Complications after Anterior Temporal Lobectomy for Medically Intractable Epilepsy: A Systematic Review and Meta-Analysis

Epilepsy is a serious worldwide health problem, affecting approximately 1% of the general population [1]. Despite optimal medical treatment, 20–40% of the epilepsy patients continue to have seizures [2]. Temporal lobe epilepsy (TLE) associated with hippocampal sclerosis constitutes the most common cause of focal, intractable epilepsy (IE) among adults [3]. It has been demonstrated, that children and adults with IE have poor long-term intellectual and psychosocial outcomes, along with a poor quality of life [4-6]. The medical treatment of these patients accounts for as high as 80% of the cumulative cost of epilepsy in the United States [7].

Two randomized controlled trials (RCTs) have well established the efficacy of surgery in the management of patients with longstanding TLE and have ultimately proved that resective surgery along with continued antiepileptic drug (AED) treatment resulted in a lower probability of seizures during the second year of follow-up than continued AED treatment alone, in selected cases [8, 9]. Anterior temporal lobectomy (ATL) and selective amygdalohippocampectomy (selAH) are the most commonly employed surgical approaches in the management of medical refractory TLE [10].

While surgical treatment is effective, many patients and their referring physicians consider a potential surgical intervention with a lot of skepticism, due to the absence of robust evidence supporting its safety [3]. A single-center RCT from India concluded that surgery, irrespective of the utilized surgical procedure, results in anticipated neurological deficits, depending on the region of brain resection [2]. In addition, a retrospective, observation study reported that ATL was associated with increased risk of complications when compared to selAH [10]. Therefore, the necessity of accurate evaluation of any ATL-associated complications becomes of paramount importance, especially nowadays with the emergence of many nonresective treatment modalities.

In an obvious need to assess the safety of ATL for the management of IE of temporal origin, we conducted a meta-analysis to estimate the postoperative mortality (Q1), overall morbidity (Q2), and individual complication rate (Q3) associated with ATL. The results were stratified according to the targeted population. The final results were interpreted by taking into consideration the overall quality of the available evidence.

We prospectively designed the search methods, eligibility criteria, and data extraction process. The working definition of the complications associated with ATL is summarized in Table 1. The search strategy is displayed in Table 2, while the summary of the protocol is presented thereunder. This meta-analysis was conducted in accordance with the PRISMA-P (Preferred Reporting Items for Systematic Review and Meta-analysis Protocol) [11], and it has been registered in the International Prospective Register of Systematic Reviews (PROSPERO), with registration number CRD42018102297.

Two review authors (T.G. and A.G.B.) independently identified candidate studies through an electronic search of five databases: PubMed, Cochrane Library, ClinicalTrials.gov, Web of Science, and Scopus. They used the following MeSH terms: “drug-resistant epilepsy,” “intractable epilepsy,” “medical refractory epilepsy,” “temporal lobe epilepsy,” “anterior temporal lobectomy,” “anterior temporal lobe surgery,” “complications,” “visual field defects,” “aphasia,” “hemiparesis,” “cranial nerve palsy,” “psychiatric disorders,” “memory,” “cognitive,” “deficit,” “hemorrhage,” “psychological,” and “infections” in any possible combination. The search period extended from 1964 until July 2018. The literature was last accessed in August 2018. Finally, the references of the resulting full texts were searched for further relevant citations.

We focused on RCTs, observational studies, and case series studies reporting at least four different types of complications after ATL for IE. The review process was limited in the English literature. In addition, we discarded editorials, reviews and meta-analyses, underpowered studies (<5 patients), and studies focusing on specific complications (Table 3).

After duplicate removal, the two review authors (A.G.B. and T.G.) assessed the retrieved articles for title and abstract relevance, independently. Then, they evaluated their full texts, according to the eligibility criteria. Disagreement between the reviewers was resolved through discussion with the senior author (K.N.F.). The remaining studies formed the basis of our systematic review and meta-analysis. The process of study selection was outlined in a flowchart (Fig. 1).

