- Top of page
Summary: Purpose: Surgical removal of the hippocampus is the standard of care of patients with drug-resistant medial temporal lobe epilepsy (MTLE). The procedure carries a success rate of ∼75%, but the reasons that some patients fail to achieve seizure control after surgery remain inexplicable. The question of whether the resection of medial temporal lobe structures in addition to the hippocampus would influence the surgical outcome in patients with MTLE was examined.
Methods: We conducted voxel-based statistical analyses of postoperative high-resolution MRI of MTLE patients who underwent anteromedial temporal resection. We applied a cost function transformation of the resection maps for each patient to a common set of spatial coordinates, and we analyzed the contribution of histologically distinct segments of the medial temporal lobe cortex to the surgical outcome. We also performed a voxel-wise mapping of surgical outcome to the temporal lobe.
Results: We observed that the extent of hippocampal removal was associated with better outcomes. However, when the resection of the hippocampus was combined with the resection of the medial temporal lobe, specifically the entorhinal cortex, a greater likelihood of higher seizure control after surgery was found.
Conclusions: Based on this finding, it is possible that the efficiency of the surgical treatment of MTLE can be improved by adjusting the procedure to include the resection of the entorhinal cortex, in addition to the resection of the hippocampus.
Anterior temporal lobe removal combined with amydalohippocampectomy is the conventional treatment for patients with drug-resistant medial temporal lobe epilepsy (MTLE) (Engel, 1997). Up to three fourths of drug-resistant MTLE patients who are submitted to surgery become seizure free after surgery (Spencer, 2002b). Nonetheless, the reason that ≥20% of these patients do not achieve complete seizure control after surgery remains unknown.
MTLE is by far the most common form of partial epilepsy (Wiebe, 2000). It is estimated that ∼100,000 patients within the United States are candidates for epilepsy surgery (Salanova et al., 2005), and 66% of these patients have MTLE (Wiebe, 2000). In the past, patients with drug-refractory MTLE were given prolonged drug therapy before surgery was attempted (Salanova et al., 2005). Lately it has been shown that patients with MTLE who do not respond to two antiepileptic drugs (AEDs) are unlikely to respond to further drug treatment (Kwan and Brodie, 2000), and a randomized controlled trial of surgery for MTLE demonstrated that surgery is superior to prolonged medical therapy (Wiebe et al., 2001; Engel et al., 2003). Anteromedial temporal lobe resection for disabling complex partial seizures generated by MTLE is now the standard of care for patents with drug-refractory MTLE (Engel et al., 2003).
Medial temporal lobe sclerosis (MTS) is the most common postoperative pathology finding in patients with MTLE (Margerison and Corselis, 1966), and MTS now can be diagnosed in vivo with high-resolution magnetic resonance imaging (MRI) in the great majority of patients with MTLE (Cendes et al., 1993). Hippocampal atrophy, whether or not associated with increased T2 signal, is the key MRI feature of MTS, because it can reliably be assessed by careful visual analysis and computer-assisted volumetric measurements (Cendes et al., 1993). The level of hippocampal atrophy correlates with the severity of the symptoms (Cendes et al., 1993) and outcome after surgery (Jack et al., 1992; Kuzniecky et al., 1993; Arruda et al., 1996). Despite recent advances concerning the diagnosis and surgical treatment of patients with MTLE, a large number of patients with MTLE due to unilateral hippocampal sclerosis who undergo surgery fail to achieve seizure control. Overall, the most important prognostic factor is believed to be the extent of hippocampal removal during surgery (Wyler et al., 1995; Arruda et al., 1996; Bonilha et al., 2004b).
When patients do not achieve a good outcome after surgery, reoperation with the intent to remove the remaining segments of the hippocampus yields freedom from seizures for up to almost two thirds of patients after repeated surgery (Hennessy et al., 2000; Salanova et al., 2005). Unfortunately, even when complete hippocampal resection is performed, surgery for MTLE does not abolish seizures for all patients. Approximately one fourth to one fifth of individuals with MTLE due to unilateral hippocampal pathology (i.e., patients who are expected to achieve the best surgical outcome) continue to experience seizures after surgery (Spencer, 2002b). Postoperative electroclinical investigation of patients who fail to achieve a good outcome despite having had complete hippocampal removal reveals that seizures after surgery arise in the hemisphere of resection, and commonly within the resected temporal lobe (Hennessy et al., 2000; Wennberg et al., 2002). This demonstrates that nonhippocampal structures within the temporal lobe are sufficient to initiate and maintain seizures. More speculatively, this finding may indicate that nonhippocampal regions play a crucial epileptogenic role in many operated-on patients with MTLE.
