Article type
Year
Abstract
Background:
Network meta-analysis (NMA) is an extension of pairwise meta-analysis that synthesizes evidence on multiple interventions. An important advantage of NMA is that it provides the ranking of all evaluated interventions for a specific outcome. To date, all ranking approaches are based solely on the NMA relative effects. However, a range of characteristics may influence the choice of the most appropriate interventions in practice. For example, the credibility of the evidence is often limited for some comparisons, or important prior knowledge about treatment performance is available from clinical experience.
Objectives:
To present a new approach for treatment ranking that combines the relative effects and other characteristics affecting the treatment choice in clinical practice.
Methods:
We used methodology from social networks and we adapted the way that websites are ranked in search engine results into treatment networks. Our starting point was a matrix containing the probability that treatment X performs better than Y for each pair of treatments (X and Y) for the studied outcome. To develop the new ranking approach we used two key components:
1) a ‘preference vector’ (P) containing the probability of selecting a treatment if they were all equally effective; and
2) a ‘weighting factor’ (w) defining the weight of relative effects and clinical preference in ranking estimation.
Results:
We will illustrate our method using exemplar NMAs with different preference criteria. We focussed on a Cochrane NMA comparing the effectiveness of 20 drugs for chronic plaque psoriasis. We defined P based on the data availability for each drug, the credibility of the evidence for every comparison and study characteristics related to the generalizability of findings. Figure 1 shows how the rankings change depending on the weight of the relative effects in ranking estimation. Ixekizumab appears to be always the most effective drug while important changes in ranking occur after w is smaller than 65%.
Conclusions:
The usefulness of conventional ranking approaches can be questionable under several settings. A treatment ranking that accounts for the potential limitations of the data and incorporates information on top of the relative effects is an important tool to decision makers.
Patient or healthcare consumer involvement:
Patients or healthcare consumers have not been involved.
Network meta-analysis (NMA) is an extension of pairwise meta-analysis that synthesizes evidence on multiple interventions. An important advantage of NMA is that it provides the ranking of all evaluated interventions for a specific outcome. To date, all ranking approaches are based solely on the NMA relative effects. However, a range of characteristics may influence the choice of the most appropriate interventions in practice. For example, the credibility of the evidence is often limited for some comparisons, or important prior knowledge about treatment performance is available from clinical experience.
Objectives:
To present a new approach for treatment ranking that combines the relative effects and other characteristics affecting the treatment choice in clinical practice.
Methods:
We used methodology from social networks and we adapted the way that websites are ranked in search engine results into treatment networks. Our starting point was a matrix containing the probability that treatment X performs better than Y for each pair of treatments (X and Y) for the studied outcome. To develop the new ranking approach we used two key components:
1) a ‘preference vector’ (P) containing the probability of selecting a treatment if they were all equally effective; and
2) a ‘weighting factor’ (w) defining the weight of relative effects and clinical preference in ranking estimation.
Results:
We will illustrate our method using exemplar NMAs with different preference criteria. We focussed on a Cochrane NMA comparing the effectiveness of 20 drugs for chronic plaque psoriasis. We defined P based on the data availability for each drug, the credibility of the evidence for every comparison and study characteristics related to the generalizability of findings. Figure 1 shows how the rankings change depending on the weight of the relative effects in ranking estimation. Ixekizumab appears to be always the most effective drug while important changes in ranking occur after w is smaller than 65%.
Conclusions:
The usefulness of conventional ranking approaches can be questionable under several settings. A treatment ranking that accounts for the potential limitations of the data and incorporates information on top of the relative effects is an important tool to decision makers.
Patient or healthcare consumer involvement:
Patients or healthcare consumers have not been involved.