Each study was identified by the name of the first author and the year of publication. Then we collected the following data: (1) the hosting country, (2) the type of study, (3) the size of patient sample and its demographic characteristics, (4) the number of patients who presented complications, including death, (5) patient characteristics related to their clinical status, and (6) technical details regarding the surgical procedure.

Two other review authors (E.K. and E.D.) performed the quality appraisal of the collected articles, independently, based on the type of the study. Thus, case series, observational studies, and RCTs were considered as “low,” “moderate,” or “high” quality studies, respectively. The quality of the overall body of evidence was assessed on each question, according to the GRADE working group as “high,” “moderate,” “low,” or “very low.” In the case of disagreement, the authors reached a consensus after consulting the senior author (K.N.F.).

Fixed- and random-effects model meta-analysis was conducted to assess the proportion estimate for each outcome individually, while the interstudy heterogeneity was measured by the I² statistic. A value of I² less than 25% was regarded as low heterogeneity, 25–75% as moderate, and greater than 75% as severe heterogeneity. The results were visualized in forest plots. We estimated the risk for publication bias using the Egger’s regression test and “trim-and-fill” funnel plots. Subgroup analysis and meta-regression was used to identify differences between adults and children. Sensitivity analysis was performed after rerunning the meta-analysis without the “low” quality studies. We used the statistical environment R for all statistical analyses [12]. Significance was set at p <0.05, and for complications associated with zero events, we used a continuity correction equal to 0.5.

The literature search identified 25 studies with 2,842 patients that fulfilled our eligibility criteria (Fig. 1; Tables 4, 5). There were 2 RCTs [8, 9] (8%), 4 observational studies [10, 13-15] (16%), and 19 case series [3, 16-32] (76%) among the retrieved articles. Eighteen studies (72%) concerned adults, 4 studies concerned children (16%), while 3 (12%) reported on mixed populations. Most studies came from the US and Canada, with 7 (28%) and 4 (16%) studies, respectively. The median length of follow-up was 3 years (IQR: 2, 5 years).

The quality of the gathered individual studies was “high” in 2 studies (8%), “moderate” in 4 (16%), and “low” in 19 (76%). In addition, the overall body of evidence was found to be of “low” quality according to the GRADE recommendations [33]. Despite the presence of 2 RCTs among the examined studies, the quality of the evidence was frequently downgraded after the identification of significant publication bias.

Sixteen eligible studies with 2,382 patients described 19 deaths after ATL for IE. The heterogeneity between studies was moderate (I² = 48%, p = 0.02), while the postoperative mortality was estimated to be as high as 0.01 (95% CI: 0.01, 0.02). This result was based on a small number of low-quality studies; however, it was robust, since it did not change after eliminating the “low”-quality studies (p = 0.486). There were no differences in terms of mortality between adult and pediatric series (p = 0.79) (Table 5; Fig. 2).

There were 458 complications recorded among the examined studies. There was significant interstudy heterogeneity (I² =92%, p < 0.01). The postoperative morbidity was estimated to be as high as 0.17 (95% CI: 0.12, 0.24). However, this result was not robust, according to our sensitivity analysis (p = 0.03), and after stratifying it by study quality. Once again, the incidence of the postoperative complications showed no difference between adults and children (p = 0.91) (Table 5; Fig. 3, 4).

