Seizure prediction algorithms no better than random predictor

Researchers assessing the usefulness prediction algorithms for anticipation of seizures have found them to be no better than a random predictor.Epilepsy is a common neurological problem second only in prevalence to stroke. The unpredictable nature of seizures is the most disabling aspect of the disease for patients, increasing their risk of injury and creating a feeling of helplessness that impacts on quality of life. Mormann et al. theorised that if it were possible to reliably predict seizure occurrence from the electrocephalogram of epilepsy patients, targeted and on-demand therapeutic advances could eventuate with the potential to move away from current preventive strategies, which involve long-term medication with antiepileptic drugs. However, understanding of mechanisms that lead to the occurrence of epileptic seizures is incomplete, and debate rages over the reproducibility of results and suitability of various approaches. In particular, the appropriateness of seizure prediction algorithms for clinical implementation remains unclear. Mormann et al. conducted a literature review on seizure prediction according to methodological standards. They found several studies used prediction algorithms in a prospective manner. However, they did not include a statistical validation or did not allow conclusions on statistical validity. The authors found that when subjected to more rigorous methodological design, measures previously considered suitable for prediction of seizures were no better than a random predictor. As such, the authors concluded that the current literature does not allow a definite conclusion as to whether seizures are predictable using prospective algorithms. They did find; however, that measures quantifying relations between different brain regions appear to perform better than random measures even with statistical validation."The current literature does not allow a definite conclusion as to whether seizures are predictable using prospective algorithms," the authors noted."Prediction algorithms must be proven to perform better than a random predictor before prospective clinical trials involving seizure intervention techniques in patients can be justified," they concluded.Reference...