Long-term memory

Long-term memory can be divided into two categories: explicit memory and implicit memory (Figure 2).^{Reference Lum and Conti-Ramsden5} Explicit memory, also known as declarative memory, refers to information that can be consciously recalled and described in detail. It has two subtypes: episodic and semantic memory. Episodic memory involves memories of specific events, such as a memorable birthday party from childhood. You can consciously recall this event and explain details like the flavor of the cake. Semantic memories pertain to facts, such as knowing that 2 + 2 equals 4. Implicit memory, also called non-declarative memory, is often used unconsciously, and its details are challenging to explain. Its primary subtype is procedural memory, which includes skills like riding a bike or driving a car. These tasks feel mostly automatic and are performed unconsciously, making it difficult to explain exactly how you balance on a bike or drive smoothly. Another form is priming, where behavior is influenced by unconscious exposure to a stimulus. For example, someone might crave doughnuts later in the day after passing a doughnut shop on the way to work without consciously noticing it. These different types of memories involve different areas of the brain. Episodic and semantic memories engage the mesial temporal and hippocampal structures, while procedural memory involves the basal ganglia, motor cortex and cerebellum.

Figure 2.

Types and subtypes of long-term memory.

Memory consolidation

When memories are newly encoded into the long-term memory store, they are vulnerable to disruption or change, especially during the first few times they are retrieved and brought back into consciousness.^{Reference Diekelmann, Wilhelm and Born6} Therefore, they must be consolidated or stabilized so that they can remain consistent for a lifetime. This consolidation occurs during periods of rest and sleep. Slow wave sleep (N3 stages 3/4) is when this process occurs, as indicated by desynchronized electrical activity recorded in the hippocampus during this period from electroencephalography recordings. In the REM (rapid eye movement) stage of sleep, more substantial consolidation of emotionally significant memories takes place. The process of consolidation is believed to be modulated by changes in hippocampal serotonin levels in the hippocampus. Raphe neurons, located in the raphe nuclei within the medial reticular formation of the brainstem, release this hormone. They decrease serotonin levels during slow wave and REM sleep, which is thought to facilitate proper consolidation.

The cortico-hippocampal circuit

The cortico-hippocampal circuit plays a crucial role in the retrieval and consolidation of episodic memories.^{Reference Shastri9} During retrieval, the neural circuits that encode the memory are reactivated. Information stored in cortical areas is transferred to the frontal lobe via the hippocampus (Figure 3). This process begins when specific sensory information enters sensory cortical areas and acts as a retrieval cue. Retrieval cues are features present at the time of memory encoding. For example, a particular smell may trigger a specific memory. These cues signal the hippocampus via the entorhinal cortex. Within the hippocampus, different regions have specialized functions related to this process:

Figure 3.

Coronal section of human brain displaying the major structures involved in the cortico-hippocampal, Papez and cortico-striatal circuits: CG = cingulate gyrus; Fx = fornix; AN = anterior nucleus; Pt = putamen; GB = globus pallidus; MB = mammillary body; DG = dentate gyrus; Sb = subiculum; EC = entorhinal cortex; PG = parahippocampal gyrus. The cortico-hippocampal pathway is shown in yellow. Inspired by Reference Crumbie10, drawn with Sketchbook (not to scale).

The dentate gyrus (DG) helps differentiate similar memories,^{Reference Schmidt, Marrone and Markus11} while the CA3 region performs pattern completion, retrieving a complete memory from a partial cue. The CA1 region performs temporal differentiation, distinguishing between similar events at different times, such as what you ate for lunch today versus yesterday.^{Reference Zou, Wanjia and Allen12} The DG contains granule cells, while the CA regions contain pyramidal neurons that facilitate excitatory activity. The CA1 region relays signals back to the entorhinal cortex via the subiculum, allowing communication with other cortical areas. The frontal cortex helps focus attention on relevant memories and distinguish between similar ones, much as the DG does.^{Reference Fletcher and Henson13} This process ultimately enables stored memories to be retrieved or brought into consciousness for further use. The CA2 region is mainly involved in social memory, specifically in distinguishing between familiar and unfamiliar individuals using aspects like their face and voice.^{Reference Tzakis and Holahan14} It receives inputs from the medial and lateral entorhinal cortices, as well as some regions of the hypothalamus, making it suited for this function. The volume of the right DG/CA4 area has been associated with free recall of verbal and visual material, highlighting its role in retrieval without cues.^{Reference Aumont, Bussy and Bedard15} The DG and CA4 are often grouped together in imaging studies due to their anatomical and functional proximity, even though they have distinct circuitry.