Frequently, there was inconsistency among the authors in reporting postoperative complications. However, we grouped them under eleven subheadings: infections, hemorrhage, hydrocephalus, hemiparesis and language deficits, visual field defects (VFDs), cranial nerve deficits (CNDs), hematomas, cognitive disorders, psychiatric disorders, medical complications, and other (Table 5, Fig. 2-6). Psychiatric disorders were the most common postoperative complication group, with an estimated frequency as high as 0.07 (95% CI: 0.04, 0.10), followed by visual field defects (0.06; 0.03, 0.11), and cognitive disorders (0.05; 0.02, 0.10). Less frequent complications included hemiparesis/hemiplegia and language disorders (0.03; 0.01, 0.06), infections (0.03; 0.02, 0.04), hemorrhage (0.02; 0.01, 0.05), cranial nerve deficits (especially trochlear nerve) (0.03; 0.02, 0.05), hydrocephalus and cerebrospinal fluid (CSF)-related disorders (0.02; 0.01, 0.04), extra-axial fluid collections (0.02; 0.01, 0.03), and medical complications (0.02, 0.01, 0.03).

The present systematic review aimed to summarize the complications associated with ATL for IE, based on an electronic search of the pertinent literature. The latter was often of low reporting quality and characterized by significant clinical heterogeneity regarding the definitions of complications and the adopted screening tools.

Two RCTs have already proved the effectiveness of ATL in the management of patients with temporal IE [8, 9]. With the current meta-analysis, we demonstrated that it is also a safe procedure, with an estimated mortality and cumulative morbidity rate as high as 1 and 17%, respectively. In addition, the majority of the complications are limited to temporary psychiatric disturbances, cognitive disorders, VFDs, hemiparesis, hemiplegia, sensory deficits, or speech disorders. Another important finding of our current meta-analysis is that there are no significant differences with regard to the mortality and the cumulative morbidity between pediatric and adult populations.

The findings of the present meta-analysis are comparable to the pertinent literature. Tebo et al. [34], performed a meta-analysis of 61 articles with a total of 5,623 patients, and reported that the overall neurological deficits, infections, and hemorrhages accounted for the majority of the postoperative complications, with a reported proportion estimate as high as 19, 1.4, and 1.3%, respectively. In addition, the authors have noticed that the complication rates had dramatically decreased over time (from 1980 to 2012), but unfortunately they remain an unavoidable consequence of resective epilepsy surgery. More specifically, neurological deficits progressively decreased from 41.8 to 5.2% among patients undergoing ATL or selAH. However, it is of note that the majority of the meta-analytic works focus on the effectiveness of the epilepsy surgery, rather than its safety [35-38].

Knowledge of the perioperative mortality and morbidity is important for both the referring neurologist and the operating surgeon, when discussing with the candidate patient. Equally important is the familiarity of the epilepsy surgeon with the relative frequency of each of the potential complications, their preoperative screening examinations, and the basic modes to timely diagnose and promptly treat them. The best available evidence, together with the surgeon’s experience, should be individually tailored to the surgical planning of each patient in order to achieve the optimal outcome [39, 40].

Hemiparesis and language disorders are the most serious complications after ATL, as they affect the quality of life and the patient’s independence and occur with an estimated frequency as high as 4% (95% CI: 3, 6) [3, 8-10, 13, 14, 16, 17, 20-30, 32, 41]. They have been attributed to either direct intraoperative injury of the pyramidal tract, causing hemiparesis, and of the arcuate and/or the uncinate among other fasciculi resulting in speech difficulties, or to the development of vasospasm secondary to manipulation of the adjacent vessels (middle cerebral artery, and/or anterior choroidal artery) [42]. The importance of employing advanced MR imaging techniques, such as diffusion tensor imaging (DTI) and fractional anisotropy, may help in outlining the corticospinal tract and the speech-involved fasciculi, preoperatively [43]. In addition, the selective employment of intraoperative micro-Doppler in high-risk patients may identify those who are predisposed to vasospasm development [44]. The early pharmacological treatment of these patients with nimodipine or other vasodilators could prevent the development of clinically symptomatic vasospasm, and thus prevent further neurological deterioration [45].

