She and her colleagues have created and validated an online risk prediction tool that clinicians can use to predict a patient’s individualized likelihood of complete freedom from seizures 2 and 5 years after undergoing resective brain surgery for drug-resistant epilepsy. The risk predictor, known as the, uses a handful of simple clinical characteristics – patient gender, pathologic cause of the seizures, the proposed type of epilepsy surgery, the presence or absence of generalized tonic-clonic seizures, epilepsy duration, and preoperative seizure frequency – and spits out the patient’s predicted seizure outcome, she explained at the congress, sponsored by the International League Against Epilepsy.
“The point here is that every patient is an individual. And to give people predictions based on 500- or 600-patient Kaplan-Meier-derived curves that just provide the average outcome for the whole cohort isn’t really going to give them what they need as far as their individualized chance of becoming seizure free,” said, a neurologist at the Cleveland Clinic.
Similarly, reliance solely upon clinical judgment is a minefield. Multiple biases prevent physicians from making objective medical predictions, she continued.
“We think of the process of medical decision-making and outcome prediction as being a process that is logical and rational, where the accumulation of knowledge improves the decisions that we make, and where past experience improves judgment, and where collective decisions are more reliable. This is what intuitively we all think. That’s why we think we are invincible as physicians. And to that I say, really? There is a wealth of literature that actually disproves each one of these points,” Dr. Jehi declared.
Outcomes of brain surgery for drug-resistant epilepsy have remained static for more than half a century: Ten years after surgery, roughly half of treated patients remain completely seizure free. The inability of clinicians to use advanced statistics to inform potential surgical candidates about their individualized chance of becoming seizure free has probably contributed to underutilization of epilepsy surgery, she added.
The Epilepsy Surgery Nomogram was developed through detailed analysis of the records of 846 patients who underwent epilepsy surgery at the Cleveland Clinic. The resultant nomogram was then validated in a cohort of 604 patients who had resective surgery at the Mayo Clinic and epilepsy surgery centers in Brazil, Italy, and France. In the development cohort, the rate of complete freedom from seizures was 57% at 2 years and 40% at 5 years. In the validation study, the nomogram had a concordance statistic of 0.60 for complete freedom from seizures, which is considered better than chance, but well below the 0.80 threshold defined as strong concordance ().
However, in an era when personalized medicine has become a catch phrase, the Epilepsy Surgery Nomogram has captured the attention of officials at the National Institutes of Health. Indeed, Dr. Jehi and her coworkers have received a $3.4 million, 5-yearfrom the NIH to improve their risk prediction model by incorporating additional variables, including EEG data, MRI findings, family history, and genetic information. The enhanced risk calculator also will include a predictor of the likelihood that an individual will experience clinically meaningful improvement in quality of life in response to epilepsy surgery, since that’s an important outcome even in the absence of 100% freedom from seizures.
Recently, Dr. Jehi and coworkers have developed and then externally validated nomograms to predict the individualized risk of clinically relevant postoperative naming decline after temporal lobe epilepsy surgery in adults. A model based upon five variables – side of surgery, sex, education, age at epilepsy onset, and age at epilepsy surgery – performed very well, with a concordance statistic of 0.81. Moreover, a second nomogram predicting moderate to severe postoperative naming decline on the basis of just three variables – side of surgery, age at epilepsy onset, and preoperative score on the Boston Naming Test – had a concordance statistic of 0.84 (Neurology. 2018 Dec 4;91:e2144-e2152.).
“Our future hopefully is one where there will always be room for gut feelings and intuition because we definitely need them. We want to honor them. But hopefully it is one where algorithms can help our guesses be more educated and where the science of algorithms and predictive modeling can help inform our outcome predictions and decision-making process,” she said.
The original Epilepsy Surgery Nomogram project was funded by the Cleveland Clinic Epilepsy Center. The postoperative naming decline nomograms project was funded by the NIH.