Applying G-methods to Account for Treatment Switching in Postmarketing Effectiveness Evaluation: A Case Study of GLP-1 Receptor Agonists for Diabetic Kidney Disease

Author(s)

Zi-Yang Peng, MS1, Huang-Yu Yang, MD, PhD2, Huang-tz Ou, PhD3.
1Institute of Clinical Pharmacy and Pharmaceutical Sciences, College of Medicine, National Cheng Kung University, Tainan, Taiwan, 2Kidney Research Center and Department of Nephrology, Chang Gung Memorial Hospital, College of Medicine, Chang Gung University, Taoyuan, Taiwan, 3Department of Pharmacy, College of Medicine, National Cheng Kung University, Tainan, Taiwan.
OBJECTIVES: Post-marketing evaluations often rely on the intention-to-treat (ITT) approach, which may inadequately reflect treatment effectiveness in the presence of frequent switching and changes common in real-world practice. G-methods, such as marginal structural methods (MSMs), account for dynamic treatment patterns and time-varying confounding but require greater analytic complexity. Despite being increasingly recommended by health technology assessment (HTA) agencies in response to the urgent need for policy-relevant effect estimates to support real-world decisions, the practical benefits and trade-offs of applying G-methods over conventional approaches remain underexplored. We compared G-method and ITT approaches using a case example evaluating the kidney effectiveness of GLP-1 receptor agonists (GLP-1RAs) versus SGLT2 inhibitors (SGLT2is) among patients with diabetic kidney disease (DKD).
METHODS: We utilized linked individual-level data from Taiwan’s National Health Insurance Research Database and Chang Gung Research Database. DKD patients initiating GLP-1RAs or SGLT2is in January/2016-June/2019 were followed for a composite kidney outcome, including estimated glomerular filtration rate less than 15 ml/min/1.73 m2, dialysis-dependent end-stage kidney disease, kidney transplantation, and all-cause death. We implemented Cox MSM (CoxMSM) analyses to account for semiannually updated treatment changes and time-varying confounders (e.g., biomarkers). Normalized and stabilized weights truncated at 90%/10%, 95%/5%, and 99%/1% tails were applied. The ITT analyses were also conducted for comparison.
RESULTS: Among 2,238 patients, the CoxMSM estimates with adjustment of truncated 90%/10%, 95%/5%, 99%/1% weights showed the hazard ratios (HRs [95% CIs]) of 0.97(0.67−1.39), 0.96(0.66−1.38), 0.81(0.55−1.17), respectively—consistently suggesting nonsignificant between-treatment difference in kidney outcomes. In contrast, the ITT analysis produced an HR (95% CI) of 0.67(0.49−0.90), indicating a potentially overstated benefit favoring GLP-1RAs.
CONCLUSIONS: By accounting for real-world patient-centered treatment changes, G-methods provide more conservative—and potentially more policy-relevant—estimates than ITT. Despite greater analytic complexity, their use can prevent generation of biased effectiveness, enhance HTA accuracy, and support equitable, evidence-based decisions in value-based healthcare systems.

Conference/Value in Health Info

2025-11, ISPOR Europe 2025, Glasgow, Scotland

Value in Health, Volume 28, Issue S2

Code

MSR30

Topic

Health Policy & Regulatory, Health Technology Assessment, Methodological & Statistical Research

Topic Subcategory

Confounding, Selection Bias Correction, Causal Inference

Disease

Diabetes/Endocrine/Metabolic Disorders (including obesity), Personalized & Precision Medicine, Urinary/Kidney Disorders

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

×