According to the Standard Consolidation Theory, the hippocampus is only involved in retrieving episodic memories and recently consolidated semantic memories.^{Reference Winocur and Moscovitch16} After repeated retrieval, the neural pathways between cortical areas and the frontal lobe strengthen so that hippocampal activity for semantic memories becomes unnecessary. However, according to the Multiple Trace Theory, the hippocampus remains continuously involved in episodic memory regardless of its age.^{Reference Nadel, Samsonovich, Ryan and Moscovitch17} The hippocampus is not a temporary locker but a co-archivist that you always need for vivid episodic recall. The more you recall, the more backup copies (traces) are made and distributed across brain networks.

The Papez circuit

The Papez circuit is involved in encoding and consolidating episodic memories, especially those with emotional significance.^{Reference Kamali, Milosavljevic and Gandhi18} It comprises the main structures of the limbic system, which generally play a role in fear and emotion. The involved structures include the hippocampus, fornix, mammillary bodies, anterior nuclei of the thalamus and the cingulate gyrus (Figure 4). The fornix is a tract (bundle of axons) that originates from the posterior part of the hippocampus and curves over the thalamus. The hippocampus connects via the fornix to the septal nuclei in front of the anterior commissure and to the mammillary bodies behind it. Neural signals travel from the hippocampus through the fornix to an area beneath the thalamus, where the hypothalamus and mammillary bodies are located, with the latter included as part of this structure. Signals then pass from the mammillary bodies to the anterior nuclei of the thalamus, a collection of neuronal cell bodies, via the mammillothalamic tract. From there, they travel to the cingulate gyrus, which is positioned anterior to the corpus callosum, and back down to the hippocampus via the entorhinal cortex. Signals can also be transmitted directly from the cingulate gyrus to the amygdala, an almond-shaped structure located anterior to the hippocampus that collaborates with it to strengthen memories with emotional significance. Damage to the amygdala has been shown to cause a reduction in hippocampal activity within this circuit, while loss of hippocampal volume can decrease activity in parts of the amygdala.^{Reference Richardson, Strange and Dolan19} This indicates that the hippocampus and amygdala are intrinsically linked, enabling them to work together to reinforce emotional memories more effectively.

Figure 4.

Sagittal (lateral) view of the brain depicting structures involved and pathway of neural impulses (pink) in the Papez circuit and pathway of communication between the hippocampus and amygdala (red). Inspired by Reference Pinel and Barnes20, drawn with Sketchbook (not to scale).

The cortico-striatal circuit

The formation of procedural memories and habit development occurs through the cortico-striatal circuit.^{Reference Haber21} Electrical signals travel from the primary motor cortex to the striatum, then to the globus pallidus and finally to the thalamus, which relays signals back to the motor area. As this circuit strengthens, it can be physically observed in shorter reaction and completion times for motor tasks as they are repeated. The amount of conscious effort needed decreases, and increased activity in the circuit takes over, enabling the task to be mainly performed automatically.

Valproate

Valproate is a broad-spectrum ASM.^{Reference Sgobio, Ghiglieri and Costa26} It works by increasing levels of γ-aminobutyric acid (GABA), an inhibitory neurotransmitter. This occurs by inhibiting GABA transaminase, an enzyme that normally degrades GABA in the synaptic cleft. Additionally, VPA inhibits the “persistent” sodium current by interfering with sodium channels. The fast sodium current triggers action potentials, while the persistent current maintains neuronal excitability. Long-term use of this medication can also lead to epigenetic changes, which modify gene expression without altering DNA sequences. It inhibits histone deacetylase, an enzyme that normally removes acetyl groups from histone proteins, around which DNA is tightly bound in eukaryotic cells. This removal usually condenses the DNA around histones and prevents gene transcription. However, VPA’s inhibition of this enzyme results in more loosely bound DNA and an increase in transcription rates. The drug can also increase the expression of brain-derived neurotrophic factors, which play a role in synaptic plasticity.

