Title: Personalized health outcome predictions and economic evaluations using individual participant data from multiple trials
Abstract: Treatment effects and cost-effectiveness may vary across patients, but conventional analyses often provide average estimates. We developed a framework combining risk modelling, network meta-analysis (NMA), decision curve analysis, and cost-effectiveness analysis to support individualized treatment decisions.
Methods: We developed a two-stage prediction model using individual participant data from multiple randomised and observational studies. First, we develop a prognostic model to estimate baseline risk; second, this risk score was used as the sole effect modifier in NMA to estimate treatment effects. We extended decision curve analysis to multiple treatment options and NMA evidence by defining risk-difference thresholds and estimating net benefit across threshold combinations. We estimated incremental cost-effectiveness ratios (ICERs) and net monetary benefits (NMBs) as functions of baseline risk. We applied the framework to three trials including 3 590 patients with relapsing-remitting multiple sclerosis, comparing natalizumab, dimethyl fumarate, glatiramer acetate, and placebo. Data from a Swiss registry were also available.
Results: Baseline risk modified treatment effects: age and disability status predicted relapse risk, influencing expected treatment benefit. Natalizumab minimized 2-year relapse risk among high-risk patients, whereas dimethyl fumarate was preferable among low-risk patients. Personalized treatment according to the prediction model performed better than, or similarly to, one-size-fits-all strategies, although advantages varied across treatment thresholds. Cost-effectiveness varied by baseline risk: the ICER for dimethyl fumarate versus glatiramer acetate increased from approximately CHF 50,000/QALY at 10% risk to CHF 240,000/QALY at 80%; at a CHF 100,000/QALY threshold, glatiramer acetate was favoured above 77% risk.
Conclusions: Integrating prognostic modelling with NMA, decision curve analysis, and economic evaluation can identify meaningful treatment heterogeneity and inform individualized treatment decisions.
Biography: Prof. Georgia Salanti leads the Biostatistics and Research Methodology Group at the Institute of Social and Preventive Medicine, University of Bern. Her research focuses on developing and applying methods for evidence synthesis, including network meta-analysis, living systematic reviews, and the responsible use of artificial intelligence in systematic review processes. She serves as the Methods Lead of the Global Alliance for Living Evidence on Anxiety, Depression and Psychosis (GALENOS), contributing to the development of innovative approaches for generating and maintaining high-quality evidence in mental health. She collaborates with international organizations and multidisciplinary research teams to improve the transparency, rigor, and efficiency of healthcare evidence synthesis. Her work aims to strengthen the translation of methodological advances into practice to support clinical, public health, and policy decision-making.
________________________________________________________________________________
Microsoft Teams meeting
Join: https://teams.microsoft.com/meet/367577605424503?p=GAgKYidbD2Q8syV6dq
Meeting ID: 367 577 605 424 503
Passcode: sA9nb2TJ