Neuromodulation of the Centromedian Thalamic Nuclei in the Treatment of Generalized Seizures and the Improvement of the Quality of Life in Patients with Lennox–Gastaut Syndrome


Address correspondence and reprint requests to Dr. A.L. Velasco at Cerrada Bosques de Moctezuma 55, La Herradura, Huixquilucan, Estado de México, CP 52780, México. E-mail:


Summary: Purpose: Our aim was to evaluate the efficacy of ESCM (electrical stimulation of the centromedian thalamic nucleus) in treatment of generalized seizures of the Lennox–Gastaut syndrome (LGS) and improvement of patient disability.

Methods: Thirteen patients with LGS were studied. They had severe generalized tonic–clonic seizures (GTC) and atypical absences (AA). All patients had at least a 6-month baseline before bilateral electrode implantation to the centromedian (CM) nuclei of the thalamus to undergo therapeutic ESCM. Once implanted, electrodes were temporally externalized through a retromastoid point for electrophysiologic confirmation of their placement. After target confirmation, stimulation parameters were set. Patients came for follow-up assessment of seizures and neurophysiologic tests every 3 months during an 18-month period of time; AED therapy was not modified.

Results: The surgical procedure as well as electrical stimulation was well tolerated by all patients. No side effects occurred with the therapeutic stimulation parameters used, and patients were not aware of device activation. Two patients were explanted because of repeated and multiple skin erosions that could not be controlled by plastic surgery procedures. Overall seizure reduction was 80%. The three patients with poorest outcomes for seizure control did not improve their ability scale score. In contrast, the two patients rendered seizure free are living a normal life at present. The remaining eight patients experienced progressive improvement, from being totally disabled to becoming independent in five cases and partially dependent in two. Patients with adequate electrode placement had a seizure reduction >87%. To consider that an electrode is correctly placed, both stereotactic placement and neurophysiologic responses are taken into account.

Conclusions: ESCM provides a nonlesional, neuromodulatory method with improvement in seizure outcome and in the abilities of patients with severe LGS.

Lennox-Gastaut syndrome (LGS) is one of most severe forms of childhood epilepsy. It is characterized by drug-resistant generalized seizures accompanied by mental deterioration. Peak onset is known to be between 1 and 7 years of age. The electroencephalogram shows characteristic slow spike–wave complexes (<2.5 per s) and bursts of rapid (10-Hz) rhythms during slow sleep. Overall prognosis is very poor; 90% of patients are mentally retarded, and >80% continue to experience seizures throughout adulthood (1–3).

Several surgical procedures have been proposed to treat generalized seizures, such as corpus callosum section (CCS). The degree of significant improvement reported for CCS is 60.9% (4). CCS may be associated with some postoperative neurologic residual effects such as lateralized cerebral deficit (5). CCS has been reported to be particularly efficient in controlling drop attacks (4,6,7) and other varieties of atypical absences (AAs) (8), but it does induce a less persistent effect on generalized tonic–clonic (GTC) seizures (9).

As for electrical stimulation, different targets have been used. Cerebellar stimulation (CS) significantly decreases GTC, tonic seizures (43 and 41%, respectively) and AA, although no reports exist on its efficacy in LGS (10,11). Vagal nerve stimulation (VNS) has also been used in LGS, showing a 42% global seizure reduction, with no patients becoming seizure free (12–14). The anterior nucleus of the thalamus has been stimulated for treatment of partial epilepsy with secondarily generalized seizures (15), showing significant improvement with respect to severity and frequency of secondary generalized seizures; no reference is made regarding LGS. Centromedian thalamic nucleus stimulation (ESCM) has been reported as effective in treatment of generalized intractable seizures (16); the best results were obtained in five patients with LGS.

The ESCM procedure for LGS patients has been refined progressively by increasing the number of treated patients and lengthening the follow-up period. In this study, a number of predictors to improve the outcome of ESCM-treated patients treated have been analyzed.


