Cardiothoracic & vascular anesthesia
Use of Machine Learning to Derive Discrete Clinical Phenotypes and Assess Treatment-Effect Heterogeneity in the Steroids in Cardiac Surgery Trial Dataset
Authors' conclusion
Despite substantial differences in preoperative risk profiles, findings were neutral for methylprednisolone across all phenotypes.
Study context
- Population
- Cardiac surgery patients in the SIRS trial; clusters derived in a Cleveland Clinic developmental dataset (30 preoperative variables)SIRS trial dataset
- Participants
- 6,836 patients
- Comparison
- Methylprednisolone treatment effect across 4 derived clinical phenotypes
- Primary endpoint
- Coprimary: 30-day mortality and composite of death and major morbidity; methylprednisolone treatment-phenotype interaction
Methodological detailsmachine learning phenotype clustering (partitioning around medoids), treatment-phenotype interaction analysis
- Machine learning phenotype clustering (partitioning around medoids)
- Treatment-phenotype interaction analysis
Primary result
As reported in the abstract
Methylprednisolone effect on coprimary outcomes by phenotype
No evidence that methylprednisolone treatment-effect differed by phenotypes
all treatment-phenotype interaction term P > .05
Further findings
| Outcome | OR (95% CI) | |
|---|---|---|
| Composite of mortality and severe morbidity by phenotype, developmental dataset (n = 16,395)Sickest, more CABG and low LVEF vs Younger and healthier | 3.4 (2.4–4.8) | |
| 0.20.5125Odds ratio (log scale) |
Mortality and severe morbidity by phenotype, SIRS trial dataset
| OR (95% CI) | ||
|---|---|---|
| Sickest, more CABG and low LVEFHighest-risk category | 2.1 (1.6–2.9) | |
| 0.20.5125Odds ratio (log scale) |
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Use of Machine Learning to Derive Discrete Clinical Phenotypes and Assess Treatment-Effect Heterogeneity in the Steroids in Cardiac Surgery Trial Dataset. Anesth Analg. Published 19 August 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →