Learning From Our Failures: When Anterior Temporal Lobectomies Fail

Abstract

Anterior temporal lobectomy (ATL) remains the standard surgical treatment for drug-resistant temporal lobe epilepsy, yet 20% to 30% of patients experience persistent seizures and/or unfavorable neuropsychological outcomes. These results highlight that postoperative success is influenced not only by the technical execution of surgery but also by the accuracy with which epileptogenic networks are characterized. As such, we consider ATL failure through 2 broad mechanisms: incomplete treatment of the presumed epileptogenic network and limitations in the initial diagnostic understanding of the epileptogenic network. It is also becoming more evident that seizure outcomes alone do not fully capture surgical success, as cognitive, psychiatric, and functional consequences play a critical role in long-term quality of life. Drawing on contemporary evidence and discussions from the Temporal Lobe Club Special Interest Group at the 2025 American Epilepsy Society Annual Meeting, we present a framework for conceptualizing, reevaluating, and managing patients following ATL failure.

Introduction

Anterior temporal lobectomy (ATL) is a well-established and highly effective surgical treatment for drug-resistant temporal lobe epilepsy (TLE), with seizure freedom rates among the best in epilepsy surgery.^1,2 Nevertheless, 20% to 30% of patients undergoing ATL are still unable to achieve seizure freedom.^3,4 A prior comprehensive review of ATL failures categorized the principal mechanisms into insufficient mesial resection, contralateral relapse, unrecognized neocortical onset, dual pathology, and extratemporal or temporal-plus epilepsy. ⁵ These patterns highlight how both technical and diagnostic challenges can undermine outcomes following ATL. We therefore aim to examine how contemporary practices are affected by these issues; and how to incorporate our growing understanding of epilepsy as a network disease.

It has also become increasingly evident that focusing only on seizures does not capture the full spectrum of postoperative outcomes. In truth, cognitive, emotional, and functional outcomes contribute to patients’ quality of life as much as seizure control. On the one hand, memory decline, language dysfunction, and mood or psychosocial impairment can substantially diminish postoperative well-being even in those who achieve seizure freedom. But on the other hand, we must not allow excessive concern about such “neuropsychological failures” to lead us to select interventions that offer inadequate seizure control, as persistent seizures can lead to further cognitive and functional decline over time. Striking an appropriate balance between maximizing seizure control and minimizing neuropsychological decline remains a central challenge when selecting among surgical options for TLE.

These considerations must also be integrated within a rapidly expanding treatment landscape that continues to reshape how we address ATL failures. When seizures persist or recur after ATL, it is becoming less of a question of if a reoperation is feasible, but rather which intervention best addresses the specific mechanism of failure. To outline practical approaches to addressing ATL failures, we present a summary of the lectures and subsequent discussion from the Temporal Lobe Club Special Interest Group at the 2025 American Epilepsy Society Annual Meeting.

Defining the Standard ATL and Its Variability

The ATL is the most common surgical procedure used to treat drug-resistant epilepsy (DRE) and remains the standard of care for TLE, with durable long-term seizure control reported in numerous studies.^1,2,6 The surgical approach to an ATL traditionally involves removal of the anterior temporal neocortex, as well as the mesial structures including the amygdala and hippocampus. Although variations in surgical technique have been proposed, the procedure fundamentally remains the same as initially established by Wilder Penfield and Murray Falconer. ⁷

Nevertheless, there is some debate regarding what a “standard” ATL entails, with substantial variability present between surgeons. While there is general agreement on removal of the limbic structures, primary differences exist regarding the extent of the neocortical and mesial structure resection. The “safe” margins for lateral neocortical resections are commonly understood to be 6.0 to 6.5 cm on the nondominant hemisphere, and 4.0 to 4.5 cm on the dominant hemisphere, with removal of more tissue thought to result in neurological deficit. ¹ Within these parameters, one notable source of variation in neocortical resection involves removal or sparing of the superior temporal gyrus (STG). In a randomized study of patients with dominant left TLE, STG resection did not result in significant differences in either visual confrontation naming or surgical outcome when compared with sparing STG. ⁸

The more general question is if a larger resection leads to better outcomes. The extent of neocortical resection has been described to be negatively associated with emotional well-being; however, a favorable seizure outcome reportedly outweighs this behavioral side effect, independent of the resection volume. ⁹ In terms of mesial resection, one randomized controlled trial found no difference in seizure outcomes when 3.5 cm versus 2.5 cm of mesial tissue was removed along the anteroposterior axis. ¹⁰ In contrast, subsequent studies have demonstrated that the superior extent of resection, particularly involving the piriform cortex, may be critical, with removal of at least half the piriform cortex associated with a 16-fold increase in the odds of seizure freedom. ¹¹ As such, an “optimal” ATL does not necessarily require maximal resection, but rather, an adequate volume resection of the appropriate structures. As our understanding of the specific structures that function as critical nodes within epileptogenic networks continues to evolve, surgical planning can increasingly shift from maximizing resection volume to selectively targeting the most relevant network components.

Understanding Temporal Lobectomy Failures Through the Lens of Temporal Lobe Networks

Multimodal, network-based analyses have emerged as a promising way to predict surgical outcomes. Epilepsy is now widely conceptualized as a network disorder. Over the last decade, quantitative network measures have been applied to routine presurgical modalities such as magnetic resonance imaging (MRI) and electroencephalography (EEG), yet clinical interpretation still relies heavily on qualitative assessment. Growing evidence suggests that quantitative analytics could substantially improve the ability to determine the most effective surgical intervention for individual patients.

A major contributing factor to ATL failures is the heterogeneity of TLE itself. One prominent hypothesis is that the distribution and organization of the epileptic network—rather than the precise location of a single seizure focus—determines surgical success. In this framework, the seizure onset zone may extend across multiple brain regions.

Network-based analyses using MRI, diffusion tensor imaging (DTI), functional MRI (fMRI), and EEG have deepened understanding of epilepsy as a network-level pathology. Graph-theoretic approaches treat brain regions as nodes and their structural or functional connections as edges, enabling assessment of both local and global network properties. Recent resting-state fMRI research has shown that unilateral TLE patients with good surgical outcomes exhibited highly integrated networks, whereas TLE patients with poor outcomes showed more segregated networks. ¹² Similar findings have emerged from diffusion-weighted imaging: high network segregation and reduced hub connectivity predict poor response to ATL. ¹³ Together, these studies support the model that distributed, highly segregated networks are associated with poorer outcomes after resective surgery.

Studies integrating diffusion imaging with intracranial EEG demonstrate that structure-function coupling may serve as a biomarker of surgical success.^14,15 While much of the focus has been on gray matter alone, emerging evidence suggests that white matter intracranial EEG recordings reflect activity from unsampled gray matter regions, and that increased white matter functional connectivity predicts poorer outcomes after laser ablation. ¹⁶ Additional research shows that focal limbic microstructural abnormalities predict good outcomes after destructive surgery in unilateral TLE, while bilateral or distributed abnormalities are linked to poorer outcomes. ¹⁷

Neuropsychological “Success” Versus “Failure” Following ATL

A neuropsychologically successful ATL is one in which the neuropsychological outcome of the procedure aligns with (or is better than) the predicted outcome of the procedure. Viewed in this light, postoperative neuropsychological declines need not represent failures, so long as they were anticipated, counseled for, and understood by the surgical candidate prior to the procedure. Many individuals will accept a degree of cognitive change postoperatively in pursuit of seizure freedom (though if these neuropsychological “costs” occur in the absence of a meaningful change in seizure control, the so-called “double hit”; this can lead to a significant deterioration in quality of life). ¹⁸ “Failure” can therefore be defined as an adverse neuropsychological outcome that was unanticipated, or had been considered a very low probability prior to the procedure.

