Interesting Statistical Lessons in Providing Real World Evidence that the ensoETM Device Protects the Esophagus from Thermal Injury During Radiofrequency Ablation

Presented by Babette Brumback, PhD Professor Emerita, Department of Biostatistics University of Florida
Wednesday, September 25, 2024, 4:00 PM, AUST 202
Webex Meeting Link
Coffee will be served at 3:30 pm in the Noether Lounge (AUST 326)

Thermal injury to the esophagus can lead to a life-threatening complication, called atrial-esophageal fistula, after ablation for atrial fibrillation. The ensoETM device (Attune Medical, Chicago, IL, USA) is routinely used to control body temperature in at risk patients in an intensive care setting and in patients whose body temperature must be lowered to protect an injured brain. As it does this by warming or cooling the lumen of the esophagus and stomach, physicians hypothesized that it could also protect the esophagus during radiofrequency (RF) ablation. In an effort to provide real world evidence in support of this hypothesis, we were tasked with analyzing two separate datasets. The first compares incidences of atrial-esophageal fistula prior to versus subsequent to introduction of the ensoETM within several medical centers. The second includes three randomized controlled studies of RF ablation with and without the ensoETM. In the first, we must choose an analysis that accomodates surprising data in the post-introduction period. In the second, data from one center leads to a statistically significant difference, whereas data from the other two do not. The primary analysis combined data across centers and failed to show a difference. However, greater RF ablation time was associated with greater thermal injury, and it also differed across center. Including RF ablation time, which has the status of a potential mediator because it is a post-treatment variable, as an effect modifier presents a different picture of the combined data, and it illustrates some of the challenges of working with an ordinal outcome that is non-zero for only 29 of 166 patients in the combined data. Finally, we present two ways to handle confounders in an analysis that compares efficacy of the ablation with and without the ensoETM.