Article type
Abstract
Background: When engaging in shared decision making (SDM) clinicians and patients need to discuss the benefits and harms of available treatment options. To avoid framing bias, decision aids should present the evidence as absolute estimates of effect, but there is no established methodology to obtain them in the context of multiple-treatment comparisons.
Objectives: 1) To provide a methodologically sound framework to calculate absolute-effect estimates of multiple interventions in the presence and absence of a network meta-analyses (NMA), and with variable sources of baseline risk; 2) To implement this framework in an online prototype that generates decision aids for the clinical encounter from evidence summaries (www.MagicApp.org).
Methods: A group of methodologists with experience in systematic reviews, network meta-analysis, and SDM brainstormed on how to obtain absolute effects in the context of network meta-analyses (NMA), tested approaches in real datasets, and incorporated feedback from experts. We used data from the studies included in the network to inform the baseline risk for the intervention chosen as the reference, and tested the impact of the choice of reference and the sequence in moving from one treatment to the next. We will also explore the feasibility of using baseline risks from other sources, and applying this approach when no NMA is available.
Results: By multiplying the mixed effects of the basic comparisons with the chosen baseline risk, we can obtain the corresponding risks for the remaining interventions. Assuming that each corresponding risk is transitive in comparisons that contain the anchor intervention, we can obtain the corresponding risks pertaining to the functional comparisons. We will present this approach at the summit using real datasets from NMA, and illustrate how it can inform the creation of decision aids for the clinical encounter in our online prototype for multiple comparisons.
Conclusions: Obtaining absolute-effect estimates in the context of multiple-treatment comparisons remains a challenge, but is critical if we want such evidence to reach patients and clinicians and support actual SDM.
Objectives: 1) To provide a methodologically sound framework to calculate absolute-effect estimates of multiple interventions in the presence and absence of a network meta-analyses (NMA), and with variable sources of baseline risk; 2) To implement this framework in an online prototype that generates decision aids for the clinical encounter from evidence summaries (www.MagicApp.org).
Methods: A group of methodologists with experience in systematic reviews, network meta-analysis, and SDM brainstormed on how to obtain absolute effects in the context of network meta-analyses (NMA), tested approaches in real datasets, and incorporated feedback from experts. We used data from the studies included in the network to inform the baseline risk for the intervention chosen as the reference, and tested the impact of the choice of reference and the sequence in moving from one treatment to the next. We will also explore the feasibility of using baseline risks from other sources, and applying this approach when no NMA is available.
Results: By multiplying the mixed effects of the basic comparisons with the chosen baseline risk, we can obtain the corresponding risks for the remaining interventions. Assuming that each corresponding risk is transitive in comparisons that contain the anchor intervention, we can obtain the corresponding risks pertaining to the functional comparisons. We will present this approach at the summit using real datasets from NMA, and illustrate how it can inform the creation of decision aids for the clinical encounter in our online prototype for multiple comparisons.
Conclusions: Obtaining absolute-effect estimates in the context of multiple-treatment comparisons remains a challenge, but is critical if we want such evidence to reach patients and clinicians and support actual SDM.