Research has shown that chronic use of VPA causes significant memory deficits in non-epileptic (wild type) mice by inducing structural changes in dendrites.^{Reference Sgobio, Ghiglieri and Costa26} Moreover, the structural damage caused by epilepsy itself, such as alterations in dendrites within the CA1 area, was not reversed in epileptic mice through VPA therapy.

In humans, the cognitive effects of valproate appear more variable and generally less severe than suggested by animal models. In an older study comparing valproate with CBZ, patients treated with valproate did not show significant declines in memory or other neuropsychological functions over 6–12 months; however, this comparison should be interpreted cautiously, as CBZ itself is associated with cognitive impairment.^{Reference Prevey, Delaney, Cramer, Cattanach, Collins and Mattson27} A review of sodium valproate studies also concluded minimal adverse cognitive effects in human subjects.^{Reference Trimble and Thompson28} A more recent comparative monotherapy study found some reduction in cognitive test scores in the valproate group relative to LEV 3 months after starting treatment, highlighting that memory effects can vary with drug choice and patient population.^{Reference El Sabaa, Ghoz, Hamdy and Sarhan29}

Carbamazepine

CBZ works to reduce excitation by blocking Na⁺ and Ca²⁺ channels and keeping them in inactivated states.^{Reference Maan, Duong and Saadabadi30} The effects of CBZ and VPA on epileptic males who experienced tonic-clonic seizures but became seizure-free due to the medication were examined.^{Reference Shehata, Bateh, Hamed, Rageh and Elsorogy31} This group was compared to a control group of epileptic patients who had never been exposed to any ASMs and a group of healthy individuals with no history of seizures. Epileptic patients treated with ASMs showed worse performance on short-term memory tasks than those not treated with ASMs. In this case, no correlation was found between drug dosage and cognitive performance. CBZ intake can cause attention and concentration deficits, which may explain its effects on short-term memory. It may also cause psychomotor slowing, referring to a reduced ability to process information. The drug was also found not to cause significant memory deficits in working, short-term or long-term memory based on dosage.^{Reference Rönnberg, Samuelsson and Söderfeldt32} However, some negative effects on recency were observed when blood levels of the drug exceeded 30 μmol. Recency, which is an indicator of short-term memory involving the retention of the most recently presented information, was 50% lower in patients with blood levels of 35–40 μmol compared to those with 20 μmol. Ultimately, these studies suggest that CBZ impacts short-term memory.

Levetiracetam

LEV is one of the relatively newer ASMs, and it works by inhibiting Ca²⁺ channels and SV2A, a vesicle protein in the synapse.^{Reference Zwierzyńska and Pietrzak33} This protein enables neurotransmitters such as glutamate to be packaged into vesicles and released into the synaptic cleft, thereby inducing excitation in postsynaptic neurons. It also has some effects that protect neurons from degeneration, such as preventing the degradation of membrane lipids and reducing neuronal inflammation. A temporary effect on emotional, fear-based memory has been observed in mice exposed to both low and high doses of this drug. However, this effect was not seen with repeated, long-term exposure. Conversely, long-term administration appears to disrupt memory formation or encoding. In humans, multiple clinical studies have demonstrated that long-term LEV treatment is not linked to significant memory impairment and may even be associated with stability or improvement in certain cognitive areas. In a one-year follow-up of patients taking LEV as an adjunct therapy, enhancements were observed in prospective memory, working memory, attention and verbal fluency.^{Reference López-Góngora, Martínez-Domeño, García and Escartín34} In patients with newly diagnosed epilepsy on LEV monotherapy, no deterioration in neuropsychological functions was recorded.^{Reference Koo, Hwang and Kim35} Additionally, an observational study involving 403 patients on LEV monotherapy reported subjective and objective cognitive gains in a substantial number of individuals.^{Reference Helmstaedter and Witt36} The author, SNA, however, believes that some of the memory improvements associated with LEV may be due to discontinuing another ASM, rather than an independent memory-enhancing effect.

Topiramate

Topiramate is used to treat both focal and generalized seizures, and it works by inhibiting Na⁺ and Ca²⁺ channels, enhancing chloride flow through GABA receptors and blocking ion channels activated by glutamate.^{Reference Marino, Pakhomov and Han37} It can inhibit AMPA receptors, but it does not directly affect NMDA receptors. However, inhibiting AMPA receptors prevents the depolarization threshold from being reached, which stops NMDA channels from producing full excitation.

Topiramate has generally been shown to cause deficits in working memory, attention and verbal fluency. About 11–20% of patients treated with this drug report negative effects on their cognition, with the primary complaint being difficulty in “word-finding.” It is hypothesized that these deficits occur due to the drug’s impact on the frontal lobe and executive function. List learning tests present patients with a list of common words and then ask them to recall as many as possible.^{Reference Thompson, Baxendale, Duncan and Sander38} Testing for recall immediately after presentation assesses short-term verbal memory, while recalling after subsequent trials evaluates long-term verbal memory. Patients who took this drug have demonstrated declines in performance between the first and second sessions compared to untreated patients, who showed improvements. It has been shown that tests involving verbal fluency are greatly affected by this drug, while nonverbal memory tasks are less affected. Therefore, short-term verbal memory appears to be the most negatively impacted. These effects were observed after approximately three months of administration, indicating a relatively short-term impact. No significant correlations have been found between memory function and the dosage of this drug. These adverse effects were shown to be mitigated by employing a longer titration rate, where the dosage is increased gradually by around 25 mg weekly.^{Reference Meador, Loring, Hulihan, Kamin and Karim39}

Benzodiazepines

BDZ are a class of psychoactive drugs with sedative, hypnotic and anticonvulsant effects.^{Reference Ochoa and Kilgo40} The ones commonly used for reducing seizure activity are diazepam, lorazepam, midazolam and clonazepam. They work mainly by acting directly on specific subunits of the GABA_A receptor. They are lipid soluble and diffuse rapidly across the blood-brain barrier, which means they have a quick effect when administered. Significant cognitive impairment has been shown with long-term BDZ administration, particularly in episodic memory, executive function and attention.^{Reference Zetsen, Schellekens, Paling, Kan and Kessels41} Smaller effects have been noted for working memory and verbal reasoning. A study involving 92 patients taking diazepam for around 3 months found 20.7% to be cognitively impaired but did not find these to be affected by dosage. Additionally, lorazepam may have some negative effects on perceptual priming.^{Reference Danion, Weingartner and File42}

VNS

VNS has shown to reduce seizure frequency by more than 50% in 25% of patients.^{Reference Lund, Kostov, Blomskjøld and Nakken51} There has been no notable difference reported in executive and cognitive performance or attention following VNS treatment compared to pretreatment.^{Reference Kong, Zhao and Zeng52} However, a longer treatment duration at moderate intensity has been shown to enhance cognitive performance and memory function in patients relative to acute stimulation. Some studies indicate that treatment during memory recollection specifically can have positive effects on memory. Performing VNS alongside the execution of a cognitive task can lead to increased blood flow to the prefrontal cortex, thereby improving cognitive function. The rate of blood flow increases proportionally with the stimulation intensity, which can be particularly important for rehabilitation purposes.

DBS

This technique involves stimulating the anterior nucleus of the thalamus (ATN-DBS) or hippocampus (Hip-DBS) through a device implanted in the chest.^{Reference Kim, Jung, Lee, Joo, Seo and Shon53} Although both the hippocampus and ATN-DBS are vital for memory and cognition, no significant changes in these functions have been reported following stimulation. One possible explanation is that ATN-DBS increases glucose metabolism in the thalamus and both hippocampi via thalamo-hippocampal connections, as demonstrated in animal studies. This modulation across both sides of the system may help preserve cognitive function. Conversely, the reduction in seizure activity due to Hip-DBS supports the normal functioning of brain networks, helping memory and other cognitive abilities remain stable. Both ATN-DBS and Hip-DBS might improve visuospatial recognition memory, although further research is needed to confirm this. Overall, DBS appears capable of decreasing seizure activity without significantly impacting cognitive functions.

