OBJECTIVES: Decisions on the cost-effectiveness of new treatments require estimates of treatment effects on generic Health related Quality of Life (HRQoL) outcomes. However, it is often the case that studies report disease specific scales, rather than the generic measure suitable for decision making, and this is especially the case in rare diseases. When there are multiple studies available, evidence synthesis of HRQoL outcomes can aid decision makers by fully reflecting the available evidence and reducing uncertainty. We aimed to explore the use of meta-analytic models for surrogate endpoint evaluation to obtain estimates of generic measures of HRQoL from disease specific measures.

METHODS: We illustrate the methods using summary data from 25 RCTs reporting effectiveness of treatments in Ankylosing Spondylitis (AS) and non-radiographic axial Spondylarthritis (nr-axSpA). Target generic measures of HRQoL are mapped by predicting the missing treatment effects on the generic outcomes. We compare the following models: bivariate random meta-analysis (BRMA), Daniels-Hughes model (D&H), and BRMA in product normal formulation (BRMA PNF). The predicted and observed treatment effects on the generic outcomes are pooled and evaluated.

RESULTS: We found associations between the generic 36-Item Short Form Survey Physical Component Summary (SF36-PCS) and the disease specific Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), the Mental Component Summary (SF36-MCS) and Functioning Index (BASFI). Cross validation confirmed the models’ prediction intervals to contain observed estimates for at least 90% of included studies. The treatment effect on BASDAI predicted the treatment effect on the SF36-PCS in all of the studies. BRMA PNF estimates were more precise than D&H estimates. Pooling estimates with BRMA PNF reduced uncertainty by 38% on SF36-PCS when compared to a univariate model.

CONCLUSIONS: Methods for synthesis of surrogate endpoints predicted and increased precision in estimates of the generic HRQoL needed for HTA. BRMA PNF gave the most precise estimates.