OBJECTIVES:

Multi-level network meta-regression (ML-NMR) has been recently introduced as a new methodology for indirect treatment comparisons in situations where individual patient data (IPD) are available for some but not all trials. ML-NMR allows accounting for effect modifiers and has the potential of treatment comparisons in target populations different from the study populations. We present a case study within the setting of large-scale randomized trials, highlighting how typical issues such as missing data can be dealt with, and how published subgroup analyses can be incorporated to provide additional information to inform population adjustment of treatment effects.

METHODS:

ML-NMR was applied in a network of three RCTs on high-doses of GLP-1 RAs (semaglutide, dulaglutide) for the treatment of type 2 diabetes (two studies with IPD (SUSTAIN 7 and SUSTAIN FORTE), one study with aggregate data (AgD) only (AWARD 11)). Missing data was imputed using multiple imputation and the multiply imputed data sets were incorporated into the ML-NMR model. Published effect estimates for subgroups of the AgD trial were used to further explore effect modification and enable robust population adjustment.

RESULTS:

CONCLUSIONS:

From an industry perspective, the ability to more fully leverage IPD in indirect treatment comparisons is important. This case study suggests that ML-NMR is well suited for application in a typical production setting, offering an accessible yet flexible option for coherently accounting for the totality of evidence available across both IPD and AgD.