Article type
Year
Abstract
Background: Ranking probabilities are one statistical step of the network meta-analysis (NMA) that rank treatments from the highest to the lowest probability of being the best treatment. However, if the certainty of the evidence if not considered in the interpretation of the results of the NMA, the inferences can mislead the clinician and the patient in choosing for a treatment with high ranking probability but very low or low certainty of evidence.
Objectives: To describe the methodological aspects of assessing the certainty of evidence using the GRADE approach (Grades of Recommendation Assessment, Development and Evaluation) for network meta-analysis (NMA) and to introduce a new framework for presenting and interpreting NMA results.
Methods: The methodology described here was used in a random Bayesian NMA of randomized controlled trials (RCTs) conducted to determine the effect of desensitizing toothpastes on dentin hypersensitivity (DH) (PROSPERO #CRD42018086815). We assessed the certainty of evidence and used a new framework proposed by the GRADE working group to present and interpret NMA. We choose a comparator as a reference category, and we separated other treatments into the following categories: 1) those that were more or less effective against the comparator and 2) those with similar effect against the comparator. We next determined the magnitude of the effects following Cohen’ classification. For NMA interpretation, we graded treatments according to certainty of evidence, and we checked consistency with effect estimates and ranking according to SUCRA.
Results: We included 90 trials in our NMA evaluating 16 treatment arms. The SUCRA ranking value suggested that arginine (87.2%) was one of the best treatments for pain relief due to tactile stimulus. However, with low certainty of evidence due to problems in risk of bias and incoherence, we cannot be confident in the final estimate. With the new approach, high-to-moderate treatments were considered effective against the comparator.
Conclusions: This NMA reported a new GRADE framework for presenting and interpreting results. The judgement that places interventions in categories relies primarily on the magnitude of the effect estimates, the certainty of evidence for those estimates, and secondarily their order in the ranking. This approach can avoid misleading inferences based solely on SUCRA ranking.
Patient or healthcare consumer involvement: Our approach demonstrates that the interpretation of data following SUCRA rankings can lead to misleading inferences. Instead, the interpretation based on certainty of evidence and the magnitude of the effect estimates can help the clinician and the patient to a shared-decision making related to the best treatment option for the patient.
Funding: CAPES; CNPq; FAPEMIG.
Objectives: To describe the methodological aspects of assessing the certainty of evidence using the GRADE approach (Grades of Recommendation Assessment, Development and Evaluation) for network meta-analysis (NMA) and to introduce a new framework for presenting and interpreting NMA results.
Methods: The methodology described here was used in a random Bayesian NMA of randomized controlled trials (RCTs) conducted to determine the effect of desensitizing toothpastes on dentin hypersensitivity (DH) (PROSPERO #CRD42018086815). We assessed the certainty of evidence and used a new framework proposed by the GRADE working group to present and interpret NMA. We choose a comparator as a reference category, and we separated other treatments into the following categories: 1) those that were more or less effective against the comparator and 2) those with similar effect against the comparator. We next determined the magnitude of the effects following Cohen’ classification. For NMA interpretation, we graded treatments according to certainty of evidence, and we checked consistency with effect estimates and ranking according to SUCRA.
Results: We included 90 trials in our NMA evaluating 16 treatment arms. The SUCRA ranking value suggested that arginine (87.2%) was one of the best treatments for pain relief due to tactile stimulus. However, with low certainty of evidence due to problems in risk of bias and incoherence, we cannot be confident in the final estimate. With the new approach, high-to-moderate treatments were considered effective against the comparator.
Conclusions: This NMA reported a new GRADE framework for presenting and interpreting results. The judgement that places interventions in categories relies primarily on the magnitude of the effect estimates, the certainty of evidence for those estimates, and secondarily their order in the ranking. This approach can avoid misleading inferences based solely on SUCRA ranking.
Patient or healthcare consumer involvement: Our approach demonstrates that the interpretation of data following SUCRA rankings can lead to misleading inferences. Instead, the interpretation based on certainty of evidence and the magnitude of the effect estimates can help the clinician and the patient to a shared-decision making related to the best treatment option for the patient.
Funding: CAPES; CNPq; FAPEMIG.