fernandez, John E. Billi, Jonathan R. Egan, Russell E. Griffin, Michael Shuster and Mary Fran Hazinski Laurie J. MorrisonLaurie J. Morrison , Lana M. GentLana M. Gent , Eddy LangEddy Lang , Mark E. NunnallyMark E. Nunnally , Melissa J. ParkerMelissa J. Parker , Clifton W. CallawayClifton W. Callaway , Vinay M. NadkarniVinay M. Nadkarni , Antonio R. FernandezAntonio R. Fernandez , John E. BilliJohn E. Billi , Jonathan R. EganJonathan R. Egan , Russell E. GriffinRussell E. Griffin , Michael ShusterMichael Shuster and Mary Fran HazinskiMary Fran Hazinski Originally published3 Nov 2015https://doi.org/10.1161/CIR.0000000000000253Circulation. 2015;132:S368–S382IntroductionThis Part describes the process of creating the 2015 American Heart Association (AHA) Guidelines Update for Cardiopulmonary Resuscitation (CPR) and Emergency Cardiovascular Care (ECC), informed by the 2015 International Consensus on CPR and ECC Science With Treatment Recommendations (CoSTR) publication.1,2 The process for the 2015 International Liaison Committee on Resuscitation (ILCOR) systematic review is quite different when compared with the process used in 2010.1–3 For the 2015 systematic review process, ILCOR used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) (www.gradeworkinggroup.org) approach to systematic reviews and guideline development. For the development of this 2015 Guidelines Update, the AHA used the ILCOR reviews as well as the AHA definition of Classes of Recommendation (COR) and Levels of Evidence (LOE) (Table 1). This Part summarizes the application of the ILCOR GRADE process to inform the creation of 2015 Guidelines Update, and the process of assigning the AHA COR and LOE.Table 1. Applying Class of Recommendations and Level of Evidence to Clinical Strategies, Interventions, Treatments, or Diagnostic Testing in Patient Care*Table 1. Applying Class of Recommendations and Level of Evidence to Clinical Strategies, Interventions, Treatments, or Diagnostic Testing in Patient Care*Development of the 2015 Consensus on Science With Treatment RecommendationsGrading of Recommendations Assessment, Development, and EvaluationThe 2015 CoSTR summarizes the published scientific evidence that was identified to answer specific resuscitation questions. ILCOR uses the GRADE system to summarize evidence and determine confidence in estimates of effect as well as to formulate treatment recommendations. GRADE is a consensus-crafted tool in wide use by many professional societies and reference organizations, including the American College of Physicians, the American Thoracic Society, and the Cochrane Collaboration, as well as the Centers for Disease Control and the World Health Organization. The choice of the GRADE approach was based on its increasingly ubiquitous use, practicality, and unique features. To our knowledge, the ILCOR evidence review process represents the largest application of the GRADE system in a healthcare-related review.GRADE is a system to review evidence to determine the confidence in the estimate of effect of an intervention or the performance of a diagnostic test and to categorize the strength of a recommendation. GRADE requires explicit documentation of the evaluation of the evidence base specific to each outcome that was chosen and ranked as critical and important before the evidence review. The evidence is assessed by multiple criteria. Questions addressed in GRADE typically follow a PICO (population, intervention, comparator, outcome) structure for ease of mapping to available evidence (Figure 1).Download figureDownload PowerPointFigure 1. Structure of questions for evidence evaluation.Confidence in the estimates of effect, synonymous with and reported more succinctly as quality, is reported by a synthesis of evidence informed by 1 or more studies as opposed to studies themselves. Quality is adjudicated by a 4-part ranking of our confidence in the estimate of effect (high, moderate, low, very low) informed by study methodology and the risk of bias. Studies start but do not necessarily end at high confidence for randomized controlled trials (RCTs), and they start but do not necessarily end at low confidence for observational studies. Studies may be downgraded for inconsistency, imprecision, indirectness, and publication bias and nonrandomized observational studies may be upgraded as the result of effect size, dose-response gradient, and plausible negative confounding; in other words, an underestimation of the association. The direction and strength of recommendations are driven by certainty of evidence effect estimates, values and preferences of patients, and, to some degree, clinicians' balance of positive and negative effects, costs and resources, equity, acceptability, and feasibility (Table 2).Table 2. From GRADE Evidence to Decision Factors for Making Strong Versus Weak RecommendationsFactorRelevant