Electroencephalographic Biomarker to Predict Postoperative Delirium

Overview

Acute post-operatory cognitive dysfunction states are one of the most important complications in older patients that underwent surgery. Among them postoperative delirium (POD) is the the most studied. Patients who develop delirium have poorer long-term outcomes, such as longer length of hospital stay, institutionalization at discharge, and even higher mortality, and consequently, the human and economic costs significantly increase for the health system. Here the research team will use an observational cohort, investigator blinded in five-center with a primary endpoint to validate intraoperative EEG analysis as a reliable biomarker of postoperative delirium.

Acute post-operatory cognitive dysfunction states are one of the most frequent complications in older patients after surgery, being POD the most important. Previous studies have shown than the incidence of POD in older patients range between 10-50%. Patients who develop POD have poorer long-term outcomes, such as longer length of hospital stay, institutionalization at discharge, and even higher mortality. Consequently, the human and economic costs associated to POD represents an important issue for health systems worldwide. A key element to diminish POD and its burden on healthcare is early diagnostic. Current risk assessment tools are centered on clinical approaches based on cognitive tests (i.e., MoCA) and/or prediction models that uses patients' clinical variables (i.e., DELPHI score). We have developed a strategy that uses intraoperative EEG features as building blocks for a new POD risk assessment predictive model. This system, called PEUMA, uses data obtained from 95 patients from a previous study (NCT04214496). This will be a multicenter (five-centers), observational study and its primary outcome will be PEUMA's ability to predict POD. To calculate the sample size, the methodology described by Riley et al was used. This method is specially designed for clinical prediction models. Such a tool is available online (https://mvansmeden.shinyapps.io/BeyondEPV/). The parameters used were the following: * Number of predictor candidates: 4 * Fraction of events: 0.22. 22% was used because it is the incidence of POD in the analysis of the preliminary data of the first stage and these are in the reporting range common worldwide. * Estimation error of the classifier: 0.06. The authors suggest prediction errors small when evaluating binary outcomes (Yes POD/No POD) The calculation indicates a sample size of 240 patients. Considering a loss of 10% (in the preliminary results of the first stage the loss was 8%), the sample size is 264 patients.