Randomised trial designs vary in moderately-to-severely active ulcerative colitis (UC). For the evaluation of maintenance phase outcomes at 1 year, older trials use a traditional parallel design (treat-through [TT] trials) and newer studies use a responder-enriched randomised withdrawal design (re-randomised responder [RRR] trials). To compare all licensed therapies, data from both study types need to be combined in a single network meta-analysis. Here, we present the methods used to enable a valid comparison across these different trial types.

To enable comparison, data from one trial type needed to be imputed to better match the other trial type. We converted outcomes from 3 TT trials into comparable outcomes from 5 RRR trials. This required less manipulation of observed data and less imputation of missing data than alternative approaches used in the literature and NICE technology appraisals and better reflected clinical practice. Observed clinical response and remission data from RRR trials were taken as is. Data from TT trials were adjusted, assuming the number of induction phase responders is a proxy for the number entering maintenance. Maintenance response from the TT trials was based on “sustained response,” thus mitigating the risk of counting maintenance phase responders who were induction phase non-responders. The number of clinical remitters at the end of follow-up as a proportion of induction phase responders was reported for active arms of some trials, but where unavailable were imputed using data from TT and RRR trials.

We have demonstrated methods for reconciling results from varied RCT designs. These conversions allowed for the synthesis of all available evidence and comparison between all licensed therapies. The results informed a subsequent cost-effectiveness model and TA and were accepted as reasonable by NICE. The pragmatic approach could be applied in other indications with similar variation in trial design.