Intriguing association of Parkinson's disease and epileptic seizures

We read with interest the article by Gruntz et al. These authors suggest that incident Parkinson's disease (PD) is associated with an increased risk of incident epileptic seizures. This observation is a very important contribution, which will be useful in future studies, and we want to acknowledge these authors for their efforts. However, some issues raise concerns and merit a more detailed examination. First, Gruntz et al’s case selection is based on a large data set from a primary care setting. The same group, using the same data set in previous articles, specified that only a small percentage (1%) of all cases were referred to a neurologist to confirm the diagnosis, and that seizure diagnosis is not formally validated in the data set. Other disorders with prominent signs and symptoms of transient loss of consciousness (TLOC), such as syncope, may be confused with seizures. The reliability of the diagnosis of the first TLOC is therefore surprisingly low, and this may particularly be the case in a primary care setting, such as the one used for this study. Second, for case selection, one of the engaged criteria was usage of antiepileptic drugs (AEDs) in combination with diagnostic codes indicating a suspected epileptic seizure as a surrogate for the diagnosis of epileptic seizure. However, AEDs can be indicated for many conditions other than epilepsy, which may lead to misclassification. In a sensitivity analysis, these authors applied a definition of incident epilepsy that they used in another study with the same data set. However, that study used criteria of AED prescriptions in conjunction with diagnostic codes for epilepsy and not with codes for suspected epileptic seizures. Finally, this study utilized a conditional logistic regression analyses to explore the “real association” between PD and the risk of epileptic seizures. Nevertheless, these authors essentially presented only the results of a univariate logistic regression. A univariate analysis should be only the first step in any model building process. Despite the fact that this study matched controls for many variables, there can still be effect modifiers, and a multivariate analysis could examine potential statistical interactions between each of the selected variables. Overall, the association of PD and epileptic seizures is intriguing, and we strongly support further investigation in this area. However, considerable caution is necessary in making causal inferences from case‐control studies.