OBJECTIVES: Due to limitations of phase III trial designs for moderate-to-severe psoriasis treatments, current NICE submission models define the proportions of patients remaining on treatment over time using a combination of short-term PASI-75 response (derived from network meta-analysis [NMA]) and constant non-treatment specific probabilities derived from published real-world evidence (RWE). However, drug survival (DS) extrapolations based on this approach may differ substantially from parametric extrapolations derived directly from the RWE. This research aims to quantify these differences, and discuss potential reasons and implications on cost-effectiveness analyses.

METHODS: DS of four biologic treatments (adalimumab, etanercept, infliximab, ustekinumab) was modelled over 10 years using: (1) short-term PASI-75 response derived from a published NMA combined with constant annual discontinuation probability of 18.7% from published RWE; (2/3) standard parametric survival models fitted directly to UK adalimumab DS data for first- and second-line biologic treatment (Warren 2015 [approach 2] and Iskandar 2018 [approach 3], respectively). Remaining interventions were modeled using adjusted hazard ratios derived from these studies.

RESULTS: 10-year DS estimates of 14.9%, 10.8%, 17.4% and 16.4% were produced for adalimumab, etanercept, infliximab and ustekinumab using approach 1 compared to 24.9%/30.7%, 10.4%/14.6%, 11.4%/15.9%, and 51.3%/56.7% for approaches 2/3. Largest differences in extrapolations were observed for ustekinumab followed by adalimumab, with approach 1 underpredicting compared to approaches 2/3 after the induction period. Approach 1 overpredicted DS for infliximab before converging with approaches 2/3 closer to 10 years, while etanercept extrapolations were most similar.

CONCLUSIONS: Differences observed in DS extrapolations may result in substantially different predictions of treatment line distributions over time, and therefore calculation of the most cost-effective treatment sequences. Further research should be conducted into comparisons of overall DS for newer biologic therapies and compared against NICE submission model drug survival extrapolations, and methods for more accurate DS prediction from short-term trial data.