Accelerated long-term forgetting in drug-resistant focal epilepsy : insight from a prospective controlled study using a multimodal associative memory paradigm

BackgroundEpisodic memory complaints are frequently reported by people with epilepsy (PWE), and often contrast with normal performance on standard neuropsychological assessment using short retention intervals (20-30 minutes). This discrepancy is consistent with the concept of accelerated long-term forgetting (ALF). However, the objective assessment of ALF remains challenging in clinical practice, and the underlying mechanisms, whether related to encoding, early or late consolidation, are still debated. This study aimed to evaluate a specifically designed associative memory task to detect episodic memory impairment and ALF in PWE who show normal performance on standard neuropsychological assessment. MethodsIn a prospective non-randomised controlled study, 40 PWE and 40 healthy controls (HC) completed an associative memory task comprising the encoding of 56 abstract word-naturalistic scene pairs, followed by retrieval at 30min and 72h. At each delay, both recollection and recognition were assessed on complementary halves of the encoded material. During encoding, participants had to provide for each pair a subjective judgement about the congruence between the word and the scene. The primary endpoint was recollection-based memory loss 72h after encoding in PWE compared to HC, analysed using a logistic mixed-effects model. ResultsMemory significantly declined 72h after encoding, with a lower recollection score at 72h than at 30min in both groups (PWE: OR [95%CI]: 0.31 [0.26;0.38]; HC: OR [95%CI]: 0.21 [0.17;0.26]). At 30-min recollection, PWE were less likely to perform correctly than HC (OR [95%CI]: 0.49 [0.38;0.63]) and despite a smaller memory loss over time, PWE continued to perform significantly worse than HC at 72h (OR [95%CI]: 0.73 [0.58;0.91]. No association was found between memory performance and epilepsy-related clinical features. A three-way interaction was observed between the congruence judgement, group and time (p=0.03), with PWE recovering normal recollection performance at 30 min for pairs judged as highly congruent during encoding. DiscussionThe results suggest that ALF in epilepsy may reflect disruption occurring early in the consolidation process, potentially during encoding itself and underline the importance of associative, multimodal episodic memory tasks to better characterise memory impairment in PWE.