Leveraging a Bayesian Approach in a Comparative Effectiveness Trial of Major Adverse Cardiovascular Events

Cara T Lwin,1 Christianne L Roumie,2– 4 Robert Alan Greevy Jr,1,2,5 Cole Beck,1,2 Kathryn Diane Snyder,3 Amber J Hackstadt1,2,5 1Department of Biostatistics, Vanderbilt University School of Medicine, Nashville, TN, USA; 2Geriatric Research Education Clinical Center (GRECC) and the VETWISE-LHS Center of Innovation, Veteran Administration Tennessee Valley VA Health Care System, Nashville, TN, USA; 3Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, USA; 4Department of Health Policy, Vanderbilt University Medical Center, Nashville, TN, USA; 5Center for Health Services Research, Vanderbilt University Medical Center, Nashville, TN, USACorrespondence: Amber J Hackstadt, Vanderbilt School of Medicine, 2525 West End Avenue, Suite 1100, Nashville, TN, 37203-1741, USA, Email amber.hackstadt@vumc.orgPurpose: We applied a Bayesian approach to further investigate the association of sodium–glucose cotransporter-2 inhibitors (SGLT2i) with the composite outcome of Major Adverse Cardiovascular Event and Heart Failure hospitalization (MACE+HF) and its individual components leveraging the ability of a Bayesian approach to incorporate prior clinical information and to make probability statements about the parameters.Methods: We use a Bayesian time-to-event model, where the covariates are directly modeled in the hazard function. Following propensity score matching, we fit three Bayesian models; one with a relatively flat, normal prior on the SGLT2i coefficient (Uninformative) and 2 with informative priors from a meta-analysis (based on a cohort with no history of cardiovascular disease [No CVD] and cohorts with a history of CVD [CVD]). We estimate the posterior distribution for the hazard ratio (HR) using a Hamiltonian Monte Carlo algorithm. It allows us to estimate the probability of a meaningful protective association (HR < 0.90) in addition to point and interval estimates.Results: The posterior means and 95% credible intervals for the HR suggested a protective association for SGLT2i versus dipeptidyl peptidase 4 inhibitors (DPP4i) for the MACE+HF outcome: No CVD: 0.82 (0.68, 0.96), CVD: 0.82 (0.71, 0.94), and Uninformative: 0.79 (0.65, 0.94). The probability of a meaningful protective association for the No CVD, CVD, and Uninformative priors were 88%, 92%, and 93%, respectively. The probability of a meaningful protective association for the HF hospitalization, CVD hospitalization and CVD death components of MACE+HF were 95%, 67%, and 93%, respectively.Conclusion: The Bayesian analysis allowed for the incorporation of prior information via an informative prior and further investigation of the association between SGLT2 and the components of the MACE+HF composite outcome. It allowed for the calculation of an easily interpretable summary measure, the probability of a meaningful protective association.Keywords: time-to-event analysis, bayesian inference, hamiltonian monte carlo, propensity score methods, cardiovascular disease, diabetes mellitus

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.