QuestionNotesPriority of problemIs the problem addressed by the question important enough to make a recommendation?Many problems may not be identified a priori as high enough importance to justify strong recommendations when weighed against other problems.Balance of benefits and harmsAcross outcomes, are the overall effects and confidence in those effects a net gain?Most interventions, prognostications, and diagnostic tests have positive and negative consequences. Confidence in these estimates must be viewed in aggregate—do positive effects outweigh negative ones? Consideration must weigh outcomes by importance.Certainty in the evidenceWhat is the overall certainty that these estimates will support a recommendation?More certainty supports stronger recommendations, and vice versa.Values and preferencesTo what extent do the values and preferences of patients regarding outcomes or interventions vary?Minimal variation and a strong endorsement of the outcomes or the interventions based on patients' values and preferences supports stronger recommendations. The lack of consistency in patients' values and preferences or a weak endorsement of the outcomes or the interventions supports weaker recommendations.Costs and resourcesAre these net results proportionate to the expenditures and demands of the recommended measure?Factors such as manpower, time, distraction from other tasks, and monetary investment are viewed through local values. Lower costs of an intervention and greater cost-effectiveness support strong recommendations, and vice versa. Analysis should account for uncertainty in the calculated costs.EquityAre the net positive effects of the measure distributed justly?Measures that improve disparities or benefit fairly may drive a stronger recommendation, and vice versa.AcceptabilityAcross stakeholders, is the measure tractable?To be strong, a recommendation ideally appeals to most.FeasibilityCan the recommendation be implemented from a practical standpoint?Something that is practical to achieve may support a strong recommendation, and vice versa.Summary: To what extent do positive and negative consequences balance in the settings in question?Negative clearly outweighs positiveNegative probably outweighs positiveNegative and positive consequences balancedPositive probably outweighs negativePositive clearly outweighs negativeStrong recommendation againstWeak recommendation againstWeak recommendation forStrong recommendation forConsiderations: Are there important subgroups that might be treated differently? Are there important concerns for implementation?The GRADE Development ToolThe GRADE Guideline Development Tool (www.guidelinedevelopment.org) provides a uniform interface in the form of standardized evidence profiles and sets forth a framework that enables the reviewer to synthesize the evidence and make a treatment recommendation.4GRADE uniquely unlocks the often rigid linkage between one's confidence in the estimate of effect from the strength of a recommendation. Although the two are related, different factors (eg, costs, values, preferences) influence the strength of the recommendation independent of one's confidence in the estimate of effect. GRADE mandates explicit reasons for judgments in a transparent structure. The GRADE Guideline Development Tool4 requires consideration of all of these factors and documentation for each decision. To qualify recommendations, an evidence-to-recommendation framework is used to document all factors that shape the recommendation. Finally, with the GRADE Guideline Development Tool, summary of evidence and evidence profile tables are created. The tables summarize effect size, confidence in the estimates of effect (quality), and the judgments made to evaluate evidence at the level of outcomes. Quality is specified across each of multiple outcomes for the same population, intervention, and comparison, with judgments documented in explanatory notes.Scientific Evidence and Evaluation Review SystemIn preparation for the 2015 systematic review process, ILCOR members, the AHA ECC staff, and compensated consultants collaborated to develop an online systematic review website. The Systematic Evidence Evaluation and Review System (SEERS) website was designed to support the management of workflow steps required to complete the ILCOR systematic reviews (in 2010, these were called worksheets) and capture the evidence extraction and evaluation data in reusable formats (Figure 2). The SEERS website facilitated the structured and consistent evidence review process, which enabled the task force members to finalize the CoSTR for each PICO question. Successful completion of the systematic review process ensured consistency in elements of the reviews from many different international reviewers.Download figureDownload PowerPointFigure 2. ILCOR 2015 Consensus on Science