Dear Editors
Introduction
The prevalence rate of depressive (up to 50%), anxiety (up to 25%) and psychotic disorders (up to 9%) is significantly higher in subjects with medically intractable epilepsy compared to general population and to patients with well-controlled seizures (Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007).
Epilepsy surgery has become an established treatment for patients with medically intractable partial epilepsy (Choi et al. Reference Choi, Sell, Lenert, Muennig, Goodman, Gilliam and Wong2008).
Despite the relatively high prevalence of psychiatric comorbidity, a psychiatric evaluation is not routinely included in pre-surgical evaluation (Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007; Pintor et al. Reference Pintor, Bailles, Fernandez-Egea, Sanchez-Gistau, Torres, Carreno, Rumia, Matrai, Boget, Raspall, Donaire, Bargallo and Setoain2007; Barry et al. Reference Barry, Ettinger, Friel, Gilliam, Harden, Herman, Kanner, Caplan, Plioplys, Salpekar, Dunn, Austin and Jones2008). Pre-surgical comorbid psychiatric disorders are relevant to epilepsy surgery because they could have an impact on seizures outcome, could influence post-surgical complications and the ability to adjust to a seizure-free life.
Post-surgical psychiatric complications mainly occur within the first 12 months of the surgery, with a higher incidence in the first 6 months (Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007).
Studies on epilepsy surgery series have reported that between one-third and one-half of patients post-operatively develop for the first time in their lives, psychiatric symptoms (Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007).
Correlation between psychiatric symptoms and site/side of surgery or post-surgical outcome is not consistent (Spencer et al. Reference Spencer, Berg, Vckrey, Sperling, Bazil, Shinnar, Langfitt, Walczak, Pacia, Ebrahimi and Frbish2003; Wrench et al. Reference Wrench, Wilson and Bladin2004; Pintor et al. Reference Pintor, Bailles, Fernandez-Egea, Sanchez-Gistau, Torres, Carreno, Rumia, Matrai, Boget, Raspall, Donaire, Bargallo and Setoain2007).
Literature data on risk factors for post-surgical depression are controversial.
The primary goal of our study was to identify risk factors for post-surgical depression in subjects operated on for drug-resistant epilepsy. Secondary goals were to confirm the high rate of depression in subjects suffering from epilepsy (prior to surgery) and to look for the first post-surgical depressive episode.
Methods
Data collection
Since January 2002 to December 2007, a consecutive series of 150 adult subjects with epilepsy undergoing surgery for severe drug-resistant focal epilepsy were examined in this study (Engel et al. Reference Engel, Van Ness, Rasmussen and Engel1993).
Inclusion criteria: age range 18–60 years, focal medically intractable epilepsy.
Exclusion criteria: severe mental retardation.
Assessment
Psychiatric evaluation
All 150 subjects were assessed by a psychiatrist as a part of routine pre- and post-surgical assessment. The first psychiatric interview was conducted 1 week to 1 month before surgery.
The same psychiatrist, at the three evaluations, evaluated the same patients.
Diagnosis was made according to DSM-IV-TR criteria. We defined ‘first depressive episode’ as the first episode in lifetime before surgery, ‘recurrent depression’ in all the following episodes. The length of psychiatric follow-up was 12 months. Data about psychopharmachological therapies were collected at baseline, 6 and 12 months. The four depressed subjects at baseline were referred for treatment to psychiatric clinics near to their place of residence. Only one depressed subject accepted an antidepressant treatment as suggested by her psychiatrist. Subjects with post-surgical depression were referred for treatment to psychiatric clinics near to their place of residence. We did not analyze data on antidepressant therapies because of the small number of subjects.
