OBJECTIVES: Re-randomization (RR) and treat-through (TT) are both recognized and recommended study designs to assess outcomes for treatments with an induction and subsequent maintenance phase such as Crohn’s Disease (CD). Heterogeneity arises when network meta analyses (NMA) include mixed trial designs. This study explores the sources of uncertainty affecting NMAs involving RR and TT trial designs and potential methodological solutions.

METHODS: A review identified indications and recent HTA submissions where TT and RR trials are used. This literature was used to explore key sources of uncertainty and methodological approaches used to address these challenges.

RESULTS: The identified TT studies do not include a comparable placebo arm during maintenance, requiring the use of an active comparator (previously studied in a RR setting) to create a connected network. Once the network is created, individual patient data (IPD) are needed for TT studies to restrict the population to induction responders, to replicate a RR design. When no IPD is available strong assumptions are needed. To mitigate potential confounding from breaking randomization when limiting to responders only, propensity score matching is recommended to estimate the probability of being in each treatment arm. An additional challenge given the nature of RR studies is that only responders in the induction active comparator arm enter the maintenance phase. Those re-randomized to receive placebo, the structure of which may differ between studies, during maintenance may experience a potential carry over effect from their respective active induction treatments, adding a further source of heterogeneity across the studies.

CONCLUSIONS: TT and RR trials differ in their purposes and the treatment goals assessed. As a result they generate maintenance outcomes that are not comparable without prior alignment. Mixed design NMAs currently utilize different methodologies of varying robustness. Methodological guidelines are needed, and detailed reporting of trial data is encouraged.