In a parallel line of research, it has been demonstrated that the neuronal damage in patients with drug-refractory MTLE extends beyond the hippocampus and affects mainly brain areas that are functionally or anatomically connected to the hippocampus and the limbic system. Conventional high-resolution MRI morphometric investigation of the medial portion of the temporal lobe demonstrated that the entorhinal cortex, which is the gate area for information reaching and leaving the hippocampus, is the most significantly atrophied area in these patients (Bonilha et al., 2003). These findings have been confirmed by automated voxel-based morphometry studies, which have also disclosed that the pattern of atrophy in the whole brain suggests that a network of damage exists that involves regions connected to the hippocampus or the limbic system (Bonilha et al., 2004c). Interestingly, electroclinical investigation, using intracranial depth electrodes in patients with MTLE before the resection of the hippocampus, has shown that seizures can be generated within the parahippocampal gyrus (i.e., within the entorhinal cortex) in ∼20% of seizures (Wennberg et al., 2002).
The converging evidence that the medial portion of the temporal lobe, more specifically the entorhinal cortex, is damaged and responsible for seizure onset in patients with MTLE has led us to hypothesize that its resection is key to achieving seizure control after surgery for MTLE. Even though surgery for MTLE is refined for the resection of the hippocampus, the extent of resection of the entorhinal cortex can vary across individuals.
In this study, we tested the prediction that if the resection of the amygdala and the hippocampus also encompasses the excision of adjacent structures to the hippocampus, in particular the entorhinal cortex, seizure freedom is achieved. We tested this hypothesis by using an automated voxel-based statistical technique using structural MRI of patients with drug-refractory unilateral MTLE who underwent surgery for the treatment of epilepsy.
- Top of page
In total, 43 patients with unilateral drug refractory MTLE were evaluated. The patient group had a mean age of 37 years (ranging from 17 to 56 years; standard deviation (SD), 10.3 years). Mean age at seizure onset was 7 years (ranging from 1 to 43 years; SD, 7.6 years). Mean duration of epilepsy was 29.8 years (ranging from 4 to 51 years; SD, 12.6 years). Mean follow-up after surgery was 40 months (ranging from 12 to 99 months; SD, 26 months). Eleven (26%) patients were submitted to surgery to the right temporal lobe, and 32 (74%) to the left.
Thirty-three (76%) patients were seizure free after surgery [i.e., were classified as Engel I (comprising patients who were classified as Engel Ia, b, or c]. Six (14%) patients had rare seizures (Engel II), two (5%) had reduction of >90% of seizures (Engel III), and two (5%) had no worthwhile improvement (Engel IV). No significant difference was found between the outcome of patients with right-sided surgery as opposed to left-sided surgery (Yates corrected χ2= 2.749; p = 0.097).
No significant association was noted between surgical outcome and age at onset of seizures (Pearson correlation, 0.17; p = 0.27), duration of epilepsy (Pearson, −0.14; p = 0.37), age at the time of surgery (Pearson, −0.05; p = 0.72), or length of follow-up time (Pearson, 0.19; p = 0.2).
We evaluated the volume of the intersection between the stereotaxic resection maps and the anatomic regions defined on the stereotaxic T1 template image. We computed linear regressions for the extent of resection of each region and the surgical outcome (Engel's score, all tests having 35 degrees of freedom). We observed a significant linear regression between outcome and the extent of resection of the hippocampus (t = 2.371; p = 0.023) and the entorhinal cortex (t = 3.286; p = 0.002) (Fig. 4). No significant linear regression occurred between outcome and the volume of the other structures (amygdala t = 0.47; p = 0.64; perirhinal cortex t = 0.076; p = 0.94; temporopolar cortex t = 1.63; p = 0.11; posterior parahippocampal cortex t = 0.54; p = 0.59).