Thirteen patients with LGS were included in this study; all fulfilled diagnostic criteria. Table 1 shows their clinical characterization. Ages varied from 4 to 22 years. All patients had seizure onset before age 7 years. Seven patients were idiopathic cases with normal magnetic resonance imaging (MRI). Six had abnormal MRIs: one had right frontal dysgenesia from birth (patient 2), another had herpetic encephalitis–associated bitemporal encephalomalacia (patient 3), another had left hemisphere atrophy from birth (patient 4), and the remaining two patients had nonevolving tuberous sclerosis (patients 7 and 11). All patients had GTC seizures and AAs. One patient (patient 9) experienced atypical absences and complex partial seizures; two patients (patients 2 and 3) also had long periods of nonconvulsive status, whereas one patient (patient 10) had four convulsive statuses per year. Seizure number per month varied from 35 to 4,300. All patients had typical electroencephalograms (EEGs) showing generalized slow spike–wave complexes. Three patients had a history of hepatic failure, and one had valproate (VPA) use-associated pancreatitis. These patients represented a high risk for major surgical procedures.

Table 1. Clinical characterization of 13 selected patients with Lennox–Gastaut syndrome
PatientAge (yr)MRISeizure typeSz/mo before ESCMAED side effects
  1. GTC, generalized tonic–clonic; AA, atypical absence; CXP, complex partial seizure; AED, antiepileptic drug; SK-W, slow spike–wave complexes; N, normal; RF DYSG, right fontal dysgenesia; BITEMP, bitemporal encephalomalacia; L ATR, left hemispheric atrophy; TbSc, tuberous sclerosis; CB INF, cerebral infarct. Sz/month before ESCM calculated by averaging the 3-month baseline immediately before starting ESCM; 4,300/NCs, 4,300 plus several nonconvulsive statuses per year; 50/Cst, 50 seizures plus several convulsive statuses per year.

1 8NYY3119Not determined
2 7RF DYSGYY4300/NC stHepatic failure
3 9BITEMPYY3780/NC stHepatic failure
413L ATRYY3030Not determined
511NYY1200Not determined
621NYY787Drowsiness ataxia
721TbScYY2576Drowsiness ataxia
822NYY168Drowsiness ataxia
919NYY56Drowsiness ataxia
1010NYY50/C stHepatic failure
11 4TbScNY150Drowsiness
1213NYY35Drowsiness ataxia
1313CB INFNY50Pancreatitis

Study design

The protocol was revised and approved by the Scientific and Ethical Committees of the General Hospital of Mexico, Mexico City. The Ethical Committee did not approve the double-blind maneuver proposed in the protocol, considering that in these critically ill patients, turning the stimulator OFF may jeopardize their clinical condition. All families were informed and signed written consent. The study was divided into five consecutive stages including baseline, electrode implantation, target confirmation, system internalization and long-term stimulation-parameter setup, and long-term follow-up.


Baseline period consisted of a 6-month follow-up before implantation. During this time, the patient's relatives were trained to keep a seizure diary for recording seizure occurrence for both seizure type and frequency. Baseline follow-up included monthly EEG recording and antiepileptic drug (AED) blood-level monitoring. Number of seizures per month referred to as baseline was estimated by averaging the 3-month seizure count immediately before implantation.

Electrode implantation

Deep brain stimulation (DBS 3387, Medtronic, Minneapolis, MN, U.S.A.) electrodes consisting of a Silastic tube with four contacts numbered from 0 (the distal) to 3 (proximal) made of platinum rings separated by a distance of 3 mm each along the axis of the electrode (DBS 3387; Medtronic, Inc., Minneapolis, MN, U.S.A.) were implanted by following the technique described by Velasco (17). Patients were operated on under general anesthesia, and electrodes were stereotactically inserted through frontal parasagittal burr holes placed 13 mm at each side of midline and immediately behind the coronal suture. With ventriculograms, electrodes were guided so that the electrode tip was placed at the intersection of anterior commissure–posterior commissure line (AC-PC) and vertical line passing at anterior border of posterior commissure (VPC) lines in the lateral radiogram and at a distance of 10 ± 2 mm at each side of midline in the AP radiogram (Fig. 1). Electrode trajectories had an angle from 45 to 60 degrees with regard to the AC-PC line. Electrodes were fixed by using a skull ring and cap system and temporally externalized through a retromastoid point for electrophysiologic confirmation of their placement. Recently, we also performed transoperative confirmation of the target by eliciting electrocortical responses and regional direct current (DC) shifts (see later). To obtain adequate transoperative recordings, we requested that the anesthesiologist bring the patient to a superficial anesthetic plane, adding muscular relaxants if necessary.

