OBJECTIVES: Cost-effectiveness analyses (CEAs) in indications with population heterogeneity may model subgroups individually, rather than an overall population, and weight the outcomes according to subgroup prevalence. This approach is uncommon in CEAs submitted to the National Institute for Care and Health Excellence (NICE), even when subgroups with distinct characteristics and outcomes are present. Our objective was to re-evaluate past NICE appraisals reporting subgroup analyses to assess the implications of a conducting a subgroup-weighted approach.

METHODS: NICE appraisals with recommendations published between 01-06-2021 and 20-06-2024 which reported incremental cost-effectiveness ratios (ICERs) for overall populations and complementary subgroups were included. ICERs for overall populations were recalculated using subgroup ICERs and prevalences. Recalculations were performed for original and updated manufacturer base cases and external assessment group (EAG)-preferred analyses when pairwise ICERS were in the north-east quadrant.

RESULTS: Nine appraisals were included, six with full positive recommendations and one with a restricted subgroup recommendation. ICERs from 26 subgroups ranged from -97.5% to +570% of the base case ICER and 39 weighted ICER recalculations were performed. Most (64%) recalculated ICERs were <10% different than the base case, 13 recalculations (33%) increased the ICER by >10%, and one recalculation (3%) reduced the ICER by >10%. Four (10%) recalculations increased the ICER above the ₤30,000 per quality-adjusted life-year threshold. One appraisal revised their base case to a subgroup-weighted approach in response to the EAG’s suggestion, while another provided a subgroup-weighted scenario considered relevant for decision making. Two appraisals were excluded since they submitted a subgroup-weighted approach as their base case.

CONCLUSIONS: Subgroup-weighted analyses have been used in some recent NICE appraisals to partially address heterogeneity challenges. Recalculated weighted results generally aligned with overall population results but tended to slightly increase ICERs. Among the limited appraisals reviewed, most received full population recommendations despite showing highly variable subgroup ICERs.