Figure 4. The relation between resection extent and clinical outcome for six medial temporal lobe regions. Regions of interest in standard space corresponding to healthy medial temporal lobe structures (the entorhinal, the perirhinal, the temporopolar, and the parahippocampal cortices, the hippocampus, and the amygdala) are color coded and shown on multislice on the left. These regions of interest were intersected with the total resection areas from each patient as a resection map transformed from the shape and size of the patient's brain to a standard stereotaxic space. The vertical axis illustrates the percentage of removal, whereas the horizontal axis shows the success of the surgery. Subjects were classified regarding their surgical outcome according to the Engel surgical scale [class I, seizure free; class II, rare seizures; class III, worthwhile improvement, with a reduction of >90% of seizures; class IV, no worthwhile improvement (<90% reduction in seizure frequency)]. Each region is plotted individually. Note that the extents of hippocampal and entorhinal removals are the strong predictors of good clinical outcome.
Download figure to PowerPoint
We also observed that no significant linear regression occurred between the overall surgical resection size and outcome (t =−0.259; p = 0.79).
We observed a significant linear regression for resection size and the resected volumes of the perirhinal cortex (t = 4.75; p < 0.001) and the temporopolar cortex (t = 2.5; p = 0.019). The volumes of resection of the other structures did not show significant linear regression with resection size (hippocampus t = 1.49; p = 0.145; amygdala t =−1.36; p = 0.18; entorhinal cortex t =−1.249; p = 0.22, posterior parahippocampal cortex t = 0.48; p = 0.66).
The results from the resection-outcome mapping analyses are shown in Fig. 5, which shows a map that is a colorized display of permutation, and Fisher's test results. Surgical outcome is demonstrated in a stereotaxic map with the probability of good outcome shown on a voxel-by-voxel basis. Similar to the results from the linear regression between the extent of resection of medial temporal structures and surgical outcome, the resection-outcome mapping analyses showed that a good surgery outcome was most affected by the combined resection of the hippocampus and the medial portion of the temporal lobe, specifically the entorhinal cortex.
Figure 5. Resection-outcome statistical maps highlight brain regions that, when resected during surgery, are associated with a better postoperative outcome. The first row shows the results of the permutation test with the Engel score as a categoric variable. It shows areas that, when resected, are associated with a higher likelihood of a small value in the Engel Outcome Scale. The middle and bottom rows show the results of the binary analyses by using the Fisher's test when patients were grouped according to their Engel scores. The middle row shows areas that are associated with seizure freedom (Engel I), as opposed to non–seizure freedom. The bottom row shows areas that, if resected, are likely to be associated with good outcome (Engel I and II), as opposed to suboptimal outcome (Engel III and IV). For top and middle rows, the scale bar shows z = scores, the right extreme being the threshold for correction for multiple comparisons. Note the association between the likelihood of better outcome or seizure freedom with resection of the entorhinal cortex and hippocampus. The comparison shown in the bottom row does not yield voxels that survive the threshold for multiple comparisons (z = 4.8) because of the low number of subjects with suboptimal outcomes. However, note a trend toward good outcome when resection involves resection of the entorhinal cortex and the hippocampus.
Download figure to PowerPoint
The results from resection-outcome mapping, when the Engel Outcome Scale was investigated as a categoric variable, shows that resection of the hippocampus and the parahippocampal gyrus, specifically the upper limits of the perirhinal cortex and the entorhinal cortex, is associated with a better outcome When the group of seizure-free patients was compared with the group that remained non–seizure free (binary variable), the resection of the hippocampus and the entorhinal cortex was associated with seizure freedom. When good outcome was compared with suboptimal outcome, no voxels survived the threshold for correction for multiple comparisons, possibly because of the reduced number of patients with Engel classes III and IV. Nonetheless, a trend suggested that hippocampal and entorhinal cortex resection were associated with a good outcome (Fig. 5).
