Scholarlink

<p>The American Statistical Association (ASA) has cautioned against overreliance on p-values, noting that rigid thresholds (e.g., <i>p</i> &lt; 0.05) can obscure statistical uncertainty and clinical meaning. This study illustrates how Bayesian methods can enrich trial interpretation by estimating the probability of treatment benefit. We reanalyzed all-cause mortality outcomes from two randomized trials—EMPULSE (empagliflozin in acute heart failure) and DanGer Shock (microaxial flow pump in cardiogenic shock)—using Bayesian hierarchical random-effects models with reference and data-derived priors. For EMPULSE, posterior probabilities of mortality benefit were high (RR &lt; 1: 90%–99%; RR &lt; 0.85: 72%–86%). In DanGer Shock, they were lower and more uncertain (RR &lt; 1: 76%–98%; RR &lt; 0.85: 15%–72%). Although both trials had similar frequentist <i>p</i>-values (0.04 and 0.05), Bayesian analysis revealed differing levels of certainty. These results highlight the value of Bayesian approaches in providing more nuanced, decision-relevant insights, particularly when trial results lie near conventional significance thresholds.</p> Graphical Abstract <p>Posterior Probabilities of Mortality in (<b>a</b>) EMPULSE and (<b>b</b>) DanGer Shock Trials Using Reference Priors.&#xa0;<i>RR</i> = Relative risk</p> <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bayesian Reanalysis of Mortality Outcomes in Cardiovascular Trials: Addressing Limitations of Traditional Significance Testing

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • Gisèle Nakhlé,
  • Jean-Claude Tardif,
  • Anick Dubois,
  • Jacques LeLorier

摘要

The American Statistical Association (ASA) has cautioned against overreliance on p-values, noting that rigid thresholds (e.g., p < 0.05) can obscure statistical uncertainty and clinical meaning. This study illustrates how Bayesian methods can enrich trial interpretation by estimating the probability of treatment benefit. We reanalyzed all-cause mortality outcomes from two randomized trials—EMPULSE (empagliflozin in acute heart failure) and DanGer Shock (microaxial flow pump in cardiogenic shock)—using Bayesian hierarchical random-effects models with reference and data-derived priors. For EMPULSE, posterior probabilities of mortality benefit were high (RR < 1: 90%–99%; RR < 0.85: 72%–86%). In DanGer Shock, they were lower and more uncertain (RR < 1: 76%–98%; RR < 0.85: 15%–72%). Although both trials had similar frequentist p-values (0.04 and 0.05), Bayesian analysis revealed differing levels of certainty. These results highlight the value of Bayesian approaches in providing more nuanced, decision-relevant insights, particularly when trial results lie near conventional significance thresholds.

Graphical Abstract

Posterior Probabilities of Mortality in (a) EMPULSE and (b) DanGer Shock Trials Using Reference Priors. RR = Relative risk