Address correspondence and reprint requests to Dr. F.G. Woermann at MRI Unit, Bethel Epilepsy Centre, Evangelisches Krankenhaus Bielefeld, Maraweg 21, Bielefeld 33617, Germany. E-mail: Friedrich.Woermann@evkb.de
Summary: Purpose: By using speech-activated functional MRI (fMRI), we investigated whether the frequency of left-sided interictal epileptic activity (IED: spikes or sharp waves on the EEG) is associated with atypical speech lateralization.
Methods: We investigated 28 patients (13 men, aged 17–59 years) with left-sided mesial temporal lobe epilepsy (MTLE) and 11 patients with right-sided MTLE as a control population. Only patients with unilateral hippocampal sclerosis with unilateral IED were included. For fMRI of individual patients, we contrasted images sampled during covert word generation with a low-level rest condition. With SPM99, an individual comparison for the contrast “word generation versus resting inactivity” was conducted. To characterize speech lateralization in individual patients, we calculated asymmetry indexes (AIs): the difference between activated left-sided and right-sided voxels was divided by all activated voxels. Analyzing long-term EEG, the first 2 min of each hour were evaluated for the frequency of IED. Univariate associations with AIs were assessed by Pearson's correlation and by t test. When testing the independent associations, multivariate linear regression was performed.
Results: The AIs in patients with left-sided MTLE were 0.40 ± 0.53 on average (range, −0.83 to +1.0), whereas in right-sided MTLE, they were 0.78 ± 0.15 (p = 0.029). For the further investigations, we included left-sided MTLE patients only. The median frequency of IED was six per hour (range, 0–240). Higher IED frequency was correlated with left-right shift of lateralization of speech fMRI activity (p = 0.002).
Conclusions: Higher left-sided spike frequency in MTLE was associated with a left–right shift of speech representation, suggesting that chronic frequent interictal activity may induce a reorganization of speech lateralization.
Brain injuries in the neighborhood of speech centers can induce atypical speech lateralization, especially if they occur at an age when neuronal plasticity allows speech reorganization. Thus it is not surprising that in patients with epilepsy, speech lateralization is more frequently atypical in comparison to that in healthy people (1,2). In left-sided temporal lobe epilepsy, 23–33% of patients have atypical speech lateralization (3–5). A recent fMRI study found that in left-sided temporal lobe epilepsy, not only an intrahemispheric, but also an intrahemispheric functional reorganization of language-related neuronal networks may be present (6).
In patients with left-sided brain injury, even if the speech centers had shifted to the right hemisphere (“escaping” from the seriously damaged areas), poor verbal cognitive performance was found by Helmstaedter et al. (7), and an incomplete right hemisphere or bilateral language dominance did not protect against verbal memory loss after left-sided temporal lobe surgery (8). Conversely, Bates et al. (9) found no difference in language production between children with left-sided versus right-sided brain damage, suggesting that a complete neural and behavioral plasticity can follow early brain damage.
By using memory-activated functional MRI (fMRI), we recently found that interictal epileptic activity may influence the lateralization of memory functions independent of the influence of lesions (10). In a recent Wada test study, we found that in patients with focal epilepsy, not only the known factors (i.e., the age at which the brain injury occurred and its localization) but also the epileptic activity may influence speech reorganization (3), as suggested by a previous study conducted by Brazdil et al. (5). The Wada test, however, is an invasive procedure, and cross-flow to the opposite anterior cerebral artery may also influence the clinical effects. Thus by Wada test, we can investigate only a minority of selected patients. In contrast, fMRI is a safe noninvasive procedure. We can investigate patients consecutively, and we can analyze the speech lateralization as a continuous phenomenon, which is much closer to the biologic reality than is the categoric division used by the Wada test.
