Regulatory-Grade Statistical Methodology for Comparative Effectiveness Analysis of Triple vs. Dual Antihypertensive Therapy Using Real-World Data
Author(s)
Sophie BIROT, PhD, Caroline Apert-Maarek, PhD, Alfonso Sartorius, MD, Khaoula Aroui, MSc, Céline Darricarrère, MSc.
Institut de Recherches Internationales Servier, Gif sur Yvette, France.
Institut de Recherches Internationales Servier, Gif sur Yvette, France.
OBJECTIVES: This study aimed to compare the effectiveness of triple therapy (bisoprolol, perindopril, and amlodipine) versus dual therapy (two among the three drugs) in reducing systolic blood pressure (SBP) at 8 weeks among uncontrolled hypertensive patients in a UK real-world setting. Designed to emulate a target trial, this study applied regulatory-grade statistical methods to rigorously address bias and confounding in support of drug registration.
METHODS: A retrospective observational cohort study was conducted using CPRD Aurum database (2005-2020). Following the target trial emulation framework, the primary endpoint was the change in SBP from baseline to 8 weeks. A propensity score matching approach was employed, with a 1:1 matching ratio, with covariates (i.e comorborbities, ethnicity, treatment adherence) identified as potential confounders from a Directed Acyclic Graph. The estimands framework was used to infer causal relationship between the model estimate and the study outcome. The primary analysis utilized a Mixed Model for Repeated Measures, including time as a continuous covariate and a treatment-by-time interaction term. Missing post-baseline SBP values due to intercurrent events were addressed using a reference-based imputation method (Baseline Mean Carried Forward), aligned with a hypothetical strategy assuming no treatment effect after discontinuation. To evaluate residual confounding, quantitative bias analyses were conducted: (1) tipping point analysis assessed the influence of a hypothetical unmeasured confounder; (2) a negative control outcome analysis examined associations with an outcome plausibly unrelated to treatment exposure.
RESULTS: The statistical strategy was presented to a EU regulatory agency, where it received favorable scientific advice in February 2024 regarding the methodological rigor and applicability of real-world data for regulatory purposes.
CONCLUSIONS: The approach aims to provide reliable evidence on additional SBP reduction achieved with triple therapy compared to dual therapy in routine clinical practice.
METHODS: A retrospective observational cohort study was conducted using CPRD Aurum database (2005-2020). Following the target trial emulation framework, the primary endpoint was the change in SBP from baseline to 8 weeks. A propensity score matching approach was employed, with a 1:1 matching ratio, with covariates (i.e comorborbities, ethnicity, treatment adherence) identified as potential confounders from a Directed Acyclic Graph. The estimands framework was used to infer causal relationship between the model estimate and the study outcome. The primary analysis utilized a Mixed Model for Repeated Measures, including time as a continuous covariate and a treatment-by-time interaction term. Missing post-baseline SBP values due to intercurrent events were addressed using a reference-based imputation method (Baseline Mean Carried Forward), aligned with a hypothetical strategy assuming no treatment effect after discontinuation. To evaluate residual confounding, quantitative bias analyses were conducted: (1) tipping point analysis assessed the influence of a hypothetical unmeasured confounder; (2) a negative control outcome analysis examined associations with an outcome plausibly unrelated to treatment exposure.
RESULTS: The statistical strategy was presented to a EU regulatory agency, where it received favorable scientific advice in February 2024 regarding the methodological rigor and applicability of real-world data for regulatory purposes.
CONCLUSIONS: The approach aims to provide reliable evidence on additional SBP reduction achieved with triple therapy compared to dual therapy in routine clinical practice.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR181
Topic
Epidemiology & Public Health, Methodological & Statistical Research, Real World Data & Information Systems
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
Cardiovascular Disorders (including MI, Stroke, Circulatory)