Postoperative cognitive [8, 15, 23, 26, 27] and psychological/psychiatric disorders [8, 15, 19, 26, 27, 29, 30, 32] are among the most common complications after ATL and occur with an estimated frequency as high as 5% (95% CI: 2, 10) and 7% (95% CI: 4, 10), respectively. However, we believe that this figure underestimates the incidence of any postoperative psychological/psychiatric disorders, as this meta-analysis excludes series focusing on specific complications.Of note, some aspects like neuropsychological side effects have been studied quite seriously only in topic-specific publications [46-48]. This complication subgroup includes the decline of preoperative verbal and visual memory (especially on dominant hemisphere cases) [48], as well as exacerbation of a preexisting or de novo development of psychosis, depression, obsession-compulsion, and anxiety disorders [49-51]. Early diagnosis of postoperative psychological and cognitive disturbances may help in their more efficacious management, and thus may improve the patients’ overall outcome.

A thorough clinical psychiatric evaluation along with the employment of the proper neuropsychological battery should be considered preoperatively in order to detect any clinically occult cases. Proper preoperative identification of any predisposing to psychopathology factors (positive family history, mood disturbances, social adjustment difficulties) may help in the early diagnosis of such psychiatric complications and may help in their prompt treatment. Previous studies have identified as predisposing factors: patient’s gender and laterality of resection, positive history of anxiety and/or depression disorders, positive family history for psychiatric disorders, presence of mood disturbances preoperatively, postoperative continuation of seizures, difficulty in postoperative psychosocial adjustment, diffuse preoperative epileptogenic area, and presence of secondary generalized, tonic-clonic seizures preoperatively [52]. The exact role of the side of resection, the extent of the mesial and neocortical temporal resection, the underlying histopathology, and the presence of previous temporal surgeries in the development of postoperative cognitive and psychiatric complications remains to be accurately defined in the future. Finally, these patients may benefit from tailored social and work therapy by improving their functional independence, thus minimizing their postoperative psychological distress.

Symptomatic homonymous quadrantanopsia is another side effect of ATL with an estimated frequency as high as 6% (95% CI: 3, 11) [3, 8, 15-17, 20, 23, 25-27, 30, 32]. Admittedly, this number seems to be a conservative representation of the actual VFD incidence. This is due to an observed heterogeneity of the definitions of the VFDs, a variability of the adopted screening tools in the pertinent literature, and an inherent weakness of our meta-analysis. The definition of VFDs was inconsistent in the literature, ranging from serious symptomatic homonymous quadrantanopsia to minor asymptomatic defects. In addition, early studies used manual kinetic perimetry examination, including Goldmann and Haimark examinations for the diagnosis of VFDs. Both allowed for a detailed mapping of the visual fields but are subjective and observer dependent. Later studies adopted automated static perimetry studies, either monocular or binocular [53-56]. The latter are more reproducible and easier to assess quantitatively [57, 58]. Finally, our meta-analysis excluded topic-specific series. The reported frequency of VFDs in these reports may be as high as 50%; however, with only 8% being symptomatic [58, 59].

The importance of symptomatic homonymous quadranopsia is related to difficulties in spatial orientation and particularly to driving issues. It occurs after injury to the anterior and inferior fibers of the Meyer’s loop. Avoidance of extensive posterior dissection may minimize the occurrence of this complication, but at the expense of the effectiveness of surgery. On the other hand, preoperative advanced MR imaging, such as DTI, may identify the visual pathway and intraoperatively protect it without compromising the extent of resection [53-55]. Additionally, the intraoperative employment of visual evoked potential monitoring and application of direct subcortical electrical stimulation may further increase the accuracy of resection along the visual pathway, and thus may minimize the chance of postoperative VFDs [60].

Surgical complications are not very common after ATL, and include postoperative hematoma formation, hydrocephalus, CSF leakage, and procedure-related medical problems. They occur with an estimated cumulative rate of 1–2%. Although these complications are rare, improvement of the surgical technique, and proper patient selection may further decrease their incidence. Moreover, early mobilization of the patient, along with deep vein thrombosis prophylaxis may further minimize the incidence of any medical complications.