work flow for all systematic reviews.Steps in the ILCOR 2015 Systematic Review ProcessILCOR created a comprehensive overview of the structured process that was used to support systematic reviews. The process was divided into 5 major categories, as outlined in Figure 2:PICO question development: systematic review question development, using the PICO format (Figure 1)Search strategy developmentEvidence reviewer article selectionGRADE evidence reviewDevelopment of CoSTRILCOR PICO Question DevelopmentShortly after the 2010 International Consensus on CPR and ECC Science With Treatment Recommendations and the 2010 AHA Guidelines for CPR and ECC were published, the 2015 ILCOR task forces reviewed the 274 PICO questions that were addressed in 2010 and generated a comprehensive list of 336 questions for potential systematic reviews in 2015. In addition, the new ILCOR task force, First Aid, developed 55 PICO questions that were initially prioritized for review. Questions were prioritized based on clinical controversy, emerging literature, and previously identified knowledge gaps. ILCOR task forces debated and eventually voted to select a focused group of questions. Of the 391 potential PICO questions generated by the task forces, a total of 165 (42%) systematic reviews were completed for 2015 (Figures 3 and 4). The number of PICO questions addressed by systematic reviews varied across task forces (Figure 4).Download figureDownload PowerPointFigure 3. ILCOR process for prioritizing PICO questions for systematic reviews.Download figureDownload PowerPointFigure 4. Comparison of the number of systematic review questions (PICO questions) addressed or deferred/not reviewed in 2015 versus 2010 reported by Part in the ILCOR International Consensus on CPR and ECC Science With Treatment Recommendations (CoSTR) publication. BLS indicates Basic Life Support; Defib: Defibrillation*; CPR Tech and Dev: Cardiopulmonary Resuscitation Techniques and Devices; ALS: Advanced Life Support; ACS: Acute Coronary Syndromes; Peds: Pediatrics; NLS: Neonatal Resuscitation; EIT, Education, Implementation, and Teams. *Note that defibrillation content (Defib) of 2010 was absorbed within the 2015 Basic Life Support, Advanced Life Support, and Pediatric CoSTR parts, and the CPR Techniques and Devices questions of 2010 were absorbed by the Advanced Life Support CoSTR part in 2015.Consistent with adopting the GRADE guideline writing process, clinical outcomes for each PICO were selected and ranked on a 9-point scale as critical and important for decision making by each task force. The GRADE evidence tables were reported by outcome, based on the priority of the clinical outcome. After task force selection of PICO questions for review in 2015, individuals without any conflicts of interest (COIs) or relevant commercial relationships were identified and selected from task force members to serve as task force question owners. Task force question owners provided the oversight control to ensure progress and completion of each systematic review.ILCOR Search Strategy DevelopmentTask force question owners worked in an iterative process with information specialists from St. Michael's Hospital Health Science Library in Toronto on contract as compensated consultants to the AHA. These information specialists created comprehensive literature search strategies. The information specialists collaborated with the task force question owners to create reproducible search strings that were customized for ease of use within the Cochrane Library (The Cochrane Collaboration, Oxford, England), PubMed (National Library of Medicine, Washington, DC), and Embase (Elsevier B.V., Amsterdam, Netherlands). Each search string was crafted with precision to meet the inclusion and exclusion criteria that were defined to balance the importance of sensitivity and specificity for a comprehensive literature search.With commitment to a transparent systematic review process for 2015, ILCOR provided an opportunity for public comment on proposed literature search strategies. Members of the public were able to review search strategies and use the search strings to view the literature that would be captured. ILCOR received 18 public comments and suggestions based on the proposed search strategies and forwarded them to the task force chairs and task force question owners for consideration. This iterative process ensured that specific articles were captured during the evaluation process that may not have been initially retrieved by the search strategy.ILCOR Evidence Reviewers' Article SelectionUpon completion of the public comment process, ILCOR invited topic experts from around the world to serve as evidence reviewers. Specialty organizations were also solicited to suggest potential evidence reviewers. The qualifications of each reviewer were assessed by the task force, and potential