In the present study, we aimed to reproduce our previous findings in a new series of patients by using a noninvasive method, speech-activated fMRI. We investigated whether the interictal epileptic activity can be responsible for the evolution of atypical speech and whether this effect is independent of other known factors. We investigated patients with mesial temporal lobe epilepsy (MTLE) who exclusively had hippocampal sclerosis (HS) but no other epileptogenic lesions. The cause of HS is thought to be an early-childhood initial precipitating injury (IPI), which induces functional and structural damage to the hippocampus, gradually evolving to HS (11). The factors that influence speech reorganization can be investigated in MTLE because these patients have the same pathology in the same location, and this pathology is located distant from the eloquent speech areas.
Our working hypothesis was that higher spike frequency is associated with an atypical speech representation in left-sided MTLE. We also hypothesized that this correlation is independent of the timing of brain injury or epilepsy onset.
We investigated 28 patients with left-sided MTLE (13 men), who consecutively underwent our adult presurgical evaluation program from 2002 to 2004, and 11 patients with right-sided MTLE (eight men, aged 20–46 years), who served as a control population. Only patients with unilateral HS detected by high-resolution MRI and with unilateral interictal activity detected by long-term EEG were included. Outside the scanner, we evaluated the word fluency by a controlled oral word association test (COW) (12). For assessing IQ, the German version of the Wechsler Adult Intelligence Scale was used. For evaluating handedness, the Edinburgh Inventory was used.
Noninvasive continuous video-EEG monitoring
All patients underwent continuous video-EEG monitoring lasting 2–6 days as a part of their presurgical evaluation. EEG recordings with 32–64 channels were used; electrodes were placed according to the 10-10 system. Sphenoidal electrodes were used in all cases. Interictal EEG samples were automatically recorded and stored by computer. In this study, the first two minutes of each hour stored automatically by the computer were evaluated for the location and frequency of interictal epileptiform discharges (IEDs) by physicians of the epilepsy-monitoring unit blinded to the goals of this study. Only spikes and sharp waves on the EEG were considered as IEDs. For evaluation of IED frequency, IEDs over the temporal lobes were considered, but we did not make a distinction according to localization within the temporal lobe (such as posterior or anterior temporal IED).
After informed consent was obtained from each patient, we used a 1.5-T Siemens Symphony scanner (Erlangen, Germany) and a EPI sequence with 16 axial 4-mm slices (field of view, 192 mm; 64 × 64 matrix; repetition time, 1,600 ms; echo time, 50 ms; flip angel, 90 degrees), for a blocked fMRI design. In each patient, we contrasted 100 sets of images sampled during 10 episodes of covert word generation (each episode/block of phonematic verbal fluency lasted 30 s) with 100 sets of images from 10 episodes of resting inactivity (each lasting 30 s), as validated and described earlier (13). For postprocessing of this blocked fMRI investigation, we used SPM99 (Wellcome Department of Cognitive Neurology, London, U.K.). Our standardized approach across all patients consisted of image realignment, spatial normalization, and smoothing, by using SPM99 default settings including a gaussian smoothing kernel of 10-mm full width at half maximum, which aimed at maximizing the yield of signal difference in this difficult patient population. In individual patients, we conducted a fixed-effects statistical comparison for the contrast “word generation versus resting inactivity” at a threshold of p < 0.001, uncorrected. This p value corresponds to a T value >3, as used in our large validation study comparing the fMRI results with Wada test results in 100 patients with epilepsy; this series provides us with test quality data showing a high concordance, especially in patients with left-sided TLE (13).
In each subject of the current study, activated voxels were counted within the lateral three fourths of each hemisphere, by using the SPM voxel counts within the MNI coordinate system (MM). The lateral three fourths were defined according to the axial view of the glass brain provided by the SPM99 software. A previous study using AIs for language-activated MRI considered only the lateral two thirds of each hemisphere (2). We chose the lateral three fourths because most parts of the language-related neural network contributing to language lateralization are represented here. Thus we avoided counting voxels in the lateral ventricles or outside the brain, which were considered artifacts. To characterize speech lateralization in individual patients, we calculated an asymmetry index (AI): activated voxels on the left–activated voxels on the right)/all activated voxels.