Figure 1.

Stereotaxic placing of the centromedian (CM) electrodes. A: Stereotaxic diagram modified from Schaltenbrand and Bailey showing the optimal stereotaxic target. CM localization is accomplished by air ventriculography. This method demonstrates anterior (AC) and posterior (PC) commissures of the third ventricle. Two lines are drawn, the AC-PC line and the vertical line perpendicular to the PC (VPC). The target point for the electrode tip was a distance 10 mm from midline and the intersection of the AC-PC line with the VPC. CM delineated by discontinuous line, optimal target delineated by the continuous line. B: Sagittal MRI showing right CM electrode implanted in patient KCMM12. The electrode has four contacts, two of which are chosen for ESCM.

Target confirmation

Postoperative imaging and neurophysiologic methods were used to confirm correct electrode placement. With regard to the imaging method, MRI was performed on day 1 after implantation by using 1.5-T Edge equipment software version 9.3 (Marconi Medical Systems, Cleveland, OH, U.S.A.) using T2-weighted fast-spin echo sequence (echo time, 11 ms; repetition time, 4,070 ms; field of view, 16.0 cm; 256 × 256 matrix), 2.5-mm-thick sections, without intersection space, oriented parallel and perpendicular to AC-PC line for axial and frontal views, respectively, and parallel to midsagittal plane for sagittal sections (17,18) (Fig. 1B). By using ventriculogram films and MRI studies, the position of each electrode contact was plotted on sagittal and frontal sections of the Schaltenbrand and Wahren Atlas (19) (Fig. 2).

Figure 2.

Plotting of pairs of contacts of electrodes used for prolonged bipolar electrical stimulation on sagittal (SI 9,0) and frontal (Fp 9,0) sections of the Schaltenbrand and Wahren anatomic atlas. A and A' plotting of cases with >80% decrease of seizures (good outcome). B and B' plotting of cases with <80% improvement (poor outcome). Note that whereas electrodes are grouped in the anterolateral part of centromedian nuclei (CM; parvocellular area) in cases with good outcome, electrodes are dispersed toward more posterior, superior and medial regions of CM (magnocellular, internal medullar laminae, and parafascicularis) in poor-outcome cases. AC, anterior commissure; PC, posterior commissure; VPC, vertical line posterior commissure; ML, midline; Ce, centromedian nucleus; mc, magnocellular; pc, parvocellular; Mfip, fibrosus posterior; Lam, lamina medullaris: Lm, medial lemniscus; M, territorium medialis thalami; Pf, parafascicularis; Pu, pulvinar: Raprl, prelemniscal radiations; Vci, ventralis caudalis internus; Vce, ventralis caudalis externus; Vcpc, ventralis caudalis parvocellularis; Vimi, ventralis intermedius; Vop, ventralis oralis posterior.

Neurophysiologic confirmation of the CM target was performed by 2-h trial of acute stimulation on days 2–3 after implantation for localization purposes. Three pairs of combinations of the four contacts in each electrode were stimulated (0–1, 1–2, and 2–3). Stimuli were delivered by a Grass S8 stimulator (Grass Instruments, Quincy, MA, U.S.A.) and isolation unit attached to the patient by means of a Tektronix CRU (Tektronix, Beaverton, OR, U.S.A.) and a comparative 10-kΩ resistor to monitor voltage (V), current flow (μA), and impedance (kΩ) of stimulated contacts within the brain tissue (20).