How then can we maximize the likelihood of successful neuropsychological outcomes following ATL? Firstly, we need good predictive models. There is an extensive literature on this, spanning from rules of thumb ¹⁹ through to regression equations ²⁰ and nomograms, ²¹ all of which consistently highlight certain features as predictive of worse cognitive outcomes:

• Removal/destruction of normal tissue

• Dominant hemisphere surgery

• Higher baseline cognitive scores: these patients have more to lose and are also more likely to be attuned to any declines

• Older age at surgery: presumably related to age dependent declines in plasticity

• Older age at seizure onset: presumably related to a greater likelihood of reorganization in response to an early insult; and conversely, less capacity to adapt to a new insult occurring within a mature, established cognitive system

• Ongoing seizures postoperatively (the “double hit” referred to above ¹⁸ ; though obviously this cannot be known with certainty until postoperatively).

For the above guidelines to be meaningfully applied, appropriate baseline data must be obtained. This demands the use of evidence-based cognitive tasks, such as supraspan word list learning and arbitrary associate learning tasks for estimating verbal memory risk and confrontation naming tasks for estimating language risk.^22,23 If an insensitive/inappropriate task is used, then the predictive value of the model is fundamentally undermined.

Neuropsychological risk must be effectively communicated and understood by surgical candidates themselves, contextualized within patient-specific educational, occupational, and life goals. Even when the postoperative cognitive profile is objectively stable, patients may subjectively evaluate their outcomes as negative or positive contingent on factors such as seizure status, psychosocial circumstances, and anticipated cognitive therapeutic effects. This intersects with the risk of postoperative psychiatric complications. Any lifetime history of psychiatric issues is an important prognostic flag. ²⁴ Psychiatric risk is an important piece of preoperative counseling, including discussion of and planning for potential postop psychiatric changes.

Even with optimal preoperative evaluation, counseling and planning, the unexpected can occur. When the unexpected does occur, the multidisciplinary team must come together to support the patient and family as best as possible, referring to rehabilitation services, counseling, local support services with ongoing monitoring.

Reevaluation for Surgical Intervention After ATL Failure

Reevaluation after ATL failure is not fundamentally novel, but largely mirrors principles applied to any patient with DRE undergoing surgical evaluation. The critical distinction lies in determining whether persistent seizures reflect a technical failure, defined by inadequate treatment of the presumed epileptic network; or a diagnostic failure, reflecting incorrect characterization of the epileptogenic network from the outset.

Incomplete Resection

Reassessment should begin with high-quality structural imaging. Epilepsy-protocol MRI, ideally at 3 T, allows objective evaluation of the true extent of resection, which may differ from surgical intent. Seizure freedom is most strongly associated with complete removal of abnormal mesial tissue rather than the absolute size of a normal hippocampal remnant.^25,26 When residual tissue is clearly localized by imaging or stereoelectroencephalography (sEEG), repeat resection or laser interstitial thermal therapy (LITT) achieves seizure freedom in approximately 50% to 65% of patients.^27,28 Of note, including areas of early propagation as defined on sEEG significantly improves the chance of seizure freedom. ²⁹

Extended or Unexpected Epilepsy Networks

When MRI confirms an anatomically adequate resection, persistent seizures should prompt reconsideration of the original diagnostic hypothesis. Functional and metabolic imaging modalities, including fluorodeoxyglucose positron emission tomography and ictal single-photon emission computed tomography (SPECT) with subtraction ictal SPECT coregistered to MRI (SISCOM), further clarify whether seizure onset localizes to the resection margin or instead reflects broader or extratemporal network involvement.

Magnetoencephalography is particularly useful in postsurgical patients because it is not affected by skull breach artifacts. DTI and fMRI can provide additional insights into structural and functional connectivity and eloquent cortex mapping. Lastly, there are several emerging diagnostic modalities not yet widely used in clinical practice such as 7 T MRI, transcranial magnetic stimulation, resting-state fMRI, and MR spectroscopy, all of which can help look for an extended epilepsy network.

