Presentation (CTI)

OBJECTIVES: The estimand framework has been adopted by regulators globally to promote clear pre-specification of target treatment effects for pivotal randomized clinical trials (RCTs). This has resulted in movement away from conventional intent-to-treat and per-protocol analyses to explicit specifications of intercurrent events (ICEs) and their strategies. Current feasibility assessment practices of network meta-analyses (NMAs) of RCTs focus on alignment of the PICO (population, intervention, comparator, and outcome) criteria, network connectivity, and clinical/methodological similarities. Limited guidance exists on incorporating the estimand framework into NMA planning. We conducted a scoping review to examine how estimands can be integrated into NMA feasibility assessments.
METHODS: Following the JBI Scoping Review Methodology guidance, we conducted a scoping review to map key estimand considerations for evidence synthesis. We searched peer-reviewed publications and methodological guidance from HTA bodies using the terms “Estimand” AND (“Evidence Synthesis” OR “Meta-Analysis”) to contextualize feasibility considerations of NMAs within the estimand framework.
RESULTS: Our search yielded five peer-reviewed publications and three technical guidelines from the CDA, the ICH, and the HTA Coordination Group. They discussed methodological concerns about bias introduced by pooling trials with differing ICE strategies or mismatched summary measures, and challenges in determining the relevance of ICEs. These documents also highlighted trade-offs between network connectivity and estimand concordance; and called for transparent reporting including specifications of key ICEs during feasibility assessments.
CONCLUSIONS: Current practices of feasibility assessments are limited and do not ensure estimand concordance. Beyond the alignment of PICO criteria and network connectivity, feasibility assessments should be extended to consider ICE handling strategies; pooling of marginal versus conditional estimands; and non-collapsible versus collapsible summary measures as potential sources of estimand discordance. Pooling estimand-discordant trials for network connectivity poses important risks for biased quantitative syntheses. Transparent reporting with targeted sensitivity analyses offer practical ways to integrate estimand considerations into evidence synthesis.