Electrocortical potentials in the form of recruiting responses and desynchronizing responses were elicited by unilateral electrical stimulation through adjacent electrode contacts (in which the cathode was always the lower contact). Unilateral stimulation was conducted at 6 Hz, 1.0 ms, 320–800 μA with 10-s duration to induce recruiting responses, and unilateral high frequency was conducted at 60 Hz, 1.0 ms, 320–800 μA to induce focalized desynchronization and negative DC shift of the EEG baseline. Scalp distribution analysis of electrocortical responses was carried out from EEG recordings taken from frontopolar (FP2, FP1), frontal (F4, F3), central (C4, C3), parietal (P4, P3), occipital (O2, O1), and frontotemporal (F8, F7), and anterior temporal (T4, T3) referred to ipsilateral ears (A2, A1) (sensitivity, 10 μA/cm; time constant, 0.35 s; paper speed, 15 mm/s) (Fig. 3).

Figure 3.

Intraoperative neurophysiologic confirmation of the centromedian (CM) target. A: Low-frequency (six per second) threshold stimulation of the parvocellular portion of the right CM elicits recruiting-like responses showing unilateral regional scalp distribution with maximal amplitude at the frontal region ipsilateral to the stimulated side; these consist of biphasic positive/negative potentials with peak latencies of 16 and 60 ms, respectively. Suprathreshold stimuli induced bilateral potentials with emphasis on the ipsilateral frontal region (not shown). B: Unilateral high-frequency (60 per second) threshold and suprathreshold stimulation of the parvocellular portion of the right CM elicit a regional EEG desynchronization consisting of a slow negative shift with bilateral regional scalp distribution with maximal amplitude at frontal region ipsilateral to the stimulated side. Note spontaneous two-per-second spike–wave riding on the DC shift.

System internalization

On day 4, the system was internalized by connecting the electrode to an extension cable and to a subcutaneous (s.c.) pulse generator (ITREL II or III IPG; Medtronic). Once the stimulation system was internalized, the following stimulation paradigm was used: alternating 1-min stimulation on each side with a 4-min interval between right and left sides during 24 h. Stimulation parameters were 130 Hz, 0.45 ms, 400–600 μA., with a charge density <3.0 μC/cm2/phase according to the formula referred by Velasco et al. (10).

Long-term follow-up period

For analysis purposes, results were evaluated on month 18 in all patients. During this 18-month period, patients came every 3 months to have an EEG, to check their stimulation system, and to turn in their seizure diary. During this time, patients received a fixed dose of AEDs that was not modified except in cases in which a useful AED had toxic effects (in patients 2, 3, 10, and 13), and the dose of the anticonvulsive was decreased to therapeutic ranges.

Every 3 months, patients had a session of electrical stimulation by means of the totally implanted system, setting up IPG transcutaneous stimulation parameters. Both 6–8/s and 60/s pulses, 6–8 V, 450 μs were delivered to induce electrocortical responses and desynchronization and DC shifts, respectively; on the scalp, EEG recording referred to the ipsilateral ear. This maneuver is important to determine electrophysiologic response integrity to ESCM (18,21,22). After 18 months, patients returned once a year, and the majority of patients returned only when stimulation was discontinued because of battery depletion and seizure reappearance.

Evaluation of seizure reduction and ability scale

Improvement in seizure number was determined for total seizure number and separately for GTC and AA. Significance in seizure number decrease from the baseline period to the end of the 18-month follow-up period for the whole group was determined by repeated measures analysis of variance (ANOVA) test.

Seizure-control impact on the patients' psychomotor condition was estimated by means of the scale of abilities, which considered the following: 0, totally disabled (i.e., patient unable to rise, walk, feed, or communicate, being totally dependent); 1, partially dependent, patient who needs assistance for the majority of his daily life activities (dressing, walking, bathing, going outdoors) and who has restricted verbal communication; 2, partially independent (i.e., the patient is self-sufficient and able to help in some housekeeping activities but needs supervision to go outdoors); 3, independent, possessing a low-profile work requirement or attendance at a special education school; and 4, is able to attend a regular school supported by special education to improve performance (23,24). Scores were processed by a Wilcoxon signed rank test for significance.