- Top of page
We demonstrated that a combined resection of the hippocampus and the upper medial temporal lobe is critical to confer seizure freedom for patients with MTLE undergoing surgical treatment. Our findings confirm a series of previous studies that demonstrated a positive correlation between the extent of hippocampal removal and the success of surgery for MTLE (Kuzniecky et al., 1993; Arruda et al., 1996; Hennessy et al., 2000; Bonilha et al., 2004b; Salanova et al., 2005). Additionally, our study demonstrates that surgical procedures that remove the entorhinal cortex lead to a better prognosis than when this region is preserved intact. We observed that a better surgical outcome is not dependent on whether the overall resection is larger, but rather on when the specific removal of the hippocampus and the entorhinal cortex is accomplished.
These findings may help explain why some patients who exhibit clear-cut severe unilateral hippocampal atrophy and have been given a complete hippocampal resection (i.e., are expected to achieve the best surgical results) fail to achieve a seizure-free status after surgery.
Well-known predictors of good surgical outcome for MTLE are the greater extent of hippocampal removal (Kuzniecky et al., 1993; Arruda et al., 1996; Hennessy et al., 2000; Bonilha et al., 2004b; Salanova et al., 2005) and a more intense preoperative degree of hippocampal atrophy (Garcia et al., 1994; Arruda et al., 1996; Wennberg et al., 2002). It is still controversial whether age at surgery, duration of preoperative epilepsy, and age at onset of seizures are determinants of poor outcome (Hennessy et al., 2000; McIntosh et al., 2004). Certainly, the existence of bilateral hippocampal atrophy or extrahippocampal pathology is associated with a greater likelihood of seizure recurrence after surgery (Jack et al., 1995; Hennessy et al., 2000). It has also been hypothesized that patients who do not achieve a seizure-free status after surgery can harbor subtle isocortical dysplastic epileptogenic lesions that are not detected in the preoperative workup (Hennessy et al., 2000). However, recent advances in diagnostic neuroimaging techniques have greatly enhanced the capability of detecting the so-called “dual pathology,” in which hippocampal atrophy coexists with focal cortical dysplasia (Montenegro et al., 2002). Even with better selection of patients for surgery, the results of the procedure for patients with unilateral hippocampal atrophy are still challenged by the somewhat consistent failure rate. In addition, the theory that poor outcome can be explained by small undetected isocortical dysplastic lesions outside the hippocampus does not support the fact that almost two thirds of patients who did not exhibit a good surgical outcome after a first procedure may achieve seizure control after a reoperation aimed to expand the resection of the hippocampus and the medial portion of the temporal lobes (Wyler et al., 1995; Salanova et al., 2005). It is more probable that the suboptimal surgery results are associated with (a) an incomplete resection of the hippocampus (Engel J, 1997; Bonilha et al., 2004b; Salanova et al., 2005); and (b) a partial resection of the medial portion of the temporal lobe, which is heavily connected to the hippocampus (the entorhinal cortex).
Our data suggest that the resection of the entorhinal cortex, associated with a complete resection of the hippocampus, is a condition of good initial surgical prognosis in patients with unilateral MTLE. This finding matches the notion that patients with MTLE do exhibit neuronal damage beyond the hippocampus, especially within the entorhinal cortex (Bernasconi et al., 2003; Bonilha et al., 2003; 2004c), and that the medial portion of the temporal lobe can generate seizures in ∼20% of patients with MTLE (Spencer, 2002a; Wennberg et al., 2002). Interestingly, the same proportion of patients (i.e., 20%) does not achieve good seizure control after surgery for MTLE.
Probably other factors could account for the success of surgery for MTLE. For instance, genetic determinants of hippocampal sclerosis, environmental factors, and patterns of hippocampal connectivity are features that are not yet well understood and that could account for the variability in surgical results. However, based on our findings, we suggest that the anatomy of the surgical resection can be one important predictor of postoperative outcome. Likewise, our findings are related to early (after 1 year of follow-up) postoperative seizure control. It is not yet clear whether the resection of the medial temporal lobe would also contribute to the long-term prognosis of MTLE surgery.
Finally, our findings also generate new questions. For example, do any presurgical signs predict the necessity of entorhinal cortex removal (for example, relatively little hippocampal atrophy as observed on MRI)? In addition, voxel-based resection mapping could be used to investigate whether quantitative removal of different medial temporal lobe brain areas implies different cognitive postsurgical-outcome profiles.