PNS318 NETWORK META-ANALYSES COMBINING RANDOMIZED AND NON-RANDOMIZED EVIDENCE; A NICE HTA REVIEW

Abstract

Randomized controlled trials (RCTs) are considered the most reliable sources of information for evidence synthesis in network meta-analyses (NMAs). However, interest in including non-randomized studies (NRSs) in NMAs is growing. NRSs may complement evidence provided by RCTs and they can be used to connect disconnected evidence networks of RCTs. Statistical techniques including Bayesian hierarchical models have been developed to account for potential bias from including RCTs and NRSs in NMAs. This review aims to assess NICE technology appraisals which included NMAs combining RCTs and NRSs and to evaluate how the risk of bias has been addressed in these NMAs. A NICE health technology assessment (HTA) review was conducted on submissions appraised between 2001-2019. NMAs combining both RCT and NRS evidence were included. NMAs constructed using indirect treatment comparison techniques to connect a single-arm pivotal trial with other comparators are excluded if they don’t include RCT evidence in the network. Two relevant submissions were identified. Unavailability of RCTs was the reason for including NRSs in the NMAs. The sources of NRSs were evidence from other indirect treatment comparisons and observational studies. Risk of bias for RCTs was evaluated using CRD's guidance for undertaking reviews in health care and the NICE user guide. For the NRSs evidence; no formal quality assessment was done in one submission, but the second submission used ROBINS-I tool. In both submissions, specific methodologies for adjusting the bias that is introduced by combining different study designs in NMAs were not implemented. Inclusion of RCTs and NRSs in NMAs submitted to NICE is limited and differences in the risk of bias was not well addressed in the included NMAs. Available methodologies for incorporating RCTs and NRSs in NMAs can benefit future submissions which would require combining RCT and NRS evidence in NMAs.