Author + information
- Kuo-Liong Chien, MD, PhD⁎ ()
- ↵⁎Institute of Preventive Medicine, College of Public Health, National Taiwan University, R517, 5F, 17, Hsu Chow Road, Taipei, Taiwan
The recent paper by Miceli et al. (1) showed that, compared with those not receiving angiotensin-converting enzyme inhibitor (ACEI) therapy, patients receiving ACEI therapy had a significant short-term risk of in-hospital death, renal dysfunction, and atrial fibrillation when they underwent a coronary artery bypass grafting. Propensity score matching was applied to control confounding factors that differed between ACEI and non-ACEI patients in this observational cohort study. However, 2 epidemiological issues should be mentioned here. First, potential bias from uncontrolled confounding may arise when ACEI use itself is a marker for a condition with a poor prognosis that will trigger clinicians to use ACEI therapy. Consequently, ACEI therapy is found to be associated with an increased risk of the outcome, as this study showed. The bias cannot be easily handled by propensity score matching, which included only known clinical measures. Second, an accelerated failure time model is more suitable for analyzing the short-term outcome data than a proportional hazards model and conditional logistic regression model due to a short follow-up period. The authors should provide detailed person-time data and check the assumption of proportional hazard for ACEI use. Descriptive statistics, including survival curve plotting and tabulation, may be helpful for readers to understand the clinical scenario. A time-dependent covariate model may be incorporated for adjusting ACEI use and clinical parameters.
Clinical observational studies provide scientific knowledge that should be tested in a randomized, controlled trial setting in which confounding and bias are controlled. Therefore, before searching the biological and clinical meanings of ACEI's effect on the risk of short-term prognosis among patients undergoing a coronary bypass grafting, the authors should first address some basic epidemiological and methodological issues, including confounding by indication and event history data analysis.
- American College of Cardiology Foundation