PROJECTING PCSK9 INHIBITOR TRIAL EFFECTS TO A US ASCVD TARGET POPULATION WITHOUT INDIVIDUAL PATIENT DATA
Author(s)
Lingyao Sun, PhD1, liuqing zhu, Undergraduate1, fenghua shen, BS1, lingli huang, Undergraduate1, Lei Tian, MSc2.
1China Pharmaceutical University, Nanjing, China, 2China Pharmaceutical University, Nanjing, China.
1China Pharmaceutical University, Nanjing, China, 2China Pharmaceutical University, Nanjing, China.
OBJECTIVES: In health technology assessment, treatment-effect estimates should reflect the decision-relevant population, yet trial populations often differ in effect-modifying covariates. Population-adjustment methods for indirect comparisons usually require individual patient data, which are often unavailable. We applied externally anchored covariate completion to project published trial subgroup evidence to a prespecified US target population using external real-world covariate information, without individual patient data from either trial.
METHODS: We conducted a comparative-effectiveness modeling study using two pivotal PCSK9 inhibitor trials: FOURIER for evolocumab and ODYSSEY OUTCOMES for alirocumab. Both trials share a placebo comparator. Subgroup hazard ratios and covariate summaries were extracted from published reports. Six population-based longitudinal cohorts from Europe, the United States, China, India, and Mexico provided the external covariate structure used to complete unreported subgroup covariate profiles. Treatment effects were projected to the US National Health and Nutrition Examination Survey 2013-2018 target population, defined by age 65 years or older, male sex, White race, stroke history, myocardial infarction history, and diabetes. Sensitivity analyses varied matching tolerance and substituted NHANES-derived correlation structures.
RESULTS: Original trial hazard ratios for major adverse cardiovascular events were approximately 0.85 for both drugs versus placebo. After projection to NHANES, estimates were 0.82 for evolocumab, with a 95% confidence interval of 0.77-0.86, and 0.83 for alirocumab, with a 95% confidence interval of 0.75-0.91. These modestly lower estimates were consistent with the older age and higher comorbidity burden of NHANES. Sensitivity analyses produced consistent results.
CONCLUSIONS: Projecting published subgroup trial evidence to a prespecified US target population yielded stable, population-targeted hazard ratio estimates without individual patient data. This approach makes population-relevance assumptions explicit and provides transparent, auditable, population-tailored evidence for health technology assessment.
METHODS: We conducted a comparative-effectiveness modeling study using two pivotal PCSK9 inhibitor trials: FOURIER for evolocumab and ODYSSEY OUTCOMES for alirocumab. Both trials share a placebo comparator. Subgroup hazard ratios and covariate summaries were extracted from published reports. Six population-based longitudinal cohorts from Europe, the United States, China, India, and Mexico provided the external covariate structure used to complete unreported subgroup covariate profiles. Treatment effects were projected to the US National Health and Nutrition Examination Survey 2013-2018 target population, defined by age 65 years or older, male sex, White race, stroke history, myocardial infarction history, and diabetes. Sensitivity analyses varied matching tolerance and substituted NHANES-derived correlation structures.
RESULTS: Original trial hazard ratios for major adverse cardiovascular events were approximately 0.85 for both drugs versus placebo. After projection to NHANES, estimates were 0.82 for evolocumab, with a 95% confidence interval of 0.77-0.86, and 0.83 for alirocumab, with a 95% confidence interval of 0.75-0.91. These modestly lower estimates were consistent with the older age and higher comorbidity burden of NHANES. Sensitivity analyses produced consistent results.
CONCLUSIONS: Projecting published subgroup trial evidence to a prespecified US target population yielded stable, population-targeted hazard ratio estimates without individual patient data. This approach makes population-relevance assumptions explicit and provides transparent, auditable, population-tailored evidence for health technology assessment.
Conference/Value in Health Info
2026-09, ISPOR Asia Pacific 2026, Bangkok, Thailand
Value in Health, Volume 55, Issue S1
Code
MSR1
Topic
Methodological & Statistical Research
Topic Subcategory
Confounding, Selection Bias Correction, Causal Inference
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory)