Temporal lobe epilepsy (TLE) is associated with behavioural and psychiatric disturbance, including major depression.1 Although surgical removal of epileptogenic tissue is a well-established method of seizure control for intractable TLE, resection may result in a de novo mood disturbance, or exacerbate pre-existing depression.2 3 The mechanisms behind this paradoxical effect remain elusive. Previous research has indicated that psychosocial adjustment difficulties may contribute to postoperative depression.4 In addition, it has been suggested that limbic system dysfunction may also play a role in the high rates of depression experienced after temporal lobe resections.2 Few studies have directly addressed this hypothesis.

In non-epilepsy populations, a large body of research examining the pathogenesis of mood disturbance has focused on the temporolimbic system and its connections to the frontal lobes. This includes findings from the animal literature and suicide postmortems that have drawn attention to the hippocampus as a key structure in the mediation of mood disorders.5 6 Further, numerous studies utilising MRI volumetry have found that patients with major depression have smaller hippocampal volumes than normal controls.7^–12

A few studies have examined hippocampal volume in TLE and have reported no differences between depressed and non-depressed participants.13 14 To date, no studies have examined the relationship between preoperatively measured hippocampal volume and depression following epilepsy surgery. Importantly, examining preoperative volumes may permit markers of postoperative behavioural change to be identified. As such, this study examined whether preoperatively measured hippocampal volume would be related to the occurrence of postoperative depression, including de novo depression. It thus aimed to explore a possible neurobiological predictor of depression following epilepsy surgery.

METHOD

Participants

The sample comprised 42 patients undergoing surgery for medically intractable complex partial seizures within the Comprehensive Epilepsy Program (CEP) of Austin Health, between 1998 and 2005. Inclusion criteria were; (1) access to high resolution preoperative MRI scans, (2) age ⩾18 years, (3) a Full Scale IQ ⩾80, (4) no previous history of neurosurgery, (5) no major peri- or postoperative complications and (6) a functional level of English that permitted completion of the study questionnaires. Twenty-six patients underwent resection of the mesial temporal lobe, including 23 standard anterior temporal lobectomies (ATL) and three parahippocampal gyrus lesionectomies. These patients comprised the mesial temporal resection (MTR) group. Of the remaining 16 patients, six underwent lateral temporal corticectomies, eight had frontal resections, and two had occipital resections. These patients formed the non-mesial temporal resection (NMTR) group, and none had dual pathology involving the mesial temporal lobe. Volumetric analyses were also conducted on 41 neurologically normal volunteers as determined by a neurological examination and a brief neurological and psychiatric screening interview. The demographics of the sample are shown in table 1. A history of febrile convulsions was more common in the MTR group (χ² = 7.111, p = 0.008) than the NMTR group. The NMTR group suffered more frequent seizures than the MTR group (t(39) = −2.183, p = 0.035). The groups did not differ significantly on the remaining demographic variables (p>0.05 for all comparisons) and were similar to the controls for age and gender (p>0.05 for all comparisons). The study was approved by the Austin Hospital and University of Melbourne research and ethics committees, and all subjects provided written informed consent.

Mood and medical variables

All participants were part of a prospective, longitudinal study of mood disturbance after epilepsy surgery. Major depression was assessed preoperatively and at 1, 3, 6 and 12 months postoperatively. Major depression was assessed clinically using the Austin CEP Interview, which is routinely conducted in the Seizure Surgery Follow-up and Rehabilitation Program at Austin Health. This is a well-validated, semistructured interview that has been used extensively by our group and has been shown to be sensitive to mood and adjustment difficulties in epilepsy patients postsurgery.15 16 It includes psychiatric evaluation according to the Diagnostic and Statistical Manual (DSM-IV) criteria. Based on this, the presence or absence of major depression was coded as a binary variable (present/absent). Patients with no preoperative history of major depression on this interview who then experienced major depression after surgery were classified as having de novo depression. In addition, patients completed the Beck Depression Inventory — Second Edition (BDI-II)17 preoperatively, and at each postoperative review. This 21-item self-report questionnaire has been well validated in numerous clinical populations.18 A review of medical files was conducted to obtain information on patient demographics, antidepressant use and seizure and surgical characterisation. Seizure outcome was coded according to the Engel postoperative seizure classification system, comparing patients assigned to Class I (seizure-free) with those assigned to Classes II–IV (not seizure-free).19

MRI and image analysis

Image acquisition

All patients underwent high-resolution three-dimensional T1-weighted 1.5 T MRI scans as part of routine presurgical clinical investigations in the CEP. Patient MRI scans performed prior to November 1998 were acquired on a Siemens Magnetom 63 (Siemens Medical Systems, Erlangen, Germany) (N = 11). The magnetisation prepared rapid gradient echo (MPRAGE) pulse sequence comprised TR 12 ms, TE 4 ms, TI 350 ms, flip angle 10°, FOV 23 cm, matrix 224×256, NEX 1. All the control scans and patient (N = 25) MRI scans performed after 1998 were acquired on a Signa Echospeed Superconducting Imaging System (General Electric Medical Systems, Milwaukee, Wisconsin). The three-dimensional spoiled gradient recalled echo acquisition (SPGR) comprised TR 10.5 ms, TE 2.2 ms, TI 350 ms, flip angle 20°, FOV 25 cm, matrix 256×256, NEX 1. There was no difference between hippocampal volumes obtained on the different scanners (p>0.05).

Image registration

All images were resampled to a slice thickness of 0.5×1×0.5 mm to provide a finer image. Images underwent automated scalp removal using purpose-written software (for MATLAB) prior to image registration. The scalped images were then registered into stereotaxic coordinate space based on the 152 subject T1-weighted average template made available from the Montreal Neurological Institute using a nine-parameter linear transformation (rotation, translation and rescaling along the principal axes) and AIR 3.0.20 This allowed whole brain volume to be controlled. Following registration into standard stereotaxic space, patient and control images were randomised, and volumetric analyses were performed by JW, who was blind to group membership and mood variables.

Volumetric analyses

Manual segmentation of the hippocampus was performed using purpose-written interactive mouse-driven software that enables simultaneous analysis of coronal, sagittal and axial images. The boundaries of the hippocampus were delineated using previously established anatomical landmarks.21 The volume of each structure was computed using a voxel counting algorithm.

Intrarater reliability of volumetric measurements was conducted on 10 randomly selected scans that were segmented on two occasions approximately 6 months apart. Intraclass correlation coefficients for left and right hippocampal volumes were 0.95 and 0.97, respectively, demonstrating a high level of intrarater reliability.

Statistical analyses

χ² analyses were used to compare the MTR and NMTR groups for differences in the qualitative measures of depression. A χ² analysis was also used to examine the influence of side of resection (left vs right) on mood outcome. Separate factorial analyses of variance (ANOVA) investigated differences in hippocampal volume as a function of mood in the MTR and NMTR groups. Contralateral and ipsilateral hippocampal volumes were the dependent variables, and seizure outcome and depression were the independent variables. Pearson correlational analyses were used to examine the relationship between BDI-II scores and contralateral hippocampal volume. A t test was conducted to determine whether contralateral hippocampal volume differed as a function of the occurrence of de novo depression in the MTR group. All analyses were run using SPSS Version 12.0.122 with p = 0.05 set as the criterion of statistical significance.

RESULTS

Characterising the occurrence of depression

Sixty-five per cent of the sample had a psychiatric history with 43% suffering depression and 26% anxiety. A similar proportion of MTR and NMTR patients had a preoperative history of depression (46% vs 38%; χ² = 0.303, p>0.05). While the proportion of patients suffering from depression in the MTR group remained relatively stable pre- to postoperatively (46% pre- to 42% postoperatively), NMTR patients experienced a decrease in depression after surgery (38% pre- to 19% postoperatively). This difference between the groups postoperatively was not statistically significant (χ² = 2.474, p>0.05). There was no relationship between the side of resection and postoperative depression in either the MTR (χ² = 0.077, p>0.05) or the NMTR group (χ² = 0.852, p>0.05).

Ten per cent of the sample suffered from de novo depression in the first 12 months following surgery. The MTR and NMTR groups were strikingly different in this regard, with no NMTR patients developing de novo depression in comparison with 15% of the MTR patients (χ² = 2.721, p = 0.09).

Preoperative mood disturbance and hippocampal volumes

ANOVA showed no relationship between left or right hippocampal volumes and a previous history of depression in the MTR group (F(1,24) = 1.102 and F(1, 24) = 0.0.774 respectively, p>0.05). There was no relationship between left or right hippocampal volumes and a history of depression in the NMTR group (F(1,14) = 1.733, F(1,14) = 1.812 respectively, p>0.05).

Postoperative depression and hippocampal volumes

MTR group

Table 2 presents the mean preoperative hippocampal volumes in the MTR and NMTR groups with or without postoperative depression. For ipsilateral hippocampal volumes in the MTR group, an ANOVA revealed no significant main effect of postoperative depression (F(1,22) = 0.673, p>0.05) or postoperative seizure occurrence within 12 months of surgery (F(1,22) = 2.044, p>0.05). There was no interaction between depression and seizure outcome (F(1,22) = 0.31, p>0.05).

