Key Features of Economic Models Used in Submissions to NICE in Dermatology Indications

OBJECTIVES: To review the key features of economic models used in submissions to the National Institute for Health and Care Excellence (NICE) in dermatology indications.

METHODS: The NICE website was searched on 24 April 2024 to identify technology appraisals (TAs) reporting cost-effectiveness analyses in dermatology.

RESULTS: Searches identified 33 TAs, 9 of which were excluded (4 cost comparisons, 4 not dermatology, 1 in development). The remaining 24 TAs were included.

The structures of 13 (all psoriasis [PsO]) of these 24 models were based on the York Model (TA103), which comprises short-term decision-tree and long-term Markov components. Five of 24 had a similar structure to the York Model without explicit citation (3 atopic dermatitis [AD], 1 alopecia areata [AA], 1 prurigo nodularis). In these 18 models, response was based on condition-specific measures, and differences included the response definition, timing of response assessment, treatment sequencing, and short-term modelling approach. The remaining 6 of 24 used alternative structures: 2 were individual-level (hand eczema and urticaria) and 4 were cohort-level (2 hidradenitis suppurativa [HS], 1 AA, 1 AD) models.

Convergence Diagnosis and Output Analysis (CODA) was used to preserve the relationship between response covariates in probabilistic analyses in 10 TAs (8 PsO, 1 AD, 1 HS). Waning of treatment effect was implemented in 3 AD models. All but 1 (in AA, citing lack of quality-of-life improvement and unacceptable level of uncertainty around cost-effectiveness outcomes) of the included TAs received a recommendation. The endpoint and threshold used to define response were the main concerns of evidence assessment groups.

CONCLUSIONS: Economic models in 24 NICE submissions in dermatology were reviewed. Most models were based on or similar to the York Model. Almost half used CODA, and a minority implemented waning of treatment effect. Endpoints and their use were the primary concern. Recommendations were not related to the model features.