To the Editor:-The response provided by Wallace regarding the study on perioperative atenolol and myocardial ischemia did not provide an adequate clarification of the statistical results in their previous reports. [2,3]Wallace et al. claimed that their previous report demonstrated that perioperative administration of atenolol decreases the incidence of death after surgery during a 2-yr follow-up period. However, their study did not provide statistical support for such a claim. In a univariate analysis (i.e., an analysis in which the effect of a single “independent” parameter is evaluated), the hazard (odds) ratio for death was 0.4 with a confidence interval of 0.2-0.9. This result suggested that the likelihood of death is 2.5-fold less when atenolol was given and that the effect of atenolol was statistically significant. However, in all analytic studies, confounding variables must always be considered as an alternative explanation for study findings, as was done by Mangano et al. In the multivariate analysis model, the hazard ratio for atenolol was 0.5 with a confidence interval of 0.2-1.1. Because the confidence interval now included 1.0, this indicated that the effect of atenolol was not different from that of placebo, i.e., no influence of atenolol on survival. In fact, in this model, diabetes mellitus proved to be the most important predictor of death with a hazard ratio of 2.8 (confidence interval, 1.4-6.2). Stated differently, when the effects of diabetes have been considered, there is no longer an effect of atenolol.
Because the message that “patients with or at risk for coronary artery disease who are treated perioperatively with [Greek small letter beta]-adrenergic blocking agents have reduced incidence of morbidity and mortality” cannot be supported by the current studies, [2,3]physicians caring for such patients should reevaluate the validity of recommending perioperative beta blockade for improved survival that was based solely on the current study findings. Furthermore, because the clinical treatment of these 200 patients after hospital discharge was completely out of the study protocol control, the effects of other confounding factors cannot be determined and assessed.
When studies involve complicated statistical modeling, it is a responsibility of the authors to understand the technique used to provide an accurate and meaningful interpretation of their results. Unfortunately, the recent popularized use of perioperative beta blockade is based on misinformation.
Jacqueline M. Leung, M.D., M.P.H.
Associate Professor of Anesthesia; University of California, San Francisco; San Francisco, California; Department of Anesthesia; UCSF Mount Zion Medical Center; San Francisco, California;firstname.lastname@example.org
(Accepted for publication November 11, 1998.)