Predicting Long Term Cancer Survival for Health Technology Assessment: A Multinational Cohort Study
Author(s)
Claire R1, Dietz J2, Koblbauer I3, Koh J4, Elvidge J5, López-Sánchez I6, Golozar A7, Pérez-Crespo L6, Palomar Cros A6, Burn E8, Robinson A8, Delmestri A9, Corby G8, Alcade Herraiz M8, Català Sabaté M8, Man WY8, Chen X8, Mayer MA10, Ramirez Anguita JM10, Leis Machin A10, Symmers N11, Vallet M11, McLean C11, Hall PS11, Enerly E12, Prinsen P13, Evers J14, Oja M15, Kolde R15, Fey E16, Taveira-Gomes T17, Fournier E18, Moreno Conde A19, Kauko T20, Marcos Gragera R21, Mosseveld M22, Verhamme K22, Duarte-Salles T6, Dawoud D23, Newby D8
1National Institute for Health and Care Excellence, Manchester, Manchester, UK, 2National Institute for Health and Care Excellence (NICE), Bristol, BST, UK, 3University of Oxford, Abingdon, OXF, UK, 4National Institute for Health and Care Excellence, Manchester, LAN, UK, 5National Institute for Health and Care Excellence (NICE), Manchester, UK, 6Fundació Institut Universitari per a la recerca a l'Atenció Primària de Salut Jordi Gol i Gurina, Barcelona, Spain, 7Odysseus Data Services, Inc., New York, NY, USA, 8University of Oxford, Oxford, Oxfordshire, UK, 9University of Oxford, Oxford, UK, 10Hospital del Mar, Barcelona, Catalonia, Spain, 11University of Edinburgh, Edinburgh, UK, 12Cancer Registry of Norway, Oslo, Oslo, Norway, 13Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht, Utrecht, Netherlands, 14IKNL, Utrecht, Utrecht, Netherlands, 15University in Tartu, Tartu, Tartu, Estonia, 16HUS Helsinki University Hospital, Helsinki, Helsinki, Finland, 17MTG Research and Development Lab, Porto, Porto, Portugal, 18Université de Genève, Geneva, Geneva, Switzerland, 19Hospital Universitario Virgen Macarena, Seville, Seville, Spain, 20The wellbeing services county of Southwest Finland (Varha), Turku, Western Finland, Finland, 21Catalan Institute of Oncology, Girona, Girona, Spain, 22Erasmus MC, Rotterdam, Netherlands, 23National Institute for Health and Care Excellence (NICE), London, LON, UK
Presentation Documents
OBJECTIVES: The European Health Data and Evidence Network (EHDEN) is a federated network of real-world data (RWD) across Europe, standardized to the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM). Using data from EHDEN data partners, this study aimed to determine the predictive accuracy of standard and flexible parametric spline models for modelling long-term cancer survival and assess their suitability for use in Health Technology Assessments (HTA).
METHODS: Fifteen databases from eight European countries (UK, Norway, Switzerland, Spain, Netherlands, Finland, Portugal, Estonia) participated in this study. Patients aged 18 years and older with a primary diagnosis of breast, pancreas, prostate, colorectal, lung, stomach, liver, or head and neck cancer from 2000 to 2019 were followed from diagnosis until death, database exit, or study end. Six standard parametric (generalized gamma, log-normal, log-logistic, exponential, Weibull, Gompertz) and two flexible parametric models (Restricted Cubic Splines with 1 and 3 knots) were fitted. Model fit was assessed using the Akaike Information Criterion (AIC), and predictive performance was evaluated by comparing the restricted mean survival time (RMST) between Kaplan-Meier and parametric model estimates at 10 years.
RESULTS: Over 1.6 million cancer patients were included in the study. Baseline characteristics were similar across databases and cancer types. Overall, flexible spline models generally provided the best fit and predictive performance of long-term survival. Among standard models, the log-normal, log-logistic, and generalized gamma performed best, while exponential, Weibull, and Gompertz models had poorer predictive performance.
CONCLUSIONS: Flexible parametric models generally offered superior fit to the observed Kaplan-Meier survival data, though assessment should be on a case-by-case basis per cancer type. These findings can help HTA agencies reduce uncertainty in decision-making by informing suitability of models for capturing survival. Further research should explore generalizability by cancer stage and data source and consider treatment-specific and time-limited distributions.
Conference/Value in Health Info
Value in Health, Volume 27, Issue 12, S2 (December 2024)
Code
RWD110
Topic
Epidemiology & Public Health, Real World Data & Information Systems
Topic Subcategory
Distributed Data & Research Networks
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Oncology