Clinical seizure semiology is subtle and identification of seizures by parents is unreliable in infants with tuberous sclerosis complex

**Background:** Tuberous sclerosis complex (TSC) is a genetic multisystem disorder with epilepsy occurring in 60-90% of patients, usually presenting in infancy. Recent studies have shown that treatment based on abnormal EEG findings—even before clinical seizures—delays seizure onset and reduces drug-resistant epilepsy. Expert consensus recommends serial EEG screening and parental education on subtle seizure recognition. However, few studies have assessed the reliability of parental seizure detection in infants with TSC or documented total seizure burden (clinical and subclinical) using prolonged vEEG.

**Methods:** This retrospective study recruited consecutive infants with antenatally diagnosed TSC from a single institution (Queensland Children's Hospital, Brisbane, Australia) over a 5-year period (May 2014–April 2019). All infants had neurologist consultation immediately after birth and serial parental education on subtle seizure signs. Infants who underwent prolonged vEEG (≥20 hours) in the first 2 years of life were included. Two investigators independently reviewed all vEEG data, classifying seizures as clinical, subclinical, or unknown. For clinical seizures, parental recognition was assessed via event button use or video/audio evidence. Seizures were defined as paroxysmal, rhythmic epileptiform activity evolving in amplitude, morphology, and location. Subclinical seizures required ≥10 seconds duration unless identical to recognized stereotyped clinical seizures.

**Key Results:** Nine infants were enrolled (6 male, 3 female; 6 TSC2, 3 TSC1 mutations). All had cardiac rhabdomyoma as the presenting fetal feature. Seven of nine had high tuber burden (≥10 tubers). A total of 24 prolonged vEEG studies were performed (total 634 h 49 min, mean 26 h 3 min per study). Vigabatrin was commenced before clinical seizures in all patients. Across 24 recordings, 674 focal seizures were identified in 8 of 9 patients. Of these, 220/674 (32.6%) were clinical seizures, 395/674 (58.6%) were subclinical, and 59/674 (8.8%) were unclassifiable. Mean focal seizure frequency was 1.1 seizures/hour. Only 63/220 clinical seizures (28.6%) were identified by parents; 157/220 (71.4%) were not identified. Of 179 awake clinical seizures, 157 (87.7%) had behavioral arrest as the initial feature. Of 41 clinical seizures during sleep, 40 (97.6%) were undetected by parents. Thirty clusters of epileptic spasms were detected in one patient: at least one clinical spasm occurred in 2/30 (6.7%), 24/30 (80%) were electrographic only, and 4/30 (13.3%) were uncertain. No clinical epileptic spasms were identified by parents.

**Clinical Implications:** This study demonstrates that infants with antenatally diagnosed TSC who meet criteria for prolonged vEEG have high seizure frequency that is substantially underestimated by parents, despite serial education on subtle seizure recognition. Clinical seizure semiology was predominantly subtle non-motor behavioral arrest (87.7% of awake seizures), which is difficult to distinguish from normal infant behavior without EEG correlation. The parental detection sensitivity of 28.6% is lower than the 43% reported in a prior mixed-etiology pediatric study, likely due to the subtle semiology in this young TSC cohort. These findings challenge the reliance on parental seizure diaries for clinical decision-making and research outcomes in TSC. The study supports the use of prolonged vEEG for accurate seizure frequency assessment and suggests that epilepsy management decisions should incorporate vEEG data rather than depending solely on parental reports. The results also support the 2021 ILAE position removing the minimum 10-second duration requirement for EEG-only seizures, as stereotyped clinical seizures were observed with ictal patterns ≤10 seconds. Limitations include small sample size (common in rare diseases), potential bias toward more severe phenotypes due to antenatal recruitment, and inability to assess specificity of parental seizure detection due to retrospective design.