Univariate associations with AIs were assessed by Pearson correlation and by Student's t test, for continuous and dichotomous variables, respectively. When testing the independent associations, multivariate linear regression was performed with AI as the response and IED, age, epilepsy onset, age at IPI, gender, performance in controlled oral-word association test, and IQ explanatory variables. Variables with skewness >1 were logarithmically transformed. The absolute value of skewness of AI was <1 (−0.94), and the Kolgomorov–Smirnov test showed that its distribution did not deviate significantly from normal distribution. Effect size was assessed by means of standardized regression coefficient values. Further to illustrate the relation of IED and speech lateralization, we performed another theoretically different statistical approach by comparing a group of left-sided TLE patients with frequent IEDs with a group of patients with left-sided TLE and a low IED frequency. The SPM99 second-level group comparison was thresholded at p < 0.05, corrected for multiple comparisons (Fig. 2).
The mean age of 28 investigated patients with left MTLE was 34.4 ± 11.9 years (range, 17–59 years); 21 were right-handed, five were left-handed, and two were ambidextrous. The mean IQ was 92.2 ± 18.4. The median age at epilepsy onset was 8 years (mean,11.1 ± 9.3 years). Data on IPI were available in 18 patients, in whom the median age at IPI was 1 year (mean, 1.53 ± 1.15 years). The median seizure frequency was four seizures/month (range, 1–24). The mean epilepsy duration was 23.4 years (range, 7–48 years). All patients had exclusively temporal IEDs. Anterior temporal IEDs occurred in all patients, and posterior temporal IEDs were present in three patients only. Dividing patients into different subgroups according to IED localization would have resulted in low statistical power; thus we did not make this distinction. The median IED frequency was six per hour (range, 0–240).
The AI in patients with left-sided MTLE was 0.40 ± 0.53 on average (range, −0.83 to +1.0), whereas in right-sided MTLE, it was 0.78 ± 0.15 (median, 0.75; range, +0.57 to +1.0), showing a significant difference (p = 0.029). For further investigations, we included left-sided MTLE patients only.
The Wada test was performed in 10 patients [our Wada test protocol has been published (3)]. Eight patients had left-sided speech lateralization by Wada test. The mean AI in them was 0.7 (range, 0.36–1.0). In two patients showing atypical speech on the Wada test, the AI was 0.1 and −0.4, respectively. Despite the small number of patients in whom the Wada test was performed, the association between the AI and the Wada test result was significant (p = 0.036).
Concerning our main question, we found that higher IED frequency was correlated with left–right shift of lateralization of speech- fMRI activity, as shown in Fig. 1 (p = 0.001).
To investigate the independent association between AI and the other variables, we performed a multivariate linear regression analysis and included IED frequency, age at epilepsy onset, gender, age, seizure frequency, Wechsler IQ, and performance in a controlled oral word-association test. We found that only the IED frequency had a statistically significant association with the AI (beta =−0.641; p = 0.002). The results are detailed in Table 1.
Table 1. Factors associated with asymmetry index of speech representation
Standardized coefficients (beta values)
aBecause of small numbers of patients who had had or had been reported to have IPI, this variable was included in a separate model.
IED, interictal epileptic discharge; IPI, initial precipitating injury.Only the spike frequency showed a significant association according to the linear regression analyses.
Age at epilepsy onset (yr)
Age at IPIa (yr)
Age at investigation (yr)
Performance in controlled oral word-association test
We also performed an SPM group comparison to demonstrate the association between IED frequency and speech lateralization by using a fundamentally different statistical approach (Fig. 2). We compared fMRI activity during word generation in a group of patients with high IED frequency (>6 IED/h; n = 16) with a group of patients with low IED frequency (<1 IED/h; n = 8). The 6 IED/h threshold for “higher IED” frequency was chosen because 6 IED/h was the median IED frequency. This SPM group comparison shows an increase of right-sided fMRI activity during word generation in patients with left-sided MTLE and frequent left-sided IED compared with left-sided TLE patients with low IED frequency (Fig. 2). Thus we included only the two extremes (groups of patients with very frequent vs. very rare spikes) for further illustration here.