During our meta-analysis, we observed significant geographical variations of the postoperative mortality and morbidity (Fig. 5). Countries with high-volume centers reported significant complication rates, while their low-volume counterparts presented a safer profile. This paradox is partially explained by the retrospective nature of the recruited studies, which is associated with a significant information bias (including recall, attrition, and misclassification). In addition, our results were largely based on simple epidemiological studies, which are characterized by important selection bias (such as patient selection and selective referral). Finally, the presence of significant publication bias draws our attention to the selective reporting of positive outcomes; a fact that contradicts complication reporting. The role of sophisticated imaging techniques and intraoperative neurophysiological monitoring in optimizing the patient’s outcome cannot be overemphasized. Likewise, surgical expertise and experience, which are associated with high-volume practice, may further mitigate the incidence of any procedure-associated complications. Thus, it would not be an overstatement that high-volume facilities are required for the safe management of these patients. It becomes evident that as global knowledge gradually accumulates, the overall morbidity after ATL decreases, a finding that supports our previous postulation (Fig. 5).

The present meta-analysis has some important limitations. As the number of RCTs on ATL was very limited, we broadened our eligibility criteria in order to include observational studies and case series as well. In contrast to the traditional two-arm meta-analysis, the proportional meta-analysis summarizes a phenomenon related to a single parameter. Thus, all three of the above-stated study types could be safely used for extracting our conclusions. Nevertheless, we performed a sensitivity analysis after rerunning our meta-analysis without the “low”-quality studies. The results were taken into consideration, while we evaluated the overall quality of evidence according to the GRADE recommendations. In addition, it needs to be emphasized that our results come from low-quality evidence.

Another limitation of this study was the presence of significant clinical heterogeneity, which was mainly related to the differences in the underlying pathology, target population, and extent of ATL among the included studies. This was verified by the identification of significant statistical heterogeneity, and interestingly, it persisted after stratifying our results according to the target population. Finally, there was no standardized method to report complications after ATL, and as a result, there were large variations among the definitions of each complication among the eligible studies. Thus, in order to deal with this form of heterogeneity, we had to redefine all complications, as presented in Table 1. However, this led to a necessity of listing heterogeneous complications as a common category. Hence, hemiparesis, hemianesthesia, and language disorders were presented as a single categ-ory, attributed to lesions of the central and speech-associated cortical areas. Similarly, there is another broad -category of complications termed “CSF-related complications,” which included both CSF leakage and hydrocephalus.

The need for consistent and uniform reporting of complications after ATL became evident during the current meta-analysis. Further evidence from high-quality studies is of paramount importance in order to improve the quality of the overall evidence, especially in pediatric populations. Finally, additional studies are needed to identify high-risk populations and technique modifications for complication prevention, timely diagnosis, and proper management.

ATL constitutes an extremely safe procedure, with mortality rates approaching zero. Postoperative cognitive and psychological deficits are the most common causes of morbidity after ATL and include a decline in the preoperative verbal and visual memory as well as exacerbation of a preexisting or a de novo development of psychosis, depression, obsessive-compulsive, and/or anxiety disorders. Their early detection may help in their more efficacious management and may improve the patients’ overall outcome. Finally, proper identification of any factors preoperatively predisposing to psychopathology may help in the early diagnosis of such psychiatric complications and may assist in their prompt treatment. Postoperative VFDs represent another common ATL-associated complication, which is systematically underreported. Application of advanced MR imaging techniques, such as DTI and fractional anisotropy, and intraoperative neurophysiological monitoring may decrease their incidence. Neurological complications, including hemiparesis, and dysphasia/aphasia occur less frequently. The employment of advanced imaging modalities for delineating major subcortical white matter tracts and intraoperative micro-Doppler for early detection of cerebral vasospasm may prevent or minimize the incidence of adverse postoperative neurological events. The importance of minimal and gentle manipulation of the adjacent vascular structures during resection cannot be overemphasized.

The authors have no ethical conflicts to disclose.

The authors declare that they have no conflicts of interest.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sector.