COIs were disclosed and evaluated by the task force co-chairs and COI co-chairs. Evidence reviewers could not have any significant COI issues pertaining to their assigned topics. If a COI was identified, the topic was assigned to a different reviewer who was free from conflict.Two evidence reviewers were invited to complete independent reviews of the literature for each PICO question. A total of 250 evidence reviewers from 39 countries completed 165 systematic reviews. The results of the search strategies were provided to the evidence reviewers. Each reviewer selected articles for inclusion, and the 2 reviewers came to agreement on articles to include before proceeding to the next step in the review process. If disagreement occurred in the selection process, the task force question owner served as a moderator to facilitate agreement between the reviewers. If necessary, the search strategy was modified and repeated based on feedback from the evidence reviewers. When final agreement was reached between the evidence reviewers on included studies, the systematic review process started.ILCOR GRADE Evidence ReviewThe bias assessment process capitalized on existing frameworks for defining the risk of systematic error in research reporting through 3 distinct approaches. The Cochrane tool was used to evaluate risk of bias in randomized trials,5,6 whereas the QUADAS-2 instrument7 was used for included studies that supported diagnostic PICO questions. For non-RCTs that drew inferences on questions of therapy or prognosis, the GRADE working group risk-of-bias criteria8 were used as a series of 4 questions that emphasized sampling bias, the integrity of predictor and outcome measurements, loss to follow-up, and adjusting for confounding influences.8,9 Occasionally an existing systematic review would be uncovered that could formally address risk of bias as it pertained to a specific outcome. However, in most instances, the task forces used an empiric approach based on an amalgamation of risk from individual studies addressing a specific outcome. The 2 (or more) reviewers were encouraged to consolidate their judgments, with adjudication from the task force if needed. Agreed bias assessments were entered into a GRADE evidence profile table.The GRADE Guideline Development Tool is a freely available online resource that includes the GRADE evidence profile table.4a GRADE Guideline Development Tool served as an invaluable aid to summarize important features, strengths, and limitations of the selected studies. To complete each cell of the evidence profile table, reviewers needed to apply judgments on the 5 dimensions of quality, including risk of bias, inconsistency, indirectness, imprecision, and other considerations (including publication bias). Quantitative data that described effect sizes and confidence intervals were also entered into the evidence profiles, although a more descriptive approach was used when pooling was deemed inappropriate. The GRADE Guideline Development Tool software calculated the quality of evidence for critical and important outcomes by row and, when therapy questions (the most common type) were addressed, generated impact estimates for groups at high, moderate, or low baseline risk as a function of the relative risk.2015 ILCOR Development of Draft Consensus on Science With Treatment RecommendationsILCOR developed a standardized template for drafting the consensus on science to capture a narrative of the evidence profile and reflect the outcome-centric approach emphasized by GRADE. The consensus on science reported (1) the importance of each outcome, (2) the quality of the evidence and (3) the confidence in estimate of effect of the treatment (or diagnostic accuracy) on each outcome, (4) the GRADE reasons for downgrading or upgrading the quality rating of the study, and (5) the effect size with confidence intervals or a description of effects when pooling was not done.The ILCOR task forces created treatment recommendations when consensus could be reached. Within the GRADE format, 4 recommendations are possible: (1) strong recommendation in favor of a treatment or diagnostic test, (2) strong recommendation against a treatment or diagnostic test, (3) weak recommendation in favor of a treatment or diagnostic test, or (4) weak recommendation against a treatment or diagnostic test. A strong recommendation is indicated by the words "we recommend" and a weak recommendation is indicated by the words "we suggest."Within the GRADE Guideline Development Tool, an evidence-to-recommendation framework assisted reviewers in making explicit the values and preferences that drove their recommendations, especially when evidence was either uncertain or was a weaker determinant of the optimal course of action. In doing so, resource considerations were invoked rarely when an economic analysis was identified