Safety and side effects

The surgical procedure as well as electrical stimulation was well tolerated by all patients. No patient had evidence of postsurgical hemorrhage or edema in the control MRI. No side effects were found with the therapeutic stimulation parameters used; patients rested were unaware of device activation. Two patients (patients 2 and 3) were explanted at months 20 and 54 because of repeated and multiple skin erosions that could not be controlled by plastic surgery procedures.

Patient selection and seizure outcome

Follow-up ranged from 23 to 132 months (average, 46 months). After 6 months of ESCM, all patients had a stable seizure reduction throughout the 18-month follow-up period included in this study; after this, patients continued to be stable throughout their entire follow-up, and only patients who had battery depletion or had to be explanted because of skin erosions had seizure increase. These cases are discussed later in this work.

Figure 4 shows total seizure, GTC seizures, and AA reduction for the group. Significance in seizure reduction from baseline to the 18-month follow-up was <0.0001. Nonconvulsive and convulsive, recurrent, epileptic-status episodes that occurred more than once per year in three patients never recurred during the 18- to 54-month follow-up period.

Figure 4.

The three graphs show seizure average for the entire group of Lennox–Gastaut syndrome (LGS) patients for 1-month baseline and 18 months with ESCM (arrow, where stimulation starts). The upper graph demonstrates seizure reduction for all seizure types; the generalized tonic–clonic (GTC) graph shows generalized tonic–clonic seizures; and AA graph, atypical absences. Note that seizure reduction is observed from month 1, but best results are observed after month 6.

EEG changes have been previously described in Velasco (20). In all patients, ictal and interictal EEG discharges number decreased, although these disappeared only in the two patients who had 100% seizure relief.

Table 4 shows the ability scale scores, which demonstrated significant improvement (p < 0.04). Improvement in seizure number was concordant with ability scale scores. The three patients with poorest outcomes for seizure control did not modify their ability scale score. In contrast, the two patients rendered seizure free are living a normal life at present. These two patients had sporadic seizures from age 3 months but were controlled with AEDs at the time and had a normal performance at school. It was not until they were age 6–7 years that they had seizure increase without response to AEDs and dramatic mental deterioration; when the seizures stopped after ESCM, they progressively regained their condition before seizure increase and, with the help of special education, returned to school within 1 year. At present, after 93 and 60 months of follow-up, respectively, these two patients are leading a normal life. The remaining eight patients had an overall ability scale score improvement of >2 points, becoming independent or partially independent.

Table 4. Improvement of ability scales after 18 months of ESCM
PatientAge (yr)Onset (yr)% Sz reductionAbility scale before ESCMAbility scale after ESCM
  1. Points of improvement were assigned according to independence in daily life activities and ability to work or study (see text). The group as a whole showed significant improvement (p = 0.004). Patient ages at ESCM and seizure onset are included. Note that patients with seizure onset at older ages and greater seizure reduction have better improvement in their abilities.

1 8710004
2 7610004
3 97 7013
4135 8003
5112 5002
6212 7003
7213 8003
8224 8003
91910 mo 6012
10105 mo 7002
11 42 mo  011
12134  022
13137  011

Stereotactic electrode placement confirmation

Figure 2 shows stimulated electrode contacts plotting, dividing patients according to stimulated contact positions into two groups; 2A and 2A' show nine patients with >80% seizure relief, whereas 2B and 2B' show four patients with <79% seizure relief. Position of the pair of contacts used for prolonged stimulation that correspond to group A cases were in general more lateral and/or anterior in CM, covering the area labeled in the stereotactic atlas as parvocellular subnucleus. Group B cases were placed in a more dorsal, posterior, and medial position, with some contacts outside of CM, which were considered incorrectly placed.