Age at IPI was available only in 18 patients. Because of the considerable loss in statistical power (the number of cases decreased by 36%, which would have resulted in an insufficient number of patients in this relatively small study), we did not include this variable in our primary multivariate models. However, the further adjustment with IPI showed no influence on the association between spike frequency and AI by means of the change in the standardized regression coefficient. Age at IPI showed no significant independent association with the AI, but the statistical power was too low to answer this question. Thirteen patients with known IPI had childhood febrile convulsions, three had meningitis in early childhood, one had encephalitis, and another one had a significant childhood brain trauma.
IED frequency shows a correlation with epilepsy duration (14,15). In the multivariate analysis, we could not include epilepsy duration as an explanatory variable because age at epilepsy onset and age also were included. Conversely, one may speculate that long-standing epilepsy (i.e., epilepsy duration and not IED frequency itself) might have an association with atypical speech representation. However, this was not the case, because Pearson's correlation coefficient between the AI and epilepsy duration was 0.18, which is nonsignificant (p = 0.92).
The functional disturbance in epilepsy may be caused by epileptogenic lesion, seizures, and interictal epileptic activity. It has been shown that frequent IEDs might result in transient cognitive impairment in epilepsy patients, whereas such impairment was not present in those periods when no IEDs appeared (16). Moreover, impairment of spatial task performance was demonstrated during right-sided IEDs, whereas during left-sided paroxysms, the verbal cognitive performance was disturbed (16). This suggests that the cognitive impairment caused by IEDs reflects the functional disturbance of the area where the spikes originate. Some patients are clearly handicapped by transient cognitive impairment and their functioning improved when the IEDs were suppressed by medication (17). Thus it is reasonable to assume that chronic frequent IEDs may cause long-term localized neuropsychological deficits, which may induce a reorganization of functions originally localized in areas to which the IEDs originate or propagate. Conversely, in another study, Aldenkamp et al. (18) analyzed the relation between focal or generalized IEDs and cognitive functions and found that neither in patients with seizures receiving antiepileptic treatment nor in seizure-free patients having IEDs, could any transient cognitive impairment be detected. They concluded that transient cognitive impairment occurs infrequently in epilepsy. We suggest that even if transient cognitive impairment does not play a significant role, interictal epileptiform activity as a continuous functional disturbance may influence the representation of cognitive functions. In the present study, we found that higher IED frequency on the left side in MTLE was associated with a left–right shift of speech representation, suggesting that functional disturbances also play a role in speech organization. By using fMRI, a new noninvasive technique, we were able to confirm our previous study in a new series of patients, controlling the other factors that may also play a role in speech lateralization (3).
IEDs in MTLE are generated in the mesiotemporal structures, but 10–50% of them propagate to the lateral temporal neocortex (19,20). Because we did not use intracranial electrodes in this study, the IED frequency was related to the degree of the interictal epileptic involvement (propagation) of the temporal neocortex. Thus our findings indicate that interictal epileptic activity, probably propagated from mesiotemporal to neocortical temporal regions, may interfere with the functions of the temporal speech-receptive areas.
