A Bayesian reanalysis of the CULPRIT-SHOCK trial

The optimal revascularization strategy for patients with acute myocardial infarction (AMI), cardiogenic shock (CS), and multivessel disease remains controversial. The CULPRIT-SHOCK trial compared culprit lesion-only vs. immediate multivessel percutaneous coronary intervention (PCI), providing important data but leaving efficacy questions unresolved. To address lingering uncertainties and gain deeper insights, we performed a Bayesian reanalysis of the CULPRIT-SHOCK trial data.

We conducted a Bayesian re-analysis of the CULPRIT-SHOCK trial data using non-informative, sceptical, and enthusiastic priors. Relative risks (RRs) with 95% highest posterior density (HPD) intervals were calculated. We defined the minimal clinically important difference (MCID) as RR < 0.84. We performed subgroup analyses for key patient characteristics and assessed secondary outcomes and safety endpoints. Probabilities of benefit, achieving MCID, and harm were computed. Results are presented as median RR with probabilities of effect sizes. Bayesian reanalysis showed a median RR of 0.82 (95% HPD 0.66-1.04) with a non-informative prior, indicating a 95% probability of benefit and 59% probability of achieving MCID. Subgroup analyses revealed potentially stronger effects in males (RR 0.78, 73% probability of MCID), patients without diabetes (RR 0.76, 79% probability of MCID), and those with non-anterior ST-segment elevation MI (STEMI; RR 0.74, 76% probability of MCID). Secondary outcomes suggested potential benefits in mortality (RR 0.85) and need for renal replacement therapy (RR 0.72) but increased risks of recurrent MI (RR 2.84) and urgent revascularization (RR 2.88).

Our Bayesian reanalysis provides intuitive insights by quantifying probabilities of treatment effect sizes, offering further evidence favouring the culprit lesion-only PCI strategy in AMI patients with CS and multivessel disease. The analysis demonstrates a high probability of overall benefit, with a notable chance of achieving a minimally clinically important difference, particularly in specific subgroups. These findings not only support the consideration of culprit lesion-only PCI in certain patient populations but also underscore the need for careful risk-benefit assessment. Furthermore, our hypothesis-generating subgroup analyses, which show varying probabilities of achieving MCID, illuminate promising avenues for future targeted investigations in this critical patient population.