RNS

RNS is a relatively newer closed-loop stimulation device used in managing medically refractory epilepsy. Some improvement in verbal memory has been observed after treatment; specifically, individuals who undergo hippocampal RNS may experience greater enhancement compared to those with neocortical RNS.^{Reference Malaga, Modiano and Haneef54} Timing this stimulation in the lateral temporal cortex with a pattern of brain activity associated with encoding can lead to improved overall memory. The strength of the connection between the stimulated area and the memory network can also influence this outcome, as stimulation alters the slow brain rhythms shared by these regions. Overall, studies have shown either improvements or no changes in memory, attention and executive functioning following RNS.

Cognitive rehabilitation

Cognitive rehabilitation involves various strategies, techniques and exercises that memory-impaired patients can use to manage or reduce the effects of cognitive deficits.^{Reference Ponds and Hendriks55} This treatment has proven effective in maintaining or improving verbal memory after epilepsy surgery.^{Reference Joplin, Stewart, Gascoigne and Lah56} Patients who received this therapy post-surgery demonstrated better verbal memory scores compared to those who did not. When designing a memory rehabilitation program, it is essential that patients with epilepsy are first and foremost well informed about the possibility of memory impairment by their medical providers.^{Reference Ponds and Hendriks55} This includes, but is not limited to, the structural effects of the disorder and the pharmacological effects of ASMs. A basic overview of memory should also be provided. Patients and their families often lack sufficient knowledge about these topics, so it is crucial to address their questions to reduce confusion. It is also important to consider any individual changes in personality or emotional stability that may accompany epilepsy. Moreover, the severity of a memory deficit, as measured by a cognitive test, may differ from the impact it has on an individual’s daily life. Patients’ beliefs and perceptions about their memory can influence their performance on memory tasks and, consequently, their progress or lack thereof through rehabilitation. Memory rehabilitation strategies can be classified into two primary categories: external and internal strategies (Figure 5). External strategies include techniques used to store information physically or digitally, remind individuals to perform tasks later or modify the external environment to depend less on memory. The first involves tools like calendars, agendas, diaries or phones to note important information. The second refers to using alarms or apps to provide reminders, while the third involves labeled or colored drawers for everyday items to avoid relying solely on memory. The two types of internal memory strategies are those that use verbal or visual imagery. An example of verbal strategies is the PQRST method for remembering reading material. It involves previewing the material (P), generating questions (Q), reading and answering these questions (R), verbalizing or recalling the answers (S) and later testing oneself (T). Visual imagery strategies include techniques like forming associations between names and faces, such as visualizing Bill King wearing a crown, which links the image to his name. Another visual technique is the Loci method, which helps remember lists or information by visualizing familiar places and mentally placing items along a path through them. For instance, to recognize a shopping list, one might imagine bananas at the front door, spilled milk on the hallway floor and bread on the bed. To recall the items, the individual mentally walks through this familiar place, with each location serving as a cue for the corresponding item.

Figure 5.

Types and subtypes of compensatory memory strategies used in cognitive rehabilitation.

An established epilepsy-specific cognitive rehabilitation program is HOBSCOTCH (Home-Based Self-management and Cognitive Training Changes Lives), which focuses on enhancing memory and executive function through education, metacognitive strategies and personalized goal setting.^{Reference Streltzov, Schmidt and Schommer57} Participation in HOBSCOTCH has been linked to improvements in subjective cognitive function and quality of life in adults with epilepsy.

Lifestyle interventions

As previously noted, sleep is crucial for memory consolidation. Therefore, maintaining a consistent sleep schedule is recommended for epileptic patients at risk of developing memory problems. Adequate rest can also reduce fatigue, thereby improving cognitive performance. The “5 principles” of good sleep health involve valuing, prioritizing, personalizing, trusting and protecting one’s sleep.^{Reference Espie58} These principles can be applied by viewing sleep as essential, making it a priority and trusting and responding to the body’s natural cues for rest. Moreover, understanding your individual sleep pattern, the number of hours needed to function effectively and the optimal window for falling asleep is essential. This can be determined through trial and error, with monitoring by keeping a sleep journal. It is advisable for patients to reflect on and record their sleep and wake times, as well as overall sleep quality, to identify any patterns. Finally, safeguarding sleep might involve avoiding screen time and caffeinated drinks before bed.