Pre-surgical evaluation
All subjects were referred to the Epilepsy and Parkinson Surgery Centre ‘C. Munari’ at the Niguarda Hospital in Milan because of drug-resistant epilepsy. Surgery was performed after an individualized pre-surgical investigation (Talairach & Bancaud, Reference Talairach and Bancaud1973; Munari et al. Reference Munari, Hoffmann, Francione, Kahane, Tassi, Lo Russo and Benabid1994; Tassi et al. Reference Tassi, Colombo, Garbelli, Francione, Lo Russo, Mai, Cardinale, Cossu, Ferrario, Galli, Bramerio, Citterio and Spreafico2002; Cossu et al. Reference Cossu, Cardinale, Castana, Citterio, Francione, Tassi, Benabid and Lo Russo2005), adapted to the patient's anatomo-electro-clinical features. The site and extent of surgical resection were based on: neurological examination and history; neuroradiological (MRI, CT, Angiography) features; long-term video EEG monitoring. Fifty-seven were studied with long-term V-SEEG (Cossu et al. Reference Cossu, Cardinale, Castana, Citterio, Francione, Tassi, Benabid and Lo Russo2005).
At baseline, all patients were on antiepileptic (AE) politherapy.
Surgery. All the subjects received a tailored surgical resection aimed at removal of the epileptogenic zone, as defined by anatomo-electro-clinical data.
Post-surgical evaluation
Nearly all subjects continued the same (AE) therapy during the follow-up. Subjects were evaluated for outcome on seizures according to Engel's classification.
Statistical analysis
The outcome variable was the presence or absence of depressive episodes during the first 12 months after surgery.
Explanatory variables were gender, age at seizure onset, age at surgery, side and site of surgery, morpheic epilepsy, histological examination of the resected tissue, outcome on seizures, family history for psychiatric disorders, lifetime psychiatric diagnosis and psychiatric diagnosis at time of surgery.
Outcome on seizures was assessed according to Engel's classification (Engel et al. Reference Engel, Van Ness, Rasmussen and Engel1993).
The site of surgical resection was categorized as a multinomial variable: (I) unilobar temporal resections; (II) unilobar frontal resections; (III) posterior resections; (IV) multilobar resections (more than two lobes involved in the resection).
The bivariate analysis was propaedeutic, and the subsequent multivariate regression analysis was performed in order to develop risk factors for post-operative depressive episodes.
For bivariate analysis, Kruskal–Wallis rank sum test was used to analyze numerical variables, and Fisher's two-tailed exact test was performed to analyze categorical (binomial or multinomial) variables. The several tests of bivariate analysis were performed as an advanced preliminary inspection, and the obtained p-values were used in two ways: as a threshold (p ≤ 0.2) to decide whether it was useful to test the explanatory variables into the multivariate model, and as a score to define their relative order into the regression model. For this reason, and because the general null hypothesis is not relevant to this study, no Bonferroni adjustment was made (Perneger, Reference Perneger1998).
For multivariate analysis, a logistic regression model (with Wald test) was fitted to assess the association of the explanatory variables with the outcome on depression in order to develop risk factors (Campbell Reference Campbell and Campbell2006). The explanatory variables were selected with a multistep process, fitting some preliminary models and a final ‘best’ model. In order to avoid multicollinearity and variance inflation, the VIF (Fox, Reference Fox and Fox2002) was every time computed (accepting VIF values <4), and the number of explanatory variables simultaneously included into the model never exceeded 3, because the events (post-surgical depressive episodes) were 33. Every variable potentially associated with the outcome (p-value at bivariate analysis >0.2) was sequentially added to the model, and the decision for the inclusion was taken according to the Best Subsets Technique (Fox, Reference Fox and Fox2002) and performing an ANOVA test between the model with the considered variable and the nested model without it. So, the final model was fitted according to the results of this multistep process, as well as to the main evidences of the literature. Relative risk was calculated with the technique described by Zhang & Yu (Reference Zhang and Yu1998).
P < 0.05 were considered as evidence of findings not attributable to chance.
Statistical analysis was performed using R2.8.1 (R Development Core Team, 2006).
