Machine learning analysis revealed treatment effect heterogeneity, with preoperative β-blockers demonstrating superior outcomes specifically in CABG patients with ventricular dysfunction and concomitant renal dysfunction (eGFR ≤ 66 mL/min/1.73 m²).
Key Findings
Results
Preoperative β-blocker therapy was associated with heterogeneous 30-day mortality outcomes among patients with ventricular dysfunction undergoing CABG.
Data from 6,492 patients in the Chinese Cardiac Surgery Registry database between 2017 and 2020 were analyzed.
An iterative causal forest (iCF) machine-learning algorithm was applied after propensity score matching to estimate individualized treatment effects (ITEs).
The iCF models showed heterogeneity in the effects of preoperative β-blockers on 30-day mortality.
The outcome measured was 30-day all-cause mortality after CABG.
Results
Among patients with eGFR ≤ 66 mL/min/1.73 m², preoperative β-blocker therapy was associated with a significantly lower risk of 30-day all-cause mortality.
Adjusted odds ratio (aOR) was 0.39 (95% CI, 0.22 to 0.67; p = 0.001).
This represents approximately a 61% reduction in the odds of 30-day all-cause mortality in this subgroup.
No significant mortality benefits were found for other subgroups outside this eGFR threshold.
The eGFR threshold of 66 mL/min/1.73 m² was identified by the iCF algorithm as a key treatment-effect modifier.
Results
The iCF algorithm identified eGFR, left ventricular end-diastolic diameter (LVEDD), and BMI as the key variables that distinguished patients with heterogeneous treatment effects from preoperative β-blocker therapy.
These three variables were identified by the machine-learning algorithm as potential treatment-effect modifiers.
eGFR was the primary variable used to define the subgroup with significant benefit.
LVEDD and BMI were also flagged as modifiers, though the abstract does not report specific threshold values or effect sizes for these variables separately.
The iterative causal forest (iCF) algorithm was used to estimate individualized treatment effects after propensity score matching.
Methods
Propensity score matching was used prior to applying the iCF algorithm to estimate individualized treatment effects in this observational registry study.
The study population consisted of patients with ventricular dysfunction undergoing CABG from the Chinese Cardiac Surgery Registry database.
The study period spanned 2017 to 2020.
Total sample size was 6,492 patients.
The design aimed to evaluate heterogeneous treatment effects rather than average treatment effects across the full population.
What This Means
This research suggests that the benefit of taking beta-blocker medications before coronary artery bypass surgery (CABG) is not the same for all patients with a weak or poorly functioning heart. Using a large registry of over 6,400 patients in China and a machine-learning technique called iterative causal forest analysis, researchers found that the overall group did not uniformly benefit, but certain patient characteristics determined who was most likely to benefit.
The most important finding was that patients who also had reduced kidney function—specifically, an estimated glomerular filtration rate (eGFR) of 66 mL/min/1.73 m² or below—had roughly 61% lower odds of dying within 30 days of surgery if they received preoperative beta-blockers compared to those who did not. For patients with better kidney function or other characteristics, no statistically significant benefit was detected. The algorithm also flagged heart size (left ventricular end-diastolic diameter) and body mass index (BMI) as potentially important factors in determining who responds to the treatment.
This research suggests that blanket policies of either giving or withholding beta-blockers before CABG in all heart-failure patients may not be optimal, and that patients with both heart and kidney dysfunction may represent a group that particularly benefits from this therapy. The findings highlight how machine-learning tools can help identify specific patient subgroups that may respond differently to a treatment, which could guide more personalized surgical care decisions in the future.
Lu W, Tang H, Wang J, Han J, Li Y, Wang X, et al.. (2026). Heterogeneous effects of preoperative beta blockers on 30-day mortality after coronary artery bypass surgery.. Annals of medicine. https://doi.org/10.1080/07853890.2026.2719256