Electrocortical responses produced by acute electrical stimulation along the CM or other structures are described elsewhere (25). Low-frequency (six per second), threshold (4–5 V = 320–400 μA) unilateral threshold stimulation of the CM parvocellular portion elicits recruiting-like responses with unilateral or bilateral regional scalp distribution with maximal amplitude at ipsilateral frontal region (Fp-F). They are biphasic positive/negative potentials with peak latency of 16 and 60 ms, respectively (Fig. 3A). Suprathreshold stimulation induced monophasic negative, long-latency (42 to 60 ms) waxing and waning responses with bilateral distribution, and higher amplitude.

Unilateral high-frequency (60 per second) threshold and suprathreshold stimulation of the right CM parvocellular portion elicited regional EEG desynchronization consisting of slow negative shift with bilateral regional scalp distribution and maximal amplitude at the frontal region ipsilateral to the stimulated side. Spontaneous two-per-second spike–wave may ride on the DC shift (Fig. 3B). These types of electrophysiologic responses were considered evidence of correct placement within the CM (25). The other type of evoked thalamocortical potentials obtained was the short-latency positive potentials and biphasic positive-negative potentials, with distribution toward more posterior and lateral cortical regions that were labeled as incorrect.

As shown in Table 3, all patients in group B (with <80% seizure reduction) did not fulfill the study's anatomic and/or electrophysiologic criteria; conversely, seven of nine patients in group A did fulfill both criteria. Of the two incorrect placements, patient 3 had incorrect placement of the left electrode, but adequate electrocortical responses, and patient 1 had the left electrode well placed.

Table 3. Stereotactic placement and electrophysiologic predictors and seizure relief obtained after 18 months of ESCM
PatientStereotactic placementElectro physiologic% Seizure reduction at
  1. C, correct; and I, incorrect stereotactic placement and localization of neurophysiologic responses (see text for details).

3CICC 95
4CCCC 95
5CCCC 95
6CCCC 91
7CCCC 89
8CCCC 87
9CCII 79
10CCCI 70
11CICC 58
12CIII 53
13ICII 30

Significant differences of correct and incorrect placements between groups A and B were determined by Fisher's Exact test that had a significance of p < 0.002 with a predictor factor of 90%. Thus to consider that an electrode is placed in a site that could render an >80% seizure reduction, both stereotactic placement and neurophysiologic responses should fulfill the criteria.

On/Off ESCM Condition

After an initial 18-month period, ESCM was discontinued temporarily or permanently in four cases for the following reasons: in patient 1, batteries were depleted once on right side at month 24 (first OFF period) and on left side, at month 55 (second OFF period). During OFF periods, seizures increased, and during ON period as well as after battery replacement, the patient became seizure free. Two patients (2 and 3) were explanted because of repeated and multiple skin erosions that could not be controlled by plastic surgical procedures. In the latter of these two patients (patient 3), seizures increased in OFF period beginning soon after explantation (Fig. 5). For the remaining patient (patient 2) who had become seizure-free without medication during 54 months of ESCM, with EEG normalization, explantation was not followed by seizure reappearance, but normalized EEG began to present 2.5-per-second spike–wave complexes similar in morphology to those observed during baseline period.

Figure 5.

Three patients in whom ESCM was temporally (patients 1 and 5) or permanently (patient 3) discontinued. The fourth patient is not shown because his seizures have not reappeared after 5 months without ESCM. X, When discontinued ESCM became evident, and arrows demonstrate when ESCM was initiated or reinitiated. Note how seizures increase progressively when ESCM is stopped without ever reaching baseline numbers. When stimulation is reinitiated, seizures decrease again (see text for broader explanation).

Another patient (patient 5) experienced two accidents related with seizure occurrence that ended with rupture of electrode lead and connector cable at the neck at 19 months after ESCM onset and scalp wound with fracture of DBS extracranial portion at month 31. Electrode replacement, extension cables, and reinitiating stimulation were followed by regaining seizure improvement.

Under the previously mentioned conditions, we averaged seizure number during the 3-month ON period before stimulation arrest and 3-month OFF stimulation periods, and found that statistical difference in seizure number between ON and OFF periods was significant (p = 0.006) as well as between OFF and baseline periods, because seizure number in the OFF periods never reached that of baseline (p = 0.002) (Fig. 5). In addition, we noted that the longer the patient has been stimulated, the shorter the time necessary for improvement after reinitiating stimulation (Fig. 5).