Results
Thirty-three out of the 150 subjects (22%) had post-surgical depressive episodes, 31 in the first 6 months, and two in the last 6 months.
Explanatory variables v. outcome on post-surgical depression are shown in Tables 1 (categorical) and 2 (numerical), together with bivariate analysis results: statistically significant associations were found only for past depressive episodes and age at surgery.
Table 1. Categorical variables and outcome on post-surgical depression: results of the bivariate statistical analysis (two tails Fisher's exact text)
*p <0.05 were considered as evidence of findings not attributable to chance.
OCD: obsessive compulsive disorder.
DNT: Dysembryoplastic Neuroepihtelial Tumor.
NOS: Not Otherwise Specified.
Table 2. Comparison of the considered numerical variables grouped by results, with bivariate statistical analysis (Kruskal–Wallis rank sum test)
*p<0.05 were considered as evidence of findings not attributable to chance.
Post-surgical depression was not associated with side/site of surgical resection, the presence of temporal mesial sclerosis and the presence of a depressive episode at baseline.
Seventy-four percent of subjects were seizure free (Engel's class I) at 12 months. Outcome on seizures was not significantly related to post-surgical depression.
Results of multivariate analysis are shown in Table 3.
Table 3. Output of the fitted logistic regression models for, respectively, pre-surgical, surgical and post-surgical variables, with post-surgical diagnosis of depression (DSM-IV-TR) being the dependent variable
*p<0.05 were considered as evidence of findings not attributable to chance.
The logistic regression model confirmed lifetime depression disorder and age at surgery as risk factors for post-surgical depression.
The site of surgery was included into the model, despite the fact that it is not associated with the outcome in our series, because of evidences from previous literature (Wrench et al. Reference Wrench, Wilson and Bladin2004; Pintor et al. Reference Pintor, Bailles, Fernandez-Egea, Sanchez-Gistau, Torres, Carreno, Rumia, Matrai, Boget, Raspall, Donaire, Bargallo and Setoain2007).
Relative risk analysis shows that one who has a personal history positive for depressive episodes has a risk for post-surgical depression 1.72 greater than others.
Twenty-nine subjects had at least a depressive episode in the past but were not depressed at baseline, three were depressed at baseline and had depressive episodes in the past, one was depressed only at baseline, for a total of 33 (22%) with depressive episodes sometimes before surgery.
After surgery, 14 subjects experienced depression for the first time in their life and nine of them were in Engel's class I.
Discussion
In our study, despite the good surgery outcome (74% Engel's class I), the percentage of subjects who were depressed during the 12 post-surgery months was higher than at baseline. Post-surgical depression is not significantly related to seizures outcome. The percentage of subjects with post-surgical depression in our study is consistent with studies in the literature and is significantly higher in those with a lifetime diagnosis of depression. Therefore, our results suggest that post-surgical depression is not significantly related to surgery outcome, but rather is strongly related to a history of previous episodes. Surgery could be a stressful event able to elicit a depressive episode in subjects at risk for depression.
Our findings are consistent with the literature on the high prevalence and incidence of depression in subjects with epilepsy (Barry et al, Reference Barry, Ettinger, Friel, Gilliam, Harden, Herman, Kanner, Caplan, Plioplys, Salpekar, Dunn, Austin and Jones2008). The most frequent psychiatric post-surgical complication is a depressive episode that develops within the first 12 months in 5–63% of subjects (Devinsky et al. Reference Devinsky, Barr, Vickrey, Berg, Bazil, Pacia, Langfitt, Walczak, Sperling, Shinnhar and Spencer2005; Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007). The relationship between post-surgical depression and seizures outcome is still an open question. Depression has been reported in subjects with good outcome as well as in subjects with unsatisfactory surgery outcome. In the Thapar et al. study (Thapar et al. Reference Thapar, Roland and Harold2005), depression score and seizure frequency were significant predictors for each other both within and across time. Depression in the Wrench et al. study (Wrench et al. Reference Wrench, Wilson and Bladin2004) did not seem to be related to seizure outcome, whereas Reuber et al. (Reference Reuber, Andersen, Elger and Helmstaedter2004) claimed that improvement in depression is related to seizure outcome. At baseline evaluation, the four subjects with depression were not on antidepressant therapy. The lack of antidepressant therapy in subjects with a depressive episode at baseline is consistent with published data that show that depression is both under-recognized and under-treated.
