OBJECTIVES

: In the absence of head-to-head comparative effectiveness data often the next best approach to take, when trying to compare treatments, is an indirect treatment comparison of individual patient data with aggregate level data. Matching adjusted indirect comparison (MAIC) methods are extremely useful in this setting, as they reduces baseline imbalances between studies, particularly upon patient characteristics that are confounded with treatment. The standard approach when conducting MAIC is that proposed by Signorovitch et al. (2010). However, there are newer, and potentially better, methods available. Our objective was to compare and contrast the strengths and weaknesses of various MAIC approaches.

METHODS

: Three different MAIC methods (Signorovitch, Entropy Balancing, Polynomial Weighting) were compared using multiple phase 3 RCTs conducted in Diabetic Retinal Edema. The matching ability of each method was assessed, alongside its ability to avoid large weights (i.e. avoiding high leverage), and maximise effective same size (ESS). Each method's overall ease of use and impact upon estimates of treatment effectiveness were also evaluated.

RESULTS

: All methods were able to precisely match the aggregate level data. However, the Entropy Balancing and Polynomial Weighting both outperformed the Signorovitch method in terms of having the lowest maximum weights and maximising the ESS. The Entropy Balancing method was arguably the most challenging to implement, whilst the Signorovitch method the least. The Polynomial Weighting method appears to provide the greatest flexibility to the user.

CONCLUSIONS

: Whilst the Signorovitch method has become almost synonymous with MAIC, the Entropy Balancing and Polynomial Weighting methods offer potentially superior performance. These new methods should provide less biased and more precise estimates of comparative effectiveness, leading to better decision making by regulators and payers.