METHODOLOGICAL LEARNINGS FROM AN INDIRECT TREATMENT COMPARISON WITH LONGITUDINAL DATA: A CASE-STUDY

Author(s)

Matthew Knowles, MSc, MMath1, Charlotte Patterson, MBChB2.
1Astellas Pharma Europe Ltd, Addlestone, United Kingdom, 2Paediatrics, York and Scarborough NHS Foundation Teaching Hospitals, Scarborough, United Kingdom.
OBJECTIVES: Longitudinal data is common in the reporting of trial results in type 1 diabetes mellitus (T1DM) for key outcomes such as glycated haemoglobin (HbA1c). Longitudinal data presents a problem for traditional indirect treatment comparison (ITC) methods such as Network Meta-Analysis (NMA), as there is no way to accurately model the intra-subject correlations between observations. The primary objective was to conduct an ITC of paediatric T1DM trials using an innovative method.
METHODS: Time-Course Model-Based NMA (TCMBNMA) fits a standard NMA model to time-course parameters from a specified time-course model, suitably chosen based on the available data. Since the selection of this time-course model may incur significant bias, the model selection was based on a systematic evaluation of the Deviation Information Criterion (DIC) score for multiple models. The three best-fitting (considered as lowest DIC score) models were evaluated in terms of model fitting using trace plots from the Bayesian sampling algorithm, and clinician input. Both Fixed Effect and Random Effects models were considered for each time-course function. Due to the sparse availability of evidence after 25 weeks, clinical interpretation was based only on predictions up to week 25.
RESULTS: For the case study, a total of seven studies were identified in the literature review. Dapagliflozin and metformin demonstrated superior efficacy as insulin-adjuvant treatment for the management of T1DM in a paediatric population when compared to insulin monotherapy.
CONCLUSIONS: The TCMBNMA model was easy to implement in R, and presents an opportunity to comprehensively synthesise longitudinal evidence. A key learning was that systematic selection of the time-course model was essential for obtaining clinically meaningful results. In addition, while TCMBNMA models allow extraction of a predicted treatment effect up to any time within the maximum follow-up the model was fit to, results become unstable after timepoints for which data becomes sparse for any treatments in the network.

Conference/Value in Health Info

2026-11, ISPOR Europe 2026, Vienna, Austria

Value in Health, Volume 29, Issue 12S

Code

SA76

Topic

Clinical Outcomes, Methodological & Statistical Research, Study Approaches

Topic Subcategory

Meta-Analysis & Indirect Comparisons

Disease

Diabetes/Endocrine/Metabolic Disorders (including obesity), Pediatrics

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×