and reviewed as germane or when the balance of risks and harms were considered by the task force to be weighed clearly against potential benefits. When there was inadequate or conflicting evidence, the task force would indicate this insufficient evidence with language such as, "The confidence in effect estimates is so low that the panel feels a recommendation to change current practice is too speculative." If economic analyses were not available, or if the task forces thought that the appropriate recommendations could differ among the resuscitation councils based on training implications or structure or resources of out-of-hospital or in-hospital resuscitation systems, then the task forces occasionally made no recommendations, leaving that to the council guidelines.The task force members reviewed, discussed and debated the evidence and developed consensus wording on the summary consensus on science statements and on the treatment recommendations during in-person meetings and after the 2015 ILCOR International Consensus on CPR and ECC Science With Treatment Recommendations Conference, held in Dallas, Texas, in February 2015. In addition, the task forces met frequently by webinar to develop the draft documents that were submitted for peer review on June 1, 2015. As in 2005 and 2010, strict COI monitoring and management continued throughout the process of developing the consensus on science statements and the treatment recommendations, as described in "Part 2: Evidence Evaluation and Management of Conflicts of Interest" in the 2015 CoSTR.10,11Public Comment on the ILCOR Draft Consensus on Science With Treatment RecommendationsAll draft recommendations were posted to allow approximately 6 weeks of public comment, including COI disclosure from those commenting. In addition, the ILCOR draft consensus on science statements and treatment recommendations developed during the January 2015 conference were posted the week after the conference, and 492 public comments were received through February 28, 2015, when the comment period closed. The CoSTR drafts were reposted to remain available through April 2015 to allow optimal stakeholder engagement and familiarity with the proposed recommendations.Development of the 2015 Guidelines UpdateThe 2015 Guidelines Update serves as an update to the 2010 Guidelines. The 2015 Guidelines Update addresses the new recommendations that arose from the 2015 ILCOR evidence reviews of the treatment of cardiac arrest and advanced life support for newborns, infants, children, and adults.Formation of the AHA Guidelines Writing GroupsThe AHA exclusively sponsors the 2015 Guidelines Update and does not accept commercial support for the development or publication. The AHA ECC Committee proposed 14 Parts of the Guidelines, which differ slightly from the 2010 Parts (Table 3).Table 3. Contents of 2010 Guidelines Compared With 2015 Guidelines Update2010 Guidelines2015 Guidelines UpdateExecutive SummaryExecutive SummaryEvidence Evaluation and Management of Potential or Perceived Conflicts of InterestEvidence Evaluation and Management of Conflicts of InterestEthicsEthical IssuesCPR OverviewSystems of Care and Continuous Quality Improvement*†Adult Basic Life SupportAdult Basic Life Support and Cardiopulmonary Resuscitation Quality*†Electrical Therapies: Automated External Defibrillators, Defibrillation, Cardioversion, and Pacing(Defibrillation content embedded in other Parts)CPR Techniques and DevicesAlternative Techniques and Ancillary Devices for Cardiopulmonary ResuscitationAdult Advanced Cardiovascular Life SupportAdult Advanced Cardiovascular Life Support‡Post–Cardiac Arrest CarePost–Cardiac Arrest CareAcute Coronary SyndromesAcute Coronary SyndromesAdult Stroke(Relevant stroke content embedded in other Parts)Cardiac Arrest in Special SituationsSpecial Circumstances of ResuscitationPediatric Basic Life SupportPediatric Basic Life Support and Cardiopulmonary Resuscitation Quality†Pediatric Advance Life SupportPediatric Advanced Life Support‡Neonatal ResuscitationNeonatal ResuscitationEducation, Implementation, and TeamsEducationFirst AidFirst Aid*Includes prehospital stroke.†Includes AED defibrillation.