The present study shows major limitations due to the inherent problems of this particular LGS population. Patients demonstrated severe baseline cognitive impairment, which renders several tests difficult to perform, such as standardized neuropsychological tests. Seizure severity and frequency made double-blind studies ethically unfeasible. However, consistency of the observations, number of cases, and length of follow-up allowed us to develop some conclusions.

Seizure control and ability scale outcomes

The present study extends previous experience in treatment of generalized seizures by ESCM (16,17). It confirms that both GTC and AA of the LGS may be significantly reduced and even abolished by this therapy, and that the beneficial effect persists as long as ESCM is maintained. The most severely affected patients (i.e., those with the greatest number of seizures per month and the most severe psychomotor deterioration) were those who responded best to ESCM.

CCS is a more-invasive procedure that produces a 60% seizure reduction but that would have represented a high surgical risk in some of our patients with high-dose VPA-related liver damage, leading to abnormal coagulation tests. The figures are considerably better than those reported for similar LGS with VNS (12–14) and cerebellar stimulation (CS) (10–14,22,25).

Nonetheless, when compared with other minimally invasive procedures for seizure control such as VNS and CS, ESCM is in general more elaborate, requiring the participation of qualified stereotactic surgeons and clinical neurophysiologists. ESCM probably also possesses higher degrees of complications than VNS. In our series, two patients required explantation 20 and 54 months after initial implantation, respectively, because of recurrent skin erosions in the stimulation systems trajectory. These patients were among the youngest in our series, and even when the operative technique included the precaution of placing the IPG under muscle fascia in the abdomen and making s.c. trajectories in scalp, neck, and chest as deep as possible, increase in patient size with consequent extension cable, increase in physical activity, and the relatively large size of stimulation systems for small children may be some factors leading to skin erosion. From this point of view, current hardware for neurostimulation appears inadequate for children, and modern nanotechnology already under clinical trial may allow extending the indication for ESCM or other neurostimulation techniques to a larger number of pediatric cases.

Neuropsychological evaluation of this group was particularly difficult in view of their deteriorated condition; several patients were in nonconvulsive status, which made it impossible to apply a battery of standardized psychological tests. Nevertheless, the ability scale used in this study shows that no patient had signs of neurologic or mental deterioration during ESCM. It appears that patients with normal development before LGS onset tend to regain their abilities regardless of convulsive syndrome severity. Added to this observation, it is possible that an early surgery could also play an important role in favoring the psychomotor outcome (patients 1 and 2). The reason for this improvement is uncertain. Nevertheless, because increase in abilities correlates with seizure reduction, this was probably the major factor underlying performance improvement. Because AEDs were maintained at the same doses as for baseline, changes related to anticonvulsive side effects may be ruled out. In previous studies, we analyzed the manner in which performance improvement of ESCM-treated patients may also be associated with increase in the frequency of background EEG activity frequency (20).

On/Off ESCM conditions

With the exception of one, all patients in whom stimulation was discontinued experienced seizure increase, seizure reappearance, or seizure number increase. The patient without seizures within 5 months after explantation had reappearance and progressive increase of 2.5 spike–wave complexes in EEG. These observations indicate that seizure control is indeed because of ESCM and not for spontaneous fluctuations in seizure occurrence. In addition, the probability of achieving spontaneous improvement time-locked with ESCM onset and maintenance in cases of patients with LGS is very remote. The expected outcome for this type of patient has been repeatedly documented as poor for both seizure occurrence and mental conditions in large series with 10- to 20-year follow-up periods (1,2,16).

Conversely, because seizure numbers during 3- to 5-month OFF periods were significantly lower than those of baseline period (p = 0.02), this suggests that prolonged ESCM induces a long-lasting residual antiepileptic effect. We have already observed this effect in a previous study (16) in which a 3-month OFF stimulation period was analyzed. Moreover, reinitiating stimulation after an OFF period that occurred in chronically stimulated patients reached maximal anticonvulsant effect quite rapidly (Fig. 4). These observations may indicate a plastic antiepileptic effect occurring at the stimulated site or anatomically related structures after long-term ES.

