Families. We tend to think that we understand familial epilepsy. The mental image is almost fixed: large pedigrees, clean inheritance, recognizable syndromes, and genes that segregate exactly as expected. But in a recent study from our epilepsy genetics program, we looked at something different. Not discovery cohorts and carefully selected multiplex families, but real-world genetic testing in 484 consecutive families. All families were evaluated in routine clinical care through ENGIN, our Epilepsy Neurogenetics Initiative, which has now evaluated more than 7,000 individuals with epilepsy and related disorders. The result was a picture of familial epilepsy that is both reassuring and unexpectedly complicated. Here is what we found.
Figure. Genetic testing in familial epilepsies in 484 consecutive families evaluated through ENGIN, our epilepsy genetics program. (1) Bar chart showing the number of diagnoses per gene among solved familial epilepsy cases. KCNQ2, NPRL3, and PRRT2 were the most frequently identified etiologies, with 46 additional genes identified in single families each. (2) Comparative bar chart showing the proportion of each gene among all genetic epilepsies versus the familial epilepsy subset, highlighting enrichment of dominantly inherited, incompletely penetrant genes including SCN1A, DEPDC5, and NPRL3. (3) Pedigree examples illustrating unexpected inheritance patterns, including de novo variants, bi-lineal inheritance, and parental mosaicism, with corresponding genetic etiologies listed. The overall diagnostic yield was 20%, and approximately one-fifth of solved families showed an inheritance pattern inconsistent with the presenting pedigree (Figure adapted from Ellis et al., 2026).
Real-world families. In our recent study led by Colin Ellis published in Epilepsia, we took a closer look at what familial epilepsy actually looks like in 2026. One of the most important aspects of this study is what this cohort represents. These families were not recruited for research but seen through routine clinical care in our epilepsy genetics clinics at CHOP and Penn. Inclusion required a clear family history, defined as the proband plus at least one first-degree relative or two more distant relatives with epilepsy, and family history was reviewed systematically as part of the clinical evaluation. Genetic testing followed real-world decision making rather than a protocol, including gene panels, exome sequencing, microarray, and stepwise testing shaped by clinical judgment, insurance approval, and timing. Real-world genetic testing is more complex than a research study where we obtain exomes or genomes on everybody. However, it reflects how genetic testing in familial epilepsies actually plays out in clinical practice.
This distinction matters. Familial epilepsy played a central role in the early years of epilepsy gene discovery, but the field later shifted toward de novo variants in developmental and epileptic encephalopathies (DEE), where gene discovery was faster, more urgent, and often genetically more straightforward. As a result, we actually know less about familial epilepsy in modern clinical practice than we sometimes assume. Therefore, our cohort reflects what familial epilepsy looks like in 2026, not in a discovery study, but in the clinic.
The yield is meaningful. The overall diagnostic yield in this cohort was 20 percent. We identified a pathogenic or likely pathogenic variant in 99 of 484 families, a number that sits right in the range many of us expected. This is reassuring given how heterogeneous these families are and was actually surprising to me, as I would have assumed a yield of 10% or less based on my own subjective experience. It is a good reminder that gut feelings and hunches are not always right.
What makes this result particularly interesting is how closely it mirrors earlier familial epilepsy cohorts. In the well-known study by Afawi and colleagues, multiplex families studied with research-based genetic analysis had a diagnostic yield of about 23 percent. Those families were often large, sometimes contributed to gene discovery itself, and were not necessarily representative of routine clinical practice. The fact that the yield in our real-world cohort is nearly identical suggests that familial epilepsy remains one of the strongest indications for genetic testing even in the current era.
This connects to something I wrote about years ago when discussing familial epilepsy cohorts. At the time, I wondered whether the shift toward disorders with de novo inheritance meant that families had generally become harder to solve. The study by Ellis et al. suggests that this is not the case. Even in 2026, familial epilepsy still carries a strong genetic signal.
Genes. The distribution of genes in this cohort looks familiar at first glance. KCNQ2, PRRT2, SCN1A, and DEPDC5 appear frequently, exactly as expected for dominantly inherited epilepsies. These are the genes that define many of the classic familial epilepsy syndromes, and it makes sense that they are enriched when we study families rather than unselected epilepsy cohorts. However, the relatively high frequency of NPRL3 families was surprising, even surpassing families with DEPDC5 variants.
When the GATOR1 genes were first described, I remember being skeptical. For the first couple of years after their discovery, I was not fully convinced that they would turn out to be valid epilepsy genes, given that they were initially identified as candidate genes rather than through linkage studies. To me, they did not feel like the kind of genes that would become common clinical diagnoses. I was wrong again. NPRL3 and the other GATOR1 genes sit in the uncomfortable middle ground between monogenic and complex inheritance. They are dominantly inherited but incompletely penetrant, often mild enough to allow transmission across generations, and variable enough to produce family histories that do not look Mendelian at first glance. In a real-world familial cohort, these are precisely the genes that accumulate.
Family structure does not predict. One of the most counterintuitive findings in this study is that pedigree structure does not influence diagnostic yield. The number of affected relatives, the presence of an affected parent, or the number of generations involved did not significantly change the likelihood of finding a genetic cause. This contradicts a long-standing intuition in epilepsy genetics that I once described as the valley of despair, the idea that very small families are hard to interpret, very large families are informative, and intermediate families are the most difficult. In this cohort, the valley largely disappears. Families with only a few affected individuals had similar yields to families with many.
Unexpected segregation. Even more striking is that about one fifth of solved families showed unexpected segregation. Variants that should have been inherited turned out to be de novo. Variants were transmitted from unaffected parents. In some families, the genetic finding did not match the pedigree at all. These patterns reflect incomplete penetrance, parental mosaicism, and sometimes the coexistence of more than one epilepsy in the same family. Familial epilepsy in 2026 is more than simple Mendelian genetics.
What you need to know. In the study by Ellis and collaborators of 484 families evaluated through ENGIN, our epilepsy genetics program that has now seen more than 7,000 individuals, familial epilepsy had a diagnostic yield of about twenty percent, very similar to earlier multiplex cohorts despite the heterogeneity of real-world clinical practice. The most common genes were dominantly inherited epilepsy genes with incomplete penetrance, including KCNQ2, PRRT2, SCN1A, DEPDC5, and NPRL3, illustrating that familial epilepsies often lie between Mendelian and complex inheritance. Pedigree structure did not predict the likelihood of a genetic diagnosis, and about one fifth of solved families showed unexpected segregation, emphasizing how often family history can mislead. The only consistent predictors of a genetic finding were early seizure onset and the presence of neurodevelopmental disorder, reinforcing that biological severity remains the strongest signal even in clearly familial cases.