Early-childhood age is thought to be the classic milestone for brain plasticity (1). Surprisingly, in our study, neither the age at IPI nor the age at epilepsy onset was correlated with the AI. Most studies found that atypical speech in TLE is strongly associated with age at epilepsy onset (1,5). Brazdil et al. (5) found that the mean patient age at the time of seizure onset in left TLE with atypical language representation was 5.6 years, but it was 13.1 years in patients with left-hemisphere language dominance (5). Another fMRI study found that although a negative correlation exists between age at brain injury and speech reorganization, no cutoff point exists after which contralateral speech reorganization does not occur in epilepsy patients. Moreover, Sabbah et al. (21) also found no correlation between atypical speech representation and age at epilepsy onset or age at brain injury. Recent studies found that shift of language lateralization can occur in some patients in adolescence or even in adulthood. Children aged 8–12 years show significantly more right-hemispheric participation in speech production than do adults (22). Patients are reported in whom a language transfer developed in adolescence because of a left-sided epileptogenic lesion or epilepsy (23,24). For example, Loddenkemper et al. (2002) reported two adolescents with initially left-hemispheric language dominance proven by Wada test. In both patients, Rasmussen encephalitis developed at ages 8 and 11 years. Because of the effect of the chronic encephalitis or epilepsia partialis continua, in both patients, a right-sided language dominance developed between 9 and 15 years and between 12 and 14 years, respectively (25). Right frontal activation during a verb-generation task appears in >50% of adult tumor patients, but this pattern is seen only in patients with extensive tumors affecting the left prefrontal or temporosuperior regions, suggesting that compensatory functional shifts of language to the right hemisphere might occur only in chronic disease (26). These studies (together with our results) affirm that the role and the nature of the plasticity may be different in acute events compared with chronic disorders, as suggested first by Liegeois et al. (27) and Voets et al. (28). We suggest that under constant functional or slowly progressive structural disturbances, such as low-grade brain tumors, chronic encephalitis, or chronic epileptic activity, the speech representation gradually shifts from left to right hemisphere without serious permanent aphasic symptoms even in adolescence or early adulthood. Conversely, in cases of acute events (for example in left-sided stroke, trauma, or hemispherectomy), the reorganization of language functions remains incomplete, and it is accompanied by speech disturbances because not enough time may exist for adapting of the right hemisphere for left–right transfer of the homologue language-related networks.
Thus we assume that a left–right shift of speech representation in MTLE can occur in later childhood, adolescence, or even in adulthood under the influence of constant and frequent chronic interictal epileptic activity.
We hypothesize that the lack of correlation between atypical language and age at IPI, as well as age at epilepsy onset in our study, may be a result of a bias for the age at IPI and development of HS in early childhood (29). In our study, most IPIs occurred before age 5 years (i.e., all known IPIs of our patients occurred when the brain plasticity certainly allows a left–right shift of speech centers) (30). Thus the age effect of brain injury is not pronounced in MTLE because all brain injuries occur at the age of maximal brain plasticity. The question is: Why did not all patients have a left–right speech shift due to the early brain injury? We hypothesize that beyond the time and localization of brain injury, the epileptic activity might be a third, independent factor that is capable of inducing speech reorganization after an early brain injury affecting the left side.
One of the major limitations of our study was that we used covert word generation as an fMRI paradigm that activates mostly the anterior frontal language-related areas, whereas posterior language cortex can be activated in only 83% (13). Thus it is possible this may not be the best task to use in this study because the left temporal lobe does support receptive language whose networks are different. Moreover, dissociated representations of language are not infrequent in epilepsy patients, with left frontal and right temporal lateralization of language networks, often resulting in “bilateral representation” at the Wada test. Thus it is possible that some patients are “underdiagnosed,” as they might be at least partly right lateralized. Conversely, in our previous fMRI study using the same word-generation paradigm (13), we investigated 63 patients with left-sided TLE, and only a 3% disconcordance was found between fMRI and Wada test results. Moreover, most other studies investigating plasticity of language lateralization also used a word-generation task (22).
Another main limitation of our study is its cross-sectional nature. Only long-term prospective studies can confirm whether the statistical association between IED frequency and atypical speech representation found in this study is a causal relation.
A third limitation of our study may be that we included 11 patients with right-sided TLE as a control population. Previous studies, however, have shown that patients with right-sided TLE do not have the same pattern of language dominance as do control subjects (31). Conversely, including healthy subjects as a control population also may have been problematic because they will certainly have a higher IQ and other strategies in word-generation tasks compared with epilepsy patients.
Conclusively, we found that chronic frequent interictal activity was associated with a right-sided shift of language lateralization in MTLE.
Acknowledgment: This work was supported by a grant from the Deutsche Forschungsgemeinschaft (DFG-Eb 111/2-2, Dr. Ebner), Humboldt Stiftung (Dr. J. Janszky), and Bolyai Scholarship (Dr. J. Janszky).