STRUCTURAL UNCERTAINTY IN MODEL-BASED ECONOMIC EVALUATIONS: A COMPARATIVE REVIEW OF TECHNOLOGY APPRAISALS FROM NICE AND HAS OVER 2014-2026
Author(s)
Florian P. Colrat, MPH, PharmD1, Alix GREAUD, MPH2, FRANCOIS-EMERY COTTE, MA, MPH, PharmD, PhD3, Sebastien Branchoux, MSc, PhD3, Xavier ARMOIRY, DrPH, PharmD4.
1HEOR Manager, Bristol Myers Squibb, Rueil Malmaison, France, 2Université Claude Bernard Lyon 1, LYON, France, 3Bristol Myers Squibb, RUEIL MALMAISON, France, 4Université Claude Bernard Lyon 1, Lyon, France.
1HEOR Manager, Bristol Myers Squibb, Rueil Malmaison, France, 2Université Claude Bernard Lyon 1, LYON, France, 3Bristol Myers Squibb, RUEIL MALMAISON, France, 4Université Claude Bernard Lyon 1, Lyon, France.
OBJECTIVES: Structural uncertainty (uncertainty arising from the core assumptions of decision-analytic models) remains under characterized in health technology assessment (HTA) despite its critical role in shaping decision-making. This study compared how NICE and the French HAS identify and categorize structural uncertainty across paired health technology appraisals.
METHODS: A targeted literature review of all paired NICE-HAS appraisal reports covering the same intervention and indication, published up to January 2026, was conducted. Structural uncertainties were extracted and classified using Bojke et al.'s taxonomy (Data, Statistical Tools, Health States/Structure, Event, Comparator, Other). A zero-shot multilingual classifier (mDeBERTa-v3) served as quality control. Inter-agency differences in category distribution and diversity were assessed via chi-square testing and normalized Shannon entropy (Pielou's J). Pair-level categorical overlap was systematically documented.
RESULTS: From 253 appraisals, 112 matched technology pairs were constituted, primarily in oncology (n=44), onco-hematology (n=22), and rare diseases (n=18). A total of 907 structural uncertainties were identified (504 NICE; 403 HAS). The most frequent categories were uncertainties around statistical Tools (23%), Data (21%), and Other (20%). Despite comparable diversity (Pielou's J: 48% vs. 53%), category distributions differed significantly (χ²=30.02, p<0.001). Only 3 of 101 pairs shared identical categories whereas 21 shared none. The "Other" category increased markedly in HAS appraisals following the 2020 methodological update (7% to 20%, p<0.001).
CONCLUSIONS: NICE and HAS diverge qualitatively rather than quantitatively in framing structural uncertainty. The expanding "Other" category signals that Bojke's taxonomy is limited and no longer fully captures the structural complexity of contemporary health economic models. A collaborative, empirically grounded update should be prioritized by the HTA community.
METHODS: A targeted literature review of all paired NICE-HAS appraisal reports covering the same intervention and indication, published up to January 2026, was conducted. Structural uncertainties were extracted and classified using Bojke et al.'s taxonomy (Data, Statistical Tools, Health States/Structure, Event, Comparator, Other). A zero-shot multilingual classifier (mDeBERTa-v3) served as quality control. Inter-agency differences in category distribution and diversity were assessed via chi-square testing and normalized Shannon entropy (Pielou's J). Pair-level categorical overlap was systematically documented.
RESULTS: From 253 appraisals, 112 matched technology pairs were constituted, primarily in oncology (n=44), onco-hematology (n=22), and rare diseases (n=18). A total of 907 structural uncertainties were identified (504 NICE; 403 HAS). The most frequent categories were uncertainties around statistical Tools (23%), Data (21%), and Other (20%). Despite comparable diversity (Pielou's J: 48% vs. 53%), category distributions differed significantly (χ²=30.02, p<0.001). Only 3 of 101 pairs shared identical categories whereas 21 shared none. The "Other" category increased markedly in HAS appraisals following the 2020 methodological update (7% to 20%, p<0.001).
CONCLUSIONS: NICE and HAS diverge qualitatively rather than quantitatively in framing structural uncertainty. The expanding "Other" category signals that Bojke's taxonomy is limited and no longer fully captures the structural complexity of contemporary health economic models. A collaborative, empirically grounded update should be prioritized by the HTA community.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR46
Topic
Economic Evaluation, Health Technology Assessment, Methodological & Statistical Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas