In Reply:— We thank Dr. Roth for his interest in our article and insightful commentary. Previous literature has demonstrated an association between intraoperative hypothermia and cardiac adverse events. However, we did not evaluate this clinical element in our analysis for several reasons. First, previous data regarding hypothermia and cardiac adverse events is limited to high-risk patients who had a preexisting diagnosis of coronary artery disease or several known risk factors for coronary artery disease undergoing high-risk thoracic, intraperitoneal, or vascular procedures.1Although our dataset included some high-risk patients, only 9.6% had a previous cardiac intervention and only 22% were undergoing high-risk surgery.2As a result, the studied population was dissimilar to previous work, and we were skeptical of being able to identify an association between hypothermia and cardiac adverse events in this more representative population. Second, although our studied dataset was large, we were only able to observe 83 events. As a result, we had to limit the number of independent variables evaluated in the logistic regression full-model fit to reduce the impact of model overfitting.3Hypothermia was one of several independent variables that we were unable to assess because of this statistical analysis constraint.
Finally, the absence of a consistent way to separate “hypothermic”versus “normothermic” groups in an observational dataset presented the final challenge. There are several ways to define hypothermia. First, we could evaluate median temperatures within 10-min epochs, similar to the presented hypotension analysis. Second, some may advocate that a single temperature measurement below 36°C would qualify as “hypothermic.” Third, others may suggest that we employ the absence of active warming to be consistent with prospective, controlled studies.
We agree that intraoperative hypothermia should be evaluated in future studies. We look forward to conducting large, multicenter observational dataset analyses that may offer us the statistical power necessary to do so.
*University of Michigan Medical School, Ann Arbor, Michigan. email@example.com