Scalp video-EEG remains an important component, as it confirms ongoing epilepsy and semiology. In some cases, sEEG may be needed to resolve discordant data and map early ictal onset, propagation pathways, and epileptic networks.^29,30

Contralateral of bilateral TLE may account for 17% to 40% of failed ATL surgeries. ³¹ As resection is not feasible in this scenario, bilateral hippocampal responsive neurostimulation (RNS) or deep brain stimulation (DBS) are preferred strategies.^32,33

Extratemporal epilepsy can mimic TLE but seizures actually arise from the insula, operculum, orbitofrontal cortex, or parietal–occipital areas and can produce temporal-like semiology after spreading to the temporal lobe. Similarly, temporal-plus epilepsy arises from a network that involves anteromesial temporal sources but extends beyond the confines of a typical ATL to include extratemporal structures.^28,29,34 A subset of the extratemporal cohort are patients with dual pathology, such as cortical dysplasia, tumor, or heterotopia. Depending on feasibility, repeat resection of the second focus may be curative; neuromodulation can serve as an alternative.^32,33 In those with dual pathology, combined treatment of all active epileptogenic lesions significantly improves seizure freedom compared with single-lesion surgery. ³⁵

Discussion

ATL failure should not be viewed as a terminal outcome but rather as an opportunity to reevaluate epileptogenic mechanisms and individualize further therapy using modern diagnostic and therapeutic tools. Addressing ATL failure demands that we revisit the technical goals of the initial operation, reassess whether the epileptogenic network extends beyond the temporal lobe, and account for the tradeoffs between seizure control and neuropsychological outcomes. These considerations are increasingly important given growing recognition that TLE itself is more heterogeneous than previously recognized, with substantial variability in network organization and capacity for reorganization.

The timing of seizure recurrence provides an important framework for classification. Early recurrence, defined as within 6 to 12 months, suggests incomplete resection or inaccurate localization of the epileptogenic zone, whereas late recurrence more often reflects progressive pathology or network reorganization. ³⁶ Here, we propose a framework for reevaluating failed ATL cases and categorizing cases by failure mechanism: by first determining whether the postoperative resection boundaries match the preoperative intent, then determining whether the presurgical hypothesis was incorrect or incomplete. The reevaluation process begins with basic diagnostic tests including MRI and EEG, but may expand to ancillary imaging tests and sEEG if the reason for the surgical failure is not easily identified.

As our surgical armamentarium expands, reduced exposure to traditional ATL techniques during residency and fellowship may have downstream implications in terms of case selection and technical expertise. Recognizing this risk underscores the need to maintain proficiency in established surgical approaches alongside newer technologies.

Epileptologists and neurosurgeons must communicate effectively and honestly with patients the proposed explanation for the TLE failure and the pros and cons of surgical options that remain. While less invasive options such as LITT, RNS, and DBS may offer relative neuropsychological advantages compared with open resection, these potential benefits must be carefully balanced against differences in seizure control when selecting among treatment options. In cases of neuropsychological failures, memory, attention, and executive dysfunction can be addressed through structured cognitive rehabilitation and compensatory strategies.^37,38 Psychological interventions targeting anxiety and depression are critical. The Home-Based Self-Management and Cognitive Training Changes Lives program, an 8-session home-based cognitive rehabilitation intervention, has demonstrated improvements in adults with epilepsy, supporting its use in postsurgical patients. ³⁹

The future of epilepsy surgical management will aim to create personalized precision presurgical care that optimizes surgical planning and estimates surgical risk with greater accuracy, integrating advanced techniques. Recent research identifying the heterogeneity in TLE network connectivity underscores the need for validated diagnostic neuroimaging-based biomarkers capable of identifying the most appropriate therapy for individual patients with drugresistant TLE. Such tools would also guide decisions about whether invasive intracranial EEG is necessary and how to design sEEG arrays. At present, however, the clinical translation of connectivity-based metrics remains a major hurdle, much of the work to date has been retrospective and further effort is required to develop reliable, prospective, and clinically actionable tools. Advancing network-based, multimodal neuroimaging approaches has the potential to improve surgical success rates (both with initial surgery and repeat surgeries) while reducing morbidity, cost, and unnecessary invasive testing.

Conclusion

ATL failure represents a critical opportunity to remap epileptogenic networks and personalize therapy. A classification-driven, multidisciplinary approach—integrating advanced imaging, hypothesis-driven SEEG, and individualized selection of resection, ablation, or neuromodulation, while predicting and counseling effectively regarding neuropsychological risk—can achieve meaningful seizure reduction and improved quality of life for many patients previously considered surgical failures.