Three criteria. You hear the phrase precision medicine quite frequently these days and might wonder what this is all about. In a nutshell, in the context of genetic epilepsies, the basic idea behind precision medicine is to use genetic patient information for treatment decisions. The broader vision behind this aims at improving the lives of individuals with epilepsy by making smarter and faster treatment decisions, which lead to better treatment response and fewer side effects. But how should we assess information on reports of precision medicine in the literature? Here are the three important criteria to assess.
Blueprint. There is an emerging body of case report, which uses the following template. (A) Exome/genome sequencing was performed in a patient with intractable epilepsy and a mutation in gene X was found. (B) Based on the genetic finding, treatment Y was initiated. (C ) The patient’s seizures improved upon this treatment decision. For example, one example was a recent publication by Pierson and collaborators, which (A) identified a GRIN2A de novo mutation in a patient with epileptic encephalopathy, (B) showed in vitro that memantine restores some of the aberrant channel function and used this for patient treatment. Upon treatment with memantine, (C) the patient’s seizures improved. In addition to systematic studies like this, there are various retrospective reports on positive responses to medications in patients, in which a genetic cause was identified through massive parallel sequencing technologies. These examples include lamotrigine in SCN2A encephalopathy or pyridoxine in congenital disorders of glycosylation (CDG).
Criteria for precision medicine. What are the quality criteria that you should apply to a case report that claims to successfully use genetic information for treatment choices? I would suggest that all three components listed above should be examined and need to to “waterproof”. These include (A) the causative nature of the mutation, (B) the reasoning of the choice of treatment and perhaps most importantly (C) the appropriate measure of the patient-related effect or outcome. Let’s examine the “CRE model” in detail (causation-reason-effect).
C – causation. The genetic alteration found in the patient needs to be causative according to established diagnostic criteria. The causative variants may include deleterious variants in known epilepsy genes. If novel genes or genes of uncertain significant are affected, we should take into account that the variant, which all further investigations are based upon, may in fact not be causative. I acknowledge that there might be modifiers that could play a particular role in the context of precision medicine, but we are not at the point yet to reliably pinpoint their role. The report by Pierson and collaborators finds a de novo mutation in GRIN2A. Even thought the phenotype is not typical, the gene is established and criterion “C” is fulfilled.
R – reason for treatment choice. The reasoning to chose a particular treatment must be stringent. In the case of the study by Pierson and collaborators, the authors managed to model the mutation in vitro, which allowed them to test various compounds. The treatment choice needs to be specific to the genetic alteration – deciding to use a antiepileptic drug may well result in a broad non-specific effect that is unrelated to the underlying mutation, suggesting a specific effect that is simply not there. In many cases, replicating the particular mutation or gene effect in a model system might not be possible – if we chose any particular treatment based on the genetic finding, we need to be aware that our reasoning might be insufficient. The same applies for using retrospective information from case reports. For example, the positive effect of lamotrigine in patients with SCN2A encephalopathy has been discussed widely – however, this only applied to a few patients reported by Nakamura and colleagues. They might have identified a particular patient subgroup or this might simply be chance.
E – effect. Finally, we need to critically assess the effect of the treatment decision. Was the follow-up period sufficiently long enough to account for a honeymoon effect? Is it possible that the authors only observed random fluctuation of a self-limiting epilepsy rather than a real effect of the treatment? There is a huge body of knowledge from clinical trials to assess the efficacy of a new drug – we need to make sure that we critically assess the treatment effect and don’t fall into the trap of wishful thinking as (A) and (B) sounded very convincing. It is a well known cognitive bias to see patterns in random data. We have to make sure that we do our best to avoid such as clustering illusion.
Next Steps. These suggested criteria will need to be refined and elaborated on. All I wanted to do in this post is raise the issue that there are at least three components to consider when assessing case reports of precision medicine in patients with genetic epilepsies. It might be worthwhile establishing a generally accepted framework that we can use to evaluate such information. Finally, there is invariably a publication bias as only positive and successful reports are published. This may then further into prominence an association which may have been arisen purely due to chance.
This blog post is on behalf of the ILAE Genetics Commission.