For contralateral hippocampal volumes, an ANOVA revealed a significant main effect of postoperative depression in the MTR group (F(1,22) = 9.708, p = 0.005). Specifically, patients suffering postoperative depression within 12 months of surgery had significantly smaller contralateral hippocampal volumes than those with no postoperative depression. There was no significant main effect for postoperative seizures within 12 months of surgery (F(1,22) = 3.120, p>0.05).

There was a trend for an interaction between postoperative depression and seizure outcome for contralateral hippocampal volumes (F(1,22) = 3.946, p = 0.06). As shown in fig 1, patients with both depression and seizures had smaller hippocampal volumes than patients with depression and no seizures after surgery. In contrast, patients with no postoperative depression showed minimal difference in hippocampal volumes as a function of seizure outcome. Qualitatively, depressed patients with or without seizures had smaller hippocampal volumes than those with no postoperative depression (approximately 40% smaller in the seizure group and 10% smaller in the seizure free group). Further, seizure outcome was not significantly associated with the development of postoperative depression (χ² = 0.257, p>0.05).

There were significant negative correlations between contralateral hippocampal volumes in the MTR group and scores on the BDI-II at 1 and 3 months postsurgery (see table 3). In other words, the more severe the postoperative depression measured on the BDI-II, the smaller the contralateral hippocampus.

De novo depression and hippocampal volumes

MTR patients with de novo depression had significantly smaller contralateral hippocampal volumes than those with no pre- or postoperative depression (t(18) = 3.082, p = 0.006; see table 4). A similar percentage of patients with de novo depression and no postoperative depression were seizure-free after surgery (χ² = 0.952, p>0.05).

NMTR group

Neither contralateral nor ipsilateral hippocampal volumes were related to postoperative depression in the NMTR group (F(1,12) = 0.59 and F(1,12) = 0.021 respectively, p>0.05). There was also no relationship between volume and seizure outcome for either ipsilateral or contralateral volume (F(1,12) = 3.234, and F(1,12) = 0.534 respectively, p>0.05) and no interaction effects (F(1,12) = 3.12 and F(1,12) = 0.671, p>0.05). These results were supported by the quantitative data, with no relationship between BDI-II scores, and contralateral or ipsilateral hippocampal volumes in this group (p>0.05 for all comparisons, see table 3).

Effects of antidepressant drug therapy

Seven (17%) patients were on antidepressant medications (all were selective serotonin-reuptake inhibitors) when their MRI scans were obtained. Of these, six (23%) were in the MTR group. There was no difference in ipsilateral (t(23) = 0.957, p>0.05) or contralateral (t(23) = 0.121, p>0.05) hippocampal volume between MTR patients taking antidepressants and the rest of the MTR group.

Comparison with control hippocampal volumes

The mean right hippocampal volume of left MTR patients with depression (ie, the contralateral hippocampal volume) and mean right hippocampal volume of controls were not significantly different (t(45) = 1.689, p>0.05). Likewise, there was no significant difference between the left mean hippocampal volume of depressed right MTR patients and the mean left hippocampal volume in controls (t(45) = −0.66, p>0.05). There were no differences in the left or right hippocampal volumes of non-depressed MTR patients and controls (p>0.05 for both comparisons).

DISCUSSION

This study found a relationship between preoperative hippocampal volume and major depression after surgery for mesial temporal lobe epilepsy. In particular, patients with postoperative depression had significantly lower hippocampal volumes contralateral to the side of resection than those with no postoperative depression. This relationship was seen when we examined MTR patients with no prior history of depression. This initial finding in de novo depression suggests that smaller hippocampal volumes may predate the onset of depression.

Clinical features of depression before and after seizure surgery

The high rates of pre- and postoperative depression experienced by our sample are consistent with those previously reported in epilepsy surgery patients.2 23 24 We note that the outcome of surgery is variable, with some patients experiencing an improvement in mood, while others developed de novo depression. The current study highlights that patients face a significant risk of developing de novo mood disturbance after seizure surgery, especially if they have undergone a mesial temporal resection. Fifteen per cent of our MTR sample developed a depressive illness of which they had no history. Previous studies have likewise indicated high rates of de novo depression following temporal lobe surgery.2 25 26 The aetiology of depression both pre- and postoperatively is likely to be multifactorial. Previous studies have highlighted the role of psychosocial factors in details.27 The current study suggests that neurobiological factors, such as having mesial temporal structures disrupted by surgery, may also play a role in de novo depression following surgery.

Role of the hippocampus in depression after epilepsy surgery

Similar to previous research, there was no relationship between preoperative depression and hippocampal volume.13 14 Postoperatively, however, contralateral hippocampal volumes were related to depression following MTR but not after NMTR. It can be postulated that removal of one mesial temporal lobe, including the hippocampus, leaves the contralateral structures to mediate mood. This study suggests that in the face of this compromise, if the contralateral hippocampus is small, patients are at increased risk of developing depression. As such, it provides further evidence that neurobiological factors play a role in the development of depression, and in particular de novo depression, following MTR.