‡Includes manual defibrillation.AED indicates automated external defibrillator; and CPR, cardiopulmonary resuscitation.In particular, content from 2010 Parts (electrical therapies, adult stroke) have been incorporated into other Parts, and a new Part that addresses systems of care and continuous quality improvement has been added. The committee nominated a slate of writing group chairs and writing group members for each Part. Writing group chairs were chosen based on their knowledge, expertise, and previous experience with the Guidelines development process. Writing group members were chosen for their knowledge and expertise relevant to their Part of the Guidelines. In addition, each writing group included at least 1 young investigator. The ECC Committee approved the composition of all writing groups before submitting them to the AHA Officers and Manuscript Oversight Committee for approval.Part 15 of the Guidelines Update, "First Aid," is jointly sponsored by the AHA and the American Red Cross. The writing group chair was selected by the AHA and the American Red Cross, and writing group members were nominated by both the AHA and the American Red Cross and approved by the ECC Committee. The evidence review for this Part was conducted through the ILCOR GRADE evidence review process.Before confirmation, all Guidelines writing group chairs and members were required to complete an AHA COI disclosure of all current healthcare-related relationships. The declarations were reviewed by AHA staff and the AHA officers. All writing group chairs and a minimum of 50% of the writing group members were required to be free of relevant COIs and relationships with industry. During the 2015 Guidelines development process, writing group members were requested to update their disclosure statements every 3 months.Classification of AHA Guidelines RecommendationsIn developing the 2015 Guidelines Update, the writing groups used the latest version of the AHA format for COR and LOE (Table 1). The COR indicates the strength that the writing group assigns the recommendation, based on the anticipated magnitude and certainty of benefit relative to risk. The LOE is assigned based on the type, quality, quantity, and consistency of scientific evidence supporting the effect of the intervention.2015 AHA Classes of RecommendationBoth the 2010 Guidelines and the 2015 Guidelines Update used the AHA Classification system that includes 3 main classes of positive recommendations: Class I, Class IIa, and Class IIb (Figure 5).Download figureDownload PowerPointFigure 5. Class of Recommendation comparison between 2010 Guidelines and 2015 Guidelines Update.A Class I recommendation is the strongest recommendation, indicating the writing group's judgment that the benefit of an intervention greatly outweighs its risk. Such recommendations are considered appropriate for the vast majority of clinicians to follow for the vast majority of patients, with infrequent exceptions based upon the judgment of practitioners in the context of the circumstances of individual cases; there is greater expectation of adherence to a Class I recommendation.Class IIa recommendations are considered moderate in strength, indicating that an intervention is reasonable and generally useful. Most clinicians will follow these recommendations most of the time, although some notable exceptions exist. Class IIb recommendations are the weakest of the positive recommendations for interventions or diagnostic studies. Class IIb recommendations are identified by language (eg, "may/might be reasonable or may/might be considered") that indicates the intervention or diagnostic study is optional because its effect is unknown or unclear. Although the clinician may consider the treatment or diagnostic study with a Class IIb recommendation, it is also reasonable to consider other approaches.The past AHA format for COR contained only 1 negative classification, a Class III recommendation. This classification indicated that the therapy or diagnostic test was not helpful, could be harmful, and should not be used. In the 2015 Guidelines Update, there are 2 types of Class III recommendations, to clearly distinguish treatments or tests that may cause harm from those that have been disproven. A Class III: Harm recommendation is a strong one, signifying that the risk of the intervention (potential harm) outweighs the benefit, and the intervention or test should not be used. The second type of Class III recommendation, the Class III: No Benefit, is a moderate recommendation, generally reserved for therapies or tests that have been shown in high-level studies (generally LOE A or B) to provide no benefit when tested against a placebo or control. This recommendation signifies that there is equal likelihood of benefit and risk, and experts agree that the intervention or test should not be used.2015 AHA Levels of EvidenceIn the 2010 Guidelines, only 3 LOEs were used to indicate the quality of the data: LOEs A, B, and C. LOE A indicated evidence from multiple populations, specifically from multiple randomized clinical trials or meta-analyses. LOE B indicated that limited populations were evaluated, and evidence was derived from a single randomized trial or nonrandomized studies. LOE C indicated that either limited populations were studied or the evidence consisted of case series or expert consensus. In this 2015 Guidelines Update, there are now 2 types of LOE B evidence, LOE B-R and LOE B-NR: LOE B-R (randomized) indicates moderate-quality evidence from 1 or more
Referência(s)