BACKGROUND: Since epilepsy surgery is an elective procedure, patients need to weigh the risks of the procedure against the likely outcome if they are to make an informed decision to proceed. The aim of this study was to examine the accuracy of multidisciplinary team predictions of postoperative outcome in epilepsy surgery candidates. METHODS: An experienced multidisciplinary team provided preoperative predictions of postoperative outcome in 94 temporal lobe epilepsy patients who subsequently proceeded to surgery and were followed up one year later. RESULTS: Team predictions of postoperative outcome were generally accurate for groups of patients judged to have a 30%, 40%, 50% or 60% chance of becoming seizure free. Team estimates of odds tended to regress towards the mean. Logistic regression analyses were more accurate than the team estimates in identifying patients with a very good (>70%) or very poor (<20%) chance of complete seizure freedom. Non localising scalp EEG, necessitating the need for an invasive EEG study prior to surgery was a significant predictor of poor postoperative outcome in this series. CONCLUSIONS: Probabilities based on logistic regression models may augment and improve the accuracy of clinical estimates of postoperative outcome in patients with a very good or very poor chance of being rendered seizure free by surgery, by counteracting the tendency of regression towards the mean in team decision making.
BACKGROUND: Since epilepsy surgery is an elective procedure, patients need to weigh the risks of the procedure against the likely outcome if they are to make an informed decision to proceed. The aim of this study was to examine the accuracy of multidisciplinary team predictions of postoperative outcome in epilepsy surgery candidates. METHODS: An experienced multidisciplinary team provided preoperative predictions of postoperative outcome in 94 temporal lobe epilepsypatients who subsequently proceeded to surgery and were followed up one year later. RESULTS: Team predictions of postoperative outcome were generally accurate for groups of patients judged to have a 30%, 40%, 50% or 60% chance of becoming seizure free. Team estimates of odds tended to regress towards the mean. Logistic regression analyses were more accurate than the team estimates in identifying patients with a very good (>70%) or very poor (<20%) chance of complete seizure freedom. Non localising scalp EEG, necessitating the need for an invasive EEG study prior to surgery was a significant predictor of poor postoperative outcome in this series. CONCLUSIONS: Probabilities based on logistic regression models may augment and improve the accuracy of clinical estimates of postoperative outcome in patients with a very good or very poor chance of being rendered seizure free by surgery, by counteracting the tendency of regression towards the mean in team decision making.