The current study focused on depression that occurred within the first year of surgery. We note that previous studies have suggested that depression frequently occurs early in the postoperative period, often within the first 3 months.2 The current findings are consistent with this, with BDI-II scores only showing significant correlations with contralateral volume in the MTR group in the first 3 months of surgery.

The small series of de novo depression cases are able to provide preliminary evidence delineating the temporal sequence of hippocampal volume reduction in depression. There has been some debate as to whether hippocampal volume reduction found in depressed non-epilepsy populations precedes or is caused by depression.9 28 This is largely due to the difficulty of systematically acquiring imaging data in individuals prior to the onset of a depressive illness. All our participants were prospectively assessed with a psychiatric interview and questionnaires both pre- and postoperatively, allowing any changes in mood to be documented. Imaging data were also obtained prior to the surgical procedure, allowing us to examine whether hippocampal volume contributed to changes in postoperative mood. The relationship between a smaller residual hippocampus and de novo depression suggests that hippocampal volume reduction may precede, and predispose to, depression. As such, a small hippocampus may be a marker of vulnerability for mood disturbance experience postoperatively.

Mechanisms of hippocampal volume reduction in depression

It has been postulated that the hippocampal volume reduction found in depressed patients is mediated by hypothalamic–pituitary–adrenal (HPA) axis dysfunction, with increased glucocorticoid levels leading to hippocampal neuronal loss.28 29 The negative correlation between hippocampal volume and lifetime duration of depression has lent indirect support to this theory.8 9 More recent research, however, has not been able to demonstrate a relationship between glucocorticoid levels and hippocampal volume or progressive hippocampal volume loss with ongoing depression.12 30 31 This is in keeping with our findings that hippocampal volume reduction appears to precede depression onset. HPA axis dysfunction may still play a pathogenic role although perhaps not by causing neuronal loss. One possibility is that hippocampal volume reduction affects mood by disrupting normal neurotransmitter function, including that of serotonin. Serotonin has been implicated in mood disorders, and 5-HT_1A receptors are present in high density in limbic system structures such as the hippocampus.32 Further, in experimental animals, increased glucocorticoid levels are associated with changes in the density of the 5-HT_1A serotonin receptor.33 34

Epilepsy and depression: common mechanisms?

Recent research has suggested there may be some commonality in the neurobiological changes associated with epilepsy and depression. First, common neuroanatomical structures have been frequently implicated in the two conditions, including the hippocampus. Second, recent studies have shown common neurotransmitter deficits in epilepsy and depression.35 These include serotonin dysfunction in patients with TLE, with decreased 5-HT_1A binding in the mesial temporal lobe both ipsilateral and contralateral to the seizure focus36 37 and changes in serotonin synthesis in the hippocampus.38 These findings have led some researchers to suggest that the presence of seizures is biologically linked to depression and vice versa.39 While we did not directly assess this hypothesis, the relationship between postoperative depression and a smaller residual hippocampus in the MTR group was strongest in patients suffering postoperative seizures. One possibility is that this relationship is partly accounted for by a potential relationship between seizure recurrence and smaller hippocampal volumes. Nonetheless, we found no relationship between postoperative seizures and the experience of either recurrent or de novo depression. In addition, even within the seizure-free group, patients suffering from postoperative depression had an approximately 10% reduction in hippocampal volume in comparison with those who were not depressed. Thus, we may consider seizure recurrence as presenting an additional risk factor for the development of depression in association with reduced hippocampal volume.

Methodological considerations

Volumetric measurements were conducted in three planes of view using the criteria of Watson et al.21 These boundaries provide reliable results, with a high intrarater reliability in the current study. The sample of patients in this study differs from those usually recruited in studies of major depression. The study design allowed us to control for the effects of brain surgery per se by examining postoperative depression in NMTR patients. In this group where both hippocampi remained intact, we found no relationship between hippocampal volumes and postoperative depression. It would be interesting for future research to examine other structures that are disrupted during MTR, such as the amygdala. This would allow the specificity of the current findings to be explored. Our sample was small, and the extent to which the findings are relevant to patients with depression in general remains to be elucidated.

Conclusion

To conclude, this study has found a relationship between depression after mesial temporal lobe surgery and pre-existing smaller hippocampi contralateral to the site of surgical resection. The findings support the importance of the hippocampus in the regulation of mood, and provide initial evidence that it may be a structural marker of vulnerability to depression following mesial temporal lobe surgery for intractable epilepsy.