Stimulation mode and parameters

The electrical stimulation cycling mode of the nervous tissue was originally proposed to avoid electrical current overcharge in areas under or around electrodes (26,27). Discontinuous and cycling ES have been successfully used in treatment of pain (28,29) and epilepsy (11,30,31). Although the main reason for using this mode of stimulation has been to save battery charge, its efficacy indicates that the beneficial effect outlasts each stimulation period. Although maximal antiepileptic effect is usually reached after 4–6 months of ON stimulation, progressive decrease in seizure number was documented from the first month; therefore a residual antiepileptic effect is probably present from the beginning of the stimulation trial.

Conversely, although adjustment of stimulation parameters is a relatively easy task when treating motor disorders because the effect on symptoms is readily recognized, setting parameters for ES in epilepsy becomes much more complicated. Information derived from animal experiments has pointed out the importance of setting stimulation parameters that take into account principally the charge density-per-phase, which for safety's sake should not exceed 4 μC/cm2/phase (31,32). In all of our cases, pulse amplitude remained between 2.0 and 3.0 V and rarely changed along follow-up. This voltage represents between 50 and 80%, which is necessary for inducing recruiting responses and DC shifts. It is important to mention that in cases with poor outcome, increasing the voltage 2 or 3 times the average did not improve ESCM efficacy.

Target localization

Although stimulation parameters do not appear to be critical for ESCM effectiveness, the precise location of DBS electrodes within the anatomic and physiologic target is critical (16,17). Some patients presented here were operated on before we concluded that anterior lateral located parvocellular CM subnucleus is the optimal stereotactic target. For the last five patients with LGS operated on from 2000 through 2003, only one had stereotactical misplacement of the electrodes, which were posterior and laterally placed with regard to the intended target. In this case, low-frequency stimulation elicited short-latency positive responses recorded from central and parietal EEG leads, whereas high frequency induced an unpleasant sensation and limited the intensity used for therapeutic stimulation. This patient (patient 13) had the poorest outcome for the group.

Cortical incremental responses have been reported as driven by low-frequency stimulation from a number of thalamic nuclei (15,33–36), as well as from extrathalamic structures in humans (36,37); therefore they cannot be taken as evidence of a precise target localization (33). However, as first described in classic experiments (38–40) detailed analysis of incremental response morphology, polarity, peak latency, and cortical distribution may aid in defining the relation of the stimulated area with specific anatomophysiologic systems within CM. Within CM, suprathreshold stimulation in parvocellular subnucleus induces monophasic negative waxing and waning potentials, with peak latencies from 40 to 60 ms, recorded bilaterally in frontal and central regions, with emphasis on the stimulation side. Outside this subnucleus, incremental responses remain mainly biphasic positive-negative, with 16- to 20-ms latency, fixed amplitudes for the positive component, and distribution extending more toward the posterior leads. ESCM posterior and basal area within boundaries with the parafascicular nucleus induces only short-latency positive potentials at low frequency and occasionally painful sensation at high frequencies (41). Recently, transoperative monitoring of electrode placement, as described by Litt and Cranston (33) and Kerrigan (15), has been found very convenient for intraoperative correction of misplacements during the same operation (Fig. 3).


LGS is a very severe form of epilepsy that affects not only the patient but also his or her entire family. It has become feasible to provide seizure control and improvement of quality of life in this condition that is highly resistant to medical treatment. ESCM provides a nonlesional, neuromodulatory method with good outcome in the majority of cases. Quality-of-life improvement was also significant. Unfortunately, ESCM possesses some disadvantages: neuromodulation systems are expensive; they require periodic follow-up visits to verify their correct functioning, and stimulators must be replaced when batteries are depleted. At present, probably the most important disadvantage is stimulation system–induced skin erosion, which often leads to explantation and lead breakage in children because of increase in motor activity and the growth process.