First depressive episode after surgery developed in 14 out of 150 (9.3%) subjects and was not related to post-surgical outcome. The percentage of first depressive episode after surgery in our study is lower than that reported in the literature (Preuter & Norra, Reference Preuter and Norra2005; Barry et al. Reference Barry, Ettinger, Friel, Gilliam, Harden, Herman, Kanner, Caplan, Plioplys, Salpekar, Dunn, Austin and Jones2008). In our study, we considered only the DSM-IV-TR depressive episodes, and therefore subjects with depressive symptoms not severe enough for a DSM-IV-TR diagnosis were not included.
In our sample, no significant relationship was found between depression and side of surgery (right v. left), site of surgery, mesial temporal sclerosis.
Published data show that most subjects who undergo surgery for medically intractable epilepsy have temporal localization (Cohen-Gadol et al. Reference Cohen-Gadol, Wilhelmi, Collignon, White, Britton, Cambier, Christianson, Marsh, Meyer and Cascino2006). Quigg et al. (Reference Quigg, Broshek, Heidal-Schultz, Maedgen and Bertram2003) found that right-sided epilepsy patients with pre-surgical depression may be particularly susceptible to post-surgical depression, but Wrench et al. (Reference Wrench, Wilson and Bladin2004) have not confirmed a relationship between depression and temporal lobe epilepsy.
Thirty-one subjects became depressed after 6 months, only two after 12 months from surgery. These data are consistent with the literature, showing most frequent depression during the first six psychiatric complications, and data about risk factors are reported only in few studies (Balabanov & Kanner, Reference Balabanov, Kanner, Etinger and Kanner2007).
A positive personal history of depression is a risk factor for post-surgical depression in patients operated for drug-resistant partial epilepsy. Predicting which patients may experience depression after surgery is a challenge. Identification of those who are already depressed and need treatment for depression before surgery could be useful for reducing post-surgical depression. Our data suggest that subjects with a positive history for depression should be informed of a higher risk of developing a post-surgical depressive episode, also if they will have a good surgical outcome, in order to get adequate therapy as soon as possible.
Prospective studies should be designed in order to validate externally the logistic regression model we fitted, and to confirm (e.g. with a case-control analysis) that first depressive episodes are caused by surgery for epilepsy and not by major surgery in general.
Limits of the study:
(1) The psychiatric interview was a part of clinical routine assessment, and therefore we did not perform a standardized clinical interview. We are aware that the lack of a standardized research interview may limit the validity of diagnoses.
(2) Our study is not a prospective study
(3) By a statistical point of view, the evidence of the significant rule of positive personal history for depression and age at surgery as risk factors for post-operative depressive episodes is definitely stronger than the ‘innocent’ rule of the other analyzed variables. In fact, the major limit of this study, strictly related to the sample size and to the accepted α-error, is represented by the probably low power of the study. The formal computation of the β-error, variable by variable, could be very difficult, if not impossible, because of the weak and inhomogeneous evidences from the literature.
(4) The peculiarity of the sample does not allow one to generalize results for other population.
Acknowledgements
Our study was supported in part by a FIRST-University of Milan Medical School grant. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Declaration of Interests
I.V.B., I.F.C., I.A.L., I.G.L.R., I.S.F., I.L.T., I.V.S., I.L.C., I.S.S. and I.O.G. declare full disclosure of any conflicts of interest.
